Merge remote branch 'origin/master' into nv50-compiler

618 files changed, 17004 insertions, 12706 deletions

diff --git a/common.py b/common.py
index f7dbe55aa8a..7a14a2c026d 100644
--- a/common.py
+++ b/common.py
@@ -87,7 +87,7 @@ def AddOptions(opts):
 	opts.Add(EnumOption('machine', 'use machine-specific assembly code', default_machine,
 											 allowed_values=('generic', 'ppc', 'x86', 'x86_64')))
 	opts.Add(EnumOption('platform', 'target platform', default_platform,
-											 allowed_values=('linux', 'cell', 'windows', 'winddk', 'wince', 'darwin', 'embedded', 'cygwin')))
+											 allowed_values=('linux', 'cell', 'windows', 'winddk', 'wince', 'darwin', 'embedded', 'cygwin', 'sunos5', 'freebsd8')))
 	opts.Add('toolchain', 'compiler toolchain', 'default')
 	opts.Add(BoolOption('llvm', 'use LLVM', default_llvm))
 	opts.Add(BoolOption('dri', 'build DRI drivers', default_dri))
diff --git a/configs/autoconf.in b/configs/autoconf.in
index 7c6f123cac0..c7611a6f781 100644
--- a/configs/autoconf.in
+++ b/configs/autoconf.in
@@ -123,7 +123,7 @@ GLUT_LIB_DEPS = -L$(TOP)/$(LIB_DIR) @GLUT_MESA_DEPS@ \
 GLW_LIB_DEPS = -L$(TOP)/$(LIB_DIR) @GLW_MESA_DEPS@ \
 	$(EXTRA_LIB_PATH) @GLW_LIB_DEPS@
 APP_LIB_DEPS = $(EXTRA_LIB_PATH) @APP_LIB_DEPS@
-GLESv1_CM_LIB_DEPS = $(EXTRA_LIB_PATH) @GLESv1_LIB_DEPS@
+GLESv1_CM_LIB_DEPS = $(EXTRA_LIB_PATH) @GLESv1_CM_LIB_DEPS@
 GLESv2_LIB_DEPS = $(EXTRA_LIB_PATH) @GLESv2_LIB_DEPS@
 VG_LIB_DEPS = $(EXTRA_LIB_PATH) @VG_LIB_DEPS@
 
diff --git a/configure.ac b/configure.ac
index 96195976a0e..9f3006230eb 100644
--- a/configure.ac
+++ b/configure.ac
@@ -643,7 +643,7 @@ AC_SUBST([GL_PC_REQ_PRIV])
 AC_SUBST([GL_PC_LIB_PRIV])
 AC_SUBST([GL_PC_CFLAGS])
 AC_SUBST([DRI_PC_REQ_PRIV])
-AC_SUBST([GLESv1_LIB_DEPS])
+AC_SUBST([GLESv1_CM_LIB_DEPS])
 AC_SUBST([GLESv1_CM_PC_LIB_PRIV])
 AC_SUBST([GLESv2_LIB_DEPS])
 AC_SUBST([GLESv2_PC_LIB_PRIV])
diff --git a/scons/gallium.py b/scons/gallium.py
index f03716bad8a..03a4ef58815 100644
--- a/scons/gallium.py
+++ b/scons/gallium.py
@@ -295,7 +295,6 @@ def generate(env):
         # - http://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html
         ccflags += [
             '-Wall',
-            '-Wmissing-field-initializers',
             '-Wno-long-long',
             '-ffast-math',
             '-fmessage-length=0', # be nice to Eclipse
@@ -304,6 +303,10 @@ def generate(env):
             '-Wmissing-prototypes',
             '-std=gnu99',
         ]
+        if distutils.version.LooseVersion(ccversion) >= distutils.version.LooseVersion('4.0'):
+            ccflags += [
+                '-Wmissing-field-initializers',
+            ]
         if distutils.version.LooseVersion(ccversion) >= distutils.version.LooseVersion('4.2'):
             ccflags += [
                 '-Werror=pointer-arith',
diff --git a/src/egl/drivers/dri2/egl_dri2.c b/src/egl/drivers/dri2/egl_dri2.c
index c6d79a64e8a..efb93bcf06b 100644
--- a/src/egl/drivers/dri2/egl_dri2.c
+++ b/src/egl/drivers/dri2/egl_dri2.c
@@ -177,9 +177,9 @@ EGLint dri2_to_egl_attribute_map[] = {
    EGL_Y_INVERTED_NOK,		/* __DRI_ATTRIB_YINVERTED */
 };
 
-static void
+static struct dri2_egl_config *
 dri2_add_config(_EGLDisplay *disp, const __DRIconfig *dri_config, int id,
-		int depth, xcb_visualtype_t *visual)
+		int depth, EGLint surface_type)
 {
    struct dri2_egl_config *conf;
    struct dri2_egl_display *dri2_dpy;
@@ -192,6 +192,10 @@ dri2_add_config(_EGLDisplay *disp, const __DRIconfig *dri_config, int id,
    _eglInitConfig(&base, disp, id);
    
    i = 0;
+   double_buffer = 0;
+   bind_to_texture_rgb = 0;
+   bind_to_texture_rgba = 0;
+
    while (dri2_dpy->core->indexConfigAttrib(dri_config, i++, &attrib, &value)) {
       switch (attrib) {
       case __DRI_ATTRIB_RENDER_TYPE:
@@ -242,35 +246,27 @@ dri2_add_config(_EGLDisplay *disp, const __DRIconfig *dri_config, int id,
     * return in the getBuffer callback to get the behaviour we want. */
 
    if (double_buffer)
-      return;
+      return NULL;
 
-   if (visual != NULL) {
-      if (depth != _eglGetConfigKey(&base, EGL_BUFFER_SIZE))
-	 return;
+   if (depth > 0 && depth != _eglGetConfigKey(&base, EGL_BUFFER_SIZE))
+      return NULL;
 
-      _eglSetConfigKey(&base, EGL_SURFACE_TYPE,
-		       EGL_WINDOW_BIT | EGL_PIXMAP_BIT | EGL_PBUFFER_BIT |
-		       EGL_SWAP_BEHAVIOR_PRESERVED_BIT);
+   _eglSetConfigKey(&base, EGL_NATIVE_RENDERABLE, EGL_TRUE);
 
-      _eglSetConfigKey(&base, EGL_NATIVE_VISUAL_ID, visual->visual_id);
-      _eglSetConfigKey(&base, EGL_NATIVE_VISUAL_TYPE, visual->_class);
-   } else {
-      _eglSetConfigKey(&base, EGL_SURFACE_TYPE,
-		       EGL_PIXMAP_BIT | EGL_PBUFFER_BIT);
+   _eglSetConfigKey(&base, EGL_SURFACE_TYPE, surface_type);
+   if (surface_type & (EGL_PIXMAP_BIT | EGL_PBUFFER_BIT)) {
+      _eglSetConfigKey(&base, EGL_BIND_TO_TEXTURE_RGB, bind_to_texture_rgb);
+      if (_eglGetConfigKey(&base, EGL_ALPHA_SIZE) > 0)
+	 _eglSetConfigKey(&base,
+			  EGL_BIND_TO_TEXTURE_RGBA, bind_to_texture_rgba);
    }
 
-   _eglSetConfigKey(&base, EGL_NATIVE_RENDERABLE, EGL_TRUE);
-   _eglSetConfigKey(&base, EGL_BIND_TO_TEXTURE_RGB, bind_to_texture_rgb);
-   if (_eglGetConfigKey(&base, EGL_ALPHA_SIZE) > 0)
-      _eglSetConfigKey(&base,
-		       EGL_BIND_TO_TEXTURE_RGBA, bind_to_texture_rgba);
-
    _eglSetConfigKey(&base, EGL_RENDERABLE_TYPE, disp->ClientAPIsMask);
    _eglSetConfigKey(&base, EGL_CONFORMANT, disp->ClientAPIsMask);
 
    if (!_eglValidateConfig(&base, EGL_FALSE)) {
       _eglLog(_EGL_DEBUG, "DRI2: failed to validate config %d", id);
-      return;
+      return NULL;
    }
 
    conf = malloc(sizeof *conf);
@@ -279,6 +275,8 @@ dri2_add_config(_EGLDisplay *disp, const __DRIconfig *dri_config, int id,
       conf->dri_config = dri_config;
       _eglAddConfig(disp, &conf->base);
    }
+
+   return conf;
 }
 
 /**
@@ -613,10 +611,19 @@ dri2_add_configs_for_visuals(struct dri2_egl_display *dri2_dpy,
    xcb_depth_iterator_t d;
    xcb_visualtype_t *visuals;
    int i, j, id;
+   struct dri2_egl_config *conf;
+   EGLint surface_type;
 
    s = xcb_setup_roots_iterator(xcb_get_setup(dri2_dpy->conn));
    d = xcb_screen_allowed_depths_iterator(s.data);
    id = 1;
+
+   surface_type =
+      EGL_WINDOW_BIT |
+      EGL_PIXMAP_BIT |
+      EGL_PBUFFER_BIT |
+      EGL_SWAP_BEHAVIOR_PRESERVED_BIT;
+
    while (d.rem > 0) {
       EGLBoolean class_added[6] = { 0, };
 
@@ -626,9 +633,16 @@ dri2_add_configs_for_visuals(struct dri2_egl_display *dri2_dpy,
 	    continue;
 
 	 class_added[visuals[i]._class] = EGL_TRUE;
-	 for (j = 0; dri2_dpy->driver_configs[j]; j++)
-	    dri2_add_config(disp, dri2_dpy->driver_configs[j],
-			    id++, d.data->depth, &visuals[i]);
+	 for (j = 0; dri2_dpy->driver_configs[j]; j++) {
+	    conf = dri2_add_config(disp, dri2_dpy->driver_configs[j],
+				   id++, d.data->depth, surface_type);
+	    if (conf == NULL)
+	       continue;
+	    _eglSetConfigKey(&conf->base,
+			     EGL_NATIVE_VISUAL_ID, visuals[i].visual_id);
+	    _eglSetConfigKey(&conf->base,
+			     EGL_NATIVE_VISUAL_TYPE, visuals[i]._class);
+	 }
       }
 
       xcb_depth_next(&d);      
@@ -738,6 +752,12 @@ dri2_create_screen(_EGLDisplay *disp)
    if (api_mask & (1 << __DRI_API_GLES2))
       disp->ClientAPIsMask |= EGL_OPENGL_ES2_BIT;
 
+   if (dri2_dpy->dri2->base.version >= 2) {
+      disp->Extensions.KHR_surfaceless_gles1 = EGL_TRUE;
+      disp->Extensions.KHR_surfaceless_gles2 = EGL_TRUE;
+      disp->Extensions.KHR_surfaceless_opengl = EGL_TRUE;
+   }
+
    return EGL_TRUE;
 
  cleanup_dri_screen:
@@ -948,6 +968,7 @@ dri2_initialize_drm(_EGLDriver *drv, _EGLDisplay *disp,
 		    EGLint *major, EGLint *minor)
 {
    struct dri2_egl_display *dri2_dpy;
+   int i;
 
    dri2_dpy = malloc(sizeof *dri2_dpy);
    if (!dri2_dpy)
@@ -970,10 +991,17 @@ dri2_initialize_drm(_EGLDriver *drv, _EGLDisplay *disp,
    if (!dri2_create_screen(disp))
       goto cleanup_driver;
 
+   for (i = 0; dri2_dpy->driver_configs[i]; i++)
+      dri2_add_config(disp, dri2_dpy->driver_configs[i], i + 1, 0, 0);
+
    disp->Extensions.KHR_image_base = EGL_TRUE;
    disp->Extensions.KHR_gl_renderbuffer_image = EGL_TRUE;
    disp->Extensions.KHR_gl_texture_2D_image = EGL_TRUE;
 
+   /* we're supporting EGL 1.4 */
+   *major = 1;
+   *minor = 4;
+
    return EGL_TRUE;
 
  cleanup_driver:
@@ -1041,6 +1069,7 @@ dri2_create_context(_EGLDriver *drv, _EGLDisplay *disp, _EGLConfig *conf,
    struct dri2_egl_display *dri2_dpy = dri2_egl_display(disp);
    struct dri2_egl_context *dri2_ctx_shared = dri2_egl_context(share_list);
    struct dri2_egl_config *dri2_config = dri2_egl_config(conf);
+   const __DRIconfig *dri_config;
    int api;
 
    dri2_ctx = malloc(sizeof *dri2_ctx);
@@ -1074,11 +1103,16 @@ dri2_create_context(_EGLDriver *drv, _EGLDisplay *disp, _EGLConfig *conf,
       return NULL;
    }
 
+   if (conf != NULL)
+      dri_config = dri2_config->dri_config;
+   else
+      dri_config = NULL;
+
    if (dri2_dpy->dri2->base.version >= 2) {
       dri2_ctx->dri_context =
 	 dri2_dpy->dri2->createNewContextForAPI(dri2_dpy->dri_screen,
 						api,
-						dri2_config->dri_config,
+						dri_config,
 						dri2_ctx_shared ? 
 						dri2_ctx_shared->dri_context : NULL,
 						dri2_ctx);
diff --git a/src/egl/main/eglapi.c b/src/egl/main/eglapi.c
index 09271140b13..4dc8707cfbc 100644
--- a/src/egl/main/eglapi.c
+++ b/src/egl/main/eglapi.c
@@ -379,7 +379,11 @@ eglCreateContext(EGLDisplay dpy, EGLConfig config, EGLContext share_list,
    _EGLContext *context;
    EGLContext ret;
 
-   _EGL_CHECK_CONFIG(disp, conf, EGL_NO_CONTEXT, drv);
+   if (config)
+      _EGL_CHECK_CONFIG(disp, conf, EGL_NO_CONTEXT, drv);
+   else
+      drv = _eglCheckDisplay(disp, __FUNCTION__);
+
    if (!share && share_list != EGL_NO_CONTEXT)
       RETURN_EGL_ERROR(disp, EGL_BAD_CONTEXT, EGL_NO_CONTEXT);
 
diff --git a/src/egl/main/eglconfig.c b/src/egl/main/eglconfig.c
index a9af3200976..ea8e47d02bb 100644
--- a/src/egl/main/eglconfig.c
+++ b/src/egl/main/eglconfig.c
@@ -314,8 +314,10 @@ _eglValidateConfig(const _EGLConfig *conf, EGLBoolean for_matching)
                    EGL_VG_ALPHA_FORMAT_PRE_BIT |
                    EGL_MULTISAMPLE_RESOLVE_BOX_BIT |
                    EGL_SWAP_BEHAVIOR_PRESERVED_BIT;
+#ifdef EGL_MESA_screen_surface
             if (conf->Display->Extensions.MESA_screen_surface)
                mask |= EGL_SCREEN_BIT_MESA;
+#endif
             break;
          case EGL_RENDERABLE_TYPE:
          case EGL_CONFORMANT:
diff --git a/src/egl/main/eglcontext.c b/src/egl/main/eglcontext.c
index d5a1e79a994..9fc529613e5 100644
--- a/src/egl/main/eglcontext.c
+++ b/src/egl/main/eglcontext.c
@@ -83,7 +83,7 @@ _eglParseContextAttribList(_EGLContext *ctx, const EGLint *attrib_list)
       }
    }
 
-   if (err == EGL_SUCCESS) {
+   if (err == EGL_SUCCESS && ctx->Config) {
       EGLint renderable_type, api_bit;
 
       renderable_type = GET_CONFIG_ATTRIB(ctx->Config, EGL_RENDERABLE_TYPE);
@@ -220,45 +220,49 @@ _eglBindContextToSurfaces(_EGLContext *newCtx,
     * surface (oldDraw), the old bindings are broken first and the new one is
     * created.
     */
-   oldCtx = newDraw->CurrentContext;
-   if (newCtx != oldCtx) {
-      if (oldCtx) {
-         assert(oldCtx->DrawSurface == newDraw);
-         oldCtx->DrawSurface = NULL;
+   if (newDraw) {
+      oldCtx = newDraw->CurrentContext;
+      if (newCtx != oldCtx) {
+	 if (oldCtx) {
+	    assert(oldCtx->DrawSurface == newDraw);
+	    oldCtx->DrawSurface = NULL;
+	 }
+      
+	 newDraw->CurrentContext = newCtx;
       }
+   }
 
-      if (newCtx) {
-         _EGLSurface *oldDraw = newCtx->DrawSurface;
-         if (oldDraw)
-            oldDraw->CurrentContext = NULL;
-
-         newCtx->DrawSurface = newDraw;
-         *draw = oldDraw;
-      }
+   if (newCtx) {
+      _EGLSurface *oldDraw = newCtx->DrawSurface;
+      if (oldDraw)
+	 oldDraw->CurrentContext = NULL;
 
-      newDraw->CurrentContext = newCtx;
+      newCtx->DrawSurface = newDraw;
+      *draw = oldDraw;
    }
 
    /* likewise */
-   if (newRead != newDraw)
+   if (newRead && newRead != newDraw) {
       oldCtx = newRead->CurrentContext;
-   if (newCtx != oldCtx) {
-      if (oldCtx) {
-         assert(oldCtx->ReadSurface == newRead);
-         oldCtx->ReadSurface = NULL;
-      }
+      if (newCtx != oldCtx) {
+	 if (oldCtx) {
+	    assert(oldCtx->ReadSurface == newRead);
+	    oldCtx->ReadSurface = NULL;
+	 }
 
-      if (newCtx) {
-         _EGLSurface *oldRead = newCtx->ReadSurface;
-         if (oldRead)
-            oldRead->CurrentContext = NULL;
-
-         newCtx->ReadSurface = newRead;
-         *read = oldRead;
+	 newRead->CurrentContext = newCtx;
       }
+   }
 
-      newRead->CurrentContext = newCtx;
+   if (newCtx) {
+      _EGLSurface *oldRead = newCtx->ReadSurface;
+      if (oldRead)
+	 oldRead->CurrentContext = NULL;
+
+      newCtx->ReadSurface = newRead;
+      *read = oldRead;
    }
+
 }
 
 
@@ -297,7 +301,9 @@ static EGLBoolean
 _eglCheckMakeCurrent(_EGLContext *ctx, _EGLSurface *draw, _EGLSurface *read)
 {
    _EGLThreadInfo *t = _eglGetCurrentThread();
+   _EGLDisplay *dpy;
    EGLint conflict_api;
+   EGLBoolean surfaceless;
 
    if (_eglIsCurrentThreadDummy())
       return _eglError(EGL_BAD_ALLOC, "eglMakeCurrent");
@@ -309,8 +315,23 @@ _eglCheckMakeCurrent(_EGLContext *ctx, _EGLSurface *draw, _EGLSurface *read)
       return EGL_TRUE;
    }
 
-   /* ctx/draw/read must be all given */
-   if (draw == NULL || read == NULL)
+   dpy = ctx->Resource.Display;
+   switch (_eglGetContextAPIBit(ctx)) {
+   case EGL_OPENGL_ES_BIT:
+      surfaceless = dpy->Extensions.KHR_surfaceless_gles1;
+      break;
+   case EGL_OPENGL_ES2_BIT:
+      surfaceless = dpy->Extensions.KHR_surfaceless_gles2;
+      break;
+   case EGL_OPENGL_BIT:
+      surfaceless = dpy->Extensions.KHR_surfaceless_opengl;
+      break;
+   default:
+      surfaceless = EGL_FALSE;
+      break;
+   }
+
+   if (!surfaceless && (draw == NULL || read == NULL))
       return _eglError(EGL_BAD_MATCH, "eglMakeCurrent");
 
    /* context stealing from another thread is not allowed */
@@ -331,12 +352,13 @@ _eglCheckMakeCurrent(_EGLContext *ctx, _EGLSurface *draw, _EGLSurface *read)
     *
     * The latter is more restrictive so we can check only the latter case.
     */
-   if ((draw->CurrentContext && draw->CurrentContext != ctx) ||
-       (read->CurrentContext && read->CurrentContext != ctx))
+   if ((draw && draw->CurrentContext && draw->CurrentContext != ctx) ||
+       (read && read->CurrentContext && read->CurrentContext != ctx))
       return _eglError(EGL_BAD_ACCESS, "eglMakeCurrent");
 
    /* simply require the configs to be equal */
-   if (draw->Config != ctx->Config || read->Config != ctx->Config)
+   if ((draw && draw->Config != ctx->Config) ||
+       (read && read->Config != ctx->Config))
       return _eglError(EGL_BAD_MATCH, "eglMakeCurrent");
 
    switch (ctx->ClientAPI) {
@@ -387,7 +409,6 @@ _eglBindContext(_EGLContext **ctx, _EGLSurface **draw, _EGLSurface **read)
 
       *draw = oldCtx->DrawSurface;
       *read = oldCtx->ReadSurface;
-      assert(*draw && *read);
 
       _eglBindContextToSurfaces(NULL, draw, read);
    }
diff --git a/src/egl/main/eglcurrent.c b/src/egl/main/eglcurrent.c
index c697bf796dc..c0e8c119a94 100644
--- a/src/egl/main/eglcurrent.c
+++ b/src/egl/main/eglcurrent.c
@@ -1,9 +1,9 @@
 #include <stdlib.h>
 #include <string.h>
-#include "eglglobals.h"
 #include "egllog.h"
 #include "eglmutex.h"
 #include "eglcurrent.h"
+#include "eglglobals.h"
 
 
 /* This should be kept in sync with _eglInitThreadInfo() */
@@ -300,12 +300,14 @@ _eglError(EGLint errCode, const char *msg)
       case EGL_BAD_SURFACE:
          s = "EGL_BAD_SURFACE";
          break;
+#ifdef EGL_MESA_screen_surface
       case EGL_BAD_SCREEN_MESA:
          s = "EGL_BAD_SCREEN_MESA";
          break;
       case EGL_BAD_MODE_MESA:
          s = "EGL_BAD_MODE_MESA";
          break;
+#endif
       default:
          s = "other";
       }
diff --git a/src/egl/main/egldisplay.h b/src/egl/main/egldisplay.h
index 0b2f26a4c07..a2cee08bf6f 100644
--- a/src/egl/main/egldisplay.h
+++ b/src/egl/main/egldisplay.h
@@ -60,6 +60,9 @@ struct _egl_extensions
    EGLBoolean KHR_gl_texture_cubemap_image;
    EGLBoolean KHR_gl_texture_3D_image;
    EGLBoolean KHR_gl_renderbuffer_image;
+   EGLBoolean KHR_surfaceless_gles1;
+   EGLBoolean KHR_surfaceless_gles2;
+   EGLBoolean KHR_surfaceless_opengl;
    EGLBoolean NOK_swap_region;
    EGLBoolean NOK_texture_from_pixmap;
 
diff --git a/src/egl/main/egldriver.c b/src/egl/main/egldriver.c
index 1e3d7d24aa7..8fc9e792b06 100644
--- a/src/egl/main/egldriver.c
+++ b/src/egl/main/egldriver.c
@@ -14,7 +14,6 @@
 #include "egldefines.h"
 #include "egldisplay.h"
 #include "egldriver.h"
-#include "eglglobals.h"
 #include "egllog.h"
 #include "eglmisc.h"
 #include "eglmode.h"
diff --git a/src/egl/main/eglglobals.c b/src/egl/main/eglglobals.c
index 725a25eca63..52eebb07f6c 100644
--- a/src/egl/main/eglglobals.c
+++ b/src/egl/main/eglglobals.c
@@ -11,7 +11,6 @@ struct _egl_global _eglGlobal =
 {
    &_eglGlobalMutex,       /* Mutex */
    NULL,                   /* DisplayList */
-   1,                      /* FreeScreenHandle */
    2,                      /* NumAtExitCalls */
    {
       /* default AtExitCalls, called in reverse order */
diff --git a/src/egl/main/eglglobals.h b/src/egl/main/eglglobals.h
index e8bf5416e2a..c3771a8ef10 100644
--- a/src/egl/main/eglglobals.h
+++ b/src/egl/main/eglglobals.h
@@ -16,8 +16,6 @@ struct _egl_global
    /* the list of all displays */
    _EGLDisplay *DisplayList;
 
-   EGLScreenMESA FreeScreenHandle;
-
    EGLint NumAtExitCalls;
    void (*AtExitCalls[10])(void);
 };
diff --git a/src/egl/main/eglmisc.c b/src/egl/main/eglmisc.c
index 281138c7523..985d1e0069d 100644
--- a/src/egl/main/eglmisc.c
+++ b/src/egl/main/eglmisc.c
@@ -97,6 +97,10 @@ _eglUpdateExtensionsString(_EGLDisplay *dpy)
    _EGL_CHECK_EXTENSION(KHR_gl_texture_3D_image);
    _EGL_CHECK_EXTENSION(KHR_gl_renderbuffer_image);
 
+   _EGL_CHECK_EXTENSION(KHR_surfaceless_gles1);
+   _EGL_CHECK_EXTENSION(KHR_surfaceless_gles2);
+   _EGL_CHECK_EXTENSION(KHR_surfaceless_opengl);
+
    _EGL_CHECK_EXTENSION(NOK_swap_region);
    _EGL_CHECK_EXTENSION(NOK_texture_from_pixmap);
 #undef _EGL_CHECK_EXTENSION
diff --git a/src/egl/main/eglmode.c b/src/egl/main/eglmode.c
index 859e9318b4a..37594cdb42d 100644
--- a/src/egl/main/eglmode.c
+++ b/src/egl/main/eglmode.c
@@ -10,6 +10,9 @@
 #include "eglstring.h"
 
 
+#ifdef EGL_MESA_screen_surface
+
+
 #define MIN2(A, B)  (((A) < (B)) ? (A) : (B))
 
 
@@ -353,3 +356,6 @@ _eglQueryModeStringMESA(_EGLDriver *drv, _EGLDisplay *dpy, _EGLMode *m)
 {
    return m->Name;
 }
+
+
+#endif /* EGL_MESA_screen_surface */
diff --git a/src/egl/main/eglmode.h b/src/egl/main/eglmode.h
index a089a5e1943..9167cbc4b9b 100644
--- a/src/egl/main/eglmode.h
+++ b/src/egl/main/eglmode.h
@@ -4,6 +4,9 @@
 #include "egltypedefs.h"
 
 
+#ifdef EGL_MESA_screen_surface
+
+
 #define EGL_NO_MODE_MESA 0
 
 
@@ -54,4 +57,7 @@ extern const char *
 _eglQueryModeStringMESA(_EGLDriver *drv, _EGLDisplay *dpy, _EGLMode *m);
 
 
+#endif /* EGL_MESA_screen_surface */
+
+
 #endif /* EGLMODE_INCLUDED */
diff --git a/src/egl/main/eglscreen.c b/src/egl/main/eglscreen.c
index 8f96fd935c7..8b8966f3ffd 100644
--- a/src/egl/main/eglscreen.c
+++ b/src/egl/main/eglscreen.c
@@ -16,7 +16,6 @@
 #include <string.h>
 
 #include "egldisplay.h"
-#include "eglglobals.h"
 #include "eglcurrent.h"
 #include "eglmode.h"
 #include "eglconfig.h"
@@ -25,6 +24,14 @@
 #include "eglmutex.h"
 
 
+#ifdef EGL_MESA_screen_surface
+
+
+/* ugh, no atomic op? */
+static _EGL_DECLARE_MUTEX(_eglNextScreenHandleMutex);
+static EGLScreenMESA _eglNextScreenHandle = 1;
+
+
 /**
  * Return a new screen handle/ID.
  * NOTE: we never reuse these!
@@ -33,10 +40,10 @@ static EGLScreenMESA
 _eglAllocScreenHandle(void)
 {
    EGLScreenMESA s;
-   
-   _eglLockMutex(_eglGlobal.Mutex);
-   s = _eglGlobal.FreeScreenHandle++;
-   _eglUnlockMutex(_eglGlobal.Mutex);
+
+   _eglLockMutex(&_eglNextScreenHandleMutex);
+   s = _eglNextScreenHandle++;
+   _eglUnlockMutex(&_eglNextScreenHandleMutex);
 
    return s;
 }
@@ -263,3 +270,5 @@ _eglDestroyScreen(_EGLScreen *scrn)
    free(scrn);
 }
 
+
+#endif /* EGL_MESA_screen_surface */
diff --git a/src/egl/main/eglscreen.h b/src/egl/main/eglscreen.h
index 0fd71f71fc8..3db20478ad6 100644
--- a/src/egl/main/eglscreen.h
+++ b/src/egl/main/eglscreen.h
@@ -5,6 +5,9 @@
 #include "egltypedefs.h"
 
 
+#ifdef EGL_MESA_screen_surface
+
+
 /**
  * Per-screen information.
  * Note that an EGL screen doesn't have a size.  A screen may be set to
@@ -86,4 +89,7 @@ PUBLIC void
 _eglDestroyScreen(_EGLScreen *scrn);
 
 
+#endif /* EGL_MESA_screen_surface */
+
+
 #endif /* EGLSCREEN_INCLUDED */
diff --git a/src/egl/main/eglsurface.c b/src/egl/main/eglsurface.c
index d46bdb0672e..52f5c240c65 100644
--- a/src/egl/main/eglsurface.c
+++ b/src/egl/main/eglsurface.c
@@ -30,6 +30,50 @@ _eglClampSwapInterval(_EGLSurface *surf, EGLint interval)
 }
 
 
+#ifdef EGL_MESA_screen_surface
+static EGLint
+_eglParseScreenSurfaceAttribList(_EGLSurface *surf, const EGLint *attrib_list)
+{
+   EGLint i, err = EGL_SUCCESS;
+
+   if (!attrib_list)
+      return EGL_SUCCESS;
+
+   for (i = 0; attrib_list[i] != EGL_NONE; i++) {
+      EGLint attr = attrib_list[i++];
+      EGLint val = attrib_list[i];
+
+      switch (attr) {
+      case EGL_WIDTH:
+         if (val < 0) {
+            err = EGL_BAD_PARAMETER;
+            break;
+         }
+         surf->Width = val;
+         break;
+      case EGL_HEIGHT:
+         if (val < 0) {
+            err = EGL_BAD_PARAMETER;
+            break;
+         }
+         surf->Height = val;
+         break;
+      default:
+         err = EGL_BAD_ATTRIBUTE;
+         break;
+      }
+
+      if (err != EGL_SUCCESS) {
+         _eglLog(_EGL_WARNING, "bad surface attribute 0x%04x", attr);
+         break;
+      }
+   }
+
+   return err;
+}
+#endif /* EGL_MESA_screen_surface */
+
+
 /**
  * Parse the list of surface attributes and return the proper error code.
  */
@@ -44,6 +88,11 @@ _eglParseSurfaceAttribList(_EGLSurface *surf, const EGLint *attrib_list)
    if (!attrib_list)
       return EGL_SUCCESS;
 
+#ifdef EGL_MESA_screen_surface
+   if (type == EGL_SCREEN_BIT_MESA)
+      return _eglParseScreenSurfaceAttribList(surf, attrib_list);
+#endif
+
    if (dpy->Extensions.NOK_texture_from_pixmap)
       texture_type |= EGL_PIXMAP_BIT;
 
@@ -52,12 +101,8 @@ _eglParseSurfaceAttribList(_EGLSurface *surf, const EGLint *attrib_list)
       EGLint val = attrib_list[i];
 
       switch (attr) {
-      /* common (except for screen surfaces) attributes */
+      /* common attributes */
       case EGL_VG_COLORSPACE:
-         if (type == EGL_SCREEN_BIT_MESA) {
-            err = EGL_BAD_ATTRIBUTE;
-            break;
-         }
          switch (val) {
          case EGL_VG_COLORSPACE_sRGB:
          case EGL_VG_COLORSPACE_LINEAR:
@@ -71,10 +116,6 @@ _eglParseSurfaceAttribList(_EGLSurface *surf, const EGLint *attrib_list)
          surf->VGColorspace = val;
          break;
       case EGL_VG_ALPHA_FORMAT:
-         if (type == EGL_SCREEN_BIT_MESA) {
-            err = EGL_BAD_ATTRIBUTE;
-            break;
-         }
          switch (val) {
          case EGL_VG_ALPHA_FORMAT_NONPRE:
          case EGL_VG_ALPHA_FORMAT_PRE:
@@ -101,7 +142,7 @@ _eglParseSurfaceAttribList(_EGLSurface *surf, const EGLint *attrib_list)
          break;
       /* pbuffer surface attributes */
       case EGL_WIDTH:
-         if (type != EGL_PBUFFER_BIT && type != EGL_SCREEN_BIT_MESA) {
+         if (type != EGL_PBUFFER_BIT) {
             err = EGL_BAD_ATTRIBUTE;
             break;
          }
@@ -112,7 +153,7 @@ _eglParseSurfaceAttribList(_EGLSurface *surf, const EGLint *attrib_list)
          surf->Width = val;
          break;
       case EGL_HEIGHT:
-         if (type != EGL_PBUFFER_BIT && type != EGL_SCREEN_BIT_MESA) {
+         if (type != EGL_PBUFFER_BIT) {
             err = EGL_BAD_ATTRIBUTE;
             break;
          }
@@ -129,6 +170,7 @@ _eglParseSurfaceAttribList(_EGLSurface *surf, const EGLint *attrib_list)
          }
          surf->LargestPbuffer = !!val;
          break;
+      /* for eglBindTexImage */
       case EGL_TEXTURE_FORMAT:
          if (!(type & texture_type)) {
             err = EGL_BAD_ATTRIBUTE;
@@ -210,10 +252,12 @@ _eglInitSurface(_EGLSurface *surf, _EGLDisplay *dpy, EGLint type,
    case EGL_PBUFFER_BIT:
       func = "eglCreatePBufferSurface";
       break;
+#ifdef EGL_MESA_screen_surface
    case EGL_SCREEN_BIT_MESA:
       func = "eglCreateScreenSurface";
       renderBuffer = EGL_SINGLE_BUFFER; /* XXX correct? */
       break;
+#endif
    default:
       _eglLog(_EGL_WARNING, "Bad type in _eglInitSurface");
       return EGL_FALSE;
diff --git a/src/gallium/Makefile.template b/src/gallium/Makefile.template
index 43203b1756e..bff399ec64f 100644
--- a/src/gallium/Makefile.template
+++ b/src/gallium/Makefile.template
@@ -23,6 +23,10 @@ INCLUDES = \
 	-I$(TOP)/src/gallium/drivers \
 	$(LIBRARY_INCLUDES)
 
+ifeq ($(MESA_LLVM),1)
+LIBRARY_DEFINES += $(LLVM_CFLAGS)
+endif
+
 
 ##### TARGETS #####
 
diff --git a/src/gallium/auxiliary/Makefile b/src/gallium/auxiliary/Makefile
index dcebab7c0f2..9544e90a965 100644
--- a/src/gallium/auxiliary/Makefile
+++ b/src/gallium/auxiliary/Makefile
@@ -131,6 +131,7 @@ C_SOURCES = \
 	util/u_sampler.c \
 	util/u_simple_shaders.c \
 	util/u_snprintf.c \
+	util/u_staging.c \
 	util/u_surface.c \
 	util/u_surfaces.c \
 	util/u_texture.c \
@@ -149,6 +150,7 @@ C_SOURCES = \
 
 GALLIVM_SOURCES = \
         gallivm/lp_bld_arit.c \
+        gallivm/lp_bld_assert.c \
         gallivm/lp_bld_const.c \
         gallivm/lp_bld_conv.c \
         gallivm/lp_bld_debug.c \
diff --git a/src/gallium/auxiliary/SConscript b/src/gallium/auxiliary/SConscript
index 72a16617db8..3124e20ce84 100644
--- a/src/gallium/auxiliary/SConscript
+++ b/src/gallium/auxiliary/SConscript
@@ -34,14 +34,14 @@ env.CodeGenerate(
     target = 'util/u_format_table.c',
     script = '#src/gallium/auxiliary/util/u_format_table.py',
     source = ['#src/gallium/auxiliary/util/u_format.csv'],
-    command = 'python $SCRIPT $SOURCE > $TARGET'
+    command = python_cmd + ' $SCRIPT $SOURCE > $TARGET'
 )
 
 env.CodeGenerate(
     target = 'util/u_half.c',
     script = 'util/u_half.py',
     source = [],
-    command = 'python $SCRIPT > $TARGET'
+    command = python_cmd + ' $SCRIPT > $TARGET'
 )
 
 env.Depends('util/u_format_table.c', [
@@ -180,6 +180,7 @@ source = [
     'util/u_sampler.c',
     'util/u_simple_shaders.c',
     'util/u_snprintf.c',
+    'util/u_staging.c',
     'util/u_surface.c',
     'util/u_surfaces.c',
     'util/u_texture.c',
@@ -198,6 +199,7 @@ source = [
 if env['llvm']:
     source += [
     'gallivm/lp_bld_arit.c',
+    'gallivm/lp_bld_assert.c',
     'gallivm/lp_bld_const.c',
     'gallivm/lp_bld_conv.c',
     'gallivm/lp_bld_debug.c',
diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c
index c127f741881..995b675b9a1 100644
--- a/src/gallium/auxiliary/draw/draw_context.c
+++ b/src/gallium/auxiliary/draw/draw_context.c
@@ -288,12 +288,19 @@ draw_set_mapped_constant_buffer(struct draw_context *draw,
                 shader_type == PIPE_SHADER_GEOMETRY);
    debug_assert(slot < PIPE_MAX_CONSTANT_BUFFERS);
 
-   if (shader_type == PIPE_SHADER_VERTEX) {
+   switch (shader_type) {
+   case PIPE_SHADER_VERTEX:
       draw->pt.user.vs_constants[slot] = buffer;
+      draw->pt.user.vs_constants_size[slot] = size;
       draw_vs_set_constants(draw, slot, buffer, size);
-   } else if (shader_type == PIPE_SHADER_GEOMETRY) {
+      break;
+   case PIPE_SHADER_GEOMETRY:
       draw->pt.user.gs_constants[slot] = buffer;
+      draw->pt.user.gs_constants_size[slot] = size;
       draw_gs_set_constants(draw, slot, buffer, size);
+      break;
+   default:
+      assert(0 && "invalid shader type in draw_set_mapped_constant_buffer");
    }
 }
 
diff --git a/src/gallium/auxiliary/draw/draw_decompose_tmp.h b/src/gallium/auxiliary/draw/draw_decompose_tmp.h
new file mode 100644
index 00000000000..a52d2b50588
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_decompose_tmp.h
@@ -0,0 +1,425 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.9
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright (C) 2010 LunarG Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+  *   Keith Whitwell <[email protected]>
+ *    Chia-I Wu <[email protected]>
+ */
+
+/* these macros are optional */
+#ifndef LOCAL_VARS
+#define LOCAL_VARS
+#endif
+#ifndef FUNC_ENTER
+#define FUNC_ENTER do {} while (0)
+#endif
+#ifndef FUNC_EXIT
+#define FUNC_EXIT do {} while (0)
+#endif
+#ifndef LINE_ADJ
+#define LINE_ADJ(flags, a0, i0, i1, a1) LINE(flags, i0, i1)
+#endif
+#ifndef TRIANGLE_ADJ
+#define TRIANGLE_ADJ(flags, i0, a0, i1, a1, i2, a2) TRIANGLE(flags, i0, i1, i2)
+#endif
+
+static void
+FUNC(FUNC_VARS)
+{
+   unsigned idx[6], i;
+   ushort flags;
+   LOCAL_VARS
+
+   FUNC_ENTER;
+
+   /* prim, count, and last_vertex_last should have been defined */
+   if (0) {
+      debug_printf("%s: prim 0x%x, count %d, last_vertex_last %d\n",
+            __FUNCTION__, prim, count, last_vertex_last);
+   }
+
+   switch (prim) {
+   case PIPE_PRIM_POINTS:
+      for (i = 0; i < count; i++) {
+         idx[0] = GET_ELT(i);
+         POINT(idx[0]);
+      }
+      break;
+
+   case PIPE_PRIM_LINES:
+      flags = DRAW_PIPE_RESET_STIPPLE;
+      for (i = 0; i + 1 < count; i += 2) {
+         idx[0] = GET_ELT(i);
+         idx[1] = GET_ELT(i + 1);
+         LINE(flags, idx[0], idx[1]);
+      }
+      break;
+
+   case PIPE_PRIM_LINE_LOOP:
+   case PIPE_PRIM_LINE_STRIP:
+      if (count >= 2) {
+         flags = DRAW_PIPE_RESET_STIPPLE;
+         idx[1] = GET_ELT(0);
+         idx[2] = idx[1];
+
+         for (i = 1; i < count; i++, flags = 0) {
+            idx[0] = idx[1];
+            idx[1] = GET_ELT(i);
+            LINE(flags, idx[0], idx[1]);
+         }
+         /* close the loop */
+         if (prim == PIPE_PRIM_LINE_LOOP)
+            LINE(flags, idx[1], idx[2]);
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLES:
+      flags = DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL;
+      for (i = 0; i + 2 < count; i += 3) {
+         idx[0] = GET_ELT(i);
+         idx[1] = GET_ELT(i + 1);
+         idx[2] = GET_ELT(i + 2);
+         TRIANGLE(flags, idx[0], idx[1], idx[2]);
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLE_STRIP:
+      if (count >= 3) {
+         flags = DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL;
+         idx[1] = GET_ELT(0);
+         idx[2] = GET_ELT(1);
+
+         if (last_vertex_last) {
+            for (i = 0; i + 2 < count; i++) {
+               idx[0] = idx[1];
+               idx[1] = idx[2];
+               idx[2] = GET_ELT(i + 2);
+               /* always emit idx[2] last */
+               if (i & 1)
+                  TRIANGLE(flags, idx[1], idx[0], idx[2]);
+               else
+                  TRIANGLE(flags, idx[0], idx[1], idx[2]);
+            }
+         }
+         else {
+            for (i = 0; i + 2 < count; i++) {
+               idx[0] = idx[1];
+               idx[1] = idx[2];
+               idx[2] = GET_ELT(i + 2);
+               /* always emit idx[0] first */
+               if (i & 1)
+                  TRIANGLE(flags, idx[0], idx[2], idx[1]);
+               else
+                  TRIANGLE(flags, idx[0], idx[1], idx[2]);
+            }
+         }
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLE_FAN:
+      if (count >= 3) {
+         flags = DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL;
+         idx[0] = GET_ELT(0);
+         idx[2] = GET_ELT(1);
+
+         /* idx[0] is neither the first nor the last vertex */
+         if (last_vertex_last) {
+            for (i = 0; i + 2 < count; i++) {
+               idx[1] = idx[2];
+               idx[2] = GET_ELT(i + 2);
+               /* always emit idx[2] last */
+               TRIANGLE(flags, idx[0], idx[1], idx[2]);
+            }
+         }
+         else {
+            for (i = 0; i + 2 < count; i++) {
+               idx[1] = idx[2];
+               idx[2] = GET_ELT(i + 2);
+               /* always emit idx[1] first */
+               TRIANGLE(flags, idx[1], idx[2], idx[0]);
+            }
+         }
+      }
+      break;
+
+   case PIPE_PRIM_QUADS:
+      if (last_vertex_last) {
+         for (i = 0; i + 3 < count; i += 4) {
+            idx[0] = GET_ELT(i);
+            idx[1] = GET_ELT(i + 1);
+            idx[2] = GET_ELT(i + 2);
+            idx[3] = GET_ELT(i + 3);
+
+            flags = DRAW_PIPE_RESET_STIPPLE |
+                    DRAW_PIPE_EDGE_FLAG_0 |
+                    DRAW_PIPE_EDGE_FLAG_2;
+            /* always emit idx[3] last */
+            TRIANGLE(flags, idx[0], idx[1], idx[3]);
+
+            flags = DRAW_PIPE_EDGE_FLAG_0 |
+                    DRAW_PIPE_EDGE_FLAG_1;
+            TRIANGLE(flags, idx[1], idx[2], idx[3]);
+         }
+      }
+      else {
+         for (i = 0; i + 3 < count; i += 4) {
+            idx[0] = GET_ELT(i);
+            idx[1] = GET_ELT(i + 1);
+            idx[2] = GET_ELT(i + 2);
+            idx[3] = GET_ELT(i + 3);
+
+            flags = DRAW_PIPE_RESET_STIPPLE |
+                    DRAW_PIPE_EDGE_FLAG_0 |
+                    DRAW_PIPE_EDGE_FLAG_1;
+            /* XXX should always emit idx[0] first */
+            /* always emit idx[3] first */
+            TRIANGLE(flags, idx[3], idx[0], idx[1]);
+
+            flags = DRAW_PIPE_EDGE_FLAG_1 |
+                    DRAW_PIPE_EDGE_FLAG_2;
+            TRIANGLE(flags, idx[3], idx[1], idx[2]);
+         }
+      }
+      break;
+
+   case PIPE_PRIM_QUAD_STRIP:
+      if (count >= 4) {
+         idx[2] = GET_ELT(0);
+         idx[3] = GET_ELT(1);
+
+         if (last_vertex_last) {
+            for (i = 0; i + 3 < count; i += 2) {
+               idx[0] = idx[2];
+               idx[1] = idx[3];
+               idx[2] = GET_ELT(i + 2);
+               idx[3] = GET_ELT(i + 3);
+
+               /* always emit idx[3] last */
+               flags = DRAW_PIPE_RESET_STIPPLE |
+                       DRAW_PIPE_EDGE_FLAG_0 |
+                       DRAW_PIPE_EDGE_FLAG_2;
+               TRIANGLE(flags, idx[2], idx[0], idx[3]);
+
+               flags = DRAW_PIPE_EDGE_FLAG_0 |
+                       DRAW_PIPE_EDGE_FLAG_1;
+               TRIANGLE(flags, idx[0], idx[1], idx[3]);
+            }
+         }
+         else {
+            for (i = 0; i + 3 < count; i += 2) {
+               idx[0] = idx[2];
+               idx[1] = idx[3];
+               idx[2] = GET_ELT(i + 2);
+               idx[3] = GET_ELT(i + 3);
+
+               flags = DRAW_PIPE_RESET_STIPPLE |
+                       DRAW_PIPE_EDGE_FLAG_0 |
+                       DRAW_PIPE_EDGE_FLAG_1;
+               /* XXX should always emit idx[0] first */
+               /* always emit idx[3] first */
+               TRIANGLE(flags, idx[3], idx[2], idx[0]);
+
+               flags = DRAW_PIPE_EDGE_FLAG_1 |
+                       DRAW_PIPE_EDGE_FLAG_2;
+               TRIANGLE(flags, idx[3], idx[0], idx[1]);
+            }
+         }
+      }
+      break;
+
+   case PIPE_PRIM_POLYGON:
+      if (count >= 3) {
+         ushort edge_next, edge_finish;
+
+         if (last_vertex_last) {
+            flags = (DRAW_PIPE_RESET_STIPPLE |
+                     DRAW_PIPE_EDGE_FLAG_2 |
+                     DRAW_PIPE_EDGE_FLAG_0);
+            edge_next = DRAW_PIPE_EDGE_FLAG_0;
+            edge_finish = DRAW_PIPE_EDGE_FLAG_1;
+         }
+         else {
+            flags = (DRAW_PIPE_RESET_STIPPLE |
+                     DRAW_PIPE_EDGE_FLAG_0 |
+                     DRAW_PIPE_EDGE_FLAG_1);
+            edge_next = DRAW_PIPE_EDGE_FLAG_1;
+            edge_finish = DRAW_PIPE_EDGE_FLAG_2;
+         }
+
+         idx[0] = GET_ELT(0);
+         idx[2] = GET_ELT(1);
+
+         for (i = 0; i + 2 < count; i++, flags = edge_next) {
+            idx[1] = idx[2];
+            idx[2] = GET_ELT(i + 2);
+
+            if (i + 3 == count)
+               flags |= edge_finish;
+
+            /* idx[0] is both the first and the last vertex */
+            if (last_vertex_last)
+               TRIANGLE(flags, idx[1], idx[2], idx[0]);
+            else
+               TRIANGLE(flags, idx[0], idx[1], idx[2]);
+         }
+      }
+      break;
+
+   case PIPE_PRIM_LINES_ADJACENCY:
+      flags = DRAW_PIPE_RESET_STIPPLE;
+      for (i = 0; i + 3 < count; i += 4) {
+         idx[0] = GET_ELT(i);
+         idx[1] = GET_ELT(i + 1);
+         idx[2] = GET_ELT(i + 2);
+         idx[3] = GET_ELT(i + 3);
+         LINE_ADJ(flags, idx[0], idx[1], idx[2], idx[3]);
+      }
+      break;
+
+   case PIPE_PRIM_LINE_STRIP_ADJACENCY:
+      if (count >= 4) {
+         flags = DRAW_PIPE_RESET_STIPPLE;
+         idx[1] = GET_ELT(0);
+         idx[2] = GET_ELT(1);
+         idx[3] = GET_ELT(2);
+
+         for (i = 1; i + 2 < count; i++, flags = 0) {
+            idx[0] = idx[1];
+            idx[1] = idx[2];
+            idx[2] = idx[3];
+            idx[3] = GET_ELT(i + 2);
+            LINE_ADJ(flags, idx[0], idx[1], idx[2], idx[3]);
+         }
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLES_ADJACENCY:
+      flags = DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL;
+      for (i = 0; i + 5 < count; i += 6) {
+         idx[0] = GET_ELT(i);
+         idx[1] = GET_ELT(i + 1);
+         idx[2] = GET_ELT(i + 2);
+         idx[3] = GET_ELT(i + 3);
+         idx[4] = GET_ELT(i + 4);
+         idx[5] = GET_ELT(i + 5);
+         TRIANGLE_ADJ(flags, idx[0], idx[1], idx[2], idx[3], idx[4], idx[5]);
+      }
+      break;
+
+   case PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY:
+      if (count >= 6) {
+         flags = DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL;
+         idx[0] = GET_ELT(1);
+         idx[2] = GET_ELT(0);
+         idx[4] = GET_ELT(2);
+         idx[3] = GET_ELT(4);
+
+         /*
+          * The vertices of the i-th triangle are stored in
+          * idx[0,2,4] = { 2*i, 2*i+2, 2*i+4 };
+          *
+          * The adjacent vertices are stored in
+          * idx[1,3,5] = { 2*i-2, 2*i+6, 2*i+3 }.
+          *
+          * However, there are two exceptions:
+          *
+          * For the first triangle, idx[1] = 1;
+          * For the  last triangle, idx[3] = 2*i+5.
+          */
+         if (last_vertex_last) {
+            for (i = 0; i + 5 < count; i += 2) {
+               idx[1] = idx[0];
+
+               idx[0] = idx[2];
+               idx[2] = idx[4];
+               idx[4] = idx[3];
+
+               idx[3] = GET_ELT(i + ((i + 7 < count) ? 6 : 5));
+               idx[5] = GET_ELT(i + 3);
+
+               /*
+                * alternate the first two vertices (idx[0] and idx[2]) and the
+                * corresponding adjacent vertices (idx[3] and idx[5]) to have
+                * the correct orientation
+                */
+               if (i & 2) {
+                  TRIANGLE_ADJ(flags,
+                        idx[2], idx[1], idx[0], idx[5], idx[4], idx[3]);
+               }
+               else {
+                  TRIANGLE_ADJ(flags,
+                        idx[0], idx[1], idx[2], idx[3], idx[4], idx[5]);
+               }
+            }
+         }
+         else {
+            for (i = 0; i + 5 < count; i += 2) {
+               idx[1] = idx[0];
+
+               idx[0] = idx[2];
+               idx[2] = idx[4];
+               idx[4] = idx[3];
+
+               idx[3] = GET_ELT(i + ((i + 7 < count) ? 6 : 5));
+               idx[5] = GET_ELT(i + 3);
+
+               /*
+                * alternate the last two vertices (idx[2] and idx[4]) and the
+                * corresponding adjacent vertices (idx[1] and idx[5]) to have
+                * the correct orientation
+                */
+               if (i & 2) {
+                  TRIANGLE_ADJ(flags,
+                        idx[0], idx[5], idx[4], idx[3], idx[2], idx[1]);
+               }
+               else {
+                  TRIANGLE_ADJ(flags,
+                        idx[0], idx[1], idx[2], idx[3], idx[4], idx[5]);
+               }
+            }
+         }
+      }
+      break;
+
+   default:
+      assert(0);
+      break;
+   }
+
+   FUNC_EXIT;
+}
+
+#undef LOCAL_VARS
+#undef FUNC_ENTER
+#undef FUNC_EXIT
+#undef LINE_ADJ
+#undef TRIANGLE_ADJ
+
+#undef FUNC
+#undef FUNC_VARS
+#undef GET_ELT
+#undef POINT
+#undef LINE
+#undef TRIANGLE
diff --git a/src/gallium/auxiliary/draw/draw_gs.c b/src/gallium/auxiliary/draw/draw_gs.c
index 79a57a67f3e..4a1013e79a5 100644
--- a/src/gallium/auxiliary/draw/draw_gs.c
+++ b/src/gallium/auxiliary/draw/draw_gs.c
@@ -1,6 +1,6 @@
 /**************************************************************************
  *
- * Copyright 2009 VMWare Inc.
+ * Copyright 2009 VMware, Inc.
  * All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -75,7 +75,10 @@ draw_gs_set_constants(struct draw_context *draw,
                       const void *constants,
                       unsigned size)
 {
-   /* noop */
+   /* noop. added here for symmetry with the VS
+    * code and in case we'll ever want to allign
+    * the constants, e.g. when we'll change to a
+    * different interpreter */
 }
 
 
@@ -370,32 +373,23 @@ static void gs_tri_adj(struct draw_geometry_shader *shader,
    gs_flush(shader, 1);
 }
 
-#define TRIANGLE(gs,i0,i1,i2) gs_tri(gs,i0,i1,i2)
-#define TRI_ADJ(gs,i0,i1,i2,i3,i4,i5)  gs_tri_adj(gs,i0,i1,i2,i3,i4,i5)
-#define LINE(gs,i0,i1)        gs_line(gs,i0,i1)
-#define LINE_ADJ(gs,i0,i1,i2,i3)    gs_line_adj(gs,i0,i1,i2,i3)
-#define POINT(gs,i0)          gs_point(gs,i0)
-#define FUNC gs_run
-#define LOCAL_VARS
+#define FUNC         gs_run
+#define GET_ELT(idx) (idx)
 #include "draw_gs_tmp.h"
 
 
-#define TRIANGLE(gs,i0,i1,i2) gs_tri(gs,elts[i0],elts[i1],elts[i2])
-#define TRI_ADJ(gs,i0,i1,i2,i3,i4,i5)           \
-   gs_tri_adj(gs,elts[i0],elts[i1],elts[i2],elts[i3], \
-              elts[i4],elts[i5])
-#define LINE(gs,i0,i1)        gs_line(gs,elts[i0],elts[i1])
-#define LINE_ADJ(gs,i0,i1,i2,i3)  gs_line_adj(gs,elts[i0],      \
-                                              elts[i1],         \
-                                              elts[i2],elts[i3])
-#define POINT(gs,i0)          gs_point(gs,elts[i0])
-#define FUNC gs_run_elts
-#define LOCAL_VARS                         \
-   const ushort *elts = input_prims->elts;
+#define FUNC         gs_run_elts
+#define LOCAL_VARS   const ushort *elts = input_prims->elts;
+#define GET_ELT(idx) (elts[idx] & ~DRAW_PIPE_FLAG_MASK)
 #include "draw_gs_tmp.h"
 
+
+/**
+ * Execute geometry shader using TGSI interpreter.
+ */
 int draw_geometry_shader_run(struct draw_geometry_shader *shader,
                              const void *constants[PIPE_MAX_CONSTANT_BUFFERS], 
+                             const unsigned constants_size[PIPE_MAX_CONSTANT_BUFFERS], 
                              const struct draw_vertex_info *input_verts,
                              const struct draw_prim_info *input_prim,
                              struct draw_vertex_info *output_verts,
@@ -405,7 +399,6 @@ int draw_geometry_shader_run(struct draw_geometry_shader *shader,
    unsigned input_stride = input_verts->vertex_size;
    unsigned vertex_size = input_verts->vertex_size;
    struct tgsi_exec_machine *machine = shader->machine;
-   unsigned int i;
    unsigned num_input_verts = input_prim->linear ?
                               input_verts->count :
                               input_prim->count;
@@ -447,9 +440,8 @@ int draw_geometry_shader_run(struct draw_geometry_shader *shader,
    }
    shader->primitive_lengths = MALLOC(max_out_prims * sizeof(unsigned));
 
-   for (i = 0; i < PIPE_MAX_CONSTANT_BUFFERS; i++) {
-      machine->Consts[i] = constants[i];
-   }
+   tgsi_exec_set_constant_buffers(machine, PIPE_MAX_CONSTANT_BUFFERS,
+                                  constants, constants_size);
 
    if (input_prim->linear)
       gs_run(shader, input_prim, input_verts,
diff --git a/src/gallium/auxiliary/draw/draw_gs.h b/src/gallium/auxiliary/draw/draw_gs.h
index 2cb634818c2..67bc1aa73ff 100644
--- a/src/gallium/auxiliary/draw/draw_gs.h
+++ b/src/gallium/auxiliary/draw/draw_gs.h
@@ -1,6 +1,6 @@
 /**************************************************************************
  * 
- * Copyright 2009 VMWare Inc.
+ * Copyright 2009 VMware, Inc.
  * All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -73,6 +73,7 @@ struct draw_geometry_shader {
  */
 int draw_geometry_shader_run(struct draw_geometry_shader *shader,
                              const void *constants[PIPE_MAX_CONSTANT_BUFFERS], 
+                             const unsigned constants_size[PIPE_MAX_CONSTANT_BUFFERS], 
                              const struct draw_vertex_info *input_verts,
                              const struct draw_prim_info *input_prim,
                              struct draw_vertex_info *output_verts,
diff --git a/src/gallium/auxiliary/draw/draw_gs_tmp.h b/src/gallium/auxiliary/draw/draw_gs_tmp.h
index 7a8683cf7c2..4a17af0dea3 100644
--- a/src/gallium/auxiliary/draw/draw_gs_tmp.h
+++ b/src/gallium/auxiliary/draw/draw_gs_tmp.h
@@ -1,152 +1,34 @@
-
-static void FUNC( struct draw_geometry_shader *shader,
-                  const struct draw_prim_info *input_prims,
-                  const struct draw_vertex_info *input_verts,
-                  struct draw_prim_info *output_prims,
-                  struct draw_vertex_info *output_verts)
-{
-   struct draw_context *draw = shader->draw;
-
-   boolean flatfirst = (draw->rasterizer->flatshade &&
-                        draw->rasterizer->flatshade_first);
-   unsigned i, j;
-   unsigned count = input_prims->count;
-   LOCAL_VARS
-
-   if (0) debug_printf("%s %d\n", __FUNCTION__, count);
-
-   debug_assert(input_prims->primitive_count == 1);
-
-   switch (input_prims->prim) {
-   case PIPE_PRIM_POINTS:
-      for (i = 0; i < count; i++) {
-	 POINT( shader, i + 0 );
-      }
-      break;
-
-   case PIPE_PRIM_LINES:
-      for (i = 0; i+1 < count; i += 2) {
-         LINE( shader , i + 0 , i + 1 );
-      }
-      break;
-
-   case PIPE_PRIM_LINE_LOOP:
-      if (count >= 2) {
-
-         for (i = 1; i < count; i++) {
-            LINE( shader, i - 1, i );
-         }
-
-	 LINE( shader, i - 1, 0 );
-      }
-      break;
-
-   case PIPE_PRIM_LINE_STRIP:
-      for (i = 1; i < count; i++) {
-         LINE( shader, i - 1, i );
-      }
-      break;
-
-   case PIPE_PRIM_TRIANGLES:
-      for (i = 0; i+2 < count; i += 3) {
-         TRIANGLE( shader, i + 0, i + 1, i + 2 );
-      }
-      break;
-
-   case PIPE_PRIM_TRIANGLE_STRIP:
-      if (flatfirst) {
-         for (i = 0; i+2 < count; i++) {
-            TRIANGLE( shader,
-                      i + 0,
-                      i + 1 + (i&1),
-                      i + 2 - (i&1) );
-         }
-      }
-      else {
-         for (i = 0; i+2 < count; i++) {
-            TRIANGLE( shader,
-                      i + 0 + (i&1),
-                      i + 1 - (i&1),
-                      i + 2 );
-         }
-      }
-      break;
-
-   case PIPE_PRIM_TRIANGLE_FAN:
-      if (count >= 3) {
-         if (flatfirst) {
-            for (i = 0; i+2 < count; i++) {
-               TRIANGLE( shader,
-                         i + 1,
-                         i + 2,
-                         0 );
-            }
-         }
-         else {
-            for (i = 0; i+2 < count; i++) {
-               TRIANGLE( shader,
-                         0,
-                         i + 1,
-                         i + 2 );
-            }
-         }
-      }
-      break;
-
-   case PIPE_PRIM_POLYGON:
-      {
-	 for (i = 0; i+2 < count; i++) {
-
-            if (flatfirst) {
-               TRIANGLE( shader, 0, i + 1, i + 2 );
-            }
-            else {
-               TRIANGLE( shader, i + 1, i + 2, 0 );
-            }
-	 }
-      }
-      break;
-
-   case PIPE_PRIM_LINES_ADJACENCY:
-      for (i = 0; i+3 < count; i += 4) {
-         LINE_ADJ( shader , i + 0 , i + 1, i + 2, i + 3 );
-      }
-      break;
-   case PIPE_PRIM_LINE_STRIP_ADJACENCY:
-      for (i = 1; i + 2 < count; i++) {
-         LINE_ADJ( shader, i - 1, i, i + 1, i + 2 );
-      }
-      break;
-
-   case PIPE_PRIM_TRIANGLES_ADJACENCY:
-      for (i = 0; i+5 < count; i += 5) {
-         TRI_ADJ( shader, i + 0, i + 1, i + 2,
-                  i + 3, i + 4, i + 5);
-      }
-      break;
-   case PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY:
-      for (i = 0, j = 0; i+5 < count; i += 2, ++j) {
-         TRI_ADJ( shader,
-                  i + 0,
-                  i + 1 + 2*(j&1),
-                  i + 2 + 2*(j&1),
-                  i + 3 - 2*(j&1),
-                  i + 4 - 2*(j&1),
-                  i + 5);
-      }
-      break;
-
-   default:
-      debug_assert(!"Unsupported primitive in geometry shader");
-      break;
-   }
-}
-
-
-#undef TRIANGLE
-#undef TRI_ADJ
-#undef POINT
-#undef LINE
-#undef LINE_ADJ
-#undef FUNC
-#undef LOCAL_VARS
+#define FUNC_VARS struct draw_geometry_shader *gs,             \
+                  const struct draw_prim_info *input_prims,    \
+                  const struct draw_vertex_info *input_verts,  \
+                  struct draw_prim_info *output_prims,         \
+                  struct draw_vertex_info *output_verts
+
+#define FUNC_ENTER                                                \
+   /* declare more local vars */                                  \
+   struct draw_context *draw = gs->draw;                          \
+   const unsigned prim = input_prims->prim;                       \
+   const unsigned count = input_prims->count;                     \
+   const boolean last_vertex_last =                               \
+      !(draw->rasterizer->flatshade &&                            \
+        draw->rasterizer->flatshade_first);                       \
+   do {                                                           \
+      debug_assert(input_prims->primitive_count == 1);            \
+      switch (prim) {                                             \
+      case PIPE_PRIM_QUADS:                                       \
+      case PIPE_PRIM_QUAD_STRIP:                                  \
+      case PIPE_PRIM_POLYGON:                                     \
+         debug_assert(!"unexpected primitive type in GS");        \
+         return;                                                  \
+      default:                                                    \
+         break;                                                   \
+      }                                                           \
+   } while (0)                                                    \
+
+#define POINT(i0)                             gs_point(gs,i0)
+#define LINE(flags,i0,i1)                     gs_line(gs,i0,i1)
+#define TRIANGLE(flags,i0,i1,i2)              gs_tri(gs,i0,i1,i2)
+#define LINE_ADJ(flags,i0,i1,i2,i3)           gs_line_adj(gs,i0,i1,i2,i3)
+#define TRIANGLE_ADJ(flags,i0,i1,i2,i3,i4,i5) gs_tri_adj(gs,i0,i1,i2,i3,i4,i5)
+
+#include "draw_decompose_tmp.h"
diff --git a/src/gallium/auxiliary/draw/draw_llvm.c b/src/gallium/auxiliary/draw/draw_llvm.c
index 19f96c37ab4..8d53601d195 100644
--- a/src/gallium/auxiliary/draw/draw_llvm.c
+++ b/src/gallium/auxiliary/draw/draw_llvm.c
@@ -37,6 +37,8 @@
 #include "gallivm/lp_bld_debug.h"
 #include "gallivm/lp_bld_tgsi.h"
 #include "gallivm/lp_bld_printf.h"
+#include "gallivm/lp_bld_intr.h"
+#include "gallivm/lp_bld_init.h"
 
 #include "tgsi/tgsi_exec.h"
 #include "tgsi/tgsi_dump.h"
@@ -681,7 +683,6 @@ draw_llvm_generate(struct draw_llvm *llvm, struct draw_llvm_variant *variant)
    unsigned i, j;
    struct lp_build_context bld;
    struct lp_build_loop_state lp_loop;
-   struct lp_type vs_type = lp_type_float_vec(32);
    const int max_vertices = 4;
    LLVMValueRef outputs[PIPE_MAX_SHADER_OUTPUTS][NUM_CHANNELS];
    void *code;
@@ -730,7 +731,7 @@ draw_llvm_generate(struct draw_llvm *llvm, struct draw_llvm_variant *variant)
    builder = LLVMCreateBuilder();
    LLVMPositionBuilderAtEnd(builder, block);
 
-   lp_build_context_init(&bld, builder, vs_type);
+   lp_build_context_init(&bld, builder, lp_type_int(32));
 
    end = lp_build_add(&bld, start, count);
 
@@ -793,6 +794,11 @@ draw_llvm_generate(struct draw_llvm *llvm, struct draw_llvm_variant *variant)
 
    sampler->destroy(sampler);
 
+#ifdef PIPE_ARCH_X86
+   /* Avoid corrupting the FPU stack on 32bit OSes. */
+   lp_build_intrinsic(builder, "llvm.x86.mmx.emms", LLVMVoidType(), NULL, 0);
+#endif
+
    LLVMBuildRetVoid(builder);
 
    LLVMDisposeBuilder(builder);
@@ -820,6 +826,7 @@ draw_llvm_generate(struct draw_llvm *llvm, struct draw_llvm_variant *variant)
    if (gallivm_debug & GALLIVM_DEBUG_ASM) {
       lp_disassemble(code);
    }
+   lp_func_delete_body(variant->function);
 }
 
 
@@ -837,9 +844,7 @@ draw_llvm_generate_elts(struct draw_llvm *llvm, struct draw_llvm_variant *varian
    struct draw_context *draw = llvm->draw;
    unsigned i, j;
    struct lp_build_context bld;
-   struct lp_build_context bld_int;
    struct lp_build_loop_state lp_loop;
-   struct lp_type vs_type = lp_type_float_vec(32);
    const int max_vertices = 4;
    LLVMValueRef outputs[PIPE_MAX_SHADER_OUTPUTS][NUM_CHANNELS];
    LLVMValueRef fetch_max;
@@ -891,8 +896,7 @@ draw_llvm_generate_elts(struct draw_llvm *llvm, struct draw_llvm_variant *varian
    builder = LLVMCreateBuilder();
    LLVMPositionBuilderAtEnd(builder, block);
 
-   lp_build_context_init(&bld, builder, vs_type);
-   lp_build_context_init(&bld_int, builder, lp_type_int(32));
+   lp_build_context_init(&bld, builder, lp_type_int(32));
 
    step = LLVMConstInt(LLVMInt32Type(), max_vertices, 0);
 
@@ -927,7 +931,7 @@ draw_llvm_generate_elts(struct draw_llvm *llvm, struct draw_llvm_variant *varian
          /* make sure we're not out of bounds which can happen
           * if fetch_count % 4 != 0, because on the last iteration
           * a few of the 4 vertex fetches will be out of bounds */
-         true_index = lp_build_min(&bld_int, true_index, fetch_max);
+         true_index = lp_build_min(&bld, true_index, fetch_max);
 
          fetch_ptr = LLVMBuildGEP(builder, fetch_elts,
                                   &true_index, 1, "");
@@ -963,6 +967,11 @@ draw_llvm_generate_elts(struct draw_llvm *llvm, struct draw_llvm_variant *varian
 
    sampler->destroy(sampler);
 
+#ifdef PIPE_ARCH_X86
+   /* Avoid corrupting the FPU stack on 32bit OSes. */
+   lp_build_intrinsic(builder, "llvm.x86.mmx.emms", LLVMVoidType(), NULL, 0);
+#endif
+
    LLVMBuildRetVoid(builder);
 
    LLVMDisposeBuilder(builder);
@@ -990,6 +999,7 @@ draw_llvm_generate_elts(struct draw_llvm *llvm, struct draw_llvm_variant *varian
    if (gallivm_debug & GALLIVM_DEBUG_ASM) {
       lp_disassemble(code);
    }
+   lp_func_delete_body(variant->function_elts);
 }
 
 void
diff --git a/src/gallium/auxiliary/draw/draw_pipe.c b/src/gallium/auxiliary/draw/draw_pipe.c
index 8cd75ecf9a3..58995e07248 100644
--- a/src/gallium/auxiliary/draw/draw_pipe.c
+++ b/src/gallium/auxiliary/draw/draw_pipe.c
@@ -169,77 +169,50 @@ static void do_triangle( struct draw_context *draw,
 /*
  * Set up macros for draw_pt_decompose.h template code.
  * This code uses vertex indexes / elements.
+ *
+ * Flags are needed by the stipple and unfilled stages.  When the two stages
+ * are active, vcache_run_extras is called and the flags are stored in the
+ * higher bits of i0.  Otherwise, flags do not matter.
  */
 
-/* emit first quad vertex as first vertex in triangles */
-#define QUAD_FIRST_PV(i0,i1,i2,i3)              \
-   do_triangle( draw,                           \
-                ( DRAW_PIPE_RESET_STIPPLE |     \
-                  DRAW_PIPE_EDGE_FLAG_0 |       \
-                  DRAW_PIPE_EDGE_FLAG_1 ),      \
-                verts + stride * (elts[i0] & ~DRAW_PIPE_FLAG_MASK),     \
-                verts + stride * (elts[i1] & ~DRAW_PIPE_FLAG_MASK),     \
-                verts + stride * (elts[i2] & ~DRAW_PIPE_FLAG_MASK));    \
-   do_triangle( draw,                           \
-                ( DRAW_PIPE_EDGE_FLAG_1 |       \
-                  DRAW_PIPE_EDGE_FLAG_2 ),      \
-                verts + stride * (elts[i0] & ~DRAW_PIPE_FLAG_MASK),     \
-                verts + stride * (elts[i2] & ~DRAW_PIPE_FLAG_MASK),     \
-                verts + stride * (elts[i3] & ~DRAW_PIPE_FLAG_MASK))
-
-/* emit last quad vertex as last vertex in triangles */
-#define QUAD_LAST_PV(i0,i1,i2,i3)               \
-   do_triangle( draw,                           \
-                ( DRAW_PIPE_RESET_STIPPLE |     \
-                  DRAW_PIPE_EDGE_FLAG_0 |       \
-                  DRAW_PIPE_EDGE_FLAG_2 ),      \
-                verts + stride * (elts[i0] & ~DRAW_PIPE_FLAG_MASK),     \
-                verts + stride * (elts[i1] & ~DRAW_PIPE_FLAG_MASK),     \
-                verts + stride * (elts[i3] & ~DRAW_PIPE_FLAG_MASK));    \
-   do_triangle( draw,                           \
-                ( DRAW_PIPE_EDGE_FLAG_0 |       \
-                  DRAW_PIPE_EDGE_FLAG_1 ),      \
-                verts + stride * (elts[i1] & ~DRAW_PIPE_FLAG_MASK),     \
-                verts + stride * (elts[i2] & ~DRAW_PIPE_FLAG_MASK),     \
-                verts + stride * (elts[i3] & ~DRAW_PIPE_FLAG_MASK))
-
-#define TRIANGLE(flags,i0,i1,i2)                                        \
-   do_triangle( draw,                                                   \
-                elts[i0],  /* flags */                                  \
-                verts + stride * (elts[i0] & ~DRAW_PIPE_FLAG_MASK),     \
-                verts + stride * (elts[i1] & ~DRAW_PIPE_FLAG_MASK),     \
-                verts + stride * (elts[i2] & ~DRAW_PIPE_FLAG_MASK) );
-
-#define LINE(flags,i0,i1)                                       \
-   do_line( draw,                                               \
-            elts[i0],                                           \
-            verts + stride * (elts[i0] & ~DRAW_PIPE_FLAG_MASK), \
-            verts + stride * (elts[i1] & ~DRAW_PIPE_FLAG_MASK) );
+#define TRIANGLE(flags,i0,i1,i2)                                  \
+   do {                                                           \
+      assert(!((i1) & DRAW_PIPE_FLAG_MASK));                      \
+      assert(!((i2) & DRAW_PIPE_FLAG_MASK));                      \
+      do_triangle( draw,                                          \
+                   i0,  /* flags */                               \
+                   verts + stride * (i0 & ~DRAW_PIPE_FLAG_MASK),  \
+                   verts + stride * (i1),                         \
+                   verts + stride * (i2) );                       \
+   } while (0)
+
+#define LINE(flags,i0,i1)                                         \
+   do {                                                           \
+      assert(!((i1) & DRAW_PIPE_FLAG_MASK));                      \
+      do_line( draw,                                              \
+               i0, /* flags */                                    \
+               verts + stride * (i0 & ~DRAW_PIPE_FLAG_MASK),      \
+               verts + stride * (i1) );                           \
+   } while (0)
 
 #define POINT(i0)                               \
-   do_point( draw,                              \
-             verts + stride * (elts[i0] & ~DRAW_PIPE_FLAG_MASK) )
+   do {                                         \
+      assert(!((i0) & DRAW_PIPE_FLAG_MASK));    \
+      do_point( draw, verts + stride * (i0) );  \
+   } while (0)
+
+#define GET_ELT(idx) (elts[idx])
 
-#define FUNC pipe_run
-#define ARGS                                    \
+#define FUNC pipe_run_elts
+#define FUNC_VARS                               \
     struct draw_context *draw,                  \
     unsigned prim,                              \
     struct vertex_header *vertices,             \
     unsigned stride,                            \
-    const ushort *elts
-
-#define LOCAL_VARS                                           \
-   char *verts = (char *)vertices;                           \
-   boolean flatfirst = (draw->rasterizer->flatshade &&       \
-                        draw->rasterizer->flatshade_first);  \
-   unsigned i;                                               \
-   ushort flags
-
-#define FLUSH
+    const ushort *elts,                         \
+    unsigned count
 
 #include "draw_pt_decompose.h"
-#undef ARGS
-#undef LOCAL_VARS
 
 
 
@@ -269,14 +242,29 @@ void draw_pipeline_run( struct draw_context *draw,
         i < prim_info->primitive_count;
         start += prim_info->primitive_lengths[i], i++)
    {
-      unsigned count = prim_info->primitive_lengths[i];
-
-      pipe_run(draw,
-               prim_info->prim,
-               vert_info->verts,
-               vert_info->stride,
-               prim_info->elts + start,
-               count);
+      const unsigned count = prim_info->primitive_lengths[i];
+
+#if DEBUG
+      /* make sure none of the element indexes go outside the vertex buffer */
+      {
+         unsigned max_index = 0x0, i;
+         /* find the largest element index */
+         for (i = 0; i < count; i++) {
+            unsigned int index = (prim_info->elts[start + i]
+                                  & ~DRAW_PIPE_FLAG_MASK);
+            if (index > max_index)
+               max_index = index;
+         }
+         assert(max_index <= vert_info->count);
+      }
+#endif
+
+      pipe_run_elts(draw,
+                    prim_info->prim,
+                    vert_info->verts,
+                    vert_info->stride,
+                    prim_info->elts + start,
+                    count);
    }
 
    draw->pipeline.verts = NULL;
@@ -289,70 +277,30 @@ void draw_pipeline_run( struct draw_context *draw,
  * This code is for non-indexed (aka linear) rendering (no elts).
  */
 
-/* emit first quad vertex as first vertex in triangles */
-#define QUAD_FIRST_PV(i0,i1,i2,i3)                               \
-   do_triangle( draw,                                            \
-                ( DRAW_PIPE_RESET_STIPPLE |                      \
-                  DRAW_PIPE_EDGE_FLAG_0 |                        \
-                  DRAW_PIPE_EDGE_FLAG_1 ),                       \
-                verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK),  \
-                verts + stride * ((i1) & ~DRAW_PIPE_FLAG_MASK),  \
-                verts + stride * ((i2) & ~DRAW_PIPE_FLAG_MASK)); \
-   do_triangle( draw,                                            \
-                ( DRAW_PIPE_EDGE_FLAG_1 |                        \
-                  DRAW_PIPE_EDGE_FLAG_2 ),                       \
-                verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK),  \
-                verts + stride * ((i2) & ~DRAW_PIPE_FLAG_MASK),  \
-                verts + stride * ((i3) & ~DRAW_PIPE_FLAG_MASK))
-
-/* emit last quad vertex as last vertex in triangles */
-#define QUAD_LAST_PV(i0,i1,i2,i3)                                \
-   do_triangle( draw,                                            \
-                ( DRAW_PIPE_RESET_STIPPLE |                      \
-                  DRAW_PIPE_EDGE_FLAG_0 |                        \
-                  DRAW_PIPE_EDGE_FLAG_2 ),                       \
-                verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK),  \
-                verts + stride * ((i1) & ~DRAW_PIPE_FLAG_MASK),  \
-                verts + stride * ((i3) & ~DRAW_PIPE_FLAG_MASK)); \
-   do_triangle( draw,                                            \
-                ( DRAW_PIPE_EDGE_FLAG_0 |                        \
-                  DRAW_PIPE_EDGE_FLAG_1 ),                       \
-                verts + stride * ((i1) & ~DRAW_PIPE_FLAG_MASK),  \
-                verts + stride * ((i2) & ~DRAW_PIPE_FLAG_MASK),  \
-                verts + stride * ((i3) & ~DRAW_PIPE_FLAG_MASK))
-
-#define TRIANGLE(flags,i0,i1,i2)                                 \
-   do_triangle( draw,                                            \
-                flags,  /* flags */                              \
-                verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK),  \
-                verts + stride * ((i1) & ~DRAW_PIPE_FLAG_MASK),  \
-                verts + stride * ((i2) & ~DRAW_PIPE_FLAG_MASK))
-
-#define LINE(flags,i0,i1)                                   \
-   do_line( draw,                                           \
-            flags,                                          \
-            verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK), \
-            verts + stride * ((i1) & ~DRAW_PIPE_FLAG_MASK))
+#define TRIANGLE(flags,i0,i1,i2)       \
+   do_triangle( draw, flags,           \
+                verts + stride * (i0), \
+                verts + stride * (i1), \
+                verts + stride * (i2) )
 
-#define POINT(i0)                               \
-   do_point( draw,                              \
-             verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK) )
+#define LINE(flags,i0,i1)              \
+   do_line( draw, flags,               \
+            verts + stride * (i0),     \
+            verts + stride * (i1) )
 
-#define FUNC pipe_run_linear
-#define ARGS                                    \
-    struct draw_context *draw,                  \
-    unsigned prim,                              \
-    struct vertex_header *vertices,             \
-    unsigned stride
+#define POINT(i0)                      \
+   do_point( draw, verts + stride * (i0) )
 
-#define LOCAL_VARS                                           \
-   char *verts = (char *)vertices;                           \
-   boolean flatfirst = (draw->rasterizer->flatshade &&       \
-                        draw->rasterizer->flatshade_first);  \
-   unsigned i;                                               \
-   ushort flags
 
-#define FLUSH
+#define GET_ELT(idx) (idx)
+
+#define FUNC pipe_run_linear
+#define FUNC_VARS                      \
+    struct draw_context *draw,         \
+    unsigned prim,                     \
+    struct vertex_header *vertices,    \
+    unsigned stride,                   \
+    unsigned count
 
 #include "draw_pt_decompose.h"
 
@@ -378,6 +326,8 @@ void draw_pipeline_run_linear( struct draw_context *draw,
       draw->pipeline.vertex_stride = vert_info->stride;
       draw->pipeline.vertex_count = count;
 
+      assert(count <= vert_info->count);
+
       pipe_run_linear(draw,
                       prim_info->prim,
                       (struct vertex_header*)verts,
diff --git a/src/gallium/auxiliary/draw/draw_pipe_clip.c b/src/gallium/auxiliary/draw/draw_pipe_clip.c
index 1cf6ee7a7f9..8a3d499febb 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_clip.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_clip.c
@@ -68,8 +68,7 @@ struct clip_stage {
 };
 
 
-/* This is a bit confusing:
- */
+/** Cast wrapper */
 static INLINE struct clip_stage *clip_stage( struct draw_stage *stage )
 {
    return (struct clip_stage *)stage;
@@ -81,18 +80,22 @@ static INLINE struct clip_stage *clip_stage( struct draw_stage *stage )
 
 /* All attributes are float[4], so this is easy:
  */
-static void interp_attr( float *fdst,
+static void interp_attr( float dst[4],
 			 float t,
-			 const float *fin,
-			 const float *fout )
+			 const float in[4],
+			 const float out[4] )
 {  
-   fdst[0] = LINTERP( t, fout[0], fin[0] );
-   fdst[1] = LINTERP( t, fout[1], fin[1] );
-   fdst[2] = LINTERP( t, fout[2], fin[2] );
-   fdst[3] = LINTERP( t, fout[3], fin[3] );
+   dst[0] = LINTERP( t, out[0], in[0] );
+   dst[1] = LINTERP( t, out[1], in[1] );
+   dst[2] = LINTERP( t, out[2], in[2] );
+   dst[3] = LINTERP( t, out[3], in[3] );
 }
 
 
+/**
+ * Copy front/back, primary/secondary colors from src vertex to dst vertex.
+ * Used when flat shading.
+ */
 static void copy_colors( struct draw_stage *stage,
 			 struct vertex_header *dst,
 			 const struct vertex_header *src )
@@ -121,20 +124,17 @@ static void interp( const struct clip_stage *clip,
 
    /* Vertex header.
     */
-   {
-      dst->clipmask = 0;
-      dst->edgeflag = 0;        /* will get overwritten later */
-      dst->pad = 0;
-      dst->vertex_id = UNDEFINED_VERTEX_ID;
-   }
+   dst->clipmask = 0;
+   dst->edgeflag = 0;        /* will get overwritten later */
+   dst->pad = 0;
+   dst->vertex_id = UNDEFINED_VERTEX_ID;
 
-   /* Clip coordinates:  interpolate normally
+   /* Interpolate the clip-space coords.
     */
-   {
-      interp_attr(dst->clip, t, in->clip, out->clip);
-   }
+   interp_attr(dst->clip, t, in->clip, out->clip);
 
-   /* Do the projective divide and insert window coordinates:
+   /* Do the projective divide and viewport transformation to get
+    * new window coordinates:
     */
    {
       const float *pos = dst->clip;
diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h
index 058aeedc17a..397d4bf653c 100644
--- a/src/gallium/auxiliary/draw/draw_private.h
+++ b/src/gallium/auxiliary/draw/draw_private.h
@@ -163,9 +163,11 @@ struct draw_context
          /** vertex arrays */
          const void *vbuffer[PIPE_MAX_ATTRIBS];
          
-         /** constant buffer (for vertex/geometry shader) */
+         /** constant buffers (for vertex/geometry shader) */
          const void *vs_constants[PIPE_MAX_CONSTANT_BUFFERS];
+         unsigned vs_constants_size[PIPE_MAX_CONSTANT_BUFFERS];
          const void *gs_constants[PIPE_MAX_CONSTANT_BUFFERS];
+         unsigned gs_constants_size[PIPE_MAX_CONSTANT_BUFFERS];
       } user;
 
       boolean test_fse;         /* enable FSE even though its not correct (eg for softpipe) */
@@ -198,6 +200,7 @@ struct draw_context
    struct pipe_viewport_state viewport;
    boolean identity_viewport;
 
+   /** Vertex shader state */
    struct {
       struct draw_vertex_shader *vertex_shader;
       uint num_vs_outputs;  /**< convenience, from vertex_shader */
@@ -227,6 +230,7 @@ struct draw_context
       struct translate_cache *emit_cache;
    } vs;
 
+   /** Geometry shader state */
    struct {
       struct draw_geometry_shader *geometry_shader;
       uint num_gs_outputs;  /**< convenience, from geometry_shader */
@@ -239,6 +243,7 @@ struct draw_context
       struct tgsi_sampler **samplers;
    } gs;
 
+   /** Stream output (vertex feedback) state */
    struct {
       struct pipe_stream_output_state state;
       void *buffers[PIPE_MAX_SO_BUFFERS];
diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c
index 92d4113b4c6..248927505da 100644
--- a/src/gallium/auxiliary/draw/draw_pt.c
+++ b/src/gallium/auxiliary/draw/draw_pt.c
@@ -259,6 +259,12 @@ draw_print_arrays(struct draw_context *draw, uint prim, int start, uint count)
       for (j = 0; j < draw->pt.nr_vertex_elements; j++) {
          uint buf = draw->pt.vertex_element[j].vertex_buffer_index;
          ubyte *ptr = (ubyte *) draw->pt.user.vbuffer[buf];
+
+         if (draw->pt.vertex_element[j].instance_divisor) {
+            ii = draw->instance_id / draw->pt.vertex_element[j].instance_divisor;
+         }
+
+         ptr += draw->pt.vertex_buffer[buf].buffer_offset;
          ptr += draw->pt.vertex_buffer[buf].stride * ii;
          ptr += draw->pt.vertex_element[j].src_offset;
 
@@ -341,19 +347,22 @@ draw_arrays_instanced(struct draw_context *draw,
    unsigned reduced_prim = u_reduced_prim(mode);
    unsigned instance;
 
+   assert(instanceCount > 0);
+
    if (reduced_prim != draw->reduced_prim) {
       draw_do_flush(draw, DRAW_FLUSH_STATE_CHANGE);
       draw->reduced_prim = reduced_prim;
    }
 
    if (0)
-      draw_print_arrays(draw, mode, start, MIN2(count, 20));
-
-   if (0) {
-      unsigned int i;
       debug_printf("draw_arrays(mode=%u start=%u count=%u):\n",
                    mode, start, count);
+
+   if (0)
       tgsi_dump(draw->vs.vertex_shader->state.tokens, 0);
+
+   if (0) {
+      unsigned int i;
       debug_printf("Elements:\n");
       for (i = 0; i < draw->pt.nr_vertex_elements; i++) {
          debug_printf("  %u: src_offset=%u  inst_div=%u   vbuf=%u  format=%s\n",
@@ -374,6 +383,9 @@ draw_arrays_instanced(struct draw_context *draw,
       }
    }
 
+   if (0)
+      draw_print_arrays(draw, mode, start, MIN2(count, 20));
+
    for (instance = 0; instance < instanceCount; instance++) {
       draw->instance_id = instance + startInstance;
       draw_pt_arrays(draw, mode, start, count);
diff --git a/src/gallium/auxiliary/draw/draw_pt_decompose.h b/src/gallium/auxiliary/draw/draw_pt_decompose.h
index 52f9593d46e..3127aad7310 100644
--- a/src/gallium/auxiliary/draw/draw_pt_decompose.h
+++ b/src/gallium/auxiliary/draw/draw_pt_decompose.h
@@ -1,194 +1,7 @@
+#define LOCAL_VARS                           \
+   char *verts = (char *) vertices;          \
+   const boolean last_vertex_last =          \
+      !(draw->rasterizer->flatshade &&       \
+        draw->rasterizer->flatshade_first);
 
-
-static void FUNC( ARGS,
-                  unsigned count )
-{
-   LOCAL_VARS;
-
-   switch (prim) {
-   case PIPE_PRIM_POINTS:
-      for (i = 0; i < count; i ++) {
-	 POINT( (i + 0) );
-      }
-      break;
-
-   case PIPE_PRIM_LINES:
-      for (i = 0; i+1 < count; i += 2) {
-         LINE( DRAW_PIPE_RESET_STIPPLE,
-               (i + 0),
-               (i + 1));
-      }
-      break;
-
-   case PIPE_PRIM_LINE_LOOP:
-      if (count >= 2) {
-         flags = DRAW_PIPE_RESET_STIPPLE;
-
-         for (i = 1; i < count; i++, flags = 0) {
-            LINE( flags,
-                  (i - 1),
-                  (i ));
-         }
-
-	 LINE( flags,
-               (i - 1),
-               (0 ));
-      }
-      break;
-
-   case PIPE_PRIM_LINE_STRIP:
-      flags = DRAW_PIPE_RESET_STIPPLE;
-      for (i = 1; i < count; i++, flags = 0) {
-         LINE( flags,
-               (i - 1),
-               (i ));
-      }
-      break;
-
-   case PIPE_PRIM_TRIANGLES:
-      for (i = 0; i+2 < count; i += 3) {
-         TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
-                   (i + 0),
-                   (i + 1),
-                   (i + 2 ));
-      }
-      break;
-
-   case PIPE_PRIM_TRIANGLE_STRIP:
-      if (flatfirst) {
-         for (i = 0; i+2 < count; i++) {
-            /* Emit first triangle vertex as first triangle vertex */
-            TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
-                      (i + 0),
-                      (i + 1 + (i&1)),
-                      (i + 2 - (i&1)) );
-         }
-      }
-      else {
-         for (i = 0; i+2 < count; i++) {
-            /* Emit last triangle vertex as last triangle vertex */
-            TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
-                      (i + 0 + (i&1)),
-                      (i + 1 - (i&1)),
-                      (i + 2 ));
-         }
-      }
-      break;
-
-   case PIPE_PRIM_TRIANGLE_FAN:
-      if (count >= 3) {
-         if (flatfirst) {
-            for (i = 0; i+2 < count; i++) {
-               TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
-                         (i + 1),
-                         (i + 2),
-                         0 );
-            }
-         }
-         else {
-            for (i = 0; i+2 < count; i++) {
-               TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL,
-                         (0),
-                         (i + 1),
-                         (i + 2 ));
-            }
-         }
-      }
-      break;
-
-
-   case PIPE_PRIM_QUADS:
-      /* GL quads don't follow provoking vertex convention */
-      if (flatfirst) {
-         for (i = 0; i+3 < count; i += 4) {
-            /* emit last quad vertex as first triangle vertex */
-            QUAD_FIRST_PV( (i + 3),
-                           (i + 0),
-                           (i + 1),
-                           (i + 2) );
-         }
-      }
-      else {
-         for (i = 0; i+3 < count; i += 4) {
-            /* emit last quad vertex as last triangle vertex */
-            QUAD_LAST_PV( (i + 0),
-                          (i + 1),
-                          (i + 2),
-                          (i + 3) );
-         }
-      }
-      break;
-
-   case PIPE_PRIM_QUAD_STRIP:
-      /* GL quad strips don't follow provoking vertex convention */
-      if (flatfirst) {
-         for (i = 0; i+3 < count; i += 2) {
-            /* emit last quad vertex as first triangle vertex */
-            QUAD_FIRST_PV( (i + 3),
-                           (i + 2),
-                           (i + 0),
-                           (i + 1) );
-
-         }
-      }
-      else {
-         for (i = 0; i+3 < count; i += 2) {
-            /* emit last quad vertex as last triangle vertex */
-            QUAD_LAST_PV( (i + 2),
-                          (i + 0),
-                          (i + 1),
-                          (i + 3) );
-         }
-      }
-      break;
-
-   case PIPE_PRIM_POLYGON:
-      /* GL polygons don't follow provoking vertex convention */
-      {
-         /* These bitflags look a little odd because we submit the
-          * vertices as (1,2,0) to satisfy flatshade requirements.
-          */
-         const ushort edge_first  = DRAW_PIPE_EDGE_FLAG_2;
-         const ushort edge_middle = DRAW_PIPE_EDGE_FLAG_0;
-         const ushort edge_last   = DRAW_PIPE_EDGE_FLAG_1;
-
-         flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle;
-
-	 for (i = 0; i+2 < count; i++, flags = edge_middle) {
-
-            if (i + 3 == count)
-               flags |= edge_last;
-
-            if (flatfirst) {
-               /* emit first polygon vertex as first triangle vertex */
-               TRIANGLE( flags,
-                         (0),
-                         (i + 1),
-                         (i + 2) );
-            }
-            else {
-               /* emit first polygon vertex as last triangle vertex */
-               TRIANGLE( flags,
-                         (i + 1),
-                         (i + 2),
-                         (0));
-            }
-	 }
-      }
-      break;
-
-   default:
-      assert(0);
-      break;
-   }
-
-   FLUSH;
-}
-
-
-#undef TRIANGLE
-#undef QUAD_FIRST_PV
-#undef QUAD_LAST_PV
-#undef POINT
-#undef LINE
-#undef FUNC
+#include "draw_decompose_tmp.h"
diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c
index 0229bcc7fe1..5568fbb9f88 100644
--- a/src/gallium/auxiliary/draw/draw_pt_emit.c
+++ b/src/gallium/auxiliary/draw/draw_pt_emit.c
@@ -182,6 +182,7 @@ void draw_pt_emit( struct pt_emit *emit,
 			 0,
 			 ~0);
 
+   /* fetch/translate vertex attribs to fill hw_verts[] */
    translate->run( translate,
 		   0, 
 		   vertex_count,
diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c
index 121dfc414a4..5b16c3788e5 100644
--- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c
+++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c
@@ -176,6 +176,7 @@ static void emit(struct pt_emit *emit,
 
 static void draw_vertex_shader_run(struct draw_vertex_shader *vshader,
                                    const void *constants[PIPE_MAX_CONSTANT_BUFFERS], 
+                                   unsigned const_size[PIPE_MAX_CONSTANT_BUFFERS],
                                    const struct draw_vertex_info *input_verts,
                                    struct draw_vertex_info *output_verts )
 {
@@ -190,6 +191,7 @@ static void draw_vertex_shader_run(struct draw_vertex_shader *vshader,
                        (const float (*)[4])input_verts->verts->data,
                        (      float (*)[4])output_verts->verts->data,
                        constants,
+                       const_size,
                        input_verts->count,
                        input_verts->vertex_size,
                        input_verts->vertex_size);
@@ -236,6 +238,7 @@ static void fetch_pipeline_generic( struct draw_pt_middle_end *middle,
    if (fpme->opt & PT_SHADE) {
       draw_vertex_shader_run(vshader,
                              draw->pt.user.vs_constants,
+                             draw->pt.user.vs_constants_size,
                              vert_info,
                              &vs_vert_info);
 
@@ -246,6 +249,7 @@ static void fetch_pipeline_generic( struct draw_pt_middle_end *middle,
    if ((fpme->opt & PT_SHADE) && gshader) {
       draw_geometry_shader_run(gshader,
                                draw->pt.user.gs_constants,
+                               draw->pt.user.gs_constants_size,
                                vert_info,
                                prim_info,
                                &gs_vert_info,
diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline_llvm.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline_llvm.c
index bc074df8c2a..4b99bee86a0 100644
--- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline_llvm.c
+++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline_llvm.c
@@ -1,6 +1,6 @@
 /**************************************************************************
  *
- * Copyright 2010 VMWare, Inc.
+ * Copyright 2010 VMware, Inc.
  * All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -254,6 +254,7 @@ llvm_pipeline_generic( struct draw_pt_middle_end *middle,
    if ((opt & PT_SHADE) && gshader) {
       draw_geometry_shader_run(gshader,
                                draw->pt.user.gs_constants,
+                               draw->pt.user.gs_constants_size,
                                vert_info,
                                prim_info,
                                &gs_vert_info,
diff --git a/src/gallium/auxiliary/draw/draw_pt_so_emit.c b/src/gallium/auxiliary/draw/draw_pt_so_emit.c
index 5d82934889b..f7f4f24d354 100644
--- a/src/gallium/auxiliary/draw/draw_pt_so_emit.c
+++ b/src/gallium/auxiliary/draw/draw_pt_so_emit.c
@@ -218,25 +218,15 @@ static void so_tri(struct pt_so_emit *so, int i0, int i1, int i2)
 }
 
 
-#define TRIANGLE(gs,i0,i1,i2) so_tri(so,i0,i1,i2)
-#define LINE(gs,i0,i1)        so_line(so,i0,i1)
-#define POINT(gs,i0)          so_point(so,i0)
-#define FUNC so_run_linear
-#define LOCAL_VARS
+#define FUNC         so_run_linear
+#define GET_ELT(idx) (start + (idx))
 #include "draw_so_emit_tmp.h"
-#undef LOCAL_VARS
-#undef FUNC
 
 
-#define TRIANGLE(gs,i0,i1,i2) so_tri(gs,elts[i0],elts[i1],elts[i2])
-#define LINE(gs,i0,i1)        so_line(gs,elts[i0],elts[i1])
-#define POINT(gs,i0)          so_point(gs,elts[i0])
-#define FUNC so_run_elts
-#define LOCAL_VARS                         \
-   const ushort *elts = input_prims->elts;
+#define FUNC         so_run_elts
+#define LOCAL_VARS   const ushort *elts = input_prims->elts;
+#define GET_ELT(idx) (elts[start + (idx)] & ~DRAW_PIPE_FLAG_MASK)
 #include "draw_so_emit_tmp.h"
-#undef LOCAL_VARS
-#undef FUNC
 
 
 void draw_pt_so_emit( struct pt_so_emit *emit,
diff --git a/src/gallium/auxiliary/draw/draw_pt_util.c b/src/gallium/auxiliary/draw/draw_pt_util.c
index 3236d38e6ab..182a597cca2 100644
--- a/src/gallium/auxiliary/draw/draw_pt_util.c
+++ b/src/gallium/auxiliary/draw/draw_pt_util.c
@@ -53,7 +53,7 @@ void draw_pt_split_prim(unsigned prim, unsigned *first, unsigned *incr)
       break;
    case PIPE_PRIM_LINES_ADJACENCY:
       *first = 4;
-      *incr = 2;
+      *incr = 4;
       break;
    case PIPE_PRIM_LINE_STRIP_ADJACENCY:
       *first = 4;
@@ -65,7 +65,7 @@ void draw_pt_split_prim(unsigned prim, unsigned *first, unsigned *incr)
       break;
    case PIPE_PRIM_TRIANGLES_ADJACENCY:
       *first = 6;
-      *incr = 3;
+      *incr = 6;
       break;
    case PIPE_PRIM_TRIANGLE_STRIP:
    case PIPE_PRIM_TRIANGLE_FAN:
@@ -75,7 +75,7 @@ void draw_pt_split_prim(unsigned prim, unsigned *first, unsigned *incr)
       break;
    case PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY:
       *first = 6;
-      *incr = 1;
+      *incr = 2;
       break;
    case PIPE_PRIM_QUADS:
       *first = 4;
diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h
index a292346be95..55e43b2a714 100644
--- a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h
+++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h
@@ -1,6 +1,11 @@
 static unsigned trim( unsigned count, unsigned first, unsigned incr )
 {
-   return count - (count - first) % incr; 
+   /*
+    * count either has been trimmed in draw_pt_arrays or is set to
+    * (driver)_fetch_max which is hopefully always larger than first.
+    */
+   assert(count >= first);
+   return count - (count - first) % incr;
 }
 
 static void FUNC(struct draw_pt_front_end *frontend,
diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c
index 8ef94c3163c..a848b54f7d2 100644
--- a/src/gallium/auxiliary/draw/draw_pt_vcache.c
+++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c
@@ -95,7 +95,7 @@ static INLINE void
 vcache_check_flush( struct vcache_frontend *vcache )
 {
    if (vcache->draw_count + 6 >= DRAW_MAX ||
-       vcache->fetch_count + 4 >= FETCH_MAX) {
+       vcache->fetch_count + 6 >= FETCH_MAX) {
       vcache_flush( vcache );
    }
 }
@@ -180,59 +180,61 @@ vcache_point( struct vcache_frontend *vcache,
 }
 
 
-static INLINE void 
-vcache_quad( struct vcache_frontend *vcache,
-             unsigned i0,
-             unsigned i1,
-             unsigned i2,
-             unsigned i3 )
+static INLINE void
+vcache_line_adj_flags( struct vcache_frontend *vcache,
+                       unsigned flags,
+                       unsigned a0, unsigned i0, unsigned i1, unsigned a1 )
 {
-   if (vcache->draw->rasterizer->flatshade_first) {
-      /* pass last quad vertex as first triangle vertex */
-      vcache_triangle( vcache, i3, i0, i1 );
-      vcache_triangle( vcache, i3, i1, i2 );
-   }
-   else {
-      /* pass last quad vertex as last triangle vertex */
-      vcache_triangle( vcache, i0, i1, i3 );
-      vcache_triangle( vcache, i1, i2, i3 );
-   }
+   vcache_elt(vcache, a0, 0);
+   vcache_elt(vcache, i0, flags);
+   vcache_elt(vcache, i1, 0);
+   vcache_elt(vcache, a1, 0);
+   vcache_check_flush(vcache);
 }
 
 
-static INLINE void 
-vcache_ef_quad( struct vcache_frontend *vcache,
-                unsigned i0,
-                unsigned i1,
-                unsigned i2,
-                unsigned i3 )
+static INLINE void
+vcache_line_adj( struct vcache_frontend *vcache,
+                 unsigned a0, unsigned i0, unsigned i1, unsigned a1 )
 {
-   if (vcache->draw->rasterizer->flatshade_first) {
-      /* pass last quad vertex as first triangle vertex */
-      vcache_triangle_flags( vcache,
-                             ( DRAW_PIPE_RESET_STIPPLE |
-                               DRAW_PIPE_EDGE_FLAG_0 |
-                               DRAW_PIPE_EDGE_FLAG_1 ),
-                             i3, i0, i1 );
-
-      vcache_triangle_flags( vcache,
-                             ( DRAW_PIPE_EDGE_FLAG_1 |
-                               DRAW_PIPE_EDGE_FLAG_2 ),
-                             i3, i1, i2 );
-   }
-   else {
-      /* pass last quad vertex as last triangle vertex */
-      vcache_triangle_flags( vcache,
-                             ( DRAW_PIPE_RESET_STIPPLE |
-                               DRAW_PIPE_EDGE_FLAG_0 |
-                               DRAW_PIPE_EDGE_FLAG_2 ),
-                             i0, i1, i3 );
-
-      vcache_triangle_flags( vcache,
-                             ( DRAW_PIPE_EDGE_FLAG_0 |
-                               DRAW_PIPE_EDGE_FLAG_1 ),
-                             i1, i2, i3 );
-   }
+   vcache_elt(vcache, a0, 0);
+   vcache_elt(vcache, i0, 0);
+   vcache_elt(vcache, i1, 0);
+   vcache_elt(vcache, a1, 0);
+   vcache_check_flush(vcache);
+}
+
+
+static INLINE void
+vcache_triangle_adj_flags( struct vcache_frontend *vcache,
+                           unsigned flags,
+                           unsigned i0, unsigned a0,
+                           unsigned i1, unsigned a1,
+                           unsigned i2, unsigned a2 )
+{
+   vcache_elt(vcache, i0, flags);
+   vcache_elt(vcache, a0, 0);
+   vcache_elt(vcache, i1, 0);
+   vcache_elt(vcache, a1, 0);
+   vcache_elt(vcache, i2, 0);
+   vcache_elt(vcache, a2, 0);
+   vcache_check_flush(vcache);
+}
+
+
+static INLINE void
+vcache_triangle_adj( struct vcache_frontend *vcache,
+                     unsigned i0, unsigned a0,
+                     unsigned i1, unsigned a1,
+                     unsigned i2, unsigned a2 )
+{
+   vcache_elt(vcache, i0, 0);
+   vcache_elt(vcache, a0, 0);
+   vcache_elt(vcache, i1, 0);
+   vcache_elt(vcache, a1, 0);
+   vcache_elt(vcache, i2, 0);
+   vcache_elt(vcache, a2, 0);
+   vcache_check_flush(vcache);
 }
 
 
@@ -240,17 +242,23 @@ vcache_ef_quad( struct vcache_frontend *vcache,
  * this.  The two paths aren't too different though - it may be
  * possible to reunify them.
  */
-#define TRIANGLE(vc,flags,i0,i1,i2) vcache_triangle_flags(vc,flags,i0,i1,i2)
-#define QUAD(vc,i0,i1,i2,i3)        vcache_ef_quad(vc,i0,i1,i2,i3)
-#define LINE(vc,flags,i0,i1)        vcache_line_flags(vc,flags,i0,i1)
-#define POINT(vc,i0)                vcache_point(vc,i0)
+#define TRIANGLE(flags,i0,i1,i2) vcache_triangle_flags(vcache,flags,i0,i1,i2)
+#define LINE(flags,i0,i1)        vcache_line_flags(vcache,flags,i0,i1)
+#define POINT(i0)                vcache_point(vcache,i0)
+#define LINE_ADJ(flags,a0,i0,i1,a1) \
+   vcache_line_adj_flags(vcache,flags,a0,i0,i1,a1)
+#define TRIANGLE_ADJ(flags,i0,a0,i1,a1,i2,a2) \
+   vcache_triangle_adj_flags(vcache,flags,i0,a0,i1,a1,i2,a2)
 #define FUNC vcache_run_extras
 #include "draw_pt_vcache_tmp.h"
 
-#define TRIANGLE(vc,flags,i0,i1,i2) vcache_triangle(vc,i0,i1,i2)
-#define QUAD(vc,i0,i1,i2,i3)        vcache_quad(vc,i0,i1,i2,i3)
-#define LINE(vc,flags,i0,i1)        vcache_line(vc,i0,i1)
-#define POINT(vc,i0)                vcache_point(vc,i0)
+#define TRIANGLE(flags,i0,i1,i2) vcache_triangle(vcache,i0,i1,i2)
+#define LINE(flags,i0,i1)        vcache_line(vcache,i0,i1)
+#define POINT(i0)                vcache_point(vcache,i0)
+#define LINE_ADJ(flags,a0,i0,i1,a1) \
+   vcache_line_adj(vcache,a0,i0,i1,a1)
+#define TRIANGLE_ADJ(flags,i0,a0,i1,a1,i2,a2) \
+   vcache_triangle_adj(vcache,i0,a0,i1,a1,i2,a2)
 #define FUNC vcache_run
 #include "draw_pt_vcache_tmp.h"
 
@@ -339,6 +347,25 @@ format_from_get_elt( pt_elt_func get_elt )
 #endif
 
 
+/**
+ * Check if any vertex attributes use instance divisors.
+ * Note that instance divisors complicate vertex fetching so we need
+ * to take the vcache path when they're in use.
+ */
+static boolean
+any_instance_divisors(const struct draw_context *draw)
+{
+   uint i;
+
+   for (i = 0; i < draw->pt.nr_vertex_elements; i++) {
+      uint div = draw->pt.vertex_element[i].instance_divisor;
+      if (div)
+         return TRUE;
+   }
+   return FALSE;
+}
+
+
 static INLINE void 
 vcache_check_run( struct draw_pt_front_end *frontend, 
                   pt_elt_func get_elt,
@@ -382,6 +409,9 @@ vcache_check_run( struct draw_pt_front_end *frontend,
    if (max_index >= (unsigned) DRAW_PIPE_MAX_VERTICES)
       goto fail;
 
+   if (any_instance_divisors(draw))
+      goto fail;
+
    fetch_count = max_index + 1 - min_index;
 
    if (0)
@@ -518,7 +548,18 @@ vcache_prepare( struct draw_pt_front_end *frontend,
     * which is a separate issue.
     */
    vcache->input_prim = in_prim;
-   vcache->output_prim = u_reduced_prim(in_prim);
+   switch (in_prim) {
+   case PIPE_PRIM_LINES_ADJACENCY:
+   case PIPE_PRIM_LINE_STRIP_ADJACENCY:
+      vcache->output_prim = PIPE_PRIM_LINES_ADJACENCY;
+      break;
+   case PIPE_PRIM_TRIANGLES_ADJACENCY:
+   case PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY:
+      vcache->output_prim = PIPE_PRIM_TRIANGLES_ADJACENCY;
+      break;
+   default:
+      vcache->output_prim = u_reduced_prim(in_prim);
+   }
 
    vcache->middle = middle;
    vcache->opt = opt;
diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h b/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h
index dac68ad4398..1a3748d5f0b 100644
--- a/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h
+++ b/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h
@@ -1,198 +1,19 @@
+#define FUNC_VARS                      \
+   struct draw_pt_front_end *frontend, \
+   pt_elt_func get_elt,                \
+   const void *elts,                   \
+   int elt_bias,                       \
+   unsigned count
 
+#define LOCAL_VARS \
+   struct vcache_frontend *vcache = (struct vcache_frontend *) frontend;   \
+   struct draw_context *draw = vcache->draw;                               \
+   const unsigned prim = vcache->input_prim;                               \
+   const boolean last_vertex_last = !(draw->rasterizer->flatshade &&       \
+                                      draw->rasterizer->flatshade_first);
 
-static void FUNC( struct draw_pt_front_end *frontend, 
-                  pt_elt_func get_elt,
-                  const void *elts,
-                  int elt_bias,
-                  unsigned count )
-{
-   struct vcache_frontend *vcache = (struct vcache_frontend *)frontend;
-   struct draw_context *draw = vcache->draw;
+#define GET_ELT(idx) (get_elt(elts, idx) + elt_bias)
 
-   boolean flatfirst = (draw->rasterizer->flatshade && 
-                        draw->rasterizer->flatshade_first);
-   unsigned i;
-   ushort flags;
+#define FUNC_EXIT do { vcache_flush(vcache); } while (0)
 
-   if (0) debug_printf("%s %d\n", __FUNCTION__, count);
-
-
-   switch (vcache->input_prim) {
-   case PIPE_PRIM_POINTS:
-      for (i = 0; i < count; i ++) {
-	 POINT( vcache,
-                get_elt(elts, i + 0) + elt_bias );
-      }
-      break;
-
-   case PIPE_PRIM_LINES:
-      for (i = 0; i+1 < count; i += 2) {
-         LINE( vcache, 
-               DRAW_PIPE_RESET_STIPPLE,
-               get_elt(elts, i + 0) + elt_bias,
-               get_elt(elts, i + 1) + elt_bias);
-      }
-      break;
-
-   case PIPE_PRIM_LINE_LOOP:  
-      if (count >= 2) {
-         flags = DRAW_PIPE_RESET_STIPPLE;
-
-         for (i = 1; i < count; i++, flags = 0) {
-            LINE( vcache, 
-                  flags,
-                  get_elt(elts, i - 1) + elt_bias,
-                  get_elt(elts, i ) + elt_bias);
-         }
-
-	 LINE( vcache, 
-               flags,
-               get_elt(elts, i - 1) + elt_bias,
-               get_elt(elts, 0 ) + elt_bias);
-      }
-      break;
-
-   case PIPE_PRIM_LINE_STRIP:
-      flags = DRAW_PIPE_RESET_STIPPLE;
-      for (i = 1; i < count; i++, flags = 0) {
-         LINE( vcache, 
-               flags,
-               get_elt(elts, i - 1) + elt_bias,
-               get_elt(elts, i ) + elt_bias);
-      }
-      break;
-
-   case PIPE_PRIM_TRIANGLES:
-      for (i = 0; i+2 < count; i += 3) {
-         TRIANGLE( vcache,
-                   DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, 
-                   get_elt(elts, i + 0) + elt_bias,
-                   get_elt(elts, i + 1) + elt_bias,
-                   get_elt(elts, i + 2 ) + elt_bias);
-      }
-      break;
-
-   case PIPE_PRIM_TRIANGLE_STRIP:
-      if (flatfirst) {
-         for (i = 0; i+2 < count; i++) {
-            TRIANGLE( vcache,
-                      DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, 
-                      get_elt(elts, i + 0) + elt_bias,
-                      get_elt(elts, i + 1 + (i&1)) + elt_bias,
-                      get_elt(elts, i + 2 - (i&1)) + elt_bias);
-         }
-      }
-      else {
-         for (i = 0; i+2 < count; i++) {
-            TRIANGLE( vcache,
-                      DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, 
-                      get_elt(elts, i + 0 + (i&1)) + elt_bias,
-                      get_elt(elts, i + 1 - (i&1)) + elt_bias,
-                      get_elt(elts, i + 2 ) + elt_bias);
-         }
-      }
-      break;
-
-   case PIPE_PRIM_TRIANGLE_FAN:
-      if (count >= 3) {
-         if (flatfirst) {
-            for (i = 0; i+2 < count; i++) {
-               TRIANGLE( vcache,
-                         DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, 
-                         get_elt(elts, i + 1) + elt_bias,
-                         get_elt(elts, i + 2) + elt_bias,
-                         get_elt(elts, 0 ) + elt_bias);
-            }
-         }
-         else {
-            for (i = 0; i+2 < count; i++) {
-               TRIANGLE( vcache,
-                         DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, 
-                         get_elt(elts, 0) + elt_bias,
-                         get_elt(elts, i + 1) + elt_bias,
-                         get_elt(elts, i + 2 ) + elt_bias);
-            }
-         }
-      }
-      break;
-
-
-   case PIPE_PRIM_QUADS:
-      for (i = 0; i+3 < count; i += 4) {
-         QUAD( vcache,
-               get_elt(elts, i + 0) + elt_bias,
-               get_elt(elts, i + 1) + elt_bias,
-               get_elt(elts, i + 2) + elt_bias,
-               get_elt(elts, i + 3) + elt_bias );
-      }
-      break;
-
-   case PIPE_PRIM_QUAD_STRIP:
-      for (i = 0; i+3 < count; i += 2) {
-         QUAD( vcache,
-               get_elt(elts, i + 2) + elt_bias,
-               get_elt(elts, i + 0) + elt_bias,
-               get_elt(elts, i + 1) + elt_bias,
-               get_elt(elts, i + 3) + elt_bias );
-      }
-      break;
-
-   case PIPE_PRIM_POLYGON:
-      {
-         /* These bitflags look a little odd because we submit the
-          * vertices as (1,2,0) to satisfy flatshade requirements.  
-          */
-         ushort edge_next, edge_finish;
-
-         if (flatfirst) {
-            flags = (DRAW_PIPE_RESET_STIPPLE |
-                     DRAW_PIPE_EDGE_FLAG_1 |
-                     DRAW_PIPE_EDGE_FLAG_2);
-            edge_next = DRAW_PIPE_EDGE_FLAG_2;
-            edge_finish = DRAW_PIPE_EDGE_FLAG_0;
-         }
-         else {
-            flags = (DRAW_PIPE_RESET_STIPPLE |
-                     DRAW_PIPE_EDGE_FLAG_2 |
-                     DRAW_PIPE_EDGE_FLAG_0);
-            edge_next = DRAW_PIPE_EDGE_FLAG_0;
-            edge_finish = DRAW_PIPE_EDGE_FLAG_1;
-         }
-
-	 for (i = 0; i+2 < count; i++, flags = edge_next) {
-
-            if (i + 3 == count)
-               flags |= edge_finish;
-
-            if (flatfirst) {
-               TRIANGLE( vcache,
-                         flags,
-                         get_elt(elts, 0) + elt_bias,
-                         get_elt(elts, i + 1) + elt_bias,
-                         get_elt(elts, i + 2) + elt_bias );
-            }
-            else {
-               TRIANGLE( vcache,
-                         flags,
-                         get_elt(elts, i + 1) + elt_bias,
-                         get_elt(elts, i + 2) + elt_bias,
-                         get_elt(elts, 0) + elt_bias);
-            }
-	 }
-      }
-      break;
-
-   default:
-      assert(0);
-      break;
-   }
-   
-   vcache_flush( vcache );
-}
-
-
-#undef TRIANGLE
-#undef QUAD
-#undef POINT
-#undef LINE
-#undef FUNC
+#include "draw_decompose_tmp.h"
diff --git a/src/gallium/auxiliary/draw/draw_so_emit_tmp.h b/src/gallium/auxiliary/draw/draw_so_emit_tmp.h
index 01212a8e536..6d8937a0b41 100644
--- a/src/gallium/auxiliary/draw/draw_so_emit_tmp.h
+++ b/src/gallium/auxiliary/draw/draw_so_emit_tmp.h
@@ -1,123 +1,33 @@
-
-static void FUNC( struct pt_so_emit *so,
-                  const struct draw_prim_info *input_prims,
-                  const struct draw_vertex_info *input_verts,
-                  unsigned start,
-                  unsigned count)
-{
-   struct draw_context *draw = so->draw;
-
-   boolean flatfirst = (draw->rasterizer->flatshade &&
-                        draw->rasterizer->flatshade_first);
-   unsigned i;
-   LOCAL_VARS
-
-   if (0) debug_printf("%s %d\n", __FUNCTION__, count);
-
-   debug_assert(input_prims->primitive_count == 1);
-
-   switch (input_prims->prim) {
-   case PIPE_PRIM_POINTS:
-      for (i = 0; i < count; i++) {
-	 POINT( so, start + i + 0 );
-      }
-      break;
-
-   case PIPE_PRIM_LINES:
-      for (i = 0; i+1 < count; i += 2) {
-         LINE( so , start + i + 0 , start + i + 1 );
-      }
-      break;
-
-   case PIPE_PRIM_LINE_LOOP:
-      if (count >= 2) {
-
-         for (i = 1; i < count; i++) {
-            LINE( so, start + i - 1, start + i );
-         }
-
-	 LINE( so, start + i - 1, start );
-      }
-      break;
-
-   case PIPE_PRIM_LINE_STRIP:
-      for (i = 1; i < count; i++) {
-         LINE( so, start + i - 1, start + i );
-      }
-      break;
-
-   case PIPE_PRIM_TRIANGLES:
-      for (i = 0; i+2 < count; i += 3) {
-         TRIANGLE( so, start + i + 0, start + i + 1, start + i + 2 );
-      }
-      break;
-
-   case PIPE_PRIM_TRIANGLE_STRIP:
-      if (flatfirst) {
-         for (i = 0; i+2 < count; i++) {
-            TRIANGLE( so,
-                      start + i + 0,
-                      start + i + 1 + (i&1),
-                      start + i + 2 - (i&1) );
-         }
-      }
-      else {
-         for (i = 0; i+2 < count; i++) {
-            TRIANGLE( so,
-                      start + i + 0 + (i&1),
-                      start + i + 1 - (i&1),
-                      start + i + 2 );
-         }
-      }
-      break;
-
-   case PIPE_PRIM_TRIANGLE_FAN:
-      if (count >= 3) {
-         if (flatfirst) {
-            for (i = 0; i+2 < count; i++) {
-               TRIANGLE( so,
-                         start + i + 1,
-                         start + i + 2,
-                         start );
-            }
-         }
-         else {
-            for (i = 0; i+2 < count; i++) {
-               TRIANGLE( so,
-                         start,
-                         start + i + 1,
-                         start + i + 2 );
-            }
-         }
-      }
-      break;
-
-   case PIPE_PRIM_POLYGON:
-      {
-         /* These bitflags look a little odd because we submit the
-          * vertices as (1,2,0) to satisfy flatshade requirements.
-          */
-
-	 for (i = 0; i+2 < count; i++) {
-
-            if (flatfirst) {
-               TRIANGLE( so, start + 0, start + i + 1, start + i + 2 );
-            }
-            else {
-               TRIANGLE( so, start + i + 1, start + i + 2, start + 0 );
-            }
-	 }
-      }
-      break;
-
-   default:
-      debug_assert(!"Unsupported primitive in stream output");
-      break;
-   }
-}
-
-
-#undef TRIANGLE
-#undef POINT
-#undef LINE
-#undef FUNC
+#define FUNC_VARS                               \
+   struct pt_so_emit *so,                       \
+   const struct draw_prim_info *input_prims,    \
+   const struct draw_vertex_info *input_verts,  \
+   unsigned start,                              \
+   unsigned count
+
+#define FUNC_ENTER                                                \
+   /* declare more local vars */                                  \
+   struct draw_context *draw = so->draw;                          \
+   const unsigned prim = input_prims->prim;                       \
+   const boolean last_vertex_last =                               \
+      !(draw->rasterizer->flatshade &&                            \
+        draw->rasterizer->flatshade_first);                       \
+   do {                                                           \
+      debug_assert(input_prims->primitive_count == 1);            \
+      switch (prim) {                                             \
+      case PIPE_PRIM_LINES_ADJACENCY:                             \
+      case PIPE_PRIM_LINE_STRIP_ADJACENCY:                        \
+      case PIPE_PRIM_TRIANGLES_ADJACENCY:                         \
+      case PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY:                    \
+         debug_assert(!"unexpected primitive type in stream output"); \
+         return;                                                  \
+      default:                                                    \
+         break;                                                   \
+      }                                                           \
+   } while (0)                                                    \
+
+#define POINT(i0)                so_point(so,i0)
+#define LINE(flags,i0,i1)        so_line(so,i0,i1)
+#define TRIANGLE(flags,i0,i1,i2) so_tri(so,i0,i1,i2)
+
+#include "draw_decompose_tmp.h"
diff --git a/src/gallium/auxiliary/draw/draw_vertex.h b/src/gallium/auxiliary/draw/draw_vertex.h
index 3af31ffe126..e63cf5f4f98 100644
--- a/src/gallium/auxiliary/draw/draw_vertex.h
+++ b/src/gallium/auxiliary/draw/draw_vertex.h
@@ -166,7 +166,7 @@ static INLINE enum pipe_format draw_translate_vinfo_format(enum attrib_emit emit
    }
 }
 
-static INLINE enum attrib_emit draw_translate_vinfo_size(enum attrib_emit emit)
+static INLINE unsigned draw_translate_vinfo_size(enum attrib_emit emit)
 {
    switch (emit) {
    case EMIT_OMIT:
diff --git a/src/gallium/auxiliary/draw/draw_vs.c b/src/gallium/auxiliary/draw/draw_vs.c
index 57ea63fc060..fb665b08fff 100644
--- a/src/gallium/auxiliary/draw/draw_vs.c
+++ b/src/gallium/auxiliary/draw/draw_vs.c
@@ -48,18 +48,30 @@
 
 DEBUG_GET_ONCE_BOOL_OPTION(gallium_dump_vs, "GALLIUM_DUMP_VS", FALSE)
 
+
+/**
+ * Set a vertex shader constant buffer.
+ * \param slot  which constant buffer in [0, PIPE_MAX_CONSTANT_BUFFERS-1]
+ * \param constants  the mapped buffer
+ * \param size  size of buffer in bytes
+ */
 void
 draw_vs_set_constants(struct draw_context *draw,
                       unsigned slot,
                       const void *constants,
                       unsigned size)
 {
-   if (((uintptr_t)constants) & 0xf) {
+   const int alignment = 16;
+
+   /* check if buffer is 16-byte aligned */
+   if (((uintptr_t)constants) & (alignment - 1)) {
+      /* if not, copy the constants into a new, 16-byte aligned buffer */
       if (size > draw->vs.const_storage_size[slot]) {
          if (draw->vs.aligned_constant_storage[slot]) {
             align_free((void *)draw->vs.aligned_constant_storage[slot]);
          }
-         draw->vs.aligned_constant_storage[slot] = align_malloc(size, 16);
+         draw->vs.aligned_constant_storage[slot] =
+            align_malloc(size, alignment);
       }
       assert(constants);
       memcpy((void *)draw->vs.aligned_constant_storage[slot],
diff --git a/src/gallium/auxiliary/draw/draw_vs.h b/src/gallium/auxiliary/draw/draw_vs.h
index a7319945234..f9a038788fb 100644
--- a/src/gallium/auxiliary/draw/draw_vs.h
+++ b/src/gallium/auxiliary/draw/draw_vs.h
@@ -133,7 +133,8 @@ struct draw_vertex_shader {
    void (*run_linear)( struct draw_vertex_shader *shader,
 		       const float (*input)[4],
 		       float (*output)[4],
-                      const void *constants[PIPE_MAX_CONSTANT_BUFFERS],
+                       const void *constants[PIPE_MAX_CONSTANT_BUFFERS],
+                       const unsigned const_size[PIPE_MAX_CONSTANT_BUFFERS],
 		       unsigned count,
 		       unsigned input_stride,
 		       unsigned output_stride );
diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c
index bc34d390dae..dab3eb1ca8e 100644
--- a/src/gallium/auxiliary/draw/draw_vs_exec.c
+++ b/src/gallium/auxiliary/draw/draw_vs_exec.c
@@ -85,7 +85,8 @@ static void
 vs_exec_run_linear( struct draw_vertex_shader *shader,
 		    const float (*input)[4],
 		    float (*output)[4],
-                   const void *constants[PIPE_MAX_CONSTANT_BUFFERS],
+                    const void *constants[PIPE_MAX_CONSTANT_BUFFERS],
+                    const unsigned const_size[PIPE_MAX_CONSTANT_BUFFERS],
 		    unsigned count,
 		    unsigned input_stride,
 		    unsigned output_stride )
@@ -95,9 +96,8 @@ vs_exec_run_linear( struct draw_vertex_shader *shader,
    unsigned int i, j;
    unsigned slot;
 
-   for (i = 0; i < PIPE_MAX_CONSTANT_BUFFERS; i++) {
-      machine->Consts[i] = constants[i];
-   }
+   tgsi_exec_set_constant_buffers(machine, PIPE_MAX_CONSTANT_BUFFERS,
+                                  constants, const_size);
 
    for (i = 0; i < count; i += MAX_TGSI_VERTICES) {
       unsigned int max_vertices = MIN2(MAX_TGSI_VERTICES, count - i);
diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c
index 6c13df79132..d13ad24fff0 100644
--- a/src/gallium/auxiliary/draw/draw_vs_llvm.c
+++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c
@@ -49,6 +49,7 @@ vs_llvm_run_linear( struct draw_vertex_shader *shader,
 		    const float (*input)[4],
 		    float (*output)[4],
                     const void *constants[PIPE_MAX_CONSTANT_BUFFERS],
+                    const unsigned constants_size[PIPE_MAX_CONSTANT_BUFFERS],
 		    unsigned count,
 		    unsigned input_stride,
 		    unsigned output_stride )
diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c
index 14c95082a9d..0b0c6077c6f 100644
--- a/src/gallium/auxiliary/draw/draw_vs_sse.c
+++ b/src/gallium/auxiliary/draw/draw_vs_sse.c
@@ -84,6 +84,7 @@ vs_sse_run_linear( struct draw_vertex_shader *base,
 		   const float (*input)[4],
 		   float (*output)[4],
                   const void *constants[PIPE_MAX_CONSTANT_BUFFERS],
+		  const unsigned const_size[PIPE_MAX_CONSTANT_BUFFERS],
 		   unsigned count,
 		   unsigned input_stride,
 		   unsigned output_stride )
diff --git a/src/gallium/auxiliary/draw/draw_vs_varient.c b/src/gallium/auxiliary/draw/draw_vs_varient.c
index 6eb26927f27..eacd1601877 100644
--- a/src/gallium/auxiliary/draw/draw_vs_varient.c
+++ b/src/gallium/auxiliary/draw/draw_vs_varient.c
@@ -149,7 +149,8 @@ static void PIPE_CDECL vsvg_run_elts( struct draw_vs_varient *varient,
    vsvg->base.vs->run_linear( vsvg->base.vs, 
                               temp_buffer,
                               temp_buffer,
-                             vsvg->base.vs->draw->pt.user.vs_constants,
+                              vsvg->base.vs->draw->pt.user.vs_constants,
+                              vsvg->base.vs->draw->pt.user.vs_constants_size,
                               count,
                               temp_vertex_stride, 
                               temp_vertex_stride);
@@ -214,7 +215,8 @@ static void PIPE_CDECL vsvg_run_linear( struct draw_vs_varient *varient,
    vsvg->base.vs->run_linear( vsvg->base.vs, 
                               temp_buffer,
                               temp_buffer,
-                             vsvg->base.vs->draw->pt.user.vs_constants,
+                              vsvg->base.vs->draw->pt.user.vs_constants,
+                              vsvg->base.vs->draw->pt.user.vs_constants_size,
                               count,
                               temp_vertex_stride, 
                               temp_vertex_stride);
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_arit.c b/src/gallium/auxiliary/gallivm/lp_bld_arit.c
index f5f2623e467..7b35dd4bb49 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_arit.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_arit.c
@@ -1,6 +1,6 @@
 /**************************************************************************
  *
- * Copyright 2009 VMware, Inc.
+ * Copyright 2009-2010 VMware, Inc.
  * All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -59,6 +59,19 @@
 #include "lp_bld_arit.h"
 
 
+/*
+ * XXX: Increasing eliminates some artifacts, but adds others, most
+ * noticeably corruption in the Earth halo in Google Earth.
+ */
+#define RCP_NEWTON_STEPS 0
+
+#define RSQRT_NEWTON_STEPS 0
+
+#define EXP_POLY_DEGREE 3
+
+#define LOG_POLY_DEGREE 5
+
+
 /**
  * Generate min(a, b)
  * No checks for special case values of a or b = 1 or 0 are done.
@@ -72,6 +85,9 @@ lp_build_min_simple(struct lp_build_context *bld,
    const char *intrinsic = NULL;
    LLVMValueRef cond;
 
+   assert(lp_check_value(type, a));
+   assert(lp_check_value(type, b));
+
    /* TODO: optimize the constant case */
 
    if(type.width * type.length == 128) {
@@ -118,6 +134,9 @@ lp_build_max_simple(struct lp_build_context *bld,
    const char *intrinsic = NULL;
    LLVMValueRef cond;
 
+   assert(lp_check_value(type, a));
+   assert(lp_check_value(type, b));
+
    /* TODO: optimize the constant case */
 
    if(type.width * type.length == 128) {
@@ -160,6 +179,8 @@ lp_build_comp(struct lp_build_context *bld,
 {
    const struct lp_type type = bld->type;
 
+   assert(lp_check_value(type, a));
+
    if(a == bld->one)
       return bld->zero;
    if(a == bld->zero)
@@ -173,9 +194,15 @@ lp_build_comp(struct lp_build_context *bld,
    }
 
    if(LLVMIsConstant(a))
-      return LLVMConstSub(bld->one, a);
+      if (type.floating)
+          return LLVMConstFSub(bld->one, a);
+      else
+          return LLVMConstSub(bld->one, a);
    else
-      return LLVMBuildSub(bld->builder, bld->one, a, "");
+      if (type.floating)
+         return LLVMBuildFSub(bld->builder, bld->one, a, "");
+      else
+         return LLVMBuildSub(bld->builder, bld->one, a, "");
 }
 
 
@@ -190,6 +217,9 @@ lp_build_add(struct lp_build_context *bld,
    const struct lp_type type = bld->type;
    LLVMValueRef res;
 
+   assert(lp_check_value(type, a));
+   assert(lp_check_value(type, b));
+
    if(a == bld->zero)
       return b;
    if(b == bld->zero)
@@ -217,9 +247,15 @@ lp_build_add(struct lp_build_context *bld,
    }
 
    if(LLVMIsConstant(a) && LLVMIsConstant(b))
-      res = LLVMConstAdd(a, b);
+      if (type.floating)
+         res = LLVMConstFAdd(a, b);
+      else
+         res = LLVMConstAdd(a, b);
    else
-      res = LLVMBuildAdd(bld->builder, a, b, "");
+      if (type.floating)
+         res = LLVMBuildFAdd(bld->builder, a, b, "");
+      else
+         res = LLVMBuildAdd(bld->builder, a, b, "");
 
    /* clamp to ceiling of 1.0 */
    if(bld->type.norm && (bld->type.floating || bld->type.fixed))
@@ -240,6 +276,8 @@ lp_build_sum_vector(struct lp_build_context *bld,
    LLVMValueRef index, res;
    unsigned i;
 
+   assert(lp_check_value(type, a));
+
    if (a == bld->zero)
       return bld->zero;
    if (a == bld->undef)
@@ -253,9 +291,16 @@ lp_build_sum_vector(struct lp_build_context *bld,
 
    for (i = 1; i < type.length; i++) {
       index = LLVMConstInt(LLVMInt32Type(), i, 0);
-      res = LLVMBuildAdd(bld->builder, res,
-                         LLVMBuildExtractElement(bld->builder, a, index, ""),
-                         "");
+      if (type.floating)
+         res = LLVMBuildFAdd(bld->builder, res,
+                            LLVMBuildExtractElement(bld->builder,
+                                                    a, index, ""),
+                            "");
+      else
+         res = LLVMBuildAdd(bld->builder, res,
+                            LLVMBuildExtractElement(bld->builder,
+                                                    a, index, ""),
+                            "");
    }
 
    return res;
@@ -273,6 +318,9 @@ lp_build_sub(struct lp_build_context *bld,
    const struct lp_type type = bld->type;
    LLVMValueRef res;
 
+   assert(lp_check_value(type, a));
+   assert(lp_check_value(type, b));
+
    if(b == bld->zero)
       return a;
    if(a == bld->undef || b == bld->undef)
@@ -300,9 +348,15 @@ lp_build_sub(struct lp_build_context *bld,
    }
 
    if(LLVMIsConstant(a) && LLVMIsConstant(b))
-      res = LLVMConstSub(a, b);
+      if (type.floating)
+         res = LLVMConstFSub(a, b);
+      else
+         res = LLVMConstSub(a, b);
    else
-      res = LLVMBuildSub(bld->builder, a, b, "");
+      if (type.floating)
+         res = LLVMBuildFSub(bld->builder, a, b, "");
+      else
+         res = LLVMBuildSub(bld->builder, a, b, "");
 
    if(bld->type.norm && (bld->type.floating || bld->type.fixed))
       res = lp_build_max_simple(bld, res, bld->zero);
@@ -360,6 +414,10 @@ lp_build_mul_u8n(LLVMBuilderRef builder,
    LLVMValueRef c8;
    LLVMValueRef ab;
 
+   assert(!i16_type.floating);
+   assert(lp_check_value(i16_type, a));
+   assert(lp_check_value(i16_type, b));
+
    c8 = lp_build_const_int_vec(i16_type, 8);
    
 #if 0
@@ -395,6 +453,9 @@ lp_build_mul(struct lp_build_context *bld,
    LLVMValueRef shift;
    LLVMValueRef res;
 
+   assert(lp_check_value(type, a));
+   assert(lp_check_value(type, b));
+
    if(a == bld->zero)
       return bld->zero;
    if(a == bld->one)
@@ -433,7 +494,10 @@ lp_build_mul(struct lp_build_context *bld,
       shift = NULL;
 
    if(LLVMIsConstant(a) && LLVMIsConstant(b)) {
-      res =  LLVMConstMul(a, b);
+      if (type.floating)
+         res = LLVMConstFMul(a, b);
+      else
+         res = LLVMConstMul(a, b);
       if(shift) {
          if(type.sign)
             res = LLVMConstAShr(res, shift);
@@ -442,7 +506,10 @@ lp_build_mul(struct lp_build_context *bld,
       }
    }
    else {
-      res = LLVMBuildMul(bld->builder, a, b, "");
+      if (type.floating)
+         res = LLVMBuildFMul(bld->builder, a, b, "");
+      else
+         res = LLVMBuildMul(bld->builder, a, b, "");
       if(shift) {
          if(type.sign)
             res = LLVMBuildAShr(bld->builder, res, shift, "");
@@ -465,6 +532,8 @@ lp_build_mul_imm(struct lp_build_context *bld,
 {
    LLVMValueRef factor;
 
+   assert(lp_check_value(bld->type, a));
+
    if(b == 0)
       return bld->zero;
 
@@ -472,7 +541,7 @@ lp_build_mul_imm(struct lp_build_context *bld,
       return a;
 
    if(b == -1)
-      return LLVMBuildNeg(bld->builder, a, "");
+      return lp_build_negate(bld, a);
 
    if(b == 2 && bld->type.floating)
       return lp_build_add(bld, a, a);
@@ -518,6 +587,9 @@ lp_build_div(struct lp_build_context *bld,
 {
    const struct lp_type type = bld->type;
 
+   assert(lp_check_value(type, a));
+   assert(lp_check_value(type, b));
+
    if(a == bld->zero)
       return bld->zero;
    if(a == bld->one)
@@ -529,13 +601,24 @@ lp_build_div(struct lp_build_context *bld,
    if(a == bld->undef || b == bld->undef)
       return bld->undef;
 
-   if(LLVMIsConstant(a) && LLVMIsConstant(b))
-      return LLVMConstFDiv(a, b);
+   if(LLVMIsConstant(a) && LLVMIsConstant(b)) {
+      if (type.floating)
+         return LLVMConstFDiv(a, b);
+      else if (type.sign)
+         return LLVMConstSDiv(a, b);
+      else
+         return LLVMConstUDiv(a, b);
+   }
 
    if(util_cpu_caps.has_sse && type.width == 32 && type.length == 4)
       return lp_build_mul(bld, a, lp_build_rcp(bld, b));
 
-   return LLVMBuildFDiv(bld->builder, a, b, "");
+   if (type.floating)
+      return LLVMBuildFDiv(bld->builder, a, b, "");
+   else if (type.sign)
+      return LLVMBuildSDiv(bld->builder, a, b, "");
+   else
+      return LLVMBuildUDiv(bld->builder, a, b, "");
 }
 
 
@@ -555,6 +638,10 @@ lp_build_lerp(struct lp_build_context *bld,
    LLVMValueRef delta;
    LLVMValueRef res;
 
+   assert(lp_check_value(bld->type, x));
+   assert(lp_check_value(bld->type, v0));
+   assert(lp_check_value(bld->type, v1));
+
    delta = lp_build_sub(bld, v1, v0);
 
    res = lp_build_mul(bld, x, delta);
@@ -596,6 +683,9 @@ lp_build_min(struct lp_build_context *bld,
              LLVMValueRef a,
              LLVMValueRef b)
 {
+   assert(lp_check_value(bld->type, a));
+   assert(lp_check_value(bld->type, b));
+
    if(a == bld->undef || b == bld->undef)
       return bld->undef;
 
@@ -624,6 +714,9 @@ lp_build_max(struct lp_build_context *bld,
              LLVMValueRef a,
              LLVMValueRef b)
 {
+   assert(lp_check_value(bld->type, a));
+   assert(lp_check_value(bld->type, b));
+
    if(a == bld->undef || b == bld->undef)
       return bld->undef;
 
@@ -653,6 +746,10 @@ lp_build_clamp(struct lp_build_context *bld,
                LLVMValueRef min,
                LLVMValueRef max)
 {
+   assert(lp_check_value(bld->type, a));
+   assert(lp_check_value(bld->type, min));
+   assert(lp_check_value(bld->type, max));
+
    a = lp_build_min(bld, a, max);
    a = lp_build_max(bld, a, min);
    return a;
@@ -669,6 +766,8 @@ lp_build_abs(struct lp_build_context *bld,
    const struct lp_type type = bld->type;
    LLVMTypeRef vec_type = lp_build_vec_type(type);
 
+   assert(lp_check_value(type, a));
+
    if(!type.sign)
       return a;
 
@@ -702,7 +801,16 @@ LLVMValueRef
 lp_build_negate(struct lp_build_context *bld,
                 LLVMValueRef a)
 {
-   return LLVMBuildNeg(bld->builder, a, "");
+   assert(lp_check_value(bld->type, a));
+
+#if HAVE_LLVM >= 0x0207
+   if (bld->type.floating)
+      a = LLVMBuildFNeg(bld->builder, a, "");
+   else
+#endif
+      a = LLVMBuildNeg(bld->builder, a, "");
+
+   return a;
 }
 
 
@@ -715,6 +823,8 @@ lp_build_sgn(struct lp_build_context *bld,
    LLVMValueRef cond;
    LLVMValueRef res;
 
+   assert(lp_check_value(type, a));
+
    /* Handle non-zero case */
    if(!type.sign) {
       /* if not zero then sign must be positive */
@@ -773,6 +883,7 @@ lp_build_set_sign(struct lp_build_context *bld,
    LLVMValueRef val, res;
 
    assert(type.floating);
+   assert(lp_check_value(type, a));
 
    /* val = reinterpret_cast<int>(a) */
    val = LLVMBuildBitCast(bld->builder, a, int_vec_type, "");
@@ -1021,7 +1132,7 @@ lp_build_iround(struct lp_build_context *bld,
       half = LLVMBuildOr(bld->builder, sign, half, "");
       half = LLVMBuildBitCast(bld->builder, half, vec_type, "");
 
-      res = LLVMBuildAdd(bld->builder, a, half, "");
+      res = LLVMBuildFAdd(bld->builder, a, half, "");
    }
 
    res = LLVMBuildFPToSI(bld->builder, res, int_vec_type, "");
@@ -1070,7 +1181,7 @@ lp_build_ifloor(struct lp_build_context *bld,
       offset = LLVMBuildAnd(bld->builder, offset, sign, "");
       offset = LLVMBuildBitCast(bld->builder, offset, vec_type, "ifloor.offset");
 
-      res = LLVMBuildAdd(bld->builder, a, offset, "ifloor.res");
+      res = LLVMBuildFAdd(bld->builder, a, offset, "ifloor.res");
    }
 
    /* round to nearest (toward zero) */
@@ -1120,7 +1231,7 @@ lp_build_iceil(struct lp_build_context *bld,
       offset = LLVMBuildAnd(bld->builder, offset, sign, "");
       offset = LLVMBuildBitCast(bld->builder, offset, vec_type, "iceil.offset");
 
-      res = LLVMBuildAdd(bld->builder, a, offset, "iceil.res");
+      res = LLVMBuildFAdd(bld->builder, a, offset, "iceil.res");
    }
 
    /* round to nearest (toward zero) */
@@ -1138,6 +1249,8 @@ lp_build_sqrt(struct lp_build_context *bld,
    LLVMTypeRef vec_type = lp_build_vec_type(type);
    char intrinsic[32];
 
+   assert(lp_check_value(type, a));
+
    /* TODO: optimize the constant case */
    /* TODO: optimize the constant case */
 
@@ -1148,12 +1261,39 @@ lp_build_sqrt(struct lp_build_context *bld,
 }
 
 
+/**
+ * Do one Newton-Raphson step to improve reciprocate precision:
+ *
+ *   x_{i+1} = x_i * (2 - a * x_i)
+ *
+ * See also:
+ * - http://en.wikipedia.org/wiki/Division_(digital)#Newton.E2.80.93Raphson_division
+ * - http://softwarecommunity.intel.com/articles/eng/1818.htm
+ */
+static INLINE LLVMValueRef
+lp_build_rcp_refine(struct lp_build_context *bld,
+                    LLVMValueRef a,
+                    LLVMValueRef rcp_a)
+{
+   LLVMValueRef two = lp_build_const_vec(bld->type, 2.0);
+   LLVMValueRef res;
+
+   res = LLVMBuildFMul(bld->builder, a, rcp_a, "");
+   res = LLVMBuildFSub(bld->builder, two, res, "");
+   res = LLVMBuildFMul(bld->builder, rcp_a, res, "");
+
+   return res;
+}
+
+
 LLVMValueRef
 lp_build_rcp(struct lp_build_context *bld,
              LLVMValueRef a)
 {
    const struct lp_type type = bld->type;
 
+   assert(lp_check_value(type, a));
+
    if(a == bld->zero)
       return bld->undef;
    if(a == bld->one)
@@ -1167,32 +1307,16 @@ lp_build_rcp(struct lp_build_context *bld,
       return LLVMConstFDiv(bld->one, a);
 
    if(util_cpu_caps.has_sse && type.width == 32 && type.length == 4) {
-      /*
-       * XXX: Added precision is not always necessary, so only enable this
-       * when we have a better system in place to track minimum precision.
-       */
-
-#if 0
-      /*
-       * Do one Newton-Raphson step to improve precision:
-       *
-       *   x1 = (2 - a * rcp(a)) * rcp(a)
-       */
-
-      LLVMValueRef two = lp_build_const_vec(bld->type, 2.0);
-      LLVMValueRef rcp_a;
       LLVMValueRef res;
+      unsigned i;
 
-      rcp_a = lp_build_intrinsic_unary(bld->builder, "llvm.x86.sse.rcp.ps", lp_build_vec_type(type), a);
+      res = lp_build_intrinsic_unary(bld->builder, "llvm.x86.sse.rcp.ps", bld->vec_type, a);
 
-      res = LLVMBuildMul(bld->builder, a, rcp_a, "");
-      res = LLVMBuildSub(bld->builder, two, res, "");
-      res = LLVMBuildMul(bld->builder, res, rcp_a, "");
+      for (i = 0; i < RCP_NEWTON_STEPS; ++i) {
+         res = lp_build_rcp_refine(bld, a, res);
+      }
 
-      return rcp_a;
-#else
-      return lp_build_intrinsic_unary(bld->builder, "llvm.x86.sse.rcp.ps", lp_build_vec_type(type), a);
-#endif
+      return res;
    }
 
    return LLVMBuildFDiv(bld->builder, bld->one, a, "");
@@ -1200,6 +1324,33 @@ lp_build_rcp(struct lp_build_context *bld,
 
 
 /**
+ * Do one Newton-Raphson step to improve rsqrt precision:
+ *
+ *   x_{i+1} = 0.5 * x_i * (3.0 - a * x_i * x_i)
+ *
+ * See also:
+ * - http://softwarecommunity.intel.com/articles/eng/1818.htm
+ */
+static INLINE LLVMValueRef
+lp_build_rsqrt_refine(struct lp_build_context *bld,
+                      LLVMValueRef a,
+                      LLVMValueRef rsqrt_a)
+{
+   LLVMValueRef half = lp_build_const_vec(bld->type, 0.5);
+   LLVMValueRef three = lp_build_const_vec(bld->type, 3.0);
+   LLVMValueRef res;
+
+   res = LLVMBuildFMul(bld->builder, rsqrt_a, rsqrt_a, "");
+   res = LLVMBuildFMul(bld->builder, a, res, "");
+   res = LLVMBuildFSub(bld->builder, three, res, "");
+   res = LLVMBuildFMul(bld->builder, rsqrt_a, res, "");
+   res = LLVMBuildFMul(bld->builder, half, res, "");
+
+   return res;
+}
+
+
+/**
  * Generate 1/sqrt(a)
  */
 LLVMValueRef
@@ -1208,10 +1359,22 @@ lp_build_rsqrt(struct lp_build_context *bld,
 {
    const struct lp_type type = bld->type;
 
+   assert(lp_check_value(type, a));
+
    assert(type.floating);
 
-   if(util_cpu_caps.has_sse && type.width == 32 && type.length == 4)
-      return lp_build_intrinsic_unary(bld->builder, "llvm.x86.sse.rsqrt.ps", lp_build_vec_type(type), a);
+   if(util_cpu_caps.has_sse && type.width == 32 && type.length == 4) {
+      LLVMValueRef res;
+      unsigned i;
+
+      res = lp_build_intrinsic_unary(bld->builder, "llvm.x86.sse.rsqrt.ps", bld->vec_type, a);
+
+      for (i = 0; i < RSQRT_NEWTON_STEPS; ++i) {
+         res = lp_build_rsqrt_refine(bld, a, res);
+      }
+
+      return res;
+   }
 
    return lp_build_rcp(bld, lp_build_sqrt(bld, a));
 }
@@ -1270,7 +1433,7 @@ lp_build_sin(struct lp_build_context *bld,
     */
    
    LLVMValueRef FOPi = lp_build_const_v4sf(1.27323954473516);
-   LLVMValueRef scale_y = LLVMBuildMul(b, x_abs, FOPi, "scale_y");
+   LLVMValueRef scale_y = LLVMBuildFMul(b, x_abs, FOPi, "scale_y");
 
    /*
     * store the integer part of y in mm0
@@ -1344,9 +1507,9 @@ lp_build_sin(struct lp_build_context *bld,
     * xmm2 = _mm_mul_ps(y, xmm2);
     * xmm3 = _mm_mul_ps(y, xmm3);
     */
-   LLVMValueRef xmm1 = LLVMBuildMul(b, y_2, DP1, "xmm1");
-   LLVMValueRef xmm2 = LLVMBuildMul(b, y_2, DP2, "xmm2");
-   LLVMValueRef xmm3 = LLVMBuildMul(b, y_2, DP3, "xmm3");
+   LLVMValueRef xmm1 = LLVMBuildFMul(b, y_2, DP1, "xmm1");
+   LLVMValueRef xmm2 = LLVMBuildFMul(b, y_2, DP2, "xmm2");
+   LLVMValueRef xmm3 = LLVMBuildFMul(b, y_2, DP3, "xmm3");
 
    /*
     * x = _mm_add_ps(x, xmm1);
@@ -1354,16 +1517,16 @@ lp_build_sin(struct lp_build_context *bld,
     * x = _mm_add_ps(x, xmm3);
     */ 
 
-   LLVMValueRef x_1 = LLVMBuildAdd(b, x_abs, xmm1, "x_1");
-   LLVMValueRef x_2 = LLVMBuildAdd(b, x_1, xmm2, "x_2");
-   LLVMValueRef x_3 = LLVMBuildAdd(b, x_2, xmm3, "x_3");
+   LLVMValueRef x_1 = LLVMBuildFAdd(b, x_abs, xmm1, "x_1");
+   LLVMValueRef x_2 = LLVMBuildFAdd(b, x_1, xmm2, "x_2");
+   LLVMValueRef x_3 = LLVMBuildFAdd(b, x_2, xmm3, "x_3");
 
    /*
     * Evaluate the first polynom  (0 <= x <= Pi/4)
     *
     * z = _mm_mul_ps(x,x);
     */
-   LLVMValueRef z = LLVMBuildMul(b, x_3, x_3, "z");
+   LLVMValueRef z = LLVMBuildFMul(b, x_3, x_3, "z");
 
    /*
     * _PS_CONST(coscof_p0,  2.443315711809948E-005);
@@ -1378,12 +1541,12 @@ lp_build_sin(struct lp_build_context *bld,
     * y = *(v4sf*)_ps_coscof_p0;
     * y = _mm_mul_ps(y, z);
     */
-   LLVMValueRef y_3 = LLVMBuildMul(b, z, coscof_p0, "y_3");
-   LLVMValueRef y_4 = LLVMBuildAdd(b, y_3, coscof_p1, "y_4");
-   LLVMValueRef y_5 = LLVMBuildMul(b, y_4, z, "y_5");
-   LLVMValueRef y_6 = LLVMBuildAdd(b, y_5, coscof_p2, "y_6");
-   LLVMValueRef y_7 = LLVMBuildMul(b, y_6, z, "y_7");
-   LLVMValueRef y_8 = LLVMBuildMul(b, y_7, z, "y_8");
+   LLVMValueRef y_3 = LLVMBuildFMul(b, z, coscof_p0, "y_3");
+   LLVMValueRef y_4 = LLVMBuildFAdd(b, y_3, coscof_p1, "y_4");
+   LLVMValueRef y_5 = LLVMBuildFMul(b, y_4, z, "y_5");
+   LLVMValueRef y_6 = LLVMBuildFAdd(b, y_5, coscof_p2, "y_6");
+   LLVMValueRef y_7 = LLVMBuildFMul(b, y_6, z, "y_7");
+   LLVMValueRef y_8 = LLVMBuildFMul(b, y_7, z, "y_8");
 
 
    /*
@@ -1392,10 +1555,10 @@ lp_build_sin(struct lp_build_context *bld,
     * y = _mm_add_ps(y, *(v4sf*)_ps_1);
     */ 
    LLVMValueRef half = lp_build_const_v4sf(0.5);
-   LLVMValueRef tmp = LLVMBuildMul(b, z, half, "tmp");
-   LLVMValueRef y_9 = LLVMBuildSub(b, y_8, tmp, "y_8");
+   LLVMValueRef tmp = LLVMBuildFMul(b, z, half, "tmp");
+   LLVMValueRef y_9 = LLVMBuildFSub(b, y_8, tmp, "y_8");
    LLVMValueRef one = lp_build_const_v4sf(1.0);
-   LLVMValueRef y_10 = LLVMBuildAdd(b, y_9, one, "y_9");
+   LLVMValueRef y_10 = LLVMBuildFAdd(b, y_9, one, "y_9");
 
    /*
     * _PS_CONST(sincof_p0, -1.9515295891E-4);
@@ -1419,13 +1582,13 @@ lp_build_sin(struct lp_build_context *bld,
     * y2 = _mm_add_ps(y2, x);
     */
 
-   LLVMValueRef y2_3 = LLVMBuildMul(b, z, sincof_p0, "y2_3");
-   LLVMValueRef y2_4 = LLVMBuildAdd(b, y2_3, sincof_p1, "y2_4");
-   LLVMValueRef y2_5 = LLVMBuildMul(b, y2_4, z, "y2_5");
-   LLVMValueRef y2_6 = LLVMBuildAdd(b, y2_5, sincof_p2, "y2_6");
-   LLVMValueRef y2_7 = LLVMBuildMul(b, y2_6, z, "y2_7");
-   LLVMValueRef y2_8 = LLVMBuildMul(b, y2_7, x_3, "y2_8");
-   LLVMValueRef y2_9 = LLVMBuildAdd(b, y2_8, x_3, "y2_9");
+   LLVMValueRef y2_3 = LLVMBuildFMul(b, z, sincof_p0, "y2_3");
+   LLVMValueRef y2_4 = LLVMBuildFAdd(b, y2_3, sincof_p1, "y2_4");
+   LLVMValueRef y2_5 = LLVMBuildFMul(b, y2_4, z, "y2_5");
+   LLVMValueRef y2_6 = LLVMBuildFAdd(b, y2_5, sincof_p2, "y2_6");
+   LLVMValueRef y2_7 = LLVMBuildFMul(b, y2_6, z, "y2_7");
+   LLVMValueRef y2_8 = LLVMBuildFMul(b, y2_7, x_3, "y2_8");
+   LLVMValueRef y2_9 = LLVMBuildFAdd(b, y2_8, x_3, "y2_9");
 
    /*
     * select the correct result from the two polynoms
@@ -1481,7 +1644,7 @@ lp_build_cos(struct lp_build_context *bld,
     */
    
    LLVMValueRef FOPi = lp_build_const_v4sf(1.27323954473516);
-   LLVMValueRef scale_y = LLVMBuildMul(b, x_abs, FOPi, "scale_y");
+   LLVMValueRef scale_y = LLVMBuildFMul(b, x_abs, FOPi, "scale_y");
 
    /*
     * store the integer part of y in mm0
@@ -1561,9 +1724,9 @@ lp_build_cos(struct lp_build_context *bld,
     * xmm2 = _mm_mul_ps(y, xmm2);
     * xmm3 = _mm_mul_ps(y, xmm3);
     */
-   LLVMValueRef xmm1 = LLVMBuildMul(b, y_2, DP1, "xmm1");
-   LLVMValueRef xmm2 = LLVMBuildMul(b, y_2, DP2, "xmm2");
-   LLVMValueRef xmm3 = LLVMBuildMul(b, y_2, DP3, "xmm3");
+   LLVMValueRef xmm1 = LLVMBuildFMul(b, y_2, DP1, "xmm1");
+   LLVMValueRef xmm2 = LLVMBuildFMul(b, y_2, DP2, "xmm2");
+   LLVMValueRef xmm3 = LLVMBuildFMul(b, y_2, DP3, "xmm3");
 
    /*
     * x = _mm_add_ps(x, xmm1);
@@ -1571,16 +1734,16 @@ lp_build_cos(struct lp_build_context *bld,
     * x = _mm_add_ps(x, xmm3);
     */ 
 
-   LLVMValueRef x_1 = LLVMBuildAdd(b, x_abs, xmm1, "x_1");
-   LLVMValueRef x_2 = LLVMBuildAdd(b, x_1, xmm2, "x_2");
-   LLVMValueRef x_3 = LLVMBuildAdd(b, x_2, xmm3, "x_3");
+   LLVMValueRef x_1 = LLVMBuildFAdd(b, x_abs, xmm1, "x_1");
+   LLVMValueRef x_2 = LLVMBuildFAdd(b, x_1, xmm2, "x_2");
+   LLVMValueRef x_3 = LLVMBuildFAdd(b, x_2, xmm3, "x_3");
 
    /*
     * Evaluate the first polynom  (0 <= x <= Pi/4)
     *
     * z = _mm_mul_ps(x,x);
     */
-   LLVMValueRef z = LLVMBuildMul(b, x_3, x_3, "z");
+   LLVMValueRef z = LLVMBuildFMul(b, x_3, x_3, "z");
 
    /*
     * _PS_CONST(coscof_p0,  2.443315711809948E-005);
@@ -1595,12 +1758,12 @@ lp_build_cos(struct lp_build_context *bld,
     * y = *(v4sf*)_ps_coscof_p0;
     * y = _mm_mul_ps(y, z);
     */
-   LLVMValueRef y_3 = LLVMBuildMul(b, z, coscof_p0, "y_3");
-   LLVMValueRef y_4 = LLVMBuildAdd(b, y_3, coscof_p1, "y_4");
-   LLVMValueRef y_5 = LLVMBuildMul(b, y_4, z, "y_5");
-   LLVMValueRef y_6 = LLVMBuildAdd(b, y_5, coscof_p2, "y_6");
-   LLVMValueRef y_7 = LLVMBuildMul(b, y_6, z, "y_7");
-   LLVMValueRef y_8 = LLVMBuildMul(b, y_7, z, "y_8");
+   LLVMValueRef y_3 = LLVMBuildFMul(b, z, coscof_p0, "y_3");
+   LLVMValueRef y_4 = LLVMBuildFAdd(b, y_3, coscof_p1, "y_4");
+   LLVMValueRef y_5 = LLVMBuildFMul(b, y_4, z, "y_5");
+   LLVMValueRef y_6 = LLVMBuildFAdd(b, y_5, coscof_p2, "y_6");
+   LLVMValueRef y_7 = LLVMBuildFMul(b, y_6, z, "y_7");
+   LLVMValueRef y_8 = LLVMBuildFMul(b, y_7, z, "y_8");
 
 
    /*
@@ -1609,10 +1772,10 @@ lp_build_cos(struct lp_build_context *bld,
     * y = _mm_add_ps(y, *(v4sf*)_ps_1);
     */ 
    LLVMValueRef half = lp_build_const_v4sf(0.5);
-   LLVMValueRef tmp = LLVMBuildMul(b, z, half, "tmp");
-   LLVMValueRef y_9 = LLVMBuildSub(b, y_8, tmp, "y_8");
+   LLVMValueRef tmp = LLVMBuildFMul(b, z, half, "tmp");
+   LLVMValueRef y_9 = LLVMBuildFSub(b, y_8, tmp, "y_8");
    LLVMValueRef one = lp_build_const_v4sf(1.0);
-   LLVMValueRef y_10 = LLVMBuildAdd(b, y_9, one, "y_9");
+   LLVMValueRef y_10 = LLVMBuildFAdd(b, y_9, one, "y_9");
 
    /*
     * _PS_CONST(sincof_p0, -1.9515295891E-4);
@@ -1636,13 +1799,13 @@ lp_build_cos(struct lp_build_context *bld,
     * y2 = _mm_add_ps(y2, x);
     */
 
-   LLVMValueRef y2_3 = LLVMBuildMul(b, z, sincof_p0, "y2_3");
-   LLVMValueRef y2_4 = LLVMBuildAdd(b, y2_3, sincof_p1, "y2_4");
-   LLVMValueRef y2_5 = LLVMBuildMul(b, y2_4, z, "y2_5");
-   LLVMValueRef y2_6 = LLVMBuildAdd(b, y2_5, sincof_p2, "y2_6");
-   LLVMValueRef y2_7 = LLVMBuildMul(b, y2_6, z, "y2_7");
-   LLVMValueRef y2_8 = LLVMBuildMul(b, y2_7, x_3, "y2_8");
-   LLVMValueRef y2_9 = LLVMBuildAdd(b, y2_8, x_3, "y2_9");
+   LLVMValueRef y2_3 = LLVMBuildFMul(b, z, sincof_p0, "y2_3");
+   LLVMValueRef y2_4 = LLVMBuildFAdd(b, y2_3, sincof_p1, "y2_4");
+   LLVMValueRef y2_5 = LLVMBuildFMul(b, y2_4, z, "y2_5");
+   LLVMValueRef y2_6 = LLVMBuildFAdd(b, y2_5, sincof_p2, "y2_6");
+   LLVMValueRef y2_7 = LLVMBuildFMul(b, y2_6, z, "y2_7");
+   LLVMValueRef y2_8 = LLVMBuildFMul(b, y2_7, x_3, "y2_8");
+   LLVMValueRef y2_9 = LLVMBuildFAdd(b, y2_8, x_3, "y2_9");
 
    /*
     * select the correct result from the two polynoms
@@ -1695,6 +1858,8 @@ lp_build_exp(struct lp_build_context *bld,
    /* log2(e) = 1/log(2) */
    LLVMValueRef log2e = lp_build_const_vec(bld->type, 1.4426950408889634);
 
+   assert(lp_check_value(bld->type, x));
+
    return lp_build_mul(bld, log2e, lp_build_exp2(bld, x));
 }
 
@@ -1709,14 +1874,12 @@ lp_build_log(struct lp_build_context *bld,
    /* log(2) */
    LLVMValueRef log2 = lp_build_const_vec(bld->type, 0.69314718055994529);
 
+   assert(lp_check_value(bld->type, x));
+
    return lp_build_mul(bld, log2, lp_build_exp2(bld, x));
 }
 
 
-#define EXP_POLY_DEGREE 3
-#define LOG_POLY_DEGREE 5
-
-
 /**
  * Generate polynomial.
  * Ex:  coeffs[0] + x * coeffs[1] + x^2 * coeffs[2].
@@ -1731,6 +1894,8 @@ lp_build_polynomial(struct lp_build_context *bld,
    LLVMValueRef res = NULL;
    unsigned i;
 
+   assert(lp_check_value(bld->type, x));
+
    /* TODO: optimize the constant case */
    if(LLVMIsConstant(x))
       debug_printf("%s: inefficient/imprecise constant arithmetic\n",
@@ -1802,6 +1967,8 @@ lp_build_exp2_approx(struct lp_build_context *bld,
    LLVMValueRef expfpart = NULL;
    LLVMValueRef res = NULL;
 
+   assert(lp_check_value(bld->type, x));
+
    if(p_exp2_int_part || p_frac_part || p_exp2) {
       /* TODO: optimize the constant case */
       if(LLVMIsConstant(x))
@@ -1817,7 +1984,7 @@ lp_build_exp2_approx(struct lp_build_context *bld,
       ipart = lp_build_floor(bld, x);
 
       /* fpart = x - ipart */
-      fpart = LLVMBuildSub(bld->builder, x, ipart, "");
+      fpart = LLVMBuildFSub(bld->builder, x, ipart, "");
    }
 
    if(p_exp2_int_part || p_exp2) {
@@ -1832,7 +1999,7 @@ lp_build_exp2_approx(struct lp_build_context *bld,
       expfpart = lp_build_polynomial(bld, fpart, lp_build_exp2_polynomial,
                                      Elements(lp_build_exp2_polynomial));
 
-      res = LLVMBuildMul(bld->builder, expipart, expfpart, "");
+      res = LLVMBuildFMul(bld->builder, expipart, expfpart, "");
    }
 
    if(p_exp2_int_part)
@@ -1915,6 +2082,8 @@ lp_build_log2_approx(struct lp_build_context *bld,
    LLVMValueRef logmant = NULL;
    LLVMValueRef res = NULL;
 
+   assert(lp_check_value(bld->type, x));
+
    if(p_exp || p_floor_log2 || p_log2) {
       /* TODO: optimize the constant case */
       if(LLVMIsConstant(x))
@@ -1945,9 +2114,9 @@ lp_build_log2_approx(struct lp_build_context *bld,
                                     Elements(lp_build_log2_polynomial));
 
       /* This effectively increases the polynomial degree by one, but ensures that log2(1) == 0*/
-      logmant = LLVMBuildMul(bld->builder, logmant, LLVMBuildSub(bld->builder, mant, bld->one, ""), "");
+      logmant = LLVMBuildFMul(bld->builder, logmant, LLVMBuildFSub(bld->builder, mant, bld->one, ""), "");
 
-      res = LLVMBuildAdd(bld->builder, logmant, logexp, "");
+      res = LLVMBuildFAdd(bld->builder, logmant, logexp, "");
    }
 
    if(p_exp) {
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_assert.c b/src/gallium/auxiliary/gallivm/lp_bld_assert.c
new file mode 100644
index 00000000000..f2ebd868a8d
--- /dev/null
+++ b/src/gallium/auxiliary/gallivm/lp_bld_assert.c
@@ -0,0 +1,101 @@
+/**************************************************************************
+ *
+ * Copyright 2010 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "util/u_debug.h"
+#include "util/u_memory.h"
+#include "lp_bld_assert.h"
+#include "lp_bld_init.h"
+#include "lp_bld_printf.h"
+
+
+/**
+ * A call to lp_build_assert() will build a function call to this function.
+ */
+static void
+lp_assert(int condition, const char *msg)
+{
+   if (!condition) {
+      debug_printf("LLVM assertion '%s' failed!\n", msg);
+      assert(condition);
+   }
+}
+
+
+
+/**
+ * lp_build_assert.
+ *
+ * Build an assertion in LLVM IR by building a function call to the
+ * lp_assert() function above.
+ *
+ * \param condition should be an 'i1' or 'i32' value
+ * \param msg  a string to print if the assertion fails.
+ */
+LLVMValueRef
+lp_build_assert(LLVMBuilderRef builder, LLVMValueRef condition,
+                const char *msg)
+{
+   LLVMModuleRef module;
+   LLVMTypeRef arg_types[2];
+   LLVMValueRef msg_string, assert_func, params[2], r;
+
+   module = LLVMGetGlobalParent(LLVMGetBasicBlockParent(
+                            LLVMGetInsertBlock(builder)));
+
+   msg_string = lp_build_const_string_variable(module, msg, strlen(msg) + 1);
+
+   arg_types[0] = LLVMInt32Type();
+   arg_types[1] = LLVMPointerType(LLVMInt8Type(), 0);
+
+   /* lookup the lp_assert function */
+   assert_func = LLVMGetNamedFunction(module, "lp_assert");
+
+   /* Create the assertion function if not found */
+   if (!assert_func) {
+      LLVMTypeRef func_type =
+         LLVMFunctionType(LLVMVoidType(), arg_types, 2, 0);
+
+      assert_func = LLVMAddFunction(module, "lp_assert", func_type);
+      LLVMSetFunctionCallConv(assert_func, LLVMCCallConv);
+      LLVMSetLinkage(assert_func, LLVMExternalLinkage);
+      LLVMAddGlobalMapping(lp_build_engine, assert_func,
+                           func_to_pointer((func_pointer)lp_assert));
+   }
+   assert(assert_func);
+
+   /* build function call param list */
+   params[0] = LLVMBuildZExt(builder, condition, arg_types[0], "");
+   params[1] = LLVMBuildBitCast(builder, msg_string, arg_types[1], "");
+
+   /* check arg types */
+   assert(LLVMTypeOf(params[0]) == arg_types[0]);
+   assert(LLVMTypeOf(params[1]) == arg_types[1]);
+
+   r = LLVMBuildCall(builder, assert_func, params, 2, "");
+
+   return r;
+}
diff --git a/src/gallium/state_trackers/dri/drm/dri2.h b/src/gallium/auxiliary/gallivm/lp_bld_assert.h
index 07adfe4f6c5..ddd879dc2c6 100644
--- a/src/gallium/state_trackers/dri/drm/dri2.h
+++ b/src/gallium/auxiliary/gallivm/lp_bld_assert.h
@@ -1,6 +1,6 @@
 /**************************************************************************
  *
- * Copyright 2009, VMware, Inc.
+ * Copyright 2010 VMware, Inc.
  * All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -25,13 +25,17 @@
  *
  **************************************************************************/
 
-#ifndef DRI2_H
-#define DRI2_H
+#ifndef LP_BLD_ASSERT_H
+#define LP_BLD_ASSERT_H
 
-#include "dri_drawable.h"
-#include "dri_wrapper.h"
 
-const __DRIconfig **
-dri2_init_screen(__DRIscreen * sPriv);
+#include "lp_bld.h"
+
+
+LLVMValueRef
+lp_build_assert(LLVMBuilderRef builder, LLVMValueRef condition,
+                const char *msg);
+
+
+#endif
 
-#endif /* DRI2_H */
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_conv.c b/src/gallium/auxiliary/gallivm/lp_bld_conv.c
index 77012f1fac6..8b477313d48 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_conv.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_conv.c
@@ -117,8 +117,8 @@ lp_build_clamped_float_to_unsigned_norm(LLVMBuilderRef builder,
    scale = (double)mask/ubound;
    bias = (double)((unsigned long long)1 << (mantissa - n));
 
-   res = LLVMBuildMul(builder, src, lp_build_const_vec(src_type, scale), "");
-   res = LLVMBuildAdd(builder, res, lp_build_const_vec(src_type, bias), "");
+   res = LLVMBuildFMul(builder, src, lp_build_const_vec(src_type, scale), "");
+   res = LLVMBuildFAdd(builder, res, lp_build_const_vec(src_type, bias), "");
    res = LLVMBuildBitCast(builder, res, int_vec_type, "");
 
    if(dst_width > n) {
@@ -175,6 +175,8 @@ lp_build_unsigned_norm_to_float(LLVMBuilderRef builder,
    double scale;
    double bias;
 
+   assert(dst_type.floating);
+
    mantissa = lp_mantissa(dst_type);
 
    n = MIN2(mantissa, src_width);
@@ -199,8 +201,8 @@ lp_build_unsigned_norm_to_float(LLVMBuilderRef builder,
 
    res = LLVMBuildBitCast(builder, res, vec_type, "");
 
-   res = LLVMBuildSub(builder, res, bias_, "");
-   res = LLVMBuildMul(builder, res, lp_build_const_vec(dst_type, scale), "");
+   res = LLVMBuildFSub(builder, res, bias_, "");
+   res = LLVMBuildFMul(builder, res, lp_build_const_vec(dst_type, scale), "");
 
    return res;
 }
@@ -296,7 +298,7 @@ lp_build_conv(LLVMBuilderRef builder,
          if (dst_scale != 1.0) {
             LLVMValueRef scale = lp_build_const_vec(tmp_type, dst_scale);
             for(i = 0; i < num_tmps; ++i)
-               tmp[i] = LLVMBuildMul(builder, tmp[i], scale, "");
+               tmp[i] = LLVMBuildFMul(builder, tmp[i], scale, "");
          }
 
          /* Use an equally sized integer for intermediate computations */
@@ -391,7 +393,7 @@ lp_build_conv(LLVMBuilderRef builder,
           if (src_scale != 1.0) {
              LLVMValueRef scale = lp_build_const_vec(tmp_type, 1.0/src_scale);
              for(i = 0; i < num_tmps; ++i)
-                tmp[i] = LLVMBuildMul(builder, tmp[i], scale, "");
+                tmp[i] = LLVMBuildFMul(builder, tmp[i], scale, "");
           }
       }
     }
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_format_aos.c b/src/gallium/auxiliary/gallivm/lp_bld_format_aos.c
index 0f01fc1d75f..247cb83ce6c 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_format_aos.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_format_aos.c
@@ -240,7 +240,7 @@ lp_build_unpack_arith_rgba_aos(LLVMBuilderRef builder,
     */
 
    if (normalized)
-      scaled = LLVMBuildMul(builder, casted, LLVMConstVector(scales, 4), "");
+      scaled = LLVMBuildFMul(builder, casted, LLVMConstVector(scales, 4), "");
    else
       scaled = casted;
 
@@ -322,7 +322,7 @@ lp_build_pack_rgba_aos(LLVMBuilderRef builder,
    }
 
    if (normalized)
-      scaled = LLVMBuildMul(builder, unswizzled, LLVMConstVector(scales, 4), "");
+      scaled = LLVMBuildFMul(builder, unswizzled, LLVMConstVector(scales, 4), "");
    else
       scaled = unswizzled;
 
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_format_soa.c b/src/gallium/auxiliary/gallivm/lp_bld_format_soa.c
index 9f405921b0a..c724a4453e6 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_format_soa.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_format_soa.c
@@ -197,7 +197,7 @@ lp_build_unpack_rgba_soa(LLVMBuilderRef builder,
             if (format_desc->channel[chan].normalized) {
                double scale = 1.0 / ((1 << (format_desc->channel[chan].size - 1)) - 1);
                LLVMValueRef scale_val = lp_build_const_vec(type, scale);
-               input = LLVMBuildMul(builder, input, scale_val, "");
+               input = LLVMBuildFMul(builder, input, scale_val, "");
             }
          }
          else {
@@ -227,7 +227,7 @@ lp_build_unpack_rgba_soa(LLVMBuilderRef builder,
             double scale = 1.0 / ((1 << (format_desc->channel[chan].size/2)) - 1);
             LLVMValueRef scale_val = lp_build_const_vec(type, scale);
             input = LLVMBuildSIToFP(builder, input, lp_build_vec_type(type), "");
-            input = LLVMBuildMul(builder, input, scale_val, "");
+            input = LLVMBuildFMul(builder, input, scale_val, "");
          }
          else {
             /* FIXME */
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_init.c b/src/gallium/auxiliary/gallivm/lp_bld_init.c
index 69353dea09e..60d8bcfa55e 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_init.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_init.c
@@ -45,6 +45,8 @@ static const struct debug_named_value lp_bld_debug_flags[] = {
    { "nopt",   GALLIVM_DEBUG_NO_OPT, NULL },
    DEBUG_NAMED_VALUE_END
 };
+
+DEBUG_GET_ONCE_FLAGS_OPTION(gallivm_debug, "GALLIVM_DEBUG", lp_bld_debug_flags, 0)
 #endif
 
 
@@ -89,7 +91,7 @@ void
 lp_build_init(void)
 {
 #ifdef DEBUG
-   gallivm_debug = debug_get_flags_option("GALLIVM_DEBUG", lp_bld_debug_flags, 0 );
+   gallivm_debug = debug_get_option_gallivm_debug();
 #endif
 
    lp_set_target_options();
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_init.h b/src/gallium/auxiliary/gallivm/lp_bld_init.h
index a32ced9b4c3..f26fdac4663 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_init.h
+++ b/src/gallium/auxiliary/gallivm/lp_bld_init.h
@@ -44,5 +44,7 @@ extern LLVMPassManagerRef lp_build_pass;
 void
 lp_build_init(void);
 
+extern void
+lp_func_delete_body(LLVMValueRef func);
 
 #endif /* !LP_BLD_INIT_H */
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_logic.c b/src/gallium/auxiliary/gallivm/lp_bld_logic.c
index 39854e43b19..7d7db3b0d92 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_logic.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_logic.c
@@ -83,6 +83,8 @@ lp_build_compare(LLVMBuilderRef builder,
 
    assert(func >= PIPE_FUNC_NEVER);
    assert(func <= PIPE_FUNC_ALWAYS);
+   assert(lp_check_value(type, a));
+   assert(lp_check_value(type, b));
 
    if(func == PIPE_FUNC_NEVER)
       return zeros;
@@ -363,9 +365,55 @@ lp_build_cmp(struct lp_build_context *bld,
 
 
 /**
+ * Return (mask & a) | (~mask & b);
+ */
+LLVMValueRef
+lp_build_select_bitwise(struct lp_build_context *bld,
+                        LLVMValueRef mask,
+                        LLVMValueRef a,
+                        LLVMValueRef b)
+{
+   struct lp_type type = bld->type;
+   LLVMValueRef res;
+
+   assert(lp_check_value(type, a));
+   assert(lp_check_value(type, b));
+
+   if (a == b) {
+      return a;
+   }
+
+   if(type.floating) {
+      LLVMTypeRef int_vec_type = lp_build_int_vec_type(type);
+      a = LLVMBuildBitCast(bld->builder, a, int_vec_type, "");
+      b = LLVMBuildBitCast(bld->builder, b, int_vec_type, "");
+   }
+
+   a = LLVMBuildAnd(bld->builder, a, mask, "");
+
+   /* This often gets translated to PANDN, but sometimes the NOT is
+    * pre-computed and stored in another constant. The best strategy depends
+    * on available registers, so it is not a big deal -- hopefully LLVM does
+    * the right decision attending the rest of the program.
+    */
+   b = LLVMBuildAnd(bld->builder, b, LLVMBuildNot(bld->builder, mask, ""), "");
+
+   res = LLVMBuildOr(bld->builder, a, b, "");
+
+   if(type.floating) {
+      LLVMTypeRef vec_type = lp_build_vec_type(type);
+      res = LLVMBuildBitCast(bld->builder, res, vec_type, "");
+   }
+
+   return res;
+}
+
+
+/**
  * Return mask ? a : b;
  *
- * mask is a bitwise mask, composed of 0 or ~0 for each element.
+ * mask is a bitwise mask, composed of 0 or ~0 for each element. Any other value
+ * will yield unpredictable results.
  */
 LLVMValueRef
 lp_build_select(struct lp_build_context *bld,
@@ -376,6 +424,9 @@ lp_build_select(struct lp_build_context *bld,
    struct lp_type type = bld->type;
    LLVMValueRef res;
 
+   assert(lp_check_value(type, a));
+   assert(lp_check_value(type, b));
+
    if(a == b)
       return a;
 
@@ -424,27 +475,7 @@ lp_build_select(struct lp_build_context *bld,
       }
    }
    else {
-      if(type.floating) {
-         LLVMTypeRef int_vec_type = lp_build_int_vec_type(type);
-         a = LLVMBuildBitCast(bld->builder, a, int_vec_type, "");
-         b = LLVMBuildBitCast(bld->builder, b, int_vec_type, "");
-      }
-
-      a = LLVMBuildAnd(bld->builder, a, mask, "");
-
-      /* This often gets translated to PANDN, but sometimes the NOT is
-       * pre-computed and stored in another constant. The best strategy depends
-       * on available registers, so it is not a big deal -- hopefully LLVM does
-       * the right decision attending the rest of the program.
-       */
-      b = LLVMBuildAnd(bld->builder, b, LLVMBuildNot(bld->builder, mask, ""), "");
-
-      res = LLVMBuildOr(bld->builder, a, b, "");
-
-      if(type.floating) {
-         LLVMTypeRef vec_type = lp_build_vec_type(type);
-         res = LLVMBuildBitCast(bld->builder, res, vec_type, "");
-      }
+      res = lp_build_select_bitwise(bld, mask, a, b);
    }
 
    return res;
@@ -461,6 +492,9 @@ lp_build_select_aos(struct lp_build_context *bld,
    const unsigned n = type.length;
    unsigned i, j;
 
+   assert(lp_check_value(type, a));
+   assert(lp_check_value(type, b));
+
    if(a == b)
       return a;
    if(cond[0] && cond[1] && cond[2] && cond[3])
@@ -516,7 +550,22 @@ lp_build_select_aos(struct lp_build_context *bld,
 LLVMValueRef
 lp_build_andc(struct lp_build_context *bld, LLVMValueRef a, LLVMValueRef b)
 {
+   const struct lp_type type = bld->type;
+
+   assert(lp_check_value(type, a));
+   assert(lp_check_value(type, b));
+
+   /* can't do bitwise ops on floating-point values */
+   if(type.floating) {
+      a = LLVMBuildBitCast(bld->builder, a, bld->int_vec_type, "");
+      b = LLVMBuildBitCast(bld->builder, b, bld->int_vec_type, "");
+   }
+
    b = LLVMBuildNot(bld->builder, b, "");
    b = LLVMBuildAnd(bld->builder, a, b, "");
+
+   if(type.floating) {
+      b = LLVMBuildBitCast(bld->builder, b, bld->vec_type, "");
+   }
    return b;
 }
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_logic.h b/src/gallium/auxiliary/gallivm/lp_bld_logic.h
index 29f9fc3b205..4e7b4c9938e 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_logic.h
+++ b/src/gallium/auxiliary/gallivm/lp_bld_logic.h
@@ -63,6 +63,11 @@ lp_build_cmp(struct lp_build_context *bld,
              LLVMValueRef a,
              LLVMValueRef b);
 
+LLVMValueRef
+lp_build_select_bitwise(struct lp_build_context *bld,
+                        LLVMValueRef mask,
+                        LLVMValueRef a,
+                        LLVMValueRef b);
 
 LLVMValueRef
 lp_build_select(struct lp_build_context *bld,
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_misc.cpp b/src/gallium/auxiliary/gallivm/lp_bld_misc.cpp
index 5a9488b5f79..6d5410d9701 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_misc.cpp
+++ b/src/gallium/auxiliary/gallivm/lp_bld_misc.cpp
@@ -39,6 +39,7 @@
 #include <llvm/Target/TargetOptions.h>
 #include <llvm/ExecutionEngine/ExecutionEngine.h>
 #include <llvm/ExecutionEngine/JITEventListener.h>
+#include <llvm/Support/CommandLine.h>
 
 #include "pipe/p_config.h"
 #include "util/u_debug.h"
@@ -141,4 +142,35 @@ lp_set_target_options(void)
 #if 0
    llvm::UnsafeFPMath = true;
 #endif
+
+#if 0
+   /*
+    * LLVM will generate MMX instructions for vectors <= 64 bits, leading to
+    * innefficient code, and in 32bit systems, to the corruption of the FPU
+    * stack given that it expects the user to generate the EMMS instructions.
+    *
+    * See also:
+    * - http://llvm.org/bugs/show_bug.cgi?id=3287
+    * - http://l4.me.uk/post/2009/06/07/llvm-wrinkle-3-configuration-what-configuration/
+    *
+    * XXX: Unfortunately this is not working.
+    */
+   static boolean first = FALSE;
+   if (first) {
+      static const char* options[] = {
+         "prog",
+         "-disable-mmx"
+      };
+      llvm::cl::ParseCommandLineOptions(2, const_cast<char**>(options));
+      first = FALSE;
+   }
+#endif
+}
+
+
+extern "C" void
+lp_func_delete_body(LLVMValueRef FF)
+{
+   llvm::Function *func = llvm::unwrap<llvm::Function>(FF);
+   func->deleteBody();
 }
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_pack.c b/src/gallium/auxiliary/gallivm/lp_bld_pack.c
index 7748f8f0999..b7b630f2e8d 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_pack.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_pack.c
@@ -171,14 +171,13 @@ lp_build_unpack2(LLVMBuilderRef builder,
       msb = lp_build_zero(src_type);
 
    /* Interleave bits */
-   if(util_cpu_caps.little_endian) {
+#ifdef PIPE_ARCH_LITTLE_ENDIAN
       *dst_lo = lp_build_interleave2(builder, src_type, src, msb, 0);
       *dst_hi = lp_build_interleave2(builder, src_type, src, msb, 1);
-   }
-   else {
+#else
       *dst_lo = lp_build_interleave2(builder, src_type, msb, src, 0);
       *dst_hi = lp_build_interleave2(builder, src_type, msb, src, 1);
-   }
+#endif
 
    /* Cast the result into the new type (twice as wide) */
 
@@ -261,13 +260,14 @@ lp_build_pack2(LLVMBuilderRef builder,
 #endif
    LLVMTypeRef dst_vec_type = lp_build_vec_type(dst_type);
    LLVMValueRef shuffle;
-   LLVMValueRef res;
+   LLVMValueRef res = NULL;
 
    assert(!src_type.floating);
    assert(!dst_type.floating);
    assert(src_type.width == dst_type.width * 2);
    assert(src_type.length * 2 == dst_type.length);
 
+   /* Check for special cases first */
    if(util_cpu_caps.has_sse2 && src_type.width * src_type.length == 128) {
       switch(src_type.width) {
       case 32:
@@ -283,8 +283,8 @@ lp_build_pack2(LLVMBuilderRef builder,
                return lp_build_intrinsic_binary(builder, "llvm.x86.sse41.packusdw", dst_vec_type, lo, hi);
             }
             else {
-               assert(0);
-               return LLVMGetUndef(dst_vec_type);
+               /* use generic shuffle below */
+               res = NULL;
             }
          }
          break;
@@ -310,10 +310,13 @@ lp_build_pack2(LLVMBuilderRef builder,
          break;
       }
 
-      res = LLVMBuildBitCast(builder, res, dst_vec_type, "");
-      return res;
+      if (res) {
+         res = LLVMBuildBitCast(builder, res, dst_vec_type, "");
+         return res;
+      }
    }
 
+   /* generic shuffle */
    lo = LLVMBuildBitCast(builder, lo, dst_vec_type, "");
    hi = LLVMBuildBitCast(builder, hi, dst_vec_type, "");
 
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_quad.c b/src/gallium/auxiliary/gallivm/lp_bld_quad.c
index ca36046d222..7b1088939b9 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_quad.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_quad.c
@@ -85,7 +85,7 @@ lp_build_scalar_ddx(struct lp_build_context *bld,
    LLVMValueRef idx_right = LLVMConstInt(LLVMInt32Type(), LP_BLD_QUAD_TOP_RIGHT, 0);
    LLVMValueRef a_left  = LLVMBuildExtractElement(bld->builder, a, idx_left, "");
    LLVMValueRef a_right = LLVMBuildExtractElement(bld->builder, a, idx_right, "");
-   return LLVMBuildSub(bld->builder, a_right, a_left, "");
+   return lp_build_sub(bld, a_right, a_left);
 }
 
 
@@ -97,5 +97,5 @@ lp_build_scalar_ddy(struct lp_build_context *bld,
    LLVMValueRef idx_bottom = LLVMConstInt(LLVMInt32Type(), LP_BLD_QUAD_BOTTOM_LEFT, 0);
    LLVMValueRef a_top    = LLVMBuildExtractElement(bld->builder, a, idx_top, "");
    LLVMValueRef a_bottom = LLVMBuildExtractElement(bld->builder, a, idx_bottom, "");
-   return LLVMBuildSub(bld->builder, a_bottom, a_top, "");
+   return lp_build_sub(bld, a_bottom, a_top);
 }
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c b/src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c
index 1a20d74cac8..806c7d56a87 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c
@@ -40,7 +40,6 @@
 #include "util/u_memory.h"
 #include "util/u_math.h"
 #include "util/u_format.h"
-#include "util/u_cpu_detect.h"
 #include "lp_bld_debug.h"
 #include "lp_bld_type.h"
 #include "lp_bld_const.h"
@@ -811,7 +810,7 @@ lp_build_minify(struct lp_build_sample_context *bld,
                 LLVMValueRef base_size,
                 LLVMValueRef level)
 {
-   LLVMValueRef size = LLVMBuildAShr(bld->builder, base_size, level, "minify");
+   LLVMValueRef size = LLVMBuildLShr(bld->builder, base_size, level, "minify");
    size = lp_build_max(&bld->int_coord_bld, size, bld->int_coord_bld.one);
    return size;
 }
@@ -888,17 +887,17 @@ lp_build_lod_selector(struct lp_build_sample_context *bld,
          /* Compute rho = max of all partial derivatives scaled by texture size.
           * XXX this could be vectorized somewhat
           */
-         rho = LLVMBuildMul(bld->builder,
+         rho = LLVMBuildFMul(bld->builder,
                             lp_build_max(float_bld, dsdx, dsdy),
                             lp_build_int_to_float(float_bld, width), "");
          if (dims > 1) {
             LLVMValueRef max;
-            max = LLVMBuildMul(bld->builder,
+            max = LLVMBuildFMul(bld->builder,
                                lp_build_max(float_bld, dtdx, dtdy),
                                lp_build_int_to_float(float_bld, height), "");
             rho = lp_build_max(float_bld, rho, max);
             if (dims > 2) {
-               max = LLVMBuildMul(bld->builder,
+               max = LLVMBuildFMul(bld->builder,
                                   lp_build_max(float_bld, drdx, drdy),
                                   lp_build_int_to_float(float_bld, depth), "");
                rho = lp_build_max(float_bld, rho, max);
@@ -912,12 +911,12 @@ lp_build_lod_selector(struct lp_build_sample_context *bld,
          if (lod_bias) {
             lod_bias = LLVMBuildExtractElement(bld->builder, lod_bias,
                                                index0, "");
-            lod = LLVMBuildAdd(bld->builder, lod, lod_bias, "shader_lod_bias");
+            lod = LLVMBuildFAdd(bld->builder, lod, lod_bias, "shader_lod_bias");
          }
       }
 
       /* add sampler lod bias */
-      lod = LLVMBuildAdd(bld->builder, lod, sampler_lod_bias, "sampler_lod_bias");
+      lod = LLVMBuildFAdd(bld->builder, lod, sampler_lod_bias, "sampler_lod_bias");
 
       /* clamp lod */
       lod = lp_build_clamp(float_bld, lod, min_lod, max_lod);
@@ -1219,8 +1218,7 @@ lp_build_cube_ima(struct lp_build_context *coord_bld, LLVMValueRef coord)
    /* ima = -0.5 / abs(coord); */
    LLVMValueRef negHalf = lp_build_const_vec(coord_bld->type, -0.5);
    LLVMValueRef absCoord = lp_build_abs(coord_bld, coord);
-   LLVMValueRef ima = lp_build_mul(coord_bld, negHalf,
-                                   lp_build_rcp(coord_bld, absCoord));
+   LLVMValueRef ima = lp_build_div(coord_bld, negHalf, absCoord);
    return ima;
 }
 
@@ -1841,7 +1839,11 @@ lp_build_sample_2d_linear_aos(struct lp_build_sample_context *bld,
       unsigned i, j;
 
       for(j = 0; j < h16.type.length; j += 4) {
-         unsigned subindex = util_cpu_caps.little_endian ? 0 : 1;
+#ifdef PIPE_ARCH_LITTLE_ENDIAN
+         unsigned subindex = 0;
+#else
+         unsigned subindex = 1;
+#endif
          LLVMValueRef index;
 
          index = LLVMConstInt(elem_type, j/2 + subindex, 0);
@@ -2029,6 +2031,8 @@ lp_build_sample_soa(LLVMBuilderRef builder,
       debug_printf("Sample from %s\n", util_format_name(fmt));
    }
 
+   assert(type.floating);
+
    /* Setup our build context */
    memset(&bld, 0, sizeof bld);
    bld.builder = builder;
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c b/src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c
index 21236839fb7..0aa64affacc 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c
@@ -489,7 +489,7 @@ get_indirect_offsets(struct lp_build_tgsi_soa_context *bld,
                               int_vec_type, "");
 
    /* addr_vec = addr_vec * 4 */
-   addr_vec = lp_build_mul(&bld->base, addr_vec, vec4);
+   addr_vec = lp_build_mul(&bld->int_bld, addr_vec, vec4);
 
    return addr_vec;
 }
@@ -533,7 +533,7 @@ emit_fetch(
                                             reg->Register.Index * 4 + swizzle);
 
          /* index_vec = index_vec + addr_vec */
-         index_vec = lp_build_add(&bld->base, index_vec, addr_vec);
+         index_vec = lp_build_add(&bld->int_bld, index_vec, addr_vec);
 
          /* Gather values from the constant buffer */
          res = build_gather(bld, bld->consts_ptr, index_vec);
@@ -612,11 +612,9 @@ emit_fetch(
    case TGSI_UTIL_SIGN_SET:
       /* TODO: Use bitwese OR for floating point */
       res = lp_build_abs( &bld->base, res );
-      res = LLVMBuildNeg( bld->base.builder, res, "" );
-      break;
-
+      /* fall through */
    case TGSI_UTIL_SIGN_TOGGLE:
-      res = LLVMBuildNeg( bld->base.builder, res, "" );
+      res = lp_build_negate( &bld->base, res );
       break;
 
    case TGSI_UTIL_SIGN_KEEP:
@@ -773,7 +771,9 @@ emit_store(
       addr = LLVMBuildExtractElement(bld->base.builder,
                                      addr, LLVMConstInt(LLVMInt32Type(), 0, 0),
                                      "");
-      addr = lp_build_mul(&bld->base, addr, LLVMConstInt(LLVMInt32Type(), 4, 0));
+      addr = LLVMBuildMul(bld->base.builder,
+                          addr, LLVMConstInt(LLVMInt32Type(), 4, 0),
+                          "");
    }
 
    switch( reg->Register.File ) {
diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c
index 5275faa5e22..298f3d0a8bb 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_exec.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c
@@ -557,6 +557,23 @@ print_temp(const struct tgsi_exec_machine *mach, uint index)
 #endif
 
 
+void
+tgsi_exec_set_constant_buffers(struct tgsi_exec_machine *mach,
+                               unsigned num_bufs,
+                               const void **bufs,
+                               const unsigned *buf_sizes)
+{
+   unsigned i;
+
+   for (i = 0; i < num_bufs; i++) {
+      mach->Consts[i] = bufs[i];
+      mach->ConstsSize[i] = buf_sizes[i];
+   }
+}
+
+
+
+
 /**
  * Check if there's a potential src/dst register data dependency when
  * using SOA execution.
@@ -632,6 +649,10 @@ tgsi_exec_machine_bind_shader(
 
    util_init_math();
 
+   if (numSamplers) {
+      assert(samplers);
+   }
+
    mach->Tokens = tokens;
    mach->Samplers = samplers;
 
@@ -1040,6 +1061,8 @@ fetch_src_file_channel(const struct tgsi_exec_machine *mach,
 {
    uint i;
 
+   assert(swizzle < 4);
+
    switch (file) {
    case TGSI_FILE_CONSTANT:
       for (i = 0; i < QUAD_SIZE; i++) {
@@ -1049,9 +1072,23 @@ fetch_src_file_channel(const struct tgsi_exec_machine *mach,
          if (index->i[i] < 0) {
             chan->u[i] = 0;
          } else {
-            const uint *p = (const uint *)mach->Consts[index2D->i[i]];
-
-            chan->u[i] = p[index->i[i] * 4 + swizzle];
+            /* NOTE: copying the const value as a uint instead of float */
+            const uint constbuf = index2D->i[i];
+            const uint *buf = (const uint *)mach->Consts[constbuf];
+            const int pos = index->i[i] * 4 + swizzle;
+            /* const buffer bounds check */
+            if (pos < 0 || pos >= mach->ConstsSize[constbuf]) {
+               if (0) {
+                  /* Debug: print warning */
+                  static int count = 0;
+                  if (count++ < 100)
+                     debug_printf("TGSI Exec: const buffer index %d"
+                                  " out of bounds\n", pos);
+               }
+               chan->u[i] = 0;
+            }
+            else
+               chan->u[i] = buf[pos];
          }
       }
       break;
@@ -1065,9 +1102,10 @@ fetch_src_file_channel(const struct tgsi_exec_machine *mach,
                          index2D->i[i] * TGSI_EXEC_MAX_INPUT_ATTRIBS + index->i[i],
                          index2D->i[i], index->i[i]);
                          }*/
-         chan->u[i] = mach->Inputs[index2D->i[i] *
-                                   TGSI_EXEC_MAX_INPUT_ATTRIBS +
-                                   index->i[i]].xyzw[swizzle].u[i];
+         int pos = index2D->i[i] * TGSI_EXEC_MAX_INPUT_ATTRIBS + index->i[i];
+         assert(pos >= 0);
+         assert(pos < Elements(mach->Inputs));
+         chan->u[i] = mach->Inputs[pos].xyzw[swizzle].u[i];
       }
       break;
 
@@ -1187,7 +1225,7 @@ fetch_source(const struct tgsi_exec_machine *mach,
       index2.i[1] =
       index2.i[2] =
       index2.i[3] = reg->Indirect.Index;
-
+      assert(reg->Indirect.File == TGSI_FILE_ADDRESS);
       /* get current value of address register[swizzle] */
       swizzle = tgsi_util_get_src_register_swizzle( &reg->Indirect, CHAN_X );
       fetch_src_file_channel(mach,
diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.h b/src/gallium/auxiliary/tgsi/tgsi_exec.h
index ccf80ca6fd9..6dee362d589 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_exec.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_exec.h
@@ -253,7 +253,10 @@ struct tgsi_exec_machine
    struct tgsi_sampler           **Samplers;
 
    unsigned                      ImmLimit;
+
    const void *Consts[PIPE_MAX_CONSTANT_BUFFERS];
+   unsigned ConstsSize[PIPE_MAX_CONSTANT_BUFFERS];
+
    const struct tgsi_token       *Tokens;   /**< Declarations, instructions */
    unsigned                      Processor; /**< TGSI_PROCESSOR_x */
 
@@ -367,6 +370,13 @@ tgsi_set_exec_mask(struct tgsi_exec_machine *mach,
 }
 
 
+extern void
+tgsi_exec_set_constant_buffers(struct tgsi_exec_machine *mach,
+                               unsigned num_bufs,
+                               const void **bufs,
+                               const unsigned *buf_sizes);
+
+
 #if defined __cplusplus
 } /* extern "C" */
 #endif
diff --git a/src/gallium/auxiliary/tgsi/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/tgsi_sanity.c
index 97148dbe233..acbff103efe 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_sanity.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_sanity.c
@@ -33,6 +33,10 @@
 #include "tgsi_info.h"
 #include "tgsi_iterate.h"
 
+
+DEBUG_GET_ONCE_BOOL_OPTION(print_sanity, "TGSI_PRINT_SANITY", FALSE)
+
+
 typedef struct {
    uint file : 28;
    /* max 2 dimensions */
@@ -54,6 +58,8 @@ struct sanity_check_ctx
    uint errors;
    uint warnings;
    uint implied_array_size;
+
+   boolean print;
 };
 
 static INLINE unsigned
@@ -148,6 +154,9 @@ report_error(
 {
    va_list args;
 
+   if (!ctx->print)
+      return;
+
    debug_printf( "Error  : " );
    va_start( args, format );
    _debug_vprintf( format, args );
@@ -164,6 +173,9 @@ report_warning(
 {
    va_list args;
 
+   if (!ctx->print)
+      return;
+
    debug_printf( "Warning: " );
    va_start( args, format );
    _debug_vprintf( format, args );
@@ -539,6 +551,7 @@ tgsi_sanity_check(
    ctx.errors = 0;
    ctx.warnings = 0;
    ctx.implied_array_size = 0;
+   ctx.print = debug_get_option_print_sanity();
 
    if (!tgsi_iterate_shader( tokens, &ctx.iter ))
       return FALSE;
diff --git a/src/gallium/auxiliary/tgsi/tgsi_sanity.h b/src/gallium/auxiliary/tgsi/tgsi_sanity.h
index 52263ff8832..73f0f414e3f 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_sanity.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_sanity.h
@@ -35,7 +35,8 @@ extern "C" {
 #endif
 
 /* Check the given token stream for errors and common mistakes.
- * Diagnostic messages are printed out to the debug output.
+ * Diagnostic messages are printed out to the debug output, and is
+ * controlled by the debug option TGSI_PRINT_SANITY (default false).
  * Returns TRUE if there are no errors, even though there could be some warnings.
  */
 boolean
diff --git a/src/gallium/auxiliary/translate/translate.c b/src/gallium/auxiliary/translate/translate.c
index a9b7253bf44..fe638e211fa 100644
--- a/src/gallium/auxiliary/translate/translate.c
+++ b/src/gallium/auxiliary/translate/translate.c
@@ -48,3 +48,8 @@ struct translate *translate_create( const struct translate_key *key )
 
    return translate_generic_create( key );
 }
+
+boolean translate_is_output_format_supported(enum pipe_format format)
+{
+   return translate_generic_is_output_format_supported(format);
+}
diff --git a/src/gallium/auxiliary/translate/translate.h b/src/gallium/auxiliary/translate/translate.h
index edd95e07882..eb6f2cc4862 100644
--- a/src/gallium/auxiliary/translate/translate.h
+++ b/src/gallium/auxiliary/translate/translate.h
@@ -105,6 +105,8 @@ struct translate *translate_lookup_or_create( struct translate_context *tctx,
 
 struct translate *translate_create( const struct translate_key *key );
 
+boolean translate_is_output_format_supported(enum pipe_format format);
+
 static INLINE int translate_keysize( const struct translate_key *key )
 {
    return 2 * sizeof(int) + key->nr_elements * sizeof(struct translate_element);
@@ -138,5 +140,6 @@ struct translate *translate_sse2_create( const struct translate_key *key );
 
 struct translate *translate_generic_create( const struct translate_key *key );
 
+boolean translate_generic_is_output_format_supported(enum pipe_format format);
 
 #endif
diff --git a/src/gallium/auxiliary/translate/translate_generic.c b/src/gallium/auxiliary/translate/translate_generic.c
index 0e43a512ee8..42cfd763e9c 100644
--- a/src/gallium/auxiliary/translate/translate_generic.c
+++ b/src/gallium/auxiliary/translate/translate_generic.c
@@ -187,9 +187,15 @@ ATTRIB( R8G8B8_SNORM,    3, char, TO_8_SNORM )
 ATTRIB( R8G8_SNORM,      2, char, TO_8_SNORM )
 ATTRIB( R8_SNORM,        1, char, TO_8_SNORM )
 
-ATTRIB( A8R8G8B8_UNORM,       4, ubyte, TO_8_UNORM )
-/*ATTRIB( R8G8B8A8_UNORM,       4, ubyte, TO_8_UNORM )*/
-
+static void
+emit_A8R8G8B8_UNORM( const float *attrib, void *ptr)
+{
+   ubyte *out = (ubyte *)ptr;
+   out[0] = TO_8_UNORM(attrib[3]);
+   out[1] = TO_8_UNORM(attrib[0]);
+   out[2] = TO_8_UNORM(attrib[1]);
+   out[3] = TO_8_UNORM(attrib[2]);
+}
 
 static void
 emit_B8G8R8A8_UNORM( const float *attrib, void *ptr)
@@ -368,23 +374,23 @@ static void PIPE_CDECL generic_run_elts( struct translate *translate,
    /* loop over vertex attributes (vertex shader inputs)
     */
    for (i = 0; i < count; i++) {
-      unsigned elt = *elts++;
+      const unsigned elt = *elts++;
 
       for (attr = 0; attr < nr_attrs; attr++) {
 	 float data[4];
-         const uint8_t *src;
-         unsigned index;
-
-	 char *dst = (vert + 
-		      tg->attrib[attr].output_offset);
+	 char *dst = vert + tg->attrib[attr].output_offset;
 
          if (tg->attrib[attr].type == TRANSLATE_ELEMENT_NORMAL) {
+            const uint8_t *src;
+            unsigned index;
+
             if (tg->attrib[attr].instance_divisor) {
                index = instance_id / tg->attrib[attr].instance_divisor;
             } else {
                index = elt;
             }
 
+            /* clamp to void going out of bounds */
             index = MIN2(index, tg->attrib[attr].max_index);
 
             src = tg->attrib[attr].input_ptr +
@@ -392,11 +398,23 @@ static void PIPE_CDECL generic_run_elts( struct translate *translate,
 
             tg->attrib[attr].fetch( data, src, 0, 0 );
 
+            if (0)
+               debug_printf("Fetch elt attr %d  from %p  stride %d  div %u  max %u  index %d:  "
+                            " %f, %f, %f, %f \n",
+                            attr,
+                            tg->attrib[attr].input_ptr,
+                            tg->attrib[attr].input_stride,
+                            tg->attrib[attr].instance_divisor,
+                            tg->attrib[attr].max_index,
+                            index,
+                            data[0], data[1],data[2], data[3]);
          } else {
             data[0] = (float)instance_id;
          }
-         if (0) debug_printf("vert %d/%d attr %d: %f %f %f %f\n",
-                             i, elt, attr, data[0], data[1], data[2], data[3]);
+
+         if (0)
+            debug_printf("vert %d/%d attr %d: %f %f %f %f\n",
+                         i, elt, attr, data[0], data[1], data[2], data[3]);
 
 	 tg->attrib[attr].emit( data, dst );
       }
@@ -425,29 +443,42 @@ static void PIPE_CDECL generic_run( struct translate *translate,
 
       for (attr = 0; attr < nr_attrs; attr++) {
 	 float data[4];
-
-	 char *dst = (vert + 
-		      tg->attrib[attr].output_offset);
+	 char *dst = vert + tg->attrib[attr].output_offset;
 
          if (tg->attrib[attr].type == TRANSLATE_ELEMENT_NORMAL) {
             const uint8_t *src;
+            unsigned index;
 
             if (tg->attrib[attr].instance_divisor) {
-               src = tg->attrib[attr].input_ptr +
-                     tg->attrib[attr].input_stride *
-                     (instance_id / tg->attrib[attr].instance_divisor);
-            } else {
-               src = tg->attrib[attr].input_ptr +
-                     tg->attrib[attr].input_stride * elt;
+               index = instance_id / tg->attrib[attr].instance_divisor;
             }
+            else {
+               index = elt;
+            }
+
+            /* clamp to void going out of bounds */
+            index = MIN2(index, tg->attrib[attr].max_index);
+
+            src = tg->attrib[attr].input_ptr +
+                  tg->attrib[attr].input_stride * index;
 
             tg->attrib[attr].fetch( data, src, 0, 0 );
+
+            if (0)
+               debug_printf("Fetch linear attr %d  from %p  stride %d  index %d: "
+                            " %f, %f, %f, %f \n",
+                            attr,
+                            tg->attrib[attr].input_ptr,
+                            tg->attrib[attr].input_stride,
+                            index,
+                            data[0], data[1],data[2], data[3]);
          } else {
             data[0] = (float)instance_id;
          }
 
-         if (0) debug_printf("vert %d attr %d: %f %f %f %f\n",
-                             i, attr, data[0], data[1], data[2], data[3]);
+         if (0)
+            debug_printf("vert %d attr %d: %f %f %f %f\n",
+                         i, attr, data[0], data[1], data[2], data[3]);
 
 	 tg->attrib[attr].emit( data, dst );
       }
@@ -523,3 +554,83 @@ struct translate *translate_generic_create( const struct translate_key *key )
 
    return &tg->translate;
 }
+
+boolean translate_generic_is_output_format_supported(enum pipe_format format)
+{
+   switch(format)
+   {
+   case PIPE_FORMAT_R64G64B64A64_FLOAT: return TRUE;
+   case PIPE_FORMAT_R64G64B64_FLOAT: return TRUE;
+   case PIPE_FORMAT_R64G64_FLOAT: return TRUE;
+   case PIPE_FORMAT_R64_FLOAT: return TRUE;
+
+   case PIPE_FORMAT_R32G32B32A32_FLOAT: return TRUE;
+   case PIPE_FORMAT_R32G32B32_FLOAT: return TRUE;
+   case PIPE_FORMAT_R32G32_FLOAT: return TRUE;
+   case PIPE_FORMAT_R32_FLOAT: return TRUE;
+
+   case PIPE_FORMAT_R32G32B32A32_USCALED: return TRUE;
+   case PIPE_FORMAT_R32G32B32_USCALED: return TRUE;
+   case PIPE_FORMAT_R32G32_USCALED: return TRUE;
+   case PIPE_FORMAT_R32_USCALED: return TRUE;
+
+   case PIPE_FORMAT_R32G32B32A32_SSCALED: return TRUE;
+   case PIPE_FORMAT_R32G32B32_SSCALED: return TRUE;
+   case PIPE_FORMAT_R32G32_SSCALED: return TRUE;
+   case PIPE_FORMAT_R32_SSCALED: return TRUE;
+
+   case PIPE_FORMAT_R32G32B32A32_UNORM: return TRUE;
+   case PIPE_FORMAT_R32G32B32_UNORM: return TRUE;
+   case PIPE_FORMAT_R32G32_UNORM: return TRUE;
+   case PIPE_FORMAT_R32_UNORM: return TRUE;
+
+   case PIPE_FORMAT_R32G32B32A32_SNORM: return TRUE;
+   case PIPE_FORMAT_R32G32B32_SNORM: return TRUE;
+   case PIPE_FORMAT_R32G32_SNORM: return TRUE;
+   case PIPE_FORMAT_R32_SNORM: return TRUE;
+
+   case PIPE_FORMAT_R16G16B16A16_USCALED: return TRUE;
+   case PIPE_FORMAT_R16G16B16_USCALED: return TRUE;
+   case PIPE_FORMAT_R16G16_USCALED: return TRUE;
+   case PIPE_FORMAT_R16_USCALED: return TRUE;
+
+   case PIPE_FORMAT_R16G16B16A16_SSCALED: return TRUE;
+   case PIPE_FORMAT_R16G16B16_SSCALED: return TRUE;
+   case PIPE_FORMAT_R16G16_SSCALED: return TRUE;
+   case PIPE_FORMAT_R16_SSCALED: return TRUE;
+
+   case PIPE_FORMAT_R16G16B16A16_UNORM: return TRUE;
+   case PIPE_FORMAT_R16G16B16_UNORM: return TRUE;
+   case PIPE_FORMAT_R16G16_UNORM: return TRUE;
+   case PIPE_FORMAT_R16_UNORM: return TRUE;
+
+   case PIPE_FORMAT_R16G16B16A16_SNORM: return TRUE;
+   case PIPE_FORMAT_R16G16B16_SNORM: return TRUE;
+   case PIPE_FORMAT_R16G16_SNORM: return TRUE;
+   case PIPE_FORMAT_R16_SNORM: return TRUE;
+
+   case PIPE_FORMAT_R8G8B8A8_USCALED: return TRUE;
+   case PIPE_FORMAT_R8G8B8_USCALED: return TRUE;
+   case PIPE_FORMAT_R8G8_USCALED: return TRUE;
+   case PIPE_FORMAT_R8_USCALED: return TRUE;
+
+   case PIPE_FORMAT_R8G8B8A8_SSCALED: return TRUE;
+   case PIPE_FORMAT_R8G8B8_SSCALED: return TRUE;
+   case PIPE_FORMAT_R8G8_SSCALED: return TRUE;
+   case PIPE_FORMAT_R8_SSCALED: return TRUE;
+
+   case PIPE_FORMAT_R8G8B8A8_UNORM: return TRUE;
+   case PIPE_FORMAT_R8G8B8_UNORM: return TRUE;
+   case PIPE_FORMAT_R8G8_UNORM: return TRUE;
+   case PIPE_FORMAT_R8_UNORM: return TRUE;
+
+   case PIPE_FORMAT_R8G8B8A8_SNORM: return TRUE;
+   case PIPE_FORMAT_R8G8B8_SNORM: return TRUE;
+   case PIPE_FORMAT_R8G8_SNORM: return TRUE;
+   case PIPE_FORMAT_R8_SNORM: return TRUE;
+
+   case PIPE_FORMAT_A8R8G8B8_UNORM: return TRUE;
+   case PIPE_FORMAT_B8G8R8A8_UNORM: return TRUE;
+   default: return FALSE;
+   }
+}
diff --git a/src/gallium/auxiliary/util/u_blitter.c b/src/gallium/auxiliary/util/u_blitter.c
index 0d94aaae956..b5b86b72142 100644
--- a/src/gallium/auxiliary/util/u_blitter.c
+++ b/src/gallium/auxiliary/util/u_blitter.c
@@ -87,6 +87,7 @@ struct blitter_context_priv
    void *dsa_write_depth_keep_stencil;
    void *dsa_keep_depth_stencil;
    void *dsa_keep_depth_write_stencil;
+   void *dsa_flush_depth_stencil;
 
    void *velem_state;
 
@@ -156,6 +157,10 @@ struct blitter_context *util_blitter_create(struct pipe_context *pipe)
    ctx->dsa_keep_depth_stencil =
       pipe->create_depth_stencil_alpha_state(pipe, &dsa);
 
+   dsa.depth.writemask = 1;
+   ctx->dsa_flush_depth_stencil =
+      pipe->create_depth_stencil_alpha_state(pipe, &dsa);
+
    dsa.depth.enabled = 1;
    dsa.depth.writemask = 1;
    dsa.depth.func = PIPE_FUNC_ALWAYS;
@@ -940,3 +945,42 @@ void util_blitter_clear_depth_stencil(struct blitter_context *blitter,
                            UTIL_BLITTER_ATTRIB_NONE, NULL);
    blitter_restore_CSOs(ctx);
 }
+
+/* Clear a region of a depth stencil surface. */
+void util_blitter_flush_depth_stencil(struct blitter_context *blitter,
+                                      struct pipe_surface *dstsurf)
+{
+   struct blitter_context_priv *ctx = (struct blitter_context_priv*)blitter;
+   struct pipe_context *pipe = ctx->base.pipe;
+   struct pipe_framebuffer_state fb_state;
+
+   assert(dstsurf->texture);
+   if (!dstsurf->texture)
+      return;
+
+   /* check the saved state */
+   blitter_check_saved_CSOs(ctx);
+   assert(blitter->saved_fb_state.nr_cbufs != ~0);
+
+   /* bind CSOs */
+   pipe->bind_blend_state(pipe, ctx->blend_keep_color);
+   pipe->bind_depth_stencil_alpha_state(pipe, ctx->dsa_flush_depth_stencil);
+
+   pipe->bind_rasterizer_state(pipe, ctx->rs_state);
+   pipe->bind_fs_state(pipe, blitter_get_fs_col(ctx, 0));
+   pipe->bind_vs_state(pipe, ctx->vs_col);
+   pipe->bind_vertex_elements_state(pipe, ctx->velem_state);
+
+   /* set a framebuffer state */
+   fb_state.width = dstsurf->width;
+   fb_state.height = dstsurf->height;
+   fb_state.nr_cbufs = 0;
+   fb_state.cbufs[0] = 0;
+   fb_state.zsbuf = dstsurf;
+   pipe->set_framebuffer_state(pipe, &fb_state);
+
+   blitter_set_dst_dimensions(ctx, dstsurf->width, dstsurf->height);
+   blitter->draw_rectangle(blitter, 0, 0, dstsurf->width, dstsurf->height, 0,
+                           UTIL_BLITTER_ATTRIB_NONE, NULL);
+   blitter_restore_CSOs(ctx);
+}
diff --git a/src/gallium/auxiliary/util/u_blitter.h b/src/gallium/auxiliary/util/u_blitter.h
index ba3f92eca85..f316587dea0 100644
--- a/src/gallium/auxiliary/util/u_blitter.h
+++ b/src/gallium/auxiliary/util/u_blitter.h
@@ -200,6 +200,8 @@ void util_blitter_clear_depth_stencil(struct blitter_context *blitter,
                                       unsigned dstx, unsigned dsty,
                                       unsigned width, unsigned height);
 
+void util_blitter_flush_depth_stencil(struct blitter_context *blitter,
+                                      struct pipe_surface *dstsurf);
 /* The functions below should be used to save currently bound constant state
  * objects inside a driver. The objects are automatically restored at the end
  * of the util_blitter_{clear, copy_region, fill_region} functions and then
diff --git a/src/gallium/auxiliary/util/u_cpu_detect.c b/src/gallium/auxiliary/util/u_cpu_detect.c
index a08241971ca..50563513072 100644
--- a/src/gallium/auxiliary/util/u_cpu_detect.c
+++ b/src/gallium/auxiliary/util/u_cpu_detect.c
@@ -38,7 +38,7 @@
 #include "u_cpu_detect.h"
 
 #if defined(PIPE_ARCH_PPC)
-#if defined(PIPE_OS_DARWIN)
+#if defined(PIPE_OS_APPLE)
 #include <sys/sysctl.h>
 #else
 #include <signal.h>
@@ -73,9 +73,15 @@
 #endif
 
 
+DEBUG_GET_ONCE_BOOL_OPTION(dump_cpu, "GALLIUM_DUMP_CPU", FALSE)
+
+
 struct util_cpu_caps util_cpu_caps;
 
+#if defined(PIPE_ARCH_X86) || defined(PIPE_ARCH_X86_64)
 static int has_cpuid(void);
+#endif
+
 
 #if defined(PIPE_ARCH_X86)
 
@@ -132,7 +138,7 @@ win32_sig_handler_sse(EXCEPTION_POINTERS* ep)
 #endif /* PIPE_ARCH_X86 */
 
 
-#if defined(PIPE_ARCH_PPC) && !defined(PIPE_OS_DARWIN)
+#if defined(PIPE_ARCH_PPC) && !defined(PIPE_OS_APPLE)
 static jmp_buf  __lv_powerpc_jmpbuf;
 static volatile sig_atomic_t __lv_powerpc_canjump = 0;
 
@@ -153,7 +159,7 @@ sigill_handler(int sig)
 static void
 check_os_altivec_support(void)
 {
-#if defined(PIPE_OS_DARWIN)
+#if defined(PIPE_OS_APPLE)
    int sels[2] = {CTL_HW, HW_VECTORUNIT};
    int has_vu = 0;
    int len = sizeof (has_vu);
@@ -166,8 +172,8 @@ check_os_altivec_support(void)
          util_cpu_caps.has_altivec = 1;
       }
    }
-#else /* !PIPE_OS_DARWIN */
-   /* no Darwin, do it the brute-force way */
+#else /* !PIPE_OS_APPLE */
+   /* not on Apple/Darwin, do it the brute-force way */
    /* this is borrowed from the libmpeg2 library */
    signal(SIGILL, sigill_handler);
    if (setjmp(__lv_powerpc_jmpbuf)) {
@@ -184,7 +190,7 @@ check_os_altivec_support(void)
       signal(SIGILL, SIG_DFL);
       util_cpu_caps.has_altivec = 1;
    }
-#endif /* PIPE_OS_DARWIN */
+#endif /* !PIPE_OS_APPLE */
 }
 #endif /* PIPE_ARCH_PPC */
 
@@ -385,23 +391,6 @@ util_cpu_detect(void)
 
    memset(&util_cpu_caps, 0, sizeof util_cpu_caps);
 
-   /* Check for arch type */
-#if defined(PIPE_ARCH_MIPS)
-   util_cpu_caps.arch = UTIL_CPU_ARCH_MIPS;
-#elif defined(PIPE_ARCH_ALPHA)
-   util_cpu_caps.arch = UTIL_CPU_ARCH_ALPHA;
-#elif defined(PIPE_ARCH_SPARC)
-   util_cpu_caps.arch = UTIL_CPU_ARCH_SPARC;
-#elif defined(PIPE_ARCH_X86) || defined(PIPE_ARCH_X86_64)
-   util_cpu_caps.arch = UTIL_CPU_ARCH_X86;
-   util_cpu_caps.little_endian = 1;
-#elif defined(PIPE_ARCH_PPC)
-   util_cpu_caps.arch = UTIL_CPU_ARCH_POWERPC;
-   util_cpu_caps.little_endian = 0;
-#else
-   util_cpu_caps.arch = UTIL_CPU_ARCH_UNKNOWN;
-#endif
-
    /* Count the number of CPUs in system */
 #if defined(PIPE_OS_WINDOWS)
    {
@@ -497,23 +486,24 @@ util_cpu_detect(void)
 #endif /* PIPE_ARCH_PPC */
 
 #ifdef DEBUG
-   debug_printf("util_cpu_caps.arch = %i\n", util_cpu_caps.arch);
-   debug_printf("util_cpu_caps.nr_cpus = %u\n", util_cpu_caps.nr_cpus);
-
-   debug_printf("util_cpu_caps.x86_cpu_type = %u\n", util_cpu_caps.x86_cpu_type);
-   debug_printf("util_cpu_caps.cacheline = %u\n", util_cpu_caps.cacheline);
-
-   debug_printf("util_cpu_caps.has_tsc = %u\n", util_cpu_caps.has_tsc);
-   debug_printf("util_cpu_caps.has_mmx = %u\n", util_cpu_caps.has_mmx);
-   debug_printf("util_cpu_caps.has_mmx2 = %u\n", util_cpu_caps.has_mmx2);
-   debug_printf("util_cpu_caps.has_sse = %u\n", util_cpu_caps.has_sse);
-   debug_printf("util_cpu_caps.has_sse2 = %u\n", util_cpu_caps.has_sse2);
-   debug_printf("util_cpu_caps.has_sse3 = %u\n", util_cpu_caps.has_sse3);
-   debug_printf("util_cpu_caps.has_ssse3 = %u\n", util_cpu_caps.has_ssse3);
-   debug_printf("util_cpu_caps.has_sse4_1 = %u\n", util_cpu_caps.has_sse4_1);
-   debug_printf("util_cpu_caps.has_3dnow = %u\n", util_cpu_caps.has_3dnow);
-   debug_printf("util_cpu_caps.has_3dnow_ext = %u\n", util_cpu_caps.has_3dnow_ext);
-   debug_printf("util_cpu_caps.has_altivec = %u\n", util_cpu_caps.has_altivec);
+   if (debug_get_option_dump_cpu()) {
+      debug_printf("util_cpu_caps.nr_cpus = %u\n", util_cpu_caps.nr_cpus);
+
+      debug_printf("util_cpu_caps.x86_cpu_type = %u\n", util_cpu_caps.x86_cpu_type);
+      debug_printf("util_cpu_caps.cacheline = %u\n", util_cpu_caps.cacheline);
+
+      debug_printf("util_cpu_caps.has_tsc = %u\n", util_cpu_caps.has_tsc);
+      debug_printf("util_cpu_caps.has_mmx = %u\n", util_cpu_caps.has_mmx);
+      debug_printf("util_cpu_caps.has_mmx2 = %u\n", util_cpu_caps.has_mmx2);
+      debug_printf("util_cpu_caps.has_sse = %u\n", util_cpu_caps.has_sse);
+      debug_printf("util_cpu_caps.has_sse2 = %u\n", util_cpu_caps.has_sse2);
+      debug_printf("util_cpu_caps.has_sse3 = %u\n", util_cpu_caps.has_sse3);
+      debug_printf("util_cpu_caps.has_ssse3 = %u\n", util_cpu_caps.has_ssse3);
+      debug_printf("util_cpu_caps.has_sse4_1 = %u\n", util_cpu_caps.has_sse4_1);
+      debug_printf("util_cpu_caps.has_3dnow = %u\n", util_cpu_caps.has_3dnow);
+      debug_printf("util_cpu_caps.has_3dnow_ext = %u\n", util_cpu_caps.has_3dnow_ext);
+      debug_printf("util_cpu_caps.has_altivec = %u\n", util_cpu_caps.has_altivec);
+   }
 #endif
 
    util_cpu_detect_initialized = TRUE;
diff --git a/src/gallium/auxiliary/util/u_cpu_detect.h b/src/gallium/auxiliary/util/u_cpu_detect.h
index 4b3dc39c342..f3bef0993c7 100644
--- a/src/gallium/auxiliary/util/u_cpu_detect.h
+++ b/src/gallium/auxiliary/util/u_cpu_detect.h
@@ -36,26 +36,15 @@
 #define _UTIL_CPU_DETECT_H
 
 #include "pipe/p_compiler.h"
-
-enum util_cpu_arch {
-   UTIL_CPU_ARCH_UNKNOWN = 0,
-   UTIL_CPU_ARCH_MIPS,
-   UTIL_CPU_ARCH_ALPHA,
-   UTIL_CPU_ARCH_SPARC,
-   UTIL_CPU_ARCH_X86,
-   UTIL_CPU_ARCH_POWERPC
-};
+#include "pipe/p_config.h"
 
 struct util_cpu_caps {
-   enum util_cpu_arch arch;
    unsigned nr_cpus;
 
    /* Feature flags */
    int x86_cpu_type;
    unsigned cacheline;
 
-   unsigned little_endian:1;
-
    unsigned has_tsc:1;
    unsigned has_mmx:1;
    unsigned has_mmx2:1;
diff --git a/src/gallium/auxiliary/util/u_debug.c b/src/gallium/auxiliary/util/u_debug.c
index ad162558bc1..504e6d2a18f 100644
--- a/src/gallium/auxiliary/util/u_debug.c
+++ b/src/gallium/auxiliary/util/u_debug.c
@@ -88,7 +88,7 @@ debug_get_option_should_print(void)
     * but its cool since we set first to false
     */
    first = FALSE;
-   value = debug_get_bool_option("GALLIUM_PRINT_OPTIONS", TRUE);
+   value = debug_get_bool_option("GALLIUM_PRINT_OPTIONS", FALSE);
    /* XXX should we print this option? Currently it wont */
    return value;
 }
diff --git a/src/gallium/auxiliary/util/u_draw.h b/src/gallium/auxiliary/util/u_draw.h
new file mode 100644
index 00000000000..2a91ea0f9ae
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_draw.h
@@ -0,0 +1,138 @@
+/**************************************************************************
+ *
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef U_DRAW_H
+#define U_DRAW_H
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_context.h"
+
+
+static INLINE void
+util_draw_init_info(struct pipe_draw_info *info)
+{
+   memset(info, 0, sizeof(*info));
+   info->instance_count = 1;
+   info->max_index = 0xffffffff;
+}
+
+
+static INLINE void
+util_draw_arrays(struct pipe_context *pipe, uint mode, uint start, uint count)
+{
+   struct pipe_draw_info info;
+
+   util_draw_init_info(&info);
+   info.mode = mode;
+   info.start = start;
+   info.count = count;
+   info.min_index = start;
+   info.max_index = start + count - 1;
+
+   pipe->draw_vbo(pipe, &info);
+}
+
+static INLINE void
+util_draw_elements(struct pipe_context *pipe, int index_bias,
+                   uint mode, uint start, uint count)
+{
+   struct pipe_draw_info info;
+
+   util_draw_init_info(&info);
+   info.indexed = TRUE;
+   info.mode = mode;
+   info.start = start;
+   info.count = count;
+   info.index_bias = index_bias;
+
+   pipe->draw_vbo(pipe, &info);
+}
+
+static INLINE void
+util_draw_arrays_instanced(struct pipe_context *pipe,
+                           uint mode, uint start, uint count,
+                           uint start_instance,
+                           uint instance_count)
+{
+   struct pipe_draw_info info;
+
+   util_draw_init_info(&info);
+   info.mode = mode;
+   info.start = start;
+   info.count = count;
+   info.start_instance = start_instance;
+   info.instance_count = instance_count;
+   info.min_index = start;
+   info.max_index = start + count - 1;
+
+   pipe->draw_vbo(pipe, &info);
+}
+
+static INLINE void
+util_draw_elements_instanced(struct pipe_context *pipe,
+                             int index_bias,
+                             uint mode, uint start, uint count,
+                             uint start_instance,
+                             uint instance_count)
+{
+   struct pipe_draw_info info;
+
+   util_draw_init_info(&info);
+   info.indexed = TRUE;
+   info.mode = mode;
+   info.start = start;
+   info.count = count;
+   info.index_bias = index_bias;
+   info.start_instance = start_instance;
+   info.instance_count = instance_count;
+
+   pipe->draw_vbo(pipe, &info);
+}
+
+static INLINE void
+util_draw_range_elements(struct pipe_context *pipe,
+                         int index_bias,
+                         uint min_index,
+                         uint max_index,
+                         uint mode, uint start, uint count)
+{
+   struct pipe_draw_info info;
+
+   util_draw_init_info(&info);
+   info.indexed = TRUE;
+   info.mode = mode;
+   info.start = start;
+   info.count = count;
+   info.index_bias = index_bias;
+   info.min_index = min_index;
+   info.max_index = max_index;
+
+   pipe->draw_vbo(pipe, &info);
+}
+
+#endif
diff --git a/src/gallium/auxiliary/util/u_draw_quad.c b/src/gallium/auxiliary/util/u_draw_quad.c
index b37b48b5aef..0b6dc5880f3 100644
--- a/src/gallium/auxiliary/util/u_draw_quad.c
+++ b/src/gallium/auxiliary/util/u_draw_quad.c
@@ -60,7 +60,7 @@ util_draw_vertex_buffer(struct pipe_context *pipe,
    /* note: vertex elements already set by caller */
 
    /* draw */
-   pipe->draw_arrays(pipe, prim_type, 0, num_verts);
+   util_draw_arrays(pipe, prim_type, 0, num_verts);
 }
 
 
diff --git a/src/gallium/auxiliary/util/u_draw_quad.h b/src/gallium/auxiliary/util/u_draw_quad.h
index 42eb1844289..52994fe05c3 100644
--- a/src/gallium/auxiliary/util/u_draw_quad.h
+++ b/src/gallium/auxiliary/util/u_draw_quad.h
@@ -29,12 +29,18 @@
 #define U_DRAWQUAD_H
 
 
+#include "pipe/p_compiler.h"
+#include "pipe/p_context.h"
+
+
 #ifdef __cplusplus
 extern "C" {
 #endif
 
 struct pipe_resource;
 
+#include "util/u_draw.h"
+
 extern void 
 util_draw_vertex_buffer(struct pipe_context *pipe,
                         struct pipe_resource *vbuf, uint offset,
diff --git a/src/gallium/auxiliary/util/u_format.h b/src/gallium/auxiliary/util/u_format.h
index 38254b1096d..8e786a390a0 100644
--- a/src/gallium/auxiliary/util/u_format.h
+++ b/src/gallium/auxiliary/util/u_format.h
@@ -631,6 +631,44 @@ util_format_has_alpha(enum pipe_format format)
 }
 
 /**
+ * Return the matching SRGB format, or PIPE_FORMAT_NONE if none.
+ */
+static INLINE enum pipe_format
+util_format_srgb(enum pipe_format format)
+{
+   switch (format) {
+   case PIPE_FORMAT_L8_UNORM:
+      return PIPE_FORMAT_L8_SRGB;
+   case PIPE_FORMAT_L8A8_UNORM:
+      return PIPE_FORMAT_L8A8_SRGB;
+   case PIPE_FORMAT_R8G8B8_UNORM:
+      return PIPE_FORMAT_R8G8B8_SRGB;
+   case PIPE_FORMAT_A8B8G8R8_UNORM:
+      return PIPE_FORMAT_A8B8G8R8_SRGB;
+   case PIPE_FORMAT_X8B8G8R8_UNORM:
+      return PIPE_FORMAT_X8B8G8R8_SRGB;
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+      return PIPE_FORMAT_B8G8R8A8_SRGB;
+   case PIPE_FORMAT_B8G8R8X8_UNORM:
+      return PIPE_FORMAT_B8G8R8X8_SRGB;
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+      return PIPE_FORMAT_A8R8G8B8_SRGB;
+   case PIPE_FORMAT_X8R8G8B8_UNORM:
+      return PIPE_FORMAT_X8R8G8B8_SRGB;
+   case PIPE_FORMAT_DXT1_RGB:
+      return PIPE_FORMAT_DXT1_SRGB;
+   case PIPE_FORMAT_DXT1_RGBA:
+      return PIPE_FORMAT_DXT1_SRGBA;
+   case PIPE_FORMAT_DXT3_RGBA:
+      return PIPE_FORMAT_DXT3_SRGBA;
+   case PIPE_FORMAT_DXT5_RGBA:
+      return PIPE_FORMAT_DXT5_SRGBA;
+   default:
+      return PIPE_FORMAT_NONE;
+   }
+}
+
+/**
  * Return the number of components stored.
  * Formats with block size != 1x1 will always have 1 component (the block).
  */
diff --git a/src/gallium/auxiliary/util/u_format_other.c b/src/gallium/auxiliary/util/u_format_other.c
index 723fa8c3bf9..fa42ec37138 100644
--- a/src/gallium/auxiliary/util/u_format_other.c
+++ b/src/gallium/auxiliary/util/u_format_other.c
@@ -121,6 +121,15 @@ util_format_r1_unorm_pack_rgba_8unorm(uint8_t *dst_row, unsigned dst_stride,
  * A.k.a. D3DFMT_CxV8U8
  */
 
+static uint8_t
+r8g8bx_derive(int16_t r, int16_t g)
+{
+   /* Derive blue from red and green components.
+    * Apparently, we must always use integers to perform calculations,
+    * otherwise the results won't match D3D's CxV8U8 definition.
+    */
+   return (uint8_t)sqrtf(0x7f * 0x7f - r * r - g * g) * 0xff / 0x7f;
+}
 
 void
 util_format_r8g8bx_snorm_unpack_rgba_float(float *dst_row, unsigned dst_stride,
@@ -145,7 +154,7 @@ util_format_r8g8bx_snorm_unpack_rgba_float(float *dst_row, unsigned dst_stride,
 
          dst[0] = (float)(r * (1.0f/0x7f)); /* r */
          dst[1] = (float)(g * (1.0f/0x7f)); /* g */
-         dst[2] = sqrtf(1.0f - dst[0] * dst[0] - dst[1] * dst[1]); /* b */
+         dst[2] = r8g8bx_derive(r, g) * (1.0f/0xff); /* b */
          dst[3] = 1.0f; /* a */
          dst += 4;
       }
@@ -177,7 +186,7 @@ util_format_r8g8bx_snorm_unpack_rgba_8unorm(uint8_t *dst_row, unsigned dst_strid
 
          dst[0] = (uint8_t)(((uint16_t)MAX2(r, 0)) * 0xff / 0x7f); /* r */
          dst[1] = (uint8_t)(((uint16_t)MAX2(g, 0)) * 0xff / 0x7f); /* g */
-         dst[2] = (uint8_t)sqrtf(0x7f*0x7f - r * r - g * g) * 0xff / 0x7f; /* b */
+         dst[2] = r8g8bx_derive(r, g); /* b */
          dst[3] = 255; /* a */
          dst += 4;
       }
@@ -262,6 +271,6 @@ util_format_r8g8bx_snorm_fetch_rgba_float(float *dst, const uint8_t *src,
 
    dst[0] = r * (1.0f/0x7f); /* r */
    dst[1] = g * (1.0f/0x7f); /* g */
-   dst[2] = sqrtf(1.0f - dst[0] * dst[0] - dst[1] * dst[1]); /* b */
+   dst[2] = r8g8bx_derive(r, g) * (1.0f/0xff); /* b */
    dst[3] = 1.0f; /* a */
 }
diff --git a/src/gallium/auxiliary/util/u_framebuffer.c b/src/gallium/auxiliary/util/u_framebuffer.c
index 768ae9ceb5d..7803ec6a8b5 100644
--- a/src/gallium/auxiliary/util/u_framebuffer.c
+++ b/src/gallium/auxiliary/util/u_framebuffer.c
@@ -85,9 +85,11 @@ util_copy_framebuffer_state(struct pipe_framebuffer_state *dst,
    dst->width = src->width;
    dst->height = src->height;
 
-   for (i = 0; i < Elements(src->cbufs); i++) {
+   for (i = 0; i < src->nr_cbufs; i++)
       pipe_surface_reference(&dst->cbufs[i], src->cbufs[i]);
-   }
+
+   for (i = src->nr_cbufs; i < dst->nr_cbufs; i++)
+      pipe_surface_reference(&dst->cbufs[i], NULL);
 
    dst->nr_cbufs = src->nr_cbufs;
 
diff --git a/src/gallium/auxiliary/util/u_mempool.c b/src/gallium/auxiliary/util/u_mempool.c
index 84e2a34acc6..1f336b39a1a 100644
--- a/src/gallium/auxiliary/util/u_mempool.c
+++ b/src/gallium/auxiliary/util/u_mempool.c
@@ -126,7 +126,6 @@ void util_mempool_set_thread_safety(struct util_mempool *pool,
    pool->threading = threading;
 
    if (threading) {
-      pipe_mutex_init(pool->mutex);
       pool->malloc = util_mempool_malloc_mt;
       pool->free = util_mempool_free_mt;
    } else {
@@ -152,6 +151,8 @@ void util_mempool_create(struct util_mempool *pool,
 
    make_empty_list(&pool->list);
 
+   pipe_mutex_init(pool->mutex);
+
    util_mempool_set_thread_safety(pool, threading);
 }
 
@@ -164,6 +165,5 @@ void util_mempool_destroy(struct util_mempool *pool)
       FREE(page);
    }
 
-   if (pool->threading)
-      pipe_mutex_destroy(pool->mutex);
+   pipe_mutex_destroy(pool->mutex);
 }
diff --git a/src/gallium/auxiliary/util/u_network.c b/src/gallium/auxiliary/util/u_network.c
index 87ee0e47685..77f2c5fc7de 100644
--- a/src/gallium/auxiliary/util/u_network.c
+++ b/src/gallium/auxiliary/util/u_network.c
@@ -6,7 +6,7 @@
 #if defined(PIPE_SUBSYSTEM_WINDOWS_USER)
 #  include <winsock2.h>
 #  include <windows.h>
-#elif defined(PIPE_OS_LINUX) || defined(PIPE_OS_HAIKU) || defined(PIPE_OS_APPLE)
+#elif defined(PIPE_OS_LINUX) || defined(PIPE_OS_HAIKU) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_CYGWIN)
 #  include <sys/socket.h>
 #  include <netinet/in.h>
 #  include <unistd.h>
diff --git a/src/gallium/auxiliary/util/u_pack_color.h b/src/gallium/auxiliary/util/u_pack_color.h
index 3ebef9fb749..5f113f742b1 100644
--- a/src/gallium/auxiliary/util/u_pack_color.h
+++ b/src/gallium/auxiliary/util/u_pack_color.h
@@ -425,6 +425,53 @@ util_pack_color(const float rgba[4], enum pipe_format format, union util_color *
    }
 }
  
+/* Integer versions of util_pack_z and util_pack_z_stencil - useful for
+ * constructing clear masks.
+ */
+static INLINE uint
+util_pack_uint_z(enum pipe_format format, unsigned z)
+{
+   switch (format) {
+   case PIPE_FORMAT_Z16_UNORM:
+      return z & 0xffff;
+   case PIPE_FORMAT_Z32_UNORM:
+   case PIPE_FORMAT_Z32_FLOAT:
+      return z;
+   case PIPE_FORMAT_Z24_UNORM_S8_USCALED:
+   case PIPE_FORMAT_Z24X8_UNORM:
+      return z & 0xffffff;
+   case PIPE_FORMAT_S8_USCALED_Z24_UNORM:
+   case PIPE_FORMAT_X8Z24_UNORM:
+      return (z & 0xffffff) << 8;
+   case PIPE_FORMAT_S8_USCALED:
+      return 0;
+   default:
+      debug_print_format("gallium: unhandled format in util_pack_z()", format);
+      assert(0);
+      return 0;
+   }
+}
+
+static INLINE uint
+util_pack_uint_z_stencil(enum pipe_format format, double z, uint s)
+{
+   unsigned packed = util_pack_uint_z(format, z);
+
+   s &= 0xff;
+
+   switch (format) {
+   case PIPE_FORMAT_Z24_UNORM_S8_USCALED:
+      return packed | (s << 24);
+   case PIPE_FORMAT_S8_USCALED_Z24_UNORM:
+      return packed | s;
+   case PIPE_FORMAT_S8_USCALED:
+      return packed | s;
+   default:
+      return packed;
+   }
+}
+
+
 
 /**
  * Note: it's assumed that z is in [0,1]
diff --git a/src/gallium/auxiliary/util/u_prim.h b/src/gallium/auxiliary/util/u_prim.h
index 606b9b5c6b9..3c851f73401 100644
--- a/src/gallium/auxiliary/util/u_prim.h
+++ b/src/gallium/auxiliary/util/u_prim.h
@@ -108,6 +108,20 @@ static INLINE boolean u_trim_pipe_prim( unsigned pipe_prim, unsigned *nr )
       ok = (*nr >= 4);
       *nr -= (*nr % 2);
       break;
+   case PIPE_PRIM_LINES_ADJACENCY:
+      ok = (*nr >= 4);
+      *nr -= (*nr % 4);
+      break;
+   case PIPE_PRIM_LINE_STRIP_ADJACENCY:
+      ok = (*nr >= 4);
+      break;
+   case PIPE_PRIM_TRIANGLES_ADJACENCY:
+      ok = (*nr >= 6);
+      *nr -= (*nr % 5);
+      break;
+   case PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY:
+      ok = (*nr >= 4);
+      break;
    default:
       ok = 0;
       break;
diff --git a/src/gallium/auxiliary/util/u_split_prim.h b/src/gallium/auxiliary/util/u_split_prim.h
new file mode 100644
index 00000000000..206e1ec3118
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_split_prim.h
@@ -0,0 +1,105 @@
+/* Originally written by Ben Skeggs for the nv50 driver*/
+#include <pipe/p_defines.h>
+
+struct util_split_prim {
+   void *priv;
+   void (*emit)(void *priv, unsigned start, unsigned count);
+   void (*edge)(void *priv, boolean enabled);
+
+   unsigned mode;
+   unsigned start;
+   unsigned p_start;
+   unsigned p_end;
+
+   uint repeat_first:1;
+   uint close_first:1;
+   uint edgeflag_off:1;
+};
+
+static INLINE void
+util_split_prim_init(struct util_split_prim *s,
+                  unsigned mode, unsigned start, unsigned count)
+{
+   if (mode == PIPE_PRIM_LINE_LOOP) {
+      s->mode = PIPE_PRIM_LINE_STRIP;
+      s->close_first = 1;
+   } else {
+      s->mode = mode;
+      s->close_first = 0;
+   }
+   s->start = start;
+   s->p_start = start;
+   s->p_end = start + count;
+   s->edgeflag_off = 0;
+   s->repeat_first = 0;
+}
+
+static INLINE boolean
+util_split_prim_next(struct util_split_prim *s, unsigned max_verts)
+{
+   int repeat = 0;
+
+   if (s->repeat_first) {
+      s->emit(s->priv, s->start, 1);
+      max_verts--;
+      if (s->edgeflag_off) {
+         s->edge(s->priv, TRUE);
+         s->edgeflag_off = FALSE;
+      }
+   }
+
+   if (s->p_start + s->close_first + max_verts >= s->p_end) {
+      s->emit(s->priv, s->p_start, s->p_end - s->p_start);
+      if (s->close_first)
+         s->emit(s->priv, s->start, 1);
+      return TRUE;
+   }
+
+   switch (s->mode) {
+   case PIPE_PRIM_LINES:
+      max_verts &= ~1;
+      break;
+   case PIPE_PRIM_LINE_STRIP:
+      repeat = 1;
+      break;
+   case PIPE_PRIM_POLYGON:
+      max_verts--;
+      s->emit(s->priv, s->p_start, max_verts);
+      s->edge(s->priv, FALSE);
+      s->emit(s->priv, s->p_start + max_verts, 1);
+      s->p_start += max_verts;
+      s->repeat_first = TRUE;
+      s->edgeflag_off = TRUE;
+      return FALSE;
+   case PIPE_PRIM_TRIANGLES:
+      max_verts = max_verts - (max_verts % 3);
+      break;
+   case PIPE_PRIM_TRIANGLE_STRIP:
+      /* to ensure winding stays correct, always split
+       * on an even number of generated triangles
+       */
+      max_verts = max_verts & ~1;
+      repeat = 2;
+      break;
+   case PIPE_PRIM_TRIANGLE_FAN:
+      s->repeat_first = TRUE;
+      repeat = 1;
+      break;
+   case PIPE_PRIM_QUADS:
+      max_verts &= ~3;
+      break;
+   case PIPE_PRIM_QUAD_STRIP:
+      max_verts &= ~1;
+      repeat = 2;
+      break;
+   case PIPE_PRIM_POINTS:
+      break;
+   default:
+      /* TODO: implement adjacency primitives */
+      assert(0);
+   }
+
+   s->emit (s->priv, s->p_start, max_verts);
+   s->p_start += (max_verts - repeat);
+   return FALSE;
+}
diff --git a/src/gallium/auxiliary/util/u_sse.h b/src/gallium/auxiliary/util/u_sse.h
index e2a8491e62c..87959ab0aab 100644
--- a/src/gallium/auxiliary/util/u_sse.h
+++ b/src/gallium/auxiliary/util/u_sse.h
@@ -41,7 +41,6 @@
 
 #if defined(PIPE_ARCH_SSE)
 
-#include <xmmintrin.h>
 #include <emmintrin.h>
 
 
@@ -72,6 +71,35 @@ _mm_castps_si128(__m128 a)
 
 #endif /* defined(_MSC_VER) && _MSC_VER < 1500 */
 
+
+#if defined(PIPE_ARCH_SSSE3)
+
+#include <tmmintrin.h>
+
+#else /* !PIPE_ARCH_SSSE3 */
+
+#include <emmintrin.h>
+
+/**
+ * Describe _mm_shuffle_epi8() with gcc extended inline assembly, for cases
+ * where -mssse3 is not supported/enabled.
+ *
+ * MSVC will never get in here as its intrinsics support do not rely on
+ * compiler command line options.
+ */
+static __inline __m128i __attribute__((__gnu_inline__, __always_inline__, __artificial__))
+_mm_shuffle_epi8(__m128i a, __m128i mask)
+{
+    __m128i result;
+    __asm__("pshufb %1, %0"
+            : "=x" (result)
+            : "xm" (mask), "0" (a));
+    return result;
+}
+
+#endif /* !PIPE_ARCH_SSSE3 */
+
+
 #endif /* PIPE_ARCH_X86 || PIPE_ARCH_X86_64 */
 
 #endif /* U_SSE_H_ */
diff --git a/src/gallium/auxiliary/util/u_staging.c b/src/gallium/auxiliary/util/u_staging.c
new file mode 100644
index 00000000000..607c31f5ee7
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_staging.c
@@ -0,0 +1,95 @@
+#include "util/u_staging.h"
+#include "pipe/p_context.h"
+#include "util/u_memory.h"
+#include "util/u_inlines.h"
+
+static void
+util_staging_resource_template(struct pipe_resource *pt, unsigned width, unsigned height, unsigned depth, struct pipe_resource *template)
+{
+   memset(template, 0, sizeof(struct pipe_resource));
+   if(pt->target != PIPE_BUFFER && depth <= 1)
+      template->target = PIPE_TEXTURE_2D;
+   else
+      template->target = pt->target;
+   template->format = pt->format;
+   template->width0 = width;
+   template->height0 = height;
+   template->depth0 = depth;
+   template->last_level = 0;
+   template->nr_samples = pt->nr_samples;
+   template->bind = 0;
+   template->usage = PIPE_USAGE_STAGING;
+   template->flags = 0;
+}
+
+struct util_staging_transfer *
+util_staging_transfer_new(struct pipe_context *pipe,
+           struct pipe_resource *pt,
+           struct pipe_subresource sr,
+           unsigned usage,
+           const struct pipe_box *box,
+           bool direct)
+{
+   struct pipe_screen *pscreen = pipe->screen;
+   struct util_staging_transfer *tx;
+   struct pipe_resource staging_resource_template;
+
+   tx = CALLOC_STRUCT(util_staging_transfer);
+   if (!tx)
+      return NULL;
+
+   pipe_resource_reference(&tx->base.resource, pt);
+   tx->base.sr = sr;
+   tx->base.usage = usage;
+   tx->base.box = *box;
+
+   if (direct)
+   {
+      tx->staging_resource = pt;
+      return tx;
+   }
+
+   util_staging_resource_template(pt, box->width, box->height, box->depth, &staging_resource_template);
+   tx->staging_resource = pscreen->resource_create(pscreen, &staging_resource_template);
+   if (!tx->staging_resource)
+   {
+      pipe_resource_reference(&tx->base.resource, NULL);
+      FREE(tx);
+      return NULL;
+   }
+
+   if (usage & PIPE_TRANSFER_READ)
+   {
+      struct pipe_subresource dstsr;
+      unsigned zi;
+      dstsr.face = 0;
+      dstsr.level = 0;
+      for(zi = 0; zi < box->depth; ++zi)
+         pipe->resource_copy_region(pipe, tx->staging_resource, dstsr, 0, 0, 0, tx->base.resource, sr, box->x, box->y, box->z + zi, box->width, box->height);
+   }
+
+   return tx;
+}
+
+void
+util_staging_transfer_destroy(struct pipe_context *pipe, struct pipe_transfer *ptx)
+{
+   struct util_staging_transfer *tx = (struct util_staging_transfer *)ptx;
+
+   if (tx->staging_resource != tx->base.resource)
+   {
+      if(tx->base.usage & PIPE_TRANSFER_WRITE) {
+         struct pipe_subresource srcsr;
+         unsigned zi;
+         srcsr.face = 0;
+         srcsr.level = 0;
+         for(zi = 0; zi < tx->base.box.depth; ++zi)
+            pipe->resource_copy_region(pipe, tx->base.resource, tx->base.sr, tx->base.box.x, tx->base.box.y, tx->base.box.z + zi, tx->staging_resource, srcsr, 0, 0, 0, tx->base.box.width, tx->base.box.height);
+      }
+
+      pipe_resource_reference(&tx->staging_resource, NULL);
+   }
+
+   pipe_resource_reference(&ptx->resource, NULL);
+   FREE(ptx);
+}
diff --git a/src/gallium/auxiliary/util/u_staging.h b/src/gallium/auxiliary/util/u_staging.h
new file mode 100644
index 00000000000..602faa2971d
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_staging.h
@@ -0,0 +1,37 @@
+/* Direct3D 10/11 has no concept of transfers. Applications instead
+ * create resources with a STAGING or DYNAMIC usage, copy between them
+ * and the real resource and use Map to map the STAGING/DYNAMIC resource.
+ *
+ * This util module allows to implement Gallium drivers as a Direct3D
+ * driver would be implemented: transfers allocate a resource with
+ * PIPE_USAGE_STAGING, and copy the data between it and the real resource
+ * with resource_copy_region.
+ */
+
+#ifndef U_STAGING_H
+#define U_STAGING_H
+
+#include "pipe/p_state.h"
+
+struct util_staging_transfer {
+   struct pipe_transfer base;
+
+   /* if direct, same as base.resource, otherwise the temporary staging resource */
+   struct pipe_resource *staging_resource;
+};
+
+/* user must be stride, slice_stride and offset */
+/* pt->usage == PIPE_USAGE_DYNAMIC should be a good value to pass for direct */
+/* staging resource is currently created with PIPE_USAGE_DYNAMIC */
+struct util_staging_transfer *
+util_staging_transfer_new(struct pipe_context *pipe,
+           struct pipe_resource *pt,
+           struct pipe_subresource sr,
+           unsigned usage,
+           const struct pipe_box *box,
+           bool direct);
+
+void
+util_staging_transfer_destroy(struct pipe_context *pipe, struct pipe_transfer *ptx);
+
+#endif
diff --git a/src/gallium/auxiliary/util/u_surfaces.c b/src/gallium/auxiliary/util/u_surfaces.c
index b5d21570d57..7733ad24d0d 100644
--- a/src/gallium/auxiliary/util/u_surfaces.c
+++ b/src/gallium/auxiliary/util/u_surfaces.c
@@ -3,40 +3,22 @@
 #include "util/u_inlines.h"
 #include "util/u_memory.h"
 
-/* TODO: ouch, util_hash_table should do these by default when passed a null function pointer
- * this indirect function call is quite bad
- */
-static unsigned
-hash(void *key)
-{
-   return (unsigned)(uintptr_t)key;
-}
-
-static int
-compare(void *key1, void *key2)
-{
-   return (unsigned)(uintptr_t)key1 - (unsigned)(uintptr_t)key2;
-}
-
 struct pipe_surface *
 util_surfaces_do_get(struct util_surfaces *us, unsigned surface_struct_size, struct pipe_screen *pscreen, struct pipe_resource *pt, unsigned face, unsigned level, unsigned zslice, unsigned flags)
 {
    struct pipe_surface *ps;
-   void *key = NULL;
 
    if(pt->target == PIPE_TEXTURE_3D || pt->target == PIPE_TEXTURE_CUBE)
-   {	/* or 2D array */
-      if(!us->u.table)
-	 us->u.table = util_hash_table_create(hash, compare);
-      key = (void *)(uintptr_t)(((zslice + face) << 8) | level);
-      /* TODO: ouch, should have a get-reference function...
-       * also, shouldn't allocate a two-pointer structure for each item... */
-      ps = util_hash_table_get(us->u.table, key);
+   {    /* or 2D array */
+      if(!us->u.hash)
+         us->u.hash = cso_hash_create();
+
+      ps = cso_hash_iter_data(cso_hash_find(us->u.hash, ((zslice + face) << 8) | level));
    }
    else
    {
       if(!us->u.array)
-	 us->u.array = CALLOC(pt->last_level + 1, sizeof(struct pipe_surface *));
+         us->u.array = CALLOC(pt->last_level + 1, sizeof(struct pipe_surface *));
       ps = us->u.array[level];
    }
 
@@ -54,7 +36,7 @@ util_surfaces_do_get(struct util_surfaces *us, unsigned surface_struct_size, str
    ps->offset = ~0;
 
    if(pt->target == PIPE_TEXTURE_3D || pt->target == PIPE_TEXTURE_CUBE)
-      util_hash_table_set(us->u.table, key, ps);
+      cso_hash_insert(us->u.hash, ((zslice + face) << 8) | level, ps);
    else
       us->u.array[level] = ps;
 
@@ -66,47 +48,44 @@ util_surfaces_do_detach(struct util_surfaces *us, struct pipe_surface *ps)
 {
    struct pipe_resource *pt = ps->texture;
    if(pt->target == PIPE_TEXTURE_3D || pt->target == PIPE_TEXTURE_CUBE)
-   {	/* or 2D array */
-      void* key = (void*)(uintptr_t)(((ps->zslice + ps->face) << 8) | ps->level);
-      util_hash_table_remove(us->u.table, key);
+   {    /* or 2D array */
+      cso_hash_erase(us->u.hash, cso_hash_find(us->u.hash, ((ps->zslice + ps->face) << 8) | ps->level));
    }
    else
       us->u.array[ps->level] = 0;
 }
 
-static enum pipe_error
-util_surfaces_destroy_callback(void *key, void *value, void *data)
-{
-   void (*destroy_surface) (struct pipe_surface * ps) = data;
-   destroy_surface((struct pipe_surface *)value);
-   return PIPE_OK;
-}
-
 void
 util_surfaces_destroy(struct util_surfaces *us, struct pipe_resource *pt, void (*destroy_surface) (struct pipe_surface *))
 {
    if(pt->target == PIPE_TEXTURE_3D || pt->target == PIPE_TEXTURE_CUBE)
-   {	/* or 2D array */
-      if(us->u.table)
+   {    /* or 2D array */
+      if(us->u.hash)
       {
-	 util_hash_table_foreach(us->u.table, util_surfaces_destroy_callback, destroy_surface);
-	 util_hash_table_destroy(us->u.table);
-	 us->u.table = NULL;
+         struct cso_hash_iter iter;
+         iter = cso_hash_first_node(us->u.hash);
+         while (!cso_hash_iter_is_null(iter)) {
+            destroy_surface(cso_hash_iter_data(iter));
+            iter = cso_hash_iter_next(iter);
+         }
+
+         cso_hash_delete(us->u.hash);
+         us->u.hash = NULL;
       }
    }
    else
    {
       if(us->u.array)
       {
-	 unsigned i;
-	 for(i = 0; i < pt->last_level; ++i)
-	 {
-	    struct pipe_surface *ps = us->u.array[i];
-	    if(ps)
-	       destroy_surface(ps);
-	 }
-	 FREE(us->u.array);
-	 us->u.array = NULL;
+         unsigned i;
+         for(i = 0; i <= pt->last_level; ++i)
+         {
+            struct pipe_surface *ps = us->u.array[i];
+            if(ps)
+               destroy_surface(ps);
+         }
+         FREE(us->u.array);
+         us->u.array = NULL;
       }
    }
 }
diff --git a/src/gallium/auxiliary/util/u_surfaces.h b/src/gallium/auxiliary/util/u_surfaces.h
index 0195bf5afba..af978c70579 100644
--- a/src/gallium/auxiliary/util/u_surfaces.h
+++ b/src/gallium/auxiliary/util/u_surfaces.h
@@ -4,15 +4,15 @@
 #include "pipe/p_compiler.h"
 #include "pipe/p_state.h"
 #include "util/u_atomic.h"
-
-struct util_hash_table;
+#include "cso_cache/cso_hash.h"
 
 struct util_surfaces
 {
    union
    {
-      struct util_hash_table *table;
+      struct cso_hash *hash;
       struct pipe_surface **array;
+      void* pv;
    } u;
 };
 
@@ -35,6 +35,18 @@ util_surfaces_get(struct util_surfaces *us, unsigned surface_struct_size, struct
    return util_surfaces_do_get(us, surface_struct_size, pscreen, pt, face, level, zslice, flags);
 }
 
+static INLINE struct pipe_surface *
+util_surfaces_peek(struct util_surfaces *us, struct pipe_resource *pt, unsigned face, unsigned level, unsigned zslice)
+{
+   if(!us->u.pv)
+      return 0;
+
+   if(unlikely(pt->target == PIPE_TEXTURE_3D || pt->target == PIPE_TEXTURE_CUBE))
+      return cso_hash_iter_data(cso_hash_find(us->u.hash, ((zslice + face) << 8) | level));
+   else
+      return us->u.array[level];
+}
+
 void util_surfaces_do_detach(struct util_surfaces *us, struct pipe_surface *ps);
 
 static INLINE void
diff --git a/src/gallium/docs/d3d11ddi.txt b/src/gallium/docs/d3d11ddi.txt
index 0954c2926df..f8155c828b1 100644
--- a/src/gallium/docs/d3d11ddi.txt
+++ b/src/gallium/docs/d3d11ddi.txt
@@ -66,9 +66,6 @@ Unordered access view: view supporting random read/write access (usually from co
 clear
 	+ Gallium supports clearing both render targets and depth/stencil with a single call
 
-draw_range_elements
-	+ Gallium supports indexed draw with explicit range
-
 fence_signalled
 fence_finish
 	+ D3D10/D3D11 don't appear to support explicit fencing; queries can often substitute though, and flushing is supported
@@ -271,31 +268,27 @@ Dispatch (D3D11 only)
 DispatchIndirect (D3D11 only)
 	- Gallium does not support compute shaders
 
-Draw -> draw_arrays
+Draw -> draw_vbo
 	! D3D11 sets primitive modes separately with IaSetTopology: it's not obvious which is better
 
 DrawAuto -> draw_auto
 
-DrawIndexed -> draw_elements
+DrawIndexed -> draw_vbo
 	! D3D11 sets primitive modes separately with IaSetTopology: it's not obvious which is better
-	* may want to add a separate set_index_buffer
-	- Gallium lacks base vertex for indexed draw calls
-	+ D3D11 lacks draw_range_elements functionality, which is required for OpenGL
+	+ D3D11 lacks explicit range, which is required for OpenGL
 
-DrawIndexedInstanced -> draw_elements_instanced
+DrawIndexedInstanced -> draw_vbo
 	! D3D11 sets primitive modes separately with IaSetTopology: it's not obvious which is better
-	* may want to add a separate set_index_buffer
-	- Gallium lacks base vertex for indexed draw calls
 
-DrawIndexedInstancedIndirect (D3D11 only) -> call draw_elements_instanced multiple times in software
-	# this allows to use an hardware buffer to specify the parameters for multiple draw_elements_instanced calls
+DrawIndexedInstancedIndirect (D3D11 only)
+	# this allows to use an hardware buffer to specify the parameters for multiple draw_vbo calls
 	- Gallium does not support draw call parameter buffers and indirect draw
 
-DrawInstanced -> draw_arrays_instanced
+DrawInstanced -> draw_vbo
 	! D3D11 sets primitive modes separately with IaSetTopology: it's not obvious which is better
 
-DrawInstancedIndirect (D3D11 only) -> call draw_arrays_instanced multiple times in software
-	# this allows to use an hardware buffer to specify the parameters for multiple draw_arrays_instanced calls
+DrawInstancedIndirect (D3D11 only)
+	# this allows to use an hardware buffer to specify the parameters for multiple draw_vbo calls
 	- Gallium does not support draw call parameter buffers and indirect draws
 
 DsSetConstantBuffers (D3D11 only)
@@ -332,10 +325,9 @@ HsSetShaderResources (D3D11 only)
 HsSetShaderWithIfaces (D3D11 only)
 	- Gallium does not support hull shaders
 
-IaSetIndexBuffer
-	! Gallium passes this to the draw_elements or draw_elements_instanced calls
+IaSetIndexBuffer -> set_index_buffer
 	+ Gallium supports 8-bit indices
-	! the D3D11 interface allows index-size-unaligned byte offsets into index buffers; it's not clear whether they actually work
+	# the D3D11 interface allows index-size-unaligned byte offsets into the index buffer; most drivers will abort with an assertion
 
 IaSetInputLayout -> bind_vertex_elements_state
 
diff --git a/src/gallium/docs/source/conf.py b/src/gallium/docs/source/conf.py
index ccc84405c41..0846e7d0ece 100644
--- a/src/gallium/docs/source/conf.py
+++ b/src/gallium/docs/source/conf.py
@@ -22,7 +22,7 @@ sys.path.append(os.path.abspath('exts'))
 
 # Add any Sphinx extension module names here, as strings. They can be extensions
 # coming with Sphinx (named 'sphinx.ext.*') or your custom ones.
-extensions = ['sphinx.ext.pngmath', 'tgsi']
+extensions = ['sphinx.ext.pngmath', 'formatting']
 
 # Add any paths that contain templates here, relative to this directory.
 templates_path = ['_templates']
@@ -38,7 +38,7 @@ master_doc = 'index'
 
 # General information about the project.
 project = u'Gallium'
-copyright = u'2009, VMWare, X.org, Nouveau'
+copyright = u'2009, VMware, X.org, Nouveau'
 
 # The version info for the project you're documenting, acts as replacement for
 # |version| and |release|, also used in various other places throughout the
@@ -176,7 +176,7 @@ htmlhelp_basename = 'Galliumdoc'
 # (source start file, target name, title, author, documentclass [howto/manual]).
 latex_documents = [
   ('index', 'Gallium.tex', u'Gallium Documentation',
-   u'VMWare, X.org, Nouveau', 'manual'),
+   u'VMware, X.org, Nouveau', 'manual'),
 ]
 
 # The name of an image file (relative to this directory) to place at the top of
diff --git a/src/gallium/docs/source/context.rst b/src/gallium/docs/source/context.rst
index 4e35a4c4082..f241411a002 100644
--- a/src/gallium/docs/source/context.rst
+++ b/src/gallium/docs/source/context.rst
@@ -45,6 +45,7 @@ buffers, surfaces) are bound to the driver.
 
 * ``set_vertex_buffers``
 
+* ``set_index_buffer``
 
 Non-CSO State
 ^^^^^^^^^^^^^
@@ -132,50 +133,26 @@ this surface need not be bound to the framebuffer.
 Drawing
 ^^^^^^^
 
-``draw_arrays`` draws a specified primitive.
+``draw_vbo`` draws a specified primitive.  The primitive mode and other
+properties are described by ``pipe_draw_info``.
 
-This command is equivalent to calling ``draw_arrays_instanced``
-with ``startInstance`` set to 0 and ``instanceCount`` set to 1.
+The ``mode``, ``start``, and ``count`` fields of ``pipe_draw_info`` specify the
+the mode of the primitive and the vertices to be fetched, in the range between
+``start`` to ``start``+``count``-1, inclusive.
 
-``draw_elements`` draws a specified primitive using an optional
-index buffer.
+Every instance with instanceID in the range between ``start_instance`` and
+``start_instance``+``instance_count``-1, inclusive, will be drawn.
 
-This command is equivalent to calling ``draw_elements_instanced``
-with ``startInstance`` set to 0 and ``instanceCount`` set to 1.
+All vertex indices must fall inside the range given by ``min_index`` and
+``max_index``.  In case non-indexed draw, ``min_index`` should be set to
+``start`` and ``max_index`` should be set to ``start``+``count``-1.
 
-``draw_range_elements``
+``index_bias`` is a value added to every vertex index before fetching vertex
+attributes.  It does not affect ``min_index`` and ``max_index``.
 
-XXX: this is (probably) a temporary entrypoint, as the range
-information should be available from the vertex_buffer state.
-Using this to quickly evaluate a specialized path in the draw
-module.
-
-``draw_arrays_instanced`` draws multiple instances of the same primitive.
-
-This command is equivalent to calling ``draw_elements_instanced``
-with ``indexBuffer`` set to NULL and ``indexSize`` set to 0.
-
-``draw_elements_instanced`` draws multiple instances of the same primitive
-using an optional index buffer.
-
-For instanceID in the range between ``startInstance``
-and ``startInstance``+``instanceCount``-1, inclusive, draw a primitive
-specified by ``mode`` and sequential numbers in the range between ``start``
-and ``start``+``count``-1, inclusive.
-
-If ``indexBuffer`` is not NULL, it specifies an index buffer with index
-byte size of ``indexSize``. The sequential numbers are used to lookup
-the index buffer and the resulting indices in turn are used to fetch
-vertex attributes.
-
-If ``indexBuffer`` is NULL, the sequential numbers are used directly
-as indices to fetch vertex attributes.
-
-``indexBias`` is a value which is added to every index read from the index 
-buffer before fetching vertex attributes.
-
-``minIndex`` and ``maxIndex`` describe minimum and maximum index contained in
-the index buffer.
+If there is an index buffer bound, and ``indexed`` field is true, all vertex
+indices will be looked up in the index buffer.  ``min_index``, ``max_index``,
+and ``index_bias`` apply after index lookup.
 
 If a given vertex element has ``instance_divisor`` set to 0, it is said
 it contains per-vertex data and effective vertex attribute address needs
diff --git a/src/gallium/docs/source/cso/rasterizer.rst b/src/gallium/docs/source/cso/rasterizer.rst
index ad1612f93e3..ee3419ccfca 100644
--- a/src/gallium/docs/source/cso/rasterizer.rst
+++ b/src/gallium/docs/source/cso/rasterizer.rst
@@ -126,11 +126,15 @@ sprite_coord_enable
 
 Specifies if a texture unit has its texture coordinates replaced or not. This
 is a packed bitfield containing the enable for all texcoords -- if all bits
-are zero, point sprites are effectively disabled. If any bit is set, then
-point_smooth and point_quad_rasterization are ignored; point smoothing is
-disabled and points are always rasterized as quads. If enabled, the four
-vertices of the resulting quad will be assigned texture coordinates,
-according to sprite_coord_mode.
+are zero, point sprites are effectively disabled. 
+
+If any bit is set, then point_smooth MUST be disabled (there are no
+round sprites) and point_quad_rasterization MUST be true (sprites are
+always rasterized as quads).  Any mismatch between these states should
+be considered a bug in the state-tracker.
+
+If enabled, the four vertices of the resulting quad will be assigned
+texture coordinates, according to sprite_coord_mode.
 
 sprite_coord_mode
 ^^^^^^^^^^^^^^^^^
@@ -141,20 +145,23 @@ have coordinates (0,0,0,1). For PIPE_SPRITE_COORD_UPPER_LEFT, the upper-left
 vertex will have coordinates (0,0,0,1).
 This state is used by :ref:`Draw` to generate texcoords.
 
-.. note::
-
-    When geometry shaders are available, a special geometry shader could be
-    used instead of this functionality, to convert incoming points into quads
-    with the proper texture coordinates.
-
 point_quad_rasterization
 ^^^^^^^^^^^^^^^^^^^^^^^^
 
-Determines if points should be rasterized as quads or points. Certain APIs,
-like Direct3D, always use quad rasterization for points, regardless of
-whether point sprites are enabled or not. If this state is enabled, point
-smoothing and antialiasing are disabled. If it is disabled, point sprite
-coordinates are not generated.
+Determines if points should be rasterized according to quad or point
+rasterization rules.
+
+OpenGL actually has quite different rasterization rules for points and
+point sprites - hence this indicates if points should be rasterized as
+points or according to point sprite (which decomposes them into quads,
+basically) rules.
+
+Additionally Direct3D will always use quad rasterization rules for
+points, regardless of whether point sprites are enabled or not.
+
+If this state is enabled, point smoothing and antialiasing are
+disabled. If it is disabled, point sprite coordinates are not
+generated.
 
 .. note::
 
diff --git a/src/gallium/docs/source/debugging.rst b/src/gallium/docs/source/debugging.rst
new file mode 100644
index 00000000000..42bda5aee93
--- /dev/null
+++ b/src/gallium/docs/source/debugging.rst
@@ -0,0 +1,101 @@
+Debugging
+=========
+
+Debugging utilities in gallium.
+
+Debug Variables
+^^^^^^^^^^^^^^^
+
+All drivers respond to a set of common debug environment variables, as well as
+some driver-specific variables. Set them as normal environment variables for
+the platform or operating system you are running. For example, for Linux this
+can be done by typing "export var=value" into a console and then running the
+program from that console.
+
+Common
+""""""
+
+.. envvar:: GALLIUM_PRINT_OPTIONS <bool> (false)
+
+This option controls if the debug variables should be printed to stderr. This
+is probably the most useful variable, since it allows you to find which
+variables a driver uses.
+
+.. envvar:: GALLIUM_RBUG <bool> (false)
+
+Controls if the :ref:`rbug` should be used.
+
+.. envvar:: GALLIUM_TRACE <string> ("")
+
+If set, this variable will cause the :ref:`Trace` output to be written to the
+specified file. Paths may be relative or absolute; relative paths are relative
+to the working directory.  For example, setting it to "trace.xml" will cause
+the trace to be written to a file of the same name in the working directory.
+
+.. envvar:: GALLIUM_DUMP_CPU <bool> (false)
+
+Dump information about the current CPU that the driver is running on.
+
+.. envvar:: TGSI_PRINT_SANITY <bool> (false)
+
+Gallium has a built-in shader sanity checker.  This option controls whether
+the shader sanity checker prints its warnings and errors to stderr.
+
+.. envvar:: DRAW_USE_LLVM <bool> (false)
+
+Whether the :ref:`Draw` module will attempt to use LLVM for vertex and geometry shaders.
+
+
+State tracker-specific
+""""""""""""""""""""""
+
+.. envvar:: ST_DEBUG <flags> (0x0)
+
+Debug :ref:`flags` for the GL state tracker.
+
+
+Driver-specific
+"""""""""""""""
+
+.. envvar:: I915_DEBUG <flags> (0x0)
+
+Debug :ref:`flags` for the i915 driver.
+
+.. envvar:: I915_NO_HW <bool> (false)
+
+Stop the i915 driver from submitting commands to the hardware.
+
+.. envvar:: I915_DUMP_CMD <bool> (false)
+
+Dump all commands going to the hardware.
+
+.. envvar:: LP_DEBUG <flags> (0x0)
+
+Debug :ref:`flags` for the llvmpipe driver.
+
+.. envvar:: LP_NUM_THREADS <int> (number of CPUs)
+
+Number of threads that the llvmpipe driver should use.
+
+
+.. _flags:
+
+Flags
+"""""
+
+The variables of type "flags" all take a string with comma-separated flags to
+enable different debugging for different parts of the drivers or state
+tracker. If set to "help", the driver will print a list of flags which the
+variable accepts. Order does not matter.
+
+
+.. _rbug:
+
+Remote Debugger
+^^^^^^^^^^^^^^^
+
+The remote debugger, commonly known as rbug, allows for runtime inspections of
+:ref:`Context`, :ref:`Screen`, :ref:`Resource` and :ref:`Shader` objects; and
+pausing and stepping of :ref:`Draw` calls. Is used with rbug-gui which is
+hosted outside of the main mesa repository. rbug is can be used over a network
+connection, so the debugger does not need to be on the same machine.
diff --git a/src/gallium/docs/source/distro.rst b/src/gallium/docs/source/distro.rst
index 6ba5a056f45..70d75b51e65 100644
--- a/src/gallium/docs/source/distro.rst
+++ b/src/gallium/docs/source/distro.rst
@@ -51,10 +51,10 @@ nVidia nv50
 
 Driver for the nVidia nv50 family of GPUs.
 
-VMWare SVGA
+VMware SVGA
 ^^^^^^^^^^^
 
-Driver for VMWare virtualized guest operating system graphics processing.
+Driver for VMware virtualized guest operating system graphics processing.
 
 ATI r300
 ^^^^^^^^
@@ -74,6 +74,11 @@ Trace
 Wrapper driver. Trace dumps an XML record of the calls made to the
 :ref:`Context` and :ref:`Screen` objects that it wraps.
 
+Rbug
+^^^^
+
+Wrapper driver. :ref:`rbug` driver used with stand alone rbug-gui.
+
 State Trackers
 --------------
 
diff --git a/src/gallium/docs/source/exts/formatting.py b/src/gallium/docs/source/exts/formatting.py
new file mode 100644
index 00000000000..14865f36033
--- /dev/null
+++ b/src/gallium/docs/source/exts/formatting.py
@@ -0,0 +1,31 @@
+# formatting.py
+# Sphinx extension providing formatting for Gallium-specific data
+# (c) Corbin Simpson 2010
+# Public domain to the extent permitted; contact author for special licensing
+
+import docutils.nodes
+import sphinx.addnodes
+
+def parse_envvar(env, sig, signode):
+    envvar, t, default = sig.split(" ", 2)
+    envvar = envvar.strip().upper()
+    t = " Type: %s" % t.strip(" <>").lower()
+    default = " Default: %s" % default.strip(" ()")
+    signode += sphinx.addnodes.desc_name(envvar, envvar)
+    signode += sphinx.addnodes.desc_type(t, t)
+    signode += sphinx.addnodes.desc_annotation(default, default)
+    return envvar
+
+def parse_opcode(env, sig, signode):
+    opcode, desc = sig.split("-", 1)
+    opcode = opcode.strip().upper()
+    desc = " (%s)" % desc.strip()
+    signode += sphinx.addnodes.desc_name(opcode, opcode)
+    signode += sphinx.addnodes.desc_annotation(desc, desc)
+    return opcode
+
+def setup(app):
+    app.add_description_unit("envvar", "envvar", "%s (environment variable)",
+        parse_envvar)
+    app.add_description_unit("opcode", "opcode", "%s (TGSI opcode)",
+        parse_opcode)
diff --git a/src/gallium/docs/source/exts/tgsi.py b/src/gallium/docs/source/exts/tgsi.py
deleted file mode 100644
index e92cd5c4d1b..00000000000
--- a/src/gallium/docs/source/exts/tgsi.py
+++ /dev/null
@@ -1,17 +0,0 @@
-# tgsi.py
-# Sphinx extension providing formatting for TGSI opcodes
-# (c) Corbin Simpson 2010
-
-import docutils.nodes
-import sphinx.addnodes
-
-def parse_opcode(env, sig, signode):
-    opcode, desc = sig.split("-", 1)
-    opcode = opcode.strip().upper()
-    desc = " (%s)" % desc.strip()
-    signode += sphinx.addnodes.desc_name(opcode, opcode)
-    signode += sphinx.addnodes.desc_annotation(desc, desc)
-    return opcode
-
-def setup(app):
-    app.add_description_unit("opcode", "opcode", "%s (TGSI opcode)", parse_opcode)
diff --git a/src/gallium/docs/source/index.rst b/src/gallium/docs/source/index.rst
index 54bc883fced..6c19842dac4 100644
--- a/src/gallium/docs/source/index.rst
+++ b/src/gallium/docs/source/index.rst
@@ -12,6 +12,7 @@ Contents:
    :maxdepth: 2
 
    intro
+   debugging
    tgsi
    screen
    context
diff --git a/src/gallium/drivers/cell/common.h b/src/gallium/drivers/cell/common.h
index bbb112fd336..a8cdde34aa7 100644
--- a/src/gallium/drivers/cell/common.h
+++ b/src/gallium/drivers/cell/common.h
@@ -230,7 +230,7 @@ struct cell_command_rasterizer
 {
    opcode_t opcode;    /**< CELL_CMD_STATE_RASTERIZER */
    struct pipe_rasterizer_state rasterizer;
-   uint32_t pad[1];
+   /*uint32_t pad[1];*/
 };
 
 
diff --git a/src/gallium/drivers/cell/ppu/cell_context.h b/src/gallium/drivers/cell/ppu/cell_context.h
index 07b6eebc69c..d1aee62ba1e 100644
--- a/src/gallium/drivers/cell/ppu/cell_context.h
+++ b/src/gallium/drivers/cell/ppu/cell_context.h
@@ -132,6 +132,7 @@ struct cell_context
    struct pipe_viewport_state viewport;
    struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS];
    uint num_vertex_buffers;
+   struct pipe_index_buffer index_buffer;
 
    ubyte *cbuf_map[PIPE_MAX_COLOR_BUFS];
    ubyte *zsbuf_map;
@@ -154,7 +155,7 @@ struct cell_context
    struct vertex_info vertex_info;
 
    /** Mapped constant buffers */
-   void *mapped_constants[PIPE_SHADER_TYPES];
+   const void *mapped_constants[PIPE_SHADER_TYPES];
 
    PIPE_ALIGN_VAR(16) struct cell_spu_function_info spu_functions;
 
diff --git a/src/gallium/drivers/cell/ppu/cell_draw_arrays.c b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c
index 6a1e4d8a646..4adef5b8c07 100644
--- a/src/gallium/drivers/cell/ppu/cell_draw_arrays.c
+++ b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c
@@ -56,16 +56,11 @@
  * XXX should the element buffer be specified/bound with a separate function?
  */
 static void
-cell_draw_range_elements(struct pipe_context *pipe,
-                         struct pipe_resource *indexBuffer,
-                         unsigned indexSize,
-                         int indexBias,
-                         unsigned min_index,
-                         unsigned max_index,
-                         unsigned mode, unsigned start, unsigned count)
+cell_draw_vbo(struct pipe_context *pipe, const struct pipe_draw_info *info)
 {
    struct cell_context *cell = cell_context(pipe);
    struct draw_context *draw = cell->draw;
+   void *mapped_indices = NULL;
    unsigned i;
 
    if (cell->dirty)
@@ -83,18 +78,20 @@ cell_draw_range_elements(struct pipe_context *pipe,
       draw_set_mapped_vertex_buffer(draw, i, buf);
    }
    /* Map index buffer, if present */
-   if (indexBuffer) {
-      void *mapped_indexes = cell_resource(indexBuffer)->data;
-      draw_set_mapped_element_buffer(draw, indexSize, indexBias, mapped_indexes);
-   }
-   else {
-      /* no index/element buffer */
-      draw_set_mapped_element_buffer(draw, 0, 0, NULL);
+   if (info->indexed && cell->index_buffer.buffer) {
+      mapped_indices = cell_resource(cell->index_buffer.buffer)->data;
+      mapped_indices += cell->index_buffer.offset;
    }
 
+   draw_set_mapped_element_buffer_range(draw, (mapped_indices) ?
+                                        lp->index_buffer.index_size : 0,
+                                        info->index_bias,
+                                        info->min_index,
+                                        info->max_index,
+                                        mapped_indices);
 
    /* draw! */
-   draw_arrays(draw, mode, start, count);
+   draw_arrays(draw, info->mode, info->start, info->count);
 
    /*
     * unmap vertex/index buffers - will cause draw module to flush
@@ -102,7 +99,7 @@ cell_draw_range_elements(struct pipe_context *pipe,
    for (i = 0; i < cell->num_vertex_buffers; i++) {
       draw_set_mapped_vertex_buffer(draw, i, NULL);
    }
-   if (indexBuffer) {
+   if (mapped_indices) {
       draw_set_mapped_element_buffer(draw, 0, 0, NULL);
    }
 
@@ -115,32 +112,9 @@ cell_draw_range_elements(struct pipe_context *pipe,
 }
 
 
-static void
-cell_draw_elements(struct pipe_context *pipe,
-                   struct pipe_resource *indexBuffer,
-                   unsigned indexSize, int indexBias,
-                   unsigned mode, unsigned start, unsigned count)
-{
-   cell_draw_range_elements( pipe, indexBuffer,
-                             indexSize, indexBias,
-                             0, 0xffffffff,
-                             mode, start, count );
-}
-
-
-static void
-cell_draw_arrays(struct pipe_context *pipe, unsigned mode,
-                     unsigned start, unsigned count)
-{
-   cell_draw_elements(pipe, NULL, 0, 0, mode, start, count);
-}
-
-
 void
 cell_init_draw_functions(struct cell_context *cell)
 {
-   cell->pipe.draw_arrays = cell_draw_arrays;
-   cell->pipe.draw_elements = cell_draw_elements;
-   cell->pipe.draw_range_elements = cell_draw_range_elements;
+   cell->pipe.draw_vbo = cell_draw_vbo;
 }
 
diff --git a/src/gallium/drivers/cell/ppu/cell_fence.c b/src/gallium/drivers/cell/ppu/cell_fence.c
index 34ca864155b..e7c9fc46d9f 100644
--- a/src/gallium/drivers/cell/ppu/cell_fence.c
+++ b/src/gallium/drivers/cell/ppu/cell_fence.c
@@ -87,6 +87,7 @@ struct cell_buffer_node
 };
 
 
+#if 0
 static void
 cell_add_buffer_to_list(struct cell_context *cell,
                         struct cell_buffer_list *list,
@@ -100,6 +101,7 @@ cell_add_buffer_to_list(struct cell_context *cell,
       list->head = node;
    }
 }
+#endif
 
 
 /**
@@ -113,7 +115,7 @@ cell_free_fenced_buffers(struct cell_context *cell,
                          struct cell_buffer_list *list)
 {
    if (list->head) {
-      struct pipe_screen *ps = cell->pipe.screen;
+      /*struct pipe_screen *ps = cell->pipe.screen;*/
       struct cell_buffer_node *node;
 
       cell_fence_finish(cell, &list->fence);
@@ -146,7 +148,7 @@ cell_free_fenced_buffers(struct cell_context *cell,
 void
 cell_add_fenced_textures(struct cell_context *cell)
 {
-   struct cell_buffer_list *list = &cell->fenced_buffers[cell->cur_batch];
+   /*struct cell_buffer_list *list = &cell->fenced_buffers[cell->cur_batch];*/
    uint i;
 
    for (i = 0; i < cell->num_textures; i++) {
diff --git a/src/gallium/drivers/cell/ppu/cell_pipe_state.c b/src/gallium/drivers/cell/ppu/cell_pipe_state.c
index 03f84d295b5..223adda48f0 100644
--- a/src/gallium/drivers/cell/ppu/cell_pipe_state.c
+++ b/src/gallium/drivers/cell/ppu/cell_pipe_state.c
@@ -281,7 +281,7 @@ cell_set_fragment_sampler_views(struct pipe_context *pipe,
          struct pipe_resource *new_tex = new_view ? new_view->texture : NULL;
 
          pipe_sampler_view_reference(&cell->fragment_sampler_views[i],
-                                     views[i]);
+                                     new_view);
          pipe_resource_reference((struct pipe_resource **) &cell->texture[i],
                                 (struct pipe_resource *) new_tex);
 
diff --git a/src/gallium/drivers/cell/ppu/cell_state_vertex.c b/src/gallium/drivers/cell/ppu/cell_state_vertex.c
index 9510ea9ac2b..4e3701cd0ac 100644
--- a/src/gallium/drivers/cell/ppu/cell_state_vertex.c
+++ b/src/gallium/drivers/cell/ppu/cell_state_vertex.c
@@ -36,7 +36,7 @@
 #include "draw/draw_context.h"
 
 
-void *
+static void *
 cell_create_vertex_elements_state(struct pipe_context *pipe,
                                   unsigned count,
                                   const struct pipe_vertex_element *attribs)
@@ -51,7 +51,7 @@ cell_create_vertex_elements_state(struct pipe_context *pipe,
    return velems;
 }
 
-void
+static void
 cell_bind_vertex_elements_state(struct pipe_context *pipe,
                                 void *velems)
 {
@@ -66,7 +66,7 @@ cell_bind_vertex_elements_state(struct pipe_context *pipe,
       draw_set_vertex_elements(cell->draw, cell_velems->count, cell_velems->velem);
 }
 
-void
+static void
 cell_delete_vertex_elements_state(struct pipe_context *pipe, void *velems)
 {
    FREE( velems );
@@ -91,10 +91,26 @@ cell_set_vertex_buffers(struct pipe_context *pipe,
 }
 
 
+static void
+cell_set_index_buffer(struct pipe_context *pipe,
+                      const struct pipe_index_buffer *ib)
+{
+   struct cell_context *cell = cell_context(pipe);
+
+   if (ib)
+      memcpy(&cell->index_buffer, ib, sizeof(cell->index_buffer));
+   else
+      memset(&cell->index_buffer, 0, sizeof(cell->index_buffer));
+
+   /* TODO make this more like a state */
+}
+
+
 void
 cell_init_vertex_functions(struct cell_context *cell)
 {
    cell->pipe.set_vertex_buffers = cell_set_vertex_buffers;
+   cell->pipe.set_index_buffer = cell_set_index_buffer;
    cell->pipe.create_vertex_elements_state = cell_create_vertex_elements_state;
    cell->pipe.bind_vertex_elements_state = cell_bind_vertex_elements_state;
    cell->pipe.delete_vertex_elements_state = cell_delete_vertex_elements_state;
diff --git a/src/gallium/drivers/failover/fo_context.c b/src/gallium/drivers/failover/fo_context.c
index 9c9c1bdc452..761a0fce721 100644
--- a/src/gallium/drivers/failover/fo_context.c
+++ b/src/gallium/drivers/failover/fo_context.c
@@ -50,13 +50,8 @@ void failover_fail_over( struct failover_context *failover )
 }
 
 
-static void failover_draw_elements( struct pipe_context *pipe,
-                                    struct pipe_resource *indexResource,
-                                    unsigned indexSize,
-                                    int indexBias,
-                                    unsigned prim, 
-                                    unsigned start, 
-                                    unsigned count)
+static void failover_draw_vbo( struct pipe_context *pipe,
+                               const struct pipe_draw_info *info)
 {
    struct failover_context *failover = failover_context( pipe );
 
@@ -70,13 +65,7 @@ static void failover_draw_elements( struct pipe_context *pipe,
    /* Try hardware:
     */
    if (failover->mode == FO_HW) {
-      failover->hw->draw_elements( failover->hw, 
-                                   indexResource, 
-                                   indexSize, 
-                                   indexBias,
-                                   prim, 
-                                   start, 
-                                   count );
+      failover->hw->draw_vbo( failover->hw, info );
    }
 
    /* Possibly try software:
@@ -88,13 +77,7 @@ static void failover_draw_elements( struct pipe_context *pipe,
 	 failover_state_emit( failover );
       }
 
-      failover->sw->draw_elements( failover->sw, 
-				   indexResource, 
-				   indexSize, 
-				   indexBias,
-				   prim, 
-				   start, 
-				   count );
+      failover->sw->draw_vbo( failover->sw, info );
 
       /* Be ready to switch back to hardware rendering without an
        * intervening flush.  Unlikely to be much performance impact to
@@ -104,13 +87,6 @@ static void failover_draw_elements( struct pipe_context *pipe,
    }
 }
 
-
-static void failover_draw_arrays( struct pipe_context *pipe,
-				     unsigned prim, unsigned start, unsigned count)
-{
-   failover_draw_elements(pipe, NULL, 0, 0, prim, start, count);
-}
-
 static unsigned int
 failover_is_resource_referenced( struct pipe_context *_pipe,
 				 struct pipe_resource *resource,
@@ -143,8 +119,7 @@ struct pipe_context *failover_create( struct pipe_context *hw,
    failover->pipe.get_paramf = hw->get_paramf;
 #endif
 
-   failover->pipe.draw_arrays = failover_draw_arrays;
-   failover->pipe.draw_elements = failover_draw_elements;
+   failover->pipe.draw_vbo = failover_draw_vbo;
    failover->pipe.clear = hw->clear;
    failover->pipe.clear_render_target = hw->clear_render_target;
    failover->pipe.clear_depth_stencil = hw->clear_depth_stencil;
diff --git a/src/gallium/drivers/failover/fo_context.h b/src/gallium/drivers/failover/fo_context.h
index 9d3e0d0dba0..1afa6c9ceed 100644
--- a/src/gallium/drivers/failover/fo_context.h
+++ b/src/gallium/drivers/failover/fo_context.h
@@ -56,6 +56,7 @@
 #define FO_NEW_VERTEX_BUFFER   0x40000
 #define FO_NEW_VERTEX_ELEMENT  0x80000
 #define FO_NEW_SAMPLE_MASK     0x100000
+#define FO_NEW_INDEX_BUFFER    0x200000
 
 
 
@@ -97,6 +98,7 @@ struct failover_context {
    struct pipe_scissor_state scissor;
    struct pipe_viewport_state viewport;
    struct pipe_vertex_buffer vertex_buffers[PIPE_MAX_ATTRIBS];
+   struct pipe_index_buffer index_buffer;
 
    uint num_vertex_buffers;
 
diff --git a/src/gallium/drivers/failover/fo_state.c b/src/gallium/drivers/failover/fo_state.c
index 12e42379f98..c265f381b67 100644
--- a/src/gallium/drivers/failover/fo_state.c
+++ b/src/gallium/drivers/failover/fo_state.c
@@ -583,6 +583,23 @@ failover_set_vertex_buffers(struct pipe_context *pipe,
 }
 
 
+static void
+failover_set_index_buffer(struct pipe_context *pipe,
+                          const struct pipe_index_buffer *ib)
+{
+   struct failover_context *failover = failover_context(pipe);
+
+   if (ib)
+      memcpy(&failover->index_buffer, ib, sizeof(failover->index_buffer));
+   else
+      memset(&failover->index_buffer, 0, sizeof(failover->index_buffer));
+
+   failover->dirty |= FO_NEW_INDEX_BUFFER;
+   failover->sw->set_index_buffer( failover->sw, ib );
+   failover->hw->set_index_buffer( failover->hw, ib );
+}
+
+
 void
 failover_set_constant_buffer(struct pipe_context *pipe,
                              uint shader, uint index,
@@ -635,6 +652,7 @@ failover_init_state_functions( struct failover_context *failover )
    failover->pipe.set_vertex_sampler_views = failover_set_vertex_sampler_views;
    failover->pipe.set_viewport_state = failover_set_viewport_state;
    failover->pipe.set_vertex_buffers = failover_set_vertex_buffers;
+   failover->pipe.set_index_buffer = failover_set_index_buffer;
    failover->pipe.set_constant_buffer = failover_set_constant_buffer;
    failover->pipe.create_sampler_view = failover_create_sampler_view;
    failover->pipe.sampler_view_destroy = failover_sampler_view_destroy;
diff --git a/src/gallium/drivers/failover/fo_state_emit.c b/src/gallium/drivers/failover/fo_state_emit.c
index 147f23269ca..7f434ff9d68 100644
--- a/src/gallium/drivers/failover/fo_state_emit.c
+++ b/src/gallium/drivers/failover/fo_state_emit.c
@@ -135,5 +135,10 @@ failover_state_emit( struct failover_context *failover )
                                         failover->vertex_buffers );
    }
 
+   if (failover->dirty & FO_NEW_INDEX_BUFFER) {
+      failover->sw->set_index_buffer( failover->sw,
+                                      &failover->index_buffer );
+   }
+
    failover->dirty = 0;
 }
diff --git a/src/gallium/drivers/galahad/glhd_context.c b/src/gallium/drivers/galahad/glhd_context.c
index ab6f17b3ab8..fe14a287efb 100644
--- a/src/gallium/drivers/galahad/glhd_context.c
+++ b/src/gallium/drivers/galahad/glhd_context.c
@@ -48,68 +48,13 @@ galahad_destroy(struct pipe_context *_pipe)
 }
 
 static void
-galahad_draw_arrays(struct pipe_context *_pipe,
-                     unsigned prim,
-                     unsigned start,
-                     unsigned count)
+galahad_draw_vbo(struct pipe_context *_pipe,
+                 const struct pipe_draw_info *info)
 {
    struct galahad_context *glhd_pipe = galahad_context(_pipe);
    struct pipe_context *pipe = glhd_pipe->pipe;
 
-   pipe->draw_arrays(pipe,
-                     prim,
-                     start,
-                     count);
-}
-
-static void
-galahad_draw_elements(struct pipe_context *_pipe,
-                       struct pipe_resource *_indexResource,
-                       unsigned indexSize,
-                       int indexBias,
-                       unsigned prim,
-                       unsigned start,
-                       unsigned count)
-{
-   struct galahad_context *glhd_pipe = galahad_context(_pipe);
-   struct galahad_resource *glhd_resource = galahad_resource(_indexResource);
-   struct pipe_context *pipe = glhd_pipe->pipe;
-   struct pipe_resource *indexResource = glhd_resource->resource;
-
-   pipe->draw_elements(pipe,
-                       indexResource,
-                       indexSize,
-                       indexBias,
-                       prim,
-                       start,
-                       count);
-}
-
-static void
-galahad_draw_range_elements(struct pipe_context *_pipe,
-                             struct pipe_resource *_indexResource,
-                             unsigned indexSize,
-                             int indexBias,
-                             unsigned minIndex,
-                             unsigned maxIndex,
-                             unsigned mode,
-                             unsigned start,
-                             unsigned count)
-{
-   struct galahad_context *glhd_pipe = galahad_context(_pipe);
-   struct galahad_resource *glhd_resource = galahad_resource(_indexResource);
-   struct pipe_context *pipe = glhd_pipe->pipe;
-   struct pipe_resource *indexResource = glhd_resource->resource;
-
-   pipe->draw_range_elements(pipe,
-                             indexResource,
-                             indexSize,
-                             indexBias,
-                             minIndex,
-                             maxIndex,
-                             mode,
-                             start,
-                             count);
+   pipe->draw_vbo(pipe, info);
 }
 
 static struct pipe_query *
@@ -650,6 +595,41 @@ galahad_set_vertex_buffers(struct pipe_context *_pipe,
                             num_buffers,
                             buffers);
 }
+
+static void
+galahad_set_index_buffer(struct pipe_context *_pipe,
+                         const struct pipe_index_buffer *_ib)
+{
+   struct galahad_context *glhd_pipe = galahad_context(_pipe);
+   struct pipe_context *pipe = glhd_pipe->pipe;
+   struct pipe_index_buffer unwrapped_ib, *ib = NULL;
+
+   if (_ib->buffer) {
+      switch (_ib->index_size) {
+      case 1:
+      case 2:
+      case 4:
+         break;
+      default:
+         glhd_warn("index buffer %p has unrecognized index size %d",
+               _ib->buffer, _ib->index_size);
+         break;
+      }
+   }
+   else if (_ib->offset || _ib->index_size) {
+      glhd_warn("non-indexed state with index offset %d and index size %d",
+            _ib->offset, _ib->index_size);
+   }
+
+   if (_ib) {
+      unwrapped_ib = *_ib;
+      unwrapped_ib.buffer = galahad_resource_unwrap(_ib->buffer);
+      ib = &unwrapped_ib;
+   }
+
+   pipe->set_index_buffer(pipe, ib);
+}
+
 static void
 galahad_resource_copy_region(struct pipe_context *_pipe,
                               struct pipe_resource *_dst,
@@ -934,9 +914,7 @@ galahad_context_create(struct pipe_screen *_screen, struct pipe_context *pipe)
    glhd_pipe->base.draw = NULL;
 
    glhd_pipe->base.destroy = galahad_destroy;
-   glhd_pipe->base.draw_arrays = galahad_draw_arrays;
-   glhd_pipe->base.draw_elements = galahad_draw_elements;
-   glhd_pipe->base.draw_range_elements = galahad_draw_range_elements;
+   glhd_pipe->base.draw_vbo = galahad_draw_vbo;
    glhd_pipe->base.create_query = galahad_create_query;
    glhd_pipe->base.destroy_query = galahad_destroy_query;
    glhd_pipe->base.begin_query = galahad_begin_query;
@@ -976,6 +954,7 @@ galahad_context_create(struct pipe_screen *_screen, struct pipe_context *pipe)
    glhd_pipe->base.set_fragment_sampler_views = galahad_set_fragment_sampler_views;
    glhd_pipe->base.set_vertex_sampler_views = galahad_set_vertex_sampler_views;
    glhd_pipe->base.set_vertex_buffers = galahad_set_vertex_buffers;
+   glhd_pipe->base.set_index_buffer = galahad_set_index_buffer;
    glhd_pipe->base.resource_copy_region = galahad_resource_copy_region;
    glhd_pipe->base.clear = galahad_clear;
    glhd_pipe->base.clear_render_target = galahad_clear_render_target;
diff --git a/src/gallium/drivers/i915/i915_context.c b/src/gallium/drivers/i915/i915_context.c
index 2af9bdac956..2beb9e3091f 100644
--- a/src/gallium/drivers/i915/i915_context.c
+++ b/src/gallium/drivers/i915/i915_context.c
@@ -45,16 +45,11 @@
 
 
 static void
-i915_draw_range_elements(struct pipe_context *pipe,
-                         struct pipe_resource *indexBuffer,
-                         unsigned indexSize,
-                         int indexBias,
-                         unsigned min_index,
-                         unsigned max_index,
-                         unsigned prim, unsigned start, unsigned count)
+i915_draw_vbo(struct pipe_context *pipe, const struct pipe_draw_info *info)
 {
    struct i915_context *i915 = i915_context(pipe);
    struct draw_context *draw = i915->draw;
+   void *mapped_indices = NULL;
    unsigned i;
 
    if (i915->dirty)
@@ -71,16 +66,18 @@ i915_draw_range_elements(struct pipe_context *pipe,
    /*
     * Map index buffer, if present
     */
-   if (indexBuffer) {
-      void *mapped_indexes = i915_buffer(indexBuffer)->data;
-      draw_set_mapped_element_buffer_range(draw, indexSize, indexBias,
-                                           min_index,
-                                           max_index,
-                                           mapped_indexes);
-   } else {
-      draw_set_mapped_element_buffer(draw, 0, 0, NULL);
+   if (info->indexed && i915->index_buffer.buffer) {
+      char *indices = (char *) i915_buffer(i915->index_buffer.buffer)->data;
+      mapped_indices = (void *) (indices + i915->index_buffer.offset);
    }
 
+   draw_set_mapped_element_buffer_range(draw, (mapped_indices) ?
+                                        i915->index_buffer.index_size : 0,
+                                        info->index_bias,
+                                        info->min_index,
+                                        info->max_index,
+                                        mapped_indices);
+
 
    draw_set_mapped_constant_buffer(draw, PIPE_SHADER_VERTEX, 0,
                                    i915->current.constants[PIPE_SHADER_VERTEX],
@@ -90,7 +87,7 @@ i915_draw_range_elements(struct pipe_context *pipe,
    /*
     * Do the drawing
     */
-   draw_arrays(i915->draw, prim, start, count);
+   draw_arrays(i915->draw, info->mode, info->start, info->count);
 
    /*
     * unmap vertex/index buffers
@@ -99,32 +96,11 @@ i915_draw_range_elements(struct pipe_context *pipe,
       draw_set_mapped_vertex_buffer(draw, i, NULL);
    }
 
-   if (indexBuffer) {
+   if (mapped_indices) {
       draw_set_mapped_element_buffer(draw, 0, 0, NULL);
    }
 }
 
-static void
-i915_draw_elements(struct pipe_context *pipe,
-                   struct pipe_resource *indexBuffer,
-                   unsigned indexSize, int indexBias,
-                   unsigned prim, unsigned start, unsigned count)
-{
-   i915_draw_range_elements(pipe, indexBuffer,
-                            indexSize, indexBias,
-                            0, 0xffffffff,
-                            prim, start, count);
-}
-
-static void
-i915_draw_arrays(struct pipe_context *pipe,
-                 unsigned prim, unsigned start, unsigned count)
-{
-   i915_draw_elements(pipe, NULL, 0, 0, prim, start, count);
-}
-
-
-
 
 /*
  * Generic context functions
@@ -168,9 +144,7 @@ i915_create_context(struct pipe_screen *screen, void *priv)
 
    i915->base.clear = i915_clear;
 
-   i915->base.draw_arrays = i915_draw_arrays;
-   i915->base.draw_elements = i915_draw_elements;
-   i915->base.draw_range_elements = i915_draw_range_elements;
+   i915->base.draw_vbo = i915_draw_vbo;
 
    /*
     * Create drawing context and plug our rendering stage into it.
diff --git a/src/gallium/drivers/i915/i915_context.h b/src/gallium/drivers/i915/i915_context.h
index b210cb130d0..3ae61d0ea70 100644
--- a/src/gallium/drivers/i915/i915_context.h
+++ b/src/gallium/drivers/i915/i915_context.h
@@ -221,6 +221,7 @@ struct i915_context
    struct pipe_sampler_view *fragment_sampler_views[PIPE_MAX_SAMPLERS];
    struct pipe_viewport_state viewport;
    struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS];
+   struct pipe_index_buffer index_buffer;
 
    unsigned dirty;
 
diff --git a/src/gallium/drivers/i915/i915_state.c b/src/gallium/drivers/i915/i915_state.c
index e767aa9f8f0..385c3b2d2d3 100644
--- a/src/gallium/drivers/i915/i915_state.c
+++ b/src/gallium/drivers/i915/i915_state.c
@@ -812,6 +812,19 @@ i915_delete_vertex_elements_state(struct pipe_context *pipe, void *velems)
    FREE( velems );
 }
 
+static void i915_set_index_buffer(struct pipe_context *pipe,
+                                  const struct pipe_index_buffer *ib)
+{
+   struct i915_context *i915 = i915_context(pipe);
+
+   if (ib)
+      memcpy(&i915->index_buffer, ib, sizeof(i915->index_buffer));
+   else
+      memset(&i915->index_buffer, 0, sizeof(i915->index_buffer));
+
+   /* TODO make this more like a state */
+}
+
 static void
 i915_set_sample_mask(struct pipe_context *pipe,
                      unsigned sample_mask)
@@ -860,4 +873,5 @@ i915_init_state_functions( struct i915_context *i915 )
    i915->base.sampler_view_destroy = i915_sampler_view_destroy;
    i915->base.set_viewport_state = i915_set_viewport_state;
    i915->base.set_vertex_buffers = i915_set_vertex_buffers;
+   i915->base.set_index_buffer = i915_set_index_buffer;
 }
diff --git a/src/gallium/drivers/i965/brw_context.h b/src/gallium/drivers/i965/brw_context.h
index 94c9c443f05..56d351f97d1 100644
--- a/src/gallium/drivers/i965/brw_context.h
+++ b/src/gallium/drivers/i965/brw_context.h
@@ -576,6 +576,7 @@ struct brw_context
        */
       struct pipe_resource *index_buffer;
       unsigned index_size;
+      unsigned index_offset;
 
       /* Updates are signalled by PIPE_NEW_INDEX_RANGE:
        */
diff --git a/src/gallium/drivers/i965/brw_draw.c b/src/gallium/drivers/i965/brw_draw.c
index 4625c2048f9..3ab9024c31e 100644
--- a/src/gallium/drivers/i965/brw_draw.c
+++ b/src/gallium/drivers/i965/brw_draw.c
@@ -142,7 +142,7 @@ static int brw_emit_prim(struct brw_context *brw,
  */
 static int
 try_draw_range_elements(struct brw_context *brw,
-			struct pipe_resource *index_buffer,
+			boolean indexed,
 			unsigned hw_prim, 
 			unsigned start, unsigned count)
 {
@@ -165,7 +165,7 @@ try_draw_range_elements(struct brw_context *brw,
    if (ret)
       return ret;
    
-   ret = brw_emit_prim(brw, start, count, index_buffer != NULL, hw_prim);
+   ret = brw_emit_prim(brw, start, count, indexed, hw_prim);
    if (ret)
       return ret;
 
@@ -177,91 +177,54 @@ try_draw_range_elements(struct brw_context *brw,
 
 
 static void
-brw_draw_range_elements(struct pipe_context *pipe,
-			struct pipe_resource *index_buffer,
-			unsigned index_size, int index_bias,
-			unsigned min_index,
-			unsigned max_index,
-			unsigned mode, unsigned start, unsigned count)
+brw_draw_vbo(struct pipe_context *pipe, const struct pipe_draw_info *info)
 {
    struct brw_context *brw = brw_context(pipe);
    int ret;
    uint32_t hw_prim;
 
-   hw_prim = brw_set_prim(brw, mode);
+   hw_prim = brw_set_prim(brw, info->mode);
 
    if (BRW_DEBUG & DEBUG_PRIMS)
       debug_printf("PRIM: %s start %d count %d index_buffer %p\n",
-                   u_prim_name(mode), start, count, (void *)index_buffer);
+                   u_prim_name(info->mode), info->start, info->count,
+                   (void *) brw->curr.index_buffer);
 
-   assert(index_bias == 0);
+   assert(info->index_bias == 0);
 
-   /* Potentially trigger upload of new index buffer.
-    *
-    * XXX: do we need to go through state validation to achieve this?
-    * Could just call upload code directly.
+   /* Potentially trigger upload of new index buffer range.
+    * XXX: do we really care?
     */
-   if (brw->curr.index_buffer != index_buffer ||
-       brw->curr.index_size != index_size) {
-      pipe_resource_reference( &brw->curr.index_buffer, index_buffer );
-      brw->curr.index_size = index_size;
-      brw->state.dirty.mesa |= PIPE_NEW_INDEX_BUFFER;
-   }
-
-   /* XXX: do we really care?
-    */
-   if (brw->curr.min_index != min_index ||
-       brw->curr.max_index != max_index) 
+   if (brw->curr.min_index != info->min_index ||
+       brw->curr.max_index != info->max_index) 
    { 
-      brw->curr.min_index = min_index;
-      brw->curr.max_index = max_index;
+      brw->curr.min_index = info->min_index;
+      brw->curr.max_index = info->max_index;
       brw->state.dirty.mesa |= PIPE_NEW_INDEX_RANGE;
    }
 
 
    /* Make a first attempt at drawing:
     */
-   ret = try_draw_range_elements(brw, index_buffer, hw_prim, start, count );
+   ret = try_draw_range_elements(brw, info->indexed,
+         hw_prim, info->start, info->count);
 
    /* Otherwise, flush and retry:
     */
    if (ret != 0) {
       brw_context_flush( brw );
-      ret = try_draw_range_elements(brw, index_buffer, hw_prim, start, count );
+      ret = try_draw_range_elements(brw, info->indexed,
+            hw_prim, info->start, info->count);
       assert(ret == 0);
    }
 }
 
-static void
-brw_draw_elements(struct pipe_context *pipe,
-		  struct pipe_resource *index_buffer,
-		  unsigned index_size, int index_bias,
-		  unsigned mode, 
-		  unsigned start, unsigned count)
-{
-   brw_draw_range_elements( pipe, index_buffer,
-                            index_size, index_bias,
-                            0, 0xffffffff,
-                            mode, 
-                            start, count );
-}
-
-static void
-brw_draw_arrays(struct pipe_context *pipe, unsigned mode,
-                     unsigned start, unsigned count)
-{
-   brw_draw_elements(pipe, NULL, 0, 0, mode, start, count);
-}
-
-
 
 boolean brw_draw_init( struct brw_context *brw )
 {
    /* Register our drawing function: 
     */
-   brw->base.draw_arrays = brw_draw_arrays;
-   brw->base.draw_elements = brw_draw_elements;
-   brw->base.draw_range_elements = brw_draw_range_elements;
+   brw->base.draw_vbo = brw_draw_vbo;
 
    /* Create helpers for uploading data in user buffers:
     */
diff --git a/src/gallium/drivers/i965/brw_draw_upload.c b/src/gallium/drivers/i965/brw_draw_upload.c
index 337eee8cd9c..ebeb1e146aa 100644
--- a/src/gallium/drivers/i965/brw_draw_upload.c
+++ b/src/gallium/drivers/i965/brw_draw_upload.c
@@ -231,7 +231,7 @@ static int brw_prepare_indices(struct brw_context *brw)
    struct pipe_resource *upload_buf = NULL;
    struct brw_winsys_buffer *bo = NULL;
    GLuint offset;
-   GLuint index_size;
+   GLuint index_size, index_offset;
    GLuint ib_size;
    int ret;
 
@@ -246,13 +246,14 @@ static int brw_prepare_indices(struct brw_context *brw)
 
    ib_size = index_buffer->width0;
    index_size = brw->curr.index_size;
+   index_offset = brw->curr.index_offset;
 
    /* Turn userbuffer into a proper hardware buffer?
     */
    if (brw_buffer_is_user_buffer(index_buffer)) {
 
       ret = u_upload_buffer( brw->vb.upload_index,
-			     0,
+			     index_offset,
 			     ib_size,
 			     index_buffer,
 			     &offset,
@@ -269,7 +270,7 @@ static int brw_prepare_indices(struct brw_context *brw)
    else {
       bo = brw_buffer(index_buffer)->bo;
       ib_size = bo->size;
-      offset = 0;
+      offset = index_offset;
    }
 
    /* Use CMD_3D_PRIM's start_vertex_offset to avoid re-uploading the
diff --git a/src/gallium/drivers/i965/brw_pipe_vertex.c b/src/gallium/drivers/i965/brw_pipe_vertex.c
index 4a120a51dad..007239efc40 100644
--- a/src/gallium/drivers/i965/brw_pipe_vertex.c
+++ b/src/gallium/drivers/i965/brw_pipe_vertex.c
@@ -274,10 +274,41 @@ static void brw_set_vertex_buffers(struct pipe_context *pipe,
 }
 
 
+static void brw_set_index_buffer(struct pipe_context *pipe,
+                                 const struct pipe_index_buffer *ib)
+{
+   struct brw_context *brw = brw_context(pipe);
+
+   if (ib) {
+      if (brw->curr.index_buffer == ib->buffer &&
+          brw->curr.index_offset == ib->offset &&
+          brw->curr.index_size == ib->index_size)
+         return;
+
+      pipe_resource_reference(&brw->curr.index_buffer, ib->buffer);
+      brw->curr.index_offset = ib->offset;
+      brw->curr.index_size = ib->index_size;
+   }
+   else {
+      if (!brw->curr.index_buffer &&
+          !brw->curr.index_offset &&
+          !brw->curr.index_size)
+         return;
+
+      pipe_resource_reference(&brw->curr.index_buffer, NULL);
+      brw->curr.index_offset = 0;
+      brw->curr.index_size = 0;
+   }
+
+   brw->state.dirty.mesa |= PIPE_NEW_INDEX_BUFFER;
+}
+
+
 void 
 brw_pipe_vertex_init( struct brw_context *brw )
 {
    brw->base.set_vertex_buffers = brw_set_vertex_buffers;
+   brw->base.set_index_buffer = brw_set_index_buffer;
    brw->base.create_vertex_elements_state = brw_create_vertex_elements_state;
    brw->base.bind_vertex_elements_state = brw_bind_vertex_elements_state;
    brw->base.delete_vertex_elements_state = brw_delete_vertex_elements_state;
diff --git a/src/gallium/drivers/identity/id_context.c b/src/gallium/drivers/identity/id_context.c
index 67be895b385..de83c249057 100644
--- a/src/gallium/drivers/identity/id_context.c
+++ b/src/gallium/drivers/identity/id_context.c
@@ -46,68 +46,13 @@ identity_destroy(struct pipe_context *_pipe)
 }
 
 static void
-identity_draw_arrays(struct pipe_context *_pipe,
-                     unsigned prim,
-                     unsigned start,
-                     unsigned count)
+identity_draw_vbo(struct pipe_context *_pipe,
+                  const struct pipe_draw_info *info)
 {
    struct identity_context *id_pipe = identity_context(_pipe);
    struct pipe_context *pipe = id_pipe->pipe;
 
-   pipe->draw_arrays(pipe,
-                     prim,
-                     start,
-                     count);
-}
-
-static void
-identity_draw_elements(struct pipe_context *_pipe,
-                       struct pipe_resource *_indexResource,
-                       unsigned indexSize,
-                       int indexBias,
-                       unsigned prim,
-                       unsigned start,
-                       unsigned count)
-{
-   struct identity_context *id_pipe = identity_context(_pipe);
-   struct identity_resource *id_resource = identity_resource(_indexResource);
-   struct pipe_context *pipe = id_pipe->pipe;
-   struct pipe_resource *indexResource = id_resource->resource;
-
-   pipe->draw_elements(pipe,
-                       indexResource,
-                       indexSize,
-                       indexBias,
-                       prim,
-                       start,
-                       count);
-}
-
-static void
-identity_draw_range_elements(struct pipe_context *_pipe,
-                             struct pipe_resource *_indexResource,
-                             unsigned indexSize,
-                             int indexBias,
-                             unsigned minIndex,
-                             unsigned maxIndex,
-                             unsigned mode,
-                             unsigned start,
-                             unsigned count)
-{
-   struct identity_context *id_pipe = identity_context(_pipe);
-   struct identity_resource *id_resource = identity_resource(_indexResource);
-   struct pipe_context *pipe = id_pipe->pipe;
-   struct pipe_resource *indexResource = id_resource->resource;
-
-   pipe->draw_range_elements(pipe,
-                             indexResource,
-                             indexSize,
-                             indexBias,
-                             minIndex,
-                             maxIndex,
-                             mode,
-                             start,
-                             count);
+   pipe->draw_vbo(pipe, info);
 }
 
 static struct pipe_query *
@@ -611,6 +556,24 @@ identity_set_vertex_buffers(struct pipe_context *_pipe,
                             num_buffers,
                             buffers);
 }
+
+static void
+identity_set_index_buffer(struct pipe_context *_pipe,
+                          const struct pipe_index_buffer *_ib)
+{
+   struct identity_context *id_pipe = identity_context(_pipe);
+   struct pipe_context *pipe = id_pipe->pipe;
+   struct pipe_index_buffer unwrapped_ib, *ib = NULL;
+
+   if (_ib) {
+      unwrapped_ib = *_ib;
+      unwrapped_ib.buffer = identity_resource_unwrap(_ib->buffer);
+      ib = &unwrapped_ib;
+   }
+
+   pipe->set_index_buffer(pipe, ib);
+}
+
 static void
 identity_resource_copy_region(struct pipe_context *_pipe,
                               struct pipe_resource *_dst,
@@ -889,9 +852,7 @@ identity_context_create(struct pipe_screen *_screen, struct pipe_context *pipe)
    id_pipe->base.draw = NULL;
 
    id_pipe->base.destroy = identity_destroy;
-   id_pipe->base.draw_arrays = identity_draw_arrays;
-   id_pipe->base.draw_elements = identity_draw_elements;
-   id_pipe->base.draw_range_elements = identity_draw_range_elements;
+   id_pipe->base.draw_vbo = identity_draw_vbo;
    id_pipe->base.create_query = identity_create_query;
    id_pipe->base.destroy_query = identity_destroy_query;
    id_pipe->base.begin_query = identity_begin_query;
@@ -931,6 +892,7 @@ identity_context_create(struct pipe_screen *_screen, struct pipe_context *pipe)
    id_pipe->base.set_fragment_sampler_views = identity_set_fragment_sampler_views;
    id_pipe->base.set_vertex_sampler_views = identity_set_vertex_sampler_views;
    id_pipe->base.set_vertex_buffers = identity_set_vertex_buffers;
+   id_pipe->base.set_index_buffer = identity_set_index_buffer;
    id_pipe->base.resource_copy_region = identity_resource_copy_region;
    id_pipe->base.clear = identity_clear;
    id_pipe->base.clear_render_target = identity_clear_render_target;
diff --git a/src/gallium/drivers/llvmpipe/SConscript b/src/gallium/drivers/llvmpipe/SConscript
index fd6ba1561ea..5583fca38e6 100644
--- a/src/gallium/drivers/llvmpipe/SConscript
+++ b/src/gallium/drivers/llvmpipe/SConscript
@@ -1,3 +1,4 @@
+from sys import executable as python_cmd
 import distutils.version
 
 Import('*')
@@ -16,7 +17,7 @@ env.CodeGenerate(
 	target = 'lp_tile_soa.c',
 	script = 'lp_tile_soa.py',
 	source = ['#src/gallium/auxiliary/util/u_format.csv'],
-	command = 'python $SCRIPT $SOURCE > $TARGET'
+	command = python_cmd + ' $SCRIPT $SOURCE > $TARGET'
 )
 
 # XXX: Our dependency scanner only finds depended modules in relative dirs.
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_depth.c b/src/gallium/drivers/llvmpipe/lp_bld_depth.c
index e05bbe5011a..99a768afd80 100644
--- a/src/gallium/drivers/llvmpipe/lp_bld_depth.c
+++ b/src/gallium/drivers/llvmpipe/lp_bld_depth.c
@@ -258,16 +258,16 @@ lp_build_stencil_op_single(struct lp_build_context *bld,
    }
 
    if (stencil->writemask != stencilMax) {
-      /* compute res = (res & mask) | (stencilVals & ~mask) */
-      LLVMValueRef mask = lp_build_const_int_vec(type, stencil->writemask);
-      LLVMValueRef cmask = LLVMBuildNot(bld->builder, mask, "notWritemask");
-      LLVMValueRef t1 = LLVMBuildAnd(bld->builder, res, mask, "t1");
-      LLVMValueRef t2 = LLVMBuildAnd(bld->builder, stencilVals, cmask, "t2");
-      res = LLVMBuildOr(bld->builder, t1, t2, "t1_or_t2");
+      /* mask &= stencil->writemask */
+      LLVMValueRef writemask = lp_build_const_int_vec(type, stencil->writemask);
+      mask = LLVMBuildAnd(bld->builder, mask, writemask, "");
+      /* res = (res & mask) | (stencilVals & ~mask) */
+      res = lp_build_select_bitwise(bld, writemask, res, stencilVals);
+   }
+   else {
+      /* res = mask ? res : stencilVals */
+      res = lp_build_select(bld, mask, res, stencilVals);
    }
-
-   /* only the update the vector elements enabled by 'mask' */
-   res = lp_build_select(bld, mask, res, stencilVals);
 
    return res;
 }
@@ -662,9 +662,9 @@ lp_build_depth_stencil_test(LLVMBuilderRef builder,
          }
 
          /* Mix the old and new Z buffer values.
-          * z_dst[i] = zselectmask[i] ? z_src[i] : z_dst[i]
+          * z_dst[i] = (zselectmask[i] & z_src[i]) | (~zselectmask[i] & z_dst[i])
           */
-         z_dst = lp_build_select(&bld, zselectmask, z_src, z_dst);
+         z_dst = lp_build_select_bitwise(&bld, zselectmask, z_src, z_dst);
       }
 
       if (stencil[0].enabled) {
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_interp.c b/src/gallium/drivers/llvmpipe/lp_bld_interp.c
index 78744da500b..2cf6f38c4b8 100644
--- a/src/gallium/drivers/llvmpipe/lp_bld_interp.c
+++ b/src/gallium/drivers/llvmpipe/lp_bld_interp.c
@@ -141,7 +141,7 @@ coeffs_init(struct lp_build_interp_soa_context *bld,
                else {
                   dadx = LLVMBuildLoad(builder, LLVMBuildGEP(builder, dadx_ptr, &index, 1, ""), "");
                   dady = LLVMBuildLoad(builder, LLVMBuildGEP(builder, dady_ptr, &index, 1, ""), "");
-                  dadxy = LLVMBuildAdd(builder, dadx, dady, "");
+                  dadxy = LLVMBuildFAdd(builder, dadx, dady, "");
                   attrib_name(dadx, attrib, chan, ".dadx");
                   attrib_name(dady, attrib, chan, ".dady");
                   attrib_name(dadxy, attrib, chan, ".dadxy");
@@ -177,7 +177,7 @@ coeffs_init(struct lp_build_interp_soa_context *bld,
              * dadq2 = 2 * dq
              */
 
-            dadq2 = LLVMBuildAdd(builder, dadq, dadq, "");
+            dadq2 = LLVMBuildFAdd(builder, dadq, dadq, "");
 
             /*
              * a = a0 + x * dadx + y * dady
@@ -193,12 +193,11 @@ coeffs_init(struct lp_build_interp_soa_context *bld,
                a = a0;
                if (interp != LP_INTERP_CONSTANT &&
                    interp != LP_INTERP_FACING) {
-                  a = LLVMBuildAdd(builder, a,
-                                   LLVMBuildMul(builder, bld->x, dadx, ""),
-                                   "");
-                  a = LLVMBuildAdd(builder, a,
-                                   LLVMBuildMul(builder, bld->y, dady, ""),
-                                   "");
+                  LLVMValueRef tmp;
+                  tmp = LLVMBuildFMul(builder, bld->x, dadx, "");
+                  a = LLVMBuildFAdd(builder, a, tmp, "");
+                  tmp = LLVMBuildFMul(builder, bld->y, dady, "");
+                  a = LLVMBuildFAdd(builder, a, tmp, "");
                }
             }
 
@@ -212,7 +211,7 @@ coeffs_init(struct lp_build_interp_soa_context *bld,
              * Compute the attrib values on the upper-left corner of each quad.
              */
 
-            a = LLVMBuildAdd(builder, a, dadq2, "");
+            a = LLVMBuildFAdd(builder, a, dadq2, "");
 
             /*
              * a    *= 1 / w
diff --git a/src/gallium/drivers/llvmpipe/lp_context.c b/src/gallium/drivers/llvmpipe/lp_context.c
index 3db4f12ebb6..7543bd7b2b0 100644
--- a/src/gallium/drivers/llvmpipe/lp_context.c
+++ b/src/gallium/drivers/llvmpipe/lp_context.c
@@ -46,6 +46,10 @@
 #include "lp_query.h"
 #include "lp_setup.h"
 
+
+DEBUG_GET_ONCE_BOOL_OPTION(lp_no_rast, "LP_NO_RAST", FALSE)
+
+
 static void llvmpipe_destroy( struct pipe_context *pipe )
 {
    struct llvmpipe_context *llvmpipe = llvmpipe_context( pipe );
@@ -130,7 +134,7 @@ llvmpipe_create_context( struct pipe_screen *screen, void *priv )
 
    /* FIXME: devise alternative to draw_texture_samplers */
 
-   if (debug_get_bool_option( "LP_NO_RAST", FALSE ))
+   if (debug_get_option_lp_no_rast())
       llvmpipe->no_rast = TRUE;
 
    llvmpipe->setup = lp_setup_create( &llvmpipe->pipe,
diff --git a/src/gallium/drivers/llvmpipe/lp_context.h b/src/gallium/drivers/llvmpipe/lp_context.h
index b2643ab33cd..50f9091c3ca 100644
--- a/src/gallium/drivers/llvmpipe/lp_context.h
+++ b/src/gallium/drivers/llvmpipe/lp_context.h
@@ -77,6 +77,7 @@ struct llvmpipe_context {
    struct pipe_sampler_view *vertex_sampler_views[PIPE_MAX_VERTEX_SAMPLERS];
    struct pipe_viewport_state viewport;
    struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS];
+   struct pipe_index_buffer index_buffer;
    struct {
       struct llvmpipe_resource *buffer[PIPE_MAX_SO_BUFFERS];
       int offset[PIPE_MAX_SO_BUFFERS];
diff --git a/src/gallium/drivers/llvmpipe/lp_draw_arrays.c b/src/gallium/drivers/llvmpipe/lp_draw_arrays.c
index 625d0c8a8c9..e73b431cb4d 100644
--- a/src/gallium/drivers/llvmpipe/lp_draw_arrays.c
+++ b/src/gallium/drivers/llvmpipe/lp_draw_arrays.c
@@ -49,20 +49,11 @@
  * the drawing to the 'draw' module.
  */
 static void
-llvmpipe_draw_range_elements_instanced(struct pipe_context *pipe,
-                                       struct pipe_resource *indexBuffer,
-                                       unsigned indexSize,
-                                       int indexBias,
-                                       unsigned minIndex,
-                                       unsigned maxIndex,
-                                       unsigned mode,
-                                       unsigned start,
-                                       unsigned count,
-                                       unsigned startInstance,
-                                       unsigned instanceCount)
+llvmpipe_draw_vbo(struct pipe_context *pipe, const struct pipe_draw_info *info)
 {
    struct llvmpipe_context *lp = llvmpipe_context(pipe);
    struct draw_context *draw = lp->draw;
+   void *mapped_indices = NULL;
    unsigned i;
 
    if (lp->dirty)
@@ -77,27 +68,25 @@ llvmpipe_draw_range_elements_instanced(struct pipe_context *pipe,
    }
 
    /* Map index buffer, if present */
-   if (indexBuffer) {
-      void *mapped_indexes = llvmpipe_resource_data(indexBuffer);
-      draw_set_mapped_element_buffer_range(draw,
-                                           indexSize,
-                                           indexBias,
-                                           minIndex,
-                                           maxIndex,
-                                           mapped_indexes);
-   }
-   else {
-      /* no index/element buffer */
-      draw_set_mapped_element_buffer_range(draw, 0, 0, start,
-                                           start + count - 1, NULL);
+   if (info->indexed && lp->index_buffer.buffer) {
+      char *indices = (char *) llvmpipe_resource_data(lp->index_buffer.buffer);
+      mapped_indices = (void *) (indices + lp->index_buffer.offset);
    }
+
+   draw_set_mapped_element_buffer_range(draw, (mapped_indices) ?
+                                        lp->index_buffer.index_size : 0,
+                                        info->index_bias,
+                                        info->min_index,
+                                        info->max_index,
+                                        mapped_indices);
+
    llvmpipe_prepare_vertex_sampling(lp,
                                     lp->num_vertex_sampler_views,
                                     lp->vertex_sampler_views);
 
    /* draw! */
-   draw_arrays_instanced(draw, mode, start, count,
-                         startInstance, instanceCount);
+   draw_arrays_instanced(draw, info->mode, info->start, info->count,
+         info->start_instance, info->instance_count);
 
    /*
     * unmap vertex/index buffers
@@ -105,7 +94,7 @@ llvmpipe_draw_range_elements_instanced(struct pipe_context *pipe,
    for (i = 0; i < lp->num_vertex_buffers; i++) {
       draw_set_mapped_vertex_buffer(draw, i, NULL);
    }
-   if (indexBuffer) {
+   if (mapped_indices) {
       draw_set_mapped_element_buffer(draw, 0, 0, NULL);
    }
    llvmpipe_cleanup_vertex_sampling(lp);
@@ -119,112 +108,8 @@ llvmpipe_draw_range_elements_instanced(struct pipe_context *pipe,
 }
 
 
-static void
-llvmpipe_draw_arrays_instanced(struct pipe_context *pipe,
-                               unsigned mode,
-                               unsigned start,
-                               unsigned count,
-                               unsigned startInstance,
-                               unsigned instanceCount)
-{
-   llvmpipe_draw_range_elements_instanced(pipe,
-                                          NULL, /* no indexBuffer */
-                                          0, 0, /* indexSize, indexBias */
-                                          0, ~0, /* minIndex, maxIndex */
-                                          mode,
-                                          start,
-                                          count,
-                                          startInstance,
-                                          instanceCount);
-}
-
-
-static void
-llvmpipe_draw_elements_instanced(struct pipe_context *pipe,
-                                 struct pipe_resource *indexBuffer,
-                                 unsigned indexSize,
-                                 int indexBias,
-                                 unsigned mode,
-                                 unsigned start,
-                                 unsigned count,
-                                 unsigned startInstance,
-                                 unsigned instanceCount)
-{
-   llvmpipe_draw_range_elements_instanced(pipe,
-                                          indexBuffer,
-                                          indexSize, indexBias,
-                                          0, ~0, /* minIndex, maxIndex */
-                                          mode,
-                                          start,
-                                          count,
-                                          startInstance,
-                                          instanceCount);
-}
-
-
-static void
-llvmpipe_draw_elements(struct pipe_context *pipe,
-                       struct pipe_resource *indexBuffer,
-                       unsigned indexSize,
-                       int indexBias,
-                       unsigned mode,
-                       unsigned start,
-                       unsigned count)
-{
-   llvmpipe_draw_range_elements_instanced(pipe,
-                                          indexBuffer,
-                                          indexSize, indexBias,
-                                          0, 0xffffffff,  /* min, maxIndex */
-                                          mode, start, count,
-                                          0,  /* startInstance */
-                                          1);  /* instanceCount */
-}
-
-
-static void
-llvmpipe_draw_range_elements(struct pipe_context *pipe,
-                             struct pipe_resource *indexBuffer,
-                             unsigned indexSize,
-                             int indexBias,
-                             unsigned min_index,
-                             unsigned max_index,
-                             unsigned mode,
-                             unsigned start,
-                             unsigned count)
-{
-   llvmpipe_draw_range_elements_instanced(pipe,
-                                          indexBuffer,
-                                          indexSize, indexBias,
-                                          min_index, max_index,
-                                          mode, start, count,
-                                          0,  /* startInstance */
-                                          1);  /* instanceCount */
-}
-
-
-static void
-llvmpipe_draw_arrays(struct pipe_context *pipe,
-                     unsigned mode,
-                     unsigned start,
-                     unsigned count)
-{
-   llvmpipe_draw_range_elements_instanced(pipe,
-                                          NULL,  /* indexBuffer */
-                                          0,  /* indexSize */
-                                          0,  /* indexBias */
-                                          0, ~0,  /* min, maxIndex */
-                                          mode, start, count,
-                                          0,  /* startInstance */
-                                          1);  /* instanceCount */
-}
-
-
 void
 llvmpipe_init_draw_funcs(struct llvmpipe_context *llvmpipe)
 {
-   llvmpipe->pipe.draw_arrays = llvmpipe_draw_arrays;
-   llvmpipe->pipe.draw_elements = llvmpipe_draw_elements;
-   llvmpipe->pipe.draw_range_elements = llvmpipe_draw_range_elements;
-   llvmpipe->pipe.draw_arrays_instanced = llvmpipe_draw_arrays_instanced;
-   llvmpipe->pipe.draw_elements_instanced = llvmpipe_draw_elements_instanced;
+   llvmpipe->pipe.draw_vbo = llvmpipe_draw_vbo;
 }
diff --git a/src/gallium/drivers/llvmpipe/lp_rast.c b/src/gallium/drivers/llvmpipe/lp_rast.c
index 654f4ea48eb..3215d0f6525 100644
--- a/src/gallium/drivers/llvmpipe/lp_rast.c
+++ b/src/gallium/drivers/llvmpipe/lp_rast.c
@@ -43,6 +43,12 @@
 #include "lp_scene.h"
 
 
+#ifdef DEBUG
+int jit_line = 0;
+const struct lp_rast_state *jit_state = NULL;
+#endif
+
+
 /**
  * Begin rasterizing a scene.
  * Called once per scene by one thread.
@@ -368,14 +374,15 @@ lp_rast_store_linear_color( struct lp_rasterizer_task *task,
 
    for (buf = 0; buf < rast->state.nr_cbufs; buf++) {
       struct pipe_surface *cbuf = scene->fb.cbufs[buf];
-      const unsigned face = cbuf->face, level = cbuf->level;
+      const unsigned face_slice = cbuf->face + cbuf->zslice;
+      const unsigned level = cbuf->level;
       struct llvmpipe_resource *lpt = llvmpipe_resource(cbuf->texture);
 
       if (!task->color_tiles[buf])
          continue;
 
       llvmpipe_unswizzle_cbuf_tile(lpt,
-                                   face,
+                                   face_slice,
                                    level,
                                    task->x, task->y,
                                    task->color_tiles[buf]);
@@ -418,6 +425,7 @@ lp_rast_shade_tile(struct lp_rasterizer_task *task,
          depth = lp_rast_get_depth_block_pointer(task, tile_x + x, tile_y + y);
 
          /* run shader on 4x4 block */
+         BEGIN_JIT_CALL(state);
          variant->jit_function[RAST_WHOLE]( &state->jit_context,
                                             tile_x + x, tile_y + y,
                                             inputs->facing,
@@ -428,6 +436,7 @@ lp_rast_shade_tile(struct lp_rasterizer_task *task,
                                             depth,
                                             0xffff,
                                             &task->vis_counter);
+         END_JIT_CALL();
       }
    }
 }
@@ -497,6 +506,7 @@ lp_rast_shade_quads_mask(struct lp_rasterizer_task *task,
    assert(lp_check_alignment(state->jit_context.blend_color, 16));
 
    /* run shader on 4x4 block */
+   BEGIN_JIT_CALL(state);
    variant->jit_function[RAST_EDGE_TEST](&state->jit_context,
                                          x, y,
                                          inputs->facing,
@@ -507,6 +517,7 @@ lp_rast_shade_quads_mask(struct lp_rasterizer_task *task,
                                          depth,
                                          mask,
                                          &task->vis_counter);
+   END_JIT_CALL();
 }
 
 
diff --git a/src/gallium/drivers/llvmpipe/lp_rast.h b/src/gallium/drivers/llvmpipe/lp_rast.h
index eaf2a6f3345..44319a0ad6f 100644
--- a/src/gallium/drivers/llvmpipe/lp_rast.h
+++ b/src/gallium/drivers/llvmpipe/lp_rast.h
@@ -104,9 +104,6 @@ struct lp_rast_plane {
 
    int dcdx;
    int dcdy;
-   
-   /* edge/step info for 3 edges and 4x4 block of pixels */
-   const int *step;
 };
 
 /**
@@ -119,8 +116,6 @@ struct lp_rast_triangle {
    /* inputs for the shader */
    struct lp_rast_shader_inputs inputs;
 
-   int step[3][16];
-
 #ifdef DEBUG
    float v[3][2];
 #endif
diff --git a/src/gallium/drivers/llvmpipe/lp_rast_priv.h b/src/gallium/drivers/llvmpipe/lp_rast_priv.h
index b4a48cfd024..fae7f6d3dc2 100644
--- a/src/gallium/drivers/llvmpipe/lp_rast_priv.h
+++ b/src/gallium/drivers/llvmpipe/lp_rast_priv.h
@@ -40,6 +40,34 @@
 #include "lp_limits.h"
 
 
+/* If we crash in a jitted function, we can examine jit_line and jit_state
+ * to get some info.  This is not thread-safe, however.
+ */
+#ifdef DEBUG
+
+extern int jit_line;
+extern const struct lp_rast_state *jit_state;
+
+#define BEGIN_JIT_CALL(state) \
+   do { \
+      jit_line = __LINE__; \
+      jit_state = state; \
+   } while (0)
+
+#define END_JIT_CALL() \
+   do { \
+      jit_line = 0; \
+      jit_state = NULL; \
+   } while (0)
+
+#else
+
+#define BEGIN_JIT_CALL(X)
+#define END_JIT_CALL()
+
+#endif
+
+
 struct lp_rasterizer;
 
 
@@ -249,6 +277,7 @@ lp_rast_shade_quads_all( struct lp_rasterizer_task *task,
    depth = lp_rast_get_depth_block_pointer(task, x, y);
 
    /* run shader on 4x4 block */
+   BEGIN_JIT_CALL(state);
    variant->jit_function[RAST_WHOLE]( &state->jit_context,
                                       x, y,
                                       inputs->facing,
@@ -259,6 +288,7 @@ lp_rast_shade_quads_all( struct lp_rasterizer_task *task,
                                       depth,
                                       0xffff,
                                       &task->vis_counter );
+   END_JIT_CALL();
 }
 
 
diff --git a/src/gallium/drivers/llvmpipe/lp_rast_tri.c b/src/gallium/drivers/llvmpipe/lp_rast_tri.c
index ebe9a8e92b4..980c18c0240 100644
--- a/src/gallium/drivers/llvmpipe/lp_rast_tri.c
+++ b/src/gallium/drivers/llvmpipe/lp_rast_tri.c
@@ -37,52 +37,6 @@
 #include "lp_tile_soa.h"
 
 
-/**
- * Map an index in [0,15] to an x,y position, multiplied by 4.
- * This is used to get the position of each subtile in a 4x4
- * grid of edge step values.
- * Note: we can use some bit twiddling to compute these values instead
- * of using a look-up table, but there's no measurable performance
- * difference.
- */
-static const int pos_table4[16][2] = {
-   { 0, 0 },
-   { 4, 0 },
-   { 0, 4 },
-   { 4, 4 },
-   { 8, 0 },
-   { 12, 0 },
-   { 8, 4 },
-   { 12, 4 },
-   { 0, 8 },
-   { 4, 8 },
-   { 0, 12 },
-   { 4, 12 },
-   { 8, 8 },
-   { 12, 8 },
-   { 8, 12 },
-   { 12, 12 }
-};
-
-
-static const int pos_table16[16][2] = {
-   { 0, 0 },
-   { 16, 0 },
-   { 0, 16 },
-   { 16, 16 },
-   { 32, 0 },
-   { 48, 0 },
-   { 32, 16 },
-   { 48, 16 },
-   { 0, 32 },
-   { 16, 32 },
-   { 0, 48 },
-   { 16, 48 },
-   { 32, 32 },
-   { 48, 32 },
-   { 32, 48 },
-   { 48, 48 }
-};
 
 
 /**
@@ -113,6 +67,68 @@ block_full_16(struct lp_rasterizer_task *task,
 	 block_full_4(task, tri, x + ix, y + iy);
 }
 
+
+static INLINE unsigned
+build_mask(int c, int dcdx, int dcdy)
+{
+   int mask = 0;
+
+   int c0 = c;
+   int c1 = c0 + dcdx;
+   int c2 = c1 + dcdx;
+   int c3 = c2 + dcdx;
+
+   mask |= ((c0 + 0 * dcdy) >> 31) & (1 << 0);
+   mask |= ((c0 + 1 * dcdy) >> 31) & (1 << 2);
+   mask |= ((c0 + 2 * dcdy) >> 31) & (1 << 8);
+   mask |= ((c0 + 3 * dcdy) >> 31) & (1 << 10);
+   mask |= ((c1 + 0 * dcdy) >> 31) & (1 << 1);
+   mask |= ((c1 + 1 * dcdy) >> 31) & (1 << 3);
+   mask |= ((c1 + 2 * dcdy) >> 31) & (1 << 9);
+   mask |= ((c1 + 3 * dcdy) >> 31) & (1 << 11); 
+   mask |= ((c2 + 0 * dcdy) >> 31) & (1 << 4);
+   mask |= ((c2 + 1 * dcdy) >> 31) & (1 << 6);
+   mask |= ((c2 + 2 * dcdy) >> 31) & (1 << 12);
+   mask |= ((c2 + 3 * dcdy) >> 31) & (1 << 14);
+   mask |= ((c3 + 0 * dcdy) >> 31) & (1 << 5);
+   mask |= ((c3 + 1 * dcdy) >> 31) & (1 << 7);
+   mask |= ((c3 + 2 * dcdy) >> 31) & (1 << 13);
+   mask |= ((c3 + 3 * dcdy) >> 31) & (1 << 15);
+  
+   return mask;
+}
+
+static INLINE unsigned
+build_mask_linear(int c, int dcdx, int dcdy)
+{
+   int mask = 0;
+
+   int c0 = c;
+   int c1 = c0 + dcdy;
+   int c2 = c1 + dcdy;
+   int c3 = c2 + dcdy;
+
+   mask |= ((c0 + 0 * dcdx) >> 31) & (1 << 0);
+   mask |= ((c0 + 1 * dcdx) >> 31) & (1 << 1);
+   mask |= ((c0 + 2 * dcdx) >> 31) & (1 << 2);
+   mask |= ((c0 + 3 * dcdx) >> 31) & (1 << 3);
+   mask |= ((c1 + 0 * dcdx) >> 31) & (1 << 4);
+   mask |= ((c1 + 1 * dcdx) >> 31) & (1 << 5);
+   mask |= ((c1 + 2 * dcdx) >> 31) & (1 << 6);
+   mask |= ((c1 + 3 * dcdx) >> 31) & (1 << 7); 
+   mask |= ((c2 + 0 * dcdx) >> 31) & (1 << 8);
+   mask |= ((c2 + 1 * dcdx) >> 31) & (1 << 9);
+   mask |= ((c2 + 2 * dcdx) >> 31) & (1 << 10);
+   mask |= ((c2 + 3 * dcdx) >> 31) & (1 << 11);
+   mask |= ((c3 + 0 * dcdx) >> 31) & (1 << 12);
+   mask |= ((c3 + 1 * dcdx) >> 31) & (1 << 13);
+   mask |= ((c3 + 2 * dcdx) >> 31) & (1 << 14);
+   mask |= ((c3 + 3 * dcdx) >> 31) & (1 << 15);
+  
+   return mask;
+}
+
+
 #define TAG(x) x##_1
 #define NR_PLANES 1
 #include "lp_rast_tri_tmp.h"
diff --git a/src/gallium/drivers/llvmpipe/lp_rast_tri_tmp.h b/src/gallium/drivers/llvmpipe/lp_rast_tri_tmp.h
index a410c611a3f..43f72d8ca8f 100644
--- a/src/gallium/drivers/llvmpipe/lp_rast_tri_tmp.h
+++ b/src/gallium/drivers/llvmpipe/lp_rast_tri_tmp.h
@@ -46,19 +46,13 @@ TAG(do_block_4)(struct lp_rasterizer_task *task,
                 int x, int y,
                 const int *c)
 {
-   unsigned mask = 0;
-   int i;
+   unsigned mask = 0xffff;
+   int j;
 
-   for (i = 0; i < 16; i++) {
-      int any_negative = 0;
-      int j;
-
-      for (j = 0; j < NR_PLANES; j++) 
-         any_negative |= (c[j] - 1 + plane[j].step[i]);
-         
-      any_negative >>= 31;
-
-      mask |= (~any_negative) & (1 << i);
+   for (j = 0; j < NR_PLANES; j++) {
+      mask &= ~build_mask(c[j] - 1, 
+			  -plane[j].dcdx,
+			  plane[j].dcdy);
    }
 
    /* Now pass to the shader:
@@ -79,24 +73,19 @@ TAG(do_block_16)(struct lp_rasterizer_task *task,
                  const int *c)
 {
    unsigned outmask, inmask, partmask, partial_mask;
-   unsigned i, j;
+   unsigned j;
 
    outmask = 0;                 /* outside one or more trivial reject planes */
    partmask = 0;                /* outside one or more trivial accept planes */
 
    for (j = 0; j < NR_PLANES; j++) {
-      const int *step = plane[j].step;
-      const int eo = plane[j].eo * 4;
-      const int ei = plane[j].ei * 4;
-      const int cox = c[j] + eo;
-      const int cio = ei - 1 - eo;
-
-      for (i = 0; i < 16; i++) {
-         int out = cox + step[i] * 4;
-         int part = out + cio;
-         outmask  |= (out >> 31) & (1 << i);
-         partmask |= (part >> 31) & (1 << i);
-      }
+      const int dcdx = -plane[j].dcdx * 4;
+      const int dcdy = plane[j].dcdy * 4;
+      const int cox = c[j] + plane[j].eo * 4;
+      const int cio = c[j] + plane[j].ei * 4 - 1;
+
+      outmask |= build_mask_linear(cox, dcdx, dcdy);
+      partmask |= build_mask_linear(cio, dcdx, dcdy);
    }
 
    if (outmask == 0xffff)
@@ -117,15 +106,19 @@ TAG(do_block_16)(struct lp_rasterizer_task *task,
     */
    while (partial_mask) {
       int i = ffs(partial_mask) - 1;
-      int px = x + pos_table4[i][0];
-      int py = y + pos_table4[i][1];
+      int ix = (i & 3) * 4;
+      int iy = (i >> 2) * 4;
+      int px = x + ix;
+      int py = y + iy; 
       int cx[NR_PLANES];
 
-      for (j = 0; j < NR_PLANES; j++)
-         cx[j] = c[j] + plane[j].step[i] * 4;
-
       partial_mask &= ~(1 << i);
 
+      for (j = 0; j < NR_PLANES; j++)
+         cx[j] = (c[j] 
+		  - plane[j].dcdx * ix
+		  + plane[j].dcdy * iy);
+
       TAG(do_block_4)(task, tri, plane, px, py, cx);
    }
 
@@ -133,8 +126,10 @@ TAG(do_block_16)(struct lp_rasterizer_task *task,
     */
    while (inmask) {
       int i = ffs(inmask) - 1;
-      int px = x + pos_table4[i][0];
-      int py = y + pos_table4[i][1];
+      int ix = (i & 3) * 4;
+      int iy = (i >> 2) * 4;
+      int px = x + ix;
+      int py = y + iy; 
 
       inmask &= ~(1 << i);
 
@@ -157,35 +152,28 @@ TAG(lp_rast_triangle)(struct lp_rasterizer_task *task,
    struct lp_rast_plane plane[NR_PLANES];
    int c[NR_PLANES];
    unsigned outmask, inmask, partmask, partial_mask;
-   unsigned i, j, nr_planes = 0;
+   unsigned j = 0;
+
+   outmask = 0;                 /* outside one or more trivial reject planes */
+   partmask = 0;                /* outside one or more trivial accept planes */
 
    while (plane_mask) {
       int i = ffs(plane_mask) - 1;
-      plane[nr_planes] = tri->plane[i];
+      plane[j] = tri->plane[i];
       plane_mask &= ~(1 << i);
-      nr_planes++;
-   };
-
-   assert(nr_planes == NR_PLANES);
-   outmask = 0;                 /* outside one or more trivial reject planes */
-   partmask = 0;                /* outside one or more trivial accept planes */
+      c[j] = plane[j].c + plane[j].dcdy * y - plane[j].dcdx * x;
 
-   for (j = 0; j < NR_PLANES; j++) {
-      const int *step = plane[j].step;
-      const int eo = plane[j].eo * 16;
-      const int ei = plane[j].ei * 16;
-      int cox, cio;
+      {
+	 const int dcdx = -plane[j].dcdx * 16;
+	 const int dcdy = plane[j].dcdy * 16;
+	 const int cox = c[j] + plane[j].eo * 16;
+	 const int cio = c[j] + plane[j].ei * 16 - 1;
 
-      c[j] = plane[j].c + plane[j].dcdy * y - plane[j].dcdx * x;
-      cox = c[j] + eo;
-      cio = ei - 1 - eo;
-
-      for (i = 0; i < 16; i++) {
-         int out = cox + step[i] * 16;
-         int part = out + cio;
-         outmask  |= (out >> 31) & (1 << i);
-         partmask |= (part >> 31) & (1 << i);
+	 outmask |= build_mask_linear(cox, dcdx, dcdy);
+	 partmask |= build_mask_linear(cio, dcdx, dcdy);
       }
+
+      j++;
    }
 
    if (outmask == 0xffff)
@@ -206,12 +194,16 @@ TAG(lp_rast_triangle)(struct lp_rasterizer_task *task,
     */
    while (partial_mask) {
       int i = ffs(partial_mask) - 1;
-      int px = x + pos_table16[i][0];
-      int py = y + pos_table16[i][1];
+      int ix = (i & 3) * 16;
+      int iy = (i >> 2) * 16;
+      int px = x + ix;
+      int py = y + iy;
       int cx[NR_PLANES];
 
       for (j = 0; j < NR_PLANES; j++)
-         cx[j] = c[j] + plane[j].step[i] * 16;
+         cx[j] = (c[j]
+		  - plane[j].dcdx * ix
+		  + plane[j].dcdy * iy);
 
       partial_mask &= ~(1 << i);
 
@@ -223,8 +215,10 @@ TAG(lp_rast_triangle)(struct lp_rasterizer_task *task,
     */
    while (inmask) {
       int i = ffs(inmask) - 1;
-      int px = x + pos_table16[i][0];
-      int py = y + pos_table16[i][1];
+      int ix = (i & 3) * 16;
+      int iy = (i >> 2) * 16;
+      int px = x + ix;
+      int py = y + iy;
 
       inmask &= ~(1 << i);
 
diff --git a/src/gallium/drivers/llvmpipe/lp_setup_tri.c b/src/gallium/drivers/llvmpipe/lp_setup_tri.c
index 7e432503c12..393533ebee4 100644
--- a/src/gallium/drivers/llvmpipe/lp_setup_tri.c
+++ b/src/gallium/drivers/llvmpipe/lp_setup_tri.c
@@ -61,36 +61,6 @@ struct tri_info {
 
 
 
-static const int step_scissor_minx[16] = {
-   0, 1, 0, 1,
-   2, 3, 2, 3,
-   0, 1, 0, 1,
-   2, 3, 2, 3
-};
-
-static const int step_scissor_maxx[16] = {
-    0, -1,  0, -1,
-   -2, -3, -2, -3,
-    0, -1,  0, -1,
-   -2, -3, -2, -3
-};
-
-static const int step_scissor_miny[16] = {
-   0, 0, 1, 1,
-   0, 0, 1, 1,
-   2, 2, 3, 3,
-   2, 2, 3, 3
-};
-
-static const int step_scissor_maxy[16] = {
-    0,  0, -1, -1,
-    0,  0, -1, -1,
-   -2, -2, -3, -3,
-   -2, -2, -3, -3
-};
-
-
-
    
 static INLINE int
 subpixel_snap(float a)
@@ -260,13 +230,13 @@ static void setup_tri_coefficients( struct lp_setup_context *setup,
 {
    unsigned fragcoord_usage_mask = TGSI_WRITEMASK_XYZ;
    unsigned slot;
+   unsigned i;
 
    /* setup interpolation for all the remaining attributes:
     */
    for (slot = 0; slot < setup->fs.nr_inputs; slot++) {
       unsigned vert_attr = setup->fs.input[slot].src_index;
       unsigned usage_mask = setup->fs.input[slot].usage_mask;
-      unsigned i;
 
       switch (setup->fs.input[slot].interp) {
       case LP_INTERP_CONSTANT:
@@ -316,6 +286,34 @@ static void setup_tri_coefficients( struct lp_setup_context *setup,
    /* The internal position input is in slot zero:
     */
    setup_fragcoord_coef(tri, info, 0, fragcoord_usage_mask);
+
+   if (0) {
+      for (i = 0; i < NUM_CHANNELS; i++) {
+         float a0   = tri->inputs.a0  [0][i];
+         float dadx = tri->inputs.dadx[0][i];
+         float dady = tri->inputs.dady[0][i];
+
+         debug_printf("POS.%c: a0 = %f, dadx = %f, dady = %f\n",
+                      "xyzw"[i],
+                      a0, dadx, dady);
+      }
+
+      for (slot = 0; slot < setup->fs.nr_inputs; slot++) {
+         unsigned usage_mask = setup->fs.input[slot].usage_mask;
+         for (i = 0; i < NUM_CHANNELS; i++) {
+            if (usage_mask & (1 << i)) {
+               float a0   = tri->inputs.a0  [1 + slot][i];
+               float dadx = tri->inputs.dadx[1 + slot][i];
+               float dady = tri->inputs.dady[1 + slot][i];
+
+               debug_printf("IN[%u].%c: a0 = %f, dadx = %f, dady = %f\n",
+                            slot,
+                            "xyzw"[i],
+                            a0, dadx, dady);
+            }
+         }
+      }
+   }
 }
 
 
@@ -525,7 +523,7 @@ do_triangle_ccw(struct lp_setup_context *setup,
    info.dx20 = info.v2[0][0] - info.v0[0][0];
    info.dy01 = info.v0[0][1] - info.v1[0][1];
    info.dy20 = info.v2[0][1] - info.v0[0][1];
-   info.oneoverarea = 1.0 / (info.dx01 * info.dy20 - info.dx20 * info.dy01);
+   info.oneoverarea = 1.0f / (info.dx01 * info.dy20 - info.dx20 * info.dy01);
    info.frontfacing = frontfacing;
 
    /* Setup parameter interpolants:
@@ -590,35 +588,6 @@ do_triangle_ccw(struct lp_setup_context *setup,
       /* Calculate trivial accept offsets from the above.
        */
       plane->ei = plane->dcdy - plane->dcdx - plane->eo;
-
-      plane->step = tri->step[i];
-
-      /* Fill in the inputs.step[][] arrays.
-       * We've manually unrolled some loops here.
-       */
-#define SETUP_STEP(j, x, y) \
-      tri->step[i][j] = y * plane->dcdy - x * plane->dcdx
-      
-      SETUP_STEP(0, 0, 0);
-      SETUP_STEP(1, 1, 0);
-      SETUP_STEP(2, 0, 1);
-      SETUP_STEP(3, 1, 1);
-
-      SETUP_STEP(4, 2, 0);
-      SETUP_STEP(5, 3, 0);
-      SETUP_STEP(6, 2, 1);
-      SETUP_STEP(7, 3, 1);
-
-      SETUP_STEP(8, 0, 2);
-      SETUP_STEP(9, 1, 2);
-      SETUP_STEP(10, 0, 3);
-      SETUP_STEP(11, 1, 3);
-
-      SETUP_STEP(12, 2, 2);
-      SETUP_STEP(13, 3, 2);
-      SETUP_STEP(14, 2, 3);
-      SETUP_STEP(15, 3, 3);
-#undef STEP
    }
 
 
@@ -641,28 +610,24 @@ do_triangle_ccw(struct lp_setup_context *setup,
     * these planes elsewhere.
     */
    if (nr_planes == 7) {
-      tri->plane[3].step = step_scissor_minx;
       tri->plane[3].dcdx = -1;
       tri->plane[3].dcdy = 0;
       tri->plane[3].c = 1-minx;
       tri->plane[3].ei = 0;
       tri->plane[3].eo = 1;
 
-      tri->plane[4].step = step_scissor_maxx;
       tri->plane[4].dcdx = 1;
       tri->plane[4].dcdy = 0;
       tri->plane[4].c = maxx;
       tri->plane[4].ei = -1;
       tri->plane[4].eo = 0;
 
-      tri->plane[5].step = step_scissor_miny;
       tri->plane[5].dcdx = 0;
       tri->plane[5].dcdy = 1;
       tri->plane[5].c = 1-miny;
       tri->plane[5].ei = 0;
       tri->plane[5].eo = 1;
 
-      tri->plane[6].step = step_scissor_maxy;
       tri->plane[6].dcdx = 0;
       tri->plane[6].dcdy = -1;
       tri->plane[6].c = maxy;
diff --git a/src/gallium/drivers/llvmpipe/lp_state_fs.c b/src/gallium/drivers/llvmpipe/lp_state_fs.c
index 5953d690a41..dbca49a2efa 100644
--- a/src/gallium/drivers/llvmpipe/lp_state_fs.c
+++ b/src/gallium/drivers/llvmpipe/lp_state_fs.c
@@ -75,6 +75,7 @@
 #include "gallivm/lp_bld_type.h"
 #include "gallivm/lp_bld_const.h"
 #include "gallivm/lp_bld_conv.h"
+#include "gallivm/lp_bld_init.h"
 #include "gallivm/lp_bld_intr.h"
 #include "gallivm/lp_bld_logic.h"
 #include "gallivm/lp_bld_tgsi.h"
@@ -676,6 +677,11 @@ generate_fragment(struct llvmpipe_context *lp,
 		     color_ptr);
    }
 
+#ifdef PIPE_ARCH_X86
+   /* Avoid corrupting the FPU stack on 32bit OSes. */
+   lp_build_intrinsic(builder, "llvm.x86.mmx.emms", LLVMVoidType(), NULL, 0);
+#endif
+
    LLVMBuildRetVoid(builder);
 
    LLVMDisposeBuilder(builder);
@@ -710,6 +716,7 @@ generate_fragment(struct llvmpipe_context *lp,
       if (gallivm_debug & GALLIVM_DEBUG_ASM) {
          lp_disassemble(f);
       }
+      lp_func_delete_body(function);
    }
 }
 
diff --git a/src/gallium/drivers/llvmpipe/lp_state_vertex.c b/src/gallium/drivers/llvmpipe/lp_state_vertex.c
index 113f13db018..d86e66b4fb8 100644
--- a/src/gallium/drivers/llvmpipe/lp_state_vertex.c
+++ b/src/gallium/drivers/llvmpipe/lp_state_vertex.c
@@ -89,6 +89,19 @@ llvmpipe_set_vertex_buffers(struct pipe_context *pipe,
 }
 
 
+static void
+llvmpipe_set_index_buffer(struct pipe_context *pipe,
+                          const struct pipe_index_buffer *ib)
+{
+   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
+
+   if (ib)
+      memcpy(&llvmpipe->index_buffer, ib, sizeof(llvmpipe->index_buffer));
+   else
+      memset(&llvmpipe->index_buffer, 0, sizeof(llvmpipe->index_buffer));
+
+   /* TODO make this more like a state */
+}
 
 void
 llvmpipe_init_vertex_funcs(struct llvmpipe_context *llvmpipe)
@@ -98,4 +111,5 @@ llvmpipe_init_vertex_funcs(struct llvmpipe_context *llvmpipe)
    llvmpipe->pipe.delete_vertex_elements_state = llvmpipe_delete_vertex_elements_state;
 
    llvmpipe->pipe.set_vertex_buffers = llvmpipe_set_vertex_buffers;
+   llvmpipe->pipe.set_index_buffer = llvmpipe_set_index_buffer;
 }
diff --git a/src/gallium/drivers/llvmpipe/lp_test_blend.c b/src/gallium/drivers/llvmpipe/lp_test_blend.c
index 0c955556551..d0389f0cb0b 100644
--- a/src/gallium/drivers/llvmpipe/lp_test_blend.c
+++ b/src/gallium/drivers/llvmpipe/lp_test_blend.c
@@ -37,6 +37,7 @@
  */
 
 
+#include "gallivm/lp_bld_init.h"
 #include "gallivm/lp_bld_type.h"
 #include "gallivm/lp_bld_debug.h"
 #include "lp_bld_blend.h"
@@ -485,8 +486,7 @@ test_one(unsigned verbose,
 {
    LLVMModuleRef module = NULL;
    LLVMValueRef func = NULL;
-   LLVMExecutionEngineRef engine = NULL;
-   LLVMModuleProviderRef provider = NULL;
+   LLVMExecutionEngineRef engine = lp_build_engine;
    LLVMPassManagerRef pass = NULL;
    char *error = NULL;
    blend_test_ptr_t blend_test_ptr;
@@ -510,15 +510,6 @@ test_one(unsigned verbose,
    }
    LLVMDisposeMessage(error);
 
-   provider = LLVMCreateModuleProviderForExistingModule(module);
-   if (LLVMCreateJITCompiler(&engine, provider, 1, &error)) {
-      if(verbose < 1)
-         dump_blend_type(stderr, blend, mode, type);
-      fprintf(stderr, "%s\n", error);
-      LLVMDisposeMessage(error);
-      abort();
-   }
-
 #if 0
    pass = LLVMCreatePassManager();
    LLVMAddTargetData(LLVMGetExecutionEngineTargetData(engine), pass);
@@ -735,7 +726,6 @@ test_one(unsigned verbose,
 
    LLVMFreeMachineCodeForFunction(engine, func);
 
-   LLVMDisposeExecutionEngine(engine);
    if(pass)
       LLVMDisposePassManager(pass);
 
diff --git a/src/gallium/drivers/llvmpipe/lp_test_conv.c b/src/gallium/drivers/llvmpipe/lp_test_conv.c
index cf41b40581f..3ba42bf11a6 100644
--- a/src/gallium/drivers/llvmpipe/lp_test_conv.c
+++ b/src/gallium/drivers/llvmpipe/lp_test_conv.c
@@ -35,6 +35,7 @@
 
 
 #include "util/u_pointer.h"
+#include "gallivm/lp_bld_init.h"
 #include "gallivm/lp_bld_type.h"
 #include "gallivm/lp_bld_const.h"
 #include "gallivm/lp_bld_conv.h"
@@ -152,8 +153,7 @@ test_one(unsigned verbose,
 {
    LLVMModuleRef module = NULL;
    LLVMValueRef func = NULL;
-   LLVMExecutionEngineRef engine = NULL;
-   LLVMModuleProviderRef provider = NULL;
+   LLVMExecutionEngineRef engine = lp_build_engine;
    LLVMPassManagerRef pass = NULL;
    char *error = NULL;
    conv_test_ptr_t conv_test_ptr;
@@ -203,15 +203,6 @@ test_one(unsigned verbose,
    }
    LLVMDisposeMessage(error);
 
-   provider = LLVMCreateModuleProviderForExistingModule(module);
-   if (LLVMCreateJITCompiler(&engine, provider, 1, &error)) {
-      if(verbose < 1)
-         dump_conv_types(stderr, src_type, dst_type);
-      fprintf(stderr, "%s\n", error);
-      LLVMDisposeMessage(error);
-      abort();
-   }
-
 #if 0
    pass = LLVMCreatePassManager();
    LLVMAddTargetData(LLVMGetExecutionEngineTargetData(engine), pass);
@@ -351,7 +342,6 @@ test_one(unsigned verbose,
 
    LLVMFreeMachineCodeForFunction(engine, func);
 
-   LLVMDisposeExecutionEngine(engine);
    if(pass)
       LLVMDisposePassManager(pass);
 
diff --git a/src/gallium/drivers/llvmpipe/lp_test_printf.c b/src/gallium/drivers/llvmpipe/lp_test_printf.c
index 21df83f9d89..4653f30e39d 100644
--- a/src/gallium/drivers/llvmpipe/lp_test_printf.c
+++ b/src/gallium/drivers/llvmpipe/lp_test_printf.c
@@ -31,6 +31,8 @@
 
 #include "util/u_pointer.h"
 #include "gallivm/lp_bld.h"
+#include "gallivm/lp_bld_init.h"
+#include "gallivm/lp_bld_assert.h"
 #include "gallivm/lp_bld_printf.h"
 
 #include <llvm-c/Analysis.h>
@@ -74,6 +76,10 @@ add_printf_test(LLVMModuleRef module)
    lp_build_printf(builder, "hello, world\n");
    lp_build_printf(builder, "print 5 6: %d %d\n", LLVMConstInt(LLVMInt32Type(), 5, 0),
 				LLVMConstInt(LLVMInt32Type(), 6, 0));
+
+   /* Also test lp_build_assert().  This should not fail. */
+   lp_build_assert(builder, LLVMConstInt(LLVMInt32Type(), 1, 0), "assert(1)");
+
    LLVMBuildRetVoid(builder);
    LLVMDisposeBuilder(builder);
    return func;
@@ -107,11 +113,16 @@ test_printf(unsigned verbose, FILE *fp, const struct printf_test_case *testcase)
    LLVMDisposeMessage(error);
 
    provider = LLVMCreateModuleProviderForExistingModule(module);
+#if 0
    if (LLVMCreateJITCompiler(&engine, provider, 1, &error)) {
       fprintf(stderr, "%s\n", error);
       LLVMDisposeMessage(error);
       abort();
    }
+#else
+   (void) provider;
+   engine = lp_build_engine;
+#endif
 
 #if 0
    pass = LLVMCreatePassManager();
diff --git a/src/gallium/drivers/llvmpipe/lp_test_round.c b/src/gallium/drivers/llvmpipe/lp_test_round.c
index f571a81a4a4..57b0ee57767 100644
--- a/src/gallium/drivers/llvmpipe/lp_test_round.c
+++ b/src/gallium/drivers/llvmpipe/lp_test_round.c
@@ -31,7 +31,7 @@
 
 #include "util/u_pointer.h"
 #include "gallivm/lp_bld.h"
-#include "gallivm/lp_bld_printf.h"
+#include "gallivm/lp_bld_init.h"
 #include "gallivm/lp_bld_arit.h"
 
 #include <llvm-c/Analysis.h>
@@ -121,8 +121,7 @@ test_round(unsigned verbose, FILE *fp)
 {
    LLVMModuleRef module = NULL;
    LLVMValueRef test_round = NULL, test_trunc, test_floor, test_ceil;
-   LLVMExecutionEngineRef engine = NULL;
-   LLVMModuleProviderRef provider = NULL;
+   LLVMExecutionEngineRef engine = lp_build_engine;
    LLVMPassManagerRef pass = NULL;
    char *error = NULL;
    test_round_t round_func, trunc_func, floor_func, ceil_func;
@@ -145,13 +144,6 @@ test_round(unsigned verbose, FILE *fp)
    }
    LLVMDisposeMessage(error);
 
-   provider = LLVMCreateModuleProviderForExistingModule(module);
-   if (LLVMCreateJITCompiler(&engine, provider, 1, &error)) {
-      fprintf(stderr, "%s\n", error);
-      LLVMDisposeMessage(error);
-      abort();
-   }
-
 #if 0
    pass = LLVMCreatePassManager();
    LLVMAddTargetData(LLVMGetExecutionEngineTargetData(engine), pass);
diff --git a/src/gallium/drivers/llvmpipe/lp_test_sincos.c b/src/gallium/drivers/llvmpipe/lp_test_sincos.c
index 1366ecddcbc..7ab357f162e 100644
--- a/src/gallium/drivers/llvmpipe/lp_test_sincos.c
+++ b/src/gallium/drivers/llvmpipe/lp_test_sincos.c
@@ -30,8 +30,9 @@
 #include <stdio.h>
 
 #include "gallivm/lp_bld.h"
-#include "gallivm/lp_bld_printf.h"
+#include "gallivm/lp_bld_init.h"
 #include "gallivm/lp_bld_arit.h"
+#include "util/u_pointer.h"
 
 #include <llvm-c/Analysis.h>
 #include <llvm-c/ExecutionEngine.h>
@@ -101,8 +102,7 @@ test_sincos(unsigned verbose, FILE *fp)
 {
    LLVMModuleRef module = NULL;
    LLVMValueRef test_sin = NULL, test_cos = NULL;
-   LLVMExecutionEngineRef engine = NULL;
-   LLVMModuleProviderRef provider = NULL;
+   LLVMExecutionEngineRef engine = lp_build_engine;
    LLVMPassManagerRef pass = NULL;
    char *error = NULL;
    test_sincos_t sin_func;
@@ -122,13 +122,6 @@ test_sincos(unsigned verbose, FILE *fp)
    }
    LLVMDisposeMessage(error);
 
-   provider = LLVMCreateModuleProviderForExistingModule(module);
-   if (LLVMCreateJITCompiler(&engine, provider, 1, &error)) {
-      fprintf(stderr, "%s\n", error);
-      LLVMDisposeMessage(error);
-      abort();
-   }
-
 #if 0
    pass = LLVMCreatePassManager();
    LLVMAddTargetData(LLVMGetExecutionEngineTargetData(engine), pass);
@@ -144,8 +137,8 @@ test_sincos(unsigned verbose, FILE *fp)
    (void)pass;
 #endif
 
-   sin_func = (test_sincos_t)LLVMGetPointerToGlobal(engine, test_sin);
-   cos_func = (test_sincos_t)LLVMGetPointerToGlobal(engine, test_cos);
+   sin_func = (test_sincos_t) pointer_to_func(LLVMGetPointerToGlobal(engine, test_sin));
+   cos_func = (test_sincos_t) pointer_to_func(LLVMGetPointerToGlobal(engine, test_cos));
 
    memset(unpacked, 0, sizeof unpacked);
 
@@ -162,7 +155,6 @@ test_sincos(unsigned verbose, FILE *fp)
    LLVMFreeMachineCodeForFunction(engine, test_sin);
    LLVMFreeMachineCodeForFunction(engine, test_cos);
 
-   LLVMDisposeExecutionEngine(engine);
    if(pass)
       LLVMDisposePassManager(pass);
 
diff --git a/src/gallium/drivers/llvmpipe/lp_tile_soa.py b/src/gallium/drivers/llvmpipe/lp_tile_soa.py
index c71ec8066c7..2ba39052aba 100644
--- a/src/gallium/drivers/llvmpipe/lp_tile_soa.py
+++ b/src/gallium/drivers/llvmpipe/lp_tile_soa.py
@@ -293,34 +293,7 @@ def generate_ssse3():
     print '''
 #if defined(PIPE_ARCH_SSE)
 
-
-#if defined(PIPE_ARCH_SSSE3)
-
-#include <tmmintrin.h>
-
-#else
-
-#include <emmintrin.h>
-
-/**
- * Describe _mm_shuffle_epi8() with gcc extended inline assembly, for cases
- * where -mssse3 is not supported/enabled.
- *
- * MSVC will never get in here as its intrinsics support do not rely on
- * compiler command line options.
- */
-static __inline __m128i __attribute__((__gnu_inline__, __always_inline__, __artificial__))
-_mm_shuffle_epi8(__m128i a, __m128i mask)
-{
-    __m128i result;
-    __asm__("pshufb %1, %0"
-            : "=x" (result)
-            : "xm" (mask), "0" (a));
-    return result;
-}
-
-#endif
-
+#include "util/u_sse.h"
 
 static void
 lp_tile_b8g8r8a8_unorm_swizzle_4ub_ssse3(uint8_t *dst,
diff --git a/src/gallium/drivers/nouveau/nouveau_stateobj.h b/src/gallium/drivers/nouveau/nouveau_stateobj.h
index f5c1c5ca2c3..e920cf9f3bc 100644
--- a/src/gallium/drivers/nouveau/nouveau_stateobj.h
+++ b/src/gallium/drivers/nouveau/nouveau_stateobj.h
@@ -151,9 +151,9 @@ so_method(struct nouveau_stateobj *so, struct nouveau_grobj *gr,
 	if (so->start_alloc <= so->cur_start) {
 		debug_printf("exceeding num_start size\n");
 		assert(0);
-	} else
+	}
 #endif /* DEBUG_NOUVEAU_STATEOBJ */
-		start = so->start;
+	start = so->start;
 
 #ifdef DEBUG_NOUVEAU_STATEOBJ
 	if (so->cur_start > 0 && start[so->cur_start - 1].size > so->cur) {
@@ -162,7 +162,6 @@ so_method(struct nouveau_stateobj *so, struct nouveau_grobj *gr,
 	}
 #endif /* DEBUG_NOUVEAU_STATEOBJ */
 
-	so->start = start;
 	start[so->cur_start].gr = gr;
 	start[so->cur_start].mthd = mthd;
 	start[so->cur_start].size = size;
@@ -193,11 +192,10 @@ so_reloc(struct nouveau_stateobj *so, struct nouveau_bo *bo,
 	if (so->reloc_alloc <= so->cur_reloc) {
 		debug_printf("exceeding num_reloc size\n");
 		assert(0);
-	} else
+	}
 #endif /* DEBUG_NOUVEAU_STATEOBJ */
-		r = so->reloc;
+	r = so->reloc;
 
-	so->reloc = r;
 	r[so->cur_reloc].bo = NULL;
 	nouveau_bo_ref(bo, &(r[so->cur_reloc].bo));
 	r[so->cur_reloc].gr = so->start[so->cur_start-1].gr;
diff --git a/src/gallium/drivers/nouveau/nouveau_util.h b/src/gallium/drivers/nouveau/nouveau_util.h
index a5e8537533e..b165f7a611a 100644
--- a/src/gallium/drivers/nouveau/nouveau_util.h
+++ b/src/gallium/drivers/nouveau/nouveau_util.h
@@ -88,104 +88,4 @@ static INLINE unsigned log2i(unsigned i)
 	return r;
 }
 
-struct u_split_prim {
-   void *priv;
-   void (*emit)(void *priv, unsigned start, unsigned count);
-   void (*edge)(void *priv, boolean enabled);
-
-   unsigned mode;
-   unsigned start;
-   unsigned p_start;
-   unsigned p_end;
-
-   uint repeat_first:1;
-   uint close_first:1;
-   uint edgeflag_off:1;
-};
-
-static INLINE void
-u_split_prim_init(struct u_split_prim *s,
-                  unsigned mode, unsigned start, unsigned count)
-{
-   if (mode == PIPE_PRIM_LINE_LOOP) {
-      s->mode = PIPE_PRIM_LINE_STRIP;
-      s->close_first = 1;
-   } else {
-      s->mode = mode;
-      s->close_first = 0;
-   }
-   s->start = start;
-   s->p_start = start;
-   s->p_end = start + count;
-   s->edgeflag_off = 0;
-   s->repeat_first = 0;
-}
-
-static INLINE boolean
-u_split_prim_next(struct u_split_prim *s, unsigned max_verts)
-{
-   int repeat = 0;
-
-   if (s->repeat_first) {
-      s->emit(s->priv, s->start, 1);
-      max_verts--;
-      if (s->edgeflag_off) {
-         s->edge(s->priv, TRUE);
-         s->edgeflag_off = FALSE;
-      }
-   }
-
-   if (s->p_start + s->close_first + max_verts >= s->p_end) {
-      s->emit(s->priv, s->p_start, s->p_end - s->p_start);
-      if (s->close_first)
-         s->emit(s->priv, s->start, 1);
-      return TRUE;
-   }
-
-   switch (s->mode) {
-   case PIPE_PRIM_LINES:
-      max_verts &= ~1;
-      break;
-   case PIPE_PRIM_LINE_STRIP:
-      repeat = 1;
-      break;
-   case PIPE_PRIM_POLYGON:
-      max_verts--;
-      s->emit(s->priv, s->p_start, max_verts);
-      s->edge(s->priv, FALSE);
-      s->emit(s->priv, s->p_start + max_verts, 1);
-      s->p_start += max_verts;
-      s->repeat_first = TRUE;
-      s->edgeflag_off = TRUE;
-      return FALSE;
-   case PIPE_PRIM_TRIANGLES:
-      max_verts = max_verts - (max_verts % 3);
-      break;
-   case PIPE_PRIM_TRIANGLE_STRIP:
-      /* to ensure winding stays correct, always split
-       * on an even number of generated triangles
-       */
-      max_verts = max_verts & ~1;
-      repeat = 2;
-      break;
-   case PIPE_PRIM_TRIANGLE_FAN:
-      s->repeat_first = TRUE;
-      repeat = 1;
-      break;
-   case PIPE_PRIM_QUADS:
-      max_verts &= ~3;
-      break;
-   case PIPE_PRIM_QUAD_STRIP:
-      max_verts &= ~1;
-      repeat = 2;
-      break;
-   default:
-      break;
-   }
-
-   s->emit (s->priv, s->p_start, max_verts);
-   s->p_start += (max_verts - repeat);
-   return FALSE;
-}
-
 #endif
diff --git a/src/gallium/drivers/nouveau/nouveau_winsys.h b/src/gallium/drivers/nouveau/nouveau_winsys.h
index df79ca89ca1..c6c93d40b8f 100644
--- a/src/gallium/drivers/nouveau/nouveau_winsys.h
+++ b/src/gallium/drivers/nouveau/nouveau_winsys.h
@@ -24,11 +24,10 @@ nouveau_screen_transfer_flags(unsigned pipe)
 		flags |= NOUVEAU_BO_WR;
 	if (pipe & PIPE_TRANSFER_DISCARD)
 		flags |= NOUVEAU_BO_INVAL;
-	if (pipe & PIPE_TRANSFER_DONTBLOCK)
-		flags |= NOUVEAU_BO_NOWAIT;
-	else
 	if (pipe & PIPE_TRANSFER_UNSYNCHRONIZED)
 		flags |= NOUVEAU_BO_NOSYNC;
+	else if (pipe & PIPE_TRANSFER_DONTBLOCK)
+		flags |= NOUVEAU_BO_NOWAIT;
 
 	return flags;
 }
diff --git a/src/gallium/drivers/nv50/nv50_context.c b/src/gallium/drivers/nv50/nv50_context.c
index 915a9254025..0874cb5e4ea 100644
--- a/src/gallium/drivers/nv50/nv50_context.c
+++ b/src/gallium/drivers/nv50/nv50_context.c
@@ -82,10 +82,7 @@ nv50_create(struct pipe_screen *pscreen, void *priv)
 
 	nv50->pipe.destroy = nv50_destroy;
 
-	nv50->pipe.draw_arrays = nv50_draw_arrays;
-	nv50->pipe.draw_arrays_instanced = nv50_draw_arrays_instanced;
-	nv50->pipe.draw_elements = nv50_draw_elements;
-	nv50->pipe.draw_elements_instanced = nv50_draw_elements_instanced;
+	nv50->pipe.draw_vbo = nv50_draw_vbo;
 	nv50->pipe.clear = nv50_clear;
 
 	nv50->pipe.flush = nv50_flush;
diff --git a/src/gallium/drivers/nv50/nv50_context.h b/src/gallium/drivers/nv50/nv50_context.h
index 12c4a93a9bd..d24d6c50ea8 100644
--- a/src/gallium/drivers/nv50/nv50_context.h
+++ b/src/gallium/drivers/nv50/nv50_context.h
@@ -148,6 +148,7 @@ struct nv50_context {
 	struct pipe_resource *constbuf[PIPE_SHADER_TYPES];
 	struct pipe_vertex_buffer vtxbuf[PIPE_MAX_ATTRIBS];
 	unsigned vtxbuf_nr;
+	struct pipe_index_buffer idxbuf;
 	struct nv50_vtxelt_stateobj *vtxelt;
 	struct nv50_sampler_stateobj *sampler[3][PIPE_MAX_SAMPLERS];
 	unsigned sampler_nr[3];
@@ -179,24 +180,8 @@ nv50_surface_do_copy(struct nv50_screen *screen, struct pipe_surface *dst,
 extern struct draw_stage *nv50_draw_render_stage(struct nv50_context *nv50);
 
 /* nv50_vbo.c */
-extern void nv50_draw_arrays(struct pipe_context *, unsigned mode,
-				unsigned start, unsigned count);
-extern void nv50_draw_arrays_instanced(struct pipe_context *, unsigned mode,
-					unsigned start, unsigned count,
-					unsigned startInstance,
-					unsigned instanceCount);
-extern void nv50_draw_elements(struct pipe_context *pipe,
-				  struct pipe_resource *indexBuffer,
-				  unsigned indexSize, int indexBias,
-				  unsigned mode, unsigned start,
-				  unsigned count);
-extern void nv50_draw_elements_instanced(struct pipe_context *pipe,
-					 struct pipe_resource *indexBuffer,
-					 unsigned indexSize, int indexBias,
-					 unsigned mode, unsigned start,
-					 unsigned count,
-					 unsigned startInstance,
-					 unsigned instanceCount);
+extern void nv50_draw_vbo(struct pipe_context *pipe,
+                          const struct pipe_draw_info *info);
 extern void nv50_vtxelt_construct(struct nv50_vtxelt_stateobj *cso);
 extern struct nouveau_stateobj *nv50_vbo_validate(struct nv50_context *nv50);
 
diff --git a/src/gallium/drivers/nv50/nv50_push.c b/src/gallium/drivers/nv50/nv50_push.c
index 481182dd8df..0091927a982 100644
--- a/src/gallium/drivers/nv50/nv50_push.c
+++ b/src/gallium/drivers/nv50/nv50_push.c
@@ -2,8 +2,8 @@
 #include "pipe/p_state.h"
 #include "util/u_inlines.h"
 #include "util/u_format.h"
+#include "util/u_split_prim.h"
 
-#include "nouveau/nouveau_util.h"
 #include "nv50_context.h"
 #include "nv50_resource.h"
 
@@ -217,7 +217,7 @@ nv50_push_elements_instanced(struct pipe_context *pipe,
                                4; /* potential edgeflag enable/disable */
    const unsigned v_overhead = 1 + /* VERTEX_DATA packet header */
                                2; /* potential edgeflag modification */
-   struct u_split_prim s;
+   struct util_split_prim s;
    unsigned vtx_size;
    boolean nzi = FALSE;
    int i;
@@ -335,7 +335,7 @@ nv50_push_elements_instanced(struct pipe_context *pipe,
          ctx.attr[i].map = (uint8_t *)ctx.attr[i].map + ctx.attr[i].stride;
       }
 
-      u_split_prim_init(&s, mode, start, count);
+      util_split_prim_init(&s, mode, start, count);
       do {
          if (AVAIL_RING(chan) < p_overhead + (6 * vtx_size)) {
             FIRE_RING(chan);
@@ -351,7 +351,7 @@ nv50_push_elements_instanced(struct pipe_context *pipe,
 
          BEGIN_RING(chan, tesla, NV50TCL_VERTEX_BEGIN, 1);
          OUT_RING  (chan, nv50_prim(s.mode) | (nzi ? (1 << 28) : 0));
-         done = u_split_prim_next(&s, max_verts);
+         done = util_split_prim_next(&s, max_verts);
          BEGIN_RING(chan, tesla, NV50TCL_VERTEX_END, 1);
          OUT_RING  (chan, 0);
       } while (!done);
diff --git a/src/gallium/drivers/nv50/nv50_state.c b/src/gallium/drivers/nv50/nv50_state.c
index 88fee3630be..3afce06557a 100644
--- a/src/gallium/drivers/nv50/nv50_state.c
+++ b/src/gallium/drivers/nv50/nv50_state.c
@@ -786,6 +786,20 @@ nv50_set_vertex_buffers(struct pipe_context *pipe, unsigned count,
 	nv50->dirty |= NV50_NEW_ARRAYS;
 }
 
+static void
+nv50_set_index_buffer(struct pipe_context *pipe,
+		      const struct pipe_index_buffer *ib)
+{
+	struct nv50_context *nv50 = nv50_context(pipe);
+
+	if (ib)
+		memcpy(&nv50->idxbuf, ib, sizeof(nv50->idxbuf));
+	else
+		memset(&nv50->idxbuf, 0, sizeof(nv50->idxbuf));
+
+	/* TODO make this more like a state */
+}
+
 static void *
 nv50_vtxelts_state_create(struct pipe_context *pipe,
 			  unsigned num_elements,
@@ -871,5 +885,6 @@ nv50_init_state_functions(struct nv50_context *nv50)
 	nv50->pipe.bind_vertex_elements_state = nv50_vtxelts_state_bind;
 
 	nv50->pipe.set_vertex_buffers = nv50_set_vertex_buffers;
+	nv50->pipe.set_index_buffer = nv50_set_index_buffer;
 }
 
diff --git a/src/gallium/drivers/nv50/nv50_vbo.c b/src/gallium/drivers/nv50/nv50_vbo.c
index 4fe0df5683d..d41a59d05db 100644
--- a/src/gallium/drivers/nv50/nv50_vbo.c
+++ b/src/gallium/drivers/nv50/nv50_vbo.c
@@ -24,8 +24,8 @@
 #include "pipe/p_state.h"
 #include "util/u_inlines.h"
 #include "util/u_format.h"
+#include "util/u_split_prim.h"
 
-#include "nouveau/nouveau_util.h"
 #include "nv50_context.h"
 #include "nv50_resource.h"
 
@@ -83,7 +83,7 @@ instance_step(struct nv50_context *nv50, struct instance *a)
 	}
 }
 
-void
+static void
 nv50_draw_arrays_instanced(struct pipe_context *pipe,
 			   unsigned mode, unsigned start, unsigned count,
 			   unsigned startInstance, unsigned instanceCount)
@@ -130,13 +130,6 @@ nv50_draw_arrays_instanced(struct pipe_context *pipe,
 	}
 }
 
-void
-nv50_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start,
-		 unsigned count)
-{
-	nv50_draw_arrays_instanced(pipe, mode, start, count, 0, 1);
-}
-
 struct inline_ctx {
 	struct nv50_context *nv50;
 	void *map;
@@ -228,7 +221,7 @@ nv50_draw_elements_inline(struct pipe_context *pipe,
 	struct pipe_transfer *transfer;
 	struct instance a[16];
 	struct inline_ctx ctx;
-	struct u_split_prim s;
+	struct util_split_prim s;
 	boolean nzi = FALSE;
 	unsigned overhead;
 
@@ -264,7 +257,7 @@ nv50_draw_elements_inline(struct pipe_context *pipe,
 		unsigned max_verts;
 		boolean done;
 
-		u_split_prim_init(&s, mode, start, count);
+		util_split_prim_init(&s, mode, start, count);
 		do {
 			if (AVAIL_RING(chan) < (overhead + 6)) {
 				FIRE_RING(chan);
@@ -283,7 +276,7 @@ nv50_draw_elements_inline(struct pipe_context *pipe,
 
 			BEGIN_RING(chan, tesla, NV50TCL_VERTEX_BEGIN, 1);
 			OUT_RING  (chan, nv50_prim(s.mode) | (nzi ? (1<<28) : 0));
-			done = u_split_prim_next(&s, max_verts);
+			done = util_split_prim_next(&s, max_verts);
 			BEGIN_RING(chan, tesla, NV50TCL_VERTEX_END, 1);
 			OUT_RING  (chan, 0);
 		} while (!done);
@@ -294,7 +287,7 @@ nv50_draw_elements_inline(struct pipe_context *pipe,
 	pipe_buffer_unmap(pipe, indexBuffer, transfer);
 }
 
-void
+static void
 nv50_draw_elements_instanced(struct pipe_context *pipe,
 			     struct pipe_resource *indexBuffer,
 			     unsigned indexSize, int indexBias,
@@ -374,13 +367,34 @@ nv50_draw_elements_instanced(struct pipe_context *pipe,
 }
 
 void
-nv50_draw_elements(struct pipe_context *pipe,
-		   struct pipe_resource *indexBuffer,
-		   unsigned indexSize, int indexBias,
-		   unsigned mode, unsigned start, unsigned count)
+nv50_draw_vbo(struct pipe_context *pipe, const struct pipe_draw_info *info)
 {
-	nv50_draw_elements_instanced(pipe, indexBuffer, indexSize, indexBias,
-				     mode, start, count, 0, 1);
+	struct nv50_context *nv50 = nv50_context(pipe);
+
+	if (info->indexed && nv50->idxbuf.buffer) {
+		unsigned offset;
+
+		assert(nv50->idxbuf.offset % nv50->idxbuf.index_size == 0);
+		offset = nv50->idxbuf.offset / nv50->idxbuf.index_size;
+
+		nv50_draw_elements_instanced(pipe,
+					     nv50->idxbuf.buffer,
+					     nv50->idxbuf.index_size,
+					     info->index_bias,
+					     info->mode,
+					     info->start + offset,
+					     info->count,
+					     info->start_instance,
+					     info->instance_count);
+	}
+	else {
+		nv50_draw_arrays_instanced(pipe,
+					   info->mode,
+					   info->start,
+					   info->count,
+					   info->start_instance,
+					   info->instance_count);
+	}
 }
 
 static INLINE boolean
diff --git a/src/gallium/drivers/nvfx/nvfx_context.c b/src/gallium/drivers/nvfx/nvfx_context.c
index 6d2dc4d5bf6..7218abff22d 100644
--- a/src/gallium/drivers/nvfx/nvfx_context.c
+++ b/src/gallium/drivers/nvfx/nvfx_context.c
@@ -55,8 +55,7 @@ nvfx_create(struct pipe_screen *pscreen, void *priv)
 	nvfx->pipe.screen = pscreen;
 	nvfx->pipe.priv = priv;
 	nvfx->pipe.destroy = nvfx_destroy;
-	nvfx->pipe.draw_arrays = nvfx_draw_arrays;
-	nvfx->pipe.draw_elements = nvfx_draw_elements;
+	nvfx->pipe.draw_vbo = nvfx_draw_vbo;
 	nvfx->pipe.clear = nvfx_clear;
 	nvfx->pipe.flush = nvfx_flush;
 
diff --git a/src/gallium/drivers/nvfx/nvfx_context.h b/src/gallium/drivers/nvfx/nvfx_context.h
index e48f9f3aa88..89f94c10bd1 100644
--- a/src/gallium/drivers/nvfx/nvfx_context.h
+++ b/src/gallium/drivers/nvfx/nvfx_context.h
@@ -121,7 +121,8 @@ struct nvfx_context {
 	struct pipe_stencil_ref stencil_ref;
 	struct pipe_viewport_state viewport;
 	struct pipe_framebuffer_state framebuffer;
-	struct pipe_resource *idxbuf;
+	struct pipe_index_buffer idxbuf;
+	struct pipe_resource *idxbuf_buffer;
 	unsigned idxbuf_format;
 	struct nvfx_sampler_state *tex_sampler[PIPE_MAX_SAMPLERS];
 	struct pipe_sampler_view *fragment_sampler_views[PIPE_MAX_SAMPLERS];
@@ -235,13 +236,8 @@ extern void nvfx_init_transfer_functions(struct nvfx_context *nvfx);
 /* nvfx_vbo.c */
 extern boolean nvfx_vbo_validate(struct nvfx_context *nvfx);
 extern void nvfx_vbo_relocate(struct nvfx_context *nvfx);
-extern void nvfx_draw_arrays(struct pipe_context *, unsigned mode,
-				unsigned start, unsigned count);
-extern void nvfx_draw_elements(struct pipe_context *pipe,
-                               struct pipe_resource *indexBuffer,
-                               unsigned indexSize, int indexBias,
-                               unsigned mode, unsigned start,
-                               unsigned count);
+extern void nvfx_draw_vbo(struct pipe_context *pipe,
+                          const struct pipe_draw_info *info);
 
 /* nvfx_vertprog.c */
 extern boolean nvfx_vertprog_validate(struct nvfx_context *nvfx);
diff --git a/src/gallium/drivers/nvfx/nvfx_fragprog.c b/src/gallium/drivers/nvfx/nvfx_fragprog.c
index 6772d9bd516..ee41f03b9b8 100644
--- a/src/gallium/drivers/nvfx/nvfx_fragprog.c
+++ b/src/gallium/drivers/nvfx/nvfx_fragprog.c
@@ -842,7 +842,6 @@ nvfx_fragprog_validate(struct nvfx_context *nvfx)
 	struct nouveau_channel* chan = nvfx->screen->base.channel;
 	struct nvfx_fragment_program *fp = nvfx->fragprog;
 	int update = 0;
-	int i;
 
 	if (!fp->translated)
 	{
@@ -895,6 +894,7 @@ nvfx_fragprog_validate(struct nvfx_context *nvfx)
 			{
 				struct nvfx_fragment_program_bo* fpbo = os_malloc_aligned(sizeof(struct nvfx_fragment_program) + fp->prog_size * fp->progs_per_bo, 16);
 				char *map, *buf;
+				int i;
 
 				if(fp->fpbo)
 				{
@@ -910,7 +910,7 @@ nvfx_fragprog_validate(struct nvfx_context *nvfx)
 
 				map = fpbo->bo->map;
 				buf = fpbo->insn;
-				for(int i = 0; i < fp->progs_per_bo; ++i)
+				for(i = 0; i < fp->progs_per_bo; ++i)
 				{
 					memcpy(buf, fp->insn, fp->insn_len * 4);
 					nvfx_fp_memcpy(map, fp->insn, fp->insn_len * 4);
@@ -931,6 +931,7 @@ nvfx_fragprog_validate(struct nvfx_context *nvfx)
 			uint32_t* map = pipe_buffer_map(&nvfx->pipe, constbuf, PIPE_TRANSFER_READ, &transfer);
 			uint32_t* fpmap = (uint32_t*)((char*)fp->fpbo->bo->map + offset);
 			uint32_t* buf = (uint32_t*)((char*)fp->fpbo->insn + offset);
+			int i;
 			for (i = 0; i < fp->nr_consts; ++i) {
 				unsigned off = fp->consts[i].offset;
 				unsigned idx = fp->consts[i].index * 4;
diff --git a/src/gallium/drivers/nvfx/nvfx_screen.c b/src/gallium/drivers/nvfx/nvfx_screen.c
index 80db28a07cb..f2525ccb38f 100644
--- a/src/gallium/drivers/nvfx/nvfx_screen.c
+++ b/src/gallium/drivers/nvfx/nvfx_screen.c
@@ -131,6 +131,8 @@ nvfx_screen_get_param(struct pipe_screen *pscreen, enum pipe_cap param)
 		return screen->is_nv4x ? 1 : 0;
 	case PIPE_CAP_GEOMETRY_SHADER4:
 		return 0;
+	case PIPE_CAP_DEPTH_CLAMP:
+		return 0; // TODO: implement depth clamp
 	default:
 		NOUVEAU_ERR("Unknown PIPE_CAP %d\n", param);
 		return 0;
diff --git a/src/gallium/drivers/nvfx/nvfx_state.c b/src/gallium/drivers/nvfx/nvfx_state.c
index 30322d46d93..cd58e439d71 100644
--- a/src/gallium/drivers/nvfx/nvfx_state.c
+++ b/src/gallium/drivers/nvfx/nvfx_state.c
@@ -555,6 +555,20 @@ nvfx_set_vertex_buffers(struct pipe_context *pipe, unsigned count,
 	nvfx->draw_dirty |= NVFX_NEW_ARRAYS;
 }
 
+static void
+nvfx_set_index_buffer(struct pipe_context *pipe,
+		      const struct pipe_index_buffer *ib)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+
+	if (ib)
+		memcpy(&nvfx->idxbuf, ib, sizeof(nvfx->idxbuf));
+	else
+		memset(&nvfx->idxbuf, 0, sizeof(nvfx->idxbuf));
+
+	/* TODO make this more like a state */
+}
+
 static void *
 nvfx_vtxelts_state_create(struct pipe_context *pipe,
 			  unsigned num_elements,
@@ -635,4 +649,5 @@ nvfx_init_state_functions(struct nvfx_context *nvfx)
 	nvfx->pipe.bind_vertex_elements_state = nvfx_vtxelts_state_bind;
 
 	nvfx->pipe.set_vertex_buffers = nvfx_set_vertex_buffers;
+	nvfx->pipe.set_index_buffer = nvfx_set_index_buffer;
 }
diff --git a/src/gallium/drivers/nvfx/nvfx_vbo.c b/src/gallium/drivers/nvfx/nvfx_vbo.c
index 520bae5aed2..4aa37938425 100644
--- a/src/gallium/drivers/nvfx/nvfx_vbo.c
+++ b/src/gallium/drivers/nvfx/nvfx_vbo.c
@@ -85,7 +85,7 @@ nvfx_vbo_set_idxbuf(struct nvfx_context *nvfx, struct pipe_resource *ib,
 	unsigned type;
 
 	if (!ib) {
-		nvfx->idxbuf = NULL;
+		nvfx->idxbuf_buffer = NULL;
 		nvfx->idxbuf_format = 0xdeadbeef;
 		return FALSE;
 	}
@@ -104,10 +104,10 @@ nvfx_vbo_set_idxbuf(struct nvfx_context *nvfx, struct pipe_resource *ib,
 		return FALSE;
 	}
 
-	if (ib != nvfx->idxbuf ||
+	if (ib != nvfx->idxbuf_buffer ||
 	    type != nvfx->idxbuf_format) {
 		nvfx->dirty |= NVFX_NEW_ARRAYS;
-		nvfx->idxbuf = ib;
+		nvfx->idxbuf_buffer = ib;
 		nvfx->idxbuf_format = type;
 	}
 
@@ -158,7 +158,7 @@ nvfx_vbo_static_attrib(struct nvfx_context *nvfx,
 	pipe_buffer_unmap(&nvfx->pipe, vb->buffer, transfer);
 }
 
-void
+static void
 nvfx_draw_arrays(struct pipe_context *pipe,
 		 unsigned mode, unsigned start, unsigned count)
 {
@@ -463,7 +463,7 @@ nvfx_draw_elements_vbo(struct pipe_context *pipe,
 	}
 }
 
-void
+static void
 nvfx_draw_elements(struct pipe_context *pipe,
 		   struct pipe_resource *indexBuffer,
 		   unsigned indexSize, int indexBias,
@@ -491,11 +491,38 @@ nvfx_draw_elements(struct pipe_context *pipe,
 	pipe->flush(pipe, 0, NULL);
 }
 
+void
+nvfx_draw_vbo(struct pipe_context *pipe, const struct pipe_draw_info *info)
+{
+	struct nvfx_context *nvfx = nvfx_context(pipe);
+
+	if (info->indexed && nvfx->idxbuf.buffer) {
+		unsigned offset;
+
+		assert(nvfx->idxbuf.offset % nvfx->idxbuf.index_size == 0);
+		offset = nvfx->idxbuf.offset / nvfx->idxbuf.index_size;
+
+		nvfx_draw_elements(pipe,
+				   nvfx->idxbuf.buffer,
+				   nvfx->idxbuf.index_size,
+				   info->index_bias,
+				   info->mode,
+				   info->start + offset,
+				   info->count);
+	}
+	else {
+		nvfx_draw_arrays(pipe,
+				info->mode,
+				info->start,
+				info->count);
+	}
+}
+
 boolean
 nvfx_vbo_validate(struct nvfx_context *nvfx)
 {
 	struct nouveau_channel* chan = nvfx->screen->base.channel;
-	struct pipe_resource *ib = nvfx->idxbuf;
+	struct pipe_resource *ib = nvfx->idxbuf_buffer;
 	unsigned ib_format = nvfx->idxbuf_format;
 	int i;
 	int elements = MAX2(nvfx->vtxelt->num_elements, nvfx->hw_vtxelt_nr);
@@ -610,10 +637,10 @@ nvfx_vbo_relocate(struct nvfx_context *nvfx)
 		}
 	}
 
-	if(nvfx->idxbuf)
+	if(nvfx->idxbuf_buffer)
 	{
 		unsigned ib_flags = nvfx->screen->index_buffer_reloc_flags | NOUVEAU_BO_RD | NOUVEAU_BO_DUMMY;
-		struct nouveau_bo* bo = nvfx_resource(nvfx->idxbuf)->bo;
+		struct nouveau_bo* bo = nvfx_resource(nvfx->idxbuf_buffer)->bo;
 
 		assert(nvfx->screen->index_buffer_reloc_flags);
 
diff --git a/src/gallium/drivers/nvfx/nvfx_vertprog.c b/src/gallium/drivers/nvfx/nvfx_vertprog.c
index 80b98b62d34..24d9846310e 100644
--- a/src/gallium/drivers/nvfx/nvfx_vertprog.c
+++ b/src/gallium/drivers/nvfx/nvfx_vertprog.c
@@ -299,7 +299,13 @@ nvfx_vp_arith(struct nvfx_context* nvfx, struct nvfx_vpc *vpc, int slot, int op,
 		  (3 << NVFX_VP(INST_COND_SWZ_W_SHIFT)));
 
 	if(!nvfx->is_nv4x) {
-		hw[1] |= (op << NV30_VP_INST_VEC_OPCODE_SHIFT);
+		if(slot == 0)
+			hw[1] |= (op << NV30_VP_INST_VEC_OPCODE_SHIFT);
+		else
+		{
+			hw[0] |= ((op >> 4) << NV30_VP_INST_SCA_OPCODEH_SHIFT);
+			hw[1] |= ((op & 0xf) << NV30_VP_INST_SCA_OPCODEL_SHIFT);
+		}
 //		hw[3] |= NVFX_VP(INST_SCA_DEST_TEMP_MASK);
 //		hw[3] |= (mask << NVFX_VP(INST_VEC_WRITEMASK_SHIFT));
 
diff --git a/src/gallium/drivers/r300/Makefile b/src/gallium/drivers/r300/Makefile
index 13152635a65..728bc40a5bb 100644
--- a/src/gallium/drivers/r300/Makefile
+++ b/src/gallium/drivers/r300/Makefile
@@ -24,6 +24,7 @@ C_SOURCES = \
 	r300_vs.c \
 	r300_vs_draw.c \
 	r300_texture.c \
+	r300_texture_desc.c \
 	r300_tgsi_to_rc.c \
 	r300_transfer.c
 
diff --git a/src/gallium/drivers/r300/SConscript b/src/gallium/drivers/r300/SConscript
index 552ed4e5bef..bf023daaa56 100644
--- a/src/gallium/drivers/r300/SConscript
+++ b/src/gallium/drivers/r300/SConscript
@@ -34,6 +34,7 @@ r300 = env.ConvenienceLibrary(
         'r300_vs.c',
         'r300_vs_draw.c',
         'r300_texture.c',
+        'r300_texture_desc.c',
         'r300_tgsi_to_rc.c',
         'r300_transfer.c',
     ] + r300compiler) + r300compiler
diff --git a/src/gallium/drivers/r300/r300_blit.c b/src/gallium/drivers/r300/r300_blit.c
index 895efaa1c4a..47ffc0cb3c6 100644
--- a/src/gallium/drivers/r300/r300_blit.c
+++ b/src/gallium/drivers/r300/r300_blit.c
@@ -21,7 +21,10 @@
  * USE OR OTHER DEALINGS IN THE SOFTWARE. */
 
 #include "r300_context.h"
+#include "r300_emit.h"
+#include "r300_hyperz.h"
 #include "r300_texture.h"
+#include "r300_winsys.h"
 
 #include "util/u_format.h"
 #include "util/u_pack_color.h"
@@ -81,7 +84,7 @@ static void r300_blitter_end(struct r300_context *r300)
 }
 
 static uint32_t r300_depth_clear_cb_value(enum pipe_format format,
-					  const float* rgba)
+                                          const float* rgba)
 {
     union util_color uc;
     util_pack_color(rgba, format, &uc);
@@ -97,29 +100,29 @@ static boolean r300_cbzb_clear_allowed(struct r300_context *r300,
 {
     struct pipe_framebuffer_state *fb =
         (struct pipe_framebuffer_state*)r300->fb_state.state;
-    struct r300_surface *surf = r300_surface(fb->cbufs[0]);
-    unsigned bpp;
 
     /* Only color clear allowed, and only one colorbuffer. */
     if (clear_buffers != PIPE_CLEAR_COLOR || fb->nr_cbufs != 1)
         return FALSE;
 
-    /* The colorbuffer must be point-sampled. */
-    if (surf->base.texture->nr_samples > 1)
-        return FALSE;
-
-    bpp = util_format_get_blocksizebits(surf->base.format);
+    return r300_surface(fb->cbufs[0])->cbzb_allowed;
+}
 
-    /* ZB can only work with the two pixel sizes. */
-    if (bpp != 16 && bpp != 32)
-        return FALSE;
+static uint32_t r300_depth_clear_value(enum pipe_format format,
+                                       double depth, unsigned stencil)
+{
+    switch (format) {
+        case PIPE_FORMAT_Z16_UNORM:
+        case PIPE_FORMAT_X8Z24_UNORM:
+            return util_pack_z(format, depth);
 
-    /* If the midpoint ZB offset is not aligned to 2048, it returns garbage
-     * with certain texture sizes. Macrotiling ensures the alignment. */
-    if (!r300_texture(surf->base.texture)->mip_macrotile[surf->base.level])
-        return FALSE;
+        case PIPE_FORMAT_S8_USCALED_Z24_UNORM:
+            return util_pack_z_stencil(format, depth, stencil);
 
-    return TRUE;
+        default:
+            assert(0);
+            return 0;
+    }
 }
 
 /* Clear currently bound buffers. */
@@ -169,8 +172,27 @@ static void r300_clear(struct pipe_context* pipe,
         (struct pipe_framebuffer_state*)r300->fb_state.state;
     struct r300_hyperz_state *hyperz =
         (struct r300_hyperz_state*)r300->hyperz_state.state;
+    struct r300_texture *zstex =
+            fb->zsbuf ? r300_texture(fb->zsbuf->texture) : NULL;
     uint32_t width = fb->width;
     uint32_t height = fb->height;
+    boolean has_hyperz = r300->rws->get_value(r300->rws, R300_CAN_HYPERZ);
+    uint32_t hyperz_dcv = hyperz->zb_depthclearvalue;
+
+    /* Enable fast Z clear.
+     * The zbuffer must be in micro-tiled mode, otherwise it locks up. */
+    if ((buffers & PIPE_CLEAR_DEPTHSTENCIL) && has_hyperz) {
+        hyperz_dcv = hyperz->zb_depthclearvalue =
+            r300_depth_clear_value(fb->zsbuf->format, depth, stencil);
+
+        r300_mark_fb_state_dirty(r300, R300_CHANGED_ZCLEAR_FLAG);
+        if (zstex->zmask_mem[fb->zsbuf->level]) {
+            r300->zmask_clear.dirty = TRUE;
+            buffers &= ~PIPE_CLEAR_DEPTHSTENCIL;
+        }
+        if (zstex->hiz_mem[fb->zsbuf->level])
+            r300->hiz_clear.dirty = TRUE;
+    }
 
     /* Enable CBZB clear. */
     if (r300_cbzb_clear_allowed(r300, buffers)) {
@@ -187,20 +209,61 @@ static void r300_clear(struct pipe_context* pipe,
     }
 
     /* Clear. */
-    r300_blitter_begin(r300, R300_CLEAR);
-    util_blitter_clear(r300->blitter,
-                       width,
-                       height,
-                       fb->nr_cbufs,
-                       buffers, rgba, depth, stencil);
-    r300_blitter_end(r300);
+    if (buffers) {
+        /* Clear using the blitter. */
+        r300_blitter_begin(r300, R300_CLEAR);
+        util_blitter_clear(r300->blitter,
+                           width,
+                           height,
+                           fb->nr_cbufs,
+                           buffers, rgba, depth, stencil);
+        r300_blitter_end(r300);
+    } else if (r300->zmask_clear.dirty) {
+        /* Just clear zmask and hiz now, this does not use a standard draw
+         * procedure. */
+        unsigned dwords;
+
+        /* Calculate zmask_clear and hiz_clear atom sizes. */
+        r300_update_hyperz_state(r300);
+        dwords = r300->zmask_clear.size +
+                 (r300->hiz_clear.dirty ? r300->hiz_clear.size : 0) +
+                 r300_get_num_cs_end_dwords(r300);
+
+        /* Reserve CS space. */
+        if (dwords > (r300->cs->ndw - r300->cs->cdw)) {
+            r300->context.flush(&r300->context, 0, NULL);
+        }
+
+        /* Emit clear packets. */
+        r300_emit_zmask_clear(r300, r300->zmask_clear.size,
+                              r300->zmask_clear.state);
+        r300->zmask_clear.dirty = FALSE;
+        if (r300->hiz_clear.dirty) {
+            r300_emit_hiz_clear(r300, r300->hiz_clear.size,
+                                r300->hiz_clear.state);
+            r300->hiz_clear.dirty = FALSE;
+        }
+    } else {
+        assert(0);
+    }
 
     /* Disable CBZB clear. */
     if (r300->cbzb_clear) {
         r300->cbzb_clear = FALSE;
+        hyperz->zb_depthclearvalue = hyperz_dcv;
         r300_mark_fb_state_dirty(r300, R300_CHANGED_CBZB_FLAG);
     }
 
+    /* Enable fastfill and/or hiz.
+     *
+     * If we cleared zmask/hiz, it's in use now. The Hyper-Z state update
+     * looks if zmask/hiz is in use and enables fastfill accordingly. */
+    if (zstex &&
+        (zstex->zmask_in_use[fb->zsbuf->level] ||
+         zstex->hiz_in_use[fb->zsbuf->level])) {
+        r300->hyperz_state.dirty = TRUE;
+    }
+
     /* XXX this flush "fixes" a hardlock in the cubestorm xscreensaver */
     if (r300->flush_counter == 0)
         pipe->flush(pipe, 0, NULL);
@@ -238,6 +301,33 @@ static void r300_clear_depth_stencil(struct pipe_context *pipe,
     r300_blitter_end(r300);
 }
 
+/* Flush a depth stencil buffer. */
+void r300_flush_depth_stencil(struct pipe_context *pipe,
+                              struct pipe_resource *dst,
+                              struct pipe_subresource subdst,
+                              unsigned zslice)
+{
+    struct r300_context *r300 = r300_context(pipe);
+    struct pipe_surface *dstsurf;
+    struct r300_texture *tex = r300_texture(dst);
+
+    if (!tex->zmask_mem[subdst.level])
+        return;
+    if (!tex->zmask_in_use[subdst.level])
+        return;
+
+    dstsurf = pipe->screen->get_tex_surface(pipe->screen, dst,
+                                            subdst.face, subdst.level, zslice,
+                                            PIPE_BIND_DEPTH_STENCIL);
+    r300->z_decomp_rd = TRUE;
+    r300_blitter_begin(r300, R300_CLEAR_SURFACE);
+    util_blitter_flush_depth_stencil(r300->blitter, dstsurf);
+    r300_blitter_end(r300);
+    r300->z_decomp_rd = FALSE;
+
+    tex->zmask_in_use[subdst.level] = FALSE;
+}
+
 /* Copy a block of pixels from one surface to another using HW. */
 static void r300_hw_copy_region(struct pipe_context* pipe,
                                 struct pipe_resource *dst,
@@ -269,7 +359,7 @@ static void r300_resource_copy_region(struct pipe_context *pipe,
 {
     enum pipe_format old_format = dst->format;
     enum pipe_format new_format = old_format;
-
+    boolean is_depth;
     if (!pipe->screen->is_format_supported(pipe->screen,
                                            old_format, src->target,
                                            src->nr_samples,
@@ -296,6 +386,10 @@ static void r300_resource_copy_region(struct pipe_context *pipe,
         }
     }
 
+    is_depth = util_format_get_component_bits(src->format, UTIL_FORMAT_COLORSPACE_ZS, 0) != 0;
+    if (is_depth) {
+        r300_flush_depth_stencil(pipe, src, subsrc, srcz);
+    }
     if (old_format != new_format) {
         dst->format = new_format;
         src->format = new_format;
diff --git a/src/gallium/drivers/r300/r300_chipset.c b/src/gallium/drivers/r300/r300_chipset.c
index 21f3b9d2610..48c24092114 100644
--- a/src/gallium/drivers/r300/r300_chipset.c
+++ b/src/gallium/drivers/r300/r300_chipset.c
@@ -36,7 +36,7 @@ void r300_parse_chipset(struct r300_capabilities* caps)
     caps->num_vert_fpus = 2;
     caps->num_tex_units = 16;
     caps->has_tcl = debug_get_bool_option("RADEON_NO_TCL", FALSE) ? FALSE : TRUE;
-    caps->has_hiz = TRUE;
+    caps->hiz_ram = 0;
     caps->is_r400 = FALSE;
     caps->is_r500 = FALSE;
     caps->high_second_pipe = FALSE;
@@ -49,6 +49,8 @@ void r300_parse_chipset(struct r300_capabilities* caps)
             caps->family = CHIP_FAMILY_R300;
             caps->high_second_pipe = TRUE;
             caps->num_vert_fpus = 4;
+            caps->hiz_ram = R300_HIZ_LIMIT;
+            caps->zmask_ram = PIPE_ZMASK_SIZE;
             break;
 
         case 0x4145:
@@ -61,6 +63,8 @@ void r300_parse_chipset(struct r300_capabilities* caps)
             caps->family = CHIP_FAMILY_R300;
             caps->high_second_pipe = TRUE;
             caps->num_vert_fpus = 4;
+            caps->hiz_ram = R300_HIZ_LIMIT;
+            caps->zmask_ram = PIPE_ZMASK_SIZE;
             break;
 
         case 0x4150:
@@ -77,8 +81,8 @@ void r300_parse_chipset(struct r300_capabilities* caps)
         case 0x4E54:
         case 0x4E56:
             caps->family = CHIP_FAMILY_RV350;
-            caps->has_hiz = FALSE;
             caps->high_second_pipe = TRUE;
+            caps->zmask_ram = RV3xx_ZMASK_SIZE;
             break;
 
         case 0x4148:
@@ -91,12 +95,16 @@ void r300_parse_chipset(struct r300_capabilities* caps)
             caps->family = CHIP_FAMILY_R350;
             caps->high_second_pipe = TRUE;
             caps->num_vert_fpus = 4;
+            caps->hiz_ram = R300_HIZ_LIMIT;
+            caps->zmask_ram = PIPE_ZMASK_SIZE;
             break;
 
         case 0x4E4A:
             caps->family = CHIP_FAMILY_R360;
             caps->high_second_pipe = TRUE;
             caps->num_vert_fpus = 4;
+            caps->hiz_ram = R300_HIZ_LIMIT;
+            caps->zmask_ram = PIPE_ZMASK_SIZE;
             break;
 
         case 0x5460:
@@ -108,8 +116,8 @@ void r300_parse_chipset(struct r300_capabilities* caps)
         case 0x5B64:
         case 0x5B65:
             caps->family = CHIP_FAMILY_RV370;
-            caps->has_hiz = FALSE;
             caps->high_second_pipe = TRUE;
+            caps->zmask_ram = RV3xx_ZMASK_SIZE;
             break;
 
         case 0x3150:
@@ -120,6 +128,8 @@ void r300_parse_chipset(struct r300_capabilities* caps)
         case 0x3E54:
             caps->family = CHIP_FAMILY_RV380;
             caps->high_second_pipe = TRUE;
+            caps->hiz_ram = R300_HIZ_LIMIT;
+            caps->zmask_ram = RV3xx_ZMASK_SIZE;
             break;
 
         case 0x4A48:
@@ -135,6 +145,8 @@ void r300_parse_chipset(struct r300_capabilities* caps)
             caps->family = CHIP_FAMILY_R420;
             caps->num_vert_fpus = 6;
             caps->is_r400 = TRUE;
+            caps->hiz_ram = R300_HIZ_LIMIT;
+            caps->zmask_ram = PIPE_ZMASK_SIZE;
             break;
 
         case 0x5548:
@@ -149,6 +161,8 @@ void r300_parse_chipset(struct r300_capabilities* caps)
             caps->family = CHIP_FAMILY_R423;
             caps->num_vert_fpus = 6;
             caps->is_r400 = TRUE;
+            caps->hiz_ram = R300_HIZ_LIMIT;
+            caps->zmask_ram = PIPE_ZMASK_SIZE;
             break;
 
         case 0x554C:
@@ -161,6 +175,8 @@ void r300_parse_chipset(struct r300_capabilities* caps)
             caps->family = CHIP_FAMILY_R430;
             caps->num_vert_fpus = 6;
             caps->is_r400 = TRUE;
+            caps->hiz_ram = R300_HIZ_LIMIT;
+            caps->zmask_ram = PIPE_ZMASK_SIZE;
             break;
 
         case 0x5D4C:
@@ -172,6 +188,8 @@ void r300_parse_chipset(struct r300_capabilities* caps)
             caps->family = CHIP_FAMILY_R480;
             caps->num_vert_fpus = 6;
             caps->is_r400 = TRUE;
+            caps->hiz_ram = R300_HIZ_LIMIT;
+            caps->zmask_ram = PIPE_ZMASK_SIZE;
             break;
 
         case 0x4B48:
@@ -182,6 +200,8 @@ void r300_parse_chipset(struct r300_capabilities* caps)
             caps->family = CHIP_FAMILY_R481;
             caps->num_vert_fpus = 6;
             caps->is_r400 = TRUE;
+            caps->hiz_ram = R300_HIZ_LIMIT;
+            caps->zmask_ram = PIPE_ZMASK_SIZE;
             break;
 
         case 0x5E4C:
@@ -199,34 +219,36 @@ void r300_parse_chipset(struct r300_capabilities* caps)
             caps->family = CHIP_FAMILY_RV410;
             caps->num_vert_fpus = 6;
             caps->is_r400 = TRUE;
+            caps->hiz_ram = R300_HIZ_LIMIT;
+            caps->zmask_ram = PIPE_ZMASK_SIZE;
             break;
 
         case 0x5954:
         case 0x5955:
             caps->family = CHIP_FAMILY_RS480;
-            caps->has_hiz = FALSE;
             caps->has_tcl = FALSE;
+            caps->zmask_ram = RV3xx_ZMASK_SIZE;
             break;
 
         case 0x5974:
         case 0x5975:
             caps->family = CHIP_FAMILY_RS482;
-            caps->has_hiz = FALSE;
             caps->has_tcl = FALSE;
+            caps->zmask_ram = RV3xx_ZMASK_SIZE;
             break;
 
         case 0x5A41:
         case 0x5A42:
             caps->family = CHIP_FAMILY_RS400;
-            caps->has_hiz = FALSE;
             caps->has_tcl = FALSE;
+            caps->zmask_ram = RV3xx_ZMASK_SIZE;
             break;
 
         case 0x5A61:
         case 0x5A62:
             caps->family = CHIP_FAMILY_RC410;
-            caps->has_hiz = FALSE;
             caps->has_tcl = FALSE;
+            caps->zmask_ram = RV3xx_ZMASK_SIZE;
             break;
 
         case 0x791E:
@@ -234,6 +256,8 @@ void r300_parse_chipset(struct r300_capabilities* caps)
             caps->family = CHIP_FAMILY_RS690;
             caps->has_tcl = FALSE;
             caps->is_r400 = TRUE;
+            caps->hiz_ram = R300_HIZ_LIMIT;
+            caps->zmask_ram = PIPE_ZMASK_SIZE;
             break;
 
         case 0x793F:
@@ -242,6 +266,8 @@ void r300_parse_chipset(struct r300_capabilities* caps)
             caps->family = CHIP_FAMILY_RS600;
             caps->has_tcl = FALSE;
             caps->is_r400 = TRUE;
+            caps->hiz_ram = R300_HIZ_LIMIT;
+            caps->zmask_ram = PIPE_ZMASK_SIZE;
             break;
 
         case 0x796C:
@@ -251,6 +277,8 @@ void r300_parse_chipset(struct r300_capabilities* caps)
             caps->family = CHIP_FAMILY_RS740;
             caps->has_tcl = FALSE;
             caps->is_r400 = TRUE;
+            caps->hiz_ram = R300_HIZ_LIMIT;
+            caps->zmask_ram = PIPE_ZMASK_SIZE;
             break;
 
         case 0x7100:
@@ -270,6 +298,8 @@ void r300_parse_chipset(struct r300_capabilities* caps)
             caps->family = CHIP_FAMILY_R520;
             caps->num_vert_fpus = 8;
             caps->is_r500 = TRUE;
+            caps->hiz_ram = R300_HIZ_LIMIT;
+            caps->zmask_ram = PIPE_ZMASK_SIZE;
             break;
 
         case 0x7140:
@@ -313,6 +343,8 @@ void r300_parse_chipset(struct r300_capabilities* caps)
             caps->family = CHIP_FAMILY_RV515;
             caps->num_vert_fpus = 2;
             caps->is_r500 = TRUE;
+            caps->hiz_ram = R300_HIZ_LIMIT;
+            caps->zmask_ram = PIPE_ZMASK_SIZE;
             break;
 
         case 0x71C0:
@@ -334,6 +366,8 @@ void r300_parse_chipset(struct r300_capabilities* caps)
             caps->family = CHIP_FAMILY_RV530;
             caps->num_vert_fpus = 5;
             caps->is_r500 = TRUE;
+            /*caps->hiz_ram = RV530_HIZ_LIMIT;*/
+            caps->zmask_ram = PIPE_ZMASK_SIZE;
             break;
 
         case 0x7240:
@@ -354,12 +388,16 @@ void r300_parse_chipset(struct r300_capabilities* caps)
             caps->family = CHIP_FAMILY_R580;
             caps->num_vert_fpus = 8;
             caps->is_r500 = TRUE;
+            caps->hiz_ram = RV530_HIZ_LIMIT;
+            caps->zmask_ram = PIPE_ZMASK_SIZE;
             break;
 
         case 0x7280:
             caps->family = CHIP_FAMILY_RV570;
             caps->num_vert_fpus = 8;
             caps->is_r500 = TRUE;
+            caps->hiz_ram = RV530_HIZ_LIMIT;
+            caps->zmask_ram = PIPE_ZMASK_SIZE;
             break;
 
         case 0x7281:
@@ -376,6 +414,8 @@ void r300_parse_chipset(struct r300_capabilities* caps)
             caps->family = CHIP_FAMILY_RV560;
             caps->num_vert_fpus = 8;
             caps->is_r500 = TRUE;
+            caps->hiz_ram = RV530_HIZ_LIMIT;
+            caps->zmask_ram = PIPE_ZMASK_SIZE;
             break;
 
         default:
diff --git a/src/gallium/drivers/r300/r300_chipset.h b/src/gallium/drivers/r300/r300_chipset.h
index 65750f54e71..e7ca642b4f3 100644
--- a/src/gallium/drivers/r300/r300_chipset.h
+++ b/src/gallium/drivers/r300/r300_chipset.h
@@ -25,6 +25,14 @@
 
 #include "pipe/p_compiler.h"
 
+/* these are sizes in dwords */
+#define R300_HIZ_LIMIT 10240
+#define RV530_HIZ_LIMIT 15360
+
+/* rv3xx have only one pipe */
+#define PIPE_ZMASK_SIZE 4096
+#define RV3xx_ZMASK_SIZE 5120
+
 /* Structure containing all the possible information about a specific Radeon
  * in the R3xx, R4xx, and R5xx families. */
 struct r300_capabilities {
@@ -42,8 +50,10 @@ struct r300_capabilities {
     unsigned num_tex_units;
     /* Whether or not TCL is physically present */
     boolean has_tcl;
-    /* Some chipsets do not have HiZ RAM. */
-    boolean has_hiz;
+    /* Some chipsets do not have HiZ RAM - other have varying amounts . */
+    int hiz_ram;
+    /*  some chipsets have zmask ram per pipe some don't */
+    int zmask_ram;
     /* Whether or not this is RV350 or newer, including all r400 and r500
      * chipsets. The differences compared to the oldest r300 chips are:
      * - Blend LTE/GTE thresholds
diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c
index df903590583..e8b6c4f7af8 100644
--- a/src/gallium/drivers/r300/r300_context.c
+++ b/src/gallium/drivers/r300/r300_context.c
@@ -30,6 +30,7 @@
 #include "r300_cb.h"
 #include "r300_context.h"
 #include "r300_emit.h"
+#include "r300_hyperz.h"
 #include "r300_screen.h"
 #include "r300_screen_buffer.h"
 #include "r300_winsys.h"
@@ -114,6 +115,10 @@ static void r300_destroy_context(struct pipe_context* context)
     u_upload_destroy(r300->upload_vb);
     u_upload_destroy(r300->upload_ib);
 
+    /* setup hyper-z mm */
+    if (r300->rws->get_value(r300->rws, R300_CAN_HYPERZ))
+        r300_hyperz_destroy_mm(r300);
+
     translate_cache_destroy(r300->tran.translate_cache);
 
     r300_release_referenced_objects(r300);
@@ -166,6 +171,9 @@ static void r300_setup_atoms(struct r300_context* r300)
     boolean is_r500 = r300->screen->caps.is_r500;
     boolean has_tcl = r300->screen->caps.has_tcl;
     boolean drm_2_3_0 = r300->rws->get_value(r300->rws, R300_VID_DRM_2_3_0);
+    boolean drm_2_6_0 = r300->rws->get_value(r300->rws, R300_VID_DRM_2_6_0);
+    boolean has_hyperz = r300->rws->get_value(r300->rws, R300_CAN_HYPERZ);
+    boolean has_hiz_ram = r300->screen->caps.hiz_ram > 0;
 
     /* Create the actual atom list.
      *
@@ -188,8 +196,8 @@ static void r300_setup_atoms(struct r300_context* r300)
     R300_INIT_ATOM(gpu_flush, 9);
     R300_INIT_ATOM(aa_state, 4);
     R300_INIT_ATOM(fb_state, 0);
+    R300_INIT_ATOM(hyperz_state, is_r500 || (is_rv350 && drm_2_6_0) ? 10 : 8);
     /* ZB (unpipelined), SC. */
-    R300_INIT_ATOM(hyperz_state, 6);
     R300_INIT_ATOM(ztop_state, 2);
     /* ZB, FG. */
     R300_INIT_ATOM(dsa_state, is_r500 ? 8 : 6);
@@ -220,6 +228,13 @@ static void r300_setup_atoms(struct r300_context* r300)
     /* TX. */
     R300_INIT_ATOM(texture_cache_inval, 2);
     R300_INIT_ATOM(textures_state, 0);
+    if (has_hyperz) {
+        /* HiZ Clear */
+        if (has_hiz_ram)
+            R300_INIT_ATOM(hiz_clear, 0);
+        /* zmask clear */
+        R300_INIT_ATOM(zmask_clear, 0);
+    }
     /* ZB (unpipelined), SU. */
     R300_INIT_ATOM(query_start, 4);
 
@@ -282,8 +297,7 @@ static void r300_init_states(struct pipe_context *pipe)
             (struct r300_vap_invariant_state*)r300->vap_invariant_state.state;
     struct r300_invariant_state *invariant =
             (struct r300_invariant_state*)r300->invariant_state.state;
-    struct r300_hyperz_state *hyperz =
-            (struct r300_hyperz_state*)r300->hyperz_state.state;
+
     CB_LOCALS;
 
     pipe->set_blend_color(pipe, &bc);
@@ -351,10 +365,20 @@ static void r300_init_states(struct pipe_context *pipe)
 
     /* Initialize the hyperz state. */
     {
-        BEGIN_CB(&hyperz->cb_begin, r300->hyperz_state.size);
+        struct r300_hyperz_state *hyperz =
+            (struct r300_hyperz_state*)r300->hyperz_state.state;
+        BEGIN_CB(&hyperz->cb_flush_begin, r300->hyperz_state.size);
+        OUT_CB_REG(R300_ZB_ZCACHE_CTLSTAT,
+                   R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE);
         OUT_CB_REG(R300_ZB_BW_CNTL, 0);
         OUT_CB_REG(R300_ZB_DEPTHCLEARVALUE, 0);
         OUT_CB_REG(R300_SC_HYPERZ, R300_SC_HYPERZ_ADJ_2);
+
+        if (r300->screen->caps.is_r500 ||
+            (r300->screen->caps.is_rv350 &&
+             r300->rws->get_value(r300->rws, R300_VID_DRM_2_6_0))) {
+            OUT_CB_REG(R300_GB_Z_PEQ_CONFIG, 0);
+        }
         END_CB;
     }
 }
@@ -415,6 +439,10 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen,
 
     rws->cs_set_flush(r300->cs, r300_flush_cb, r300);
 
+    /* setup hyper-z mm */
+    if (r300->rws->get_value(r300->rws, R300_CAN_HYPERZ))
+        r300_hyperz_init_mm(r300);
+
     r300->upload_ib = u_upload_create(&r300->context,
 				      32 * 1024, 16,
 				      PIPE_BIND_INDEX_BUFFER);
diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h
index b9c96d5bdd8..6fa7f470f98 100644
--- a/src/gallium/drivers/r300/r300_context.h
+++ b/src/gallium/drivers/r300/r300_context.h
@@ -106,13 +106,19 @@ struct r300_dsa_state {
 };
 
 struct r300_hyperz_state {
+    int current_func; /* -1 after a clear before first op */
+    int flush;
     /* This is actually a command buffer with named dwords. */
+    uint32_t cb_flush_begin;
+    uint32_t zb_zcache_ctlstat;     /* R300_ZB_CACHE_CNTL */
     uint32_t cb_begin;
     uint32_t zb_bw_cntl;            /* R300_ZB_BW_CNTL */
     uint32_t cb_reg1;
     uint32_t zb_depthclearvalue;    /* R300_ZB_DEPTHCLEARVALUE */
     uint32_t cb_reg2;
     uint32_t sc_hyperz;             /* R300_SC_HYPERZ */
+    uint32_t cb_reg3;
+    uint32_t gb_z_peq_config;       /* R300_GB_Z_PEQ_CONFIG: 0x4028 */
 };
 
 struct r300_gpu_flush {
@@ -318,29 +324,39 @@ struct r300_surface {
     uint32_t cbzb_midpoint_offset;  /* DEPTHOFFSET. */
     uint32_t cbzb_pitch;            /* DEPTHPITCH. */
     uint32_t cbzb_format;           /* ZB_FORMAT. */
+
+    /* Whether the CBZB clear is allowed on the surface. */
+    boolean cbzb_allowed;
+
 };
 
-struct r300_texture {
-    /* Parent class */
+struct r300_texture_desc {
+    /* Parent class. */
     struct u_resource b;
 
-    enum r300_buffer_domain domain;
+    /* Buffer tiling.
+     * Macrotiling is specified per-level because small mipmaps cannot
+     * be macrotiled. */
+    enum r300_buffer_tiling microtile;
+    enum r300_buffer_tiling macrotile[R300_MAX_TEXTURE_LEVELS];
 
     /* Offsets into the buffer. */
-    unsigned offset[R300_MAX_TEXTURE_LEVELS];
+    unsigned offset_in_bytes[R300_MAX_TEXTURE_LEVELS];
 
-    /* A pitch for each mip-level */
-    unsigned pitch[R300_MAX_TEXTURE_LEVELS];
+    /* Strides for each mip-level. */
+    unsigned stride_in_pixels[R300_MAX_TEXTURE_LEVELS];
+    unsigned stride_in_bytes[R300_MAX_TEXTURE_LEVELS];
 
-    /* A pitch multiplied by blockwidth as hardware wants
-     * the number of pixels instead of the number of blocks. */
-    unsigned hwpitch[R300_MAX_TEXTURE_LEVELS];
+    /* Size of one zslice or face or 2D image based on the texture target. */
+    unsigned layer_size_in_bytes[R300_MAX_TEXTURE_LEVELS];
 
-    /* Size of one zslice or face based on the texture target */
-    unsigned layer_size[R300_MAX_TEXTURE_LEVELS];
+    /* Total size of this texture, in bytes,
+     * derived from the texture properties. */
+    unsigned size_in_bytes;
 
-    /* Whether the mipmap level is macrotiled. */
-    enum r300_buffer_tiling mip_macrotile[R300_MAX_TEXTURE_LEVELS];
+    /* Total size of the buffer backing this texture, in bytes.
+     * It must be >= size. */
+    unsigned buffer_size_in_bytes;
 
     /**
      * If non-zero, override the natural texture layout with
@@ -350,16 +366,24 @@ struct r300_texture {
      *
      * \sa r300_texture_get_stride
      */
-    unsigned stride_override;
+    unsigned stride_in_bytes_override;
 
-    /* Total size of this texture, in bytes. */
-    unsigned size;
+    /* Whether this texture has non-power-of-two dimensions.
+     * It can be either a regular texture or a rectangle one. */
+    boolean is_npot;
 
-    /* Whether this texture has non-power-of-two dimensions
-     * or a user-specified pitch.
-     * It can be either a regular texture or a rectangle one.
-     */
-    boolean uses_pitch;
+    /* This flag says that hardware must use the stride for addressing
+     * instead of the width. */
+    boolean uses_stride_addressing;
+
+    /* Whether CBZB fast color clear is allowed on the miplevel. */
+    boolean cbzb_allowed[R300_MAX_TEXTURE_LEVELS];
+};
+
+struct r300_texture {
+    struct r300_texture_desc desc;
+
+    enum r300_buffer_domain domain;
 
     /* Pipe buffer backing this texture. */
     struct r300_winsys_buffer *buffer;
@@ -370,8 +394,11 @@ struct r300_texture {
     /* All bits should be filled in. */
     struct r300_texture_fb_state fb_state;
 
-    /* Buffer tiling */
-    enum r300_buffer_tiling microtile, macrotile;
+    /* hyper-z memory allocs */
+    struct mem_block *hiz_mem[R300_MAX_TEXTURE_LEVELS];
+    struct mem_block *zmask_mem[R300_MAX_TEXTURE_LEVELS];
+    boolean zmask_in_use[R300_MAX_TEXTURE_LEVELS];
+    boolean hiz_in_use[R300_MAX_TEXTURE_LEVELS];
 
     /* This is the level tiling flags were last time set for.
      * It's used to prevent redundant tiling-flags changes from happening.*/
@@ -498,6 +525,10 @@ struct r300_context {
     struct r300_atom texture_cache_inval;
     /* GPU flush. */
     struct r300_atom gpu_flush;
+    /* HiZ clear */
+    struct r300_atom hiz_clear;
+    /* zmask clear */
+    struct r300_atom zmask_clear;
 
     /* Invariant state. This must be emitted to get the engine started. */
     struct r300_atom invariant_state;
@@ -510,6 +541,8 @@ struct r300_context {
     struct r300_vertex_element_state *velems;
     bool any_user_vbs;
 
+    struct pipe_index_buffer index_buffer;
+
     /* Vertex info for Draw. */
     struct vertex_info vertex_info;
 
@@ -533,8 +566,16 @@ struct r300_context {
     boolean two_sided_color;
     /* Incompatible vertex buffer layout? (misaligned stride or buffer_offset) */
     boolean incompatible_vb_layout;
-
+#define R300_Z_COMPRESS_44 1
+#define RV350_Z_COMPRESS_88 2
+    int z_compression;
     boolean cbzb_clear;
+    boolean z_decomp_rd;
+
+    /* two mem block managers for hiz/zmask ram space */
+    struct mem_block *hiz_mm;
+    struct mem_block *zmask_mm;
+
     /* upload managers */
     struct u_upload_mgr *upload_vb;
     struct u_upload_mgr *upload_ib;
@@ -586,6 +627,12 @@ void r300_init_render_functions(struct r300_context *r300);
 void r300_init_state_functions(struct r300_context* r300);
 void r300_init_resource_functions(struct r300_context* r300);
 
+/* r300_blit.c */
+void r300_flush_depth_stencil(struct pipe_context *pipe,
+                              struct pipe_resource *dst,
+                              struct pipe_subresource subdst,
+                              unsigned zslice);
+
 /* r300_query.c */
 void r300_resume_query(struct r300_context *r300,
                        struct r300_query *query);
@@ -605,7 +652,8 @@ void r300_plug_in_stencil_ref_fallback(struct r300_context *r300);
 /* r300_state.c */
 enum r300_fb_state_change {
     R300_CHANGED_FB_STATE = 0,
-    R300_CHANGED_CBZB_FLAG
+    R300_CHANGED_CBZB_FLAG,
+    R300_CHANGED_ZCLEAR_FLAG
 };
 
 void r300_mark_fb_state_dirty(struct r300_context *r300,
diff --git a/src/gallium/drivers/r300/r300_cs.h b/src/gallium/drivers/r300/r300_cs.h
index 3beb625d430..c194d6a1b08 100644
--- a/src/gallium/drivers/r300/r300_cs.h
+++ b/src/gallium/drivers/r300/r300_cs.h
@@ -136,8 +136,8 @@
 
 #define WRITE_CS_TABLE(values, count) do { \
     CS_DEBUG(assert(cs_count == 0);) \
-    memcpy(cs_copy->ptr + cs_copy->cdw, values, count * 4); \
-    cs_copy->cdw += count; \
+    memcpy(cs_copy->ptr + cs_copy->cdw, (values), (count) * 4); \
+    cs_copy->cdw += (count); \
 } while (0)
 
 #endif /* R300_CS_H */
diff --git a/src/gallium/drivers/r300/r300_debug.c b/src/gallium/drivers/r300/r300_debug.c
index 31d4e14681b..c3e157e99af 100644
--- a/src/gallium/drivers/r300/r300_debug.c
+++ b/src/gallium/drivers/r300/r300_debug.c
@@ -38,11 +38,13 @@ static const struct debug_named_value debug_options[] = {
     { "fall", DBG_FALL, "Fallbacks (for debugging)" },
     { "rs", DBG_RS, "Rasterizer (for debugging)" },
     { "fb", DBG_FB, "Framebuffer (for debugging)" },
+    { "cbzb", DBG_CBZB, "Fast color clear info (for debugging)" },
     { "fakeocc", DBG_FAKE_OCC, "Use fake occlusion queries (for debugging)" },
     { "anisohq", DBG_ANISOHQ, "High quality anisotropic filtering (for benchmarking)" },
     { "notiling", DBG_NO_TILING, "Disable tiling (for benchmarking)" },
     { "noimmd", DBG_NO_IMMD, "Disable immediate mode (for benchmarking)" },
     { "stats", DBG_STATS, "Gather statistics" },
+    { "hyperz", DBG_HYPERZ, "HyperZ (for debugging)" },
 
     /* must be last */
     DEBUG_NAMED_VALUE_END
diff --git a/src/gallium/drivers/r300/r300_defines.h b/src/gallium/drivers/r300/r300_defines.h
index d510d80a7bb..896aeef395d 100644
--- a/src/gallium/drivers/r300/r300_defines.h
+++ b/src/gallium/drivers/r300/r300_defines.h
@@ -36,7 +36,10 @@
 enum r300_buffer_tiling {
     R300_BUFFER_LINEAR = 0,
     R300_BUFFER_TILED,
-    R300_BUFFER_SQUARETILED
+    R300_BUFFER_SQUARETILED,
+
+    R300_BUFFER_UNKNOWN,
+    R300_BUFFER_SELECT_LAYOUT = R300_BUFFER_UNKNOWN
 };
 
 enum r300_buffer_domain { /* bitfield */
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 36a26a78717..d0fd45349e3 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -25,6 +25,7 @@
 
 #include "util/u_format.h"
 #include "util/u_math.h"
+#include "util/u_mm.h"
 #include "util/u_simple_list.h"
 
 #include "r300_context.h"
@@ -329,6 +330,7 @@ void r300_emit_fb_state(struct r300_context* r300, unsigned size, void* state)
     struct pipe_framebuffer_state* fb = (struct pipe_framebuffer_state*)state;
     struct r300_surface* surf;
     unsigned i;
+    boolean has_hyperz = r300->rws->get_value(r300->rws, R300_CAN_HYPERZ);
     CS_LOCALS(r300);
 
     BEGIN_CS(size);
@@ -364,6 +366,10 @@ void r300_emit_fb_state(struct r300_context* r300, unsigned size, void* state)
 
         OUT_CS_REG_SEQ(R300_ZB_DEPTHPITCH, 1);
         OUT_CS_RELOC(surf->buffer, surf->cbzb_pitch, 0, surf->domain);
+
+        DBG(r300, DBG_CBZB,
+            "CBZB clearing cbuf %08x %08x\n", surf->cbzb_format,
+            surf->cbzb_pitch);
     }
     /* Set up a zbuffer. */
     else if (fb->zsbuf) {
@@ -377,15 +383,32 @@ void r300_emit_fb_state(struct r300_context* r300, unsigned size, void* state)
         OUT_CS_REG_SEQ(R300_ZB_DEPTHPITCH, 1);
         OUT_CS_RELOC(surf->buffer, surf->pitch, 0, surf->domain);
 
-        /* HiZ RAM. */
-        if (r300->screen->caps.has_hiz) {
-            OUT_CS_REG(R300_ZB_HIZ_OFFSET, 0);
-            OUT_CS_REG(R300_ZB_HIZ_PITCH, 0);
+        if (has_hyperz) {
+            uint32_t surf_pitch;
+            struct r300_texture *tex;
+            int level = surf->base.level;
+            tex = r300_texture(surf->base.texture);
+
+            surf_pitch = surf->pitch & R300_DEPTHPITCH_MASK;
+            /* HiZ RAM. */
+            if (r300->screen->caps.hiz_ram) {
+                if (tex->hiz_mem[level]) {
+                    OUT_CS_REG(R300_ZB_HIZ_OFFSET, tex->hiz_mem[level]->ofs << 2);
+                    OUT_CS_REG(R300_ZB_HIZ_PITCH, surf_pitch);
+                } else {
+                    OUT_CS_REG(R300_ZB_HIZ_OFFSET, 0);
+                    OUT_CS_REG(R300_ZB_HIZ_PITCH, 0);
+                }
+            }
+            /* Z Mask RAM. (compressed zbuffer) */
+            if (tex->zmask_mem[level]) {
+                OUT_CS_REG(R300_ZB_ZMASK_OFFSET, tex->zmask_mem[level]->ofs << 2);
+                OUT_CS_REG(R300_ZB_ZMASK_PITCH, surf_pitch);
+            } else {
+                OUT_CS_REG(R300_ZB_ZMASK_OFFSET, 0);
+                OUT_CS_REG(R300_ZB_ZMASK_PITCH, 0);
+            }
         }
-
-        /* Z Mask RAM. (compressed zbuffer) */
-        OUT_CS_REG(R300_ZB_ZMASK_OFFSET, 0);
-        OUT_CS_REG(R300_ZB_ZMASK_PITCH, 0);
     }
 
     END_CS;
@@ -394,8 +417,12 @@ void r300_emit_fb_state(struct r300_context* r300, unsigned size, void* state)
 void r300_emit_hyperz_state(struct r300_context *r300,
                             unsigned size, void *state)
 {
+    struct r300_hyperz_state *z = state;
     CS_LOCALS(r300);
-    WRITE_CS_TABLE(state, size);
+    if (z->flush)
+        WRITE_CS_TABLE(&z->cb_flush_begin, size);
+    else
+        WRITE_CS_TABLE(&z->cb_begin, size - 2);
 }
 
 void r300_emit_hyperz_end(struct r300_context *r300)
@@ -403,9 +430,11 @@ void r300_emit_hyperz_end(struct r300_context *r300)
     struct r300_hyperz_state z =
             *(struct r300_hyperz_state*)r300->hyperz_state.state;
 
+    z.flush = 1;
     z.zb_bw_cntl = 0;
     z.zb_depthclearvalue = 0;
     z.sc_hyperz = R300_SC_HYPERZ_ADJ_2;
+    z.gb_z_peq_config = 0;
 
     r300_emit_hyperz_state(r300, r300->hyperz_state.size, &z);
 }
@@ -907,6 +936,22 @@ void r300_emit_vs_state(struct r300_context* r300, unsigned size, void* state)
             OUT_CS_TABLE(data, 4);
         }
     }
+
+    /* Emit flow control instructions. */
+    if (code->num_fc_ops) {
+
+        OUT_CS_REG(R300_VAP_PVS_FLOW_CNTL_OPC, code->fc_ops);
+        if (r300screen->caps.is_r500) {
+            OUT_CS_REG_SEQ(R500_VAP_PVS_FLOW_CNTL_ADDRS_LW_0, code->num_fc_ops * 2);
+            OUT_CS_TABLE(code->fc_op_addrs.r500, code->num_fc_ops * 2);
+        } else {
+            OUT_CS_REG_SEQ(R300_VAP_PVS_FLOW_CNTL_ADDRS_0, code->num_fc_ops);
+            OUT_CS_TABLE(code->fc_op_addrs.r300, code->num_fc_ops);
+        }
+        OUT_CS_REG_SEQ(R300_VAP_PVS_FLOW_CNTL_LOOP_INDEX_0, code->num_fc_ops);
+        OUT_CS_TABLE(code->fc_loop_index, code->num_fc_ops);
+    }
+
     END_CS;
 }
 
@@ -943,6 +988,111 @@ void r300_emit_viewport_state(struct r300_context* r300,
     END_CS;
 }
 
+static void r300_emit_hiz_line_clear(struct r300_context *r300, int start, uint16_t count, uint32_t val)
+{
+    CS_LOCALS(r300);
+    BEGIN_CS(4);
+    OUT_CS_PKT3(R300_PACKET3_3D_CLEAR_HIZ, 2);
+    OUT_CS(start);
+    OUT_CS(count);
+    OUT_CS(val);
+    END_CS;
+}
+
+static void r300_emit_zmask_line_clear(struct r300_context *r300, int start, uint16_t count, uint32_t val)
+{
+    CS_LOCALS(r300);
+    BEGIN_CS(4);
+    OUT_CS_PKT3(R300_PACKET3_3D_CLEAR_ZMASK, 2);
+    OUT_CS(start);
+    OUT_CS(count);
+    OUT_CS(val);
+    END_CS;
+}
+
+#define ALIGN_DIVUP(x, y) (((x) + (y) - 1) / (y))
+
+void r300_emit_hiz_clear(struct r300_context *r300, unsigned size, void *state)
+{
+    struct pipe_framebuffer_state *fb =
+        (struct pipe_framebuffer_state*)r300->fb_state.state;
+    struct r300_hyperz_state *z =
+        (struct r300_hyperz_state*)r300->hyperz_state.state;
+    struct r300_screen* r300screen = r300->screen;
+    uint32_t stride, offset = 0, height, offset_shift;
+    struct r300_texture* tex;
+    int i;
+
+    tex = r300_texture(fb->zsbuf->texture);
+
+    offset = tex->hiz_mem[fb->zsbuf->level]->ofs;
+    stride = tex->desc.stride_in_pixels[fb->zsbuf->level];
+
+    /* convert from pixels to 4x4 blocks */
+    stride = ALIGN_DIVUP(stride, 4);
+
+    stride = ALIGN_DIVUP(stride, r300screen->caps.num_frag_pipes);    
+    /* there are 4 blocks per dwords */
+    stride = ALIGN_DIVUP(stride, 4);
+
+    height = ALIGN_DIVUP(fb->zsbuf->height, 4);
+
+    offset_shift = 2;
+    offset_shift += (r300screen->caps.num_frag_pipes / 2);
+
+    for (i = 0; i < height; i++) {
+        offset = i * stride;
+        offset <<= offset_shift;
+        r300_emit_hiz_line_clear(r300, offset, stride, 0xffffffff);
+    }
+    z->current_func = -1;
+
+    /* Mark the current zbuffer's hiz ram as in use. */
+    tex->hiz_in_use[fb->zsbuf->level] = TRUE;
+}
+
+void r300_emit_zmask_clear(struct r300_context *r300, unsigned size, void *state)
+{
+    struct pipe_framebuffer_state *fb =
+        (struct pipe_framebuffer_state*)r300->fb_state.state;
+    struct r300_screen* r300screen = r300->screen;
+    uint32_t stride, offset = 0;
+    struct r300_texture* tex;
+    uint32_t i, height;
+    int mult, offset_shift;
+
+    tex = r300_texture(fb->zsbuf->texture);
+    stride = tex->desc.stride_in_pixels[fb->zsbuf->level];
+
+    offset = tex->zmask_mem[fb->zsbuf->level]->ofs;
+
+    if (r300->z_compression == RV350_Z_COMPRESS_88)
+        mult = 8;
+    else
+        mult = 4;
+
+    height = ALIGN_DIVUP(fb->zsbuf->height, mult);
+
+    offset_shift = 4;
+    offset_shift += (r300screen->caps.num_frag_pipes / 2);
+    stride = ALIGN_DIVUP(stride, r300screen->caps.num_frag_pipes);
+
+    /* okay have width in pixels - divide by block width */
+    stride = ALIGN_DIVUP(stride, mult);
+    /* have width in blocks - divide by number of fragment pipes screen width */
+    /* 16 blocks per dword */
+    stride = ALIGN_DIVUP(stride, 16);
+
+    for (i = 0; i < height; i++) {
+        offset = i * stride;
+        offset <<= offset_shift;
+        r300_emit_zmask_line_clear(r300, offset, stride, 0x0);//0xffffffff);
+    }
+
+    /* Mark the current zbuffer's zmask as in use. */
+    tex->zmask_in_use[fb->zsbuf->level] = TRUE;
+}
+
 void r300_emit_ztop_state(struct r300_context* r300,
                           unsigned size, void* state)
 {
@@ -1062,6 +1212,17 @@ unsigned r300_get_num_dirty_dwords(struct r300_context *r300)
     return dwords;
 }
 
+unsigned r300_get_num_cs_end_dwords(struct r300_context *r300)
+{
+    unsigned dwords = 0;
+
+    /* Emitted in flush. */
+    dwords += 26; /* emit_query_end */
+    dwords += r300->hyperz_state.size + 2; /* emit_hyperz_end + zcache flush */
+
+    return dwords;
+}
+
 /* Emit all dirty state. */
 void r300_emit_dirty_state(struct r300_context* r300)
 {
diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h
index 5d05039669f..bae25256346 100644
--- a/src/gallium/drivers/r300/r300_emit.h
+++ b/src/gallium/drivers/r300/r300_emit.h
@@ -112,7 +112,11 @@ void r300_emit_texture_cache_inval(struct r300_context* r300, unsigned size, voi
 void r300_emit_invariant_state(struct r300_context *r300,
                                unsigned size, void *state);
 
+void r300_emit_hiz_clear(struct r300_context *r300, unsigned size, void *state);
+void r300_emit_zmask_clear(struct r300_context *r300, unsigned size, void *state);
+
 unsigned r300_get_num_dirty_dwords(struct r300_context *r300);
+unsigned r300_get_num_cs_end_dwords(struct r300_context *r300);
 
 /* Emit all dirty state. */
 void r300_emit_dirty_state(struct r300_context* r300);
diff --git a/src/gallium/drivers/r300/r300_flush.c b/src/gallium/drivers/r300/r300_flush.c
index ae7b5759e78..fe182b6615b 100644
--- a/src/gallium/drivers/r300/r300_flush.c
+++ b/src/gallium/drivers/r300/r300_flush.c
@@ -43,14 +43,6 @@ static void r300_flush(struct pipe_context* pipe,
     u_upload_flush(r300->upload_vb);
     u_upload_flush(r300->upload_ib);
 
-    /* We probably need to flush Draw, but we may have been called from
-     * within Draw. This feels kludgy, but it might be the best thing.
-     *
-     * Of course, the best thing is to kill Draw with fire. :3 */
-    if (r300->draw && !r300->draw->flushing) {
-        draw_flush(r300->draw);
-    }
-
     if (r300->dirty_hw) {
         r300_emit_hyperz_end(r300);
         r300_emit_query_end(r300);
diff --git a/src/gallium/drivers/r300/r300_fs.c b/src/gallium/drivers/r300/r300_fs.c
index b145ded6399..2a0c30620ad 100644
--- a/src/gallium/drivers/r300/r300_fs.c
+++ b/src/gallium/drivers/r300/r300_fs.c
@@ -72,6 +72,11 @@ void r300_shader_read_fs_inputs(struct tgsi_shader_info* info,
                 fs_inputs->wpos = i;
                 break;
 
+            case TGSI_SEMANTIC_FACE:
+                assert(index == 0);
+                fs_inputs->face = i;
+                break;
+
             default:
                 fprintf(stderr, "r300: FP: Unknown input semantic: %i\n",
                         info->input_semantic_name[i]);
@@ -120,6 +125,9 @@ static void allocate_hardware_inputs(
             allocate(mydata, inputs->color[i], reg++);
         }
     }
+    if (inputs->face != ATTR_UNUSED) {
+        allocate(mydata, inputs->face, reg++);
+    }
     for (i = 0; i < ATTR_GENERIC_COUNT; i++) {
         if (inputs->generic[i] != ATTR_UNUSED) {
             allocate(mydata, inputs->generic[i], reg++);
@@ -173,7 +181,7 @@ static void get_external_state(
             t = (struct r300_texture*)texstate->sampler_views[i]->base.texture;
 
             /* XXX this should probably take into account STR, not just S. */
-            if (t->uses_pitch) {
+            if (t->desc.is_npot) {
                 switch (s->state.wrap_s) {
                     case PIPE_TEX_WRAP_REPEAT:
                         state->unit[i].wrap_mode = RC_WRAP_REPEAT;
@@ -248,13 +256,18 @@ static void r300_emit_fs_code_to_buffer(
 
         shader->cb_code_size = 19 +
                                ((code->inst_end + 1) * 6) +
-                               imm_count * 7;
+                               imm_count * 7 +
+			       code->int_constant_count * 2;
 
         NEW_CB(shader->cb_code, shader->cb_code_size);
         OUT_CB_REG(R500_US_CONFIG, R500_ZERO_TIMES_ANYTHING_EQUALS_ZERO);
         OUT_CB_REG(R500_US_PIXSIZE, code->max_temp_idx);
         OUT_CB_REG(R500_US_FC_CTRL, code->us_fc_ctrl);
-        OUT_CB_REG(R500_US_CODE_RANGE,
+        for(i = 0; i < code->int_constant_count; i++){
+		OUT_CB_REG(R500_US_FC_INT_CONST_0 + (i * 4),
+						code->int_constants[i]);
+	}
+	OUT_CB_REG(R500_US_CODE_RANGE,
                    R500_US_CODE_RANGE_ADDR(0) | R500_US_CODE_RANGE_SIZE(code->inst_end));
         OUT_CB_REG(R500_US_CODE_OFFSET, 0);
         OUT_CB_REG(R500_US_CODE_ADDR,
@@ -355,13 +368,14 @@ static void r300_translate_fragment_shader(
 {
     struct r300_fragment_program_compiler compiler;
     struct tgsi_to_rc ttr;
-    int wpos;
+    int wpos, face;
     unsigned i;
 
     tgsi_scan_shader(tokens, &shader->info);
     r300_shader_read_fs_inputs(&shader->info, &shader->inputs);
 
     wpos = shader->inputs.wpos;
+    face = shader->inputs.face;
 
     /* Setup the compiler. */
     memset(&compiler, 0, sizeof(compiler));
@@ -378,7 +392,7 @@ static void r300_translate_fragment_shader(
     find_output_registers(&compiler, shader);
 
     if (compiler.Base.Debug) {
-        debug_printf("r300: Initial fragment program\n");
+        DBG(r300, DBG_FP, "r300: Initial fragment program\n");
         tgsi_dump(tokens, 0);
     }
 
@@ -401,6 +415,10 @@ static void r300_translate_fragment_shader(
         rc_transform_fragment_wpos(&compiler.Base, wpos, wpos, TRUE);
     }
 
+    if (face != ATTR_UNUSED) {
+        rc_transform_fragment_face(&compiler.Base, face);
+    }
+
     /* Invoke the compiler */
     r3xx_compile_fragment_program(&compiler);
 
@@ -413,7 +431,7 @@ static void r300_translate_fragment_shader(
     }
 
     if (compiler.Base.Error) {
-        fprintf(stderr, "r300 FP: Compiler Error:\n%sUsing a dummy shader"
+        DBG(r300, DBG_FP, "r300 FP: Compiler Error:\n%sUsing a dummy shader"
                 " instead.\nIf there's an 'unknown opcode' message, please"
                 " file a bug report and attach this log.\n", compiler.Base.ErrorMsg);
 
diff --git a/src/gallium/drivers/r300/r300_hyperz.c b/src/gallium/drivers/r300/r300_hyperz.c
index e9528956019..811b5646e16 100644
--- a/src/gallium/drivers/r300/r300_hyperz.c
+++ b/src/gallium/drivers/r300/r300_hyperz.c
@@ -25,21 +25,170 @@
 #include "r300_hyperz.h"
 #include "r300_reg.h"
 #include "r300_fs.h"
+#include "r300_winsys.h"
 
+#include "util/u_format.h"
+#include "util/u_mm.h"
+
+/*
+  HiZ rules - taken from various docs 
+   1. HiZ only works on depth values
+   2. Cannot HiZ if stencil fail or zfail is !KEEP
+   3. on R300/400, HiZ is disabled if depth test is EQUAL
+   4. comparison changes without clears usually mean disabling HiZ
+*/
 /*****************************************************************************/
 /* The HyperZ setup                                                          */
 /*****************************************************************************/
 
+static bool r300_get_sc_hz_max(struct r300_context *r300)
+{
+    struct r300_dsa_state *dsa_state = r300->dsa_state.state;
+    int func = dsa_state->z_stencil_control & 0x7;
+    int ret = R300_SC_HYPERZ_MIN;
+
+    if (func >= 4 && func <= 7)
+       ret = R300_SC_HYPERZ_MAX;
+    return ret;
+}
+
+static bool r300_zfunc_same_direction(int func1, int func2)
+{
+    /* func1 is less/lessthan */
+    if (func1 == 1 || func1 == 2)
+        if (func2 == 3 || func2 == 4 || func2 == 5)
+            return FALSE;
+
+    if (func2 == 1 || func2 == 2)
+        if (func1 == 4 || func1 == 5)
+            return FALSE;
+    return TRUE;
+}
+    
+static int r300_get_hiz_min(struct r300_context *r300)
+{
+    struct r300_dsa_state *dsa_state = r300->dsa_state.state;
+    int func = dsa_state->z_stencil_control & 0x7;
+    int ret = R300_HIZ_MIN;
+
+    if (func == 1 || func == 2)
+       ret = R300_HIZ_MAX;
+    return ret;
+}
+
+static boolean r300_dsa_stencil_op_not_keep(struct pipe_stencil_state *s)
+{
+    if (s->enabled && (s->fail_op != PIPE_STENCIL_OP_KEEP ||
+                       s->zfail_op != PIPE_STENCIL_OP_KEEP))
+        return TRUE;
+    return FALSE;
+}
+
+static boolean r300_can_hiz(struct r300_context *r300)
+{
+    struct r300_dsa_state *dsa_state = r300->dsa_state.state;
+    struct pipe_depth_stencil_alpha_state *dsa = &dsa_state->dsa;
+    struct r300_screen* r300screen = r300->screen;
+    struct r300_hyperz_state *z = r300->hyperz_state.state;
+
+    /* shader writes depth - no HiZ */
+    if (r300_fragment_shader_writes_depth(r300_fs(r300))) /* (5) */
+        return FALSE;
+
+    if (r300->query_current)
+        return FALSE;
+    /* if stencil fail/zfail op is not KEEP */
+    if (r300_dsa_stencil_op_not_keep(&dsa->stencil[0]) ||
+        r300_dsa_stencil_op_not_keep(&dsa->stencil[1]))
+        return FALSE;
+
+    if (dsa->depth.enabled) {
+        /* if depth func is EQUAL pre-r500 */
+        if (dsa->depth.func == PIPE_FUNC_EQUAL && !r300screen->caps.is_r500)
+            return FALSE;
+        /* if depth func is NOTEQUAL */
+        if (dsa->depth.func == PIPE_FUNC_NOTEQUAL)
+            return FALSE;
+    }
+    /* depth comparison function - if just cleared save and return okay */
+    if (z->current_func == -1) {
+        int func = dsa_state->z_stencil_control & 0x7;
+        if (func != 0 && func != 7)
+            z->current_func = dsa_state->z_stencil_control & 0x7;
+    } else {
+        /* simple don't change */
+        if (!r300_zfunc_same_direction(z->current_func, (dsa_state->z_stencil_control & 0x7))) {
+            DBG(r300, DBG_HYPERZ, "z func changed direction - disabling hyper-z %d -> %d\n", z->current_func, dsa_state->z_stencil_control);
+            return FALSE;
+        }
+    }    
+    return TRUE;
+}
+
 static void r300_update_hyperz(struct r300_context* r300)
 {
     struct r300_hyperz_state *z =
         (struct r300_hyperz_state*)r300->hyperz_state.state;
+    struct pipe_framebuffer_state *fb =
+        (struct pipe_framebuffer_state*)r300->fb_state.state;
+    struct r300_texture *zstex =
+            fb->zsbuf ? r300_texture(fb->zsbuf->texture) : NULL;
+    boolean zmask_in_use = FALSE;
+    boolean hiz_in_use = FALSE;
 
+    z->gb_z_peq_config = 0;
     z->zb_bw_cntl = 0;
     z->sc_hyperz = R300_SC_HYPERZ_ADJ_2;
+    z->flush = 0;
 
-    if (r300->cbzb_clear)
+    if (r300->cbzb_clear) {
         z->zb_bw_cntl |= R300_ZB_CB_CLEAR_CACHE_LINE_WRITE_ONLY;
+        return;
+    }
+
+    if (!zstex)
+        return;
+
+    if (!r300->rws->get_value(r300->rws, R300_CAN_HYPERZ))
+        return;
+
+    zmask_in_use = zstex->zmask_in_use[fb->zsbuf->level];
+    hiz_in_use = zstex->hiz_in_use[fb->zsbuf->level];
+
+    /* Z fastfill. */
+    if (zmask_in_use) {
+        z->zb_bw_cntl |= R300_FAST_FILL_ENABLE; /*  | R300_FORCE_COMPRESSED_STENCIL_VALUE_ENABLE;*/
+    }
+
+    /* Zbuffer compression. */
+    if (zmask_in_use && r300->z_compression) {
+        z->zb_bw_cntl |= R300_RD_COMP_ENABLE;
+        if (r300->z_decomp_rd == false)
+            z->zb_bw_cntl |= R300_WR_COMP_ENABLE;
+    }
+    /* RV350 and up optimizations. */
+    /* The section 10.4.9 in the docs is a lie. */
+    if (r300->z_compression == RV350_Z_COMPRESS_88)
+        z->gb_z_peq_config |= R300_GB_Z_PEQ_CONFIG_Z_PEQ_SIZE_8_8;
+
+    if (hiz_in_use) {
+        bool can_hiz = r300_can_hiz(r300);
+        if (can_hiz) {
+            z->zb_bw_cntl |= R300_HIZ_ENABLE;
+            z->sc_hyperz |= R300_SC_HYPERZ_ENABLE;
+            z->sc_hyperz |= r300_get_sc_hz_max(r300);
+            z->zb_bw_cntl |= r300_get_hiz_min(r300);
+        }
+    }
+
+    /* R500-specific features and optimizations. */
+    if (r300->screen->caps.is_r500) {
+        z->zb_bw_cntl |= R500_HIZ_FP_EXP_BITS_3;
+        z->zb_bw_cntl |=
+                R500_HIZ_EQUAL_REJECT_ENABLE |
+                R500_PEQ_PACKING_ENABLE |
+                R500_COVERED_PTR_MASKING_ENABLE;
+    }
 }
 
 /*****************************************************************************/
@@ -126,15 +275,121 @@ static void r300_update_ztop(struct r300_context* r300)
     } else {
         ztop_state->z_buffer_top = R300_ZTOP_ENABLE;
     }
-
     if (ztop_state->z_buffer_top != old_ztop)
         r300->ztop_state.dirty = TRUE;
 }
 
+#define ALIGN_DIVUP(x, y) (((x) + (y) - 1) / (y))
+
+static void r300_update_hiz_clear(struct r300_context *r300)
+{
+    struct pipe_framebuffer_state *fb =
+        (struct pipe_framebuffer_state*)r300->fb_state.state;
+    uint32_t height;
+
+    height = ALIGN_DIVUP(fb->zsbuf->height, 4);
+    r300->hiz_clear.size = height * 4;
+}
+
+static void r300_update_zmask_clear(struct r300_context *r300)
+{
+    struct pipe_framebuffer_state *fb =
+        (struct pipe_framebuffer_state*)r300->fb_state.state;
+    uint32_t height;
+    int mult;
+
+    if (r300->z_compression == RV350_Z_COMPRESS_88)
+        mult = 8;
+    else
+        mult = 4;
+
+    height = ALIGN_DIVUP(fb->zsbuf->height, mult);
+
+    r300->zmask_clear.size = height * 4;
+}
+
 void r300_update_hyperz_state(struct r300_context* r300)
 {
     r300_update_ztop(r300);
     if (r300->hyperz_state.dirty) {
         r300_update_hyperz(r300);
     }
+
+    if (r300->hiz_clear.dirty) {
+       r300_update_hiz_clear(r300);
+    }
+    if (r300->zmask_clear.dirty) {
+       r300_update_zmask_clear(r300);
+    }
+}
+
+void r300_hiz_alloc_block(struct r300_context *r300, struct r300_surface *surf)
+{
+    struct r300_texture *tex;
+    uint32_t zsize, ndw;
+    int level = surf->base.level;
+
+    tex = r300_texture(surf->base.texture);
+
+    if (tex->hiz_mem[level])
+        return;
+
+    zsize = tex->desc.layer_size_in_bytes[level];
+    zsize /= util_format_get_blocksize(tex->desc.b.b.format);
+    ndw = ALIGN_DIVUP(zsize, 64);
+
+    tex->hiz_mem[level] = u_mmAllocMem(r300->hiz_mm, ndw, 0, 0);
+    return;
+}
+
+void r300_zmask_alloc_block(struct r300_context *r300, struct r300_surface *surf, int compress)
+{
+    int bsize = 256;
+    uint32_t zsize, ndw;
+    int level = surf->base.level;
+    struct r300_texture *tex;
+
+    tex = r300_texture(surf->base.texture);
+
+    /* We currently don't handle decompression for 3D textures and cubemaps
+     * correctly. */
+    if (tex->desc.b.b.target != PIPE_TEXTURE_1D &&
+        tex->desc.b.b.target != PIPE_TEXTURE_2D)
+        return;
+
+    if (tex->zmask_mem[level])
+        return;
+
+    zsize = tex->desc.layer_size_in_bytes[level];
+    zsize /= util_format_get_blocksize(tex->desc.b.b.format);
+
+    /* each zmask dword represents 16 4x4 blocks - which is 256 pixels
+       or 16 8x8 depending on the gb peq flag = 1024 pixels */
+    if (compress == RV350_Z_COMPRESS_88)
+        bsize = 1024;
+
+    ndw = ALIGN_DIVUP(zsize, bsize);
+    tex->zmask_mem[level] = u_mmAllocMem(r300->zmask_mm, ndw, 0, 0);
+    return;
+}
+
+void r300_hyperz_init_mm(struct r300_context *r300)
+{
+    struct r300_screen* r300screen = r300->screen;
+    int frag_pipes = r300screen->caps.num_frag_pipes;
+
+    if (r300screen->caps.hiz_ram)
+      r300->hiz_mm = u_mmInit(0, r300screen->caps.hiz_ram * frag_pipes);
+
+    r300->zmask_mm = u_mmInit(0, r300screen->caps.zmask_ram * frag_pipes);
+}
+
+void r300_hyperz_destroy_mm(struct r300_context *r300)
+{
+    struct r300_screen* r300screen = r300->screen;
+
+    if (r300screen->caps.hiz_ram)
+      u_mmDestroy(r300->hiz_mm);
+
+    u_mmDestroy(r300->zmask_mm);
 }
diff --git a/src/gallium/drivers/r300/r300_hyperz.h b/src/gallium/drivers/r300/r300_hyperz.h
index 3df5053b896..09e1ff6625c 100644
--- a/src/gallium/drivers/r300/r300_hyperz.h
+++ b/src/gallium/drivers/r300/r300_hyperz.h
@@ -27,4 +27,9 @@ struct r300_context;
 
 void r300_update_hyperz_state(struct r300_context* r300);
 
+void r300_hiz_alloc_block(struct r300_context *r300, struct r300_surface *surf);
+void r300_zmask_alloc_block(struct r300_context *r300, struct r300_surface *surf, int compress);
+
+void r300_hyperz_init_mm(struct r300_context *r300);
+void r300_hyperz_destroy_mm(struct r300_context *r300);
 #endif
diff --git a/src/gallium/drivers/r300/r300_reg.h b/src/gallium/drivers/r300/r300_reg.h
index 2acc1a903e8..60d3b600cb7 100644
--- a/src/gallium/drivers/r300/r300_reg.h
+++ b/src/gallium/drivers/r300/r300_reg.h
@@ -496,6 +496,12 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define R300_VAP_GB_HORZ_CLIP_ADJ                   0x2228
 #define R300_VAP_GB_HORZ_DISC_ADJ                   0x222c
 
+#define R300_VAP_PVS_FLOW_CNTL_ADDRS_0      0x2230
+#define R300_PVS_FC_ACT_ADRS(x)             ((x) << 0)
+#define R300_PVS_FC_LOOP_CNT_JMP_INST(x)    ((x) << 8)
+#define R300_PVS_FC_LAST_INST(x)            ((x) << 16)
+#define R300_PVS_FC_RTN_INST(x)             ((x) << 24)
+
 /* gap */
 
 /* Sometimes, END_OF_PKT and 0x2284=0 are the only commands sent between
@@ -514,6 +520,10 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #       define R300_2288_R300                    0x00750000 /* -- nh */
 #       define R300_2288_RV350                   0x0000FFFF /* -- Vladimir */
 
+#define R300_VAP_PVS_FLOW_CNTL_LOOP_INDEX_0 0x2290
+#define R300_PVS_FC_LOOP_INIT_VAL(x)        ((x) << 0)
+#define R300_PVS_FC_LOOP_STEP_VAL(x)        ((x) << 8)
+
 /* gap */
 
 /* Addresses are relative to the vertex program instruction area of the
@@ -548,6 +558,9 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define R300_VAP_PVS_CODE_CNTL_1	    0x22D8
 #       define R300_PVS_LAST_VTX_SRC_INST_SHIFT  0
 #define R300_VAP_PVS_FLOW_CNTL_OPC          0x22DC
+#define R300_VAP_PVS_FC_OPC_JUMP(x)         (1 << (2 * (x)))
+#define R300_VAP_PVS_FC_OPC_LOOP(x)         (2 << (2 * (x)))
+#define R300_VAP_PVS_FC_OPC_JSR(x)          (3 << (2 * (x)))
 
 /* The entire range from 0x2300 to 0x2AC inclusive seems to be used for
  * immediate vertices
@@ -564,6 +577,14 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 /* write 0 to indicate end of packet? */
 #define R300_VAP_VTX_END_OF_PKT             0x24AC
 
+#define R500_VAP_PVS_FLOW_CNTL_ADDRS_LW_0   0x2500
+#define R500_PVS_FC_ACT_ADRS(x)             ((x) << 0)
+#define R500_PVS_FC_LOOP_CNT_JMP_INST(x)    ((x) << 16)
+
+#define R500_VAP_PVS_FLOW_CNTL_ADDRS_UW_0   0x2504
+#define R500_PVS_FC_LAST_INST(x)            ((x) << 0)
+#define R500_PVS_FC_RTN_INST(x)             ((x) << 16)
+
 /* gap */
 
 /* These are values from r300_reg/r300_reg.h - they are known to be correct
@@ -3436,6 +3457,7 @@ enum {
 #   define R300_VBPNTR_SIZE1(x)    (((x) >> 2) << 16)
 #   define R300_VBPNTR_STRIDE1(x)  (((x) >> 2) << 24)
 
+#define R300_PACKET3_3D_CLEAR_ZMASK         0x00003200
 #define R300_PACKET3_INDX_BUFFER            0x00003300
 #    define R300_INDX_BUFFER_DST_SHIFT          0
 #    define R300_INDX_BUFFER_SKIP_SHIFT         16
diff --git a/src/gallium/drivers/r300/r300_render.c b/src/gallium/drivers/r300/r300_render.c
index bae02135da9..86b11ca0458 100644
--- a/src/gallium/drivers/r300/r300_render.c
+++ b/src/gallium/drivers/r300/r300_render.c
@@ -186,20 +186,14 @@ enum r300_prepare_flags {
  * \param cs_dwords     The number of dwords to reserve in CS.
  * \param aos_offset    The offset passed to emit_aos.
  * \param index_bias    The index bias to emit.
- * \param end_cs_dwords The number of free dwords which must be available
- *                      at the end of CS after drawing in case the CS space
- *                      management is performed by a draw_* function manually.
- *                      The parameter may be NULL.
  */
 static void r300_prepare_for_rendering(struct r300_context *r300,
                                        enum r300_prepare_flags flags,
                                        struct pipe_resource *index_buffer,
                                        unsigned cs_dwords,
                                        int aos_offset,
-                                       int index_bias,
-                                       unsigned *end_cs_dwords)
+                                       int index_bias)
 {
-    unsigned end_dwords    = 0;
     boolean flushed        = FALSE;
     boolean first_draw     = flags & PREP_FIRST_DRAW;
     boolean emit_aos       = flags & PREP_EMIT_AOS;
@@ -221,11 +215,7 @@ static void r300_prepare_for_rendering(struct r300_context *r300,
             cs_dwords += 7; /* emit_aos_swtcl */
     }
 
-    /* Emitted in flush. */
-    end_dwords += 26; /* emit_query_end */
-    end_dwords += r300->hyperz_state.size; /* emit_hyperz_end */
-
-    cs_dwords += end_dwords;
+    cs_dwords += r300_get_num_cs_end_dwords(r300);
 
     /* Reserve requested CS space. */
     if (cs_dwords > (r300->cs->ndw - r300->cs->cdw)) {
@@ -250,9 +240,6 @@ static void r300_prepare_for_rendering(struct r300_context *r300,
         if (emit_aos_swtcl)
             r300_emit_aos_swtcl(r300, indexed);
     }
-
-    if (end_cs_dwords)
-        *end_cs_dwords = end_dwords;
 }
 
 static boolean immd_is_good_idea(struct r300_context *r300,
@@ -353,7 +340,7 @@ static void r300_emit_draw_arrays_immediate(struct r300_context *r300,
 
     dwords = 9 + count * vertex_size;
 
-    r300_prepare_for_rendering(r300, PREP_FIRST_DRAW, NULL, dwords, 0, 0, NULL);
+    r300_prepare_for_rendering(r300, PREP_FIRST_DRAW, NULL, dwords, 0, 0);
 
     BEGIN_CS(dwords);
     OUT_CS_REG(R300_GA_COLOR_CONTROL,
@@ -533,7 +520,7 @@ static void r300_draw_range_elements(struct pipe_context* pipe,
     /* 15 dwords for emit_draw_elements */
     r300_prepare_for_rendering(r300,
         PREP_FIRST_DRAW | PREP_VALIDATE_VBOS | PREP_EMIT_AOS | PREP_INDEXED,
-        indexBuffer, 15, buffer_offset, indexBias, NULL);
+        indexBuffer, 15, buffer_offset, indexBias);
 
     if (alt_num_verts || count <= 65535) {
         r300_emit_draw_elements(r300, indexBuffer, indexSize,
@@ -552,7 +539,7 @@ static void r300_draw_range_elements(struct pipe_context* pipe,
             if (count) {
                 r300_prepare_for_rendering(r300,
                     PREP_VALIDATE_VBOS | PREP_EMIT_AOS | PREP_INDEXED,
-                    indexBuffer, 15, buffer_offset, indexBias, NULL);
+                    indexBuffer, 15, buffer_offset, indexBias);
             }
         } while (count);
     }
@@ -566,19 +553,6 @@ static void r300_draw_range_elements(struct pipe_context* pipe,
     }
 }
 
-/* Simple helpers for context setup. Should probably be moved to util. */
-static void r300_draw_elements(struct pipe_context* pipe,
-                               struct pipe_resource* indexBuffer,
-                               unsigned indexSize, int indexBias, unsigned mode,
-                               unsigned start, unsigned count)
-{
-    struct r300_context *r300 = r300_context(pipe);
-
-    pipe->draw_range_elements(pipe, indexBuffer, indexSize, indexBias,
-                              0, r300->vertex_buffer_max_index,
-                              mode, start, count);
-}
-
 static void r300_draw_arrays(struct pipe_context* pipe, unsigned mode,
                              unsigned start, unsigned count)
 {
@@ -610,7 +584,7 @@ static void r300_draw_arrays(struct pipe_context* pipe, unsigned mode,
     } else {
         /* 9 spare dwords for emit_draw_arrays. */
         r300_prepare_for_rendering(r300, PREP_FIRST_DRAW | PREP_VALIDATE_VBOS | PREP_EMIT_AOS,
-                               NULL, 9, start, 0, NULL);
+                               NULL, 9, start, 0);
 
         if (alt_num_verts || count <= 65535) {
             r300_emit_draw_arrays(r300, mode, count);
@@ -626,7 +600,7 @@ static void r300_draw_arrays(struct pipe_context* pipe, unsigned mode,
                 if (count) {
                     r300_prepare_for_rendering(r300,
                         PREP_VALIDATE_VBOS | PREP_EMIT_AOS, NULL, 9,
-                        start, 0, NULL);
+                        start, 0);
                 }
             } while (count);
         }
@@ -638,111 +612,104 @@ static void r300_draw_arrays(struct pipe_context* pipe, unsigned mode,
     }
 }
 
-/****************************************************************************
- * The rest of this file is for SW TCL rendering only. Please be polite and *
- * keep these functions separated so that they are easier to locate. ~C.    *
- ***************************************************************************/
-
-/* SW TCL arrays, using Draw. */
-static void r300_swtcl_draw_arrays(struct pipe_context* pipe,
-                                   unsigned mode,
-                                   unsigned start,
-                                   unsigned count)
+static void r300_draw_vbo(struct pipe_context* pipe,
+                          const struct pipe_draw_info *info)
 {
     struct r300_context* r300 = r300_context(pipe);
-    struct pipe_transfer *vb_transfer[PIPE_MAX_ATTRIBS];
-    int i;
 
-    if (r300->skip_rendering) {
-        return;
+    if (info->indexed && r300->index_buffer.buffer) {
+        unsigned offset;
+
+        assert(r300->index_buffer.offset % r300->index_buffer.index_size == 0);
+        offset = r300->index_buffer.offset / r300->index_buffer.index_size;
+
+        r300_draw_range_elements(pipe,
+                                 r300->index_buffer.buffer,
+                                 r300->index_buffer.index_size,
+                                 info->index_bias,
+                                 info->min_index,
+                                 info->max_index,
+                                 info->mode,
+                                 info->start + offset,
+                                 info->count);
     }
-
-    if (!u_trim_pipe_prim(mode, &count)) {
-        return;
-    }
-
-    r300_update_derived_state(r300);
-
-    for (i = 0; i < r300->vertex_buffer_count; i++) {
-        void* buf = pipe_buffer_map(pipe,
-                                    r300->vertex_buffer[i].buffer,
-                                    PIPE_TRANSFER_READ,
-				    &vb_transfer[i]);
-        draw_set_mapped_vertex_buffer(r300->draw, i, buf);
-    }
-
-    draw_set_mapped_element_buffer(r300->draw, 0, 0, NULL);
-
-    draw_arrays(r300->draw, mode, start, count);
-
-    /* XXX Not sure whether this is the best fix.
-     * It prevents CS from being rejected and weird assertion failures. */
-    draw_flush(r300->draw);
-
-    for (i = 0; i < r300->vertex_buffer_count; i++) {
-        pipe_buffer_unmap(pipe, r300->vertex_buffer[i].buffer,
-			  vb_transfer[i]);
-        draw_set_mapped_vertex_buffer(r300->draw, i, NULL);
+    else {
+        r300_draw_arrays(pipe,
+                         info->mode,
+                         info->start,
+                         info->count);
     }
 }
 
+/****************************************************************************
+ * The rest of this file is for SW TCL rendering only. Please be polite and *
+ * keep these functions separated so that they are easier to locate. ~C.    *
+ ***************************************************************************/
+
 /* SW TCL elements, using Draw. */
-static void r300_swtcl_draw_range_elements(struct pipe_context* pipe,
-                                           struct pipe_resource* indexBuffer,
-                                           unsigned indexSize,
-                                           int indexBias,
-                                           unsigned minIndex,
-                                           unsigned maxIndex,
-                                           unsigned mode,
-                                           unsigned start,
-                                           unsigned count)
+static void r300_swtcl_draw_vbo(struct pipe_context* pipe,
+                                const struct pipe_draw_info *info)
 {
     struct r300_context* r300 = r300_context(pipe);
     struct pipe_transfer *vb_transfer[PIPE_MAX_ATTRIBS];
-    struct pipe_transfer *ib_transfer;
+    struct pipe_transfer *ib_transfer = NULL;
+    unsigned count = info->count;
     int i;
-    void* indices;
+    void* indices = NULL;
 
     if (r300->skip_rendering) {
         return;
     }
 
-    if (!u_trim_pipe_prim(mode, &count)) {
+    if (!u_trim_pipe_prim(info->mode, &count)) {
         return;
     }
 
     r300_update_derived_state(r300);
 
     for (i = 0; i < r300->vertex_buffer_count; i++) {
-        void* buf = pipe_buffer_map(pipe,
-                                    r300->vertex_buffer[i].buffer,
-                                    PIPE_TRANSFER_READ,
-				    &vb_transfer[i]);
-        draw_set_mapped_vertex_buffer(r300->draw, i, buf);
+        if (r300->vertex_buffer[i].buffer) {
+            void *buf = pipe_buffer_map(pipe,
+                                  r300->vertex_buffer[i].buffer,
+                                  PIPE_TRANSFER_READ,
+                                  &vb_transfer[i]);
+            draw_set_mapped_vertex_buffer(r300->draw, i, buf);
+        }
     }
 
-    indices = pipe_buffer_map(pipe, indexBuffer,
-                              PIPE_TRANSFER_READ, &ib_transfer);
-    draw_set_mapped_element_buffer_range(r300->draw, indexSize, indexBias,
-                                         minIndex, maxIndex, indices);
+    if (info->indexed && r300->index_buffer.buffer) {
+        indices = pipe_buffer_map(pipe, r300->index_buffer.buffer,
+                                  PIPE_TRANSFER_READ, &ib_transfer);
+        if (indices)
+            indices = (void *) ((char *) indices + r300->index_buffer.offset);
+    }
 
-    draw_arrays(r300->draw, mode, start, count);
+    draw_set_mapped_element_buffer_range(r300->draw, (indices) ?
+                                         r300->index_buffer.index_size : 0,
+                                         info->index_bias,
+                                         info->min_index,
+                                         info->max_index,
+                                         indices);
+
+    draw_arrays(r300->draw, info->mode, info->start, count);
 
     /* XXX Not sure whether this is the best fix.
      * It prevents CS from being rejected and weird assertion failures. */
     draw_flush(r300->draw);
 
     for (i = 0; i < r300->vertex_buffer_count; i++) {
-        pipe_buffer_unmap(pipe, r300->vertex_buffer[i].buffer,
-			  vb_transfer[i]);
-        draw_set_mapped_vertex_buffer(r300->draw, i, NULL);
+        if (r300->vertex_buffer[i].buffer) {
+            pipe_buffer_unmap(pipe, r300->vertex_buffer[i].buffer,
+                              vb_transfer[i]);
+            draw_set_mapped_vertex_buffer(r300->draw, i, NULL);
+        }
     }
 
-    pipe_buffer_unmap(pipe, indexBuffer,
-		      ib_transfer);
-    draw_set_mapped_element_buffer_range(r300->draw, 0, 0,
-                                         start, start + count - 1,
-                                         NULL);
+    if (ib_transfer) {
+        pipe_buffer_unmap(pipe, r300->index_buffer.buffer, ib_transfer);
+        draw_set_mapped_element_buffer_range(r300->draw, 0, 0, info->start,
+                info->start + count - 1, NULL);
+    }
 }
 
 /* Object for rendering using Draw. */
@@ -820,6 +787,8 @@ static void* r300_render_map_vertices(struct vbuf_render* render)
                                           PIPE_TRANSFER_WRITE,
 					  &r300render->vbo_transfer);
 
+    assert(r300render->vbo_ptr);
+
     return ((uint8_t*)r300render->vbo_ptr + r300render->vbo_offset);
 }
 
@@ -872,7 +841,7 @@ static void r300_render_draw_arrays(struct vbuf_render* render,
     (void) i; (void) ptr;
 
     r300_prepare_for_rendering(r300, PREP_FIRST_DRAW | PREP_EMIT_AOS_SWTCL,
-                               NULL, dwords, 0, 0, NULL);
+                               NULL, dwords, 0, 0);
 
     DBG(r300, DBG_DRAW, "r300: render_draw_arrays (count: %d)\n", count);
 
@@ -925,7 +894,8 @@ static void r300_render_draw_elements(struct vbuf_render* render,
      * indices than it can fit in CS. */
     r300_prepare_for_rendering(r300,
         PREP_FIRST_DRAW | PREP_EMIT_AOS_SWTCL | PREP_INDEXED,
-        NULL, 256, 0, 0, &end_cs_dwords);
+        NULL, 256, 0, 0);
+    end_cs_dwords = r300_get_num_cs_end_dwords(r300);
 
     while (count) {
         free_dwords = r300->cs->ndw - r300->cs->cdw;
@@ -955,7 +925,8 @@ static void r300_render_draw_elements(struct vbuf_render* render,
         if (count) {
             r300_prepare_for_rendering(r300,
                 PREP_EMIT_AOS_SWTCL | PREP_INDEXED,
-                NULL, 256, 0, 0, &end_cs_dwords);
+                NULL, 256, 0, 0);
+            end_cs_dwords = r300_get_num_cs_end_dwords(r300);
         }
     }
 }
@@ -1049,7 +1020,7 @@ static void r300_blitter_draw_rectangle(struct blitter_context *blitter,
     r300->clip_state.dirty = FALSE;
     r300->viewport_state.dirty = FALSE;
 
-    r300_prepare_for_rendering(r300, PREP_FIRST_DRAW, NULL, dwords, 0, 0, NULL);
+    r300_prepare_for_rendering(r300, PREP_FIRST_DRAW, NULL, dwords, 0, 0);
 
     DBG(r300, DBG_DRAW, "r300: draw_rectangle\n");
 
@@ -1141,16 +1112,11 @@ static void r300_resource_resolve(struct pipe_context* pipe,
 
 void r300_init_render_functions(struct r300_context *r300)
 {
-    /* Set generic functions. */
-    r300->context.draw_elements = r300_draw_elements;
-
     /* Set draw functions based on presence of HW TCL. */
     if (r300->screen->caps.has_tcl) {
-        r300->context.draw_arrays = r300_draw_arrays;
-        r300->context.draw_range_elements = r300_draw_range_elements;
+        r300->context.draw_vbo = r300_draw_vbo;
     } else {
-        r300->context.draw_arrays = r300_swtcl_draw_arrays;
-        r300->context.draw_range_elements = r300_swtcl_draw_range_elements;
+        r300->context.draw_vbo = r300_swtcl_draw_vbo;
     }
 
     r300->context.resource_resolve = r300_resource_resolve;
diff --git a/src/gallium/drivers/r300/r300_render_stencilref.c b/src/gallium/drivers/r300/r300_render_stencilref.c
index 9a6b4e12ff1..1f035d64a28 100644
--- a/src/gallium/drivers/r300/r300_render_stencilref.c
+++ b/src/gallium/drivers/r300/r300_render_stencilref.c
@@ -34,13 +34,8 @@
 #include "r300_reg.h"
 
 struct r300_stencilref_context {
-    void (*draw_arrays)(struct pipe_context *pipe,
-                        unsigned mode, unsigned start, unsigned count);
-
-    void (*draw_range_elements)(
-        struct pipe_context *pipe, struct pipe_resource *indexBuffer,
-        unsigned indexSize, int indexBias, unsigned minIndex, unsigned maxIndex,
-        unsigned mode, unsigned start, unsigned count);
+    void (*draw_vbo)(struct pipe_context *pipe,
+                     const struct pipe_draw_info *info);
 
     uint32_t rs_cull_mode;
     uint32_t zb_stencilrefmask;
@@ -105,41 +100,19 @@ static void r300_stencilref_end(struct r300_context *r300)
     r300->dsa_state.dirty = TRUE;
 }
 
-static void r300_stencilref_draw_arrays(struct pipe_context *pipe, unsigned mode,
-                                        unsigned start, unsigned count)
-{
-    struct r300_context *r300 = r300_context(pipe);
-    struct r300_stencilref_context *sr = r300->stencilref_fallback;
-
-    if (!r300_stencilref_needed(r300)) {
-        sr->draw_arrays(pipe, mode, start, count);
-    } else {
-        r300_stencilref_begin(r300);
-        sr->draw_arrays(pipe, mode, start, count);
-        r300_stencilref_switch_side(r300);
-        sr->draw_arrays(pipe, mode, start, count);
-        r300_stencilref_end(r300);
-    }
-}
-
-static void r300_stencilref_draw_range_elements(
-    struct pipe_context *pipe, struct pipe_resource *indexBuffer,
-    unsigned indexSize, int indexBias, unsigned minIndex, unsigned maxIndex,
-    unsigned mode, unsigned start, unsigned count)
+static void r300_stencilref_draw_vbo(struct pipe_context *pipe,
+                                     const struct pipe_draw_info *info)
 {
     struct r300_context *r300 = r300_context(pipe);
     struct r300_stencilref_context *sr = r300->stencilref_fallback;
 
     if (!r300_stencilref_needed(r300)) {
-        sr->draw_range_elements(pipe, indexBuffer, indexSize, indexBias,
-                                minIndex, maxIndex, mode, start, count);
+        sr->draw_vbo(pipe, info);
     } else {
         r300_stencilref_begin(r300);
-        sr->draw_range_elements(pipe, indexBuffer, indexSize, indexBias,
-                                minIndex, maxIndex, mode, start, count);
+        sr->draw_vbo(pipe, info);
         r300_stencilref_switch_side(r300);
-        sr->draw_range_elements(pipe, indexBuffer, indexSize, indexBias,
-                                minIndex, maxIndex, mode, start, count);
+        sr->draw_vbo(pipe, info);
         r300_stencilref_end(r300);
     }
 }
@@ -148,11 +121,9 @@ void r300_plug_in_stencil_ref_fallback(struct r300_context *r300)
 {
     r300->stencilref_fallback = CALLOC_STRUCT(r300_stencilref_context);
 
-    /* Save original draw functions. */
-    r300->stencilref_fallback->draw_arrays = r300->context.draw_arrays;
-    r300->stencilref_fallback->draw_range_elements = r300->context.draw_range_elements;
+    /* Save original draw function. */
+    r300->stencilref_fallback->draw_vbo = r300->context.draw_vbo;
 
-    /* Override the draw functions. */
-    r300->context.draw_arrays = r300_stencilref_draw_arrays;
-    r300->context.draw_range_elements = r300_stencilref_draw_range_elements;
+    /* Override the draw function. */
+    r300->context.draw_vbo = r300_stencilref_draw_vbo;
 }
diff --git a/src/gallium/drivers/r300/r300_screen.c b/src/gallium/drivers/r300/r300_screen.c
index 5a11b98eb6c..1e4edcdbc31 100644
--- a/src/gallium/drivers/r300/r300_screen.c
+++ b/src/gallium/drivers/r300/r300_screen.c
@@ -115,7 +115,6 @@ static int r300_get_param(struct pipe_screen* pscreen, enum pipe_cap param)
         case PIPE_CAP_TEXTURE_MIRROR_REPEAT:
         case PIPE_CAP_BLEND_EQUATION_SEPARATE:
         case PIPE_CAP_TEXTURE_SWIZZLE:
-        case PIPE_CAP_DEPTH_CLAMP:
             return 1;
 
         /* Unsupported features (boolean caps). */
@@ -124,6 +123,8 @@ static int r300_get_param(struct pipe_screen* pscreen, enum pipe_cap param)
         case PIPE_CAP_TGSI_CONT_SUPPORTED:
         case PIPE_CAP_INDEP_BLEND_ENABLE:
         case PIPE_CAP_INDEP_BLEND_FUNC:
+        case PIPE_CAP_DEPTH_CLAMP: /* XXX implemented, but breaks Regnum Online */
+        case PIPE_CAP_DEPTHSTENCIL_CLEAR_SEPARATE:
             return 0;
 
         /* Texturing. */
@@ -150,9 +151,6 @@ static int r300_get_param(struct pipe_screen* pscreen, enum pipe_cap param)
         case PIPE_CAP_MAX_CONST_BUFFER_SIZE:
             return 256;
 
-        case PIPE_CAP_DEPTHSTENCIL_CLEAR_SEPARATE:
-            return 1;
-
         /* Fragment coordinate conventions. */
         case PIPE_CAP_TGSI_FS_COORD_ORIGIN_UPPER_LEFT:
         case PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_HALF_INTEGER:
@@ -257,8 +255,6 @@ static boolean r300_is_format_supported(struct pipe_screen* screen,
     uint32_t retval = 0;
     boolean is_r500 = r300_screen(screen)->caps.is_r500;
     boolean is_r400 = r300_screen(screen)->caps.is_r400;
-    boolean is_z24 = format == PIPE_FORMAT_X8Z24_UNORM ||
-                     format == PIPE_FORMAT_S8_USCALED_Z24_UNORM;
     boolean is_color2101010 = format == PIPE_FORMAT_R10G10B10A2_UNORM ||
                               format == PIPE_FORMAT_R10G10B10X2_SNORM ||
                               format == PIPE_FORMAT_B10G10R10A2_UNORM ||
@@ -281,11 +277,14 @@ static boolean r300_is_format_supported(struct pipe_screen* screen,
         case 3:
         case 4:
         case 6:
+            return FALSE;
+#if 0
             if (usage != PIPE_BIND_RENDER_TARGET ||
                 !util_format_is_rgba8_variant(
                     util_format_description(format))) {
                 return FALSE;
             }
+#endif
             break;
         default:
             return FALSE;
@@ -293,8 +292,6 @@ static boolean r300_is_format_supported(struct pipe_screen* screen,
 
     /* Check sampler format support. */
     if ((usage & PIPE_BIND_SAMPLER_VIEW) &&
-        /* Z24 cannot be sampled from on non-r5xx. */
-        (is_r500 || !is_z24) &&
         /* ATI1N is r5xx-only. */
         (is_r500 || !is_ati1n) &&
         /* ATI2N is supported on r4xx-r5xx. */
diff --git a/src/gallium/drivers/r300/r300_screen.h b/src/gallium/drivers/r300/r300_screen.h
index edc494ff6ca..13a3320b992 100644
--- a/src/gallium/drivers/r300/r300_screen.h
+++ b/src/gallium/drivers/r300/r300_screen.h
@@ -90,6 +90,8 @@ r300_winsys_screen(struct pipe_screen *screen) {
 #define DBG_FALL        (1 << 8)
 #define DBG_FB          (1 << 9)
 #define DBG_RS_BLOCK    (1 << 10)
+#define DBG_CBZB        (1 << 11)
+#define DBG_HYPERZ      (1 << 12)
 /* Features. */
 #define DBG_ANISOHQ     (1 << 16)
 #define DBG_NO_TILING   (1 << 17)
diff --git a/src/gallium/drivers/r300/r300_shader_semantics.h b/src/gallium/drivers/r300/r300_shader_semantics.h
index cb7a37033f3..4be23e64ce7 100644
--- a/src/gallium/drivers/r300/r300_shader_semantics.h
+++ b/src/gallium/drivers/r300/r300_shader_semantics.h
@@ -38,6 +38,7 @@ struct r300_shader_semantics {
     int psize;
     int color[ATTR_COLOR_COUNT];
     int bcolor[ATTR_COLOR_COUNT];
+    int face;
     int generic[ATTR_GENERIC_COUNT];
     int fog;
     int wpos;
@@ -50,6 +51,7 @@ static INLINE void r300_shader_semantics_reset(
 
     info->pos = ATTR_UNUSED;
     info->psize = ATTR_UNUSED;
+    info->face = ATTR_UNUSED;
     info->fog = ATTR_UNUSED;
     info->wpos = ATTR_UNUSED;
 
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
index f52265b1c03..239edd98e32 100644
--- a/src/gallium/drivers/r300/r300_state.c
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -25,6 +25,7 @@
 
 #include "util/u_blitter.h"
 #include "util/u_math.h"
+#include "util/u_mm.h"
 #include "util/u_memory.h"
 #include "util/u_pack_color.h"
 
@@ -43,6 +44,7 @@
 #include "r300_texture.h"
 #include "r300_vs.h"
 #include "r300_winsys.h"
+#include "r300_hyperz.h"
 
 /* r300_state: Functions used to intialize state context by translating
  * Gallium state objects into semi-native r300 state objects. */
@@ -446,7 +448,6 @@ static void r300_set_clip_state(struct pipe_context* pipe,
 
         r300->clip_state.dirty = TRUE;
     } else {
-        draw_flush(r300->draw);
         draw_set_clip_state(r300->draw, state);
     }
 }
@@ -473,14 +474,14 @@ static void*
 
     dsa->dsa = *state;
 
-    /* Depth test setup. */
+    /* Depth test setup. - separate write mask depth for decomp flush */
+    if (state->depth.writemask) {
+        dsa->z_buffer_control |= R300_Z_WRITE_ENABLE;
+    }
+
     if (state->depth.enabled) {
         dsa->z_buffer_control |= R300_Z_ENABLE;
 
-        if (state->depth.writemask) {
-            dsa->z_buffer_control |= R300_Z_WRITE_ENABLE;
-        }
-
         dsa->z_stencil_control |=
             (r300_translate_depth_stencil_function(state->depth.func) <<
                 R300_Z_FUNC_SHIFT);
@@ -593,6 +594,7 @@ static void r300_bind_dsa_state(struct pipe_context* pipe,
 
     UPDATE_STATE(state, r300->dsa_state);
 
+    r300->hyperz_state.dirty = TRUE; /* Will be updated before the emission. */
     r300_dsa_inject_stencilref(r300);
 }
 
@@ -619,7 +621,8 @@ static void r300_tex_set_tiling_flags(struct r300_context *r300,
 {
     /* Check if the macrotile flag needs to be changed.
      * Skip changing the flags otherwise. */
-    if (tex->mip_macrotile[tex->surface_level] != tex->mip_macrotile[level]) {
+    if (tex->desc.macrotile[tex->surface_level] !=
+        tex->desc.macrotile[level]) {
         /* Tiling determines how DRM treats the buffer data.
          * We must flush CS when changing it if the buffer is referenced. */
         if (r300->rws->cs_is_buffer_referenced(r300->cs,
@@ -627,8 +630,8 @@ static void r300_tex_set_tiling_flags(struct r300_context *r300,
             r300->context.flush(&r300->context, 0, NULL);
 
         r300->rws->buffer_set_tiling(r300->rws, tex->buffer,
-                tex->microtile, tex->mip_macrotile[level],
-                tex->pitch[0] * util_format_get_blocksize(tex->b.b.format));
+                tex->desc.microtile, tex->desc.macrotile[level],
+                tex->desc.stride_in_bytes[0]);
 
         tex->surface_level = level;
     }
@@ -670,8 +673,10 @@ static void r300_print_fb_surf_info(struct pipe_surface *surf, unsigned index,
             surf->zslice, surf->face, surf->level,
             util_format_short_name(surf->format),
 
-            rtex->macrotile ? "YES" : " NO", rtex->microtile ? "YES" : " NO",
-            rtex->hwpitch[0], tex->width0, tex->height0, tex->depth0,
+            rtex->desc.macrotile[0] ? "YES" : " NO",
+            rtex->desc.microtile ? "YES" : " NO",
+            rtex->desc.stride_in_pixels[0],
+            tex->width0, tex->height0, tex->depth0,
             tex->last_level, util_format_short_name(tex->format));
 }
 
@@ -679,6 +684,7 @@ void r300_mark_fb_state_dirty(struct r300_context *r300,
                               enum r300_fb_state_change change)
 {
     struct pipe_framebuffer_state *state = r300->fb_state.state;
+    boolean has_hyperz = r300->rws->get_value(r300->rws, R300_CAN_HYPERZ);
 
     /* What is marked as dirty depends on the enum r300_fb_state_change. */
     r300->gpu_flush.dirty = TRUE;
@@ -695,8 +701,11 @@ void r300_mark_fb_state_dirty(struct r300_context *r300,
 
     if (r300->cbzb_clear)
         r300->fb_state.size += 10;
-    else if (state->zsbuf)
-        r300->fb_state.size += r300->screen->caps.has_hiz ? 18 : 14;
+    else if (state->zsbuf) {
+        r300->fb_state.size += 10;
+        if (has_hyperz)
+            r300->fb_state.size += r300->screen->caps.hiz_ram ? 8 : 4;
+    }
 
     /* The size of the rest of atoms stays the same. */
 }
@@ -708,8 +717,10 @@ static void
     struct r300_context* r300 = r300_context(pipe);
     struct r300_aa_state *aa = (struct r300_aa_state*)r300->aa_state.state;
     struct pipe_framebuffer_state *old_state = r300->fb_state.state;
+    boolean has_hyperz = r300->rws->get_value(r300->rws, R300_CAN_HYPERZ);
     unsigned max_width, max_height, i;
     uint32_t zbuffer_bpp = 0;
+    int blocksize;
 
     if (r300->screen->caps.is_r500) {
         max_width = max_height = 4096;
@@ -725,10 +736,6 @@ static void
         return;
     }
 
-    if (r300->draw) {
-        draw_flush(r300->draw);
-    }
-
     /* If nr_cbufs is changed from zero to non-zero or vice versa... */
     if (!!old_state->nr_cbufs != !!state->nr_cbufs) {
         r300->blend_state.dirty = TRUE;
@@ -745,20 +752,50 @@ static void
 
     r300_mark_fb_state_dirty(r300, R300_CHANGED_FB_STATE);
 
-    /* Polygon offset depends on the zbuffer bit depth. */
-    if (state->zsbuf && r300->polygon_offset_enabled) {
-        switch (util_format_get_blocksize(state->zsbuf->texture->format)) {
-            case 2:
-                zbuffer_bpp = 16;
-                break;
-            case 4:
-                zbuffer_bpp = 24;
-                break;
+    r300->z_compression = false;
+    
+    if (state->zsbuf) {
+        blocksize = util_format_get_blocksize(state->zsbuf->texture->format);
+        switch (blocksize) {
+        case 2:
+            zbuffer_bpp = 16;
+            break;
+        case 4:
+            zbuffer_bpp = 24;
+            break;
+        }
+        if (has_hyperz) {
+            struct r300_surface *zs_surf = r300_surface(state->zsbuf);
+            struct r300_texture *tex;
+            int compress = r300->screen->caps.is_rv350 ? RV350_Z_COMPRESS_88 : R300_Z_COMPRESS_44;
+            int level = zs_surf->base.level;
+
+            tex = r300_texture(zs_surf->base.texture);
+
+            /* work out whether we can support hiz on this buffer */
+            r300_hiz_alloc_block(r300, zs_surf);
+        
+            /* work out whether we can support zmask features on this buffer */
+            r300_zmask_alloc_block(r300, zs_surf, compress);
+
+            if (tex->zmask_mem[level]) {
+                /* compression causes hangs on 16-bit */
+                if (zbuffer_bpp == 24)
+                    r300->z_compression = compress;
+            }
+            DBG(r300, DBG_HYPERZ,
+                "hyper-z features: hiz: %d @ %08x z-compression: %d z-fastfill: %d @ %08x\n", tex->hiz_mem[level] ? 1 : 0,
+                tex->hiz_mem[level] ? tex->hiz_mem[level]->ofs : 0xdeadbeef,
+                r300->z_compression, tex->zmask_mem[level] ? 1 : 0,
+                tex->zmask_mem[level] ? tex->zmask_mem[level]->ofs : 0xdeadbeef);
         }
 
+        /* Polygon offset depends on the zbuffer bit depth. */
         if (r300->zbuffer_bpp != zbuffer_bpp) {
             r300->zbuffer_bpp = zbuffer_bpp;
-            r300->rs_state.dirty = TRUE;
+
+            if (r300->polygon_offset_enabled)
+                r300->rs_state.dirty = TRUE;
         }
     }
 
@@ -1016,7 +1053,7 @@ static void* r300_create_rs_state(struct pipe_context* pipe,
 	for (i = 0; i < 8; i++) {
 	    if (state->sprite_coord_enable & (1 << i))
                 stuffing_enable |=
-		    R300_GB_TEX_STR << (R300_GB_TEX0_SOURCE_SHIFT + (i*2));
+                    R300_GB_TEX_ST << (R300_GB_TEX0_SOURCE_SHIFT + (i*2));
 	}
 
         point_texcoord_left = 0.0f;
@@ -1093,14 +1130,11 @@ static void r300_bind_rs_state(struct pipe_context* pipe, void* state)
     boolean last_two_sided_color = r300->two_sided_color;
 
     if (r300->draw && rs) {
-        draw_flush(r300->draw);
         draw_set_rasterizer_state(r300->draw, &rs->rs_draw, state);
     }
 
     if (rs) {
-        r300->polygon_offset_enabled = (rs->rs.offset_point ||
-                                        rs->rs.offset_line ||
-                                        rs->rs.offset_tri);
+        r300->polygon_offset_enabled = rs->polygon_offset_enable;
         r300->sprite_coord_enable = rs->rs.sprite_coord_enable;
         r300->two_sided_color = rs->rs.light_twoside;
     } else {
@@ -1293,7 +1327,7 @@ static void r300_set_fragment_sampler_views(struct pipe_context* pipe,
             /* Set the texrect factor in the fragment shader.
              * Needed for RECT and NPOT fallback. */
             texture = r300_texture(views[i]->texture);
-            if (texture->uses_pitch) {
+            if (texture->desc.is_npot) {
                 r300->fs_rc_constant_state.dirty = TRUE;
             }
 
@@ -1327,6 +1361,7 @@ r300_create_sampler_view(struct pipe_context *pipe,
 {
     struct r300_sampler_view *view = CALLOC_STRUCT(r300_sampler_view);
     struct r300_texture *tex = r300_texture(texture);
+    boolean is_r500 = r300_screen(pipe->screen)->caps.is_r500;
 
     if (view) {
         view->base = *templ;
@@ -1342,8 +1377,9 @@ r300_create_sampler_view(struct pipe_context *pipe,
 
         view->format = tex->tx_format;
         view->format.format1 |= r300_translate_texformat(templ->format,
-                                                         view->swizzle);
-        if (r300_screen(pipe->screen)->caps.is_r500) {
+                                                         view->swizzle,
+                                                         is_r500);
+        if (is_r500) {
             view->format.format2 |= r500_tx_format_msb_bit(templ->format);
         }
     }
@@ -1380,7 +1416,6 @@ static void r300_set_viewport_state(struct pipe_context* pipe,
     r300->viewport = *state;
 
     if (r300->draw) {
-        draw_flush(r300->draw);
         draw_set_viewport_state(r300->draw, state);
         viewport->vte_control = R300_VTX_XY_FMT | R300_VTX_Z_FMT;
         return;
@@ -1481,7 +1516,6 @@ static void r300_set_vertex_buffers(struct pipe_context* pipe,
 
     } else {
         /* SW TCL. */
-        draw_flush(r300->draw);
         draw_set_vertex_buffers(r300->draw, count, buffers);
     }
 
@@ -1500,6 +1534,23 @@ static void r300_set_vertex_buffers(struct pipe_context* pipe,
     r300->vertex_buffer_count = count;
 }
 
+static void r300_set_index_buffer(struct pipe_context* pipe,
+                                  const struct pipe_index_buffer *ib)
+{
+    struct r300_context* r300 = r300_context(pipe);
+
+    if (ib) {
+        pipe_resource_reference(&r300->index_buffer.buffer, ib->buffer);
+        memcpy(&r300->index_buffer, ib, sizeof(r300->index_buffer));
+    }
+    else {
+        pipe_resource_reference(&r300->index_buffer.buffer, NULL);
+        memset(&r300->index_buffer, 0, sizeof(r300->index_buffer));
+    }
+
+    /* TODO make this more like a state */
+}
+
 /* Initialize the PSC tables. */
 static void r300_vertex_psc(struct r300_vertex_element_state *velems)
 {
@@ -1649,7 +1700,6 @@ static void r300_bind_vertex_elements_state(struct pipe_context *pipe,
     r300->velems = velems;
 
     if (r300->draw) {
-        draw_flush(r300->draw);
         draw_set_vertex_elements(r300->draw, velems->count, velems->velem);
         return;
     }
@@ -1701,10 +1751,12 @@ static void r300_bind_vs_state(struct pipe_context* pipe, void* shader)
     r300->rs_block_state.dirty = TRUE; /* Will be updated before the emission. */
 
     if (r300->screen->caps.has_tcl) {
+        unsigned fc_op_dwords = r300->screen->caps.is_r500 ? 3 : 2;
         r300->vs_state.dirty = TRUE;
         r300->vs_state.size =
                 vs->code.length + 9 +
-                (vs->immediates_count ? vs->immediates_count * 4 + 3 : 0);
+                (vs->immediates_count ? vs->immediates_count * 4 + 3 : 0) +
+        (vs->code.num_fc_ops ? vs->code.num_fc_ops * fc_op_dwords + 4 : 0);
 
         if (vs->externals_count) {
             r300->vs_constants.dirty = TRUE;
@@ -1715,7 +1767,6 @@ static void r300_bind_vs_state(struct pipe_context* pipe, void* shader)
 
         r300->pvs_flush.dirty = TRUE;
     } else {
-        draw_flush(r300->draw);
         draw_bind_vertex_shader(r300->draw,
                 (struct draw_vertex_shader*)vs->draw_vs);
     }
@@ -1847,6 +1898,7 @@ void r300_init_state_functions(struct r300_context* r300)
     r300->context.set_viewport_state = r300_set_viewport_state;
 
     r300->context.set_vertex_buffers = r300_set_vertex_buffers;
+    r300->context.set_index_buffer = r300_set_index_buffer;
 
     r300->context.create_vertex_elements_state = r300_create_vertex_elements_state;
     r300->context.bind_vertex_elements_state = r300_bind_vertex_elements_state;
diff --git a/src/gallium/drivers/r300/r300_state_derived.c b/src/gallium/drivers/r300/r300_state_derived.c
index 2ef9766578c..4a63ed7fc13 100644
--- a/src/gallium/drivers/r300/r300_state_derived.c
+++ b/src/gallium/drivers/r300/r300_state_derived.c
@@ -46,6 +46,11 @@ enum r300_rs_swizzle {
     SWIZ_0001,
 };
 
+enum r300_rs_col_write_type {
+    WRITE_COLOR = 0,
+    WRITE_FACE
+};
+
 static void r300_draw_emit_attrib(struct r300_context* r300,
                                   enum attrib_emit emit,
                                   enum interp_mode interp,
@@ -203,8 +208,10 @@ static void r300_rs_col(struct r300_rs_block* rs, int id, int ptr,
     rs->inst[id] |= R300_RS_INST_COL_ID(id);
 }
 
-static void r300_rs_col_write(struct r300_rs_block* rs, int id, int fp_offset)
+static void r300_rs_col_write(struct r300_rs_block* rs, int id, int fp_offset,
+                              enum r300_rs_col_write_type type)
 {
+    assert(type != WRITE_COLOR);
     rs->inst[id] |= R300_RS_INST_COL_CN_WRITE |
                     R300_RS_INST_COL_ADDR(fp_offset);
 }
@@ -213,19 +220,19 @@ static void r300_rs_tex(struct r300_rs_block* rs, int id, int ptr,
                         enum r300_rs_swizzle swiz)
 {
     if (swiz == SWIZ_X001) {
-        rs->ip[id] |= R300_RS_TEX_PTR(ptr*4) |
+        rs->ip[id] |= R300_RS_TEX_PTR(ptr) |
                       R300_RS_SEL_S(R300_RS_SEL_C0) |
                       R300_RS_SEL_T(R300_RS_SEL_K0) |
                       R300_RS_SEL_R(R300_RS_SEL_K0) |
                       R300_RS_SEL_Q(R300_RS_SEL_K1);
     } else if (swiz == SWIZ_XY01) {
-        rs->ip[id] |= R300_RS_TEX_PTR(ptr*4) |
+        rs->ip[id] |= R300_RS_TEX_PTR(ptr) |
                       R300_RS_SEL_S(R300_RS_SEL_C0) |
                       R300_RS_SEL_T(R300_RS_SEL_C1) |
                       R300_RS_SEL_R(R300_RS_SEL_K0) |
                       R300_RS_SEL_Q(R300_RS_SEL_K1);
     } else {
-        rs->ip[id] |= R300_RS_TEX_PTR(ptr*4) |
+        rs->ip[id] |= R300_RS_TEX_PTR(ptr) |
                       R300_RS_SEL_S(R300_RS_SEL_C0) |
                       R300_RS_SEL_T(R300_RS_SEL_C1) |
                       R300_RS_SEL_R(R300_RS_SEL_C2) |
@@ -252,32 +259,36 @@ static void r500_rs_col(struct r300_rs_block* rs, int id, int ptr,
     rs->inst[id] |= R500_RS_INST_COL_ID(id);
 }
 
-static void r500_rs_col_write(struct r300_rs_block* rs, int id, int fp_offset)
+static void r500_rs_col_write(struct r300_rs_block* rs, int id, int fp_offset,
+                              enum r300_rs_col_write_type type)
 {
-    rs->inst[id] |= R500_RS_INST_COL_CN_WRITE |
-                    R500_RS_INST_COL_ADDR(fp_offset);
+    if (type == WRITE_FACE)
+        rs->inst[id] |= R500_RS_INST_COL_CN_WRITE_BACKFACE |
+                        R500_RS_INST_COL_ADDR(fp_offset);
+    else
+        rs->inst[id] |= R500_RS_INST_COL_CN_WRITE |
+                        R500_RS_INST_COL_ADDR(fp_offset);
+
 }
 
 static void r500_rs_tex(struct r300_rs_block* rs, int id, int ptr,
 			enum r300_rs_swizzle swiz)
 {
-    int rs_tex_comp = ptr*4;
-
     if (swiz == SWIZ_X001) {
-        rs->ip[id] |= R500_RS_SEL_S(rs_tex_comp) |
+        rs->ip[id] |= R500_RS_SEL_S(ptr) |
                       R500_RS_SEL_T(R500_RS_IP_PTR_K0) |
                       R500_RS_SEL_R(R500_RS_IP_PTR_K0) |
                       R500_RS_SEL_Q(R500_RS_IP_PTR_K1);
     } else if (swiz == SWIZ_XY01) {
-        rs->ip[id] |= R500_RS_SEL_S(rs_tex_comp) |
-                      R500_RS_SEL_T(rs_tex_comp + 1) |
+        rs->ip[id] |= R500_RS_SEL_S(ptr) |
+                      R500_RS_SEL_T(ptr + 1) |
                       R500_RS_SEL_R(R500_RS_IP_PTR_K0) |
                       R500_RS_SEL_Q(R500_RS_IP_PTR_K1);
     } else {
-        rs->ip[id] |= R500_RS_SEL_S(rs_tex_comp) |
-                      R500_RS_SEL_T(rs_tex_comp + 1) |
-                      R500_RS_SEL_R(rs_tex_comp + 2) |
-                      R500_RS_SEL_Q(rs_tex_comp + 3);
+        rs->ip[id] |= R500_RS_SEL_S(ptr) |
+                      R500_RS_SEL_T(ptr + 1) |
+                      R500_RS_SEL_R(ptr + 2) |
+                      R500_RS_SEL_Q(ptr + 3);
     }
     rs->inst[id] |= R500_RS_INST_TEX_ID(id);
 }
@@ -305,9 +316,9 @@ static void r300_update_rs_block(struct r300_context *r300)
     struct r300_shader_semantics *vs_outputs = &vs->outputs;
     struct r300_shader_semantics *fs_inputs = &r300_fs(r300)->shader->inputs;
     struct r300_rs_block rs = {0};
-    int i, col_count = 0, tex_count = 0, fp_offset = 0, count, loc = 0;
+    int i, col_count = 0, tex_count = 0, fp_offset = 0, count, loc = 0, tex_ptr = 0;
     void (*rX00_rs_col)(struct r300_rs_block*, int, int, enum r300_rs_swizzle);
-    void (*rX00_rs_col_write)(struct r300_rs_block*, int, int);
+    void (*rX00_rs_col_write)(struct r300_rs_block*, int, int, enum r300_rs_col_write_type);
     void (*rX00_rs_tex)(struct r300_rs_block*, int, int, enum r300_rs_swizzle);
     void (*rX00_rs_tex_write)(struct r300_rs_block*, int, int);
     boolean any_bcolor_used = vs_outputs->bcolor[0] != ATTR_UNUSED ||
@@ -326,6 +337,11 @@ static void r300_update_rs_block(struct r300_context *r300)
         rX00_rs_tex_write = r300_rs_tex_write;
     }
 
+    /* 0x5555 copied from classic, which means:
+     * Select user color 0 for COLOR0 up to COLOR7.
+     * What the hell does that mean? */
+    rs.vap_vtx_state_cntl = 0x5555;
+
     /* The position is always present in VAP. */
     rs.vap_vsm_vtx_assm |= R300_INPUT_CNTL_POS;
     rs.vap_out_vtx_fmt[0] |= R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT;
@@ -352,7 +368,7 @@ static void r300_update_rs_block(struct r300_context *r300)
 
             /* Write it to the FS input register if it's needed by the FS. */
             if (fs_inputs->color[i] != ATTR_UNUSED) {
-                rX00_rs_col_write(&rs, col_count, fp_offset);
+                rX00_rs_col_write(&rs, col_count, fp_offset, WRITE_COLOR);
                 fp_offset++;
 
                 DBG(r300, DBG_RS,
@@ -393,12 +409,31 @@ static void r300_update_rs_block(struct r300_context *r300)
                 stream_loc_notcl[loc++] = 6 + tex_count;
 
                 /* Rasterize it. */
-                rX00_rs_tex(&rs, tex_count, tex_count, SWIZ_XYZW);
+                rX00_rs_tex(&rs, tex_count, tex_ptr, SWIZ_XYZW);
                 tex_count++;
+                tex_ptr += 4;
             }
         }
     }
 
+    /* gl_FrontFacing.
+     * Note that we can use either the two-sided color selection based on
+     * the front and back vertex shader colors, or gl_FrontFacing,
+     * but not both! It locks up otherwise.
+     *
+     * In Direct3D 9, the two-sided color selection can be used
+     * with shaders 2.0 only, while gl_FrontFacing can be used
+     * with shaders 3.0 only. The hardware apparently hasn't been designed
+     * to support both at the same time. */
+    if (r300->screen->caps.is_r500 && fs_inputs->face != ATTR_UNUSED &&
+        !(any_bcolor_used && r300->two_sided_color)) {
+        rX00_rs_col(&rs, col_count, col_count, SWIZ_XYZW);
+        rX00_rs_col_write(&rs, col_count, fp_offset, WRITE_FACE);
+        fp_offset++;
+        col_count++;
+        DBG(r300, DBG_RS, "r300: Rasterized FACE written to FS.\n");
+    }
+
     /* Rasterize texture coordinates. */
     for (i = 0; i < ATTR_GENERIC_COUNT && tex_count < 8; i++) {
 	bool sprite_coord = !!(r300->sprite_coord_enable & (1 << i));
@@ -412,7 +447,7 @@ static void r300_update_rs_block(struct r300_context *r300)
             }
 
             /* Rasterize it. */
-            rX00_rs_tex(&rs, tex_count, tex_count,
+            rX00_rs_tex(&rs, tex_count, tex_ptr,
 			sprite_coord ? SWIZ_XY01 : SWIZ_XYZW);
 
             /* Write it to the FS input register if it's needed by the FS. */
@@ -429,6 +464,7 @@ static void r300_update_rs_block(struct r300_context *r300)
                     i, sprite_coord ? " (sprite coord)" : "");
             }
             tex_count++;
+            tex_ptr += sprite_coord ? 2 : 4;
         } else {
             /* Skip the FS input register, leave it uninitialized. */
             /* If we try to set it to (0,0,0,1), it will lock up. */
@@ -449,7 +485,7 @@ static void r300_update_rs_block(struct r300_context *r300)
         stream_loc_notcl[loc++] = 6 + tex_count;
 
         /* Rasterize it. */
-        rX00_rs_tex(&rs, tex_count, tex_count, SWIZ_X001);
+        rX00_rs_tex(&rs, tex_count, tex_ptr, SWIZ_X001);
 
         /* Write it to the FS input register if it's needed by the FS. */
         if (fs_inputs->fog != ATTR_UNUSED) {
@@ -461,6 +497,7 @@ static void r300_update_rs_block(struct r300_context *r300)
             DBG(r300, DBG_RS, "r300: Rasterized fog unused.\n");
         }
         tex_count++;
+        tex_ptr += 4;
     } else {
         /* Skip the FS input register, leave it uninitialized. */
         /* If we try to set it to (0,0,0,1), it will lock up. */
@@ -480,7 +517,7 @@ static void r300_update_rs_block(struct r300_context *r300)
         stream_loc_notcl[loc++] = 6 + tex_count;
 
         /* Rasterize it. */
-        rX00_rs_tex(&rs, tex_count, tex_count, SWIZ_XYZW);
+        rX00_rs_tex(&rs, tex_count, tex_ptr, SWIZ_XYZW);
 
         /* Write it to the FS input register. */
         rX00_rs_tex_write(&rs, tex_count, fp_offset);
@@ -489,6 +526,7 @@ static void r300_update_rs_block(struct r300_context *r300)
 
         fp_offset++;
         tex_count++;
+        tex_ptr += 4;
     }
 
     /* Invalidate the rest of the no-TCL (GA) stream locations. */
@@ -507,7 +545,7 @@ static void r300_update_rs_block(struct r300_context *r300)
     DBG(r300, DBG_RS, "r300: --- Rasterizer status ---: colors: %i, "
         "generics: %i.\n", col_count, tex_count);
 
-    rs.count = (tex_count*4) | (col_count << R300_IC_COUNT_SHIFT) |
+    rs.count = MIN2(tex_ptr, 32) | (col_count << R300_IC_COUNT_SHIFT) |
         R300_HIRES_EN;
 
     count = MAX3(col_count, tex_count, 1);
@@ -528,15 +566,9 @@ static void r300_merge_textures_and_samplers(struct r300_context* r300)
     struct r300_sampler_state *sampler;
     struct r300_sampler_view *view;
     struct r300_texture *tex;
-    unsigned min_level, max_level, i, size;
+    unsigned min_level, max_level, i, j, size;
     unsigned count = MIN2(state->sampler_view_count,
                           state->sampler_state_count);
-    unsigned char depth_swizzle[4] = {
-        UTIL_FORMAT_SWIZZLE_X,
-        UTIL_FORMAT_SWIZZLE_X,
-        UTIL_FORMAT_SWIZZLE_X,
-        UTIL_FORMAT_SWIZZLE_X
-    };
 
     /* The KIL opcode fix, see below. */
     if (!count && !r300->screen->caps.is_r500)
@@ -563,14 +595,29 @@ static void r300_merge_textures_and_samplers(struct r300_context* r300)
             /* Assign a texture cache region. */
             texstate->format.format1 |= view->texcache_region;
 
-            /* If compare mode is disabled, the sampler view swizzles
-             * are stored in the format.
-             * Otherwise, swizzles must be applied after the compare mode
-             * in the fragment shader. */
-            if (util_format_is_depth_or_stencil(tex->b.b.format)) {
+            /* Depth textures are kinda special. */
+            if (util_format_is_depth_or_stencil(tex->desc.b.b.format)) {
+                unsigned char depth_swizzle[4];
+
+                if (!r300->screen->caps.is_r500 &&
+                    util_format_get_blocksizebits(tex->desc.b.b.format) == 32) {
+                    /* X24x8 is sampled as Y16X16 on r3xx-r4xx.
+                     * The depth here is at the Y component. */
+                    for (j = 0; j < 4; j++)
+                        depth_swizzle[j] = UTIL_FORMAT_SWIZZLE_Y;
+                } else {
+                    for (j = 0; j < 4; j++)
+                        depth_swizzle[j] = UTIL_FORMAT_SWIZZLE_X;
+                }
+
+                /* If compare mode is disabled, sampler view swizzles
+                 * are stored in the format.
+                 * Otherwise, the swizzles must be applied after the compare
+                 * mode in the fragment shader. */
                 if (sampler->state.compare_mode == PIPE_TEX_COMPARE_NONE) {
                     texstate->format.format1 |=
-                        r300_get_swizzle_combined(depth_swizzle, view->swizzle);
+                        r300_get_swizzle_combined(depth_swizzle,
+                                                  view->swizzle);
                 } else {
                     texstate->format.format1 |=
                         r300_get_swizzle_combined(depth_swizzle, 0);
@@ -578,12 +625,12 @@ static void r300_merge_textures_and_samplers(struct r300_context* r300)
             }
 
             /* to emulate 1D textures through 2D ones correctly */
-            if (tex->b.b.target == PIPE_TEXTURE_1D) {
+            if (tex->desc.b.b.target == PIPE_TEXTURE_1D) {
                 texstate->filter0 &= ~R300_TX_WRAP_T_MASK;
                 texstate->filter0 |= R300_TX_WRAP_T(R300_TX_CLAMP_TO_EDGE);
             }
 
-            if (tex->uses_pitch) {
+            if (tex->desc.is_npot) {
                 /* NPOT textures don't support mip filter, unfortunately.
                  * This prevents incorrect rendering. */
                 texstate->filter0 &= ~R300_TX_MIN_FILTER_MIP_MASK;
@@ -610,7 +657,7 @@ static void r300_merge_textures_and_samplers(struct r300_context* r300)
                 /* determine min/max levels */
                 /* the MAX_MIP level is the largest (finest) one */
                 max_level = MIN3(sampler->max_lod + view->base.first_level,
-                                 tex->b.b.last_level, view->base.last_level);
+                                 tex->desc.b.b.last_level, view->base.last_level);
                 min_level = MIN2(sampler->min_lod + view->base.first_level,
                                  max_level);
                 texstate->format.format0 |= R300_TX_NUM_LEVELS(max_level);
@@ -665,8 +712,44 @@ static void r300_merge_textures_and_samplers(struct r300_context* r300)
     }
 }
 
+/* We can't use compressed zbuffers as samplers. */
+static void r300_flush_depth_textures(struct r300_context *r300)
+{
+    struct r300_textures_state *state =
+        (struct r300_textures_state*)r300->textures_state.state;
+    unsigned i, level;
+    unsigned count = MIN2(state->sampler_view_count,
+                          state->sampler_state_count);
+
+    if (r300->z_decomp_rd)
+        return;
+
+    for (i = 0; i < count; i++)
+        if (state->sampler_views[i] && state->sampler_states[i]) {
+            struct pipe_resource *tex = state->sampler_views[i]->base.texture;
+
+            if (tex->target == PIPE_TEXTURE_3D ||
+                tex->target == PIPE_TEXTURE_CUBE)
+                continue;
+
+            /* Ignore non-depth textures.
+             * Also ignore reinterpreted depth textures, e.g. resource_copy. */
+            if (!util_format_is_depth_or_stencil(tex->format))
+                continue;
+
+            for (level = 0; level <= tex->last_level; level++)
+                if (r300_texture(tex)->zmask_in_use[level]) {
+                    /* We don't handle 3D textures and cubemaps yet. */
+                    r300_flush_depth_stencil(&r300->context, tex,
+                                             u_subresource(0, level), 0);
+                }
+        }
+}
+
 void r300_update_derived_state(struct r300_context* r300)
 {
+    r300_flush_depth_textures(r300);
+
     if (r300->textures_state.dirty) {
         r300_merge_textures_and_samplers(r300);
     }
diff --git a/src/gallium/drivers/r300/r300_texture.c b/src/gallium/drivers/r300/r300_texture.c
index 5750bc4329f..da8eadd3b53 100644
--- a/src/gallium/drivers/r300/r300_texture.c
+++ b/src/gallium/drivers/r300/r300_texture.c
@@ -26,6 +26,7 @@
 
 #include "r300_context.h"
 #include "r300_reg.h"
+#include "r300_texture_desc.h"
 #include "r300_transfer.h"
 #include "r300_screen.h"
 #include "r300_winsys.h"
@@ -34,14 +35,10 @@
 #include "util/u_format_s3tc.h"
 #include "util/u_math.h"
 #include "util/u_memory.h"
+#include "util/u_mm.h"
 
 #include "pipe/p_screen.h"
 
-enum r300_dim {
-    DIM_WIDTH  = 0,
-    DIM_HEIGHT = 1
-};
-
 unsigned r300_get_swizzle_combined(const unsigned char *swizzle_format,
                                    const unsigned char *swizzle_view)
 {
@@ -109,7 +106,8 @@ unsigned r300_get_swizzle_combined(const unsigned char *swizzle_format,
  * The FORMAT specifies how the texture sampler will treat the texture, and
  * makes available X, Y, Z, W, ZERO, and ONE for swizzling. */
 uint32_t r300_translate_texformat(enum pipe_format format,
-                                  const unsigned char *swizzle_view)
+                                  const unsigned char *swizzle_view,
+                                  boolean is_r500)
 {
     uint32_t result = 0;
     const struct util_format_description *desc;
@@ -134,7 +132,10 @@ uint32_t r300_translate_texformat(enum pipe_format format,
                     return R300_TX_FORMAT_X16;
                 case PIPE_FORMAT_X8Z24_UNORM:
                 case PIPE_FORMAT_S8_USCALED_Z24_UNORM:
-                    return R500_TX_FORMAT_Y8X24;
+                    if (is_r500)
+                        return R500_TX_FORMAT_Y8X24;
+                    else
+                        return R300_TX_FORMAT_Y16X16;
                 default:
                     return ~0; /* Unsupported. */
             }
@@ -537,26 +538,27 @@ boolean r300_is_zs_format_supported(enum pipe_format format)
 
 boolean r300_is_sampler_format_supported(enum pipe_format format)
 {
-    return r300_translate_texformat(format, 0) != ~0;
+    return r300_translate_texformat(format, 0, TRUE) != ~0;
 }
 
 static void r300_texture_setup_immutable_state(struct r300_screen* screen,
                                                struct r300_texture* tex)
 {
     struct r300_texture_format_state* f = &tex->tx_format;
-    struct pipe_resource *pt = &tex->b.b;
+    struct pipe_resource *pt = &tex->desc.b.b;
     boolean is_r500 = screen->caps.is_r500;
 
     /* Set sampler state. */
     f->format0 = R300_TX_WIDTH((pt->width0 - 1) & 0x7ff) |
                  R300_TX_HEIGHT((pt->height0 - 1) & 0x7ff);
 
-    if (tex->uses_pitch) {
+    if (tex->desc.uses_stride_addressing) {
         /* rectangles love this */
         f->format0 |= R300_TX_PITCH_EN;
-        f->format2 = (tex->hwpitch[0] - 1) & 0x1fff;
+        f->format2 = (tex->desc.stride_in_pixels[0] - 1) & 0x1fff;
     } else {
-        /* power of two textures (3D, mipmaps, and no pitch) */
+        /* Power of two textures (3D, mipmaps, and no pitch),
+         * also NPOT textures with a width being POT. */
         f->format0 |= R300_TX_DEPTH(util_logbase2(pt->depth0) & 0xf);
     }
 
@@ -579,8 +581,8 @@ static void r300_texture_setup_immutable_state(struct r300_screen* screen,
         }
     }
 
-    f->tile_config = R300_TXO_MACRO_TILE(tex->macrotile) |
-                     R300_TXO_MICRO_TILE(tex->microtile);
+    f->tile_config = R300_TXO_MACRO_TILE(tex->desc.macrotile[0]) |
+                     R300_TXO_MICRO_TILE(tex->desc.microtile);
 }
 
 static void r300_texture_setup_fb_state(struct r300_screen* screen,
@@ -589,23 +591,23 @@ static void r300_texture_setup_fb_state(struct r300_screen* screen,
     unsigned i;
 
     /* Set framebuffer state. */
-    if (util_format_is_depth_or_stencil(tex->b.b.format)) {
-        for (i = 0; i <= tex->b.b.last_level; i++) {
+    if (util_format_is_depth_or_stencil(tex->desc.b.b.format)) {
+        for (i = 0; i <= tex->desc.b.b.last_level; i++) {
             tex->fb_state.pitch[i] =
-                tex->hwpitch[i] |
-                R300_DEPTHMACROTILE(tex->mip_macrotile[i]) |
-                R300_DEPTHMICROTILE(tex->microtile);
+                tex->desc.stride_in_pixels[i] |
+                R300_DEPTHMACROTILE(tex->desc.macrotile[i]) |
+                R300_DEPTHMICROTILE(tex->desc.microtile);
         }
-        tex->fb_state.format = r300_translate_zsformat(tex->b.b.format);
+        tex->fb_state.format = r300_translate_zsformat(tex->desc.b.b.format);
     } else {
-        for (i = 0; i <= tex->b.b.last_level; i++) {
+        for (i = 0; i <= tex->desc.b.b.last_level; i++) {
             tex->fb_state.pitch[i] =
-                tex->hwpitch[i] |
-                r300_translate_colorformat(tex->b.b.format) |
-                R300_COLOR_TILE(tex->mip_macrotile[i]) |
-                R300_COLOR_MICROTILE(tex->microtile);
+                tex->desc.stride_in_pixels[i] |
+                r300_translate_colorformat(tex->desc.b.b.format) |
+                R300_COLOR_TILE(tex->desc.macrotile[i]) |
+                R300_COLOR_MICROTILE(tex->desc.microtile);
         }
-        tex->fb_state.format = r300_translate_out_fmt(tex->b.b.format);
+        tex->fb_state.format = r300_translate_out_fmt(tex->desc.b.b.format);
     }
 }
 
@@ -625,282 +627,6 @@ void r300_texture_reinterpret_format(struct pipe_screen *screen,
     r300_texture_setup_fb_state(r300_screen(screen), r300_texture(tex));
 }
 
-unsigned r300_texture_get_offset(struct r300_texture* tex, unsigned level,
-                                 unsigned zslice, unsigned face)
-{
-    unsigned offset = tex->offset[level];
-
-    switch (tex->b.b.target) {
-        case PIPE_TEXTURE_3D:
-            assert(face == 0);
-            return offset + zslice * tex->layer_size[level];
-
-        case PIPE_TEXTURE_CUBE:
-            assert(zslice == 0);
-            return offset + face * tex->layer_size[level];
-
-        default:
-            assert(zslice == 0 && face == 0);
-            return offset;
-    }
-}
-
-/* Returns the number of pixels that the texture should be aligned to
- * in the given dimension. */
-static unsigned r300_get_pixel_alignment(struct r300_texture *tex,
-                                         enum r300_buffer_tiling macrotile,
-                                         enum r300_dim dim)
-{
-    static const unsigned table[2][5][3][2] =
-    {
-        {
-    /* Macro: linear    linear    linear
-       Micro: linear    tiled  square-tiled */
-            {{ 32, 1}, { 8,  4}, { 0,  0}}, /*   8 bits per pixel */
-            {{ 16, 1}, { 8,  2}, { 4,  4}}, /*  16 bits per pixel */
-            {{  8, 1}, { 4,  2}, { 0,  0}}, /*  32 bits per pixel */
-            {{  4, 1}, { 0,  0}, { 2,  2}}, /*  64 bits per pixel */
-            {{  2, 1}, { 0,  0}, { 0,  0}}  /* 128 bits per pixel */
-        },
-        {
-    /* Macro: tiled     tiled     tiled
-       Micro: linear    tiled  square-tiled */
-            {{256, 8}, {64, 32}, { 0,  0}}, /*   8 bits per pixel */
-            {{128, 8}, {64, 16}, {32, 32}}, /*  16 bits per pixel */
-            {{ 64, 8}, {32, 16}, { 0,  0}}, /*  32 bits per pixel */
-            {{ 32, 8}, { 0,  0}, {16, 16}}, /*  64 bits per pixel */
-            {{ 16, 8}, { 0,  0}, { 0,  0}}  /* 128 bits per pixel */
-        }
-    };
-    static const unsigned aa_block[2] = {4, 8};
-    unsigned res = 0;
-    unsigned pixsize = util_format_get_blocksize(tex->b.b.format);
-
-    assert(macrotile <= R300_BUFFER_TILED);
-    assert(tex->microtile <= R300_BUFFER_SQUARETILED);
-    assert(pixsize <= 16);
-    assert(dim <= DIM_HEIGHT);
-
-    if (tex->b.b.nr_samples > 1) {
-        /* Multisampled textures have their own alignment scheme. */
-        if (pixsize == 4)
-            res = aa_block[dim];
-    } else {
-        /* Standard alignment. */
-        res = table[macrotile][util_logbase2(pixsize)][tex->microtile][dim];
-    }
-
-    assert(res);
-    return res;
-}
-
-/* Return true if macrotiling should be enabled on the miplevel. */
-static boolean r300_texture_macro_switch(struct r300_texture *tex,
-                                         unsigned level,
-                                         boolean rv350_mode,
-                                         enum r300_dim dim)
-{
-    unsigned tile, texdim;
-
-    tile = r300_get_pixel_alignment(tex, R300_BUFFER_TILED, dim);
-    if (dim == DIM_WIDTH) {
-        texdim = u_minify(tex->b.b.width0, level);
-    } else {
-        texdim = u_minify(tex->b.b.height0, level);
-    }
-
-    /* See TX_FILTER1_n.MACRO_SWITCH. */
-    if (rv350_mode) {
-        return texdim >= tile;
-    } else {
-        return texdim > tile;
-    }
-}
-
-/**
- * Return the stride, in bytes, of the texture images of the given texture
- * at the given level.
- */
-unsigned r300_texture_get_stride(struct r300_screen* screen,
-                                 struct r300_texture* tex, unsigned level)
-{
-    unsigned tile_width, width, stride;
-
-    if (tex->stride_override)
-        return tex->stride_override;
-
-    /* Check the level. */
-    if (level > tex->b.b.last_level) {
-        SCREEN_DBG(screen, DBG_TEX, "%s: level (%u) > last_level (%u)\n",
-                   __FUNCTION__, level, tex->b.b.last_level);
-        return 0;
-    }
-
-    width = u_minify(tex->b.b.width0, level);
-
-    if (util_format_is_plain(tex->b.b.format)) {
-        tile_width = r300_get_pixel_alignment(tex, tex->mip_macrotile[level],
-                                              DIM_WIDTH);
-        width = align(width, tile_width);
-
-        stride = util_format_get_stride(tex->b.b.format, width);
-
-        /* Some IGPs need a minimum stride of 64 bytes, hmm...
-         * This doesn't seem to apply to tiled textures, according to r300c. */
-        if (!tex->microtile && !tex->mip_macrotile[level] &&
-            (screen->caps.family == CHIP_FAMILY_RS600 ||
-             screen->caps.family == CHIP_FAMILY_RS690 ||
-             screen->caps.family == CHIP_FAMILY_RS740)) {
-            return stride < 64 ? 64 : stride;
-        }
-
-        /* The alignment to 32 bytes is sort of implied by the layout... */
-        return stride;
-    } else {
-        return align(util_format_get_stride(tex->b.b.format, width), 32);
-    }
-}
-
-static unsigned r300_texture_get_nblocksy(struct r300_texture* tex,
-                                          unsigned level)
-{
-    unsigned height, tile_height;
-
-    height = u_minify(tex->b.b.height0, level);
-
-    if (util_format_is_plain(tex->b.b.format)) {
-        tile_height = r300_get_pixel_alignment(tex, tex->mip_macrotile[level],
-                                               DIM_HEIGHT);
-        height = align(height, tile_height);
-
-        /* This is needed for the kernel checker, unfortunately. */
-        height = util_next_power_of_two(height);
-    }
-
-    return util_format_get_nblocksy(tex->b.b.format, height);
-}
-
-static void r300_texture_3d_fix_mipmapping(struct r300_screen *screen,
-                                           struct r300_texture *tex)
-{
-    /* The kernels <= 2.6.34-rc4 compute the size of mipmapped 3D textures
-     * incorrectly. This is a workaround to prevent CS from being rejected. */
-
-    unsigned i, size;
-
-    if (!screen->rws->get_value(screen->rws, R300_VID_DRM_2_3_0) &&
-        tex->b.b.target == PIPE_TEXTURE_3D &&
-        tex->b.b.last_level > 0) {
-        size = 0;
-
-        for (i = 0; i <= tex->b.b.last_level; i++) {
-            size += r300_texture_get_stride(screen, tex, i) *
-                    r300_texture_get_nblocksy(tex, i);
-        }
-
-        size *= tex->b.b.depth0;
-        tex->size = size;
-    }
-}
-
-static void r300_setup_miptree(struct r300_screen* screen,
-                               struct r300_texture* tex)
-{
-    struct pipe_resource* base = &tex->b.b;
-    unsigned stride, size, layer_size, nblocksy, i;
-    boolean rv350_mode = screen->caps.is_rv350;
-
-    SCREEN_DBG(screen, DBG_TEXALLOC,
-        "r300: Making miptree for texture, format %s\n",
-        util_format_short_name(base->format));
-
-    for (i = 0; i <= base->last_level; i++) {
-        /* Let's see if this miplevel can be macrotiled. */
-        tex->mip_macrotile[i] =
-            (tex->macrotile == R300_BUFFER_TILED &&
-             r300_texture_macro_switch(tex, i, rv350_mode, DIM_WIDTH) &&
-             r300_texture_macro_switch(tex, i, rv350_mode, DIM_HEIGHT)) ?
-             R300_BUFFER_TILED : R300_BUFFER_LINEAR;
-
-        stride = r300_texture_get_stride(screen, tex, i);
-        nblocksy = r300_texture_get_nblocksy(tex, i);
-        layer_size = stride * nblocksy;
-
-        if (base->nr_samples) {
-            layer_size *= base->nr_samples;
-        }
-
-        if (base->target == PIPE_TEXTURE_CUBE)
-            size = layer_size * 6;
-        else
-            size = layer_size * u_minify(base->depth0, i);
-
-        tex->offset[i] = tex->size;
-        tex->size = tex->offset[i] + size;
-        tex->layer_size[i] = layer_size;
-        tex->pitch[i] = stride / util_format_get_blocksize(base->format);
-        tex->hwpitch[i] =
-                tex->pitch[i] * util_format_get_blockwidth(base->format);
-
-        SCREEN_DBG(screen, DBG_TEXALLOC, "r300: Texture miptree: Level %d "
-                "(%dx%dx%d px, pitch %d bytes) %d bytes total, macrotiled %s\n",
-                i, u_minify(base->width0, i), u_minify(base->height0, i),
-                u_minify(base->depth0, i), stride, tex->size,
-                tex->mip_macrotile[i] ? "TRUE" : "FALSE");
-    }
-}
-
-static void r300_setup_flags(struct r300_texture* tex)
-{
-    tex->uses_pitch = !util_is_power_of_two(tex->b.b.width0) ||
-                      !util_is_power_of_two(tex->b.b.height0) ||
-                      tex->stride_override;
-}
-
-static void r300_setup_tiling(struct pipe_screen *screen,
-                              struct r300_texture *tex)
-{
-    struct r300_winsys_screen *rws = (struct r300_winsys_screen *)screen->winsys;
-    enum pipe_format format = tex->b.b.format;
-    boolean rv350_mode = r300_screen(screen)->caps.is_rv350;
-    boolean is_zb = util_format_is_depth_or_stencil(format);
-    boolean dbg_no_tiling = SCREEN_DBG_ON(r300_screen(screen), DBG_NO_TILING);
-
-    if (!util_format_is_plain(format)) {
-        return;
-    }
-
-    /* If height == 1, disable microtiling except for zbuffer. */
-    if (!is_zb && (tex->b.b.height0 == 1 || dbg_no_tiling)) {
-        return;
-    }
-
-    /* Set microtiling. */
-    switch (util_format_get_blocksize(format)) {
-        case 1:
-        case 4:
-            tex->microtile = R300_BUFFER_TILED;
-            break;
-
-        case 2:
-        case 8:
-            if (rws->get_value(rws, R300_VID_SQUARE_TILING_SUPPORT)) {
-                tex->microtile = R300_BUFFER_SQUARETILED;
-            }
-            break;
-    }
-
-    if (dbg_no_tiling) {
-        return;
-    }
-
-    /* Set macrotiling. */
-    if (r300_texture_macro_switch(tex, 0, rv350_mode, DIM_WIDTH) &&
-        r300_texture_macro_switch(tex, 0, rv350_mode, DIM_HEIGHT)) {
-        tex->macrotile = R300_BUFFER_TILED;
-    }
-}
-
 static unsigned r300_texture_is_referenced(struct pipe_context *context,
 					 struct pipe_resource *texture,
 					 unsigned face, unsigned level)
@@ -920,8 +646,16 @@ static void r300_texture_destroy(struct pipe_screen *screen,
 {
     struct r300_texture* tex = (struct r300_texture*)texture;
     struct r300_winsys_screen *rws = (struct r300_winsys_screen *)texture->screen->winsys;
+    int i;
 
     rws->buffer_reference(rws, &tex->buffer, NULL);
+    for (i = 0; i < R300_MAX_TEXTURE_LEVELS; i++) {
+        if (tex->hiz_mem[i])
+            u_mmFreeMem(tex->hiz_mem[i]);
+        if (tex->zmask_mem[i])
+            u_mmFreeMem(tex->zmask_mem[i]);
+    }
+
     FREE(tex);
 }
 
@@ -937,11 +671,10 @@ static boolean r300_texture_get_handle(struct pipe_screen* screen,
     }
 
     return rws->buffer_get_handle(rws, tex->buffer,
-                    r300_texture_get_stride(r300_screen(screen), tex, 0),
-                    whandle);
+                                  tex->desc.stride_in_bytes[0], whandle);
 }
 
-struct u_resource_vtbl r300_texture_vtbl = 
+struct u_resource_vtbl r300_texture_vtbl =
 {
    r300_texture_get_handle,	      /* get_handle */
    r300_texture_destroy,	      /* resource_destroy */
@@ -954,17 +687,69 @@ struct u_resource_vtbl r300_texture_vtbl =
    u_default_transfer_inline_write    /* transfer_inline_write */
 };
 
-/* Create a new texture. */
-struct pipe_resource* r300_texture_create(struct pipe_screen* screen,
-                                          const struct pipe_resource* base)
+/* The common texture constructor. */
+static struct r300_texture*
+r300_texture_create_object(struct r300_screen *rscreen,
+                           const struct pipe_resource *base,
+                           enum r300_buffer_tiling microtile,
+                           enum r300_buffer_tiling macrotile,
+                           unsigned stride_in_bytes_override,
+                           unsigned max_buffer_size,
+                           struct r300_winsys_buffer *buffer)
 {
-    struct r300_texture* tex = CALLOC_STRUCT(r300_texture);
-    struct r300_screen* rscreen = r300_screen(screen);
-    struct r300_winsys_screen *rws = (struct r300_winsys_screen *)screen->winsys;
-
+    struct r300_winsys_screen *rws = rscreen->rws;
+    struct r300_texture *tex = CALLOC_STRUCT(r300_texture);
     if (!tex) {
+        if (buffer)
+            rws->buffer_reference(rws, &buffer, NULL);
+        return NULL;
+    }
+
+    /* Initialize the descriptor. */
+    if (!r300_texture_desc_init(rscreen, &tex->desc, base,
+                                microtile, macrotile,
+                                stride_in_bytes_override,
+                                max_buffer_size)) {
+        if (buffer)
+            rws->buffer_reference(rws, &buffer, NULL);
+        FREE(tex);
         return NULL;
     }
+    /* Initialize the hardware state. */
+    r300_texture_setup_immutable_state(rscreen, tex);
+    r300_texture_setup_fb_state(rscreen, tex);
+
+    tex->desc.b.vtbl = &r300_texture_vtbl;
+    pipe_reference_init(&tex->desc.b.b.reference, 1);
+    tex->domain = base->flags & R300_RESOURCE_FLAG_TRANSFER ?
+                  R300_DOMAIN_GTT :
+                  R300_DOMAIN_VRAM | R300_DOMAIN_GTT;
+    tex->buffer = buffer;
+
+    /* Create the backing buffer if needed. */
+    if (!tex->buffer) {
+        tex->buffer = rws->buffer_create(rws, tex->desc.size_in_bytes, 2048,
+                                         base->bind, base->usage, tex->domain);
+
+        if (!tex->buffer) {
+            FREE(tex);
+            return NULL;
+        }
+    }
+
+    rws->buffer_set_tiling(rws, tex->buffer,
+            tex->desc.microtile, tex->desc.macrotile[0],
+            tex->desc.stride_in_bytes[0]);
+
+    return tex;
+}
+
+/* Create a new texture. */
+struct pipe_resource *r300_texture_create(struct pipe_screen *screen,
+                                          const struct pipe_resource *base)
+{
+    struct r300_screen *rscreen = r300_screen(screen);
+    enum r300_buffer_tiling microtile, macrotile;
 
     /* Refuse to create a texture with size 0. */
     if (!base->width0 ||
@@ -974,58 +759,70 @@ struct pipe_resource* r300_texture_create(struct pipe_screen* screen,
         fprintf(stderr, "r300: texture_create: "
                 "Got invalid texture dimensions: %ix%ix%i\n",
                 base->width0, base->height0, base->depth0);
-        FREE(tex);
         return NULL;
     }
 
-    tex->b.b = *base;
-    tex->b.vtbl = &r300_texture_vtbl;
-    pipe_reference_init(&tex->b.b.reference, 1);
-    tex->b.b.screen = screen;
+    if ((base->flags & R300_RESOURCE_FLAG_TRANSFER) ||
+        (base->bind & PIPE_BIND_SCANOUT)) {
+        microtile = R300_BUFFER_LINEAR;
+        macrotile = R300_BUFFER_LINEAR;
+    } else {
+        microtile = R300_BUFFER_SELECT_LAYOUT;
+        macrotile = R300_BUFFER_SELECT_LAYOUT;
+    }
+
+    return (struct pipe_resource*)
+           r300_texture_create_object(rscreen, base, microtile, macrotile,
+                                      0, 0, NULL);
+}
+
+struct pipe_resource *r300_texture_from_handle(struct pipe_screen *screen,
+                                               const struct pipe_resource *base,
+                                               struct winsys_handle *whandle)
+{
+    struct r300_winsys_screen *rws = (struct r300_winsys_screen*)screen->winsys;
+    struct r300_screen *rscreen = r300_screen(screen);
+    struct r300_winsys_buffer *buffer;
+    enum r300_buffer_tiling microtile, macrotile;
+    unsigned stride, size;
 
-    r300_setup_flags(tex);
-    if (!(base->flags & R300_RESOURCE_FLAG_TRANSFER) &&
-        !(base->bind & PIPE_BIND_SCANOUT)) {
-        r300_setup_tiling(screen, tex);
+    /* Support only 2D textures without mipmaps */
+    if (base->target != PIPE_TEXTURE_2D ||
+        base->depth0 != 1 ||
+        base->last_level != 0) {
+        return NULL;
     }
-    r300_setup_miptree(rscreen, tex);
-    r300_texture_3d_fix_mipmapping(rscreen, tex);
-    r300_texture_setup_immutable_state(rscreen, tex);
-    r300_texture_setup_fb_state(rscreen, tex);
 
-    SCREEN_DBG(rscreen, DBG_TEX,
-               "r300: texture_create: Macro: %s, Micro: %s, Pitch: %i, "
-               "Dim: %ix%ix%i, LastLevel: %i, Size: %i, Format: %s\n",
-               tex->macrotile ? "YES" : " NO",
-               tex->microtile ? "YES" : " NO",
-               tex->hwpitch[0],
-               base->width0, base->height0, base->depth0, base->last_level,
-               tex->size,
-               util_format_short_name(base->format));
+    buffer = rws->buffer_from_handle(rws, whandle, &stride, &size);
+    if (!buffer)
+        return NULL;
 
-    tex->domain = base->flags & R300_RESOURCE_FLAG_TRANSFER ?
-                  R300_DOMAIN_GTT :
-                  R300_DOMAIN_VRAM | R300_DOMAIN_GTT;
+    rws->buffer_get_tiling(rws, buffer, &microtile, &macrotile);
 
-    tex->buffer = rws->buffer_create(rws, tex->size, 2048, base->bind,
-                                     base->usage, tex->domain);
+    /* Enforce a microtiled zbuffer. */
+    if (util_format_is_depth_or_stencil(base->format) &&
+        microtile == R300_BUFFER_LINEAR) {
+        switch (util_format_get_blocksize(base->format)) {
+            case 4:
+                microtile = R300_BUFFER_TILED;
+                break;
 
-    if (!tex->buffer) {
-	FREE(tex);
-	return NULL;
+            case 2:
+                if (rws->get_value(rws, R300_VID_SQUARE_TILING_SUPPORT))
+                    microtile = R300_BUFFER_SQUARETILED;
+                break;
+        }
     }
 
-    rws->buffer_set_tiling(rws, tex->buffer,
-            tex->microtile, tex->macrotile,
-            tex->pitch[0] * util_format_get_blocksize(tex->b.b.format));
-
-    return (struct pipe_resource*)tex;
+    return (struct pipe_resource*)
+           r300_texture_create_object(rscreen, base, microtile, macrotile,
+                                      stride, size, buffer);
 }
 
 /* Not required to implement u_resource_vtbl, consider moving to another file:
  */
 struct pipe_surface* r300_get_tex_surface(struct pipe_screen* screen,
-					  struct pipe_resource* texture,
+                                          struct pipe_resource* texture,
 					  unsigned face,
 					  unsigned level,
 					  unsigned zslice,
@@ -1035,7 +832,7 @@ struct pipe_surface* r300_get_tex_surface(struct pipe_screen* screen,
     struct r300_surface* surface = CALLOC_STRUCT(r300_surface);
 
     if (surface) {
-        uint32_t stride, offset, tile_height;
+        uint32_t offset, tile_height;
 
         pipe_reference_init(&surface->base.reference, 1);
         pipe_resource_reference(&surface->base.texture, texture);
@@ -1054,23 +851,29 @@ struct pipe_surface* r300_get_tex_surface(struct pipe_screen* screen,
         if (surface->domain & R300_DOMAIN_VRAM)
             surface->domain &= ~R300_DOMAIN_GTT;
 
-        surface->offset = r300_texture_get_offset(tex, level, zslice, face);
+        surface->offset = r300_texture_get_offset(&tex->desc,
+                                                  level, zslice, face);
         surface->pitch = tex->fb_state.pitch[level];
         surface->format = tex->fb_state.format;
 
         /* Parameters for the CBZB clear. */
+        surface->cbzb_allowed = tex->desc.cbzb_allowed[level];
         surface->cbzb_width = align(surface->base.width, 64);
 
         /* Height must be aligned to the size of a tile. */
-        tile_height = r300_get_pixel_alignment(tex, tex->mip_macrotile[level],
+        tile_height = r300_get_pixel_alignment(tex->desc.b.b.format,
+                                               tex->desc.b.b.nr_samples,
+                                               tex->desc.microtile,
+                                               tex->desc.macrotile[level],
                                                DIM_HEIGHT);
+
         surface->cbzb_height = align((surface->base.height + 1) / 2,
                                      tile_height);
 
         /* Offset must be aligned to 2K and must point at the beginning
          * of a scanline. */
-        stride = r300_texture_get_stride(r300_screen(screen), tex, level);
-        offset = surface->offset + stride * surface->cbzb_height;
+        offset = surface->offset +
+                 tex->desc.stride_in_bytes[level] * surface->cbzb_height;
         surface->cbzb_midpoint_offset = offset & ~2047;
 
         surface->cbzb_pitch = surface->pitch & 0x1ffffc;
@@ -1080,11 +883,11 @@ struct pipe_surface* r300_get_tex_surface(struct pipe_screen* screen,
         else
             surface->cbzb_format = R300_DEPTHFORMAT_16BIT_INT_Z;
 
-        SCREEN_DBG(r300_screen(screen), DBG_TEX,
+        SCREEN_DBG(r300_screen(screen), DBG_CBZB,
                    "CBZB Dim: %ix%i, Misalignment: %i, Macro: %s\n",
                    surface->cbzb_width, surface->cbzb_height,
                    offset & 2047,
-                   tex->mip_macrotile[level] ? "YES" : " NO");
+                   tex->desc.macrotile[level] ? "YES" : " NO");
     }
 
     return &surface->base;
@@ -1097,88 +900,3 @@ void r300_tex_surface_destroy(struct pipe_surface* s)
     pipe_resource_reference(&s->texture, NULL);
     FREE(s);
 }
-
-struct pipe_resource*
-r300_texture_from_handle(struct pipe_screen* screen,
-			  const struct pipe_resource* base,
-			  struct winsys_handle *whandle)
-{
-    struct r300_winsys_screen *rws = (struct r300_winsys_screen*)screen->winsys;
-    struct r300_screen* rscreen = r300_screen(screen);
-    struct r300_winsys_buffer *buffer;
-    struct r300_texture* tex;
-    unsigned stride;
-    boolean override_zb_flags;
-
-    /* Support only 2D textures without mipmaps */
-    if (base->target != PIPE_TEXTURE_2D ||
-        base->depth0 != 1 ||
-        base->last_level != 0) {
-        return NULL;
-    }
-
-    buffer = rws->buffer_from_handle(rws, whandle, &stride);
-    if (!buffer) {
-        return NULL;
-    }
-
-    tex = CALLOC_STRUCT(r300_texture);
-    if (!tex) {
-        return NULL;
-    }
-
-    tex->b.b = *base;
-    tex->b.vtbl = &r300_texture_vtbl;
-    pipe_reference_init(&tex->b.b.reference, 1);
-    tex->b.b.screen = screen;
-    tex->domain = R300_DOMAIN_VRAM;
-
-    tex->stride_override = stride;
-
-    /* one ref already taken */
-    tex->buffer = buffer;
-
-    rws->buffer_get_tiling(rws, buffer, &tex->microtile, &tex->macrotile);
-    r300_setup_flags(tex);
-    SCREEN_DBG(rscreen, DBG_TEX,
-               "r300: texture_from_handle: Macro: %s, Micro: %s, "
-               "Pitch: % 4i, Dim: %ix%i, Format: %s\n",
-               tex->macrotile ? "YES" : " NO",
-               tex->microtile ? "YES" : " NO",
-               stride / util_format_get_blocksize(base->format),
-               base->width0, base->height0,
-               util_format_short_name(base->format));
-
-    /* Enforce microtiled zbuffer. */
-    override_zb_flags = util_format_is_depth_or_stencil(base->format) &&
-                        tex->microtile == R300_BUFFER_LINEAR;
-
-    if (override_zb_flags) {
-        switch (util_format_get_blocksize(base->format)) {
-            case 4:
-                tex->microtile = R300_BUFFER_TILED;
-                break;
-
-            case 2:
-                if (rws->get_value(rws, R300_VID_SQUARE_TILING_SUPPORT)) {
-                    tex->microtile = R300_BUFFER_SQUARETILED;
-                    break;
-                }
-                /* Pass through. */
-
-            default:
-                override_zb_flags = FALSE;
-        }
-    }
-
-    r300_setup_miptree(rscreen, tex);
-    r300_texture_setup_immutable_state(rscreen, tex);
-    r300_texture_setup_fb_state(rscreen, tex);
-
-    if (override_zb_flags) {
-        rws->buffer_set_tiling(rws, tex->buffer,
-                tex->microtile, tex->macrotile,
-                tex->pitch[0] * util_format_get_blocksize(tex->b.b.format));
-    }
-    return (struct pipe_resource*)tex;
-}
diff --git a/src/gallium/drivers/r300/r300_texture.h b/src/gallium/drivers/r300/r300_texture.h
index 99e7694254e..a4524320fda 100644
--- a/src/gallium/drivers/r300/r300_texture.h
+++ b/src/gallium/drivers/r300/r300_texture.h
@@ -35,16 +35,11 @@ unsigned r300_get_swizzle_combined(const unsigned char *swizzle_format,
                                    const unsigned char *swizzle_view);
 
 uint32_t r300_translate_texformat(enum pipe_format format,
-                                  const unsigned char *swizzle_view);
+                                  const unsigned char *swizzle_view,
+                                  boolean is_r500);
 
 uint32_t r500_tx_format_msb_bit(enum pipe_format format);
 
-unsigned r300_texture_get_stride(struct r300_screen* screen,
-                                 struct r300_texture* tex, unsigned level);
-
-unsigned r300_texture_get_offset(struct r300_texture* tex, unsigned level,
-                                 unsigned zslice, unsigned face);
-
 void r300_texture_reinterpret_format(struct pipe_screen *screen,
                                      struct pipe_resource *tex,
                                      enum pipe_format new_format);
diff --git a/src/gallium/drivers/r300/r300_texture_desc.c b/src/gallium/drivers/r300/r300_texture_desc.c
new file mode 100644
index 00000000000..5d690e8c332
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_texture_desc.c
@@ -0,0 +1,479 @@
+/*
+ * Copyright 2008 Corbin Simpson <[email protected]>
+ * Copyright 2010 Marek Olšák <[email protected]>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#include "r300_texture_desc.h"
+
+#include "r300_context.h"
+#include "r300_winsys.h"
+
+#include "util/u_format.h"
+
+/* Returns the number of pixels that the texture should be aligned to
+ * in the given dimension. */
+unsigned r300_get_pixel_alignment(enum pipe_format format,
+                                  unsigned num_samples,
+                                  enum r300_buffer_tiling microtile,
+                                  enum r300_buffer_tiling macrotile,
+                                  enum r300_dim dim)
+{
+    static const unsigned table[2][5][3][2] =
+    {
+        {
+    /* Macro: linear    linear    linear
+       Micro: linear    tiled  square-tiled */
+            {{ 32, 1}, { 8,  4}, { 0,  0}}, /*   8 bits per pixel */
+            {{ 16, 1}, { 8,  2}, { 4,  4}}, /*  16 bits per pixel */
+            {{  8, 1}, { 4,  2}, { 0,  0}}, /*  32 bits per pixel */
+            {{  4, 1}, { 0,  0}, { 2,  2}}, /*  64 bits per pixel */
+            {{  2, 1}, { 0,  0}, { 0,  0}}  /* 128 bits per pixel */
+        },
+        {
+    /* Macro: tiled     tiled     tiled
+       Micro: linear    tiled  square-tiled */
+            {{256, 8}, {64, 32}, { 0,  0}}, /*   8 bits per pixel */
+            {{128, 8}, {64, 16}, {32, 32}}, /*  16 bits per pixel */
+            {{ 64, 8}, {32, 16}, { 0,  0}}, /*  32 bits per pixel */
+            {{ 32, 8}, { 0,  0}, {16, 16}}, /*  64 bits per pixel */
+            {{ 16, 8}, { 0,  0}, { 0,  0}}  /* 128 bits per pixel */
+        }
+    };
+    static const unsigned aa_block[2] = {4, 8};
+    unsigned tile = 0;
+    unsigned pixsize = util_format_get_blocksize(format);
+
+    assert(macrotile <= R300_BUFFER_TILED);
+    assert(microtile <= R300_BUFFER_SQUARETILED);
+    assert(pixsize <= 16);
+    assert(dim <= DIM_HEIGHT);
+
+    if (num_samples > 1) {
+        /* Multisampled textures have their own alignment scheme. */
+        if (pixsize == 4)
+            tile = aa_block[dim];
+        /* XXX FP16 AA. */
+    } else {
+        /* Standard alignment. */
+        tile = table[macrotile][util_logbase2(pixsize)][microtile][dim];
+    }
+
+    assert(tile);
+    return tile;
+}
+
+/* Return true if macrotiling should be enabled on the miplevel. */
+static boolean r300_texture_macro_switch(struct r300_texture_desc *desc,
+                                         unsigned level,
+                                         boolean rv350_mode,
+                                         enum r300_dim dim)
+{
+    unsigned tile, texdim;
+
+    tile = r300_get_pixel_alignment(desc->b.b.format, desc->b.b.nr_samples,
+                                    desc->microtile, R300_BUFFER_TILED, dim);
+    if (dim == DIM_WIDTH) {
+        texdim = u_minify(desc->b.b.width0, level);
+    } else {
+        texdim = u_minify(desc->b.b.height0, level);
+    }
+
+    /* See TX_FILTER1_n.MACRO_SWITCH. */
+    if (rv350_mode) {
+        return texdim >= tile;
+    } else {
+        return texdim > tile;
+    }
+}
+
+/**
+ * Return the stride, in bytes, of the texture image of the given texture
+ * at the given level.
+ */
+static unsigned r300_texture_get_stride(struct r300_screen *screen,
+                                        struct r300_texture_desc *desc,
+                                        unsigned level)
+{
+    unsigned tile_width, width, stride;
+
+    if (desc->stride_in_bytes_override)
+        return desc->stride_in_bytes_override;
+
+    /* Check the level. */
+    if (level > desc->b.b.last_level) {
+        SCREEN_DBG(screen, DBG_TEX, "%s: level (%u) > last_level (%u)\n",
+                   __FUNCTION__, level, desc->b.b.last_level);
+        return 0;
+    }
+
+    width = u_minify(desc->b.b.width0, level);
+
+    if (util_format_is_plain(desc->b.b.format)) {
+        tile_width = r300_get_pixel_alignment(desc->b.b.format,
+                                              desc->b.b.nr_samples,
+                                              desc->microtile,
+                                              desc->macrotile[level],
+                                              DIM_WIDTH);
+        width = align(width, tile_width);
+
+        stride = util_format_get_stride(desc->b.b.format, width);
+
+        /* Some IGPs need a minimum stride of 64 bytes, hmm... */
+        if (!desc->macrotile[level] &&
+            (screen->caps.family == CHIP_FAMILY_RS600 ||
+             screen->caps.family == CHIP_FAMILY_RS690 ||
+             screen->caps.family == CHIP_FAMILY_RS740)) {
+            unsigned min_stride;
+
+            if (desc->microtile) {
+                unsigned tile_height =
+                        r300_get_pixel_alignment(desc->b.b.format,
+                                                 desc->b.b.nr_samples,
+                                                 desc->microtile,
+                                                 desc->macrotile[level],
+                                                 DIM_HEIGHT);
+
+                min_stride = 64 / tile_height;
+            } else {
+                min_stride = 64;
+            }
+
+            return stride < min_stride ? min_stride : stride;
+        }
+
+        /* The alignment to 32 bytes is sort of implied by the layout... */
+        return stride;
+    } else {
+        return align(util_format_get_stride(desc->b.b.format, width), 32);
+    }
+}
+
+static unsigned r300_texture_get_nblocksy(struct r300_texture_desc *desc,
+                                          unsigned level,
+                                          boolean *out_aligned_for_cbzb)
+{
+    unsigned height, tile_height;
+
+    height = u_minify(desc->b.b.height0, level);
+
+    if (util_format_is_plain(desc->b.b.format)) {
+        tile_height = r300_get_pixel_alignment(desc->b.b.format,
+                                               desc->b.b.nr_samples,
+                                               desc->microtile,
+                                               desc->macrotile[level],
+                                               DIM_HEIGHT);
+        height = align(height, tile_height);
+
+        /* This is needed for the kernel checker, unfortunately. */
+        if ((desc->b.b.target != PIPE_TEXTURE_1D &&
+             desc->b.b.target != PIPE_TEXTURE_2D) ||
+            desc->b.b.last_level != 0) {
+            height = util_next_power_of_two(height);
+        }
+
+        /* See if the CBZB clear can be used on the buffer,
+         * taking the texture size into account. */
+        if (out_aligned_for_cbzb) {
+            if (desc->macrotile[level]) {
+                /* When clearing, the layer (width*height) is horizontally split
+                 * into two, and the upper and lower halves are cleared by the CB
+                 * and ZB units, respectively. Therefore, the number of macrotiles
+                 * in the Y direction must be even. */
+
+                /* Align the height so that there is an even number of macrotiles.
+                 * Do so for 3 or more macrotiles in the Y direction. */
+                if (level == 0 && desc->b.b.last_level == 0 &&
+                    (desc->b.b.target == PIPE_TEXTURE_1D ||
+                     desc->b.b.target == PIPE_TEXTURE_2D) &&
+                    height >= tile_height * 3) {
+                    height = align(height, tile_height * 2);
+                }
+
+                *out_aligned_for_cbzb = height % (tile_height * 2) == 0;
+            } else {
+                *out_aligned_for_cbzb = FALSE;
+            }
+        }
+    }
+
+    return util_format_get_nblocksy(desc->b.b.format, height);
+}
+
+static void r300_texture_3d_fix_mipmapping(struct r300_screen *screen,
+                                           struct r300_texture_desc *desc)
+{
+    /* The kernels <= 2.6.34-rc4 compute the size of mipmapped 3D textures
+     * incorrectly. This is a workaround to prevent CS from being rejected. */
+
+    unsigned i, size;
+
+    if (!screen->rws->get_value(screen->rws, R300_VID_DRM_2_3_0) &&
+        desc->b.b.target == PIPE_TEXTURE_3D &&
+        desc->b.b.last_level > 0) {
+        size = 0;
+
+        for (i = 0; i <= desc->b.b.last_level; i++) {
+            size += desc->stride_in_bytes[i] *
+                    r300_texture_get_nblocksy(desc, i, FALSE);
+        }
+
+        size *= desc->b.b.depth0;
+        desc->size_in_bytes = size;
+    }
+}
+
+/* Get a width in pixels from a stride in bytes. */
+static unsigned stride_to_width(enum pipe_format format,
+                                unsigned stride_in_bytes)
+{
+    return (stride_in_bytes / util_format_get_blocksize(format)) *
+            util_format_get_blockwidth(format);
+}
+
+static void r300_setup_miptree(struct r300_screen *screen,
+                               struct r300_texture_desc *desc,
+                               boolean align_for_cbzb)
+{
+    struct pipe_resource *base = &desc->b.b;
+    unsigned stride, size, layer_size, nblocksy, i;
+    boolean rv350_mode = screen->caps.is_rv350;
+    boolean aligned_for_cbzb;
+
+    desc->size_in_bytes = 0;
+
+    SCREEN_DBG(screen, DBG_TEXALLOC,
+        "r300: Making miptree for texture, format %s\n",
+        util_format_short_name(base->format));
+
+    for (i = 0; i <= base->last_level; i++) {
+        /* Let's see if this miplevel can be macrotiled. */
+        desc->macrotile[i] =
+            (desc->macrotile[0] == R300_BUFFER_TILED &&
+             r300_texture_macro_switch(desc, i, rv350_mode, DIM_WIDTH) &&
+             r300_texture_macro_switch(desc, i, rv350_mode, DIM_HEIGHT)) ?
+             R300_BUFFER_TILED : R300_BUFFER_LINEAR;
+
+        stride = r300_texture_get_stride(screen, desc, i);
+
+        /* Compute the number of blocks in Y, see if the CBZB clear can be
+         * used on the texture. */
+        aligned_for_cbzb = FALSE;
+        if (align_for_cbzb && desc->cbzb_allowed[i])
+            nblocksy = r300_texture_get_nblocksy(desc, i, &aligned_for_cbzb);
+        else
+            nblocksy = r300_texture_get_nblocksy(desc, i, NULL);
+
+        layer_size = stride * nblocksy;
+
+        if (base->nr_samples) {
+            layer_size *= base->nr_samples;
+        }
+
+        if (base->target == PIPE_TEXTURE_CUBE)
+            size = layer_size * 6;
+        else
+            size = layer_size * u_minify(base->depth0, i);
+
+        desc->offset_in_bytes[i] = desc->size_in_bytes;
+        desc->size_in_bytes = desc->offset_in_bytes[i] + size;
+        desc->layer_size_in_bytes[i] = layer_size;
+        desc->stride_in_bytes[i] = stride;
+        desc->stride_in_pixels[i] = stride_to_width(desc->b.b.format, stride);
+        desc->cbzb_allowed[i] = desc->cbzb_allowed[i] && aligned_for_cbzb;
+
+        SCREEN_DBG(screen, DBG_TEXALLOC, "r300: Texture miptree: Level %d "
+                "(%dx%dx%d px, pitch %d bytes) %d bytes total, macrotiled %s\n",
+                i, u_minify(base->width0, i), u_minify(base->height0, i),
+                u_minify(base->depth0, i), stride, desc->size_in_bytes,
+                desc->macrotile[i] ? "TRUE" : "FALSE");
+    }
+}
+
+static void r300_setup_flags(struct r300_texture_desc *desc)
+{
+    desc->uses_stride_addressing =
+        !util_is_power_of_two(desc->b.b.width0) ||
+        !util_is_power_of_two(desc->b.b.height0) ||
+        (desc->stride_in_bytes_override &&
+         stride_to_width(desc->b.b.format,
+                         desc->stride_in_bytes_override) != desc->b.b.width0);
+
+    desc->is_npot =
+        desc->uses_stride_addressing ||
+        !util_is_power_of_two(desc->b.b.height0);
+}
+
+static void r300_setup_cbzb_flags(struct r300_screen *rscreen,
+                                  struct r300_texture_desc *desc)
+{
+    unsigned i, bpp;
+    boolean first_level_valid;
+
+    bpp = util_format_get_blocksizebits(desc->b.b.format);
+
+    /* 1) The texture must be point-sampled,
+     * 2) The depth must be 16 or 32 bits.
+     * 3) If the midpoint ZB offset is not aligned to 2048, it returns garbage
+     *    with certain texture sizes. Macrotiling ensures the alignment. */
+    first_level_valid = desc->b.b.nr_samples <= 1 &&
+                       (bpp == 16 || bpp == 32) &&
+                       desc->macrotile[0];
+
+    for (i = 0; i <= desc->b.b.last_level; i++)
+        desc->cbzb_allowed[i] = first_level_valid && desc->macrotile[i];
+}
+
+static void r300_setup_tiling(struct r300_screen *screen,
+                              struct r300_texture_desc *desc)
+{
+    struct r300_winsys_screen *rws = screen->rws;
+    enum pipe_format format = desc->b.b.format;
+    boolean rv350_mode = screen->caps.is_rv350;
+    boolean is_zb = util_format_is_depth_or_stencil(format);
+    boolean dbg_no_tiling = SCREEN_DBG_ON(screen, DBG_NO_TILING);
+
+    if (!util_format_is_plain(format)) {
+        return;
+    }
+
+    /* If height == 1, disable microtiling except for zbuffer. */
+    if (!is_zb && (desc->b.b.height0 == 1 || dbg_no_tiling)) {
+        return;
+    }
+
+    /* Set microtiling. */
+    switch (util_format_get_blocksize(format)) {
+        case 1:
+        case 4:
+            desc->microtile = R300_BUFFER_TILED;
+            break;
+
+        case 2:
+        case 8:
+            if (rws->get_value(rws, R300_VID_SQUARE_TILING_SUPPORT)) {
+                desc->microtile = R300_BUFFER_SQUARETILED;
+            }
+            break;
+    }
+
+    if (dbg_no_tiling) {
+        return;
+    }
+
+    /* Set macrotiling. */
+    if (r300_texture_macro_switch(desc, 0, rv350_mode, DIM_WIDTH) &&
+        r300_texture_macro_switch(desc, 0, rv350_mode, DIM_HEIGHT)) {
+        desc->macrotile[0] = R300_BUFFER_TILED;
+    }
+}
+
+static void r300_tex_print_info(struct r300_screen *rscreen,
+                                struct r300_texture_desc *desc,
+                                const char *func)
+{
+    fprintf(stderr,
+            "r300: %s: Macro: %s, Micro: %s, Pitch: %i, Dim: %ix%ix%i, "
+            "LastLevel: %i, Size: %i, Format: %s\n",
+            func,
+            desc->macrotile[0] ? "YES" : " NO",
+            desc->microtile ? "YES" : " NO",
+            desc->stride_in_pixels[0],
+            desc->b.b.width0, desc->b.b.height0, desc->b.b.depth0,
+            desc->b.b.last_level, desc->size_in_bytes,
+            util_format_short_name(desc->b.b.format));
+}
+
+boolean r300_texture_desc_init(struct r300_screen *rscreen,
+                               struct r300_texture_desc *desc,
+                               const struct pipe_resource *base,
+                               enum r300_buffer_tiling microtile,
+                               enum r300_buffer_tiling macrotile,
+                               unsigned stride_in_bytes_override,
+                               unsigned max_buffer_size)
+{
+    desc->b.b = *base;
+    desc->b.b.screen = &rscreen->screen;
+
+    desc->stride_in_bytes_override = stride_in_bytes_override;
+
+    if (microtile == R300_BUFFER_SELECT_LAYOUT ||
+        macrotile == R300_BUFFER_SELECT_LAYOUT) {
+        r300_setup_tiling(rscreen, desc);
+    } else {
+        desc->microtile = microtile;
+        desc->macrotile[0] = macrotile;
+        assert(desc->b.b.last_level == 0);
+    }
+
+    r300_setup_flags(desc);
+    r300_setup_cbzb_flags(rscreen, desc);
+
+    /* Setup the miptree description. */
+    r300_setup_miptree(rscreen, desc, TRUE);
+    /* If the required buffer size is larger the given max size,
+     * try again without the alignment for the CBZB clear. */
+    if (max_buffer_size && desc->size_in_bytes > max_buffer_size) {
+        r300_setup_miptree(rscreen, desc, FALSE);
+    }
+
+    r300_texture_3d_fix_mipmapping(rscreen, desc);
+
+    if (max_buffer_size) {
+        /* Make sure the buffer we got is large enough. */
+        if (desc->size_in_bytes > max_buffer_size) {
+            fprintf(stderr, "r300: texture_from_handle: The buffer is not "
+                            "large enough. Got: %i, Need: %i, Info:\n",
+                            max_buffer_size, desc->size_in_bytes);
+            r300_tex_print_info(rscreen, desc, "texture_from_handle");
+            return FALSE;
+        }
+
+        desc->buffer_size_in_bytes = max_buffer_size;
+    } else {
+        desc->buffer_size_in_bytes = desc->size_in_bytes;
+    }
+
+    if (SCREEN_DBG_ON(rscreen, DBG_TEX))
+        r300_tex_print_info(rscreen, desc, "texture_from_handle");
+
+    return TRUE;
+}
+
+unsigned r300_texture_get_offset(struct r300_texture_desc *desc,
+                                 unsigned level, unsigned zslice,
+                                 unsigned face)
+{
+    unsigned offset = desc->offset_in_bytes[level];
+
+    switch (desc->b.b.target) {
+        case PIPE_TEXTURE_3D:
+            assert(face == 0);
+            return offset + zslice * desc->layer_size_in_bytes[level];
+
+        case PIPE_TEXTURE_CUBE:
+            assert(zslice == 0);
+            return offset + face * desc->layer_size_in_bytes[level];
+
+        default:
+            assert(zslice == 0 && face == 0);
+            return offset;
+    }
+}
diff --git a/src/gallium/drivers/r300/r300_texture_desc.h b/src/gallium/drivers/r300/r300_texture_desc.h
new file mode 100644
index 00000000000..95de66f6549
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_texture_desc.h
@@ -0,0 +1,57 @@
+/*
+ * Copyright 2008 Corbin Simpson <[email protected]>
+ * Copyright 2010 Marek Olšák <[email protected]>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#ifndef R300_TEXTURE_DESC_H
+#define R300_TEXTURE_DESC_H
+
+#include "r300_defines.h"
+
+struct pipe_resource;
+struct r300_screen;
+struct r300_texture_desc;
+struct r300_texture;
+
+enum r300_dim {
+    DIM_WIDTH  = 0,
+    DIM_HEIGHT = 1
+};
+
+unsigned r300_get_pixel_alignment(enum pipe_format format,
+                                  unsigned num_samples,
+                                  enum r300_buffer_tiling microtile,
+                                  enum r300_buffer_tiling macrotile,
+                                  enum r300_dim dim);
+
+boolean r300_texture_desc_init(struct r300_screen *rscreen,
+                               struct r300_texture_desc *desc,
+                               const struct pipe_resource *base,
+                               enum r300_buffer_tiling microtile,
+                               enum r300_buffer_tiling macrotile,
+                               unsigned stride_in_bytes_override,
+                               unsigned max_buffer_size);
+
+unsigned r300_texture_get_offset(struct r300_texture_desc *desc,
+                                 unsigned level, unsigned zslice,
+                                 unsigned face);
+
+#endif
diff --git a/src/gallium/drivers/r300/r300_tgsi_to_rc.c b/src/gallium/drivers/r300/r300_tgsi_to_rc.c
index 51b2c555502..dd697b9c374 100644
--- a/src/gallium/drivers/r300/r300_tgsi_to_rc.c
+++ b/src/gallium/drivers/r300/r300_tgsi_to_rc.c
@@ -126,7 +126,7 @@ static unsigned translate_opcode(unsigned opcode)
      /* case TGSI_OPCODE_SAD: return RC_OPCODE_SAD; */
      /* case TGSI_OPCODE_TXF: return RC_OPCODE_TXF; */
      /* case TGSI_OPCODE_TXQ: return RC_OPCODE_TXQ; */
-     /* case TGSI_OPCODE_CONT: return RC_OPCODE_CONT; */
+        case TGSI_OPCODE_CONT: return RC_OPCODE_CONT;
      /* case TGSI_OPCODE_EMIT: return RC_OPCODE_EMIT; */
      /* case TGSI_OPCODE_ENDPRIM: return RC_OPCODE_ENDPRIM; */
      /* case TGSI_OPCODE_BGNLOOP2: return RC_OPCODE_BGNLOOP2; */
diff --git a/src/gallium/drivers/r300/r300_transfer.c b/src/gallium/drivers/r300/r300_transfer.c
index 3cc4c8c9582..e9333b35ef5 100644
--- a/src/gallium/drivers/r300/r300_transfer.c
+++ b/src/gallium/drivers/r300/r300_transfer.c
@@ -22,7 +22,7 @@
  * USE OR OTHER DEALINGS IN THE SOFTWARE. */
 
 #include "r300_transfer.h"
-#include "r300_texture.h"
+#include "r300_texture_desc.h"
 #include "r300_screen_buffer.h"
 
 #include "util/u_memory.h"
@@ -35,8 +35,8 @@ struct r300_transfer {
     /* Offset from start of buffer. */
     unsigned offset;
 
-    /* Detiled texture. */
-    struct r300_texture *detiled_texture;
+    /* Linear texture. */
+    struct r300_texture *linear_texture;
 };
 
 /* Convenience cast wrapper. */
@@ -57,7 +57,7 @@ static void r300_copy_from_tiled_texture(struct pipe_context *ctx,
     subdst.face = 0;
     subdst.level = 0;
 
-    ctx->resource_copy_region(ctx, &r300transfer->detiled_texture->b.b, subdst,
+    ctx->resource_copy_region(ctx, &r300transfer->linear_texture->desc.b.b, subdst,
 			      0, 0, 0,
 			      tex, transfer->sr,
 			      transfer->box.x, transfer->box.y, transfer->box.z,
@@ -77,7 +77,7 @@ static void r300_copy_into_tiled_texture(struct pipe_context *ctx,
 
     ctx->resource_copy_region(ctx, tex, transfer->sr,
 			      transfer->box.x, transfer->box.y, transfer->box.z,
-			      &r300transfer->detiled_texture->b.b, subsrc,
+                              &r300transfer->linear_texture->desc.b.b, subsrc,
 			      0, 0, 0,
 			      transfer->box.width, transfer->box.height);
 
@@ -93,7 +93,6 @@ r300_texture_get_transfer(struct pipe_context *ctx,
 {
     struct r300_context *r300 = r300_context(ctx);
     struct r300_texture *tex = r300_texture(texture);
-    struct r300_screen *r300screen = r300_screen(ctx->screen);
     struct r300_transfer *trans;
     struct pipe_resource base;
     boolean referenced_cs, referenced_hw, blittable;
@@ -124,7 +123,7 @@ r300_texture_get_transfer(struct pipe_context *ctx,
         /* If the texture is tiled, we must create a temporary detiled texture
          * for this transfer.
          * Also make write transfers pipelined. */
-        if (tex->microtile || tex->macrotile ||
+        if (tex->desc.microtile || tex->desc.macrotile[sr.level] ||
             ((referenced_hw & !(usage & PIPE_TRANSFER_READ)) && blittable)) {
             base.target = PIPE_TEXTURE_2D;
             base.format = texture->format;
@@ -149,23 +148,23 @@ r300_texture_get_transfer(struct pipe_context *ctx,
             }
 
             /* Create the temporary texture. */
-            trans->detiled_texture = r300_texture(
+            trans->linear_texture = r300_texture(
                ctx->screen->resource_create(ctx->screen,
                                             &base));
 
-            if (!trans->detiled_texture) {
+            if (!trans->linear_texture) {
                 /* Oh crap, the thing can't create the texture.
                  * Let's flush and try again. */
                 ctx->flush(ctx, 0, NULL);
 
-                trans->detiled_texture = r300_texture(
+                trans->linear_texture = r300_texture(
                    ctx->screen->resource_create(ctx->screen,
                                                 &base));
 
-                if (!trans->detiled_texture) {
+                if (!trans->linear_texture) {
                     /* For linear textures, it's safe to fallback to
                      * an unpipelined transfer. */
-                    if (!tex->microtile && !tex->macrotile) {
+                    if (!tex->desc.microtile && !tex->desc.macrotile[sr.level]) {
                         goto unpipelined;
                     }
 
@@ -177,8 +176,8 @@ r300_texture_get_transfer(struct pipe_context *ctx,
                 }
             }
 
-            assert(!trans->detiled_texture->microtile &&
-                   !trans->detiled_texture->macrotile);
+            assert(!trans->linear_texture->desc.microtile &&
+                   !trans->linear_texture->desc.macrotile[0]);
 
             /* Set the stride.
 	     *
@@ -188,7 +187,7 @@ r300_texture_get_transfer(struct pipe_context *ctx,
 	     * right thing internally.
 	     */
             trans->transfer.stride =
-                r300_texture_get_stride(r300screen, trans->detiled_texture, 0);
+                    trans->linear_texture->desc.stride_in_bytes[0];
 
             if (usage & PIPE_TRANSFER_READ) {
                 /* We cannot map a tiled texture directly because the data is
@@ -203,9 +202,9 @@ r300_texture_get_transfer(struct pipe_context *ctx,
 
     unpipelined:
         /* Unpipelined transfer. */
-        trans->transfer.stride =
-                r300_texture_get_stride(r300screen, tex, sr.level);
-        trans->offset = r300_texture_get_offset(tex, sr.level, box->z, sr.face);
+        trans->transfer.stride = tex->desc.stride_in_bytes[sr.level];
+        trans->offset = r300_texture_get_offset(&tex->desc,
+                                                sr.level, box->z, sr.face);
 
         if (referenced_cs)
             ctx->flush(ctx, PIPE_FLUSH_RENDER_CACHE, NULL);
@@ -219,13 +218,13 @@ void r300_texture_transfer_destroy(struct pipe_context *ctx,
 {
     struct r300_transfer *r300transfer = r300_transfer(trans);
 
-    if (r300transfer->detiled_texture) {
+    if (r300transfer->linear_texture) {
         if (trans->usage & PIPE_TRANSFER_WRITE) {
             r300_copy_into_tiled_texture(ctx, r300transfer);
         }
 
         pipe_resource_reference(
-            (struct pipe_resource**)&r300transfer->detiled_texture, NULL);
+            (struct pipe_resource**)&r300transfer->linear_texture, NULL);
     }
     pipe_resource_reference(&trans->resource, NULL);
     FREE(trans);
@@ -239,13 +238,13 @@ void* r300_texture_transfer_map(struct pipe_context *ctx,
     struct r300_transfer *r300transfer = r300_transfer(transfer);
     struct r300_texture *tex = r300_texture(transfer->resource);
     char *map;
-    enum pipe_format format = tex->b.b.format;
+    enum pipe_format format = tex->desc.b.b.format;
 
-    if (r300transfer->detiled_texture) {
+    if (r300transfer->linear_texture) {
         /* The detiled texture is of the same size as the region being mapped
          * (no offset needed). */
         return rws->buffer_map(rws,
-                               r300transfer->detiled_texture->buffer,
+                               r300transfer->linear_texture->buffer,
                                r300->cs,
                                transfer->usage);
     } else {
@@ -270,8 +269,8 @@ void r300_texture_transfer_unmap(struct pipe_context *ctx,
     struct r300_transfer *r300transfer = r300_transfer(transfer);
     struct r300_texture *tex = r300_texture(transfer->resource);
 
-    if (r300transfer->detiled_texture) {
-	rws->buffer_unmap(rws, r300transfer->detiled_texture->buffer);
+    if (r300transfer->linear_texture) {
+        rws->buffer_unmap(rws, r300transfer->linear_texture->buffer);
     } else {
         rws->buffer_unmap(rws, tex->buffer);
     }
diff --git a/src/gallium/drivers/r300/r300_vs.c b/src/gallium/drivers/r300/r300_vs.c
index b25c786d6b3..54c8de12419 100644
--- a/src/gallium/drivers/r300/r300_vs.c
+++ b/src/gallium/drivers/r300/r300_vs.c
@@ -207,7 +207,7 @@ void r300_translate_vertex_shader(struct r300_context *r300,
     compiler.Base.max_temp_regs = 32;
 
     if (compiler.Base.Debug) {
-        debug_printf("r300: Initial vertex program\n");
+        DBG(r300, DBG_VP, "r300: Initial vertex program\n");
         tgsi_dump(vs->state.tokens, 0);
     }
 
@@ -227,8 +227,7 @@ void r300_translate_vertex_shader(struct r300_context *r300,
     /* Invoke the compiler */
     r3xx_compile_vertex_program(&compiler);
     if (compiler.Base.Error) {
-        /* XXX We should fallback using Draw. */
-        fprintf(stderr, "r300 VP: Compiler error:\n%sUsing a dummy shader"
+        DBG(r300, DBG_VP, "r300 VP: Compiler error:\n%sUsing a dummy shader"
                 " instead.\nIf there's an 'unknown opcode' message, please"
                 " file a bug report and attach this log.\n", compiler.Base.ErrorMsg);
 
diff --git a/src/gallium/drivers/r300/r300_winsys.h b/src/gallium/drivers/r300/r300_winsys.h
index 7e115c2d624..187780750fa 100644
--- a/src/gallium/drivers/r300/r300_winsys.h
+++ b/src/gallium/drivers/r300/r300_winsys.h
@@ -49,6 +49,8 @@ enum r300_value_id {
     R300_VID_Z_PIPES,
     R300_VID_SQUARE_TILING_SUPPORT,
     R300_VID_DRM_2_3_0,
+    R300_VID_DRM_2_6_0,
+    R300_CAN_HYPERZ,
 };
 
 enum r300_reference_domain { /* bitfield */
@@ -184,12 +186,13 @@ struct r300_winsys_screen {
      * \param ws        The winsys this function is called from.
      * \param whandle   A winsys handle pointer as was received from a state
      *                  tracker.
-     * \param stride    A pointer to the stride return variable.
-     *                  The stride is in bytes.
+     * \param stride    The returned buffer stride in bytes.
+     * \param size      The returned buffer size.
      */
     struct r300_winsys_buffer *(*buffer_from_handle)(struct r300_winsys_screen *ws,
                                                      struct winsys_handle *whandle,
-                                                     unsigned *stride);
+                                                     unsigned *stride,
+                                                     unsigned *size);
 
     /**
      * Get a winsys handle from a winsys buffer. The internal structure
diff --git a/src/gallium/drivers/r600/Makefile b/src/gallium/drivers/r600/Makefile
index aae31a6a6eb..fc94ae71f4a 100644
--- a/src/gallium/drivers/r600/Makefile
+++ b/src/gallium/drivers/r600/Makefile
@@ -9,6 +9,7 @@ LIBRARY_INCLUDES = \
 C_SOURCES = \
 	r600_buffer.c \
 	r600_context.c \
+	r600_shader.c \
 	r600_draw.c \
 	r600_blit.c \
 	r600_helper.c \
@@ -17,11 +18,7 @@ C_SOURCES = \
 	r600_screen.c \
 	r600_state.c \
 	r600_texture.c \
-	r600_shader.c \
-	r600_compiler.c \
-	r600_compiler_tgsi.c \
-	r600_compiler_dump.c \
-	r600_compiler_r600.c \
-	r600_compiler_r700.c
+	r600_asm.c \
+	r700_asm.c
 
 include ../../Makefile.template
diff --git a/src/gallium/drivers/r600/SConscript b/src/gallium/drivers/r600/SConscript
index 26e2f1941cc..99c8644e026 100644
--- a/src/gallium/drivers/r600/SConscript
+++ b/src/gallium/drivers/r600/SConscript
@@ -27,11 +27,8 @@ r600 = env.ConvenienceLibrary(
         'r600_state.c',
         'r600_texture.c',
         'r600_shader.c',
-        'r600_compiler.c',
-        'r600_compiler_tgsi.c',
-        'r600_compiler_dump.c',
-        'r600_compiler_r600.c',
-        'r600_compiler_r700.c'
+        'r600_asm.c',
+        'r700_asm.c',
     ])
 
 Export('r600')
diff --git a/src/gallium/drivers/r600/r600_asm.c b/src/gallium/drivers/r600/r600_asm.c
new file mode 100644
index 00000000000..9ea9d4354d6
--- /dev/null
+++ b/src/gallium/drivers/r600/r600_asm.c
@@ -0,0 +1,476 @@
+/*
+ * Copyright 2010 Jerome Glisse <[email protected]>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+#include "r600_asm.h"
+#include "r600_context.h"
+#include "util/u_memory.h"
+#include "r600_sq.h"
+#include <stdio.h>
+#include <errno.h>
+
+int r700_bc_alu_build(struct r600_bc *bc, struct r600_bc_alu *alu, unsigned id);
+
+static struct r600_bc_cf *r600_bc_cf(void)
+{
+	struct r600_bc_cf *cf = CALLOC_STRUCT(r600_bc_cf);
+
+	if (cf == NULL)
+		return NULL;
+	LIST_INITHEAD(&cf->list);
+	LIST_INITHEAD(&cf->alu);
+	LIST_INITHEAD(&cf->vtx);
+	LIST_INITHEAD(&cf->tex);
+	return cf;
+}
+
+static struct r600_bc_alu *r600_bc_alu(void)
+{
+	struct r600_bc_alu *alu = CALLOC_STRUCT(r600_bc_alu);
+
+	if (alu == NULL)
+		return NULL;
+	LIST_INITHEAD(&alu->list);
+	return alu;
+}
+
+static struct r600_bc_vtx *r600_bc_vtx(void)
+{
+	struct r600_bc_vtx *vtx = CALLOC_STRUCT(r600_bc_vtx);
+
+	if (vtx == NULL)
+		return NULL;
+	LIST_INITHEAD(&vtx->list);
+	return vtx;
+}
+
+static struct r600_bc_tex *r600_bc_tex(void)
+{
+	struct r600_bc_tex *tex = CALLOC_STRUCT(r600_bc_tex);
+
+	if (tex == NULL)
+		return NULL;
+	LIST_INITHEAD(&tex->list);
+	return tex;
+}
+
+int r600_bc_init(struct r600_bc *bc, enum radeon_family family)
+{
+	LIST_INITHEAD(&bc->cf);
+	bc->family = family;
+	return 0;
+}
+
+static int r600_bc_add_cf(struct r600_bc *bc)
+{
+	struct r600_bc_cf *cf = r600_bc_cf();
+
+	if (cf == NULL)
+		return -ENOMEM;
+	LIST_ADDTAIL(&cf->list, &bc->cf);
+	if (bc->cf_last)
+		cf->id = bc->cf_last->id + 2;
+	bc->cf_last = cf;
+	bc->ncf++;
+	bc->ndw += 2;
+	bc->force_add_cf = 0;
+	return 0;
+}
+
+int r600_bc_add_output(struct r600_bc *bc, const struct r600_bc_output *output)
+{
+	int r;
+
+	r = r600_bc_add_cf(bc);
+	if (r)
+		return r;
+	bc->cf_last->inst = output->inst;
+	memcpy(&bc->cf_last->output, output, sizeof(struct r600_bc_output));
+	return 0;
+}
+
+int r600_bc_add_alu(struct r600_bc *bc, const struct r600_bc_alu *alu)
+{
+	struct r600_bc_alu *nalu = r600_bc_alu();
+	struct r600_bc_alu *lalu;
+	int i, r;
+
+	if (nalu == NULL)
+		return -ENOMEM;
+	memcpy(nalu, alu, sizeof(struct r600_bc_alu));
+	nalu->nliteral = 0;
+
+	/* cf can contains only alu or only vtx or only tex */
+	if (bc->cf_last == NULL || bc->cf_last->inst != (V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU << 3) ||
+		bc->force_add_cf) {
+		/* at most 128 slots, one add alu can add 4 slots + 4 constant worst case */
+		r = r600_bc_add_cf(bc);
+		if (r) {
+			free(nalu);
+			return r;
+		}
+		bc->cf_last->inst = V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU << 3;
+	}
+	if (alu->last && (bc->cf_last->ndw >> 1) >= 124) {
+		bc->force_add_cf = 1;
+	}
+	/* number of gpr == the last gpr used in any alu */
+	for (i = 0; i < 3; i++) {
+		if (alu->src[i].sel >= bc->ngpr && alu->src[i].sel < 128) {
+			bc->ngpr = alu->src[i].sel + 1;
+		}
+		/* compute how many literal are needed
+		 * either 2 or 4 literals
+		 */
+		if (alu->src[i].sel == 253) {
+			if (((alu->src[i].chan + 2) & 0x6) > nalu->nliteral) {
+				nalu->nliteral = (alu->src[i].chan + 2) & 0x6;
+			}
+		}
+	}
+	if (!LIST_IS_EMPTY(&bc->cf_last->alu)) {
+		lalu = LIST_ENTRY(struct r600_bc_alu, bc->cf_last->alu.prev, list);
+		if (!lalu->last && lalu->nliteral > nalu->nliteral) {
+			nalu->nliteral = lalu->nliteral;
+		}
+	}
+	if (alu->dst.sel >= bc->ngpr) {
+		bc->ngpr = alu->dst.sel + 1;
+	}
+	LIST_ADDTAIL(&nalu->list, &bc->cf_last->alu);
+	/* each alu use 2 dwords */
+	bc->cf_last->ndw += 2;
+	bc->ndw += 2;
+	return 0;
+}
+
+int r600_bc_add_literal(struct r600_bc *bc, const u32 *value)
+{
+	struct r600_bc_alu *alu;
+
+	if (bc->cf_last == NULL) {
+		return 0;
+	}
+	if (bc->cf_last->inst == V_SQ_CF_WORD1_SQ_CF_INST_TEX) {
+		return 0;
+	}
+	if (bc->cf_last->inst != (V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU << 3) ||
+		LIST_IS_EMPTY(&bc->cf_last->alu)) {
+		R600_ERR("last CF is not ALU (%p)\n", bc->cf_last);
+		return -EINVAL;
+	}
+	alu = LIST_ENTRY(struct r600_bc_alu, bc->cf_last->alu.prev, list);
+	if (!alu->last || !alu->nliteral || alu->literal_added) {
+		return 0;
+	}
+	memcpy(alu->value, value, 4 * 4);
+	bc->cf_last->ndw += alu->nliteral;
+	bc->ndw += alu->nliteral;
+	alu->literal_added = 1;
+	return 0;
+}
+
+int r600_bc_add_vtx(struct r600_bc *bc, const struct r600_bc_vtx *vtx)
+{
+	struct r600_bc_vtx *nvtx = r600_bc_vtx();
+	int r;
+
+	if (nvtx == NULL)
+		return -ENOMEM;
+	memcpy(nvtx, vtx, sizeof(struct r600_bc_vtx));
+
+	/* cf can contains only alu or only vtx or only tex */
+	if (bc->cf_last == NULL ||
+		(bc->cf_last->inst != V_SQ_CF_WORD1_SQ_CF_INST_VTX &&
+		 bc->cf_last->inst != V_SQ_CF_WORD1_SQ_CF_INST_VTX_TC)) {
+		r = r600_bc_add_cf(bc);
+		if (r) {
+			free(nvtx);
+			return r;
+		}
+		bc->cf_last->inst = V_SQ_CF_WORD1_SQ_CF_INST_VTX;
+	}
+	LIST_ADDTAIL(&nvtx->list, &bc->cf_last->vtx);
+	/* each fetch use 4 dwords */
+	bc->cf_last->ndw += 4;
+	bc->ndw += 4;
+	return 0;
+}
+
+int r600_bc_add_tex(struct r600_bc *bc, const struct r600_bc_tex *tex)
+{
+	struct r600_bc_tex *ntex = r600_bc_tex();
+	int r;
+
+	if (ntex == NULL)
+		return -ENOMEM;
+	memcpy(ntex, tex, sizeof(struct r600_bc_tex));
+
+	/* cf can contains only alu or only vtx or only tex */
+	if (bc->cf_last == NULL ||
+		bc->cf_last->inst != V_SQ_CF_WORD1_SQ_CF_INST_TEX) {
+		r = r600_bc_add_cf(bc);
+		if (r) {
+			free(ntex);
+			return r;
+		}
+		bc->cf_last->inst = V_SQ_CF_WORD1_SQ_CF_INST_TEX;
+	}
+	LIST_ADDTAIL(&ntex->list, &bc->cf_last->tex);
+	/* each texture fetch use 4 dwords */
+	bc->cf_last->ndw += 4;
+	bc->ndw += 4;
+	return 0;
+}
+
+static int r600_bc_vtx_build(struct r600_bc *bc, struct r600_bc_vtx *vtx, unsigned id)
+{
+	bc->bytecode[id++] = S_SQ_VTX_WORD0_BUFFER_ID(vtx->buffer_id) |
+				S_SQ_VTX_WORD0_SRC_GPR(vtx->src_gpr) |
+				S_SQ_VTX_WORD0_SRC_SEL_X(vtx->src_sel_x) |
+				S_SQ_VTX_WORD0_MEGA_FETCH_COUNT(vtx->mega_fetch_count);
+	bc->bytecode[id++] = S_SQ_VTX_WORD1_DST_SEL_X(vtx->dst_sel_x) |
+				S_SQ_VTX_WORD1_DST_SEL_Y(vtx->dst_sel_y) |
+				S_SQ_VTX_WORD1_DST_SEL_Z(vtx->dst_sel_z) |
+				S_SQ_VTX_WORD1_DST_SEL_W(vtx->dst_sel_w) |
+				S_SQ_VTX_WORD1_USE_CONST_FIELDS(1) |
+				S_SQ_VTX_WORD1_GPR_DST_GPR(vtx->dst_gpr);
+	bc->bytecode[id++] = S_SQ_VTX_WORD2_MEGA_FETCH(1);
+	bc->bytecode[id++] = 0;
+	return 0;
+}
+
+static int r600_bc_tex_build(struct r600_bc *bc, struct r600_bc_tex *tex, unsigned id)
+{
+	bc->bytecode[id++] = S_SQ_TEX_WORD0_TEX_INST(tex->inst) |
+				S_SQ_TEX_WORD0_RESOURCE_ID(tex->resource_id) |
+				S_SQ_TEX_WORD0_SRC_GPR(tex->src_gpr) |
+				S_SQ_TEX_WORD0_SRC_REL(tex->src_rel);
+	bc->bytecode[id++] = S_SQ_TEX_WORD1_DST_GPR(tex->dst_gpr) |
+				S_SQ_TEX_WORD1_DST_REL(tex->dst_rel) |
+				S_SQ_TEX_WORD1_DST_SEL_X(tex->dst_sel_x) |
+				S_SQ_TEX_WORD1_DST_SEL_Y(tex->dst_sel_y) |
+				S_SQ_TEX_WORD1_DST_SEL_Z(tex->dst_sel_z) |
+				S_SQ_TEX_WORD1_DST_SEL_W(tex->dst_sel_w) |
+				S_SQ_TEX_WORD1_LOD_BIAS(tex->lod_bias) |
+				S_SQ_TEX_WORD1_COORD_TYPE_X(tex->coord_type_x) |
+				S_SQ_TEX_WORD1_COORD_TYPE_Y(tex->coord_type_y) |
+				S_SQ_TEX_WORD1_COORD_TYPE_Z(tex->coord_type_z) |
+				S_SQ_TEX_WORD1_COORD_TYPE_W(tex->coord_type_w);
+	bc->bytecode[id++] = S_SQ_TEX_WORD2_OFFSET_X(tex->offset_x) |
+				S_SQ_TEX_WORD2_OFFSET_Y(tex->offset_y) |
+				S_SQ_TEX_WORD2_OFFSET_Z(tex->offset_z) |
+				S_SQ_TEX_WORD2_SAMPLER_ID(tex->sampler_id) |
+				S_SQ_TEX_WORD2_SRC_SEL_X(tex->src_sel_x) |
+				S_SQ_TEX_WORD2_SRC_SEL_Y(tex->src_sel_y) |
+				S_SQ_TEX_WORD2_SRC_SEL_Z(tex->src_sel_z) |
+				S_SQ_TEX_WORD2_SRC_SEL_W(tex->src_sel_w);
+	bc->bytecode[id++] = 0;
+	return 0;
+}
+
+static int r600_bc_alu_build(struct r600_bc *bc, struct r600_bc_alu *alu, unsigned id)
+{
+	unsigned i;
+
+	/* don't replace gpr by pv or ps for destination register */
+	if (alu->is_op3) {
+		bc->bytecode[id++] = S_SQ_ALU_WORD0_SRC0_SEL(alu->src[0].sel) |
+					S_SQ_ALU_WORD0_SRC0_CHAN(alu->src[0].chan) |
+					S_SQ_ALU_WORD0_SRC1_SEL(alu->src[1].sel) |
+					S_SQ_ALU_WORD0_SRC1_CHAN(alu->src[1].chan) |
+					S_SQ_ALU_WORD0_LAST(alu->last);
+		bc->bytecode[id++] = S_SQ_ALU_WORD1_DST_GPR(alu->dst.sel) |
+					S_SQ_ALU_WORD1_DST_CHAN(alu->dst.chan) |
+					S_SQ_ALU_WORD1_CLAMP(alu->dst.clamp) |
+					S_SQ_ALU_WORD1_OP3_SRC2_SEL(alu->src[2].sel) |
+					S_SQ_ALU_WORD1_OP3_SRC2_CHAN(alu->src[2].chan) |
+					S_SQ_ALU_WORD1_OP3_SRC2_NEG(alu->src[2].neg) |
+					S_SQ_ALU_WORD1_OP3_ALU_INST(alu->inst) |
+					S_SQ_ALU_WORD1_BANK_SWIZZLE(0);
+	} else {
+		bc->bytecode[id++] = S_SQ_ALU_WORD0_SRC0_SEL(alu->src[0].sel) |
+					S_SQ_ALU_WORD0_SRC0_CHAN(alu->src[0].chan) |
+					S_SQ_ALU_WORD0_SRC0_NEG(alu->src[0].neg) |
+					S_SQ_ALU_WORD0_SRC1_SEL(alu->src[1].sel) |
+					S_SQ_ALU_WORD0_SRC1_CHAN(alu->src[1].chan) |
+					S_SQ_ALU_WORD0_SRC1_NEG(alu->src[1].neg) |
+					S_SQ_ALU_WORD0_LAST(alu->last);
+		bc->bytecode[id++] = S_SQ_ALU_WORD1_DST_GPR(alu->dst.sel) |
+					S_SQ_ALU_WORD1_DST_CHAN(alu->dst.chan) |
+					S_SQ_ALU_WORD1_CLAMP(alu->dst.clamp) |
+					S_SQ_ALU_WORD1_OP2_SRC0_ABS(alu->src[0].abs) |
+					S_SQ_ALU_WORD1_OP2_SRC1_ABS(alu->src[1].abs) |
+					S_SQ_ALU_WORD1_OP2_WRITE_MASK(alu->dst.write) |
+					S_SQ_ALU_WORD1_OP2_ALU_INST(alu->inst) |
+					S_SQ_ALU_WORD1_BANK_SWIZZLE(0);
+	}
+	if (alu->last) {
+		for (i = 0; i < alu->nliteral; i++) {
+			bc->bytecode[id++] = alu->value[i];
+		}
+	}
+	return 0;
+}
+
+static int r600_bc_cf_build(struct r600_bc *bc, struct r600_bc_cf *cf)
+{
+	unsigned id = cf->id;
+
+	switch (cf->inst) {
+	case (V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU << 3):
+		bc->bytecode[id++] = S_SQ_CF_ALU_WORD0_ADDR(cf->addr >> 1);
+		bc->bytecode[id++] = S_SQ_CF_ALU_WORD1_CF_INST(cf->inst >> 3) |
+					S_SQ_CF_ALU_WORD1_BARRIER(1) |
+					S_SQ_CF_ALU_WORD1_COUNT((cf->ndw / 2) - 1);
+		break;
+	case V_SQ_CF_WORD1_SQ_CF_INST_TEX:
+	case V_SQ_CF_WORD1_SQ_CF_INST_VTX:
+	case V_SQ_CF_WORD1_SQ_CF_INST_VTX_TC:
+		bc->bytecode[id++] = S_SQ_CF_WORD0_ADDR(cf->addr >> 1);
+		bc->bytecode[id++] = S_SQ_CF_WORD1_CF_INST(cf->inst) |
+					S_SQ_CF_WORD1_BARRIER(1) |
+					S_SQ_CF_WORD1_COUNT((cf->ndw / 4) - 1);
+		break;
+	case V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT:
+	case V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT_DONE:
+		bc->bytecode[id++] = S_SQ_CF_ALLOC_EXPORT_WORD0_RW_GPR(cf->output.gpr) |
+			S_SQ_CF_ALLOC_EXPORT_WORD0_ELEM_SIZE(cf->output.elem_size) |
+			S_SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE(cf->output.array_base) |
+			S_SQ_CF_ALLOC_EXPORT_WORD0_TYPE(cf->output.type);
+		bc->bytecode[id++] = S_SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_X(cf->output.swizzle_x) |
+			S_SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Y(cf->output.swizzle_y) |
+			S_SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Z(cf->output.swizzle_z) |
+			S_SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_W(cf->output.swizzle_w) |
+			S_SQ_CF_ALLOC_EXPORT_WORD1_BARRIER(cf->output.barrier) |
+			S_SQ_CF_ALLOC_EXPORT_WORD1_CF_INST(cf->output.inst) |
+			S_SQ_CF_ALLOC_EXPORT_WORD1_END_OF_PROGRAM(cf->output.end_of_program);
+		break;
+	default:
+		R600_ERR("unsupported CF instruction (0x%X)\n", cf->inst);
+		return -EINVAL;
+	}
+	return 0;
+}
+
+int r600_bc_build(struct r600_bc *bc)
+{
+	struct r600_bc_cf *cf;
+	struct r600_bc_alu *alu;
+	struct r600_bc_vtx *vtx;
+	struct r600_bc_tex *tex;
+	unsigned addr;
+	int r;
+
+
+	/* first path compute addr of each CF block */
+	/* addr start after all the CF instructions */
+	addr = bc->cf_last->id + 2;
+	LIST_FOR_EACH_ENTRY(cf, &bc->cf, list) {
+		switch (cf->inst) {
+		case (V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU << 3):
+			break;
+		case V_SQ_CF_WORD1_SQ_CF_INST_TEX:
+		case V_SQ_CF_WORD1_SQ_CF_INST_VTX:
+		case V_SQ_CF_WORD1_SQ_CF_INST_VTX_TC:
+			/* fetch node need to be 16 bytes aligned*/
+			addr += 3;
+			addr &= 0xFFFFFFFCUL;
+			break;
+		case V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT:
+		case V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT_DONE:
+			break;
+		default:
+			R600_ERR("unsupported CF instruction (0x%X)\n", cf->inst);
+			return -EINVAL;
+		}
+		cf->addr = addr;
+		addr += cf->ndw;
+		bc->ndw = cf->addr + cf->ndw;
+	}
+	free(bc->bytecode);
+	bc->bytecode = calloc(1, bc->ndw * 4);
+	if (bc->bytecode == NULL)
+		return -ENOMEM;
+	LIST_FOR_EACH_ENTRY(cf, &bc->cf, list) {
+		addr = cf->addr;
+		r = r600_bc_cf_build(bc, cf);
+		if (r)
+			return r;
+		switch (cf->inst) {
+		case (V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU << 3):
+			LIST_FOR_EACH_ENTRY(alu, &cf->alu, list) {
+				switch (bc->family) {
+				case CHIP_R600:
+				case CHIP_RV610:
+				case CHIP_RV630:
+				case CHIP_RV670:
+				case CHIP_RV620:
+				case CHIP_RV635:
+				case CHIP_RS780:
+				case CHIP_RS880:
+					r = r600_bc_alu_build(bc, alu, addr);
+					break;
+				case CHIP_RV770:
+				case CHIP_RV730:
+				case CHIP_RV710:
+				case CHIP_RV740:
+					r = r700_bc_alu_build(bc, alu, addr);
+					break;
+				default:
+					R600_ERR("unknown family %d\n", bc->family);
+					return -EINVAL;
+				}
+				if (r)
+					return r;
+				addr += 2;
+				if (alu->last) {
+					addr += alu->nliteral;
+				}
+			}
+			break;
+		case V_SQ_CF_WORD1_SQ_CF_INST_VTX:
+		case V_SQ_CF_WORD1_SQ_CF_INST_VTX_TC:
+			LIST_FOR_EACH_ENTRY(vtx, &cf->vtx, list) {
+				r = r600_bc_vtx_build(bc, vtx, addr);
+				if (r)
+					return r;
+				addr += 4;
+			}
+			break;
+		case V_SQ_CF_WORD1_SQ_CF_INST_TEX:
+			LIST_FOR_EACH_ENTRY(tex, &cf->tex, list) {
+				r = r600_bc_tex_build(bc, tex, addr);
+				if (r)
+					return r;
+				addr += 4;
+			}
+			break;
+		case V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT:
+		case V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT_DONE:
+			break;
+		default:
+			R600_ERR("unsupported CF instruction (0x%X)\n", cf->inst);
+			return -EINVAL;
+		}
+	}
+	return 0;
+}
diff --git a/src/gallium/drivers/r600/r600_asm.h b/src/gallium/drivers/r600/r600_asm.h
new file mode 100644
index 00000000000..10d98afaf00
--- /dev/null
+++ b/src/gallium/drivers/r600/r600_asm.h
@@ -0,0 +1,143 @@
+/*
+ * Copyright 2010 Jerome Glisse <[email protected]>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+#ifndef R600_ASM_H
+#define R600_ASM_H
+
+#include "radeon.h"
+#include "util/u_double_list.h"
+
+struct r600_bc_alu_src {
+	unsigned			sel;
+	unsigned			chan;
+	unsigned			neg;
+	unsigned			abs;
+};
+
+struct r600_bc_alu_dst {
+	unsigned			sel;
+	unsigned			chan;
+	unsigned			clamp;
+	unsigned			write;
+};
+
+struct r600_bc_alu {
+	struct list_head		list;
+	struct r600_bc_alu_src		src[3];
+	struct r600_bc_alu_dst		dst;
+	unsigned			inst;
+	unsigned			last;
+	unsigned			is_op3;
+	unsigned			nliteral;
+	unsigned			literal_added;
+	u32				value[4];
+};
+
+struct r600_bc_tex {
+	struct list_head		list;
+	unsigned			inst;
+	unsigned			resource_id;
+	unsigned			src_gpr;
+	unsigned			src_rel;
+	unsigned			dst_gpr;
+	unsigned			dst_rel;
+	unsigned			dst_sel_x;
+	unsigned			dst_sel_y;
+	unsigned			dst_sel_z;
+	unsigned			dst_sel_w;
+	unsigned			lod_bias;
+	unsigned			coord_type_x;
+	unsigned			coord_type_y;
+	unsigned			coord_type_z;
+	unsigned			coord_type_w;
+	unsigned			offset_x;
+	unsigned			offset_y;
+	unsigned			offset_z;
+	unsigned			sampler_id;
+	unsigned			src_sel_x;
+	unsigned			src_sel_y;
+	unsigned			src_sel_z;
+	unsigned			src_sel_w;
+};
+
+struct r600_bc_vtx {
+	struct list_head		list;
+	unsigned			inst;
+	unsigned			fetch_type;
+	unsigned			buffer_id;
+	unsigned			src_gpr;
+	unsigned			src_sel_x;
+	unsigned			mega_fetch_count;
+	unsigned			dst_gpr;
+	unsigned			dst_sel_x;
+	unsigned			dst_sel_y;
+	unsigned			dst_sel_z;
+	unsigned			dst_sel_w;
+};
+
+struct r600_bc_output {
+	unsigned			array_base;
+	unsigned			type;
+	unsigned			end_of_program;
+	unsigned			inst;
+	unsigned			elem_size;
+	unsigned			gpr;
+	unsigned			swizzle_x;
+	unsigned			swizzle_y;
+	unsigned			swizzle_z;
+	unsigned			swizzle_w;
+	unsigned			barrier;
+};
+
+struct r600_bc_cf {
+	struct list_head		list;
+	unsigned			inst;
+	unsigned			addr;
+	unsigned			ndw;
+	unsigned			id;
+	struct list_head		alu;
+	struct list_head		tex;
+	struct list_head		vtx;
+	struct r600_bc_output		output;
+};
+
+struct r600_bc {
+	enum radeon_family		family;
+	struct list_head		cf;
+	struct r600_bc_cf		*cf_last;
+	unsigned			ndw;
+	unsigned			ncf;
+	unsigned			ngpr;
+	unsigned			nresource;
+	unsigned			force_add_cf;
+	u32				*bytecode;
+};
+
+int r600_bc_init(struct r600_bc *bc, enum radeon_family family);
+int r600_bc_add_alu(struct r600_bc *bc, const struct r600_bc_alu *alu);
+int r600_bc_add_literal(struct r600_bc *bc, const u32 *value);
+int r600_bc_add_vtx(struct r600_bc *bc, const struct r600_bc_vtx *vtx);
+int r600_bc_add_tex(struct r600_bc *bc, const struct r600_bc_tex *tex);
+int r600_bc_add_output(struct r600_bc *bc, const struct r600_bc_output *output);
+int r600_bc_build(struct r600_bc *bc);
+
+#endif
diff --git a/src/gallium/drivers/r600/r600_blit.c b/src/gallium/drivers/r600/r600_blit.c
index 1dcb19babc9..f4eedfe4cb1 100644
--- a/src/gallium/drivers/r600/r600_blit.c
+++ b/src/gallium/drivers/r600/r600_blit.c
@@ -34,31 +34,37 @@
 
 static void r600_blitter_save_states(struct r600_context *rctx)
 {
-	util_blitter_save_blend(rctx->blitter,
-					rctx->draw->state[R600_BLEND]);
-	util_blitter_save_depth_stencil_alpha(rctx->blitter,
-					rctx->draw->state[R600_DSA]);
-	util_blitter_save_stencil_ref(rctx->blitter, &rctx->stencil_ref);
-	util_blitter_save_rasterizer(rctx->blitter,
-					rctx->draw->state[R600_RASTERIZER]);
-	util_blitter_save_fragment_shader(rctx->blitter,
-					rctx->ps_shader);
-	util_blitter_save_vertex_shader(rctx->blitter,
-					rctx->vs_shader);
-	util_blitter_save_vertex_elements(rctx->blitter,
-					rctx->vertex_elements);
-	util_blitter_save_viewport(rctx->blitter,
-					&rctx->viewport);
+	util_blitter_save_blend(rctx->blitter, rctx->blend);
+	util_blitter_save_depth_stencil_alpha(rctx->blitter, rctx->dsa);
+	if (rctx->stencil_ref) {
+		util_blitter_save_stencil_ref(rctx->blitter,
+					&rctx->stencil_ref->state.stencil_ref);
+	}
+	util_blitter_save_rasterizer(rctx->blitter, rctx->rasterizer);
+	util_blitter_save_fragment_shader(rctx->blitter, rctx->ps_shader);
+	util_blitter_save_vertex_shader(rctx->blitter, rctx->vs_shader);
+	util_blitter_save_vertex_elements(rctx->blitter, rctx->vertex_elements);
+	if (rctx->viewport) {
+		util_blitter_save_viewport(rctx->blitter, &rctx->viewport->state.viewport);
+	}
 	/* XXX util_blitter_save_clip(rctx->blitter, &rctx->clip); */
 	util_blitter_save_vertex_buffers(rctx->blitter, rctx->nvertex_buffer,
-					 rctx->vertex_buffer);
+					rctx->vertex_buffer);
+
+	/* remove ptr so they don't get deleted */
+	rctx->blend = NULL;
+	rctx->vs_shader = NULL;
+	rctx->ps_shader = NULL;
+	rctx->rasterizer = NULL;
+	rctx->dsa = NULL;
+	rctx->vertex_elements = NULL;
 }
 
 static void r600_clear(struct pipe_context *ctx, unsigned buffers,
 		       const float *rgba, double depth, unsigned stencil)
 {
 	struct r600_context *rctx = r600_context(ctx);
-	struct pipe_framebuffer_state *fb = &rctx->fb_state;
+	struct pipe_framebuffer_state *fb = &rctx->framebuffer->state.framebuffer;
 
 	r600_blitter_save_states(rctx);
 	util_blitter_clear(rctx->blitter, fb->width, fb->height,
@@ -73,9 +79,10 @@ static void r600_clear_render_target(struct pipe_context *pipe,
 				     unsigned width, unsigned height)
 {
 	struct r600_context *rctx = r600_context(pipe);
+	struct pipe_framebuffer_state *fb = &rctx->framebuffer->state.framebuffer;
 
 	r600_blitter_save_states(rctx);
-	util_blitter_save_framebuffer(rctx->blitter, &rctx->fb_state);
+	util_blitter_save_framebuffer(rctx->blitter, fb);
 
 	util_blitter_clear_render_target(rctx->blitter, dst, rgba,
 					 dstx, dsty, width, height);
@@ -90,9 +97,10 @@ static void r600_clear_depth_stencil(struct pipe_context *pipe,
 				     unsigned width, unsigned height)
 {
 	struct r600_context *rctx = r600_context(pipe);
+	struct pipe_framebuffer_state *fb = &rctx->framebuffer->state.framebuffer;
 
 	r600_blitter_save_states(rctx);
-	util_blitter_save_framebuffer(rctx->blitter, &rctx->fb_state);
+	util_blitter_save_framebuffer(rctx->blitter, fb);
 
 	util_blitter_clear_depth_stencil(rctx->blitter, dst, clear_flags, depth, stencil,
 					 dstx, dsty, width, height);
diff --git a/src/gallium/drivers/r600/r600_buffer.c b/src/gallium/drivers/r600/r600_buffer.c
index bc6e336ba7b..7829a479c2e 100644
--- a/src/gallium/drivers/r600/r600_buffer.c
+++ b/src/gallium/drivers/r600/r600_buffer.c
@@ -32,10 +32,11 @@
 #include "state_tracker/drm_driver.h"
 #include "r600_screen.h"
 #include "r600_context.h"
+#include "r600_resource.h"
 
 extern struct u_resource_vtbl r600_buffer_vtbl;
 
-static u32 r600_domain_from_usage(unsigned usage)
+u32 r600_domain_from_usage(unsigned usage)
 {
 	u32 domain = RADEON_GEM_DOMAIN_GTT;
 
@@ -63,47 +64,47 @@ struct pipe_resource *r600_buffer_create(struct pipe_screen *screen,
 					 const struct pipe_resource *templ)
 {
 	struct r600_screen *rscreen = r600_screen(screen);
-	struct r600_buffer *rbuffer;
+	struct r600_resource *rbuffer;
 	struct radeon_bo *bo;
 	struct pb_desc desc;
 	/* XXX We probably want a different alignment for buffers and textures. */
 	unsigned alignment = 4096;
 
-	rbuffer = CALLOC_STRUCT(r600_buffer);
+	rbuffer = CALLOC_STRUCT(r600_resource);
 	if (rbuffer == NULL)
 		return NULL;
 
-	rbuffer->b.b = *templ;
-	pipe_reference_init(&rbuffer->b.b.reference, 1);
-	rbuffer->b.b.screen = screen;
-	rbuffer->b.vtbl = &r600_buffer_vtbl;
+	rbuffer->base.b = *templ;
+	pipe_reference_init(&rbuffer->base.b.reference, 1);
+	rbuffer->base.b.screen = screen;
+	rbuffer->base.vtbl = &r600_buffer_vtbl;
 
-	if (rbuffer->b.b.bind & PIPE_BIND_CONSTANT_BUFFER) {
+	if (rbuffer->base.b.bind & PIPE_BIND_CONSTANT_BUFFER) {
 		desc.alignment = alignment;
-		desc.usage = rbuffer->b.b.bind;
-		rbuffer->pb = pb_malloc_buffer_create(rbuffer->b.b.width0,
+		desc.usage = rbuffer->base.b.bind;
+		rbuffer->pb = pb_malloc_buffer_create(rbuffer->base.b.width0,
 						      &desc);
 		if (rbuffer->pb == NULL) {
 			free(rbuffer);
 			return NULL;
 		}
-		return &rbuffer->b.b;
+		return &rbuffer->base.b;
 	}
-	rbuffer->domain = r600_domain_from_usage(rbuffer->b.b.bind);
-	bo = radeon_bo(rscreen->rw, 0, rbuffer->b.b.width0, alignment, NULL);
+	rbuffer->domain = r600_domain_from_usage(rbuffer->base.b.bind);
+	bo = radeon_bo(rscreen->rw, 0, rbuffer->base.b.width0, alignment, NULL);
 	if (bo == NULL) {
 		FREE(rbuffer);
 		return NULL;
 	}
 	rbuffer->bo = bo;
-	return &rbuffer->b.b;
+	return &rbuffer->base.b;
 }
 
 struct pipe_resource *r600_user_buffer_create(struct pipe_screen *screen,
 					      void *ptr, unsigned bytes,
 					      unsigned bind)
 {
-	struct r600_buffer *rbuffer;
+	struct r600_resource *rbuffer;
 	struct r600_screen *rscreen = r600_screen(screen);
 	struct pipe_resource templ;
 
@@ -116,20 +117,20 @@ struct pipe_resource *r600_user_buffer_create(struct pipe_screen *screen,
 	templ.height0 = 1;
 	templ.depth0 = 1;
 
-	rbuffer = (struct r600_buffer*)r600_buffer_create(screen, &templ);
+	rbuffer = (struct r600_resource*)r600_buffer_create(screen, &templ);
 	if (rbuffer == NULL) {
 		return NULL;
 	}
 	radeon_bo_map(rscreen->rw, rbuffer->bo);
 	memcpy(rbuffer->bo->data, ptr, bytes);
 	radeon_bo_unmap(rscreen->rw, rbuffer->bo);
-	return &rbuffer->b.b;
+	return &rbuffer->base.b;
 }
 
 static void r600_buffer_destroy(struct pipe_screen *screen,
 				struct pipe_resource *buf)
 {
-	struct r600_buffer *rbuffer = (struct r600_buffer*)buf;
+	struct r600_resource *rbuffer = (struct r600_resource*)buf;
 	struct r600_screen *rscreen = r600_screen(screen);
 
 	if (rbuffer->pb) {
@@ -140,13 +141,14 @@ static void r600_buffer_destroy(struct pipe_screen *screen,
 	if (rbuffer->bo) {
 		radeon_bo_decref(rscreen->rw, rbuffer->bo);
 	}
+	memset(rbuffer, 0, sizeof(struct r600_resource));
 	FREE(rbuffer);
 }
 
 static void *r600_buffer_transfer_map(struct pipe_context *pipe,
 				      struct pipe_transfer *transfer)
 {
-	struct r600_buffer *rbuffer = (struct r600_buffer*)transfer->resource;
+	struct r600_resource *rbuffer = (struct r600_resource*)transfer->resource;
 	struct r600_screen *rscreen = r600_screen(pipe->screen);
 	int write = 0;
 
@@ -166,9 +168,9 @@ static void *r600_buffer_transfer_map(struct pipe_context *pipe,
 }
 
 static void r600_buffer_transfer_unmap(struct pipe_context *pipe,
-				       struct pipe_transfer *transfer)
+					struct pipe_transfer *transfer)
 {
-	struct r600_buffer *rbuffer = (struct r600_buffer*)transfer->resource;
+	struct r600_resource *rbuffer = (struct r600_resource*)transfer->resource;
 	struct r600_screen *rscreen = r600_screen(pipe->screen);
 
 	if (rbuffer->pb) {
@@ -188,7 +190,7 @@ unsigned r600_buffer_is_referenced_by_cs(struct pipe_context *context,
 					 struct pipe_resource *buf,
 					 unsigned face, unsigned level)
 {
-	/* XXX */
+	/* FIXME */
 	return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE;
 }
 
@@ -196,7 +198,7 @@ struct pipe_resource *r600_buffer_from_handle(struct pipe_screen *screen,
 					      struct winsys_handle *whandle)
 {
 	struct radeon *rw = (struct radeon*)screen->winsys;
-	struct r600_buffer *rbuffer;
+	struct r600_resource *rbuffer;
 	struct radeon_bo *bo = NULL;
 
 	bo = radeon_bo(rw, whandle->handle, 0, 0, NULL);
@@ -204,18 +206,18 @@ struct pipe_resource *r600_buffer_from_handle(struct pipe_screen *screen,
 		return NULL;
 	}
 
-	rbuffer = CALLOC_STRUCT(r600_buffer);
+	rbuffer = CALLOC_STRUCT(r600_resource);
 	if (rbuffer == NULL) {
 		radeon_bo_decref(rw, bo);
 		return NULL;
 	}
 
-	pipe_reference_init(&rbuffer->b.b.reference, 1);
-	rbuffer->b.b.target = PIPE_BUFFER;
-	rbuffer->b.b.screen = screen;
-	rbuffer->b.vtbl = &r600_buffer_vtbl;
+	pipe_reference_init(&rbuffer->base.b.reference, 1);
+	rbuffer->base.b.target = PIPE_BUFFER;
+	rbuffer->base.b.screen = screen;
+	rbuffer->base.vtbl = &r600_buffer_vtbl;
 	rbuffer->bo = bo;
-	return &rbuffer->b.b;
+	return &rbuffer->base.b;
 }
 
 struct u_resource_vtbl r600_buffer_vtbl =
diff --git a/src/gallium/drivers/r600/r600_compiler.c b/src/gallium/drivers/r600/r600_compiler.c
deleted file mode 100644
index 1804b86d24e..00000000000
--- a/src/gallium/drivers/r600/r600_compiler.c
+++ /dev/null
@@ -1,447 +0,0 @@
-/*
- * Copyright 2010 Jerome Glisse <[email protected]>
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * on the rights to use, copy, modify, merge, publish, distribute, sub
- * license, and/or sell copies of the Software, and to permit persons to whom
- * the Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
- * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
- * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
- * USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-#include <stdlib.h>
-#include <string.h>
-#include <stdint.h>
-#include <stdio.h>
-#include <errno.h>
-#include "r600_compiler.h"
-
-struct c_vector *c_vector_new(void)
-{
-	struct c_vector *v = calloc(1, sizeof(struct c_vector));
-
-	if (v == NULL) {
-		return NULL;
-	}
-	LIST_INITHEAD(&v->head);
-	return v;
-}
-
-static unsigned c_opcode_is_alu(unsigned opcode)
-{
-	switch (opcode) {
-	case C_OPCODE_MOV:
-	case C_OPCODE_MUL:
-	case C_OPCODE_MAD:
-	case C_OPCODE_ARL:
-	case C_OPCODE_LIT:
-	case C_OPCODE_RCP:
-	case C_OPCODE_RSQ:
-	case C_OPCODE_EXP:
-	case C_OPCODE_LOG:
-	case C_OPCODE_ADD:
-	case C_OPCODE_DP3:
-	case C_OPCODE_DP4:
-	case C_OPCODE_DST:
-	case C_OPCODE_MIN:
-	case C_OPCODE_MAX:
-	case C_OPCODE_SLT:
-	case C_OPCODE_SGE:
-	case C_OPCODE_SUB:
-	case C_OPCODE_LRP:
-	case C_OPCODE_CND:
-	case C_OPCODE_DP2A:
-	case C_OPCODE_FRC:
-	case C_OPCODE_CLAMP:
-	case C_OPCODE_FLR:
-	case C_OPCODE_ROUND:
-	case C_OPCODE_EX2:
-	case C_OPCODE_LG2:
-	case C_OPCODE_POW:
-	case C_OPCODE_XPD:
-	case C_OPCODE_ABS:
-	case C_OPCODE_RCC:
-	case C_OPCODE_DPH:
-	case C_OPCODE_COS:
-	case C_OPCODE_DDX:
-	case C_OPCODE_DDY:
-	case C_OPCODE_PK2H:
-	case C_OPCODE_PK2US:
-	case C_OPCODE_PK4B:
-	case C_OPCODE_PK4UB:
-	case C_OPCODE_RFL:
-	case C_OPCODE_SEQ:
-	case C_OPCODE_SFL:
-	case C_OPCODE_SGT:
-	case C_OPCODE_SIN:
-	case C_OPCODE_SLE:
-	case C_OPCODE_SNE:
-	case C_OPCODE_STR:
-	case C_OPCODE_UP2H:
-	case C_OPCODE_UP2US:
-	case C_OPCODE_UP4B:
-	case C_OPCODE_UP4UB:
-	case C_OPCODE_X2D:
-	case C_OPCODE_ARA:
-	case C_OPCODE_ARR:
-	case C_OPCODE_BRA:
-	case C_OPCODE_SSG:
-	case C_OPCODE_CMP:
-	case C_OPCODE_SCS:
-	case C_OPCODE_NRM:
-	case C_OPCODE_DIV:
-	case C_OPCODE_DP2:
-	case C_OPCODE_CEIL:
-	case C_OPCODE_I2F:
-	case C_OPCODE_NOT:
-	case C_OPCODE_TRUNC:
-	case C_OPCODE_SHL:
-	case C_OPCODE_AND:
-	case C_OPCODE_OR:
-	case C_OPCODE_MOD:
-	case C_OPCODE_XOR:
-	case C_OPCODE_SAD:
-	case C_OPCODE_NRM4:
-	case C_OPCODE_F2I:
-	case C_OPCODE_IDIV:
-	case C_OPCODE_IMAX:
-	case C_OPCODE_IMIN:
-	case C_OPCODE_INEG:
-	case C_OPCODE_ISGE:
-	case C_OPCODE_ISHR:
-	case C_OPCODE_ISLT:
-	case C_OPCODE_F2U:
-	case C_OPCODE_U2F:
-	case C_OPCODE_UADD:
-	case C_OPCODE_UDIV:
-	case C_OPCODE_UMAD:
-	case C_OPCODE_UMAX:
-	case C_OPCODE_UMIN:
-	case C_OPCODE_UMOD:
-	case C_OPCODE_UMUL:
-	case C_OPCODE_USEQ:
-	case C_OPCODE_USGE:
-	case C_OPCODE_USHR:
-	case C_OPCODE_USLT:
-	case C_OPCODE_USNE:
-		return 1;
-	case C_OPCODE_END:
-	case C_OPCODE_VFETCH:
-	case C_OPCODE_KILP:
-	case C_OPCODE_CAL:
-	case C_OPCODE_RET:
-	case C_OPCODE_TXB:
-	case C_OPCODE_TXL:
-	case C_OPCODE_BRK:
-	case C_OPCODE_IF:
-	case C_OPCODE_BGNFOR:
-	case C_OPCODE_REP:
-	case C_OPCODE_ELSE:
-	case C_OPCODE_ENDIF:
-	case C_OPCODE_ENDFOR:
-	case C_OPCODE_ENDREP:
-	case C_OPCODE_PUSHA:
-	case C_OPCODE_POPA:
-	case C_OPCODE_TXF:
-	case C_OPCODE_TXQ:
-	case C_OPCODE_CONT:
-	case C_OPCODE_EMIT:
-	case C_OPCODE_ENDPRIM:
-	case C_OPCODE_BGNLOOP:
-	case C_OPCODE_BGNSUB:
-	case C_OPCODE_ENDLOOP:
-	case C_OPCODE_ENDSUB:
-	case C_OPCODE_NOP:
-	case C_OPCODE_CALLNZ:
-	case C_OPCODE_IFC:
-	case C_OPCODE_BREAKC:
-	case C_OPCODE_KIL:
-	case C_OPCODE_TEX:
-	case C_OPCODE_TXD:
-	case C_OPCODE_TXP:
-	case C_OPCODE_SWITCH:
-	case C_OPCODE_CASE:
-	case C_OPCODE_DEFAULT:
-	case C_OPCODE_ENDSWITCH:
-	default:
-		return 0;
-	}
-}
-
-
-/* NEW */
-void c_node_init(struct c_node *node)
-{
-	memset(node, 0, sizeof(struct c_node));
-	LIST_INITHEAD(&node->predecessors);
-	LIST_INITHEAD(&node->successors);
-	LIST_INITHEAD(&node->childs);
-	LIST_INITHEAD(&node->insts);
-	node->parent = NULL;
-}
-
-static struct c_node_link *c_node_link_new(struct c_node *node)
-{
-	struct c_node_link *link;
-
-	link = calloc(1, sizeof(struct c_node_link));
-	if (link == NULL)
-		return NULL;
-	LIST_INITHEAD(&link->head);
-	link->node = node;
-	return link;
-}
-
-int c_node_cfg_link(struct c_node *predecessor, struct c_node *successor)
-{
-	struct c_node_link *pedge, *sedge;
-
-	pedge = c_node_link_new(successor);
-	sedge = c_node_link_new(predecessor);
-	if (sedge == NULL || pedge == NULL) {
-		free(sedge);
-		free(pedge);
-		return -ENOMEM;
-	}
-	LIST_ADDTAIL(&pedge->head, &predecessor->successors);
-	LIST_ADDTAIL(&sedge->head, &successor->predecessors);
-
-	return 0;
-}
-
-int c_node_add_new_instruction_head(struct c_node *node, struct c_instruction *instruction)
-{
-	struct c_instruction *inst = malloc(sizeof(struct c_instruction));
-
-	if (inst == NULL)
-		return -ENOMEM;
-	memcpy(inst, instruction, sizeof(struct c_instruction));
-	LIST_ADD(&inst->head, &node->insts);
-	return 0;
-}
-
-int c_node_add_new_instruction(struct c_node *node, struct c_instruction *instruction)
-{
-	struct c_instruction *inst = malloc(sizeof(struct c_instruction));
-
-	if (inst == NULL)
-		return -ENOMEM;
-	memcpy(inst, instruction, sizeof(struct c_instruction));
-	LIST_ADDTAIL(&inst->head, &node->insts);
-	return 0;
-}
-
-struct c_node *c_shader_cfg_new_node_after(struct c_shader *shader, struct c_node *predecessor)
-{
-	struct c_node *node = calloc(1, sizeof(struct c_node));
-
-	if (node == NULL)
-		return NULL;
-	c_node_init(node);
-	if (c_node_cfg_link(predecessor, node)) {
-		free(node);
-		return NULL;
-	}
-	LIST_ADDTAIL(&node->head, &shader->nodes);
-	return node;
-}
-
-int c_shader_init(struct c_shader *shader, unsigned type)
-{
-	unsigned i;
-	int r;
-
-	shader->type = type;
-	for (i = 0; i < C_FILE_COUNT; i++) {
-		shader->files[i].nvectors = 0;
-		LIST_INITHEAD(&shader->files[i].vectors);
-	}
-	LIST_INITHEAD(&shader->nodes);
-	c_node_init(&shader->entry);
-	c_node_init(&shader->end);
-	shader->entry.opcode = C_OPCODE_ENTRY;
-	shader->end.opcode = C_OPCODE_END;
-	r = c_node_cfg_link(&shader->entry, &shader->end);
-	if (r)
-		return r;
-	return 0;
-}
-
-struct c_vector *c_shader_vector_new(struct c_shader *shader, unsigned file, unsigned name, int sid)
-{
-	struct c_vector *v = calloc(1, sizeof(struct c_vector));
-	int i;
-
-	if (v == NULL) {
-		return NULL;
-	}
-	for (i = 0; i < 4; i++) {
-		v->channel[i] = calloc(1, sizeof(struct c_channel));
-		if (v->channel[i] == NULL)
-			goto out_err;
-		v->channel[i]->vindex = i;
-		v->channel[i]->vector = v;
-	}
-	v->file = file;
-	v->name = name;
-	v->sid = sid;
-	shader->files[v->file].nvectors++;
-	v->id = shader->nvectors++;
-	LIST_ADDTAIL(&v->head, &shader->files[v->file].vectors);
-	return v;
-out_err:
-	for (i = 0; i < 4; i++) {
-		free(v->channel[i]);
-	}
-	free(v);
-	return NULL;
-}
-
-static void c_node_remove_link(struct list_head *head, struct c_node *node)
-{
-	struct c_node_link *link, *tmp;
-
-	LIST_FOR_EACH_ENTRY_SAFE(link, tmp, head, head) {
-		if (link->node == node) {
-			LIST_DEL(&link->head);
-			free(link);
-		}
-	}
-}
-
-static void c_node_destroy(struct c_node *node)
-{
-	struct c_instruction *i, *ni;
-	struct c_node_link *link, *tmp;
-
-	LIST_FOR_EACH_ENTRY_SAFE(i, ni, &node->insts, head) {
-		LIST_DEL(&i->head);
-		free(i);
-	}
-	if (node->parent)
-		c_node_remove_link(&node->parent->childs, node);
-	node->parent = NULL;
-	LIST_FOR_EACH_ENTRY_SAFE(link, tmp, &node->predecessors, head) {
-		c_node_remove_link(&link->node->successors, node);
-		LIST_DEL(&link->head);
-		free(link);
-	}
-	LIST_FOR_EACH_ENTRY_SAFE(link, tmp, &node->successors, head) {
-		c_node_remove_link(&link->node->predecessors, node);
-		LIST_DEL(&link->head);
-		free(link);
-	}
-	LIST_FOR_EACH_ENTRY_SAFE(link, tmp, &node->childs, head) {
-		link->node->parent = NULL;
-		LIST_DEL(&link->head);
-		free(link);
-	}
-}
-
-void c_shader_destroy(struct c_shader *shader)
-{
-	struct c_node *n, *nn;
-	struct c_vector *v, *nv;
-	unsigned i;
-
-	for (i = 0; i < C_FILE_COUNT; i++) {
-		shader->files[i].nvectors = 0;
-		LIST_FOR_EACH_ENTRY_SAFE(v, nv, &shader->files[i].vectors, head) {
-			LIST_DEL(&v->head);
-			free(v->channel[0]);
-			free(v->channel[1]);
-			free(v->channel[2]);
-			free(v->channel[3]);
-			free(v);
-		}
-	}
-	LIST_FOR_EACH_ENTRY_SAFE(n, nn, &shader->nodes, head) {
-		LIST_DEL(&n->head);
-		c_node_destroy(n);
-	}
-	memset(shader, 0, sizeof(struct c_shader));
-}
-
-static void c_shader_dfs_without_rec(struct c_node *entry, struct c_node *node)
-{
-	struct c_node_link *link;
-
-	if (entry == node || entry->visited)
-		return;
-	entry->visited = 1;
-	LIST_FOR_EACH_ENTRY(link, &entry->successors, head) {
-		c_shader_dfs_without_rec(link->node, node);
-	}
-}
-
-static void c_shader_dfs_without(struct c_shader *shader, struct c_node *node)
-{
-	struct c_node *n;
-
-	shader->entry.visited = 0;
-	shader->end.visited = 0;
-	LIST_FOR_EACH_ENTRY(n, &shader->nodes, head) {
-		n->visited = 0;
-	}
-	c_shader_dfs_without_rec(&shader->entry, node);
-}
-
-static int c_shader_build_dominator_tree_rec(struct c_shader *shader, struct c_node *node)
-{
-	struct c_node_link *link, *nlink;
-	unsigned found = 0;
-	int r;
-
-	if (node->done)
-		return 0;
-	node->done = 1;
-	LIST_FOR_EACH_ENTRY(link, &node->predecessors, head) {
-		/* if we remove this predecessor can we reach the current node ? */
-		c_shader_dfs_without(shader, link->node);
-		if (node->visited == 0) {
-			/* we were unable to visit current node thus current
-			 * predecessor  is the immediate dominator of node, as
-			 * their can be only one immediate dominator we break
-			 */
-			node->parent = link->node;
-			nlink = c_node_link_new(node);
-			if (nlink == NULL)
-				return -ENOMEM;
-			LIST_ADDTAIL(&nlink->head, &link->node->childs);
-			found = 1;
-			break;
-		}
-	}
-	/* this shouldn't happen there should at least be 1 denominator for each node */
-	if (!found && node->opcode != C_OPCODE_ENTRY) {
-		fprintf(stderr, "invalid flow control graph node %p (%d) has no immediate dominator\n",
-			node, node->opcode);
-		return -EINVAL;
-	}
-	LIST_FOR_EACH_ENTRY(link, &node->predecessors, head) {
-		r = c_shader_build_dominator_tree_rec(shader, link->node);
-		if (r)
-			return r;
-	}
-	return 0;
-}
-
-int c_shader_build_dominator_tree(struct c_shader *shader)
-{
-	struct c_node *node;
-	LIST_FOR_EACH_ENTRY(node, &shader->nodes, head) {
-		node->done = 0;
-	}
-	return c_shader_build_dominator_tree_rec(shader, &shader->end);
-}
diff --git a/src/gallium/drivers/r600/r600_compiler.h b/src/gallium/drivers/r600/r600_compiler.h
deleted file mode 100644
index 77230aed73b..00000000000
--- a/src/gallium/drivers/r600/r600_compiler.h
+++ /dev/null
@@ -1,320 +0,0 @@
-/*
- * Copyright 2010 Jerome Glisse <[email protected]>
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * on the rights to use, copy, modify, merge, publish, distribute, sub
- * license, and/or sell copies of the Software, and to permit persons to whom
- * the Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
- * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
- * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
- * USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-#ifndef R600_COMPILER_H
-#define R600_COMPILER_H
-
-#include "util/u_double_list.h"
-
-struct c_vector;
-
-/* operand are the basic source/destination of each operation */
-struct c_channel {
-	struct list_head        head;
-	unsigned		vindex;		/**< index in vector X,Y,Z,W (0,1,2,3) */
-	unsigned		value;		/**< immediate value 32bits */
-	struct c_vector		*vector;	/**< vector to which it belongs */
-};
-
-/* in GPU world most of the time operand are grouped into vector
- * of 4 component this structure is mostly and handler to group
- * operand into a same vector
- */
-struct c_vector {
-	struct list_head        head;
-	unsigned		id;		/**< vector uniq id */
-	unsigned		name;		/**< semantic name */
-	unsigned		file;		/**< operand file C_FILE_* */
-	int			sid;		/**< semantic id */
-	struct c_channel	*channel[4];	/**< operands */
-};
-
-#define C_PROGRAM_TYPE_VS	0
-#define C_PROGRAM_TYPE_FS	1
-#define C_PROGRAM_TYPE_COUNT	2
-
-#define C_NODE_FLAG_ALU		1
-#define C_NODE_FLAG_FETCH	2
-
-#define C_SWIZZLE_X		0
-#define C_SWIZZLE_Y		1
-#define C_SWIZZLE_Z		2
-#define C_SWIZZLE_W		3
-#define C_SWIZZLE_0		4
-#define C_SWIZZLE_1		5
-#define C_SWIZZLE_D		6
-
-#define C_FILE_NULL		0
-#define C_FILE_CONSTANT		1
-#define C_FILE_INPUT		2
-#define C_FILE_OUTPUT		3
-#define C_FILE_TEMPORARY	4
-#define C_FILE_SAMPLER		5
-#define C_FILE_ADDRESS		6
-#define C_FILE_IMMEDIATE	7
-#define C_FILE_LOOP		8
-#define C_FILE_PREDICATE	9
-#define C_FILE_SYSTEM_VALUE	10
-#define C_FILE_RESOURCE		11
-#define C_FILE_COUNT		12
-
-#define C_SEMANTIC_POSITION	0
-#define C_SEMANTIC_COLOR	1
-#define C_SEMANTIC_BCOLOR	2  /**< back-face color */
-#define C_SEMANTIC_FOG		3
-#define C_SEMANTIC_PSIZE	4
-#define C_SEMANTIC_GENERIC	5
-#define C_SEMANTIC_NORMAL	6
-#define C_SEMANTIC_FACE		7
-#define C_SEMANTIC_EDGEFLAG	8
-#define C_SEMANTIC_PRIMID	9
-#define C_SEMANTIC_INSTANCEID	10
-#define C_SEMANTIC_VERTEXID	11
-#define C_SEMANTIC_COUNT	12 /**< number of semantic values */
-
-#define C_OPCODE_NOP		0
-#define C_OPCODE_MOV		1
-#define C_OPCODE_LIT		2
-#define C_OPCODE_RCP		3
-#define C_OPCODE_RSQ		4
-#define C_OPCODE_EXP		5
-#define C_OPCODE_LOG		6
-#define C_OPCODE_MUL		7
-#define C_OPCODE_ADD		8
-#define C_OPCODE_DP3		9
-#define C_OPCODE_DP4		10
-#define C_OPCODE_DST		11
-#define C_OPCODE_MIN		12
-#define C_OPCODE_MAX		13
-#define C_OPCODE_SLT		14
-#define C_OPCODE_SGE		15
-#define C_OPCODE_MAD		16
-#define C_OPCODE_SUB		17
-#define C_OPCODE_LRP		18
-#define C_OPCODE_CND		19
-/* gap */
-#define C_OPCODE_DP2A		21
-/* gap */
-#define C_OPCODE_FRC		24
-#define C_OPCODE_CLAMP		25
-#define C_OPCODE_FLR		26
-#define C_OPCODE_ROUND		27
-#define C_OPCODE_EX2		28
-#define C_OPCODE_LG2		29
-#define C_OPCODE_POW		30
-#define C_OPCODE_XPD		31
-/* gap */
-#define C_OPCODE_ABS		33
-#define C_OPCODE_RCC		34
-#define C_OPCODE_DPH		35
-#define C_OPCODE_COS		36
-#define C_OPCODE_DDX		37
-#define C_OPCODE_DDY		38
-#define C_OPCODE_KILP		39		/* predicated kill */
-#define C_OPCODE_PK2H		40
-#define C_OPCODE_PK2US		41
-#define C_OPCODE_PK4B		42
-#define C_OPCODE_PK4UB		43
-#define C_OPCODE_RFL		44
-#define C_OPCODE_SEQ		45
-#define C_OPCODE_SFL		46
-#define C_OPCODE_SGT		47
-#define C_OPCODE_SIN		48
-#define C_OPCODE_SLE		49
-#define C_OPCODE_SNE		50
-#define C_OPCODE_STR		51
-#define C_OPCODE_TEX		52
-#define C_OPCODE_TXD		53
-#define C_OPCODE_TXP		54
-#define C_OPCODE_UP2H		55
-#define C_OPCODE_UP2US		56
-#define C_OPCODE_UP4B		57
-#define C_OPCODE_UP4UB		58
-#define C_OPCODE_X2D		59
-#define C_OPCODE_ARA		60
-#define C_OPCODE_ARR		61
-#define C_OPCODE_BRA		62
-#define C_OPCODE_CAL		63
-#define C_OPCODE_RET		64
-#define C_OPCODE_SSG		65		/* SGN */
-#define C_OPCODE_CMP		66
-#define C_OPCODE_SCS		67
-#define C_OPCODE_TXB		68
-#define C_OPCODE_NRM		69
-#define C_OPCODE_DIV		70
-#define C_OPCODE_DP2		71
-#define C_OPCODE_TXL		72
-#define C_OPCODE_BRK		73
-#define C_OPCODE_IF		74
-#define C_OPCODE_BGNFOR		75
-#define C_OPCODE_REP		76
-#define C_OPCODE_ELSE		77
-#define C_OPCODE_ENDIF		78
-#define C_OPCODE_ENDFOR		79
-#define C_OPCODE_ENDREP		80
-#define C_OPCODE_PUSHA		81
-#define C_OPCODE_POPA		82
-#define C_OPCODE_CEIL		83
-#define C_OPCODE_I2F		84
-#define C_OPCODE_NOT		85
-#define C_OPCODE_TRUNC		86
-#define C_OPCODE_SHL		87
-/* gap */
-#define C_OPCODE_AND		89
-#define C_OPCODE_OR		90
-#define C_OPCODE_MOD		91
-#define C_OPCODE_XOR		92
-#define C_OPCODE_SAD		93
-#define C_OPCODE_TXF		94
-#define C_OPCODE_TXQ		95
-#define C_OPCODE_CONT		96
-#define C_OPCODE_EMIT		97
-#define C_OPCODE_ENDPRIM	98
-#define C_OPCODE_BGNLOOP	99
-#define C_OPCODE_BGNSUB		100
-#define C_OPCODE_ENDLOOP	101
-#define C_OPCODE_ENDSUB		102
-/* gap */
-#define C_OPCODE_NRM4		112
-#define C_OPCODE_CALLNZ		113
-#define C_OPCODE_IFC		114
-#define C_OPCODE_BREAKC		115
-#define C_OPCODE_KIL		116	/* conditional kill */
-#define C_OPCODE_END		117	/* aka HALT */
-/* gap */
-#define C_OPCODE_F2I		119
-#define C_OPCODE_IDIV		120
-#define C_OPCODE_IMAX		121
-#define C_OPCODE_IMIN		122
-#define C_OPCODE_INEG		123
-#define C_OPCODE_ISGE		124
-#define C_OPCODE_ISHR		125
-#define C_OPCODE_ISLT		126
-#define C_OPCODE_F2U		127
-#define C_OPCODE_U2F		128
-#define C_OPCODE_UADD		129
-#define C_OPCODE_UDIV		130
-#define C_OPCODE_UMAD		131
-#define C_OPCODE_UMAX		132
-#define C_OPCODE_UMIN		133
-#define C_OPCODE_UMOD		134
-#define C_OPCODE_UMUL		135
-#define C_OPCODE_USEQ		136
-#define C_OPCODE_USGE		137
-#define C_OPCODE_USHR		138
-#define C_OPCODE_USLT		139
-#define C_OPCODE_USNE		140
-#define C_OPCODE_SWITCH		141
-#define C_OPCODE_CASE		142
-#define C_OPCODE_DEFAULT	143
-#define C_OPCODE_ENDSWITCH	144
-#define C_OPCODE_VFETCH		145
-#define C_OPCODE_ENTRY		146
-#define C_OPCODE_ARL		147
-#define C_OPCODE_LAST		148
-
-#define C_OPERAND_FLAG_ABS		(1 << 0)
-#define C_OPERAND_FLAG_NEG		(1 << 1)
-
-struct c_operand {
-	struct c_vector		*vector;
-	unsigned		swizzle;
-	unsigned		flag;
-};
-
-struct c_op {
-	unsigned		ninput;
-	struct c_operand	input[3];
-	struct c_operand	output;
-	unsigned		opcode;
-};
-
-struct c_instruction {
-	struct list_head        head;
-	unsigned		nop;
-	struct c_op		op[5];
-};
-
-struct c_node;
-
-struct c_node_link {
-	struct list_head        head;
-	struct c_node		*node;
-};
-
-/**
- * struct c_node
- *
- * @next:		all node are in a double linked list, this point to
- * 			next node
- * @next:		all node are in a double linked list, this point to
- * 			previous node
- * @predecessors:	list of all predecessor nodes in the flow graph
- * @successors:		list of all sucessor nodes in the flow graph
- * @parent:		parent node in the depth first walk tree
- * @childs:		child nodes in the depth first walk tree
- */
-struct c_node {
-	struct list_head        head;
-	struct list_head        predecessors;
-	struct list_head        successors;
-	struct list_head        childs;
-	struct c_node		*parent;
-	struct list_head        insts;
-	unsigned		opcode;
-	unsigned		visited;
-	unsigned		done;
-	void			*backend;
-};
-
-struct c_file {
-	unsigned		nvectors;
-	struct list_head        vectors;
-};
-
-struct c_shader {
-	unsigned			nvectors;
-	struct c_file			files[C_FILE_COUNT];
-	struct list_head                nodes;
-	struct c_node			entry;
-	struct c_node			end;
-	unsigned			type;
-};
-
-int c_shader_init(struct c_shader *shader, unsigned type);
-void c_shader_destroy(struct c_shader *shader);
-struct c_vector *c_shader_vector_new(struct c_shader *shader, unsigned file, unsigned name, int sid);
-int c_shader_build_dominator_tree(struct c_shader *shader);
-void c_shader_dump(struct c_shader *shader);
-
-void c_node_init(struct c_node *node);
-int c_node_add_new_instruction(struct c_node *node, struct c_instruction *instruction);
-int c_node_add_new_instruction_head(struct c_node *node, struct c_instruction *instruction);
-
-/* control flow graph functions */
-int c_node_cfg_link(struct c_node *predecessor, struct c_node *successor);
-struct c_node *c_node_cfg_new_after(struct c_node *predecessor);
-struct c_node *c_shader_cfg_new_node_after(struct c_shader *shader, struct c_node *predecessor);
-
-struct c_vector *c_vector_new(void);
-
-#endif
diff --git a/src/gallium/drivers/r600/r600_compiler_dump.c b/src/gallium/drivers/r600/r600_compiler_dump.c
deleted file mode 100644
index bb022b7c298..00000000000
--- a/src/gallium/drivers/r600/r600_compiler_dump.c
+++ /dev/null
@@ -1,267 +0,0 @@
-/*
- * Copyright 2010 Jerome Glisse <[email protected]>
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * on the rights to use, copy, modify, merge, publish, distribute, sub
- * license, and/or sell copies of the Software, and to permit persons to whom
- * the Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
- * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
- * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
- * USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-#include <stdlib.h>
-#include <string.h>
-#include <stdint.h>
-#include <stdio.h>
-#include "r600_compiler.h"
-
-static const char *c_file_swz[] = {
-	"x",
-	"y",
-	"z",
-	"w",
-	"0",
-	"1",
-	".",
-};
-
-static const char *c_file_str[] = {
-	"NULL",
-	"CONSTANT",
-	"INPUT",
-	"OUTPUT",
-	"TEMPORARY",
-	"SAMPLER",
-	"ADDRESS",
-	"IMMEDIATE",
-	"LOOP",
-	"PREDICATE",
-	"SYSTEM_VALUE",
-};
-
-static const char *c_semantic_str[] = {
-	"POSITION",
-	"COLOR",
-	"BCOLOR",
-	"FOG",
-	"PSIZE",
-	"GENERIC",
-	"NORMAL",
-	"FACE",
-	"EDGEFLAG",
-	"PRIMID",
-	"INSTANCEID",
-};
-
-static const char *c_opcode_str[] = {
-	"ARL",
-	"MOV",
-	"LIT",
-	"RCP",
-	"RSQ",
-	"EXP",
-	"LOG",
-	"MUL",
-	"ADD",
-	"DP3",
-	"DP4",
-	"DST",
-	"MIN",
-	"MAX",
-	"SLT",
-	"SGE",
-	"MAD",
-	"SUB",
-	"LRP",
-	"CND",
-	"(INVALID)",
-	"DP2A",
-	"(INVALID)",
-	"(INVALID)",
-	"FRC",
-	"CLAMP",
-	"FLR",
-	"ROUND",
-	"EX2",
-	"LG2",
-	"POW",
-	"XPD",
-	"(INVALID)",
-	"ABS",
-	"RCC",
-	"DPH",
-	"COS",
-	"DDX",
-	"DDY",
-	"KILP",
-	"PK2H",
-	"PK2US",
-	"PK4B",
-	"PK4UB",
-	"RFL",
-	"SEQ",
-	"SFL",
-	"SGT",
-	"SIN",
-	"SLE",
-	"SNE",
-	"STR",
-	"TEX",
-	"TXD",
-	"TXP",
-	"UP2H",
-	"UP2US",
-	"UP4B",
-	"UP4UB",
-	"X2D",
-	"ARA",
-	"ARR",
-	"BRA",
-	"CAL",
-	"RET",
-	"SSG",
-	"CMP",
-	"SCS",
-	"TXB",
-	"NRM",
-	"DIV",
-	"DP2",
-	"TXL",
-	"BRK",
-	"IF",
-	"BGNFOR",
-	"REP",
-	"ELSE",
-	"ENDIF",
-	"ENDFOR",
-	"ENDREP",
-	"PUSHA",
-	"POPA",
-	"CEIL",
-	"I2F",
-	"NOT",
-	"TRUNC",
-	"SHL",
-	"(INVALID)",
-	"AND",
-	"OR",
-	"MOD",
-	"XOR",
-	"SAD",
-	"TXF",
-	"TXQ",
-	"CONT",
-	"EMIT",
-	"ENDPRIM",
-	"BGNLOOP",
-	"BGNSUB",
-	"ENDLOOP",
-	"ENDSUB",
-	"(INVALID)",
-	"(INVALID)",
-	"(INVALID)",
-	"(INVALID)",
-	"NOP",
-	"(INVALID)",
-	"(INVALID)",
-	"(INVALID)",
-	"(INVALID)",
-	"NRM4",
-	"CALLNZ",
-	"IFC",
-	"BREAKC",
-	"KIL",
-	"END",
-	"(INVALID)",
-	"F2I",
-	"IDIV",
-	"IMAX",
-	"IMIN",
-	"INEG",
-	"ISGE",
-	"ISHR",
-	"ISLT",
-	"F2U",
-	"U2F",
-	"UADD",
-	"UDIV",
-	"UMAD",
-	"UMAX",
-	"UMIN",
-	"UMOD",
-	"UMUL",
-	"USEQ",
-	"USGE",
-	"USHR",
-	"USLT",
-	"USNE",
-	"SWITCH",
-	"CASE",
-	"DEFAULT",
-	"ENDSWITCH",
-	"VFETCH",
-	"ENTRY",
-};
-
-static inline const char *c_get_name(const char *name[], unsigned i)
-{
-	return name[i];
-}
-
-static void pindent(unsigned indent)
-{
-	unsigned i;
-	for (i = 0; i < indent; i++)
-		fprintf(stderr, " ");
-}
-
-static void c_node_dump(struct c_node *node, unsigned indent)
-{
-	struct c_instruction *i;
-	unsigned j, k;
-
-	pindent(indent); fprintf(stderr, "# node %s\n", c_get_name(c_opcode_str, node->opcode));
-	LIST_FOR_EACH_ENTRY(i, &node->insts, head) {
-		for (k = 0; k < i->nop; k++) {
-			pindent(indent);
-			fprintf(stderr, "%s", c_get_name(c_opcode_str, i->op[k].opcode));
-			fprintf(stderr, " %s[%d][%s]",
-				c_get_name(c_file_str, i->op[k].output.vector->file),
-				i->op[k].output.vector->id,
-				c_get_name(c_file_swz, i->op[k].output.swizzle));
-			for (j = 0; j < i->op[k].ninput; j++) {
-				fprintf(stderr, " %s[%d][%s]",
-						c_get_name(c_file_str, i->op[k].input[j].vector->file),
-						i->op[k].input[j].vector->id,
-						c_get_name(c_file_swz, i->op[k].input[j].swizzle));
-			}
-			fprintf(stderr, ";\n");
-		}
-	}
-}
-
-static void c_shader_dump_rec(struct c_shader *shader, struct c_node *node, unsigned indent)
-{
-	struct c_node_link *link;
-
-	c_node_dump(node, indent);
-	LIST_FOR_EACH_ENTRY(link, &node->childs, head) {
-		c_shader_dump_rec(shader, link->node, indent + 1);
-	}
-}
-
-void c_shader_dump(struct c_shader *shader)
-{
-	c_shader_dump_rec(shader, &shader->entry, 0);
-}
diff --git a/src/gallium/drivers/r600/r600_compiler_r600.c b/src/gallium/drivers/r600/r600_compiler_r600.c
deleted file mode 100644
index 27ad8f1a182..00000000000
--- a/src/gallium/drivers/r600/r600_compiler_r600.c
+++ /dev/null
@@ -1,972 +0,0 @@
-/*
- * Copyright 2010 Jerome Glisse <[email protected]>
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * on the rights to use, copy, modify, merge, publish, distribute, sub
- * license, and/or sell copies of the Software, and to permit persons to whom
- * the Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
- * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
- * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
- * USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-#include <stdlib.h>
-#include <string.h>
-#include <stdint.h>
-#include <stdio.h>
-#include <errno.h>
-#include <util/u_format.h>
-#include "r600_screen.h"
-#include "r600_context.h"
-#include "r600_sq.h"
-
-
-struct r600_alu_instruction {
-	unsigned			copcode;
-	enum r600_instruction		instruction;
-};
-
-static int r600_shader_alu_translate(struct r600_shader *rshader,
-					struct r600_shader_node *node,
-					struct c_instruction *instruction);
-struct r600_alu_instruction r600_alu_instruction[C_OPCODE_LAST];
-struct r600_instruction_info r600_instruction_info[];
-
-int r600_shader_insert_fetch(struct c_shader *shader)
-{
-	struct c_vector *vi, *vr, *v, *nv;
-	struct c_instruction instruction;
-	int r;
-
-	if (shader->type != C_PROGRAM_TYPE_VS)
-		return 0;
-	vi = c_shader_vector_new(shader, C_FILE_INPUT, C_SEMANTIC_VERTEXID, -1);
-	if (vi == NULL)
-		return -ENOMEM;
-	LIST_FOR_EACH_ENTRY_SAFE(v, nv, &shader->files[C_FILE_INPUT].vectors, head) {
-		if (v == vi)
-			continue;
-		vr = c_shader_vector_new(shader, C_FILE_RESOURCE, C_SEMANTIC_GENERIC, -1);
-		if (vr == NULL)
-			return -ENOMEM;
-		memset(&instruction, 0, sizeof(struct c_instruction));
-		instruction.nop = 4;
-		instruction.op[0].opcode = C_OPCODE_VFETCH;
-		instruction.op[1].opcode = C_OPCODE_VFETCH;
-		instruction.op[2].opcode = C_OPCODE_VFETCH;
-		instruction.op[3].opcode = C_OPCODE_VFETCH;
-		instruction.op[0].ninput = 2;
-		instruction.op[1].ninput = 2;
-		instruction.op[2].ninput = 2;
-		instruction.op[3].ninput = 2;
-		instruction.op[0].output.vector = v;
-		instruction.op[1].output.vector = v;
-		instruction.op[2].output.vector = v;
-		instruction.op[3].output.vector = v;
-		instruction.op[0].input[0].vector = vi;
-		instruction.op[0].input[1].vector = vr;
-		instruction.op[1].input[0].vector = vi;
-		instruction.op[1].input[1].vector = vr;
-		instruction.op[2].input[0].vector = vi;
-		instruction.op[2].input[1].vector = vr;
-		instruction.op[3].input[0].vector = vi;
-		instruction.op[3].input[1].vector = vr;
-		instruction.op[0].output.swizzle = C_SWIZZLE_X;
-		instruction.op[1].output.swizzle = C_SWIZZLE_Y;
-		instruction.op[2].output.swizzle = C_SWIZZLE_Z;
-		instruction.op[3].output.swizzle = C_SWIZZLE_W;
-		r = c_node_add_new_instruction_head(&shader->entry, &instruction);
-		if (r)
-			return r;
-		LIST_DEL(&v->head);
-		shader->files[C_FILE_INPUT].nvectors--;
-		LIST_ADDTAIL(&v->head, &shader->files[C_FILE_TEMPORARY].vectors);
-		shader->files[C_FILE_TEMPORARY].nvectors++;
-		v->file = C_FILE_TEMPORARY;
-	}
-	return 0;
-}
-
-void r600_shader_cleanup(struct r600_shader *rshader)
-{
-	struct r600_shader_node *n, *nn;
-	struct r600_shader_vfetch *vf, *nvf;
-	struct r600_shader_alu *alu, *nalu;
-	int i;
-
-	if (rshader == NULL)
-		return;
-	if (rshader->gpr) {
-		for (i = 0; i < rshader->nvector; i++) {
-			free(rshader->gpr[i]);
-		}
-		free(rshader->gpr);
-		rshader->gpr = NULL;
-	}
-	LIST_FOR_EACH_ENTRY_SAFE(n, nn, &rshader->nodes, head) {
-		LIST_DEL(&n->head);
-		LIST_FOR_EACH_ENTRY_SAFE(vf, nvf, &n->vfetch, head) {
-			LIST_DEL(&vf->head);
-			free(vf);
-		}
-		LIST_FOR_EACH_ENTRY_SAFE(alu, nalu, &n->alu,  head) {
-			LIST_DEL(&alu->head);
-			free(alu);
-		}
-		free(n);
-	}
-	free(rshader->bcode);
-	return;
-}
-
-int r600_shader_vfetch_bytecode(struct r600_shader *rshader,
-				struct r600_shader_node *rnode,
-				struct r600_shader_vfetch *vfetch,
-				unsigned *cid)
-{
-	unsigned id = *cid;
-
-	vfetch->cf_addr = id;
-	rshader->bcode[id++] = S_SQ_VTX_WORD0_BUFFER_ID(vfetch->src[1].sel) |
-				S_SQ_VTX_WORD0_SRC_GPR(vfetch->src[0].sel) |
-				S_SQ_VTX_WORD0_SRC_SEL_X(vfetch->src[0].sel) |
-				S_SQ_VTX_WORD0_MEGA_FETCH_COUNT(0x1F);
-	rshader->bcode[id++] = S_SQ_VTX_WORD1_DST_SEL_X(vfetch->dst[0].chan) |
-				S_SQ_VTX_WORD1_DST_SEL_Y(vfetch->dst[1].chan) |
-				S_SQ_VTX_WORD1_DST_SEL_Z(vfetch->dst[2].chan) |
-				S_SQ_VTX_WORD1_DST_SEL_W(vfetch->dst[3].chan) |
-				S_SQ_VTX_WORD1_USE_CONST_FIELDS(1) |
-				S_SQ_VTX_WORD1_GPR_DST_GPR(vfetch->dst[0].sel);
-	rshader->bcode[id++] = S_SQ_VTX_WORD2_MEGA_FETCH(1);
-	rshader->bcode[id++] = 0;
-	*cid = id;
-	return 0;
-}
-
-int r600_shader_update(struct r600_shader *rshader, enum pipe_format *resource_format)
-{
-	struct r600_shader_node *rnode;
-	struct r600_shader_vfetch *vfetch;
-	unsigned i;
-
-	memcpy(rshader->resource_format, resource_format,
-		rshader->nresource * sizeof(enum pipe_format));
-	LIST_FOR_EACH_ENTRY(rnode, &rshader->nodes, head) {
-		LIST_FOR_EACH_ENTRY(vfetch, &rnode->vfetch, head) {
-			const struct util_format_description *desc;
-			i = vfetch->cf_addr + 1;
-			rshader->bcode[i] &= C_SQ_VTX_WORD1_DST_SEL_X;
-			rshader->bcode[i] &= C_SQ_VTX_WORD1_DST_SEL_Y;
-			rshader->bcode[i] &= C_SQ_VTX_WORD1_DST_SEL_Z;
-			rshader->bcode[i] &= C_SQ_VTX_WORD1_DST_SEL_W;
-			desc = util_format_description(resource_format[vfetch->src[1].sel]);
-			if (desc == NULL) {
-				fprintf(stderr, "%s unknown format %d\n", __func__, resource_format[vfetch->src[1].sel]);
-				continue;
-			}
-			/* WARNING so far TGSI swizzle match R600 ones */
-			rshader->bcode[i] |= S_SQ_VTX_WORD1_DST_SEL_X(desc->swizzle[0]);
-			rshader->bcode[i] |= S_SQ_VTX_WORD1_DST_SEL_Y(desc->swizzle[1]);
-			rshader->bcode[i] |= S_SQ_VTX_WORD1_DST_SEL_Z(desc->swizzle[2]);
-			rshader->bcode[i] |= S_SQ_VTX_WORD1_DST_SEL_W(desc->swizzle[3]);
-		}
-	}
-	return 0;
-}
-
-int r600_shader_register(struct r600_shader *rshader)
-{
-	struct c_vector *v, *nv;
-	unsigned tid, cid, rid, i;
-
-	rshader->nvector = rshader->cshader.nvectors;
-	rshader->gpr = calloc(rshader->nvector, sizeof(void*));
-	if (rshader->gpr == NULL)
-		return -ENOMEM;
-	tid = 0;
-	cid = 0;
-	rid = 0;
-	/* alloc input first */
-	LIST_FOR_EACH_ENTRY(v, &rshader->cshader.files[C_FILE_INPUT].vectors, head) {
-		nv = c_vector_new();
-		if (nv == NULL) {
-			return -ENOMEM;
-		}
-		memcpy(nv, v, sizeof(struct c_vector));
-		nv->id = tid++;
-		rshader->gpr[v->id] = nv;
-	}
-	for (i = 0; i < C_FILE_COUNT; i++) {
-		if (i == C_FILE_INPUT || i == C_FILE_IMMEDIATE)
-			continue;
-		LIST_FOR_EACH_ENTRY(v, &rshader->cshader.files[i].vectors, head) {
-			switch (v->file) {
-			case C_FILE_OUTPUT:
-			case C_FILE_TEMPORARY:
-				nv = c_vector_new();
-				if (nv == NULL) {
-					return -ENOMEM;
-				}
-				memcpy(nv, v, sizeof(struct c_vector));
-				nv->id = tid++;
-				rshader->gpr[v->id] = nv;
-				break;
-			case C_FILE_CONSTANT:
-				nv = c_vector_new();
-				if (nv == NULL) {
-					return -ENOMEM;
-				}
-				memcpy(nv, v, sizeof(struct c_vector));
-				nv->id = (cid++) + 256;
-				rshader->gpr[v->id] = nv;
-				break;
-			case C_FILE_RESOURCE:
-				nv = c_vector_new();
-				if (nv == NULL) {
-					return -ENOMEM;
-				}
-				memcpy(nv, v, sizeof(struct c_vector));
-				nv->id = (rid++);
-				rshader->gpr[v->id] = nv;
-				break;
-			default:
-				fprintf(stderr, "%s:%d unsupported file %d\n", __func__, __LINE__, v->file);
-				return -EINVAL;
-			}
-		}
-	}
-	rshader->ngpr = tid;
-	rshader->nconstant = cid;
-	rshader->nresource = rid;
-	return 0;
-}
-
-int r600_shader_find_gpr(struct r600_shader *rshader, struct c_vector *v, unsigned swizzle,
-			struct r600_shader_operand *operand)
-{
-	struct c_vector *tmp;
-
-	/* Values [0,127] correspond to GPR[0..127]. 
-	 * Values [256,511] correspond to cfile constants c[0..255]. 
-	 * Other special values are shown in the list below.
-	 * 248	SQ_ALU_SRC_0: special constant 0.0.
-	 * 249	SQ_ALU_SRC_1: special constant 1.0 float.
-	 * 250	SQ_ALU_SRC_1_INT: special constant 1 integer.
-	 * 251	SQ_ALU_SRC_M_1_INT: special constant -1 integer.
-	 * 252	SQ_ALU_SRC_0_5: special constant 0.5 float.
-	 * 253	SQ_ALU_SRC_LITERAL: literal constant.
-	 * 254	SQ_ALU_SRC_PV: previous vector result.
-	 * 255	SQ_ALU_SRC_PS: previous scalar result.
-	 */
-	operand->vector = v;
-	operand->sel = 248;
-	operand->chan = 0;
-	operand->neg = 0;
-	operand->abs = 0;
-	if (v == NULL)
-		return 0;
-	if (v->file == C_FILE_IMMEDIATE) {
-		operand->sel = 253;
-	} else {
-		tmp = rshader->gpr[v->id];
-		if (tmp == NULL) {
-			fprintf(stderr, "%s %d unknown register\n", __FILE__, __LINE__);
-			return -EINVAL;
-		}
-		operand->sel = tmp->id;
-	}
-	operand->chan = swizzle;
-	switch (swizzle) {
-	case C_SWIZZLE_X:
-	case C_SWIZZLE_Y:
-	case C_SWIZZLE_Z:
-	case C_SWIZZLE_W:
-		break;
-	case C_SWIZZLE_0:
-		operand->sel = 248;
-		operand->chan = 0;
-		break;
-	case C_SWIZZLE_1:
-		operand->sel = 249;
-		operand->chan = 0;
-		break;
-	default:
-		fprintf(stderr, "%s %d invalid swizzle %d\n", __FILE__, __LINE__, swizzle);
-		return -EINVAL;
-	}
-	return 0;
-}
-
-static struct r600_shader_node *r600_shader_new_node(struct r600_shader *rshader, struct c_node *node)
-{
-	struct r600_shader_node *rnode;
-
-	rnode = CALLOC_STRUCT(r600_shader_node);
-	if (rnode == NULL)
-		return NULL;
-	rnode->node = node;
-	LIST_INITHEAD(&rnode->vfetch);
-fprintf(stderr, "------------------------ new node (%p %p)\n", &rnode->vfetch, rnode->vfetch.next);
-	LIST_INITHEAD(&rnode->alu);
-	LIST_ADDTAIL(&rnode->head, &rshader->nodes);
-	return rnode;
-}
-
-static int r600_shader_add_vfetch(struct r600_shader *rshader,
-				struct r600_shader_node *node,
-				struct c_instruction *instruction)
-{
-	struct r600_shader_vfetch *vfetch;
-	struct r600_shader_node *rnode;
-	int r;
-
-	if (instruction == NULL)
-		return 0;
-	if (instruction->op[0].opcode != C_OPCODE_VFETCH)
-		return 0;
-	if (!LIST_IS_EMPTY(&node->alu)) {
-		rnode = r600_shader_new_node(rshader, node->node);
-		if (rnode == NULL)
-			return -ENOMEM;
-		node = rnode;
-	}
-	vfetch = calloc(1, sizeof(struct r600_shader_vfetch));
-	if (vfetch == NULL)
-		return -ENOMEM;
-	r = r600_shader_find_gpr(rshader, instruction->op[0].output.vector, 0, &vfetch->dst[0]);
-	if (r)
-		return r;
-	r = r600_shader_find_gpr(rshader, instruction->op[0].input[0].vector, 0, &vfetch->src[0]);
-	if (r)
-		return r;
-	r = r600_shader_find_gpr(rshader, instruction->op[0].input[1].vector, 0, &vfetch->src[1]);
-	if (r)
-		return r;
-	vfetch->dst[0].chan = C_SWIZZLE_X;
-	vfetch->dst[1].chan = C_SWIZZLE_Y;
-	vfetch->dst[2].chan = C_SWIZZLE_Z;
-	vfetch->dst[3].chan = C_SWIZZLE_W;
-	LIST_ADDTAIL(&vfetch->head, &node->vfetch);
-	node->nslot += 2;
-	return 0;
-}
-
-static int r600_node_translate(struct r600_shader *rshader, struct c_node *node)
-{
-	struct c_instruction *instruction;
-	struct r600_shader_node *rnode;
-	int r;
-
-	rnode = r600_shader_new_node(rshader, node);
-	if (rnode == NULL)
-		return -ENOMEM;
-	LIST_FOR_EACH_ENTRY(instruction, &node->insts, head) {
-		switch (instruction->op[0].opcode) {
-		case C_OPCODE_VFETCH:
-			r = r600_shader_add_vfetch(rshader, rnode, instruction);
-			if (r) {
-				fprintf(stderr, "%s %d vfetch failed\n", __func__, __LINE__);
-				return r;
-			}
-			break;
-		default:
-			r = r600_shader_alu_translate(rshader, rnode, instruction);
-			if (r) {
-				fprintf(stderr, "%s %d alu failed\n", __func__, __LINE__);
-				return r;
-			}
-			break;
-		}
-	}
-	return 0;
-}
-
-int r600_shader_translate_rec(struct r600_shader *rshader, struct c_node *node)
-{
-	struct c_node_link *link;
-	int r;
-
-	if (node->opcode == C_OPCODE_END)
-		return 0;
-	r = r600_node_translate(rshader, node);
-	if (r)
-		return r;
-	LIST_FOR_EACH_ENTRY(link, &node->childs, head) {
-		r = r600_shader_translate_rec(rshader, link->node);
-		if (r)
-			return r;
-	}
-	return 0;
-}
-
-static struct r600_shader_alu *r600_shader_insert_alu(struct r600_shader *rshader, struct r600_shader_node *node)
-{
-	struct r600_shader_alu *alu;
-
-	alu = CALLOC_STRUCT(r600_shader_alu);
-	if (alu == NULL)
-		return NULL;
-	alu->alu[0].inst = INST_NOP;
-	alu->alu[1].inst = INST_NOP;
-	alu->alu[2].inst = INST_NOP;
-	alu->alu[3].inst = INST_NOP;
-	alu->alu[4].inst = INST_NOP;
-	alu->alu[0].opcode = V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP;
-	alu->alu[1].opcode = V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP;
-	alu->alu[2].opcode = V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP;
-	alu->alu[3].opcode = V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP;
-	alu->alu[4].opcode = V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP;
-	alu->alu[1].dst.chan = 1;
-	alu->alu[2].dst.chan = 2;
-	alu->alu[3].dst.chan = 3;
-	LIST_ADDTAIL(&alu->head, &node->alu);
-	return alu;
-}
-
-static int r600_shader_alu_translate(struct r600_shader *rshader,
-					struct r600_shader_node *node,
-					struct c_instruction *instruction)
-{
-	struct r600_shader_node *rnode;
-	struct r600_shader_alu *alu;
-	int i, j, r, litteral_lastcomp = -1;
-
-	if (!LIST_IS_EMPTY(&node->vfetch)) {
-fprintf(stderr, "------------------------ add node (%p %p)\n", &node->vfetch, node->vfetch.next);
-		rnode = r600_shader_new_node(rshader, node->node);
-		if (rnode == NULL) {
-			fprintf(stderr, "%s %d new node failed\n", __func__, __LINE__);
-			return -ENOMEM;
-		}
-		node = rnode;
-	}
-
-	/* initialize alu */
-	alu = r600_shader_insert_alu(rshader, node);
-
-	/* check special operation like lit */
-
-	/* go through operation */
-	for (i = 0; i < instruction->nop; i++) {
-		struct r600_alu_instruction *ainfo = &r600_alu_instruction[instruction->op[i].opcode];
-		struct r600_instruction_info *iinfo = &r600_instruction_info[ainfo->instruction];
-		unsigned comp;
-
-		/* check that output is a valid component */
-		comp = instruction->op[i].output.swizzle;
-		switch (comp) {
-		case C_SWIZZLE_X:
-		case C_SWIZZLE_Y:
-		case C_SWIZZLE_Z:
-		case C_SWIZZLE_W:
-			break;
-		case C_SWIZZLE_0:
-		case C_SWIZZLE_1:
-		default:
-			fprintf(stderr, "%s %d invalid output %d\n", __func__, __LINE__, comp);
-			return -EINVAL;
-		}
-		alu->alu[comp].inst = ainfo->instruction;
-		alu->alu[comp].opcode = iinfo->opcode;
-		alu->alu[comp].is_op3 = iinfo->is_op3;
-		for (j = 0; j < instruction->op[i].ninput; j++) {
-			r = r600_shader_find_gpr(rshader, instruction->op[i].input[j].vector,
-					instruction->op[i].input[j].swizzle, &alu->alu[comp].src[j]);
-			if (r) {
-				fprintf(stderr, "%s %d register failed\n", __FILE__, __LINE__);
-				return r;
-			}
-			if (instruction->op[i].input[j].vector->file == C_FILE_IMMEDIATE) {
-				r = instruction->op[i].input[j].swizzle;
-				switch (r) {
-				case C_SWIZZLE_X:
-				case C_SWIZZLE_Y:
-				case C_SWIZZLE_Z:
-				case C_SWIZZLE_W:
-					break;
-				case C_SWIZZLE_0:
-				case C_SWIZZLE_1:
-				default:
-					fprintf(stderr, "%s %d invalid input\n", __func__, __LINE__);
-					return -EINVAL;
-				}
-				alu->literal[r] = instruction->op[i].input[j].vector->channel[r]->value;
-				if (r > litteral_lastcomp) {
-					litteral_lastcomp = r;
-				}
-			}
-		}
-		r = r600_shader_find_gpr(rshader, instruction->op[i].output.vector,
-				instruction->op[i].output.swizzle, &alu->alu[comp].dst);
-		if (r) {
-			fprintf(stderr, "%s %d register failed\n", __FILE__, __LINE__);
-			return r;
-		}
-	}
-	switch (litteral_lastcomp) {
-	case 0:
-	case 1:
-		alu->nliteral = 2;
-		break;
-	case 2:
-	case 3:
-		alu->nliteral = 4;
-		break;
-	case -1:
-	default:
-		break;
-	}
-	for (i = 4; i >= 0; i--) {
-		if (alu->alu[i].inst != INST_NOP) {
-			alu->alu[i].last = 1;
-			alu->nalu = i + 1;
-			break;
-		}
-	}
-	return 0;
-}
-
-void r600_shader_node_place(struct r600_shader *rshader)
-{
-	struct r600_shader_node *node, *nnode;
-	struct r600_shader_alu *alu, *nalu;
-	struct r600_shader_vfetch *vfetch, *nvfetch;
-	unsigned cf_id = 0, cf_addr = 0;
-
-	rshader->ncf = 0;
-	rshader->nslot = 0;
-	LIST_FOR_EACH_ENTRY_SAFE(node, nnode, &rshader->nodes, head) {
-		LIST_FOR_EACH_ENTRY_SAFE(alu, nalu, &node->alu, head) {
-			node->nslot += alu->nalu;
-			node->nslot += alu->nliteral >> 1;
-		}
-		node->nfetch = 0;
-		LIST_FOR_EACH_ENTRY_SAFE(vfetch, nvfetch, &node->vfetch, head) {
-			node->nslot += 2;
-			node->nfetch += 1;
-		}
-		if (!LIST_IS_EMPTY(&node->vfetch)) {
-			/* fetch node need to be 16 bytes aligned*/
-			cf_addr += 1;
-			cf_addr &= 0xFFFFFFFEUL;
-		}
-		node->cf_id = cf_id;
-		node->cf_addr = cf_addr;
-		cf_id += 2;
-		cf_addr += node->nslot * 2;
-		rshader->ncf++;
-	}
-	rshader->nslot = cf_addr;
-	LIST_FOR_EACH_ENTRY_SAFE(node, nnode, &rshader->nodes, head) {
-		node->cf_addr += cf_id * 2;
-	}
-	rshader->ncf += rshader->cshader.files[C_FILE_OUTPUT].nvectors;
-	rshader->ndw = rshader->ncf * 2 + rshader->nslot * 2;
-}
-
-int r600_shader_legalize(struct r600_shader *rshader)
-{
-	return 0;
-}
-
-
-static int r600_cshader_legalize_rec(struct c_shader *shader, struct c_node *node)
-{
-	struct c_node_link *link;
-	struct c_instruction *i, *n;
-	struct c_operand operand;
-	unsigned k, inst;
-	int r;
-
-	LIST_FOR_EACH_ENTRY(i, &node->insts, head) {
-		for (k = 0; k < i->nop; k++) {
-			switch (i->op[k].opcode) {
-			case C_OPCODE_SLT:
-				i->op[k].opcode = C_OPCODE_SGT;
-				memcpy(&operand, &i->op[k].input[0], sizeof(struct c_operand));
-				memcpy(&i->op[k].input[0], &i->op[k].input[1], sizeof(struct c_operand));
-				memcpy(&i->op[k].input[1], &operand, sizeof(struct c_operand));
-				break;
-			default:
-				break;
-			}
-			inst = r600_alu_instruction[i->op[k].opcode].instruction;
-			if (r600_instruction_info[inst].is_trans && k < (i->nop -1)) {
-				/* split trans opcode */
-				n = CALLOC_STRUCT(c_instruction);
-				if (n == NULL)
-					return -ENOMEM;
-				for (n->nop = 0, k = k + 1; k < i->nop; k++, n->nop++) {
-					memcpy(&n->op[n->nop - 0], &i->op[k], sizeof(struct c_op));
-				}
-				i->nop -= n->nop;
-				LIST_ADD(&n->head, &i->head);
-			}
-		}
-	}
-	LIST_FOR_EACH_ENTRY(link, &node->childs, head) {
-		r = r600_cshader_legalize_rec(shader, link->node);
-		if (r) {
-			return r;
-		}
-	}
-	return 0;
-}
-
-int r600_cshader_legalize(struct c_shader *shader)
-{
-	return r600_cshader_legalize_rec(shader, &shader->entry);
-}
-
-
-struct r600_instruction_info r600_instruction_info[] = {
-	{INST_ADD,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD,			0, 0},
-	{INST_MUL,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL,			0, 0},
-	{INST_MUL_IEEE,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL_IEEE,		0, 0},
-	{INST_MAX,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MAX,			0, 0},
-	{INST_MIN,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MIN,			0, 0},
-	{INST_MAX_DX10,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MAX_DX10,		0, 0},
-	{INST_MIN_DX10,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MIN_DX10,		0, 0},
-	{INST_SETE,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETE,			0, 0},
-	{INST_SETGT,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGT,			0, 0},
-	{INST_SETGE,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGE,			0, 0},
-	{INST_SETNE,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETNE,			0, 0},
-	{INST_SETE_DX10,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETE_DX10,		0, 0},
-	{INST_SETGT_DX10,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGT_DX10,		0, 0},
-	{INST_SETGE_DX10,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGE_DX10,		0, 0},
-	{INST_SETNE_DX10,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETNE_DX10,		0, 0},
-	{INST_FRACT,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FRACT,			0, 0},
-	{INST_TRUNC,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_TRUNC,			0, 0},
-	{INST_CEIL,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_CEIL,			0, 0},
-	{INST_RNDNE,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RNDNE,			0, 0},
-	{INST_FLOOR,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLOOR,			0, 0},
-	{INST_MOVA,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOVA,			0, 0},
-	{INST_MOVA_FLOOR,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOVA_FLOOR,		0, 0},
-	{INST_MOVA_INT,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOVA_INT,		0, 0},
-	{INST_MOV,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV,			0, 0},
-	{INST_NOP,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP,			0, 0},
-	{INST_PRED_SETGT_UINT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETGT_UINT,		0, 0},
-	{INST_PRED_SETGE_UINT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETGE_UINT,		0, 0},
-	{INST_PRED_SETE,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETE,		0, 0},
-	{INST_PRED_SETGT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETGT,		0, 0},
-	{INST_PRED_SETGE,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETGE,		0, 0},
-	{INST_PRED_SETNE,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETNE,		0, 0},
-	{INST_PRED_SET_INV,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SET_INV,		0, 0},
-	{INST_PRED_SET_POP,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SET_POP,		0, 0},
-	{INST_PRED_SET_CLR,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SET_CLR,		0, 0},
-	{INST_PRED_SET_RESTORE,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SET_RESTORE,	0, 0},
-	{INST_PRED_SETE_PUSH,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETE_PUSH,		0, 0},
-	{INST_PRED_SETGT_PUSH,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETGT_PUSH,		0, 0},
-	{INST_PRED_SETGE_PUSH,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETGE_PUSH,		0, 0},
-	{INST_PRED_SETNE_PUSH,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETNE_PUSH,		0, 0},
-	{INST_KILLE,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLE,			0, 0},
-	{INST_KILLGT,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLGT,			0, 0},
-	{INST_KILLGE,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLGE,			0, 0},
-	{INST_KILLNE,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLNE,			0, 0},
-	{INST_AND_INT,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_AND_INT,			0, 0},
-	{INST_OR_INT,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_OR_INT,			0, 0},
-	{INST_XOR_INT,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_XOR_INT,			0, 0},
-	{INST_NOT_INT,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOT_INT,			0, 0},
-	{INST_ADD_INT,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD_INT,			0, 0},
-	{INST_SUB_INT,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SUB_INT,			0, 0},
-	{INST_MAX_INT,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MAX_INT,			0, 0},
-	{INST_MIN_INT,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MIN_INT,			0, 0},
-	{INST_MAX_UINT,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MAX_UINT,		0, 0},
-	{INST_MIN_UINT,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MIN_UINT,		0, 0},
-	{INST_SETE_INT,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETE_INT,		0, 0},
-	{INST_SETGT_INT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGT_INT,		0, 0},
-	{INST_SETGE_INT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGE_INT,		0, 0},
-	{INST_SETNE_INT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETNE_INT,		0, 0},
-	{INST_SETGT_UINT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGT_UINT,		0, 0},
-	{INST_SETGE_UINT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGE_UINT,		0, 0},
-	{INST_KILLGT_UINT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLGT_UINT,		0, 0},
-	{INST_KILLGE_UINT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLGE_UINT,		0, 0},
-	{INST_PRED_SETE_INT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETE_INT,		0, 0},
-	{INST_PRED_SETGT_INT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETGT_INT,		0, 0},
-	{INST_PRED_SETGE_INT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETGE_INT,		0, 0},
-	{INST_PRED_SETNE_INT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETNE_INT,		0, 0},
-	{INST_KILLE_INT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLE_INT,		0, 0},
-	{INST_KILLGT_INT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLGT_INT,		0, 0},
-	{INST_KILLGE_INT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLGE_INT,		0, 0},
-	{INST_KILLNE_INT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLNE_INT,		0, 0},
-	{INST_PRED_SETE_PUSH_INT,	V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETE_PUSH_INT,	0, 0},
-	{INST_PRED_SETGT_PUSH_INT,	V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETGT_PUSH_INT,	0, 0},
-	{INST_PRED_SETGE_PUSH_INT,	V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETGE_PUSH_INT,	0, 0},
-	{INST_PRED_SETNE_PUSH_INT,	V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETNE_PUSH_INT,	0, 0},
-	{INST_PRED_SETLT_PUSH_INT,	V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETLT_PUSH_INT,	0, 0},
-	{INST_PRED_SETLE_PUSH_INT,	V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETLE_PUSH_INT,	0, 0},
-	{INST_DOT4,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4,			0, 0},
-	{INST_DOT4_IEEE,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4_IEEE,		0, 0},
-	{INST_CUBE,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_CUBE,			0, 0},
-	{INST_MAX4,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MAX4,			0, 0},
-	{INST_MOVA_GPR_INT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOVA_GPR_INT,		0, 0},
-	{INST_EXP_IEEE,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE,		1, 0},
-	{INST_LOG_CLAMPED,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_CLAMPED,		1, 0},
-	{INST_LOG_IEEE,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE,		1, 0},
-	{INST_RECIP_CLAMPED,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_CLAMPED,		1, 0},
-	{INST_RECIP_FF,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_FF,		1, 0},
-	{INST_RECIP_IEEE,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE,		1, 0},
-	{INST_RECIPSQRT_CLAMPED,	V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIPSQRT_CLAMPED,	1, 0},
-	{INST_RECIPSQRT_FF,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIPSQRT_FF,		1, 0},
-	{INST_RECIPSQRT_IEEE,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIPSQRT_IEEE,		1, 0},
-	{INST_SQRT_IEEE,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SQRT_IEEE,		1, 0},
-	{INST_FLT_TO_INT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLT_TO_INT,		1, 0},
-	{INST_INT_TO_FLT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_INT_TO_FLT,		1, 0},
-	{INST_UINT_TO_FLT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_UINT_TO_FLT,		1, 0},
-	{INST_SIN,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SIN,			1, 0},
-	{INST_COS,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_COS,			1, 0},
-	{INST_ASHR_INT,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ASHR_INT,		1, 0},
-	{INST_LSHR_INT,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LSHR_INT,		1, 0},
-	{INST_LSHL_INT,			V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LSHL_INT,		1, 0},
-	{INST_MULLO_INT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MULLO_INT,		1, 0},
-	{INST_MULHI_INT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MULHI_INT,		1, 0},
-	{INST_MULLO_UINT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MULLO_UINT,		1, 0},
-	{INST_MULHI_UINT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MULHI_UINT,		1, 0},
-	{INST_RECIP_INT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_INT,		1, 0},
-	{INST_RECIP_UINT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_UINT,		1, 0},
-	{INST_FLT_TO_UINT,		V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLT_TO_UINT,		1, 0},
-	{INST_MUL_LIT,			V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MUL_LIT,			1, 1},
-	{INST_MUL_LIT_M2,		V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MUL_LIT_M2,		1, 1},
-	{INST_MUL_LIT_M4,		V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MUL_LIT_M4,		1, 1},
-	{INST_MUL_LIT_D2,		V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MUL_LIT_D2,		1, 1},
-	{INST_MULADD,			V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD,			0, 1},
-	{INST_MULADD_M2,		V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD_M2,		0, 1},
-	{INST_MULADD_M4,		V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD_M4,		0, 1},
-	{INST_MULADD_D2,		V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD_D2,		0, 1},
-	{INST_MULADD_IEEE,		V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD_IEEE,		0, 1},
-	{INST_MULADD_IEEE_M2,		V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD_IEEE_M2,		0, 1},
-	{INST_MULADD_IEEE_M4,		V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD_IEEE_M4,		0, 1},
-	{INST_MULADD_IEEE_D2,		V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD_IEEE_D2,		0, 1},
-	{INST_CNDE,			V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_CNDE,			0, 1},
-	{INST_CNDGT,			V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_CNDGT,			0, 1},
-	{INST_CNDGE,			V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_CNDGE,			0, 1},
-	{INST_CNDE_INT,			V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_CNDE_INT,		0, 1},
-	{INST_CNDGT_INT,		V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_CNDGT_INT,		0, 1},
-	{INST_CNDGE_INT,		V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_CNDGE_INT,		0, 1},
-};
-
-struct r600_alu_instruction r600_alu_instruction[C_OPCODE_LAST] = {
-	{C_OPCODE_NOP,		INST_NOP},
-	{C_OPCODE_MOV,		INST_MOV},
-	{C_OPCODE_LIT,		INST_NOP},
-	{C_OPCODE_RCP,		INST_RECIP_IEEE},
-	{C_OPCODE_RSQ,		INST_RECIPSQRT_IEEE},
-	{C_OPCODE_EXP,		INST_EXP_IEEE},
-	{C_OPCODE_LOG,		INST_LOG_IEEE},
-	{C_OPCODE_MUL,		INST_MUL},
-	{C_OPCODE_ADD,		INST_ADD},
-	{C_OPCODE_DP3,		INST_DOT4},
-	{C_OPCODE_DP4,		INST_DOT4},
-	{C_OPCODE_DST,		INST_NOP},
-	{C_OPCODE_MIN,		INST_MIN},
-	{C_OPCODE_MAX,		INST_MAX},
-	{C_OPCODE_SLT,		INST_NOP},
-	{C_OPCODE_SGE,		INST_NOP},
-	{C_OPCODE_MAD,		INST_MULADD},
-	{C_OPCODE_SUB,		INST_COUNT},
-	{C_OPCODE_LRP,		INST_NOP},
-	{C_OPCODE_CND,		INST_NOP},
-	{20,			INST_NOP},
-	{C_OPCODE_DP2A,		INST_NOP},
-	{22,			INST_NOP},
-	{23,			INST_NOP},
-	{C_OPCODE_FRC,		INST_NOP},
-	{C_OPCODE_CLAMP,	INST_NOP},
-	{C_OPCODE_FLR,		INST_NOP},
-	{C_OPCODE_ROUND,	INST_NOP},
-	{C_OPCODE_EX2,		INST_NOP},
-	{C_OPCODE_LG2,		INST_NOP},
-	{C_OPCODE_POW,		INST_NOP},
-	{C_OPCODE_XPD,		INST_NOP},
-	{32,			INST_NOP},
-	{C_OPCODE_ABS,		INST_COUNT},
-	{C_OPCODE_RCC,		INST_NOP},
-	{C_OPCODE_DPH,		INST_NOP},
-	{C_OPCODE_COS,		INST_COS},
-	{C_OPCODE_DDX,		INST_NOP},
-	{C_OPCODE_DDY,		INST_NOP},
-	{C_OPCODE_KILP,		INST_NOP},
-	{C_OPCODE_PK2H,		INST_NOP},
-	{C_OPCODE_PK2US,	INST_NOP},
-	{C_OPCODE_PK4B,		INST_NOP},
-	{C_OPCODE_PK4UB,	INST_NOP},
-	{C_OPCODE_RFL,		INST_NOP},
-	{C_OPCODE_SEQ,		INST_NOP},
-	{C_OPCODE_SFL,		INST_NOP},
-	{C_OPCODE_SGT,		INST_SETGT},
-	{C_OPCODE_SIN,		INST_SIN},
-	{C_OPCODE_SLE,		INST_NOP},
-	{C_OPCODE_SNE,		INST_NOP},
-	{C_OPCODE_STR,		INST_NOP},
-	{C_OPCODE_TEX,		INST_NOP},
-	{C_OPCODE_TXD,		INST_NOP},
-	{C_OPCODE_TXP,		INST_NOP},
-	{C_OPCODE_UP2H,		INST_NOP},
-	{C_OPCODE_UP2US,	INST_NOP},
-	{C_OPCODE_UP4B,		INST_NOP},
-	{C_OPCODE_UP4UB,	INST_NOP},
-	{C_OPCODE_X2D,		INST_NOP},
-	{C_OPCODE_ARA,		INST_NOP},
-	{C_OPCODE_ARR,		INST_NOP},
-	{C_OPCODE_BRA,		INST_NOP},
-	{C_OPCODE_CAL,		INST_NOP},
-	{C_OPCODE_RET,		INST_NOP},
-	{C_OPCODE_SSG,		INST_NOP},
-	{C_OPCODE_CMP,		INST_NOP},
-	{C_OPCODE_SCS,		INST_NOP},
-	{C_OPCODE_TXB,		INST_NOP},
-	{C_OPCODE_NRM,		INST_NOP},
-	{C_OPCODE_DIV,		INST_NOP},
-	{C_OPCODE_DP2,		INST_NOP},
-	{C_OPCODE_TXL,		INST_NOP},
-	{C_OPCODE_BRK,		INST_NOP},
-	{C_OPCODE_IF,		INST_NOP},
-	{C_OPCODE_BGNFOR,	INST_NOP},
-	{C_OPCODE_REP,		INST_NOP},
-	{C_OPCODE_ELSE,		INST_NOP},
-	{C_OPCODE_ENDIF,	INST_NOP},
-	{C_OPCODE_ENDFOR,	INST_NOP},
-	{C_OPCODE_ENDREP,	INST_NOP},
-	{C_OPCODE_PUSHA,	INST_NOP},
-	{C_OPCODE_POPA,		INST_NOP},
-	{C_OPCODE_CEIL,		INST_NOP},
-	{C_OPCODE_I2F,		INST_NOP},
-	{C_OPCODE_NOT,		INST_NOP},
-	{C_OPCODE_TRUNC,	INST_NOP},
-	{C_OPCODE_SHL,		INST_NOP},
-	{88,			INST_NOP},
-	{C_OPCODE_AND,		INST_NOP},
-	{C_OPCODE_OR,		INST_NOP},
-	{C_OPCODE_MOD,		INST_NOP},
-	{C_OPCODE_XOR,		INST_NOP},
-	{C_OPCODE_SAD,		INST_NOP},
-	{C_OPCODE_TXF,		INST_NOP},
-	{C_OPCODE_TXQ,		INST_NOP},
-	{C_OPCODE_CONT,		INST_NOP},
-	{C_OPCODE_EMIT,		INST_NOP},
-	{C_OPCODE_ENDPRIM,	INST_NOP},
-	{C_OPCODE_BGNLOOP,	INST_NOP},
-	{C_OPCODE_BGNSUB,	INST_NOP},
-	{C_OPCODE_ENDLOOP,	INST_NOP},
-	{C_OPCODE_ENDSUB,	INST_NOP},
-	{103,			INST_NOP},
-	{104,			INST_NOP},
-	{105,			INST_NOP},
-	{106,			INST_NOP},
-	{107,			INST_NOP},
-	{108,			INST_NOP},
-	{109,			INST_NOP},
-	{110,			INST_NOP},
-	{111,			INST_NOP},
-	{C_OPCODE_NRM4,		INST_NOP},
-	{C_OPCODE_CALLNZ,	INST_NOP},
-	{C_OPCODE_IFC,		INST_NOP},
-	{C_OPCODE_BREAKC,	INST_NOP},
-	{C_OPCODE_KIL,		INST_NOP},
-	{C_OPCODE_END,		INST_NOP},
-	{118,			INST_NOP},
-	{C_OPCODE_F2I,		INST_NOP},
-	{C_OPCODE_IDIV,		INST_NOP},
-	{C_OPCODE_IMAX,		INST_NOP},
-	{C_OPCODE_IMIN,		INST_NOP},
-	{C_OPCODE_INEG,		INST_NOP},
-	{C_OPCODE_ISGE,		INST_NOP},
-	{C_OPCODE_ISHR,		INST_NOP},
-	{C_OPCODE_ISLT,		INST_NOP},
-	{C_OPCODE_F2U,		INST_NOP},
-	{C_OPCODE_U2F,		INST_NOP},
-	{C_OPCODE_UADD,		INST_NOP},
-	{C_OPCODE_UDIV,		INST_NOP},
-	{C_OPCODE_UMAD,		INST_NOP},
-	{C_OPCODE_UMAX,		INST_NOP},
-	{C_OPCODE_UMIN,		INST_NOP},
-	{C_OPCODE_UMOD,		INST_NOP},
-	{C_OPCODE_UMUL,		INST_NOP},
-	{C_OPCODE_USEQ,		INST_NOP},
-	{C_OPCODE_USGE,		INST_NOP},
-	{C_OPCODE_USHR,		INST_NOP},
-	{C_OPCODE_USLT,		INST_NOP},
-	{C_OPCODE_USNE,		INST_NOP},
-	{C_OPCODE_SWITCH,	INST_NOP},
-	{C_OPCODE_CASE,		INST_NOP},
-	{C_OPCODE_DEFAULT,	INST_NOP},
-	{C_OPCODE_ENDSWITCH,	INST_NOP},
-	{C_OPCODE_VFETCH,	INST_NOP},
-	{C_OPCODE_ENTRY,	INST_NOP},
-	{C_OPCODE_ARL,		INST_NOP},
-};
-
-
-static int r600_shader_alu_bytecode(struct r600_shader *rshader,
-					struct r600_shader_node *rnode,
-					struct r600_shader_inst *alu,
-					unsigned *cid)
-{
-	unsigned id = *cid;
-
-	/* don't replace gpr by pv or ps for destination register */
-	if (alu->is_op3) {
-		rshader->bcode[id++] = S_SQ_ALU_WORD0_SRC0_SEL(alu->src[0].sel) |
-					S_SQ_ALU_WORD0_SRC0_CHAN(alu->src[0].chan) |
-					S_SQ_ALU_WORD0_SRC1_SEL(alu->src[1].sel) |
-					S_SQ_ALU_WORD0_SRC1_CHAN(alu->src[1].chan) |
-					S_SQ_ALU_WORD0_LAST(alu->last);
-		rshader->bcode[id++] = S_SQ_ALU_WORD1_DST_GPR(alu->dst.sel) |
-					S_SQ_ALU_WORD1_DST_CHAN(alu->dst.chan) |
-					S_SQ_ALU_WORD1_OP3_SRC2_SEL(alu->src[2].sel) |
-					S_SQ_ALU_WORD1_OP3_SRC2_CHAN(alu->src[2].chan) |
-					S_SQ_ALU_WORD1_OP3_SRC2_NEG(alu->src[2].neg) |
-					S_SQ_ALU_WORD1_OP3_ALU_INST(alu->opcode) |
-					S_SQ_ALU_WORD1_BANK_SWIZZLE(0);
-	} else {
-		rshader->bcode[id++] = S_SQ_ALU_WORD0_SRC0_SEL(alu->src[0].sel) |
-					S_SQ_ALU_WORD0_SRC0_CHAN(alu->src[0].chan) |
-					S_SQ_ALU_WORD0_SRC0_NEG(alu->src[0].neg) |
-					S_SQ_ALU_WORD0_SRC1_SEL(alu->src[1].sel) |
-					S_SQ_ALU_WORD0_SRC1_CHAN(alu->src[1].chan) |
-					S_SQ_ALU_WORD0_SRC1_NEG(alu->src[1].neg) |
-					S_SQ_ALU_WORD0_LAST(alu->last);
-		rshader->bcode[id++] = S_SQ_ALU_WORD1_DST_GPR(alu->dst.sel) |
-					S_SQ_ALU_WORD1_DST_CHAN(alu->dst.chan) |
-					S_SQ_ALU_WORD1_OP2_SRC0_ABS(alu->src[0].abs) |
-					S_SQ_ALU_WORD1_OP2_SRC1_ABS(alu->src[1].abs) |
-					S_SQ_ALU_WORD1_OP2_WRITE_MASK(1) |
-					S_SQ_ALU_WORD1_OP2_ALU_INST(alu->opcode) |
-					S_SQ_ALU_WORD1_BANK_SWIZZLE(0);
-	}
-	*cid = id;
-	return 0;
-}
-
-int r6xx_shader_alu_translate(struct r600_shader *rshader,
-			      struct r600_shader_node *rnode,
-			      unsigned *cid)
-{
-	struct r600_shader_alu *alu;
-	unsigned id = *cid;
-	int i;
-	int r = 0;
-	LIST_FOR_EACH_ENTRY(alu, &rnode->alu, head) {
-		for (i = 0; i < alu->nalu; i++) {
-			r = r600_shader_alu_bytecode(rshader, rnode, &alu->alu[i], &id);
-			if (r)
-				goto out;
-		}
-		for (i = 0; i < alu->nliteral; i++) {
-			rshader->bcode[id++] = alu->literal[i];
-		}
-	}
-out:
-	*cid = id;
-	return r;
-}
diff --git a/src/gallium/drivers/r600/r600_compiler_r700.c b/src/gallium/drivers/r600/r600_compiler_r700.c
deleted file mode 100644
index 0b439428662..00000000000
--- a/src/gallium/drivers/r600/r600_compiler_r700.c
+++ /dev/null
@@ -1,233 +0,0 @@
-/*
- * Copyright 2010 Jerome Glisse <[email protected]>
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * on the rights to use, copy, modify, merge, publish, distribute, sub
- * license, and/or sell copies of the Software, and to permit persons to whom
- * the Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
- * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
- * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
- * USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-#include <stdlib.h>
-#include <string.h>
-#include <stdint.h>
-#include <stdio.h>
-#include <errno.h>
-#include "r600_context.h"
-#include "r700_sq.h"
-
-static int r700_shader_cf_node_bytecode(struct r600_shader *rshader,
-					struct r600_shader_node *rnode,
-					unsigned *cid)
-{
-	unsigned id = *cid;
-
-	if (rnode->nfetch) {
-		rshader->bcode[id++] = S_SQ_CF_WORD0_ADDR(rnode->cf_addr >> 1);
-		rshader->bcode[id++] = S_SQ_CF_WORD1_CF_INST(V_SQ_CF_WORD1_SQ_CF_INST_VTX) |
-					S_SQ_CF_WORD1_BARRIER(1) |
-					S_SQ_CF_WORD1_COUNT(rnode->nfetch - 1);
-	} else {
-		rshader->bcode[id++] = S_SQ_CF_ALU_WORD0_ADDR(rnode->cf_addr >> 1);
-		rshader->bcode[id++] = S_SQ_CF_ALU_WORD1_CF_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU) |
-					S_SQ_CF_ALU_WORD1_BARRIER(1) |
-					S_SQ_CF_ALU_WORD1_COUNT(rnode->nslot - 1);
-	}
-	*cid = id;
-	return 0;
-}
-
-static int r700_shader_cf_output_bytecode(struct r600_shader *rshader,
-						struct c_vector *v,
-						unsigned *cid,
-						unsigned end)
-{
-	struct r600_shader_operand out;
-	unsigned id = *cid;
-	int r;
-
-	r = r600_shader_find_gpr(rshader, v, 0, &out);
-	if (r)
-		return r;
-	rshader->bcode[id + 0] = S_SQ_CF_ALLOC_EXPORT_WORD0_RW_GPR(out.sel) |
-				S_SQ_CF_ALLOC_EXPORT_WORD0_ELEM_SIZE(3);
-	rshader->bcode[id + 1] = S_SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_X(0) |
-		S_SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Y(1) |
-		S_SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Z(2) |
-		S_SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_W(3) |
-		S_SQ_CF_ALLOC_EXPORT_WORD1_BARRIER(1) |
-		S_SQ_CF_ALLOC_EXPORT_WORD1_CF_INST(V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT_DONE) |
-		S_SQ_CF_ALLOC_EXPORT_WORD1_END_OF_PROGRAM(end);
-	switch (v->name) {
-	case C_SEMANTIC_POSITION:
-		rshader->bcode[id + 0] |= S_SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE(60) |
-			S_SQ_CF_ALLOC_EXPORT_WORD0_TYPE(V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_POS);
-		break;
-	case C_SEMANTIC_COLOR:
-		if (rshader->cshader.type == C_PROGRAM_TYPE_VS) {
-			rshader->output[rshader->noutput].gpr = out.sel;
-			rshader->output[rshader->noutput].sid = v->sid;
-			rshader->output[rshader->noutput].name = v->name;
-			rshader->bcode[id + 0] |= S_SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE(rshader->noutput++) |
-				S_SQ_CF_ALLOC_EXPORT_WORD0_TYPE(V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PARAM);
-		} else {
-			rshader->bcode[id + 0] |= S_SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE(0) |
-				S_SQ_CF_ALLOC_EXPORT_WORD0_TYPE(V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL);
-		}
-		break;
-	case C_SEMANTIC_GENERIC:
-		rshader->output[rshader->noutput].gpr = out.sel;
-		rshader->output[rshader->noutput].sid = v->sid;
-		rshader->output[rshader->noutput].name = v->name;
-		rshader->bcode[id + 0] |= S_SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE(rshader->noutput++) |
-			S_SQ_CF_ALLOC_EXPORT_WORD0_TYPE(V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PARAM);
-		break;
-	default:
-		fprintf(stderr, "%s:%d unsupported\n", __func__, __LINE__);
-		return -EINVAL;
-	}
-	*cid = id + 2;
-	return 0;
-}
-
-static int r700_shader_alu_bytecode(struct r600_shader *rshader,
-					struct r600_shader_node *rnode,
-					struct r600_shader_inst *alu,
-					unsigned *cid)
-{
-	unsigned id = *cid;
-
-	/* don't replace gpr by pv or ps for destination register */
-	if (alu->is_op3) {
-		rshader->bcode[id++] = S_SQ_ALU_WORD0_SRC0_SEL(alu->src[0].sel) |
-					S_SQ_ALU_WORD0_SRC0_CHAN(alu->src[0].chan) |
-					S_SQ_ALU_WORD0_SRC1_SEL(alu->src[1].sel) |
-					S_SQ_ALU_WORD0_SRC1_CHAN(alu->src[1].chan) |
-					S_SQ_ALU_WORD0_LAST(alu->last);
-		rshader->bcode[id++] = S_SQ_ALU_WORD1_DST_GPR(alu->dst.sel) |
-					S_SQ_ALU_WORD1_DST_CHAN(alu->dst.chan) |
-					S_SQ_ALU_WORD1_OP3_SRC2_SEL(alu->src[2].sel) |
-					S_SQ_ALU_WORD1_OP3_SRC2_CHAN(alu->src[2].chan) |
-					S_SQ_ALU_WORD1_OP3_SRC2_NEG(alu->src[2].neg) |
-					S_SQ_ALU_WORD1_OP3_ALU_INST(alu->opcode) |
-					S_SQ_ALU_WORD1_BANK_SWIZZLE(0);
-	} else {
-		rshader->bcode[id++] = S_SQ_ALU_WORD0_SRC0_SEL(alu->src[0].sel) |
-					S_SQ_ALU_WORD0_SRC0_CHAN(alu->src[0].chan) |
-					S_SQ_ALU_WORD0_SRC0_NEG(alu->src[0].neg) |
-					S_SQ_ALU_WORD0_SRC1_SEL(alu->src[1].sel) |
-					S_SQ_ALU_WORD0_SRC1_CHAN(alu->src[1].chan) |
-					S_SQ_ALU_WORD0_SRC1_NEG(alu->src[1].neg) |
-					S_SQ_ALU_WORD0_LAST(alu->last);
-		rshader->bcode[id++] = S_SQ_ALU_WORD1_DST_GPR(alu->dst.sel) |
-					S_SQ_ALU_WORD1_DST_CHAN(alu->dst.chan) |
-					S_SQ_ALU_WORD1_OP2_SRC0_ABS(alu->src[0].abs) |
-					S_SQ_ALU_WORD1_OP2_SRC1_ABS(alu->src[1].abs) |
-					S_SQ_ALU_WORD1_OP2_WRITE_MASK(1) |
-					S_SQ_ALU_WORD1_OP2_ALU_INST(alu->opcode) |
-					S_SQ_ALU_WORD1_BANK_SWIZZLE(0);
-	}
-	*cid = id;
-	return 0;
-}
-
-static int r700_shader_alu_translate(struct r600_shader *rshader,
-				     struct r600_shader_node *rnode,
-				     unsigned *cid)
-				     
-{
-	struct r600_shader_alu *alu;
-	unsigned id = *cid;
-	int i;
-	int r = 0;
-	LIST_FOR_EACH_ENTRY(alu, &rnode->alu, head) {
-		for (i = 0; i < alu->nalu; i++) {
-			r = r700_shader_alu_bytecode(rshader, rnode, &alu->alu[i], &id);
-			if (r)
-				goto out;
-		}
-		for (i = 0; i < alu->nliteral; i++) {
-			rshader->bcode[id++] = alu->literal[i];
-		}
-	}
- out:
-	*cid = id;
-	return r;
-}
-
-int r700_shader_translate(struct r600_shader *rshader)
-{
-	struct c_shader *shader = &rshader->cshader;
-	struct r600_shader_node *rnode;
-	struct r600_shader_vfetch *vfetch;
-	struct c_vector *v;
-	unsigned id, end;
-	int r;
-
-	r = r600_shader_register(rshader);
-	if (r) {
-		fprintf(stderr, "%s %d register allocation failed\n", __FILE__, __LINE__);
-		return r;
-	}
-	r = r600_shader_translate_rec(rshader, &shader->entry);
-	if (r) {
-		fprintf(stderr, "%s %d translation failed\n", __FILE__, __LINE__);
-		return r;
-	}
-	r = r600_shader_legalize(rshader);
-	if (r) {
-		fprintf(stderr, "%s %d legalize failed\n", __FILE__, __LINE__);
-		return r;
-	}
-	r600_shader_node_place(rshader);
-	rshader->bcode = malloc(rshader->ndw * 4);
-	if (rshader->bcode == NULL)
-		return -ENOMEM;
-	LIST_FOR_EACH_ENTRY(rnode, &rshader->nodes, head) {
-		id = rnode->cf_addr;
-		LIST_FOR_EACH_ENTRY(vfetch, &rnode->vfetch, head) {
-			r = r600_shader_vfetch_bytecode(rshader, rnode, vfetch, &id);
-			if (r)
-				return r;
-		}
-		if (rshader->r6xx_compile)
-			r = r6xx_shader_alu_translate(rshader, rnode, &id);
-		else
-			r = r700_shader_alu_translate(rshader, rnode, &id);
-		if (r)
-			return r;
-	}
-	id = 0;
-	LIST_FOR_EACH_ENTRY(rnode, &rshader->nodes, head) {
-		r = r700_shader_cf_node_bytecode(rshader, rnode, &id);
-		if (r)
-			return r;
-	}
-	LIST_FOR_EACH_ENTRY(v, &rshader->cshader.files[C_FILE_OUTPUT].vectors, head) {
-		end = 0;
-		if (v->head.next == &rshader->cshader.files[C_FILE_OUTPUT].vectors)
-			end = 1;
-		r = r700_shader_cf_output_bytecode(rshader, v, &id, end);
-		if (r)
-			return r;
-	}
-	LIST_FOR_EACH_ENTRY(v, &rshader->cshader.files[C_FILE_INPUT].vectors, head) {
-		rshader->input[rshader->ninput].gpr = rshader->ninput;
-		rshader->input[rshader->ninput].sid = v->sid;
-		rshader->input[rshader->ninput].name = v->name;
-		rshader->ninput++;
-	}
-	return 0;
-}
diff --git a/src/gallium/drivers/r600/r600_compiler_tgsi.c b/src/gallium/drivers/r600/r600_compiler_tgsi.c
deleted file mode 100644
index 172cf154a34..00000000000
--- a/src/gallium/drivers/r600/r600_compiler_tgsi.c
+++ /dev/null
@@ -1,730 +0,0 @@
-/*
- * Copyright 2010 Jerome Glisse <[email protected]>
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * on the rights to use, copy, modify, merge, publish, distribute, sub
- * license, and/or sell copies of the Software, and to permit persons to whom
- * the Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
- * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
- * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
- * USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-#include <stdlib.h>
-#include <string.h>
-#include <stdint.h>
-#include <stdio.h>
-#include <errno.h>
-#include <tgsi/tgsi_parse.h>
-#include <tgsi/tgsi_scan.h>
-#include "r600_shader.h"
-#include "r600_context.h"
-
-struct tgsi_shader {
-	struct c_vector				**v[TGSI_FILE_COUNT];
-	struct tgsi_shader_info			info;
-	struct tgsi_parse_context		parser;
-	const struct tgsi_token			*tokens;
-	struct c_shader				*shader;
-	struct c_node				*node;
-};
-
-static unsigned tgsi_file_to_c_file(unsigned file);
-static unsigned tgsi_sname_to_c_sname(unsigned sname);
-static int tgsi_opcode_to_c_opcode(unsigned opcode, unsigned *copcode);
-
-static int tgsi_shader_init(struct tgsi_shader *ts,
-				const struct tgsi_token *tokens,
-				struct c_shader *shader)
-{
-	int i;
-
-	ts->shader = shader;
-	ts->tokens = tokens;
-	tgsi_scan_shader(ts->tokens, &ts->info);
-	tgsi_parse_init(&ts->parser, ts->tokens);
-	/* initialize to NULL in case of error */
-	for (i = 0; i < C_FILE_COUNT; i++) {
-		ts->v[i] = NULL;
-	}
-	for (i = 0; i < TGSI_FILE_COUNT; i++) {
-		if (ts->info.file_count[i] > 0) {
-			ts->v[i] = calloc(ts->info.file_count[i], sizeof(void*));
-			if (ts->v[i] == NULL) {
-				fprintf(stderr, "%s:%d unsupported %d %d\n", __func__, __LINE__, i, ts->info.file_count[i]);
-				return -ENOMEM;
-			}
-		}
-	}
-	return 0;
-}
-
-static void tgsi_shader_destroy(struct tgsi_shader *ts)
-{
-	int i;
-
-	for (i = 0; i < TGSI_FILE_COUNT; i++) {
-		free(ts->v[i]);
-	}
-	tgsi_parse_free(&ts->parser);
-}
-
-static int ntransform_declaration(struct tgsi_shader *ts)
-{
-	struct tgsi_full_declaration *fd = &ts->parser.FullToken.FullDeclaration;
-	struct c_vector *v;
-	unsigned file;
-	unsigned name;
-	int sid;
-	int i;
-
-	if (fd->Declaration.Dimension) {
-		fprintf(stderr, "%s:%d unsupported\n", __func__, __LINE__);
-		return -EINVAL;
-	}
-	for (i = fd->Range.First ; i <= fd->Range.Last; i++) {
-		sid = i;
-		name = C_SEMANTIC_GENERIC;
-		file = tgsi_file_to_c_file(fd->Declaration.File);
-		if (file == TGSI_FILE_NULL) {
-			fprintf(stderr, "%s:%d unsupported\n", __func__, __LINE__);
-			return -EINVAL;
-		}
-		if (fd->Declaration.Semantic) {
-			name = tgsi_sname_to_c_sname(fd->Semantic.Name);
-			sid = fd->Semantic.Index;
-		}
-		v = c_shader_vector_new(ts->shader, file, name, sid);
-		if (v == NULL) {
-			fprintf(stderr, "%s:%d unsupported\n", __func__, __LINE__);
-			return -ENOMEM;
-		}
-		ts->v[fd->Declaration.File][i] = v;
-	}
-	return 0;
-}
-
-static int ntransform_immediate(struct tgsi_shader *ts)
-{
-	struct tgsi_full_immediate *fd = &ts->parser.FullToken.FullImmediate;
-	struct c_vector *v;
-	unsigned file;
-	unsigned name;
-
-	if (fd->Immediate.DataType != TGSI_IMM_FLOAT32) {
-		fprintf(stderr, "%s:%d unsupported\n", __func__, __LINE__);
-		return -EINVAL;
-	}
-	name = C_SEMANTIC_GENERIC;
-	file = C_FILE_IMMEDIATE;
-	v = c_shader_vector_new(ts->shader, file, name, 0);
-	if (v == NULL) {
-		fprintf(stderr, "%s:%d unsupported\n", __func__, __LINE__);
-			return -ENOMEM;
-	}
-	v->channel[0]->value = fd->u[0].Uint;
-	v->channel[1]->value = fd->u[1].Uint;
-	v->channel[2]->value = fd->u[2].Uint;
-	v->channel[3]->value = fd->u[3].Uint;
-	ts->v[TGSI_FILE_IMMEDIATE][0] = v;
-	return 0;
-}
-
-static int ntransform_instruction(struct tgsi_shader *ts)
-{
-	struct tgsi_full_instruction *fi = &ts->parser.FullToken.FullInstruction;
-	struct c_shader *shader = ts->shader;
-	struct c_instruction instruction;
-	unsigned opcode;
-	int i, j, r;
-
-	if (fi->Instruction.NumDstRegs > 1) {
-		fprintf(stderr, "%s %d unsupported\n", __func__, __LINE__);
-		return -EINVAL;
-	}
-	if (fi->Instruction.Saturate) {
-		fprintf(stderr, "%s %d unsupported\n", __func__, __LINE__);
-		return -EINVAL;
-	}
-	if (fi->Instruction.Predicate) {
-		fprintf(stderr, "%s %d unsupported\n", __func__, __LINE__);
-		return -EINVAL;
-	}
-	if (fi->Instruction.Label) {
-		fprintf(stderr, "%s %d unsupported\n", __func__, __LINE__);
-		return -EINVAL;
-	}
-	if (fi->Instruction.Texture) {
-		fprintf(stderr, "%s %d unsupported\n", __func__, __LINE__);
-		return -EINVAL;
-	}
-	for (i = 0; i < fi->Instruction.NumSrcRegs; i++) {
-		if (fi->Src[i].Register.Indirect ||
-			fi->Src[i].Register.Dimension ||
-			fi->Src[i].Register.Absolute) {
-			fprintf(stderr, "%s %d unsupported\n", __func__, __LINE__);
-			return -EINVAL;
-		}
-	}
-	for (i = 0; i < fi->Instruction.NumDstRegs; i++) {
-		if (fi->Dst[i].Register.Indirect || fi->Dst[i].Register.Dimension) {
-			fprintf(stderr, "%s %d unsupported\n", __func__, __LINE__);
-			return -EINVAL;
-		}
-	}
-	r = tgsi_opcode_to_c_opcode(fi->Instruction.Opcode, &opcode);
-	if (r) {
-		fprintf(stderr, "%s:%d unsupported\n", __func__, __LINE__);
-		return r;
-	}
-	if (opcode == C_OPCODE_END) {
-		return c_node_cfg_link(ts->node, &shader->end);
-	}
-	/* FIXME add flow instruction handling */
-	memset(&instruction, 0, sizeof(struct c_instruction));
-	instruction.nop = 0;
-	for (j = 0; j < 4; j++) {
-		instruction.op[instruction.nop].opcode = opcode;
-		instruction.op[instruction.nop].ninput = fi->Instruction.NumSrcRegs;
-		for (i = 0; i < fi->Instruction.NumSrcRegs; i++) {
-			instruction.op[instruction.nop].input[i].vector = ts->v[fi->Src[i].Register.File][fi->Src[i].Register.Index];
-			switch (j) {
-			case 0:
-				instruction.op[instruction.nop].input[i].swizzle = fi->Src[i].Register.SwizzleX;
-				break;
-			case 1:
-				instruction.op[instruction.nop].input[i].swizzle = fi->Src[i].Register.SwizzleY;
-				break;
-			case 2:
-				instruction.op[instruction.nop].input[i].swizzle = fi->Src[i].Register.SwizzleZ;
-				break;
-			case 3:
-				instruction.op[instruction.nop].input[i].swizzle = fi->Src[i].Register.SwizzleW;
-				break;
-			default:
-				return -EINVAL;
-			}
-		}
-		instruction.op[instruction.nop].output.vector = ts->v[fi->Dst[0].Register.File][fi->Dst[0].Register.Index];
-		switch (j) {
-		case 0:
-			instruction.op[instruction.nop].output.swizzle = (fi->Dst[0].Register.WriteMask & 0x1) ? C_SWIZZLE_X : C_SWIZZLE_D;
-			break;
-		case 1:
-			instruction.op[instruction.nop].output.swizzle = (fi->Dst[0].Register.WriteMask & 0x1) ? C_SWIZZLE_Y : C_SWIZZLE_D;
-			break;
-		case 2:
-			instruction.op[instruction.nop].output.swizzle = (fi->Dst[0].Register.WriteMask & 0x1) ? C_SWIZZLE_Z : C_SWIZZLE_D;
-			break;
-		case 3:
-			instruction.op[instruction.nop].output.swizzle = (fi->Dst[0].Register.WriteMask & 0x1) ? C_SWIZZLE_W : C_SWIZZLE_D;
-			break;
-		default:
-			return -EINVAL;
-		}
-		instruction.nop++;
-	}
-	return c_node_add_new_instruction(ts->node, &instruction);
-}
-
-int c_shader_from_tgsi(struct c_shader *shader, unsigned type,
-			const struct tgsi_token *tokens)
-{
-	struct tgsi_shader ts;
-	int r = 0;
-
-	c_shader_init(shader, type);
-	r = tgsi_shader_init(&ts, tokens, shader);
-	if (r)
-		goto out_err;
-	ts.shader = shader;
-	ts.node = &shader->entry;
-	while (!tgsi_parse_end_of_tokens(&ts.parser)) {
-		tgsi_parse_token(&ts.parser);
-		switch (ts.parser.FullToken.Token.Type) {
-		case TGSI_TOKEN_TYPE_IMMEDIATE:
-			r = ntransform_immediate(&ts);
-			if (r)
-				goto out_err;
-			break;
-		case TGSI_TOKEN_TYPE_DECLARATION:
-			r = ntransform_declaration(&ts);
-			if (r)
-				goto out_err;
-			break;
-		case TGSI_TOKEN_TYPE_INSTRUCTION:
-			r = ntransform_instruction(&ts);
-			if (r)
-				goto out_err;
-			break;
-		default:
-			r = -EINVAL;
-			goto out_err;
-		}
-	}
-	tgsi_shader_destroy(&ts);
-	return 0;
-out_err:
-	c_shader_destroy(shader);
-	tgsi_shader_destroy(&ts);
-	return r;
-}
-
-static unsigned tgsi_file_to_c_file(unsigned file)
-{
-	switch (file) {
-	case TGSI_FILE_CONSTANT:
-		return C_FILE_CONSTANT;
-	case TGSI_FILE_INPUT:
-		return C_FILE_INPUT;
-	case TGSI_FILE_OUTPUT:
-		return C_FILE_OUTPUT;
-	case TGSI_FILE_TEMPORARY:
-		return C_FILE_TEMPORARY;
-	case TGSI_FILE_SAMPLER:
-		return C_FILE_SAMPLER;
-	case TGSI_FILE_ADDRESS:
-		return C_FILE_ADDRESS;
-	case TGSI_FILE_IMMEDIATE:
-		return C_FILE_IMMEDIATE;
-	case TGSI_FILE_PREDICATE:
-		return C_FILE_PREDICATE;
-	case TGSI_FILE_SYSTEM_VALUE:
-		return C_FILE_SYSTEM_VALUE;
-	case TGSI_FILE_NULL:
-		return C_FILE_NULL;
-	default:
-		fprintf(stderr, "%s:%d unsupported file %d\n", __func__, __LINE__, file);
-		return C_FILE_NULL;
-	}
-}
-
-static unsigned tgsi_sname_to_c_sname(unsigned sname)
-{
-	switch (sname) {
-	case TGSI_SEMANTIC_POSITION:
-		return C_SEMANTIC_POSITION;
-	case TGSI_SEMANTIC_COLOR:
-		return C_SEMANTIC_COLOR;
-	case TGSI_SEMANTIC_BCOLOR:
-		return C_SEMANTIC_BCOLOR;
-	case TGSI_SEMANTIC_FOG:
-		return C_SEMANTIC_FOG;
-	case TGSI_SEMANTIC_PSIZE:
-		return C_SEMANTIC_PSIZE;
-	case TGSI_SEMANTIC_GENERIC:
-		return C_SEMANTIC_GENERIC;
-	case TGSI_SEMANTIC_NORMAL:
-		return C_SEMANTIC_NORMAL;
-	case TGSI_SEMANTIC_FACE:
-		return C_SEMANTIC_FACE;
-	case TGSI_SEMANTIC_EDGEFLAG:
-		return C_SEMANTIC_EDGEFLAG;
-	case TGSI_SEMANTIC_PRIMID:
-		return C_SEMANTIC_PRIMID;
-	case TGSI_SEMANTIC_INSTANCEID:
-		return C_SEMANTIC_INSTANCEID;
-	default:
-		return C_SEMANTIC_GENERIC;
-	}
-}
-
-static int tgsi_opcode_to_c_opcode(unsigned opcode, unsigned *copcode)
-{
-	switch (opcode) {
-	case TGSI_OPCODE_MOV:
-		*copcode = C_OPCODE_MOV;
-		return 0;
-	case TGSI_OPCODE_MUL:
-		*copcode = C_OPCODE_MUL;
-		return 0;
-	case TGSI_OPCODE_MAD:
-		*copcode = C_OPCODE_MAD;
-		return 0;
-	case TGSI_OPCODE_END:
-		*copcode = C_OPCODE_END;
-		return 0;
-	case TGSI_OPCODE_ARL:
-		*copcode = C_OPCODE_ARL;
-		return 0;
-	case TGSI_OPCODE_LIT:
-		*copcode = C_OPCODE_LIT;
-		return 0;
-	case TGSI_OPCODE_RCP:
-		*copcode = C_OPCODE_RCP;
-		return 0;
-	case TGSI_OPCODE_RSQ:
-		*copcode = C_OPCODE_RSQ;
-		return 0;
-	case TGSI_OPCODE_EXP:
-		*copcode = C_OPCODE_EXP;
-		return 0;
-	case TGSI_OPCODE_LOG:
-		*copcode = C_OPCODE_LOG;
-		return 0;
-	case TGSI_OPCODE_ADD:
-		*copcode = C_OPCODE_ADD;
-		return 0;
-	case TGSI_OPCODE_DP3:
-		*copcode = C_OPCODE_DP3;
-		return 0;
-	case TGSI_OPCODE_DP4:
-		*copcode = C_OPCODE_DP4;
-		return 0;
-	case TGSI_OPCODE_DST:
-		*copcode = C_OPCODE_DST;
-		return 0;
-	case TGSI_OPCODE_MIN:
-		*copcode = C_OPCODE_MIN;
-		return 0;
-	case TGSI_OPCODE_MAX:
-		*copcode = C_OPCODE_MAX;
-		return 0;
-	case TGSI_OPCODE_SLT:
-		*copcode = C_OPCODE_SLT;
-		return 0;
-	case TGSI_OPCODE_SGE:
-		*copcode = C_OPCODE_SGE;
-		return 0;
-	case TGSI_OPCODE_SUB:
-		*copcode = C_OPCODE_SUB;
-		return 0;
-	case TGSI_OPCODE_LRP:
-		*copcode = C_OPCODE_LRP;
-		return 0;
-	case TGSI_OPCODE_CND:
-		*copcode = C_OPCODE_CND;
-		return 0;
-	case TGSI_OPCODE_DP2A:
-		*copcode = C_OPCODE_DP2A;
-		return 0;
-	case TGSI_OPCODE_FRC:
-		*copcode = C_OPCODE_FRC;
-		return 0;
-	case TGSI_OPCODE_CLAMP:
-		*copcode = C_OPCODE_CLAMP;
-		return 0;
-	case TGSI_OPCODE_FLR:
-		*copcode = C_OPCODE_FLR;
-		return 0;
-	case TGSI_OPCODE_ROUND:
-		*copcode = C_OPCODE_ROUND;
-		return 0;
-	case TGSI_OPCODE_EX2:
-		*copcode = C_OPCODE_EX2;
-		return 0;
-	case TGSI_OPCODE_LG2:
-		*copcode = C_OPCODE_LG2;
-		return 0;
-	case TGSI_OPCODE_POW:
-		*copcode = C_OPCODE_POW;
-		return 0;
-	case TGSI_OPCODE_XPD:
-		*copcode = C_OPCODE_XPD;
-		return 0;
-	case TGSI_OPCODE_ABS:
-		*copcode = C_OPCODE_ABS;
-		return 0;
-	case TGSI_OPCODE_RCC:
-		*copcode = C_OPCODE_RCC;
-		return 0;
-	case TGSI_OPCODE_DPH:
-		*copcode = C_OPCODE_DPH;
-		return 0;
-	case TGSI_OPCODE_COS:
-		*copcode = C_OPCODE_COS;
-		return 0;
-	case TGSI_OPCODE_DDX:
-		*copcode = C_OPCODE_DDX;
-		return 0;
-	case TGSI_OPCODE_DDY:
-		*copcode = C_OPCODE_DDY;
-		return 0;
-	case TGSI_OPCODE_KILP:
-		*copcode = C_OPCODE_KILP;
-		return 0;
-	case TGSI_OPCODE_PK2H:
-		*copcode = C_OPCODE_PK2H;
-		return 0;
-	case TGSI_OPCODE_PK2US:
-		*copcode = C_OPCODE_PK2US;
-		return 0;
-	case TGSI_OPCODE_PK4B:
-		*copcode = C_OPCODE_PK4B;
-		return 0;
-	case TGSI_OPCODE_PK4UB:
-		*copcode = C_OPCODE_PK4UB;
-		return 0;
-	case TGSI_OPCODE_RFL:
-		*copcode = C_OPCODE_RFL;
-		return 0;
-	case TGSI_OPCODE_SEQ:
-		*copcode = C_OPCODE_SEQ;
-		return 0;
-	case TGSI_OPCODE_SFL:
-		*copcode = C_OPCODE_SFL;
-		return 0;
-	case TGSI_OPCODE_SGT:
-		*copcode = C_OPCODE_SGT;
-		return 0;
-	case TGSI_OPCODE_SIN:
-		*copcode = C_OPCODE_SIN;
-		return 0;
-	case TGSI_OPCODE_SLE:
-		*copcode = C_OPCODE_SLE;
-		return 0;
-	case TGSI_OPCODE_SNE:
-		*copcode = C_OPCODE_SNE;
-		return 0;
-	case TGSI_OPCODE_STR:
-		*copcode = C_OPCODE_STR;
-		return 0;
-	case TGSI_OPCODE_TEX:
-		*copcode = C_OPCODE_TEX;
-		return 0;
-	case TGSI_OPCODE_TXD:
-		*copcode = C_OPCODE_TXD;
-		return 0;
-	case TGSI_OPCODE_TXP:
-		*copcode = C_OPCODE_TXP;
-		return 0;
-	case TGSI_OPCODE_UP2H:
-		*copcode = C_OPCODE_UP2H;
-		return 0;
-	case TGSI_OPCODE_UP2US:
-		*copcode = C_OPCODE_UP2US;
-		return 0;
-	case TGSI_OPCODE_UP4B:
-		*copcode = C_OPCODE_UP4B;
-		return 0;
-	case TGSI_OPCODE_UP4UB:
-		*copcode = C_OPCODE_UP4UB;
-		return 0;
-	case TGSI_OPCODE_X2D:
-		*copcode = C_OPCODE_X2D;
-		return 0;
-	case TGSI_OPCODE_ARA:
-		*copcode = C_OPCODE_ARA;
-		return 0;
-	case TGSI_OPCODE_ARR:
-		*copcode = C_OPCODE_ARR;
-		return 0;
-	case TGSI_OPCODE_BRA:
-		*copcode = C_OPCODE_BRA;
-		return 0;
-	case TGSI_OPCODE_CAL:
-		*copcode = C_OPCODE_CAL;
-		return 0;
-	case TGSI_OPCODE_RET:
-		*copcode = C_OPCODE_RET;
-		return 0;
-	case TGSI_OPCODE_SSG:
-		*copcode = C_OPCODE_SSG;
-		return 0;
-	case TGSI_OPCODE_CMP:
-		*copcode = C_OPCODE_CMP;
-		return 0;
-	case TGSI_OPCODE_SCS:
-		*copcode = C_OPCODE_SCS;
-		return 0;
-	case TGSI_OPCODE_TXB:
-		*copcode = C_OPCODE_TXB;
-		return 0;
-	case TGSI_OPCODE_NRM:
-		*copcode = C_OPCODE_NRM;
-		return 0;
-	case TGSI_OPCODE_DIV:
-		*copcode = C_OPCODE_DIV;
-		return 0;
-	case TGSI_OPCODE_DP2:
-		*copcode = C_OPCODE_DP2;
-		return 0;
-	case TGSI_OPCODE_TXL:
-		*copcode = C_OPCODE_TXL;
-		return 0;
-	case TGSI_OPCODE_BRK:
-		*copcode = C_OPCODE_BRK;
-		return 0;
-	case TGSI_OPCODE_IF:
-		*copcode = C_OPCODE_IF;
-		return 0;
-	case TGSI_OPCODE_ELSE:
-		*copcode = C_OPCODE_ELSE;
-		return 0;
-	case TGSI_OPCODE_ENDIF:
-		*copcode = C_OPCODE_ENDIF;
-		return 0;
-	case TGSI_OPCODE_PUSHA:
-		*copcode = C_OPCODE_PUSHA;
-		return 0;
-	case TGSI_OPCODE_POPA:
-		*copcode = C_OPCODE_POPA;
-		return 0;
-	case TGSI_OPCODE_CEIL:
-		*copcode = C_OPCODE_CEIL;
-		return 0;
-	case TGSI_OPCODE_I2F:
-		*copcode = C_OPCODE_I2F;
-		return 0;
-	case TGSI_OPCODE_NOT:
-		*copcode = C_OPCODE_NOT;
-		return 0;
-	case TGSI_OPCODE_TRUNC:
-		*copcode = C_OPCODE_TRUNC;
-		return 0;
-	case TGSI_OPCODE_SHL:
-		*copcode = C_OPCODE_SHL;
-		return 0;
-	case TGSI_OPCODE_AND:
-		*copcode = C_OPCODE_AND;
-		return 0;
-	case TGSI_OPCODE_OR:
-		*copcode = C_OPCODE_OR;
-		return 0;
-	case TGSI_OPCODE_MOD:
-		*copcode = C_OPCODE_MOD;
-		return 0;
-	case TGSI_OPCODE_XOR:
-		*copcode = C_OPCODE_XOR;
-		return 0;
-	case TGSI_OPCODE_SAD:
-		*copcode = C_OPCODE_SAD;
-		return 0;
-	case TGSI_OPCODE_TXF:
-		*copcode = C_OPCODE_TXF;
-		return 0;
-	case TGSI_OPCODE_TXQ:
-		*copcode = C_OPCODE_TXQ;
-		return 0;
-	case TGSI_OPCODE_CONT:
-		*copcode = C_OPCODE_CONT;
-		return 0;
-	case TGSI_OPCODE_EMIT:
-		*copcode = C_OPCODE_EMIT;
-		return 0;
-	case TGSI_OPCODE_ENDPRIM:
-		*copcode = C_OPCODE_ENDPRIM;
-		return 0;
-	case TGSI_OPCODE_BGNLOOP:
-		*copcode = C_OPCODE_BGNLOOP;
-		return 0;
-	case TGSI_OPCODE_BGNSUB:
-		*copcode = C_OPCODE_BGNSUB;
-		return 0;
-	case TGSI_OPCODE_ENDLOOP:
-		*copcode = C_OPCODE_ENDLOOP;
-		return 0;
-	case TGSI_OPCODE_ENDSUB:
-		*copcode = C_OPCODE_ENDSUB;
-		return 0;
-	case TGSI_OPCODE_NOP:
-		*copcode = C_OPCODE_NOP;
-		return 0;
-	case TGSI_OPCODE_NRM4:
-		*copcode = C_OPCODE_NRM4;
-		return 0;
-	case TGSI_OPCODE_CALLNZ:
-		*copcode = C_OPCODE_CALLNZ;
-		return 0;
-	case TGSI_OPCODE_IFC:
-		*copcode = C_OPCODE_IFC;
-		return 0;
-	case TGSI_OPCODE_BREAKC:
-		*copcode = C_OPCODE_BREAKC;
-		return 0;
-	case TGSI_OPCODE_KIL:
-		*copcode = C_OPCODE_KIL;
-		return 0;
-	case TGSI_OPCODE_F2I:
-		*copcode = C_OPCODE_F2I;
-		return 0;
-	case TGSI_OPCODE_IDIV:
-		*copcode = C_OPCODE_IDIV;
-		return 0;
-	case TGSI_OPCODE_IMAX:
-		*copcode = C_OPCODE_IMAX;
-		return 0;
-	case TGSI_OPCODE_IMIN:
-		*copcode = C_OPCODE_IMIN;
-		return 0;
-	case TGSI_OPCODE_INEG:
-		*copcode = C_OPCODE_INEG;
-		return 0;
-	case TGSI_OPCODE_ISGE:
-		*copcode = C_OPCODE_ISGE;
-		return 0;
-	case TGSI_OPCODE_ISHR:
-		*copcode = C_OPCODE_ISHR;
-		return 0;
-	case TGSI_OPCODE_ISLT:
-		*copcode = C_OPCODE_ISLT;
-		return 0;
-	case TGSI_OPCODE_F2U:
-		*copcode = C_OPCODE_F2U;
-		return 0;
-	case TGSI_OPCODE_U2F:
-		*copcode = C_OPCODE_U2F;
-		return 0;
-	case TGSI_OPCODE_UADD:
-		*copcode = C_OPCODE_UADD;
-		return 0;
-	case TGSI_OPCODE_UDIV:
-		*copcode = C_OPCODE_UDIV;
-		return 0;
-	case TGSI_OPCODE_UMAD:
-		*copcode = C_OPCODE_UMAD;
-		return 0;
-	case TGSI_OPCODE_UMAX:
-		*copcode = C_OPCODE_UMAX;
-		return 0;
-	case TGSI_OPCODE_UMIN:
-		*copcode = C_OPCODE_UMIN;
-		return 0;
-	case TGSI_OPCODE_UMOD:
-		*copcode = C_OPCODE_UMOD;
-		return 0;
-	case TGSI_OPCODE_UMUL:
-		*copcode = C_OPCODE_UMUL;
-		return 0;
-	case TGSI_OPCODE_USEQ:
-		*copcode = C_OPCODE_USEQ;
-		return 0;
-	case TGSI_OPCODE_USGE:
-		*copcode = C_OPCODE_USGE;
-		return 0;
-	case TGSI_OPCODE_USHR:
-		*copcode = C_OPCODE_USHR;
-		return 0;
-	case TGSI_OPCODE_USLT:
-		*copcode = C_OPCODE_USLT;
-		return 0;
-	case TGSI_OPCODE_USNE:
-		*copcode = C_OPCODE_USNE;
-		return 0;
-	case TGSI_OPCODE_SWITCH:
-		*copcode = C_OPCODE_SWITCH;
-		return 0;
-	case TGSI_OPCODE_CASE:
-		*copcode = C_OPCODE_CASE;
-		return 0;
-	case TGSI_OPCODE_DEFAULT:
-		*copcode = C_OPCODE_DEFAULT;
-		return 0;
-	case TGSI_OPCODE_ENDSWITCH:
-		*copcode = C_OPCODE_ENDSWITCH;
-		return 0;
-	default:
-		fprintf(stderr, "%s:%d unsupported opcode %d\n", __func__, __LINE__, opcode);
-		return -EINVAL;
-	}
-}
diff --git a/src/gallium/drivers/r600/r600_context.c b/src/gallium/drivers/r600/r600_context.c
index 05575b57676..edde80c660a 100644
--- a/src/gallium/drivers/r600/r600_context.c
+++ b/src/gallium/drivers/r600/r600_context.c
@@ -29,9 +29,9 @@
 #include <util/u_format.h>
 #include <util/u_memory.h>
 #include <util/u_blitter.h>
-#include "r600_resource.h"
 #include "r600_screen.h"
 #include "r600_context.h"
+#include "r600_resource.h"
 #include "r600d.h"
 
 static void r600_destroy_context(struct pipe_context *context)
@@ -41,26 +41,31 @@ static void r600_destroy_context(struct pipe_context *context)
 	FREE(rctx);
 }
 
-static void r600_flush(struct pipe_context *ctx, unsigned flags,
+void r600_flush(struct pipe_context *ctx, unsigned flags,
 			struct pipe_fence_handle **fence)
 {
 	struct r600_context *rctx = r600_context(ctx);
 	struct r600_screen *rscreen = rctx->screen;
 	static int dc = 0;
+	char dname[256];
 
 	if (radeon_ctx_pm4(rctx->ctx))
 		return;
 	/* FIXME dumping should be removed once shader support instructions
 	 * without throwing bad code
 	 */
-	if (!dc)
-		radeon_ctx_dump_bof(rctx->ctx, "gallium.bof");
-#if 0
+	if (!rctx->ctx->cpm4)
+		goto out;
+	sprintf(dname, "gallium-%08d.bof", dc);
+	if (dc < 1)
+		radeon_ctx_dump_bof(rctx->ctx, dname);
+#if 1
 	radeon_ctx_submit(rctx->ctx);
 #endif
+	dc++;
+out:
 	rctx->ctx = radeon_ctx_decref(rctx->ctx);
 	rctx->ctx = radeon_ctx(rscreen->rw);
-	dc++;
 }
 
 static void r600_init_config(struct r600_context *rctx)
@@ -202,27 +207,9 @@ static void r600_init_config(struct r600_context *rctx)
 		num_es_stack_entries = 0;
 		break;
 	}
-	printf("ps_prio : %d\n", ps_prio);
-	printf("vs_prio : %d\n", vs_prio);
-	printf("gs_prio : %d\n", gs_prio);
-	printf("es_prio : %d\n", es_prio);
-	printf("num_ps_gprs : %d\n", num_ps_gprs);
-	printf("num_vs_gprs : %d\n", num_vs_gprs);
-	printf("num_gs_gprs : %d\n", num_gs_gprs);
-	printf("num_es_gprs : %d\n", num_es_gprs);
-	printf("num_temp_gprs : %d\n", num_temp_gprs);
-	printf("num_ps_threads : %d\n", num_ps_threads);
-	printf("num_vs_threads : %d\n", num_vs_threads);
-	printf("num_gs_threads : %d\n", num_gs_threads);
-	printf("num_es_threads : %d\n", num_es_threads);
-	printf("num_ps_stack_entries : %d\n", num_ps_stack_entries);
-	printf("num_vs_stack_entries : %d\n", num_vs_stack_entries);
-	printf("num_gs_stack_entries : %d\n", num_gs_stack_entries);
-	printf("num_es_stack_entries : %d\n", num_es_stack_entries);
-
-	rctx->config = radeon_state(rctx->rw, R600_CONFIG_TYPE, R600_CONFIG);
+	rctx->hw_states.config = radeon_state(rctx->rw, R600_CONFIG_TYPE, R600_CONFIG);
 
-	rctx->config->states[R600_CONFIG__SQ_CONFIG] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__SQ_CONFIG] = 0x00000000;
 	switch (family) {
 	case CHIP_RV610:
 	case CHIP_RV620:
@@ -231,75 +218,75 @@ static void r600_init_config(struct r600_context *rctx)
 	case CHIP_RV710:
 		break;
 	default:
-		rctx->config->states[R600_CONFIG__SQ_CONFIG] |= S_008C00_VC_ENABLE(1);
+		rctx->hw_states.config->states[R600_CONFIG__SQ_CONFIG] |= S_008C00_VC_ENABLE(1);
 		break;
 	}
-	rctx->config->states[R600_CONFIG__SQ_CONFIG] |= S_008C00_DX9_CONSTS(1);
-	rctx->config->states[R600_CONFIG__SQ_CONFIG] |= S_008C00_ALU_INST_PREFER_VECTOR(1);
-	rctx->config->states[R600_CONFIG__SQ_CONFIG] |= S_008C00_PS_PRIO(ps_prio);
-	rctx->config->states[R600_CONFIG__SQ_CONFIG] |= S_008C00_VS_PRIO(vs_prio);
-	rctx->config->states[R600_CONFIG__SQ_CONFIG] |= S_008C00_GS_PRIO(gs_prio);
-	rctx->config->states[R600_CONFIG__SQ_CONFIG] |= S_008C00_ES_PRIO(es_prio);
+	rctx->hw_states.config->states[R600_CONFIG__SQ_CONFIG] |= S_008C00_DX9_CONSTS(1);
+	rctx->hw_states.config->states[R600_CONFIG__SQ_CONFIG] |= S_008C00_ALU_INST_PREFER_VECTOR(1);
+	rctx->hw_states.config->states[R600_CONFIG__SQ_CONFIG] |= S_008C00_PS_PRIO(ps_prio);
+	rctx->hw_states.config->states[R600_CONFIG__SQ_CONFIG] |= S_008C00_VS_PRIO(vs_prio);
+	rctx->hw_states.config->states[R600_CONFIG__SQ_CONFIG] |= S_008C00_GS_PRIO(gs_prio);
+	rctx->hw_states.config->states[R600_CONFIG__SQ_CONFIG] |= S_008C00_ES_PRIO(es_prio);
 
-	rctx->config->states[R600_CONFIG__SQ_GPR_RESOURCE_MGMT_1] = 0;
-	rctx->config->states[R600_CONFIG__SQ_GPR_RESOURCE_MGMT_1] |= S_008C04_NUM_PS_GPRS(num_ps_gprs);
-	rctx->config->states[R600_CONFIG__SQ_GPR_RESOURCE_MGMT_1] |= S_008C04_NUM_VS_GPRS(num_vs_gprs);
-	rctx->config->states[R600_CONFIG__SQ_GPR_RESOURCE_MGMT_1] |= S_008C04_NUM_CLAUSE_TEMP_GPRS(num_temp_gprs);
+	rctx->hw_states.config->states[R600_CONFIG__SQ_GPR_RESOURCE_MGMT_1] = 0;
+	rctx->hw_states.config->states[R600_CONFIG__SQ_GPR_RESOURCE_MGMT_1] |= S_008C04_NUM_PS_GPRS(num_ps_gprs);
+	rctx->hw_states.config->states[R600_CONFIG__SQ_GPR_RESOURCE_MGMT_1] |= S_008C04_NUM_VS_GPRS(num_vs_gprs);
+	rctx->hw_states.config->states[R600_CONFIG__SQ_GPR_RESOURCE_MGMT_1] |= S_008C04_NUM_CLAUSE_TEMP_GPRS(num_temp_gprs);
 
-	rctx->config->states[R600_CONFIG__SQ_GPR_RESOURCE_MGMT_2] = 0;
-	rctx->config->states[R600_CONFIG__SQ_GPR_RESOURCE_MGMT_2] |= S_008C08_NUM_GS_GPRS(num_gs_gprs);
-	rctx->config->states[R600_CONFIG__SQ_GPR_RESOURCE_MGMT_2] |= S_008C08_NUM_GS_GPRS(num_es_gprs);
+	rctx->hw_states.config->states[R600_CONFIG__SQ_GPR_RESOURCE_MGMT_2] = 0;
+	rctx->hw_states.config->states[R600_CONFIG__SQ_GPR_RESOURCE_MGMT_2] |= S_008C08_NUM_GS_GPRS(num_gs_gprs);
+	rctx->hw_states.config->states[R600_CONFIG__SQ_GPR_RESOURCE_MGMT_2] |= S_008C08_NUM_GS_GPRS(num_es_gprs);
 
-	rctx->config->states[R600_CONFIG__SQ_THREAD_RESOURCE_MGMT] = 0;
-	rctx->config->states[R600_CONFIG__SQ_THREAD_RESOURCE_MGMT] |= S_008C0C_NUM_PS_THREADS(num_ps_threads);
-	rctx->config->states[R600_CONFIG__SQ_THREAD_RESOURCE_MGMT] |= S_008C0C_NUM_VS_THREADS(num_vs_threads);
-	rctx->config->states[R600_CONFIG__SQ_THREAD_RESOURCE_MGMT] |= S_008C0C_NUM_GS_THREADS(num_gs_threads);
-	rctx->config->states[R600_CONFIG__SQ_THREAD_RESOURCE_MGMT] |= S_008C0C_NUM_ES_THREADS(num_es_threads);
+	rctx->hw_states.config->states[R600_CONFIG__SQ_THREAD_RESOURCE_MGMT] = 0;
+	rctx->hw_states.config->states[R600_CONFIG__SQ_THREAD_RESOURCE_MGMT] |= S_008C0C_NUM_PS_THREADS(num_ps_threads);
+	rctx->hw_states.config->states[R600_CONFIG__SQ_THREAD_RESOURCE_MGMT] |= S_008C0C_NUM_VS_THREADS(num_vs_threads);
+	rctx->hw_states.config->states[R600_CONFIG__SQ_THREAD_RESOURCE_MGMT] |= S_008C0C_NUM_GS_THREADS(num_gs_threads);
+	rctx->hw_states.config->states[R600_CONFIG__SQ_THREAD_RESOURCE_MGMT] |= S_008C0C_NUM_ES_THREADS(num_es_threads);
 
-	rctx->config->states[R600_CONFIG__SQ_STACK_RESOURCE_MGMT_1] = 0;
-	rctx->config->states[R600_CONFIG__SQ_STACK_RESOURCE_MGMT_1] |= S_008C10_NUM_PS_STACK_ENTRIES(num_ps_stack_entries);
-	rctx->config->states[R600_CONFIG__SQ_STACK_RESOURCE_MGMT_1] |= S_008C10_NUM_VS_STACK_ENTRIES(num_vs_stack_entries);
+	rctx->hw_states.config->states[R600_CONFIG__SQ_STACK_RESOURCE_MGMT_1] = 0;
+	rctx->hw_states.config->states[R600_CONFIG__SQ_STACK_RESOURCE_MGMT_1] |= S_008C10_NUM_PS_STACK_ENTRIES(num_ps_stack_entries);
+	rctx->hw_states.config->states[R600_CONFIG__SQ_STACK_RESOURCE_MGMT_1] |= S_008C10_NUM_VS_STACK_ENTRIES(num_vs_stack_entries);
 
-	rctx->config->states[R600_CONFIG__SQ_STACK_RESOURCE_MGMT_2] = 0;
-	rctx->config->states[R600_CONFIG__SQ_STACK_RESOURCE_MGMT_2] |= S_008C14_NUM_GS_STACK_ENTRIES(num_gs_stack_entries);
-	rctx->config->states[R600_CONFIG__SQ_STACK_RESOURCE_MGMT_2] |= S_008C14_NUM_ES_STACK_ENTRIES(num_es_stack_entries);
+	rctx->hw_states.config->states[R600_CONFIG__SQ_STACK_RESOURCE_MGMT_2] = 0;
+	rctx->hw_states.config->states[R600_CONFIG__SQ_STACK_RESOURCE_MGMT_2] |= S_008C14_NUM_GS_STACK_ENTRIES(num_gs_stack_entries);
+	rctx->hw_states.config->states[R600_CONFIG__SQ_STACK_RESOURCE_MGMT_2] |= S_008C14_NUM_ES_STACK_ENTRIES(num_es_stack_entries);
 
-	rctx->config->states[R600_CONFIG__SQ_DYN_GPR_CNTL_PS_FLUSH_REQ] = 0x00004000;
-	rctx->config->states[R600_CONFIG__TA_CNTL_AUX] = 0x07000002;
-	rctx->config->states[R600_CONFIG__VC_ENHANCE] = 0x00000000;
-	rctx->config->states[R600_CONFIG__DB_DEBUG] = 0x00000000;
-	rctx->config->states[R600_CONFIG__DB_WATERMARKS] = 0x00420204;
-	rctx->config->states[R600_CONFIG__SX_MISC] = 0x00000000;
-	rctx->config->states[R600_CONFIG__SPI_THREAD_GROUPING] = 0x00000001;
-	rctx->config->states[R600_CONFIG__CB_SHADER_CONTROL] = 0x00000003;
-	rctx->config->states[R600_CONFIG__SQ_ESGS_RING_ITEMSIZE] = 0x00000000;
-	rctx->config->states[R600_CONFIG__SQ_GSVS_RING_ITEMSIZE] = 0x00000000;
-	rctx->config->states[R600_CONFIG__SQ_ESTMP_RING_ITEMSIZE] = 0x00000000;
-	rctx->config->states[R600_CONFIG__SQ_GSTMP_RING_ITEMSIZE] = 0x00000000;
-	rctx->config->states[R600_CONFIG__SQ_VSTMP_RING_ITEMSIZE] = 0x00000000;
-	rctx->config->states[R600_CONFIG__SQ_PSTMP_RING_ITEMSIZE] = 0x00000000;
-	rctx->config->states[R600_CONFIG__SQ_FBUF_RING_ITEMSIZE] = 0x00000000;
-	rctx->config->states[R600_CONFIG__SQ_REDUC_RING_ITEMSIZE] = 0x00000000;
-	rctx->config->states[R600_CONFIG__SQ_GS_VERT_ITEMSIZE] = 0x00000000;
-	rctx->config->states[R600_CONFIG__VGT_OUTPUT_PATH_CNTL] = 0x00000000;
-	rctx->config->states[R600_CONFIG__VGT_HOS_CNTL] = 0x00000000;
-	rctx->config->states[R600_CONFIG__VGT_HOS_MAX_TESS_LEVEL] = 0x00000000;
-	rctx->config->states[R600_CONFIG__VGT_HOS_MIN_TESS_LEVEL] = 0x00000000;
-	rctx->config->states[R600_CONFIG__VGT_HOS_REUSE_DEPTH] = 0x00000000;
-	rctx->config->states[R600_CONFIG__VGT_GROUP_PRIM_TYPE] = 0x00000000;
-	rctx->config->states[R600_CONFIG__VGT_GROUP_FIRST_DECR] = 0x00000000;
-	rctx->config->states[R600_CONFIG__VGT_GROUP_DECR] = 0x00000000;
-	rctx->config->states[R600_CONFIG__VGT_GROUP_VECT_0_CNTL] = 0x00000000;
-	rctx->config->states[R600_CONFIG__VGT_GROUP_VECT_1_CNTL] = 0x00000000;
-	rctx->config->states[R600_CONFIG__VGT_GROUP_VECT_0_FMT_CNTL] = 0x00000000;
-	rctx->config->states[R600_CONFIG__VGT_GROUP_VECT_1_FMT_CNTL] = 0x00000000;
-	rctx->config->states[R600_CONFIG__VGT_GS_MODE] = 0x00000000;
-	rctx->config->states[R600_CONFIG__PA_SC_MODE_CNTL] = 0x00514000;
-	rctx->config->states[R600_CONFIG__VGT_STRMOUT_EN] = 0x00000000;
-	rctx->config->states[R600_CONFIG__VGT_REUSE_OFF] = 0x00000001;
-	rctx->config->states[R600_CONFIG__VGT_VTX_CNT_EN] = 0x00000000;
-	rctx->config->states[R600_CONFIG__VGT_STRMOUT_BUFFER_EN] = 0x00000000;
-	radeon_state_pm4(rctx->config);
+	rctx->hw_states.config->states[R600_CONFIG__SQ_DYN_GPR_CNTL_PS_FLUSH_REQ] = 0x00004000;
+	rctx->hw_states.config->states[R600_CONFIG__TA_CNTL_AUX] = 0x07000002;
+	rctx->hw_states.config->states[R600_CONFIG__VC_ENHANCE] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__DB_DEBUG] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__DB_WATERMARKS] = 0x00420204;
+	rctx->hw_states.config->states[R600_CONFIG__SX_MISC] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__SPI_THREAD_GROUPING] = 0x00000001;
+	rctx->hw_states.config->states[R600_CONFIG__CB_SHADER_CONTROL] = 0x00000003;
+	rctx->hw_states.config->states[R600_CONFIG__SQ_ESGS_RING_ITEMSIZE] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__SQ_GSVS_RING_ITEMSIZE] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__SQ_ESTMP_RING_ITEMSIZE] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__SQ_GSTMP_RING_ITEMSIZE] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__SQ_VSTMP_RING_ITEMSIZE] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__SQ_PSTMP_RING_ITEMSIZE] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__SQ_FBUF_RING_ITEMSIZE] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__SQ_REDUC_RING_ITEMSIZE] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__SQ_GS_VERT_ITEMSIZE] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__VGT_OUTPUT_PATH_CNTL] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__VGT_HOS_CNTL] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__VGT_HOS_MAX_TESS_LEVEL] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__VGT_HOS_MIN_TESS_LEVEL] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__VGT_HOS_REUSE_DEPTH] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__VGT_GROUP_PRIM_TYPE] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__VGT_GROUP_FIRST_DECR] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__VGT_GROUP_DECR] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__VGT_GROUP_VECT_0_CNTL] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__VGT_GROUP_VECT_1_CNTL] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__VGT_GROUP_VECT_0_FMT_CNTL] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__VGT_GROUP_VECT_1_FMT_CNTL] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__VGT_GS_MODE] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__PA_SC_MODE_CNTL] = 0x00514000;
+	rctx->hw_states.config->states[R600_CONFIG__VGT_STRMOUT_EN] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__VGT_REUSE_OFF] = 0x00000001;
+	rctx->hw_states.config->states[R600_CONFIG__VGT_VTX_CNT_EN] = 0x00000000;
+	rctx->hw_states.config->states[R600_CONFIG__VGT_STRMOUT_BUFFER_EN] = 0x00000000;
+	radeon_state_pm4(rctx->hw_states.config);
 }
 
 struct pipe_context *r600_create_context(struct pipe_screen *screen, void *priv)
@@ -313,9 +300,7 @@ struct pipe_context *r600_create_context(struct pipe_screen *screen, void *priv)
 	rctx->context.screen = screen;
 	rctx->context.priv = priv;
 	rctx->context.destroy = r600_destroy_context;
-	rctx->context.draw_arrays = r600_draw_arrays;
-	rctx->context.draw_elements = r600_draw_elements;
-	rctx->context.draw_range_elements = r600_draw_range_elements;
+	rctx->context.draw_vbo = r600_draw_vbo;
 	rctx->context.flush = r600_flush;
 
 	/* Easy accessing of screen/winsys. */
@@ -333,20 +318,6 @@ struct pipe_context *r600_create_context(struct pipe_screen *screen, void *priv)
 		return NULL;
 	}
 
-	rctx->cb_cntl = radeon_state(rscreen->rw, R600_CB_CNTL_TYPE, R600_CB_CNTL);
-	rctx->cb_cntl->states[R600_CB_CNTL__CB_SHADER_MASK] = 0x0000000F;
-	rctx->cb_cntl->states[R600_CB_CNTL__CB_TARGET_MASK] = 0x0000000F;
-	rctx->cb_cntl->states[R600_CB_CNTL__CB_COLOR_CONTROL] = 0x00CC0000;
-	rctx->cb_cntl->states[R600_CB_CNTL__PA_SC_AA_CONFIG] = 0x00000000;
-	rctx->cb_cntl->states[R600_CB_CNTL__PA_SC_AA_SAMPLE_LOCS_MCTX] = 0x00000000;
-	rctx->cb_cntl->states[R600_CB_CNTL__PA_SC_AA_SAMPLE_LOCS_8S_WD1_MCTX] = 0x00000000;
-	rctx->cb_cntl->states[R600_CB_CNTL__CB_CLRCMP_CONTROL] = 0x01000000;
-	rctx->cb_cntl->states[R600_CB_CNTL__CB_CLRCMP_SRC] = 0x00000000;
-	rctx->cb_cntl->states[R600_CB_CNTL__CB_CLRCMP_DST] = 0x000000FF;
-	rctx->cb_cntl->states[R600_CB_CNTL__CB_CLRCMP_MSK] = 0xFFFFFFFF;
-	rctx->cb_cntl->states[R600_CB_CNTL__PA_SC_AA_MASK] = 0xFFFFFFFF;
-	radeon_state_pm4(rctx->cb_cntl);
-
 	r600_init_config(rctx);
 
 	rctx->ctx = radeon_ctx(rscreen->rw);
diff --git a/src/gallium/drivers/r600/r600_context.h b/src/gallium/drivers/r600/r600_context.h
index f27ff58ed46..76d5de86532 100644
--- a/src/gallium/drivers/r600/r600_context.h
+++ b/src/gallium/drivers/r600/r600_context.h
@@ -23,6 +23,7 @@
 #ifndef R600_CONTEXT_H
 #define R600_CONTEXT_H
 
+#include <stdio.h>
 #include <pipe/p_state.h>
 #include <pipe/p_context.h>
 #include <tgsi/tgsi_scan.h>
@@ -33,17 +34,74 @@
 #include "r600_shader.h"
 
 /* XXX move this to a more appropriate place */
-struct r600_vertex_elements_state
+union pipe_states {
+	struct pipe_rasterizer_state		rasterizer;
+	struct pipe_poly_stipple		poly_stipple;
+	struct pipe_scissor_state		scissor;
+	struct pipe_clip_state			clip;
+	struct pipe_shader_state		shader;
+	struct pipe_depth_state			depth;
+	struct pipe_stencil_state		stencil;
+	struct pipe_alpha_state			alpha;
+	struct pipe_depth_stencil_alpha_state	dsa;
+	struct pipe_blend_state			blend;
+	struct pipe_blend_color			blend_color;
+	struct pipe_stencil_ref			stencil_ref;
+	struct pipe_framebuffer_state		framebuffer;
+	struct pipe_sampler_state		sampler;
+	struct pipe_sampler_view		sampler_view;
+	struct pipe_viewport_state		viewport;
+};
+
+enum pipe_state_type {
+	pipe_rasterizer_type = 1,
+	pipe_poly_stipple_type,
+	pipe_scissor_type,
+	pipe_clip_type,
+	pipe_shader_type,
+	pipe_depth_type,
+	pipe_stencil_type,
+	pipe_alpha_type,
+	pipe_dsa_type,
+	pipe_blend_type,
+	pipe_stencil_ref_type,
+	pipe_framebuffer_type,
+	pipe_sampler_type,
+	pipe_sampler_view_type,
+	pipe_viewport_type,
+	pipe_type_count
+};
+
+struct r600_context_state {
+	union pipe_states		state;
+	unsigned			refcount;
+	unsigned			type;
+	struct radeon_state		*rstate;
+	struct r600_shader		shader;
+	struct radeon_bo		*bo;
+};
+
+struct r600_vertex_element
 {
-	unsigned count;
-	struct pipe_vertex_element elements[32];
+	unsigned			refcount;
+	unsigned			count;
+	struct pipe_vertex_element	elements[32];
 };
 
-struct r600_pipe_shader {
-	unsigned				type;
-	struct r600_shader			shader;
-	struct radeon_bo			*bo;
-	struct radeon_state			*state;
+struct r600_context_hw_states {
+	struct radeon_state	*rasterizer;
+	struct radeon_state	*scissor;
+	struct radeon_state	*dsa;
+	struct radeon_state	*blend;
+	struct radeon_state	*viewport;
+	struct radeon_state	*cb[8];
+	struct radeon_state	*config;
+	struct radeon_state	*cb_cntl;
+	struct radeon_state	*db;
+	unsigned		ps_nresource;
+	unsigned		ps_nsampler;
+	struct radeon_state	*ps_resource[160];
+	struct radeon_state	*ps_sampler[16];
 };
 
 struct r600_context {
@@ -51,20 +109,39 @@ struct r600_context {
 	struct r600_screen		*screen;
 	struct radeon			*rw;
 	struct radeon_ctx		*ctx;
-	struct radeon_state		*cb_cntl;
-	struct radeon_state		*db;
-	struct radeon_state		*config;
-	struct r600_pipe_shader		*ps_shader;
-	struct r600_pipe_shader		*vs_shader;
+	struct blitter_context		*blitter;
+	struct radeon_draw		*draw;
+	/* hw states */
+	struct r600_context_hw_states	hw_states;
+	/* pipe states */
 	unsigned			flat_shade;
+	unsigned			ps_nsampler;
+	unsigned			vs_nsampler;
+	unsigned			ps_nsampler_view;
+	unsigned			vs_nsampler_view;
 	unsigned			nvertex_buffer;
-	struct r600_vertex_elements_state *vertex_elements;
+	struct r600_context_state	*rasterizer;
+	struct r600_context_state	*poly_stipple;
+	struct r600_context_state	*scissor;
+	struct r600_context_state	*clip;
+	struct r600_context_state	*ps_shader;
+	struct r600_context_state	*vs_shader;
+	struct r600_context_state	*depth;
+	struct r600_context_state	*stencil;
+	struct r600_context_state	*alpha;
+	struct r600_context_state	*dsa;
+	struct r600_context_state	*blend;
+	struct r600_context_state	*stencil_ref;
+	struct r600_context_state	*viewport;
+	struct r600_context_state	*framebuffer;
+	struct r600_context_state	*ps_sampler[PIPE_MAX_ATTRIBS];
+	struct r600_context_state	*vs_sampler[PIPE_MAX_ATTRIBS];
+	struct r600_context_state	*ps_sampler_view[PIPE_MAX_ATTRIBS];
+	struct r600_context_state	*vs_sampler_view[PIPE_MAX_ATTRIBS];
+	struct r600_vertex_element	*vertex_elements;
 	struct pipe_vertex_buffer	vertex_buffer[PIPE_MAX_ATTRIBS];
-	struct blitter_context		*blitter;
-	struct pipe_stencil_ref		stencil_ref;
-	struct pipe_framebuffer_state	fb_state;
-	struct radeon_draw		*draw;
-	struct pipe_viewport_state	viewport;
+	struct pipe_index_buffer	index_buffer;
+	struct pipe_blend_color         blend_color;
 };
 
 /* Convenience cast wrapper. */
@@ -73,27 +150,32 @@ static INLINE struct r600_context *r600_context(struct pipe_context *pipe)
     return (struct r600_context*)pipe;
 }
 
-void r600_draw_arrays(struct pipe_context *ctx, unsigned mode,
-			unsigned start, unsigned count);
-void r600_draw_elements(struct pipe_context *ctx,
-		struct pipe_resource *index_buffer,
-		unsigned index_size, int index_bias, unsigned mode,
-		unsigned start, unsigned count);
-void r600_draw_range_elements(struct pipe_context *ctx,
-		struct pipe_resource *index_buffer,
-		unsigned index_size, int index_bias, unsigned min_index,
-		unsigned max_index, unsigned mode,
-		unsigned start, unsigned count);
+struct r600_context_state *r600_context_state(struct r600_context *rctx, unsigned type, const void *state);
+struct r600_context_state *r600_context_state_incref(struct r600_context_state *rstate);
+struct r600_context_state *r600_context_state_decref(struct r600_context_state *rstate);
+void r600_flush(struct pipe_context *ctx, unsigned flags,
+			struct pipe_fence_handle **fence);
+
+int r600_context_hw_states(struct r600_context *rctx);
+
+void r600_draw_vbo(struct pipe_context *ctx,
+                   const struct pipe_draw_info *info);
 
 void r600_init_blit_functions(struct r600_context *rctx);
 void r600_init_state_functions(struct r600_context *rctx);
 void r600_init_query_functions(struct r600_context* rctx);
 struct pipe_context *r600_create_context(struct pipe_screen *screen, void *priv);
 
-void r600_pipe_shader_destroy(struct pipe_context *ctx, struct r600_pipe_shader *rpshader);
-struct r600_pipe_shader *r600_pipe_shader_create(struct pipe_context *ctx,
-						unsigned type,
-						const struct tgsi_token *tokens);
-int r600_pipe_shader_update(struct pipe_context *ctx, struct r600_pipe_shader *rpshader);
+extern int r600_pipe_shader_create(struct pipe_context *ctx,
+			struct r600_context_state *rstate,
+			const struct tgsi_token *tokens);
+extern int r600_pipe_shader_update(struct pipe_context *ctx,
+				struct r600_context_state *rstate);
+
+#define R600_ERR(fmt, args...) \
+	fprintf(stderr, "EE %s/%s:%d - "fmt, __FILE__, __func__, __LINE__, ##args)
 
+uint32_t r600_translate_texformat(enum pipe_format format,
+				  const unsigned char *swizzle_view, 
+				  uint32_t *word4_p, uint32_t *yuv_format_p);
 #endif
diff --git a/src/gallium/drivers/r600/r600_draw.c b/src/gallium/drivers/r600/r600_draw.c
index 724fb6c988e..f0584551620 100644
--- a/src/gallium/drivers/r600/r600_draw.c
+++ b/src/gallium/drivers/r600/r600_draw.c
@@ -33,7 +33,8 @@
 #include <util/u_memory.h>
 #include "r600_screen.h"
 #include "r600_context.h"
-#include "r600d.h"
+#include "r600_resource.h"
+#include "r600_state_inlines.h"
 
 struct r600_draw {
 	struct pipe_context	*ctx;
@@ -51,11 +52,15 @@ static int r600_draw_common(struct r600_draw *draw)
 	struct r600_context *rctx = r600_context(draw->ctx);
 	struct r600_screen *rscreen = rctx->screen;
 	struct radeon_state *vs_resource;
-	struct r600_buffer *rbuffer;
+	struct r600_resource *rbuffer;
 	unsigned i, j, offset, format, prim;
 	u32 vgt_dma_index_type, vgt_draw_initiator;
+	struct pipe_vertex_buffer *vertex_buffer;
 	int r;
 
+	r = r600_context_hw_states(rctx);
+	if (r)
+		return r;
 	switch (draw->index_size) {
 	case 2:
 		vgt_draw_initiator = 0;
@@ -83,29 +88,19 @@ static int r600_draw_common(struct r600_draw *draw)
 	r = r600_pipe_shader_update(draw->ctx, rctx->ps_shader);
 	if (r)
 		return r;
-	r = radeon_draw_set(rctx->draw, rctx->vs_shader->state);
-	if (r)
-		return r;
-	r = radeon_draw_set(rctx->draw, rctx->ps_shader->state);
+	r = radeon_draw_set(rctx->draw, rctx->vs_shader->rstate);
 	if (r)
 		return r;
-	r = radeon_draw_set(rctx->draw, rctx->cb_cntl);
-	if (r)
-		return r;
-	r = radeon_draw_set(rctx->draw, rctx->db);
-	if (r)
-		return r;
-	r = radeon_draw_set(rctx->draw, rctx->config);
+	r = radeon_draw_set(rctx->draw, rctx->ps_shader->rstate);
 	if (r)
 		return r;
 
 	for (i = 0 ; i < rctx->vertex_elements->count; i++) {
 		j = rctx->vertex_elements->elements[i].vertex_buffer_index;
-		rbuffer = (struct r600_buffer*)rctx->vertex_buffer[j].buffer;
-		offset = rctx->vertex_elements->elements[i].src_offset + rctx->vertex_buffer[j].buffer_offset;
-		r = r600_conv_pipe_format(rctx->vertex_elements->elements[i].src_format, &format);
-		if (r)
-			return r;
+		vertex_buffer = &rctx->vertex_buffer[j];
+		rbuffer = (struct r600_resource*)vertex_buffer->buffer;
+		offset = rctx->vertex_elements->elements[i].src_offset + vertex_buffer->buffer_offset;
+		format = r600_translate_colorformat(rctx->vertex_elements->elements[i].src_format);
 		vs_resource = radeon_state(rscreen->rw, R600_VS_RESOURCE_TYPE, R600_VS_RESOURCE + i);
 		if (vs_resource == NULL)
 			return -ENOMEM;
@@ -113,7 +108,7 @@ static int r600_draw_common(struct r600_draw *draw)
 		vs_resource->nbo = 1;
 		vs_resource->states[R600_PS_RESOURCE__RESOURCE0_WORD0] = offset;
 		vs_resource->states[R600_PS_RESOURCE__RESOURCE0_WORD1] = rbuffer->bo->size - offset;
-		vs_resource->states[R600_PS_RESOURCE__RESOURCE0_WORD2] = S_038008_STRIDE(rctx->vertex_buffer[j].stride) |
+		vs_resource->states[R600_PS_RESOURCE__RESOURCE0_WORD2] = S_038008_STRIDE(vertex_buffer->stride) |
 								S_038008_DATA_FORMAT(format);
 		vs_resource->states[R600_PS_RESOURCE__RESOURCE0_WORD3] = 0x00000000;
 		vs_resource->states[R600_PS_RESOURCE__RESOURCE0_WORD4] = 0x00000000;
@@ -132,7 +127,7 @@ static int r600_draw_common(struct r600_draw *draw)
 	draw->draw->states[R600_DRAW__VGT_NUM_INDICES] = draw->count;
 	draw->draw->states[R600_DRAW__VGT_DRAW_INITIATOR] = vgt_draw_initiator;
 	if (draw->index_buffer) {
-		rbuffer = (struct r600_buffer*)draw->index_buffer;
+		rbuffer = (struct r600_resource*)draw->index_buffer;
 		draw->draw->bo[0] = radeon_bo_incref(rscreen->rw, rbuffer->bo);
 		draw->draw->placement[0] = RADEON_GEM_DOMAIN_GTT;
 		draw->draw->placement[1] = RADEON_GEM_DOMAIN_GTT;
@@ -160,58 +155,39 @@ static int r600_draw_common(struct r600_draw *draw)
 		return r;
 	/* FIXME */
 	r = radeon_ctx_set_draw_new(rctx->ctx, rctx->draw);
+	if (r == -EBUSY) {
+		r600_flush(draw->ctx, 0, NULL);
+		r = radeon_ctx_set_draw_new(rctx->ctx, rctx->draw);
+	}
 	if (r)
 		return r;
 	rctx->draw = radeon_draw_duplicate(rctx->draw);
 	return 0;
 }
 
-void r600_draw_range_elements(struct pipe_context *ctx,
-		struct pipe_resource *index_buffer,
-		unsigned index_size, int index_bias, unsigned min_index,
-		unsigned max_index, unsigned mode,
-		unsigned start, unsigned count)
+void r600_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *info)
 {
+	struct r600_context *rctx = r600_context(ctx);
 	struct r600_draw draw;
-	assert(index_bias == 0);
 
-	draw.ctx = ctx;
-	draw.mode = mode;
-	draw.start = start;
-	draw.count = count;
-	draw.index_size = index_size;
-	draw.index_buffer = index_buffer;
-printf("index_size %d min %d max %d  start %d  count %d\n", index_size, min_index, max_index, start, count);
-	r600_draw_common(&draw);
-}
-
-void r600_draw_elements(struct pipe_context *ctx,
-		struct pipe_resource *index_buffer,
-		unsigned index_size, int index_bias, unsigned mode,
-		unsigned start, unsigned count)
-{
-	struct r600_draw draw;
-	assert(index_bias == 0);
+	assert(info->index_bias == 0);
 
 	draw.ctx = ctx;
-	draw.mode = mode;
-	draw.start = start;
-	draw.count = count;
-	draw.index_size = index_size;
-	draw.index_buffer = index_buffer;
-	r600_draw_common(&draw);
-}
+	draw.mode = info->mode;
+	draw.start = info->start;
+	draw.count = info->count;
+	if (info->indexed && rctx->index_buffer.buffer) {
+		draw.index_size = rctx->index_buffer.index_size;
+		draw.index_buffer = rctx->index_buffer.buffer;
 
-void r600_draw_arrays(struct pipe_context *ctx, unsigned mode,
-			unsigned start, unsigned count)
-{
-	struct r600_draw draw;
-
-	draw.ctx = ctx;
-	draw.mode = mode;
-	draw.start = start;
-	draw.count = count;
-	draw.index_size = 0;
-	draw.index_buffer = NULL;
+		assert(rctx->index_buffer.offset %
+				rctx->index_buffer.index_size == 0);
+		draw.start += rctx->index_buffer.offset /
+			rctx->index_buffer.index_size;
+	}
+	else {
+		draw.index_size = 0;
+		draw.index_buffer = NULL;
+	}
 	r600_draw_common(&draw);
 }
diff --git a/src/gallium/drivers/r600/r600_helper.c b/src/gallium/drivers/r600/r600_helper.c
index e3175b627aa..5e0e0aab570 100644
--- a/src/gallium/drivers/r600/r600_helper.c
+++ b/src/gallium/drivers/r600/r600_helper.c
@@ -27,95 +27,9 @@
 #include <errno.h>
 #include <util/u_inlines.h>
 #include "r600_screen.h"
+#include "r600_context.h"
 #include "r600d.h"
 
-int r600_conv_pipe_format(unsigned pformat, unsigned *format)
-{
-	switch (pformat) {
-	case PIPE_FORMAT_R32G32B32_FLOAT:
-		*format = 0x30;
-		return 0;
-	case PIPE_FORMAT_R32G32B32A32_FLOAT:
-		*format = V_0280A0_COLOR_32_32_32_32_FLOAT;
-		return 0;
-	case PIPE_FORMAT_A8R8G8B8_UNORM:
-	case PIPE_FORMAT_X8R8G8B8_UNORM:
-	case PIPE_FORMAT_B8G8R8A8_UNORM:
-	case PIPE_FORMAT_B8G8R8X8_UNORM:
-	case PIPE_FORMAT_R8G8B8A8_UNORM:
-	case PIPE_FORMAT_R8G8B8X8_UNORM:
-	case PIPE_FORMAT_R8G8B8A8_USCALED:
-	case PIPE_FORMAT_R8G8B8A8_SNORM:
-	case PIPE_FORMAT_R8G8B8A8_SSCALED:
-		*format = V_0280A0_COLOR_8_8_8_8;
-		return 0;
-	case PIPE_FORMAT_L8_UNORM:
-	case PIPE_FORMAT_A8_UNORM:
-	case PIPE_FORMAT_I8_UNORM:
-	case PIPE_FORMAT_L16_UNORM:
-	case PIPE_FORMAT_Z16_UNORM:
-	case PIPE_FORMAT_Z32_UNORM:
-	case PIPE_FORMAT_Z32_FLOAT:
-	case PIPE_FORMAT_R64_FLOAT:
-	case PIPE_FORMAT_R64G64_FLOAT:
-	case PIPE_FORMAT_R64G64B64_FLOAT:
-	case PIPE_FORMAT_R64G64B64A64_FLOAT:
-	case PIPE_FORMAT_R32_FLOAT:
-	case PIPE_FORMAT_R32G32_FLOAT:
-	case PIPE_FORMAT_R32_UNORM:
-	case PIPE_FORMAT_R32G32_UNORM:
-	case PIPE_FORMAT_R32G32B32_UNORM:
-	case PIPE_FORMAT_R32G32B32A32_UNORM:
-	case PIPE_FORMAT_R32_USCALED:
-	case PIPE_FORMAT_R32G32_USCALED:
-	case PIPE_FORMAT_R32G32B32_USCALED:
-	case PIPE_FORMAT_R32G32B32A32_USCALED:
-	case PIPE_FORMAT_R32_SNORM:
-	case PIPE_FORMAT_R32G32_SNORM:
-	case PIPE_FORMAT_R32G32B32_SNORM:
-	case PIPE_FORMAT_R32G32B32A32_SNORM:
-	case PIPE_FORMAT_R32_SSCALED:
-	case PIPE_FORMAT_R32G32_SSCALED:
-	case PIPE_FORMAT_R32G32B32_SSCALED:
-	case PIPE_FORMAT_R32G32B32A32_SSCALED:
-	case PIPE_FORMAT_R16_UNORM:
-	case PIPE_FORMAT_R16G16_UNORM:
-	case PIPE_FORMAT_R16G16B16_UNORM:
-	case PIPE_FORMAT_R16G16B16A16_UNORM:
-	case PIPE_FORMAT_R16_USCALED:
-	case PIPE_FORMAT_R16G16_USCALED:
-	case PIPE_FORMAT_R16G16B16_USCALED:
-	case PIPE_FORMAT_R16G16B16A16_USCALED:
-	case PIPE_FORMAT_R16_SNORM:
-	case PIPE_FORMAT_R16G16_SNORM:
-	case PIPE_FORMAT_R16G16B16_SNORM:
-	case PIPE_FORMAT_R16G16B16A16_SNORM:
-	case PIPE_FORMAT_R16_SSCALED:
-	case PIPE_FORMAT_R16G16_SSCALED:
-	case PIPE_FORMAT_R16G16B16_SSCALED:
-	case PIPE_FORMAT_R16G16B16A16_SSCALED:
-	case PIPE_FORMAT_R8_UNORM:
-	case PIPE_FORMAT_R8G8_UNORM:
-	case PIPE_FORMAT_R8G8B8_UNORM:
-	case PIPE_FORMAT_R8_USCALED:
-	case PIPE_FORMAT_R8G8_USCALED:
-	case PIPE_FORMAT_R8G8B8_USCALED:
-	case PIPE_FORMAT_R8_SNORM:
-	case PIPE_FORMAT_R8G8_SNORM:
-	case PIPE_FORMAT_R8G8B8_SNORM:
-	case PIPE_FORMAT_R8_SSCALED:
-	case PIPE_FORMAT_R8G8_SSCALED:
-	case PIPE_FORMAT_R8G8B8_SSCALED:
-	case PIPE_FORMAT_R32_FIXED:
-	case PIPE_FORMAT_R32G32_FIXED:
-	case PIPE_FORMAT_R32G32B32_FIXED:
-	case PIPE_FORMAT_R32G32B32A32_FIXED:
-	default:
-		fprintf(stderr, "%s:%d unsupported %d\n", __func__, __LINE__, pformat);
-		return -EINVAL;
-	}
-}
-
 int r600_conv_pipe_prim(unsigned pprim, unsigned *prim)
 {
 	switch (pprim) {
diff --git a/src/gallium/drivers/r600/r600_resource.c b/src/gallium/drivers/r600/r600_resource.c
index d9aa1df04f0..8dc411ef409 100644
--- a/src/gallium/drivers/r600/r600_resource.c
+++ b/src/gallium/drivers/r600/r600_resource.c
@@ -24,45 +24,44 @@
 #include "r600_context.h"
 #include "r600_resource.h"
 #include "r600_screen.h"
-#include "r600_texture.h"
 
-static struct pipe_resource *
-r600_resource_create(struct pipe_screen *screen,
-                    const struct pipe_resource *templ)
+static struct pipe_resource *r600_resource_create(struct pipe_screen *screen,
+						const struct pipe_resource *templ)
 {
-   if (templ->target == PIPE_BUFFER)
-      return r600_buffer_create(screen, templ);
-   else
-      return r600_texture_create(screen, templ);
+	if (templ->target == PIPE_BUFFER) {
+		return r600_buffer_create(screen, templ);
+	} else {
+		return r600_texture_create(screen, templ);
+	}
 }
 
-static struct pipe_resource *
-r600_resource_from_handle(struct pipe_screen * screen,
-			 const struct pipe_resource *templ,
-			 struct winsys_handle *whandle)
+static struct pipe_resource *r600_resource_from_handle(struct pipe_screen * screen,
+							const struct pipe_resource *templ,
+							struct winsys_handle *whandle)
 {
-   if (templ->target == PIPE_BUFFER)
-      return NULL;
-   else
-      return r600_texture_from_handle(screen, templ, whandle);
+	if (templ->target == PIPE_BUFFER) {
+		return NULL;
+	} else {
+		return r600_texture_from_handle(screen, templ, whandle);
+	}
 }
 
 void r600_init_context_resource_functions(struct r600_context *r600)
 {
-   r600->context.get_transfer = u_get_transfer_vtbl;
-   r600->context.transfer_map = u_transfer_map_vtbl;
-   r600->context.transfer_flush_region = u_transfer_flush_region_vtbl;
-   r600->context.transfer_unmap = u_transfer_unmap_vtbl;
-   r600->context.transfer_destroy = u_transfer_destroy_vtbl;
-   r600->context.transfer_inline_write = u_transfer_inline_write_vtbl;
-   r600->context.is_resource_referenced = u_is_resource_referenced_vtbl;
+	r600->context.get_transfer = u_get_transfer_vtbl;
+	r600->context.transfer_map = u_transfer_map_vtbl;
+	r600->context.transfer_flush_region = u_transfer_flush_region_vtbl;
+	r600->context.transfer_unmap = u_transfer_unmap_vtbl;
+	r600->context.transfer_destroy = u_transfer_destroy_vtbl;
+	r600->context.transfer_inline_write = u_transfer_inline_write_vtbl;
+	r600->context.is_resource_referenced = u_is_resource_referenced_vtbl;
 }
 
 void r600_init_screen_resource_functions(struct r600_screen *r600screen)
 {
-   r600screen->screen.resource_create = r600_resource_create;
-   r600screen->screen.resource_from_handle = r600_resource_from_handle;
-   r600screen->screen.resource_get_handle = u_resource_get_handle_vtbl;
-   r600screen->screen.resource_destroy = u_resource_destroy_vtbl;
-   r600screen->screen.user_buffer_create = r600_user_buffer_create;
+	r600screen->screen.resource_create = r600_resource_create;
+	r600screen->screen.resource_from_handle = r600_resource_from_handle;
+	r600screen->screen.resource_get_handle = u_resource_get_handle_vtbl;
+	r600screen->screen.resource_destroy = u_resource_destroy_vtbl;
+	r600screen->screen.user_buffer_create = r600_user_buffer_create;
 }
diff --git a/src/gallium/drivers/r600/r600_resource.h b/src/gallium/drivers/r600/r600_resource.h
index 95084a371bd..bb90e76fb78 100644
--- a/src/gallium/drivers/r600/r600_resource.h
+++ b/src/gallium/drivers/r600/r600_resource.h
@@ -20,14 +20,47 @@
  * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
  * USE OR OTHER DEALINGS IN THE SOFTWARE.
  */
-
 #ifndef R600_RESOURCE_H
 #define R600_RESOURCE_H
 
+#include "util/u_transfer.h"
+
 struct r600_context;
 struct r600_screen;
 
+/* This gets further specialized into either buffer or texture
+ * structures. Use the vtbl struct to choose between the two
+ * underlying implementations.
+ */
+struct r600_resource {
+	struct u_resource		base;
+	struct radeon_bo		*bo;
+	u32				domain;
+	u32				flink;
+	struct pb_buffer		*pb;
+};
+
+struct r600_resource_texture {
+	struct r600_resource		resource;
+	unsigned long			offset[PIPE_MAX_TEXTURE_LEVELS];
+	unsigned long			pitch[PIPE_MAX_TEXTURE_LEVELS];
+	unsigned long			layer_size[PIPE_MAX_TEXTURE_LEVELS];
+	unsigned long			pitch_override;
+	unsigned long			bpt;
+	unsigned long			size;
+};
+
 void r600_init_context_resource_functions(struct r600_context *r600);
 void r600_init_screen_resource_functions(struct r600_screen *r600screen);
 
+/* r600_buffer */
+u32 r600_domain_from_usage(unsigned usage);
+
+/* r600_texture */
+struct pipe_resource *r600_texture_create(struct pipe_screen *screen,
+					const struct pipe_resource *templ);
+struct pipe_resource *r600_texture_from_handle(struct pipe_screen *screen,
+						const struct pipe_resource *base,
+						struct winsys_handle *whandle);
+
 #endif
diff --git a/src/gallium/drivers/r600/r600_screen.c b/src/gallium/drivers/r600/r600_screen.c
index dec6fa8d272..cdaca9ed7db 100644
--- a/src/gallium/drivers/r600/r600_screen.c
+++ b/src/gallium/drivers/r600/r600_screen.c
@@ -24,15 +24,15 @@
  *      Jerome Glisse
  *      Corbin Simpson
  */
-#include <util/u_inlines.h>
-#include <util/u_format.h>
-#include <util/u_memory.h>
-#include "r600_resource.h"
+#include <stdio.h>
+#include "util/u_inlines.h"
+#include "util/u_format.h"
+#include "util/u_memory.h"
 #include "r600_screen.h"
-#include "r600_texture.h"
 #include "r600_context.h"
 #include "r600_public.h"
-#include <stdio.h>
+#include "r600_resource.h"
+#include "r600_state_inlines.h"
 
 static const char* r600_get_vendor(struct pipe_screen* pscreen)
 {
@@ -53,61 +53,102 @@ static const char* r600_get_name(struct pipe_screen* pscreen)
 static int r600_get_param(struct pipe_screen* pscreen, enum pipe_cap param)
 {
 	switch (param) {
-	case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS:
-	case PIPE_CAP_MAX_COMBINED_SAMPLERS:
-		return 16;
+	/* Supported features (boolean caps). */
 	case PIPE_CAP_NPOT_TEXTURES:
-		return 1;
 	case PIPE_CAP_TWO_SIDED_STENCIL:
-		return 1;
 	case PIPE_CAP_GLSL:
-		return 1;
 	case PIPE_CAP_DUAL_SOURCE_BLEND:
-		return 1;
 	case PIPE_CAP_ANISOTROPIC_FILTER:
-		return 1;
 	case PIPE_CAP_POINT_SPRITE:
-		return 1;
-	case PIPE_CAP_MAX_RENDER_TARGETS:
-		/* FIXME some r6xx are buggy and can only do 4 */
-		return 8;
 	case PIPE_CAP_OCCLUSION_QUERY:
-		return 1;
 	case PIPE_CAP_TEXTURE_SHADOW_MAP:
-		return 1;
-	case PIPE_CAP_MAX_TEXTURE_2D_LEVELS:
-	case PIPE_CAP_MAX_TEXTURE_3D_LEVELS:
-	case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS:
-		/* FIXME not sure here */
-		return 13;
 	case PIPE_CAP_TEXTURE_MIRROR_CLAMP:
-		return 1;
 	case PIPE_CAP_TEXTURE_MIRROR_REPEAT:
-		return 1;
-	case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS:
-		/* FIXME allow this once infrastructure is there */
-		return 0;
-	case PIPE_CAP_TGSI_CONT_SUPPORTED:
-		return 0;
 	case PIPE_CAP_BLEND_EQUATION_SEPARATE:
-		return 1;
 	case PIPE_CAP_SM3:
-		return 1;
+	case PIPE_CAP_TEXTURE_SWIZZLE:
 	case PIPE_CAP_INDEP_BLEND_ENABLE:
-		return 1;
-	case PIPE_CAP_INDEP_BLEND_FUNC:
-		/* FIXME allow this */
-		return 0;
 	case PIPE_CAP_DEPTHSTENCIL_CLEAR_SEPARATE:
 		return 1;
+
+	/* Unsupported features (boolean caps). */
+	case PIPE_CAP_TIMER_QUERY:
+	case PIPE_CAP_TGSI_CONT_SUPPORTED:
+	case PIPE_CAP_STREAM_OUTPUT:
+	case PIPE_CAP_INDEP_BLEND_FUNC: /* FIXME allow this */
+	case PIPE_CAP_GEOMETRY_SHADER4:
+	case PIPE_CAP_DEPTH_CLAMP: /* FIXME allow this */
+		return 0;
+
+	/* Texturing. */
+	case PIPE_CAP_MAX_TEXTURE_2D_LEVELS:
+	case PIPE_CAP_MAX_TEXTURE_3D_LEVELS:
+	case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS:
+		return 14;
+	case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS:
+		/* FIXME allow this once infrastructure is there */
+		return 0;
+	case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS:
+	case PIPE_CAP_MAX_COMBINED_SAMPLERS:
+		return 16;
+
+	/* Render targets. */
+	case PIPE_CAP_MAX_RENDER_TARGETS:
+		/* FIXME some r6xx are buggy and can only do 4 */
+		return 8;
+
+	/* Fragment coordinate conventions. */
 	case PIPE_CAP_TGSI_FS_COORD_ORIGIN_UPPER_LEFT:
 	case PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_HALF_INTEGER:
 		return 1;
 	case PIPE_CAP_TGSI_FS_COORD_ORIGIN_LOWER_LEFT:
 	case PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_INTEGER:
 		return 0;
+
+	/* Shader limits. */
+	case PIPE_CAP_MAX_VS_INSTRUCTIONS:
+		return 16384;  //max native instructions, not greater than max instructions
+	case PIPE_CAP_MAX_VS_ALU_INSTRUCTIONS:
+	case PIPE_CAP_MAX_VS_TEX_INSTRUCTIONS:
+	case PIPE_CAP_MAX_VS_TEX_INDIRECTIONS:
+		return 16384;
+	case PIPE_CAP_MAX_FS_INSTRUCTIONS:
+		return 16384; //max program native instructions
+	case PIPE_CAP_MAX_FS_ALU_INSTRUCTIONS:
+		return 16384; //max program native ALU instructions
+	case PIPE_CAP_MAX_FS_TEX_INSTRUCTIONS:
+		return 16384; //max program native texture instructions
+	case PIPE_CAP_MAX_FS_TEX_INDIRECTIONS:
+		return 2048; //max program native texture indirections
+	case PIPE_CAP_MAX_VS_CONTROL_FLOW_DEPTH:
+	case PIPE_CAP_MAX_FS_CONTROL_FLOW_DEPTH:
+		return 8; /* FIXME */
+	case PIPE_CAP_MAX_VS_INPUTS:
+		return 16; //max native attributes
+	case PIPE_CAP_MAX_FS_INPUTS:
+		return 10; //max native attributes
+	case PIPE_CAP_MAX_VS_TEMPS:
+		return 256; //max native temporaries
+	case PIPE_CAP_MAX_FS_TEMPS:
+		return 256; //max native temporaries
+	case PIPE_CAP_MAX_VS_ADDRS:
+	case PIPE_CAP_MAX_FS_ADDRS:
+		return 1; //max native address registers/* FIXME Isn't this equal to TEMPS? */
+	case PIPE_CAP_MAX_VS_CONSTS:
+		return 256; //max native parameters
+	case PIPE_CAP_MAX_FS_CONSTS:
+		return 256; //max program native parameters
+	case PIPE_CAP_MAX_CONST_BUFFERS:
+		return 1;
+	case PIPE_CAP_MAX_CONST_BUFFER_SIZE: /* in bytes */
+		return 4096;
+	case PIPE_CAP_MAX_PREDICATE_REGISTERS:
+	case PIPE_CAP_MAX_VS_PREDS:
+	case PIPE_CAP_MAX_FS_PREDS:
+		return 0; /* FIXME */
+
 	default:
-		debug_printf("r600: unknown param %d\n", param);
+		R600_ERR("r600: unknown param %d\n", param);
 		return 0;
 	}
 }
@@ -125,7 +166,7 @@ static float r600_get_paramf(struct pipe_screen* pscreen, enum pipe_cap param)
 	case PIPE_CAP_MAX_TEXTURE_LOD_BIAS:
 		return 16.0f;
 	default:
-		debug_printf("r600: unsupported paramf %d\n", param);
+		R600_ERR("r600: unsupported paramf %d\n", param);
 		return 0.0f;
 	}
 }
@@ -134,108 +175,51 @@ static boolean r600_is_format_supported(struct pipe_screen* screen,
 					enum pipe_format format,
 					enum pipe_texture_target target,
 					unsigned sample_count,
-					unsigned bindings,
+					unsigned usage,
 					unsigned geom_flags)
 {
+	unsigned retval = 0;
 	if (target >= PIPE_MAX_TEXTURE_TYPES) {
-		debug_printf("r600: unsupported texture type %d\n", target);
+		R600_ERR("r600: unsupported texture type %d\n", target);
 		return FALSE;
 	}
-	switch (format) {
-	case PIPE_FORMAT_B4G4R4A4_UNORM:
-	case PIPE_FORMAT_B5G6R5_UNORM:
-	case PIPE_FORMAT_B5G5R5A1_UNORM:
-	case PIPE_FORMAT_A8_UNORM:
-	case PIPE_FORMAT_L8_UNORM:
-	case PIPE_FORMAT_A8R8G8B8_SRGB:
-	case PIPE_FORMAT_R8G8B8A8_SRGB:
-	case PIPE_FORMAT_DXT1_RGB:
-	case PIPE_FORMAT_DXT1_RGBA:
-	case PIPE_FORMAT_DXT3_RGBA:
-	case PIPE_FORMAT_DXT5_RGBA:
-	case PIPE_FORMAT_UYVY:
-	case PIPE_FORMAT_L8_SRGB:
-	case PIPE_FORMAT_L8A8_SRGB:
-	case PIPE_FORMAT_L8A8_UNORM:
-	case PIPE_FORMAT_A8R8G8B8_UNORM:
-	case PIPE_FORMAT_X8R8G8B8_UNORM:
-	case PIPE_FORMAT_R8G8B8A8_UNORM:
-	case PIPE_FORMAT_R8G8B8X8_UNORM:
-	case PIPE_FORMAT_B8G8R8A8_UNORM:
-	case PIPE_FORMAT_B8G8R8X8_UNORM:
-	case PIPE_FORMAT_A8B8G8R8_SRGB:
-	case PIPE_FORMAT_B8G8R8A8_SRGB:
-	case PIPE_FORMAT_I8_UNORM:
-	case PIPE_FORMAT_Z16_UNORM:
-	case PIPE_FORMAT_X8Z24_UNORM:
-	case PIPE_FORMAT_S8_USCALED_Z24_UNORM:
-	case PIPE_FORMAT_Z32_UNORM:
-	case PIPE_FORMAT_Z24_UNORM_S8_USCALED:
-	case PIPE_FORMAT_Z24X8_UNORM:
-		return TRUE;
-	default:
-		/* Unknown format... */
-		break;
-	}
-	return FALSE;
-}
-
-struct pipe_transfer* r600_texture_get_transfer(struct pipe_context *ctx,
-						struct pipe_resource *texture,
-						struct pipe_subresource sr,
-						unsigned usage,
-						const struct pipe_box *box)
-{
-	struct r600_texture *rtex = (struct r600_texture*)texture;
-	struct r600_transfer *trans;
 
-	trans = CALLOC_STRUCT(r600_transfer);
-	if (trans == NULL)
-		return NULL;
-	pipe_resource_reference(&trans->transfer.resource, texture);
-	trans->transfer.sr = sr;
-	trans->transfer.usage = usage;
-	trans->transfer.box = *box;
-	trans->transfer.stride = rtex->stride[sr.level];
-	trans->offset = r600_texture_get_offset(rtex, sr.level, box->z, sr.face);
-	return &trans->transfer;
-}
-
-void r600_texture_transfer_destroy(struct pipe_context *ctx,
-				   struct pipe_transfer *trans)
-{
-	pipe_resource_reference(&trans->resource, NULL);
-	FREE(trans);
-}
+	/* Multisample */
+	if (sample_count > 1)
+		return FALSE;
 
-void* r600_texture_transfer_map(struct pipe_context *ctx,
-				struct pipe_transfer* transfer)
-{
-	struct r600_transfer *rtransfer = (struct r600_transfer*)transfer;
-	struct r600_texture *rtex = (struct r600_texture*)transfer->resource;
-	char *map;
-	enum pipe_format format = rtex->b.b.format;
+	if ((usage & PIPE_BIND_SAMPLER_VIEW) &&
+	    r600_is_sampler_format_supported(format)) {
+		retval |= PIPE_BIND_SAMPLER_VIEW;
+	}
 
-	map = pipe_buffer_map(ctx, rtex->buffer,
-			      transfer->usage,
-			      &rtransfer->buffer_transfer);
+	if ((usage & (PIPE_BIND_RENDER_TARGET |
+                  PIPE_BIND_DISPLAY_TARGET |
+                  PIPE_BIND_SCANOUT |
+                  PIPE_BIND_SHARED)) &&
+	    r600_is_colorbuffer_format_supported(format)) {
+		retval |= usage &
+			(PIPE_BIND_RENDER_TARGET |
+			 PIPE_BIND_DISPLAY_TARGET |
+			 PIPE_BIND_SCANOUT |
+			 PIPE_BIND_SHARED);
+	}
 
-	if (!map) {
-		return NULL;
+	if ((usage & PIPE_BIND_DEPTH_STENCIL) &&
+	    r600_is_zs_format_supported(format)) {
+		retval |= PIPE_BIND_DEPTH_STENCIL;
 	}
 
-	return map + rtransfer->offset +
-		transfer->box.y / util_format_get_blockheight(format) * transfer->stride +
-		transfer->box.x / util_format_get_blockwidth(format) * util_format_get_blocksize(format);
-}
+	if ((usage & PIPE_BIND_VERTEX_BUFFER) &&
+	    r600_is_vertex_format_supported(format))
+		retval |= PIPE_BIND_VERTEX_BUFFER;
 
-void r600_texture_transfer_unmap(struct pipe_context *ctx,
-				 struct pipe_transfer* transfer)
-{
-	struct r600_transfer *rtransfer = (struct r600_transfer*)transfer;
-	struct r600_texture *rtex = (struct r600_texture*)transfer->resource;
+	if (usage & PIPE_BIND_TRANSFER_READ)
+		retval |= PIPE_BIND_TRANSFER_READ;
+	if (usage & PIPE_BIND_TRANSFER_WRITE)
+		retval |= PIPE_BIND_TRANSFER_WRITE;
 
-	pipe_buffer_unmap(ctx, rtex->buffer, rtransfer->buffer_transfer);
+	return retval == usage;
 }
 
 static void r600_destroy_screen(struct pipe_screen* pscreen)
diff --git a/src/gallium/drivers/r600/r600_screen.h b/src/gallium/drivers/r600/r600_screen.h
index 0a0286d96bc..53b560c617f 100644
--- a/src/gallium/drivers/r600/r600_screen.h
+++ b/src/gallium/drivers/r600/r600_screen.h
@@ -40,14 +40,6 @@ struct r600_transfer {
 	unsigned			offset;
 };
 
-struct r600_buffer {
-	struct u_resource		b;
-	struct radeon_bo		*bo;
-	u32				domain;
-	u32				flink;
-	struct pb_buffer		*pb;
-};
-
 struct r600_screen {
 	struct pipe_screen		screen;
 	struct radeon			*rw;
@@ -88,17 +80,6 @@ void r600_texture_transfer_unmap(struct pipe_context *ctx,
 int r600_conv_pipe_format(unsigned pformat, unsigned *format);
 int r600_conv_pipe_prim(unsigned pprim, unsigned *prim);
 
-union r600_float_to_u32_u {
-	u32	u;
-	float	f;
-};
-
-static inline u32 r600_float_to_u32(float f)
-{
-	union r600_float_to_u32_u c;
-
-	c.f = f;
-	return c.u;
-}
+void r600_init_screen_texture_functions(struct pipe_screen *screen);
 
 #endif
diff --git a/src/gallium/drivers/r600/r600_shader.c b/src/gallium/drivers/r600/r600_shader.c
index f7d6e106638..956c7e7930c 100644
--- a/src/gallium/drivers/r600/r600_shader.c
+++ b/src/gallium/drivers/r600/r600_shader.c
@@ -19,78 +19,190 @@
  * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
  * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
  * USE OR OTHER DEALINGS IN THE SOFTWARE.
- *
- * Authors:
- *      Jerome Glisse
  */
-#include <stdio.h>
-#include <errno.h>
-#include <util/u_inlines.h>
-#include <util/u_format.h>
-#include <util/u_memory.h>
-#include <tgsi/tgsi_dump.h>
+#include "pipe/p_shader_tokens.h"
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_scan.h"
+#include "tgsi/tgsi_dump.h"
+#include "util/u_format.h"
 #include "r600_screen.h"
 #include "r600_context.h"
+#include "r600_shader.h"
+#include "r600_asm.h"
+#include "r600_sq.h"
 #include "r600d.h"
+#include <stdio.h>
+#include <errno.h>
+
+
+struct r600_shader_tgsi_instruction;
+
+struct r600_shader_ctx {
+	struct tgsi_shader_info			info;
+	struct tgsi_parse_context		parse;
+	const struct tgsi_token			*tokens;
+	unsigned				type;
+	unsigned				file_offset[TGSI_FILE_COUNT];
+	unsigned				temp_reg;
+	struct r600_shader_tgsi_instruction	*inst_info;
+	struct r600_bc				*bc;
+	struct r600_shader			*shader;
+	u32					value[4];
+};
+
+struct r600_shader_tgsi_instruction {
+	unsigned	tgsi_opcode;
+	unsigned	is_op3;
+	unsigned	r600_opcode;
+	int (*process)(struct r600_shader_ctx *ctx);
+};
+
+static struct r600_shader_tgsi_instruction r600_shader_tgsi_instruction[];
+static int r600_shader_from_tgsi(const struct tgsi_token *tokens, struct r600_shader *shader);
+
+static int r600_shader_update(struct pipe_context *ctx, struct r600_shader *shader)
+{
+	struct r600_context *rctx = r600_context(ctx);
+	const struct util_format_description *desc;
+	enum pipe_format resource_format[160];
+	unsigned i, nresources = 0;
+	struct r600_bc *bc = &shader->bc;
+	struct r600_bc_cf *cf;
+	struct r600_bc_vtx *vtx;
+
+	if (shader->processor_type != TGSI_PROCESSOR_VERTEX)
+		return 0;
+	for (i = 0; i < rctx->vertex_elements->count; i++) {
+		resource_format[nresources++] = rctx->vertex_elements->elements[i].src_format;
+	}
+	LIST_FOR_EACH_ENTRY(cf, &bc->cf, list) {
+		switch (cf->inst) {
+		case V_SQ_CF_WORD1_SQ_CF_INST_VTX:
+		case V_SQ_CF_WORD1_SQ_CF_INST_VTX_TC:
+			LIST_FOR_EACH_ENTRY(vtx, &cf->vtx, list) {
+				desc = util_format_description(resource_format[vtx->buffer_id]);
+				if (desc == NULL) {
+					R600_ERR("unknown format %d\n", resource_format[vtx->buffer_id]);
+					return -EINVAL;
+				}
+				vtx->dst_sel_x = desc->swizzle[0];
+				vtx->dst_sel_y = desc->swizzle[1];
+				vtx->dst_sel_z = desc->swizzle[2];
+				vtx->dst_sel_w = desc->swizzle[3];
+			}
+			break;
+		default:
+			break;
+		}
+	}
+	return r600_bc_build(&shader->bc);
+}
+
+int r600_pipe_shader_create(struct pipe_context *ctx,
+			struct r600_context_state *rpshader,
+			const struct tgsi_token *tokens)
+{
+	struct r600_screen *rscreen = r600_screen(ctx->screen);
+	int r;
+
+fprintf(stderr, "--------------------------------------------------------------\n");
+tgsi_dump(tokens, 0);
+	if (rpshader == NULL)
+		return -ENOMEM;
+	rpshader->shader.family = radeon_get_family(rscreen->rw);
+	r = r600_shader_from_tgsi(tokens, &rpshader->shader);
+	if (r) {
+		R600_ERR("translation from TGSI failed !\n");
+		return r;
+	}
+	r = r600_bc_build(&rpshader->shader.bc);
+	if (r) {
+		R600_ERR("building bytecode failed !\n");
+		return r;
+	}
+fprintf(stderr, "______________________________________________________________\n");
+	return 0;
+}
 
-static int r600_pipe_shader_vs(struct pipe_context *ctx, struct r600_pipe_shader *rpshader)
+static int r600_pipe_shader_vs(struct pipe_context *ctx, struct r600_context_state *rpshader)
 {
 	struct r600_screen *rscreen = r600_screen(ctx->screen);
 	struct r600_shader *rshader = &rpshader->shader;
 	struct radeon_state *state;
 	unsigned i, tmp;
 
-	rpshader->state = radeon_state_decref(rpshader->state);
+	rpshader->rstate = radeon_state_decref(rpshader->rstate);
 	state = radeon_state(rscreen->rw, R600_VS_SHADER_TYPE, R600_VS_SHADER);
 	if (state == NULL)
 		return -ENOMEM;
-	for (i = 0; i < rshader->noutput; i += 4) {
-		tmp = rshader->output[i].sid;
-		tmp |= rshader->output[i + 1].sid << 8;
-		tmp |= rshader->output[i + 2].sid << 16;
-		tmp |= rshader->output[i + 3].sid << 24;
-		state->states[R600_VS_SHADER__SPI_VS_OUT_ID_0 + i / 4] = tmp;
-	}
-	state->states[R600_VS_SHADER__SPI_VS_OUT_CONFIG] = S_0286C4_VS_EXPORT_COUNT(rshader->noutput - 1);
-	state->states[R600_VS_SHADER__SQ_PGM_RESOURCES_VS] = S_028868_NUM_GPRS(rshader->ngpr);
-	rpshader->state = state;
-	rpshader->state->bo[0] = radeon_bo_incref(rscreen->rw, rpshader->bo);
-	rpshader->state->bo[1] = radeon_bo_incref(rscreen->rw, rpshader->bo);
-	rpshader->state->nbo = 2;
-	rpshader->state->placement[0] = RADEON_GEM_DOMAIN_GTT;
+	for (i = 0; i < 10; i++) {
+		state->states[R600_VS_SHADER__SPI_VS_OUT_ID_0 + i] = 0;
+	}
+	/* so far never got proper semantic id from tgsi */
+	for (i = 0; i < 32; i++) {
+		tmp = i << ((i & 3) * 8);
+		state->states[R600_VS_SHADER__SPI_VS_OUT_ID_0 + i / 4] |= tmp;
+	}
+	state->states[R600_VS_SHADER__SPI_VS_OUT_CONFIG] = S_0286C4_VS_EXPORT_COUNT(rshader->noutput - 2);
+	state->states[R600_VS_SHADER__SQ_PGM_RESOURCES_VS] = S_028868_NUM_GPRS(rshader->bc.ngpr);
+	rpshader->rstate = state;
+	rpshader->rstate->bo[0] = radeon_bo_incref(rscreen->rw, rpshader->bo);
+	rpshader->rstate->bo[1] = radeon_bo_incref(rscreen->rw, rpshader->bo);
+	rpshader->rstate->nbo = 2;
+	rpshader->rstate->placement[0] = RADEON_GEM_DOMAIN_GTT;
 	return radeon_state_pm4(state);
 }
 
-static int r600_pipe_shader_ps(struct pipe_context *ctx, struct r600_pipe_shader *rpshader)
+static int r600_pipe_shader_ps(struct pipe_context *ctx, struct r600_context_state *rpshader)
 {
+	const struct pipe_rasterizer_state *rasterizer;
 	struct r600_screen *rscreen = r600_screen(ctx->screen);
 	struct r600_shader *rshader = &rpshader->shader;
+	struct r600_context *rctx = r600_context(ctx);
 	struct radeon_state *state;
-	unsigned i, tmp;
+	unsigned i, tmp, exports_ps, num_cout;
 
-	rpshader->state = radeon_state_decref(rpshader->state);
+	rasterizer = &rctx->rasterizer->state.rasterizer;
+	rpshader->rstate = radeon_state_decref(rpshader->rstate);
 	state = radeon_state(rscreen->rw, R600_PS_SHADER_TYPE, R600_PS_SHADER);
 	if (state == NULL)
 		return -ENOMEM;
 	for (i = 0; i < rshader->ninput; i++) {
-		tmp = S_028644_SEMANTIC(rshader->input[i].sid);
+		tmp = S_028644_SEMANTIC(i);
 		tmp |= S_028644_SEL_CENTROID(1);
-		tmp |= S_028644_FLAT_SHADE(rshader->flat_shade);
+		if (rshader->input[i].name == TGSI_SEMANTIC_COLOR ||
+			rshader->input[i].name == TGSI_SEMANTIC_BCOLOR) {
+			tmp |= S_028644_FLAT_SHADE(rshader->flat_shade);
+		}
+		if (rasterizer->sprite_coord_enable & (1 << i)) {
+			tmp |= S_028644_PT_SPRITE_TEX(1);
+		}
 		state->states[R600_PS_SHADER__SPI_PS_INPUT_CNTL_0 + i] = tmp;
 	}
+
+	exports_ps = 0;
+	num_cout = 0;
+	for (i = 0; i < rshader->noutput; i++) {
+		if (rshader->output[i].name == TGSI_SEMANTIC_POSITION)
+			exports_ps |= 1;
+		else if (rshader->output[i].name == TGSI_SEMANTIC_COLOR) {
+			exports_ps |= (1 << (num_cout+1));
+			num_cout++;
+		}
+	}
 	state->states[R600_PS_SHADER__SPI_PS_IN_CONTROL_0] = S_0286CC_NUM_INTERP(rshader->ninput) |
 							S_0286CC_PERSP_GRADIENT_ENA(1);
 	state->states[R600_PS_SHADER__SPI_PS_IN_CONTROL_1] = 0x00000000;
-	state->states[R600_PS_SHADER__SQ_PGM_RESOURCES_PS] = S_028868_NUM_GPRS(rshader->ngpr);
-	state->states[R600_PS_SHADER__SQ_PGM_EXPORTS_PS] = 0x00000002;
-	rpshader->state = state;
-	rpshader->state->bo[0] = radeon_bo_incref(rscreen->rw, rpshader->bo);
-	rpshader->state->nbo = 1;
-	rpshader->state->placement[0] = RADEON_GEM_DOMAIN_GTT;
+	state->states[R600_PS_SHADER__SQ_PGM_RESOURCES_PS] = S_028868_NUM_GPRS(rshader->bc.ngpr);
+	state->states[R600_PS_SHADER__SQ_PGM_EXPORTS_PS] = exports_ps;
+	rpshader->rstate = state;
+	rpshader->rstate->bo[0] = radeon_bo_incref(rscreen->rw, rpshader->bo);
+	rpshader->rstate->nbo = 1;
+	rpshader->rstate->placement[0] = RADEON_GEM_DOMAIN_GTT;
 	return radeon_state_pm4(state);
 }
 
-static int r600_pipe_shader(struct pipe_context *ctx, struct r600_pipe_shader *rpshader)
+static int r600_pipe_shader(struct pipe_context *ctx, struct r600_context_state *rpshader)
 {
 	struct r600_screen *rscreen = r600_screen(ctx->screen);
 	struct r600_context *rctx = r600_context(ctx);
@@ -100,21 +212,21 @@ static int r600_pipe_shader(struct pipe_context *ctx, struct r600_pipe_shader *r
 	/* copy new shader */
 	radeon_bo_decref(rscreen->rw, rpshader->bo);
 	rpshader->bo = NULL;
-	rpshader->bo = radeon_bo(rscreen->rw, 0, rshader->ndw * 4,
+	rpshader->bo = radeon_bo(rscreen->rw, 0, rshader->bc.ndw * 4,
 				4096, NULL);
 	if (rpshader->bo == NULL) {
 		return -ENOMEM;
 	}
 	radeon_bo_map(rscreen->rw, rpshader->bo);
-	memcpy(rpshader->bo->data, rshader->bcode, rshader->ndw * 4);
+	memcpy(rpshader->bo->data, rshader->bc.bytecode, rshader->bc.ndw * 4);
 	radeon_bo_unmap(rscreen->rw, rpshader->bo);
 	/* build state */
 	rshader->flat_shade = rctx->flat_shade;
-	switch (rpshader->type) {
-	case C_PROGRAM_TYPE_VS:
+	switch (rshader->processor_type) {
+	case TGSI_PROCESSOR_VERTEX:
 		r = r600_pipe_shader_vs(ctx, rpshader);
 		break;
-	case C_PROGRAM_TYPE_FS:
+	case TGSI_PROCESSOR_FRAGMENT:
 		r = r600_pipe_shader_ps(ctx, rpshader);
 		break;
 	default:
@@ -124,104 +236,1129 @@ static int r600_pipe_shader(struct pipe_context *ctx, struct r600_pipe_shader *r
 	return r;
 }
 
-struct r600_pipe_shader *r600_pipe_shader_create(struct pipe_context *ctx, unsigned type, const struct tgsi_token *tokens)
+int r600_pipe_shader_update(struct pipe_context *ctx, struct r600_context_state *rpshader)
 {
-	struct r600_screen *rscreen = r600_screen(ctx->screen);
-	struct r600_pipe_shader *rpshader = CALLOC_STRUCT(r600_pipe_shader);
-	struct r600_shader *rshader = &rpshader->shader;
+	struct r600_context *rctx = r600_context(ctx);
 	int r;
-	enum radeon_family family;
 
 	if (rpshader == NULL)
-		return NULL;
-	rpshader->type = type;
-	family = radeon_get_family(rscreen->rw);
-	rshader->r6xx_compile = (family >= CHIP_R600 && family < CHIP_RV770);
-	LIST_INITHEAD(&rshader->nodes);
-	fprintf(stderr, "<< %s\n", rshader->r6xx_compile ? "R600" : "R700");
-	tgsi_dump(tokens, 0);
-	fprintf(stderr, "--------------------------------------------------------------\n");
-	r = c_shader_from_tgsi(&rshader->cshader, type, tokens);
-	if (r) {
-		r600_pipe_shader_destroy(ctx, rpshader);
-		fprintf(stderr, "ERROR(%s %d)>>\n\n", __func__, __LINE__);
-		return NULL;
+		return -EINVAL;
+	/* there should be enough input */
+	if (rctx->vertex_elements->count < rpshader->shader.bc.nresource) {
+		R600_ERR("%d resources provided, expecting %d\n",
+			rctx->vertex_elements->count, rpshader->shader.bc.nresource);
+		return -EINVAL;
 	}
-	r = r600_shader_insert_fetch(&rshader->cshader);
-	if (r) {
-		r600_pipe_shader_destroy(ctx, rpshader);
-		fprintf(stderr, "ERROR(%s %d)>>\n\n", __func__, __LINE__);
-		return NULL;
+	r = r600_shader_update(ctx, &rpshader->shader);
+	if (r)
+		return r;
+	return r600_pipe_shader(ctx, rpshader);
+}
+
+static int tgsi_is_supported(struct r600_shader_ctx *ctx)
+{
+	struct tgsi_full_instruction *i = &ctx->parse.FullToken.FullInstruction;
+	int j;
+
+	if (i->Instruction.NumDstRegs > 1) {
+		R600_ERR("too many dst (%d)\n", i->Instruction.NumDstRegs);
+		return -EINVAL;
 	}
-	r = c_shader_build_dominator_tree(&rshader->cshader);
-	if (r) {
-		r600_pipe_shader_destroy(ctx, rpshader);
-		fprintf(stderr, "ERROR(%s %d)>>\n\n", __func__, __LINE__);
-		return NULL;
+	if (i->Instruction.Predicate) {
+		R600_ERR("predicate unsupported\n");
+		return -EINVAL;
 	}
-	r = r600_cshader_legalize(&rshader->cshader);
-	if (r) {
-		r600_pipe_shader_destroy(ctx, rpshader);
-		fprintf(stderr, "ERROR(%s %d)>>\n\n", __func__, __LINE__);
-		return NULL;
+	if (i->Instruction.Label) {
+		R600_ERR("label unsupported\n");
+		return -EINVAL;
 	}
-	c_shader_dump(&rshader->cshader);
-	r = r700_shader_translate(rshader);
-	if (r) {
-		r600_pipe_shader_destroy(ctx, rpshader);
-		fprintf(stderr, "ERROR(%s %d)>>\n\n", __func__, __LINE__);
-		return NULL;
+	for (j = 0; j < i->Instruction.NumSrcRegs; j++) {
+		if (i->Src[j].Register.Indirect ||
+			i->Src[j].Register.Dimension ||
+			i->Src[j].Register.Absolute) {
+			R600_ERR("unsupported src (indirect|dimension|absolute)\n");
+			return -EINVAL;
+		}
 	}
-#if 1
-#if 0
-	fprintf(stderr, "--------------------------------------------------------------\n");
-	for (int i = 0; i < rshader->ndw; i++) {
-		fprintf(stderr, "0x%08X\n", rshader->bcode[i]);
+	for (j = 0; j < i->Instruction.NumDstRegs; j++) {
+		if (i->Dst[j].Register.Indirect || i->Dst[j].Register.Dimension) {
+			R600_ERR("unsupported dst (indirect|dimension)\n");
+			return -EINVAL;
+		}
 	}
-#endif
-	fprintf(stderr, ">>\n\n");
-#endif
-	return rpshader;
+	return 0;
 }
 
-void r600_pipe_shader_destroy(struct pipe_context *ctx, struct r600_pipe_shader *rpshader)
+static int tgsi_declaration(struct r600_shader_ctx *ctx)
 {
-	struct r600_screen *rscreen = r600_screen(ctx->screen);
+	struct tgsi_full_declaration *d = &ctx->parse.FullToken.FullDeclaration;
+	struct r600_bc_vtx vtx;
+	unsigned i;
+	int r;
 
-	if (rpshader == NULL)
-		return;
-	radeon_bo_decref(rscreen->rw, rpshader->bo);
-	rpshader->bo = NULL;
-	r600_shader_cleanup(&rpshader->shader);
-	FREE(rpshader);
+	switch (d->Declaration.File) {
+	case TGSI_FILE_INPUT:
+		i = ctx->shader->ninput++;
+		ctx->shader->input[i].name = d->Semantic.Name;
+		ctx->shader->input[i].sid = d->Semantic.Index;
+		ctx->shader->input[i].interpolate = d->Declaration.Interpolate;
+		ctx->shader->input[i].gpr = ctx->file_offset[TGSI_FILE_INPUT] + i;
+		if (ctx->type == TGSI_PROCESSOR_VERTEX) {
+			/* turn input into fetch */
+			memset(&vtx, 0, sizeof(struct r600_bc_vtx));
+			vtx.inst = 0;
+			vtx.fetch_type = 0;
+			vtx.buffer_id = i;
+			/* register containing the index into the buffer */
+			vtx.src_gpr = 0;
+			vtx.src_sel_x = 0;
+			vtx.mega_fetch_count = 0x1F;
+			vtx.dst_gpr = ctx->shader->input[i].gpr;
+			vtx.dst_sel_x = 0;
+			vtx.dst_sel_y = 1;
+			vtx.dst_sel_z = 2;
+			vtx.dst_sel_w = 3;
+			r = r600_bc_add_vtx(ctx->bc, &vtx);
+			if (r)
+				return r;
+		}
+		break;
+	case TGSI_FILE_OUTPUT:
+		i = ctx->shader->noutput++;
+		ctx->shader->output[i].name = d->Semantic.Name;
+		ctx->shader->output[i].sid = d->Semantic.Index;
+		ctx->shader->output[i].gpr = ctx->file_offset[TGSI_FILE_OUTPUT] + i;
+		ctx->shader->output[i].interpolate = d->Declaration.Interpolate;
+		break;
+	case TGSI_FILE_CONSTANT:
+	case TGSI_FILE_TEMPORARY:
+	case TGSI_FILE_SAMPLER:
+		break;
+	default:
+		R600_ERR("unsupported file %d declaration\n", d->Declaration.File);
+		return -EINVAL;
+	}
+	return 0;
 }
 
-int r600_pipe_shader_update(struct pipe_context *ctx, struct r600_pipe_shader *rpshader)
+int r600_shader_from_tgsi(const struct tgsi_token *tokens, struct r600_shader *shader)
 {
-	struct r600_context *rctx = r600_context(ctx);
-	struct r600_shader *rshader;
-	enum pipe_format resource_format[160];
-	unsigned i, nresources = 0;
-	int r;
+	struct tgsi_full_immediate *immediate;
+	struct r600_shader_ctx ctx;
+	struct r600_bc_output output[32];
+	unsigned output_done, noutput;
+	unsigned opcode;
+	int i, r = 0, pos0;
 
-	if (rpshader == NULL)
-		return -EINVAL;
-	rshader = &rpshader->shader;
-	switch (rpshader->type) {
-	case C_PROGRAM_TYPE_VS:
-		for (i = 0; i < rctx->vertex_elements->count; i++) {
-			resource_format[nresources++] = rctx->vertex_elements->elements[i].src_format;
+	ctx.bc = &shader->bc;
+	ctx.shader = shader;
+	r = r600_bc_init(ctx.bc, shader->family);
+	if (r)
+		return r;
+	ctx.tokens = tokens;
+	tgsi_scan_shader(tokens, &ctx.info);
+	tgsi_parse_init(&ctx.parse, tokens);
+	ctx.type = ctx.parse.FullHeader.Processor.Processor;
+	shader->processor_type = ctx.type;
+
+	/* register allocations */
+	/* Values [0,127] correspond to GPR[0..127]. 
+	 * Values [256,511] correspond to cfile constants c[0..255]. 
+	 * Other special values are shown in the list below.
+	 * 248	SQ_ALU_SRC_0: special constant 0.0.
+	 * 249	SQ_ALU_SRC_1: special constant 1.0 float.
+	 * 250	SQ_ALU_SRC_1_INT: special constant 1 integer.
+	 * 251	SQ_ALU_SRC_M_1_INT: special constant -1 integer.
+	 * 252	SQ_ALU_SRC_0_5: special constant 0.5 float.
+	 * 253	SQ_ALU_SRC_LITERAL: literal constant.
+	 * 254	SQ_ALU_SRC_PV: previous vector result.
+	 * 255	SQ_ALU_SRC_PS: previous scalar result.
+	 */
+	for (i = 0; i < TGSI_FILE_COUNT; i++) {
+		ctx.file_offset[i] = 0;
+	}
+	if (ctx.type == TGSI_PROCESSOR_VERTEX) {
+		ctx.file_offset[TGSI_FILE_INPUT] = 1;
+	}
+	ctx.file_offset[TGSI_FILE_OUTPUT] = ctx.file_offset[TGSI_FILE_INPUT] +
+						ctx.info.file_count[TGSI_FILE_INPUT];
+	ctx.file_offset[TGSI_FILE_TEMPORARY] = ctx.file_offset[TGSI_FILE_OUTPUT] +
+						ctx.info.file_count[TGSI_FILE_OUTPUT];
+	ctx.file_offset[TGSI_FILE_CONSTANT] = 256;
+	ctx.file_offset[TGSI_FILE_IMMEDIATE] = 253;
+	ctx.temp_reg = ctx.file_offset[TGSI_FILE_TEMPORARY] +
+			ctx.info.file_count[TGSI_FILE_TEMPORARY];
+
+	while (!tgsi_parse_end_of_tokens(&ctx.parse)) {
+		tgsi_parse_token(&ctx.parse);
+		switch (ctx.parse.FullToken.Token.Type) {
+		case TGSI_TOKEN_TYPE_IMMEDIATE:
+			immediate = &ctx.parse.FullToken.FullImmediate;
+			ctx.value[0] = immediate->u[0].Uint;
+			ctx.value[1] = immediate->u[1].Uint;
+			ctx.value[2] = immediate->u[2].Uint;
+			ctx.value[3] = immediate->u[3].Uint;
+			break;
+		case TGSI_TOKEN_TYPE_DECLARATION:
+			r = tgsi_declaration(&ctx);
+			if (r)
+				goto out_err;
+			break;
+		case TGSI_TOKEN_TYPE_INSTRUCTION:
+			r = tgsi_is_supported(&ctx);
+			if (r)
+				goto out_err;
+			opcode = ctx.parse.FullToken.FullInstruction.Instruction.Opcode;
+			ctx.inst_info = &r600_shader_tgsi_instruction[opcode];
+			r = ctx.inst_info->process(&ctx);
+			if (r)
+				goto out_err;
+			r = r600_bc_add_literal(ctx.bc, ctx.value);
+			if (r)
+				goto out_err;
+			break;
+		default:
+			R600_ERR("unsupported token type %d\n", ctx.parse.FullToken.Token.Type);
+			r = -EINVAL;
+			goto out_err;
 		}
-		break;
+	}
+	/* export output */
+	noutput = shader->noutput;
+	for (i = 0, pos0 = 0; i < noutput; i++) {
+		memset(&output[i], 0, sizeof(struct r600_bc_output));
+		output[i].gpr = shader->output[i].gpr;
+		output[i].elem_size = 3;
+		output[i].swizzle_x = 0;
+		output[i].swizzle_y = 1;
+		output[i].swizzle_z = 2;
+		output[i].swizzle_w = 3;
+		output[i].barrier = 1;
+		output[i].type = V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PARAM;
+		output[i].array_base = i - pos0;
+		output[i].inst = V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT;
+		switch (ctx.type) {
+		case TGSI_PROCESSOR_VERTEX:
+			if (shader->output[i].name == TGSI_SEMANTIC_POSITION) {
+				output[i].array_base = 60;
+				output[i].type = V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_POS;
+				/* position doesn't count in array_base */
+				pos0++;
+			}
+			if (shader->output[i].name == TGSI_SEMANTIC_PSIZE) {
+				output[i].array_base = 61;
+				output[i].type = V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_POS;
+				/* position doesn't count in array_base */
+				pos0++;
+			}
+			break;
+		case TGSI_PROCESSOR_FRAGMENT:
+			if (shader->output[i].name == TGSI_SEMANTIC_COLOR) {
+				output[i].array_base = shader->output[i].sid;
+				output[i].type = V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL;
+			} else if (shader->output[i].name == TGSI_SEMANTIC_POSITION) {
+				output[i].array_base = 61;
+				output[i].type = V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL;
+			} else {
+				R600_ERR("unsupported fragment output name %d\n", shader->output[i].name);
+				r = -EINVAL;
+				goto out_err;
+			}
+			break;
+		default:
+			R600_ERR("unsupported processor type %d\n", ctx.type);
+			r = -EINVAL;
+			goto out_err;
+		}
+	}
+	/* add fake param output for vertex shader if no param is exported */
+	if (ctx.type == TGSI_PROCESSOR_VERTEX) {
+		for (i = 0, pos0 = 0; i < noutput; i++) {
+			if (output[i].type == V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PARAM) {
+				pos0 = 1;
+				break;
+			}
+		}
+		if (!pos0) {
+			memset(&output[i], 0, sizeof(struct r600_bc_output));
+			output[i].gpr = 0;
+			output[i].elem_size = 3;
+			output[i].swizzle_x = 0;
+			output[i].swizzle_y = 1;
+			output[i].swizzle_z = 2;
+			output[i].swizzle_w = 3;
+			output[i].barrier = 1;
+			output[i].type = V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PARAM;
+			output[i].array_base = 0;
+			output[i].inst = V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT;
+			noutput++;
+		}
+	}
+	/* add fake pixel export */
+	if (ctx.type == TGSI_PROCESSOR_FRAGMENT && !noutput) {
+		memset(&output[0], 0, sizeof(struct r600_bc_output));
+		output[0].gpr = 0;
+		output[0].elem_size = 3;
+		output[0].swizzle_x = 7;
+		output[0].swizzle_y = 7;
+		output[0].swizzle_z = 7;
+		output[0].swizzle_w = 7;
+		output[0].barrier = 1;
+		output[0].type = V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PARAM;
+		output[0].array_base = 0;
+		output[0].inst = V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT;
+		noutput++;
+	}
+	/* set export done on last export of each type */
+	for (i = noutput - 1, output_done = 0; i >= 0; i--) {
+		if (i == (noutput - 1)) {
+			output[i].end_of_program = 1;
+		}
+		if (!(output_done & (1 << output[i].type))) {
+			output_done |= (1 << output[i].type);
+			output[i].inst = V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT_DONE;
+		}
+	}
+	/* add output to bytecode */
+	for (i = 0; i < noutput; i++) {
+		r = r600_bc_add_output(ctx.bc, &output[i]);
+		if (r)
+			goto out_err;
+	}
+	tgsi_parse_free(&ctx.parse);
+	return 0;
+out_err:
+	tgsi_parse_free(&ctx.parse);
+	return r;
+}
+
+static int tgsi_unsupported(struct r600_shader_ctx *ctx)
+{
+	R600_ERR("%d tgsi opcode unsupported\n", ctx->inst_info->tgsi_opcode);
+	return -EINVAL;
+}
+
+static int tgsi_end(struct r600_shader_ctx *ctx)
+{
+	return 0;
+}
+
+static int tgsi_src(struct r600_shader_ctx *ctx,
+			const struct tgsi_full_src_register *tgsi_src,
+			struct r600_bc_alu_src *r600_src)
+{
+	memset(r600_src, 0, sizeof(struct r600_bc_alu_src));
+	r600_src->sel = tgsi_src->Register.Index;
+	if (tgsi_src->Register.File == TGSI_FILE_IMMEDIATE) {
+		r600_src->sel = 0;
+	}
+	r600_src->neg = tgsi_src->Register.Negate;
+	r600_src->sel += ctx->file_offset[tgsi_src->Register.File];
+	return 0;
+}
+
+static int tgsi_dst(struct r600_shader_ctx *ctx,
+			const struct tgsi_full_dst_register *tgsi_dst,
+			unsigned swizzle,
+			struct r600_bc_alu_dst *r600_dst)
+{
+	struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
+
+	r600_dst->sel = tgsi_dst->Register.Index;
+	r600_dst->sel += ctx->file_offset[tgsi_dst->Register.File];
+	r600_dst->chan = swizzle;
+	r600_dst->write = 1;
+	if (inst->Instruction.Saturate) {
+		r600_dst->clamp = 1;
+	}
+	return 0;
+}
+
+static unsigned tgsi_chan(const struct tgsi_full_src_register *tgsi_src, unsigned swizzle)
+{
+	switch (swizzle) {
+	case 0:
+		return tgsi_src->Register.SwizzleX;
+	case 1:
+		return tgsi_src->Register.SwizzleY;
+	case 2:
+		return tgsi_src->Register.SwizzleZ;
+	case 3:
+		return tgsi_src->Register.SwizzleW;
 	default:
-		break;
+		return 0;
 	}
-	/* there should be enough input */
-	if (nresources < rshader->nresource)
-		return -EINVAL;
-	/* FIXME compare resources */
-	r = r600_shader_update(rshader, resource_format);
+}
+
+static int tgsi_split_constant(struct r600_shader_ctx *ctx, struct r600_bc_alu_src r600_src[3])
+{
+	struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
+	struct r600_bc_alu alu;
+	int i, j, k, nconst, r;
+
+	for (i = 0, nconst = 0; i < inst->Instruction.NumSrcRegs; i++) {
+		if (inst->Src[i].Register.File == TGSI_FILE_CONSTANT) {
+			nconst++;
+		}
+		r = tgsi_src(ctx, &inst->Src[i], &r600_src[i]);
+		if (r) {
+			return r;
+		}
+	}
+	for (i = 0, j = nconst - 1; i < inst->Instruction.NumSrcRegs; i++) {
+		if (inst->Src[j].Register.File == TGSI_FILE_CONSTANT && j > 0) {
+			for (k = 0; k < 4; k++) {
+				memset(&alu, 0, sizeof(struct r600_bc_alu));
+				alu.inst = V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV;
+				alu.src[0].sel = r600_src[0].sel;
+				alu.src[0].chan = k;
+				alu.dst.sel = ctx->temp_reg + j;
+				alu.dst.chan = k;
+				alu.dst.write = 1;
+				if (k == 3)
+					alu.last = 1;
+				r = r600_bc_add_alu(ctx->bc, &alu);
+				if (r)
+					return r;
+			}
+			r600_src[0].sel = ctx->temp_reg + j;
+			j--;
+		}
+	}
+	return 0;
+}
+
+static int tgsi_op2(struct r600_shader_ctx *ctx)
+{
+	struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
+	struct r600_bc_alu_src r600_src[3];
+	struct r600_bc_alu alu;
+	int i, j, r;
+
+	r = tgsi_split_constant(ctx, r600_src);
 	if (r)
 		return r;
-	return r600_pipe_shader(ctx, rpshader);
+	for (i = 0; i < 4; i++) {
+		memset(&alu, 0, sizeof(struct r600_bc_alu));
+		if (!(inst->Dst[0].Register.WriteMask & (1 << i))) {
+			alu.inst = V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP;
+			alu.dst.chan = i;
+		} else {
+			alu.inst = ctx->inst_info->r600_opcode;
+			for (j = 0; j < inst->Instruction.NumSrcRegs; j++) {
+				alu.src[j] = r600_src[j];
+				alu.src[j].chan = tgsi_chan(&inst->Src[j], i);
+			}
+			r = tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst);
+			if (r)
+				return r;
+		}
+		/* handle some special cases */
+		switch (ctx->inst_info->tgsi_opcode) {
+		case TGSI_OPCODE_SUB:
+			alu.src[1].neg = 1;
+			break;
+		case TGSI_OPCODE_ABS:
+			alu.src[0].abs = 1;
+			break;
+		default:
+			break;
+		}
+		if (i == 3) {
+			alu.last = 1;
+		}
+		r = r600_bc_add_alu(ctx->bc, &alu);
+		if (r)
+			return r;
+	}
+	return 0;
+}
+
+static int tgsi_kill(struct r600_shader_ctx *ctx)
+{
+	struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
+	struct r600_bc_alu alu;
+	int i, r;
+
+	for (i = 0; i < 4; i++) {
+		memset(&alu, 0, sizeof(struct r600_bc_alu));
+		alu.inst = ctx->inst_info->r600_opcode;
+		alu.dst.chan = i;
+		alu.src[0].sel = 248;
+		r = tgsi_src(ctx, &inst->Src[0], &alu.src[1]);
+		if (r)
+			return r;
+		alu.src[1].chan = tgsi_chan(&inst->Src[0], i);
+		if (i == 3) {
+			alu.last = 1;
+		}
+		r = r600_bc_add_alu(ctx->bc, &alu);
+		if (r)
+			return r;
+	}
+	return 0;
+}
+
+static int tgsi_slt(struct r600_shader_ctx *ctx)
+{
+	struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
+	struct r600_bc_alu_src r600_src[3];
+	struct r600_bc_alu alu;
+	int i, r;
+
+	r = tgsi_split_constant(ctx, r600_src);
+	if (r)
+		return r;
+	for (i = 0; i < 4; i++) {
+		memset(&alu, 0, sizeof(struct r600_bc_alu));
+		if (!(inst->Dst[0].Register.WriteMask & (1 << i))) {
+			alu.inst = V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP;
+			alu.dst.chan = i;
+		} else {
+			alu.inst = ctx->inst_info->r600_opcode;
+			alu.src[1] = r600_src[0];
+			alu.src[1].chan = tgsi_chan(&inst->Src[0], i);
+			alu.src[0] = r600_src[1];
+			alu.src[0].chan = tgsi_chan(&inst->Src[1], i);
+			r = tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst);
+			if (r)
+				return r;
+		}
+		if (i == 3) {
+			alu.last = 1;
+		}
+		r = r600_bc_add_alu(ctx->bc, &alu);
+		if (r)
+			return r;
+	}
+	return 0;
+}
+
+static int tgsi_lit(struct r600_shader_ctx *ctx)
+{
+	struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
+	struct r600_bc_alu alu;
+	int r;
+
+	/* dst.x, <- 1.0  */
+	memset(&alu, 0, sizeof(struct r600_bc_alu));
+	alu.inst = V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV;
+	alu.src[0].sel  = 249; /*1.0*/
+	alu.src[0].chan = 0;
+	r = tgsi_dst(ctx, &inst->Dst[0], 0, &alu.dst);
+	if (r)
+		return r;
+	alu.dst.write = (inst->Dst[0].Register.WriteMask >> 0) & 1;
+	r = r600_bc_add_alu(ctx->bc, &alu);
+	if (r)
+		return r;
+
+	/* dst.y = max(src.x, 0.0) */
+	memset(&alu, 0, sizeof(struct r600_bc_alu));
+	alu.inst = V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MAX;
+	r = tgsi_src(ctx, &inst->Src[0], &alu.src[0]);
+	if (r)
+		return r;
+	alu.src[1].sel  = 248; /*0.0*/
+	alu.src[1].chan = tgsi_chan(&inst->Src[0], 0);
+	r = tgsi_dst(ctx, &inst->Dst[0], 1, &alu.dst);
+	if (r)
+		return r;
+	alu.dst.write = (inst->Dst[0].Register.WriteMask >> 1) & 1;
+	r = r600_bc_add_alu(ctx->bc, &alu);
+	if (r)
+		return r;
+
+	/* dst.z = NOP - fill Z slot */
+	memset(&alu, 0, sizeof(struct r600_bc_alu));
+	alu.inst = V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP;
+	alu.dst.chan = 2;
+	r = r600_bc_add_alu(ctx->bc, &alu);
+	if (r)
+		return r;
+
+	/* dst.w, <- 1.0  */
+	memset(&alu, 0, sizeof(struct r600_bc_alu));
+	alu.inst = V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV;
+	alu.src[0].sel  = 249;
+	alu.src[0].chan = 0;
+	r = tgsi_dst(ctx, &inst->Dst[0], 3, &alu.dst);
+	if (r)
+		return r;
+	alu.dst.write = (inst->Dst[0].Register.WriteMask >> 3) & 1;
+	alu.last = 1;
+	r = r600_bc_add_alu(ctx->bc, &alu);
+	if (r)
+		return r;
+
+	if (inst->Dst[0].Register.WriteMask & (1 << 2))
+	{
+		int chan;
+		int sel;
+
+		/* dst.z = log(src.y) */
+		memset(&alu, 0, sizeof(struct r600_bc_alu));
+		alu.inst = V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_CLAMPED;
+		r = tgsi_src(ctx, &inst->Src[0], &alu.src[0]);
+		if (r)
+			return r;
+		alu.src[0].chan = tgsi_chan(&inst->Src[0], 1);
+		r = tgsi_dst(ctx, &inst->Dst[0], 2, &alu.dst);
+		if (r)
+			return r;
+		alu.last = 1;
+		r = r600_bc_add_alu(ctx->bc, &alu);
+		if (r)
+			return r;
+
+		chan = alu.dst.chan;
+		sel = alu.dst.sel;
+
+		/* tmp.x = amd MUL_LIT(src.w, dst.z, src.x ) */
+		memset(&alu, 0, sizeof(struct r600_bc_alu));
+		alu.inst = V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MUL_LIT;
+		r = tgsi_src(ctx, &inst->Src[0], &alu.src[0]);
+		if (r)
+			return r;
+		alu.src[0].chan = tgsi_chan(&inst->Src[0], 3);
+		alu.src[1].sel  = sel;
+		alu.src[1].chan = chan;
+		r = tgsi_src(ctx, &inst->Src[0], &alu.src[2]);
+		if (r)
+			return r;
+		alu.src[2].chan = tgsi_chan(&inst->Src[0], 0);
+		alu.dst.sel = ctx->temp_reg;
+		alu.dst.chan = 0;
+		alu.dst.write = 1;
+		alu.is_op3 = 1;
+		alu.last = 1;
+		r = r600_bc_add_alu(ctx->bc, &alu);
+		if (r)
+			return r;
+
+		/* dst.z = exp(tmp.x) */
+		memset(&alu, 0, sizeof(struct r600_bc_alu));
+		alu.inst = V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE;
+		alu.src[0].sel = ctx->temp_reg;
+		alu.src[0].chan = 0;
+		r = tgsi_dst(ctx, &inst->Dst[0], 2, &alu.dst);
+		if (r)
+			return r;
+		alu.last = 1;
+		r = r600_bc_add_alu(ctx->bc, &alu);
+		if (r)
+			return r;
+	}
+	return 0;
+}
+
+static int tgsi_trans(struct r600_shader_ctx *ctx)
+{
+	struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
+	struct r600_bc_alu alu;
+	int i, j, r;
+
+	for (i = 0; i < 4; i++) {
+		memset(&alu, 0, sizeof(struct r600_bc_alu));
+		if (inst->Dst[0].Register.WriteMask & (1 << i)) {
+			alu.inst = ctx->inst_info->r600_opcode;
+			for (j = 0; j < inst->Instruction.NumSrcRegs; j++) {
+				r = tgsi_src(ctx, &inst->Src[j], &alu.src[j]);
+				if (r)
+					return r;
+				alu.src[j].chan = tgsi_chan(&inst->Src[j], i);
+			}
+			r = tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst);
+			if (r)
+				return r;
+			alu.last = 1;
+			r = r600_bc_add_alu(ctx->bc, &alu);
+			if (r)
+				return r;
+		}
+	}
+	return 0;
+}
+
+static int tgsi_trans_srcx_replicate(struct r600_shader_ctx *ctx)
+{
+	struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
+	struct r600_bc_alu alu;
+	int i, j, r;
+
+	memset(&alu, 0, sizeof(struct r600_bc_alu));
+	alu.inst = ctx->inst_info->r600_opcode;
+	for (j = 0; j < inst->Instruction.NumSrcRegs; j++) {
+		r = tgsi_src(ctx, &inst->Src[j], &alu.src[j]);
+		if (r)
+			return r;
+		alu.src[j].chan = tgsi_chan(&inst->Src[j], 0);
+	}
+	alu.dst.sel = ctx->temp_reg;
+	alu.dst.write = 1;
+	alu.last = 1;
+	r = r600_bc_add_alu(ctx->bc, &alu);
+	if (r)
+		return r;
+	/* replicate result */
+	for (i = 0; i < 4; i++) {
+		memset(&alu, 0, sizeof(struct r600_bc_alu));
+		alu.src[0].sel = ctx->temp_reg;
+		alu.inst = V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV;
+		alu.dst.chan = i;
+		r = tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst);
+		if (r)
+			return r;
+		alu.dst.write = (inst->Dst[0].Register.WriteMask >> i) & 1;
+		if (i == 3)
+			alu.last = 1;
+		r = r600_bc_add_alu(ctx->bc, &alu);
+		if (r)
+			return r;
+	}
+	return 0;
+}
+
+static int tgsi_helper_copy(struct r600_shader_ctx *ctx, struct tgsi_full_instruction *inst)
+{
+	struct r600_bc_alu alu;
+	int i, r;
+
+	r = r600_bc_add_literal(ctx->bc, ctx->value);
+	if (r)
+		return r;
+	for (i = 0; i < 4; i++) {
+		memset(&alu, 0, sizeof(struct r600_bc_alu));
+		if (!(inst->Dst[0].Register.WriteMask & (1 << i))) {
+			alu.inst = V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP;
+			alu.dst.chan = i;
+		} else {
+			alu.inst = V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV;
+			r = tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst);
+			if (r)
+				return r;
+			alu.src[0].sel = ctx->temp_reg;
+			alu.src[0].chan = i;
+		}
+		if (i == 3) {
+			alu.last = 1;
+		}
+		r = r600_bc_add_alu(ctx->bc, &alu);
+		if (r)
+			return r;
+	}
+	return 0;
+}
+
+static int tgsi_op3(struct r600_shader_ctx *ctx)
+{
+	struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
+	struct r600_bc_alu_src r600_src[3];
+	struct r600_bc_alu alu;
+	int i, j, r;
+
+	r = tgsi_split_constant(ctx, r600_src);
+	if (r)
+		return r;
+	/* do it in 2 step as op3 doesn't support writemask */
+	for (i = 0; i < 4; i++) {
+		memset(&alu, 0, sizeof(struct r600_bc_alu));
+		alu.inst = ctx->inst_info->r600_opcode;
+		for (j = 0; j < inst->Instruction.NumSrcRegs; j++) {
+			alu.src[j] = r600_src[j];
+			alu.src[j].chan = tgsi_chan(&inst->Src[j], i);
+		}
+		alu.dst.sel = ctx->temp_reg;
+		alu.dst.chan = i;
+		alu.dst.write = 1;
+		alu.is_op3 = 1;
+		if (i == 3) {
+			alu.last = 1;
+		}
+		r = r600_bc_add_alu(ctx->bc, &alu);
+		if (r)
+			return r;
+	}
+	return tgsi_helper_copy(ctx, inst);
 }
+
+static int tgsi_dp(struct r600_shader_ctx *ctx)
+{
+	struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
+	struct r600_bc_alu_src r600_src[3];
+	struct r600_bc_alu alu;
+	int i, j, r;
+
+	r = tgsi_split_constant(ctx, r600_src);
+	if (r)
+		return r;
+	for (i = 0; i < 4; i++) {
+		memset(&alu, 0, sizeof(struct r600_bc_alu));
+		alu.inst = ctx->inst_info->r600_opcode;
+		for (j = 0; j < inst->Instruction.NumSrcRegs; j++) {
+			alu.src[j] = r600_src[j];
+			alu.src[j].chan = tgsi_chan(&inst->Src[j], i);
+		}
+		alu.dst.sel = ctx->temp_reg;
+		alu.dst.chan = i;
+		alu.dst.write = 1;
+		/* handle some special cases */
+		switch (ctx->inst_info->tgsi_opcode) {
+		case TGSI_OPCODE_DP2:
+			if (i > 1) {
+				alu.src[0].sel = alu.src[1].sel = 248;
+				alu.src[0].chan = alu.src[1].chan = 0;
+			}
+			break;
+		case TGSI_OPCODE_DP3:
+			if (i > 2) {
+				alu.src[0].sel = alu.src[1].sel = 248;
+				alu.src[0].chan = alu.src[1].chan = 0;
+			}
+			break;
+		default:
+			break;
+		}
+		if (i == 3) {
+			alu.last = 1;
+		}
+		r = r600_bc_add_alu(ctx->bc, &alu);
+		if (r)
+			return r;
+	}
+	return tgsi_helper_copy(ctx, inst);
+}
+
+static int tgsi_tex(struct r600_shader_ctx *ctx)
+{
+	struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
+	struct r600_bc_tex tex;
+	struct r600_bc_alu alu;
+	unsigned src_gpr;
+	int r;
+
+	src_gpr = ctx->file_offset[inst->Src[0].Register.File] + inst->Src[0].Register.Index;
+
+	/* Add perspective divide */
+	memset(&alu, 0, sizeof(struct r600_bc_alu));
+	alu.inst = V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE;
+	alu.src[0].sel = src_gpr;
+	alu.src[0].chan = tgsi_chan(&inst->Src[0], 3);
+	alu.dst.sel = ctx->temp_reg;
+	alu.dst.chan = 3;
+	alu.last = 1;
+	alu.dst.write = 1;
+	r = r600_bc_add_alu(ctx->bc, &alu);
+	if (r)
+		return r;
+
+	memset(&alu, 0, sizeof(struct r600_bc_alu));
+	alu.inst = V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL;
+	alu.src[0].sel = ctx->temp_reg;
+	alu.src[0].chan = 3;
+	alu.src[1].sel = src_gpr;
+	alu.src[1].chan = tgsi_chan(&inst->Src[0], 0);
+	alu.dst.sel = ctx->temp_reg;
+	alu.dst.chan = 0;
+	alu.dst.write = 1;
+	r = r600_bc_add_alu(ctx->bc, &alu);
+	if (r)
+		return r;
+	memset(&alu, 0, sizeof(struct r600_bc_alu));
+	alu.inst = V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL;
+	alu.src[0].sel = ctx->temp_reg;
+	alu.src[0].chan = 3;
+	alu.src[1].sel = src_gpr;
+	alu.src[1].chan = tgsi_chan(&inst->Src[0], 1);
+	alu.dst.sel = ctx->temp_reg;
+	alu.dst.chan = 1;
+	alu.dst.write = 1;
+	r = r600_bc_add_alu(ctx->bc, &alu);
+	if (r)
+		return r;
+	memset(&alu, 0, sizeof(struct r600_bc_alu));
+	alu.inst = V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL;
+	alu.src[0].sel = ctx->temp_reg;
+	alu.src[0].chan = 3;
+	alu.src[1].sel = src_gpr;
+	alu.src[1].chan = tgsi_chan(&inst->Src[0], 2);
+	alu.dst.sel = ctx->temp_reg;
+	alu.dst.chan = 2;
+	alu.dst.write = 1;
+	r = r600_bc_add_alu(ctx->bc, &alu);
+	if (r)
+		return r;
+	memset(&alu, 0, sizeof(struct r600_bc_alu));
+	alu.inst = V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV;
+	alu.src[0].sel = 249;
+	alu.src[0].chan = 0;
+	alu.dst.sel = ctx->temp_reg;
+	alu.dst.chan = 3;
+	alu.last = 1;
+	alu.dst.write = 1;
+	r = r600_bc_add_alu(ctx->bc, &alu);
+	if (r)
+		return r;
+	src_gpr = ctx->temp_reg;
+
+	/* TODO use temp if src_gpr is not a temporary reg (File != TEMPORARY) */
+	memset(&tex, 0, sizeof(struct r600_bc_tex));
+	tex.inst = ctx->inst_info->r600_opcode;
+	tex.resource_id = ctx->file_offset[inst->Src[1].Register.File] + inst->Src[1].Register.Index;
+	tex.sampler_id = tex.resource_id;
+	tex.src_gpr = src_gpr;
+	tex.dst_gpr = ctx->file_offset[inst->Dst[0].Register.File] + inst->Dst[0].Register.Index;
+	tex.dst_sel_x = 0;
+	tex.dst_sel_y = 1;
+	tex.dst_sel_z = 2;
+	tex.dst_sel_w = 3;
+	tex.src_sel_x = 0;
+	tex.src_sel_y = 1;
+	tex.src_sel_z = 2;
+	tex.src_sel_w = 3;
+
+	if (inst->Texture.Texture != TGSI_TEXTURE_RECT) {
+		tex.coord_type_x = 1;
+		tex.coord_type_y = 1;
+		tex.coord_type_z = 1;
+		tex.coord_type_w = 1;
+	}
+	return r600_bc_add_tex(ctx->bc, &tex);
+}
+
+static int tgsi_lrp(struct r600_shader_ctx *ctx)
+{
+	struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction;
+	struct r600_bc_alu_src r600_src[3];
+	struct r600_bc_alu alu;
+	unsigned i;
+	int r;
+
+	r = tgsi_split_constant(ctx, r600_src);
+	if (r)
+		return r;
+	/* 1 - src0 */
+	for (i = 0; i < 4; i++) {
+		memset(&alu, 0, sizeof(struct r600_bc_alu));
+		alu.inst = V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD;
+		alu.src[0].sel = 249;
+		alu.src[0].chan = 0;
+		alu.src[1] = r600_src[0];
+		alu.src[1].chan = tgsi_chan(&inst->Src[0], i);
+		alu.src[1].neg = 1;
+		alu.dst.sel = ctx->temp_reg;
+		alu.dst.chan = i;
+		if (i == 3) {
+			alu.last = 1;
+		}
+		alu.dst.write = 1;
+		r = r600_bc_add_alu(ctx->bc, &alu);
+		if (r)
+			return r;
+	}
+	r = r600_bc_add_literal(ctx->bc, ctx->value);
+	if (r)
+		return r;
+
+	/* (1 - src0) * src2 */
+	for (i = 0; i < 4; i++) {
+		memset(&alu, 0, sizeof(struct r600_bc_alu));
+		alu.inst = V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL;
+		alu.src[0].sel = ctx->temp_reg;
+		alu.src[0].chan = i;
+		alu.src[1] = r600_src[2];
+		alu.src[1].chan = tgsi_chan(&inst->Src[2], i);
+		alu.dst.sel = ctx->temp_reg;
+		alu.dst.chan = i;
+		if (i == 3) {
+			alu.last = 1;
+		}
+		alu.dst.write = 1;
+		r = r600_bc_add_alu(ctx->bc, &alu);
+		if (r)
+			return r;
+	}
+	r = r600_bc_add_literal(ctx->bc, ctx->value);
+	if (r)
+		return r;
+
+	/* src0 * src1 + (1 - src0) * src2 */
+	for (i = 0; i < 4; i++) {
+		memset(&alu, 0, sizeof(struct r600_bc_alu));
+		alu.inst = V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD;
+		alu.is_op3 = 1;
+		alu.src[0] = r600_src[0];
+		alu.src[0].chan = tgsi_chan(&inst->Src[0], i);
+		alu.src[1] = r600_src[1];
+		alu.src[1].chan = tgsi_chan(&inst->Src[1], i);
+		alu.src[2].sel = ctx->temp_reg;
+		alu.src[2].chan = i;
+		alu.dst.sel = ctx->temp_reg;
+		alu.dst.chan = i;
+		if (i == 3) {
+			alu.last = 1;
+		}
+		r = r600_bc_add_alu(ctx->bc, &alu);
+		if (r)
+			return r;
+	}
+	return tgsi_helper_copy(ctx, inst);
+}
+
+static struct r600_shader_tgsi_instruction r600_shader_tgsi_instruction[] = {
+	{TGSI_OPCODE_ARL,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_MOV,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV, tgsi_op2},
+	{TGSI_OPCODE_LIT,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_lit},
+	{TGSI_OPCODE_RCP,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE, tgsi_trans_srcx_replicate},
+	{TGSI_OPCODE_RSQ,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIPSQRT_IEEE, tgsi_trans_srcx_replicate},
+	{TGSI_OPCODE_EXP,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_LOG,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_MUL,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL, tgsi_op2},
+	{TGSI_OPCODE_ADD,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD, tgsi_op2},
+	{TGSI_OPCODE_DP3,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4, tgsi_dp},
+	{TGSI_OPCODE_DP4,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4, tgsi_dp},
+	{TGSI_OPCODE_DST,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_MIN,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MIN, tgsi_op2},
+	{TGSI_OPCODE_MAX,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MAX, tgsi_op2},
+	{TGSI_OPCODE_SLT,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGT, tgsi_slt},
+	{TGSI_OPCODE_SGE,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_MAD,	1, V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD, tgsi_op3},
+	{TGSI_OPCODE_SUB,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD, tgsi_op2},
+	{TGSI_OPCODE_LRP,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_lrp},
+	{TGSI_OPCODE_CND,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	/* gap */
+	{20,			0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_DP2A,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	/* gap */
+	{22,			0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{23,			0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_FRC,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_CLAMP,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_FLR,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_ROUND,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_EX2,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE, tgsi_trans_srcx_replicate},
+	{TGSI_OPCODE_LG2,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_POW,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_XPD,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	/* gap */
+	{32,			0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_ABS,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV, tgsi_op2},
+	{TGSI_OPCODE_RCC,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_DPH,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_COS,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_DDX,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_DDY,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_KILP,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},  /* predicated kill */
+	{TGSI_OPCODE_PK2H,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_PK2US,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_PK4B,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_PK4UB,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_RFL,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_SEQ,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_SFL,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_SGT,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_SIN,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_SLE,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_SNE,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_STR,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_TEX,	0, 0x10, tgsi_tex},
+	{TGSI_OPCODE_TXD,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_TXP,	0, 0x10, tgsi_tex},
+	{TGSI_OPCODE_UP2H,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_UP2US,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_UP4B,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_UP4UB,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_X2D,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_ARA,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_ARR,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_BRA,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_CAL,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_RET,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_SSG,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported}, /* SGN */
+	{TGSI_OPCODE_CMP,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_SCS,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_TXB,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_NRM,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_DIV,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_DP2,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4, tgsi_dp},
+	{TGSI_OPCODE_TXL,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_BRK,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_IF,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	/* gap */
+	{75,			0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{76,			0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_ELSE,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_ENDIF,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	/* gap */
+	{79,			0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{80,			0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_PUSHA,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_POPA,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_CEIL,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_I2F,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_NOT,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_TRUNC,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_SHL,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	/* gap */
+	{88,			0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_AND,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_OR,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_MOD,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_XOR,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_SAD,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_TXF,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_TXQ,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_CONT,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_EMIT,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_ENDPRIM,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_BGNLOOP,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_BGNSUB,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_ENDLOOP,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_ENDSUB,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	/* gap */
+	{103,			0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{104,			0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{105,			0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{106,			0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_NOP,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	/* gap */
+	{108,			0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{109,			0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{110,			0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{111,			0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_NRM4,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_CALLNZ,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_IFC,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_BREAKC,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_KIL,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLGT, tgsi_kill},  /* conditional kill */
+	{TGSI_OPCODE_END,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_end},  /* aka HALT */
+	/* gap */
+	{118,			0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_F2I,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_IDIV,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_IMAX,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_IMIN,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_INEG,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_ISGE,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_ISHR,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_ISLT,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_F2U,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_U2F,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_UADD,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_UDIV,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_UMAD,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_UMAX,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_UMIN,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_UMOD,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_UMUL,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_USEQ,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_USGE,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_USHR,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_USLT,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_USNE,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_SWITCH,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_CASE,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_DEFAULT,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_ENDSWITCH,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+	{TGSI_OPCODE_LAST,	0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported},
+};
diff --git a/src/gallium/drivers/r600/r600_shader.h b/src/gallium/drivers/r600/r600_shader.h
index 40064ba8a99..2ee7780ead0 100644
--- a/src/gallium/drivers/r600/r600_shader.h
+++ b/src/gallium/drivers/r600/r600_shader.h
@@ -23,243 +23,25 @@
 #ifndef R600_SHADER_H
 #define R600_SHADER_H
 
-#include "r600_compiler.h"
-#include "radeon.h"
-
-struct r600_shader_operand {
-	struct c_vector			*vector;
-	unsigned			sel;
-	unsigned			chan;
-	unsigned			neg;
-	unsigned			abs;
-};
-
-struct r600_shader_vfetch {
-	struct list_head                head;
-	unsigned			cf_addr;
-	struct r600_shader_operand	src[2];
-	struct r600_shader_operand	dst[4];
-};
-
-struct r600_shader_inst {
-	unsigned			is_op3;
-	unsigned			opcode;
-	unsigned			inst;
-	struct r600_shader_operand	src[3];
-	struct r600_shader_operand	dst;
-	unsigned			last;
-};
-
-struct r600_shader_alu {
-	struct list_head                head;
-	unsigned			nalu;
-	unsigned			nliteral;
-	unsigned			nconstant;
-	struct r600_shader_inst		alu[5];
-	u32				literal[4];
-};
-
-struct r600_shader_node {
-	struct list_head                head;
-	unsigned			cf_id;		/**< cf index (in dw) in byte code */
-	unsigned			cf_addr;	/**< instructions index (in dw) in byte code */
-	unsigned			nslot;		/**< number of slot (2 dw) needed by this node */
-	unsigned			nfetch;
-	struct c_node			*node;		/**< compiler node from which this node originate */
-	struct list_head                vfetch;         /**< list of vfetch instructions */
-	struct list_head                alu;            /**< list of alu instructions */
-};
+#include "r600_asm.h"
 
 struct r600_shader_io {
-	unsigned	name;
-	unsigned	gpr;
-	int		sid;
+	unsigned		name;
+	unsigned		gpr;
+	unsigned		done;
+	int			sid;
+	unsigned		interpolate;
 };
 
 struct r600_shader {
-	unsigned			stack_size;		/**< stack size needed by this shader */
-	unsigned			ngpr;			/**< number of GPR needed by this shader */
-	unsigned			nconstant;		/**< number of constants used by this shader */
-	unsigned			nresource;		/**< number of resources used by this shader */
-	unsigned			noutput;
-	unsigned			ninput;
-	unsigned			nvector;
-	unsigned			ncf;			/**< total number of cf clauses */
-	unsigned			nslot;			/**< total number of slots (2 dw) */
-	unsigned			flat_shade;		/**< are we flat shading */
-	struct list_head		nodes;			/**< list of node */
-	struct r600_shader_io		input[32];
-	struct r600_shader_io		output[32];
-	/* TODO replace GPR by some better register allocator */
-	struct c_vector			**gpr;
-	unsigned			ndw;			/**< bytes code size in dw */
-	u32				*bcode;			/**< bytes code */
-	enum pipe_format		resource_format[160];	/**< format of resource */
-	struct c_shader			cshader;
-	boolean r6xx_compile;
+	unsigned		processor_type;
+	struct r600_bc		bc;
+	boolean			flat_shade;
+	unsigned		ninput;
+	unsigned		noutput;
+	struct r600_shader_io	input[32];
+	struct r600_shader_io	output[32];
+	enum radeon_family	family;
 };
 
-void r600_shader_cleanup(struct r600_shader *rshader);
-int r600_shader_register(struct r600_shader *rshader);
-int r600_shader_node(struct r600_shader *shader);
-void r600_shader_node_place(struct r600_shader *rshader);
-int r600_shader_find_gpr(struct r600_shader *rshader, struct c_vector *v, unsigned swizzle,
-			struct r600_shader_operand *operand);
-int r600_shader_vfetch_bytecode(struct r600_shader *rshader,
-				struct r600_shader_node *rnode,
-				struct r600_shader_vfetch *vfetch,
-				unsigned *cid);
-int r600_shader_update(struct r600_shader *rshader,
-			enum pipe_format *resource_format);
-int r600_shader_legalize(struct r600_shader *rshader);
-int r600_cshader_legalize(struct c_shader *shader);
-
-int r700_shader_translate(struct r600_shader *rshader);
-
-int c_shader_from_tgsi(struct c_shader *shader, unsigned type,
-			const struct tgsi_token *tokens);
-int r600_shader_register(struct r600_shader *rshader);
-int r600_shader_translate_rec(struct r600_shader *rshader, struct c_node *node);
-int r700_shader_translate(struct r600_shader *rshader);
-int r600_shader_insert_fetch(struct c_shader *shader);
-
-int r6xx_shader_alu_translate(struct r600_shader *rshader,
-			      struct r600_shader_node *rnode,
-			      unsigned *cid);
-
-enum r600_instruction {
-	INST_ADD			= 0,
-	INST_MUL			= 1,
-	INST_MUL_IEEE			= 2,
-	INST_MAX			= 3,
-	INST_MIN			= 4,
-	INST_MAX_DX10			= 5,
-	INST_MIN_DX10			= 6,
-	INST_SETE			= 7,
-	INST_SETGT			= 8,
-	INST_SETGE			= 9,
-	INST_SETNE			= 10,
-	INST_SETE_DX10			= 11,
-	INST_SETGT_DX10			= 12,
-	INST_SETGE_DX10			= 13,
-	INST_SETNE_DX10			= 14,
-	INST_FRACT			= 15,
-	INST_TRUNC			= 16,
-	INST_CEIL			= 17,
-	INST_RNDNE			= 18,
-	INST_FLOOR			= 19,
-	INST_MOVA			= 20,
-	INST_MOVA_FLOOR			= 21,
-	INST_MOVA_INT			= 22,
-	INST_MOV			= 23,
-	INST_NOP			= 24,
-	INST_PRED_SETGT_UINT		= 25,
-	INST_PRED_SETGE_UINT		= 26,
-	INST_PRED_SETE			= 27,
-	INST_PRED_SETGT			= 28,
-	INST_PRED_SETGE			= 29,
-	INST_PRED_SETNE			= 30,
-	INST_PRED_SET_INV		= 31,
-	INST_PRED_SET_POP		= 32,
-	INST_PRED_SET_CLR		= 33,
-	INST_PRED_SET_RESTORE		= 34,
-	INST_PRED_SETE_PUSH		= 35,
-	INST_PRED_SETGT_PUSH		= 36,
-	INST_PRED_SETGE_PUSH		= 37,
-	INST_PRED_SETNE_PUSH		= 38,
-	INST_KILLE			= 39,
-	INST_KILLGT			= 40,
-	INST_KILLGE			= 41,
-	INST_KILLNE			= 42,
-	INST_AND_INT			= 43,
-	INST_OR_INT			= 44,
-	INST_XOR_INT			= 45,
-	INST_NOT_INT			= 46,
-	INST_ADD_INT			= 47,
-	INST_SUB_INT			= 48,
-	INST_MAX_INT			= 49,
-	INST_MIN_INT			= 50,
-	INST_MAX_UINT			= 51,
-	INST_MIN_UINT			= 52,
-	INST_SETE_INT			= 53,
-	INST_SETGT_INT			= 54,
-	INST_SETGE_INT			= 55,
-	INST_SETNE_INT			= 56,
-	INST_SETGT_UINT			= 57,
-	INST_SETGE_UINT			= 58,
-	INST_KILLGT_UINT		= 59,
-	INST_KILLGE_UINT		= 60,
-	INST_PRED_SETE_INT		= 61,
-	INST_PRED_SETGT_INT		= 62,
-	INST_PRED_SETGE_INT		= 63,
-	INST_PRED_SETNE_INT		= 64,
-	INST_KILLE_INT			= 65,
-	INST_KILLGT_INT			= 66,
-	INST_KILLGE_INT			= 67,
-	INST_KILLNE_INT			= 68,
-	INST_PRED_SETE_PUSH_INT		= 69,
-	INST_PRED_SETGT_PUSH_INT	= 70,
-	INST_PRED_SETGE_PUSH_INT	= 71,
-	INST_PRED_SETNE_PUSH_INT	= 72,
-	INST_PRED_SETLT_PUSH_INT	= 73,
-	INST_PRED_SETLE_PUSH_INT	= 74,
-	INST_DOT4			= 75,
-	INST_DOT4_IEEE			= 76,
-	INST_CUBE			= 77,
-	INST_MAX4			= 78,
-	INST_MOVA_GPR_INT		= 79,
-	INST_EXP_IEEE			= 80,
-	INST_LOG_CLAMPED		= 81,
-	INST_LOG_IEEE			= 82,
-	INST_RECIP_CLAMPED		= 83,
-	INST_RECIP_FF			= 84,
-	INST_RECIP_IEEE			= 85,
-	INST_RECIPSQRT_CLAMPED		= 86,
-	INST_RECIPSQRT_FF		= 87,
-	INST_RECIPSQRT_IEEE		= 88,
-	INST_SQRT_IEEE			= 89,
-	INST_FLT_TO_INT			= 90,
-	INST_INT_TO_FLT			= 91,
-	INST_UINT_TO_FLT		= 92,
-	INST_SIN			= 93,
-	INST_COS			= 94,
-	INST_ASHR_INT			= 95,
-	INST_LSHR_INT			= 96,
-	INST_LSHL_INT			= 97,
-	INST_MULLO_INT			= 98,
-	INST_MULHI_INT			= 99,
-	INST_MULLO_UINT			= 100,
-	INST_MULHI_UINT			= 101,
-	INST_RECIP_INT			= 102,
-	INST_RECIP_UINT			= 103,
-	INST_FLT_TO_UINT		= 104,
-	INST_MUL_LIT			= 105,
-	INST_MUL_LIT_M2			= 106,
-	INST_MUL_LIT_M4			= 107,
-	INST_MUL_LIT_D2			= 108,
-	INST_MULADD			= 109,
-	INST_MULADD_M2			= 110,
-	INST_MULADD_M4			= 111,
-	INST_MULADD_D2			= 112,
-	INST_MULADD_IEEE		= 113,
-	INST_MULADD_IEEE_M2		= 114,
-	INST_MULADD_IEEE_M4		= 115,
-	INST_MULADD_IEEE_D2		= 116,
-	INST_CNDE			= 117,
-	INST_CNDGT			= 118,
-	INST_CNDGE			= 119,
-	INST_CNDE_INT			= 120,
-	INST_CNDGT_INT			= 121,
-	INST_CNDGE_INT			= 122,
-	INST_COUNT
-};
-
-struct r600_instruction_info {
-	enum r600_instruction		instruction;
-	unsigned			opcode;
-	unsigned			is_trans;
-	unsigned			is_op3;
-};
-
-
 #endif
diff --git a/src/gallium/drivers/r600/r600_sq.h b/src/gallium/drivers/r600/r600_sq.h
index 447ba98f004..002660c654a 100644
--- a/src/gallium/drivers/r600/r600_sq.h
+++ b/src/gallium/drivers/r600/r600_sq.h
@@ -87,9 +87,9 @@
 #define   G_SQ_CF_WORD1_BARRIER(x)                                   (((x) >> 31) & 0x1)
 #define   C_SQ_CF_WORD1_BARRIER                                      0x7FFFFFFF
 #define P_SQ_CF_ALU_WORD0
-#define   S_SQ_CF_ALU_WORD0_ALU_ADDR(x)                              (((x) & 0x3FFFFF) << 0)
-#define   G_SQ_CF_ALU_WORD0_ALU_ADDR(x)                              (((x) >> 0) & 0x3FFFFF)
-#define   C_SQ_CF_ALU_WORD0_ALU_ADDR                                 0xFFC00000
+#define   S_SQ_CF_ALU_WORD0_ADDR(x)                                  (((x) & 0x3FFFFF) << 0)
+#define   G_SQ_CF_ALU_WORD0_ADDR(x)                                  (((x) >> 0) & 0x3FFFFF)
+#define   C_SQ_CF_ALU_WORD0_ADDR                                     0xFFC00000
 #define   S_SQ_CF_ALU_WORD0_KCACHE_BANK0(x)                          (((x) & 0xF) << 22)
 #define   G_SQ_CF_ALU_WORD0_KCACHE_BANK0(x)                          (((x) >> 22) & 0xF)
 #define   C_SQ_CF_ALU_WORD0_KCACHE_BANK0                             0xFC3FFFFF
@@ -109,15 +109,15 @@
 #define   S_SQ_CF_ALU_WORD1_KCACHE_ADDR1(x)                          (((x) & 0xFF) << 10)
 #define   G_SQ_CF_ALU_WORD1_KCACHE_ADDR1(x)                          (((x) >> 10) & 0xFF)
 #define   C_SQ_CF_ALU_WORD1_KCACHE_ADDR1                             0xFFFC03FF
-#define   S_SQ_CF_ALU_WORD1_ALU_COUNT(x)                             (((x) & 0x7F) << 18)
-#define   G_SQ_CF_ALU_WORD1_ALU_COUNT(x)                             (((x) >> 18) & 0x7F)
-#define   C_SQ_CF_ALU_WORD1_ALU_COUNT                                0xFE03FFFF
+#define   S_SQ_CF_ALU_WORD1_COUNT(x)                                 (((x) & 0x7F) << 18)
+#define   G_SQ_CF_ALU_WORD1_COUNT(x)                                 (((x) >> 18) & 0x7F)
+#define   C_SQ_CF_ALU_WORD1_COUNT                                    0xFE03FFFF
 #define   S_SQ_CF_ALU_WORD1_USES_WATERFALL(x)                        (((x) & 0x1) << 25)
 #define   G_SQ_CF_ALU_WORD1_USES_WATERFALL(x)                        (((x) >> 25) & 0x1)
 #define   C_SQ_CF_ALU_WORD1_USES_WATERFALL                           0xFDFFFFFF
-#define   S_SQ_CF_ALU_WORD1_CF_ALU_INST(x)                           (((x) & 0xF) << 26)
-#define   G_SQ_CF_ALU_WORD1_CF_ALU_INST(x)                           (((x) >> 26) & 0xF)
-#define   C_SQ_CF_ALU_WORD1_CF_ALU_INST                              0xC3FFFFFF
+#define   S_SQ_CF_ALU_WORD1_CF_INST(x)                               (((x) & 0xF) << 26)
+#define   G_SQ_CF_ALU_WORD1_CF_INST(x)                               (((x) >> 26) & 0xF)
+#define   C_SQ_CF_ALU_WORD1_CF_INST                                  0xC3FFFFFF
 #define     V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU                         0x00000008
 #define     V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU_PUSH_BEFORE             0x00000009
 #define     V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU_POP_AFTER               0x0000000A
@@ -546,6 +546,8 @@
 #define   S_SQ_TEX_WORD1_COORD_TYPE_X(x)                             (((x) & 0x1) << 28)
 #define   G_SQ_TEX_WORD1_COORD_TYPE_X(x)                             (((x) >> 28) & 0x1)
 #define   C_SQ_TEX_WORD1_COORD_TYPE_X                                0xEFFFFFFF
+#define     V_SQ_TEX_WORD1_COORD_UNNORMALIZED                        0x00000000
+#define     V_SQ_TEX_WORD1_COORD_NORMALIZED                          0x00000001
 #define   S_SQ_TEX_WORD1_COORD_TYPE_Y(x)                             (((x) & 0x1) << 29)
 #define   G_SQ_TEX_WORD1_COORD_TYPE_Y(x)                             (((x) >> 29) & 0x1)
 #define   C_SQ_TEX_WORD1_COORD_TYPE_Y                                0xDFFFFFFF
diff --git a/src/gallium/drivers/r600/r600_state.c b/src/gallium/drivers/r600/r600_state.c
index 4150f88785e..3efd409ae0d 100644
--- a/src/gallium/drivers/r600/r600_state.c
+++ b/src/gallium/drivers/r600/r600_state.c
@@ -24,36 +24,605 @@
  *      Jerome Glisse
  */
 #include <stdio.h>
-#include <util/u_inlines.h>
-#include <util/u_format.h>
-#include <util/u_memory.h>
+#include <errno.h>
+#include "util/u_inlines.h"
+#include "util/u_format.h"
+#include "util/u_memory.h"
 #include "r600_screen.h"
-#include "r600_texture.h"
 #include "r600_context.h"
+#include "r600_resource.h"
 #include "r600d.h"
+#include "r600_state_inlines.h"
 
+static void *r600_create_blend_state(struct pipe_context *ctx,
+					const struct pipe_blend_state *state)
+{
+	struct r600_context *rctx = r600_context(ctx);
+
+	return r600_context_state(rctx, pipe_blend_type, state);
+}
+
+static void *r600_create_dsa_state(struct pipe_context *ctx,
+					const struct pipe_depth_stencil_alpha_state *state)
+{
+	struct r600_context *rctx = r600_context(ctx);
+
+	return r600_context_state(rctx, pipe_dsa_type, state);
+}
+
+static void *r600_create_rs_state(struct pipe_context *ctx,
+					const struct pipe_rasterizer_state *state)
+{
+	struct r600_context *rctx = r600_context(ctx);
+
+	return r600_context_state(rctx, pipe_rasterizer_type, state);
+}
+
+static void *r600_create_sampler_state(struct pipe_context *ctx,
+					const struct pipe_sampler_state *state)
+{
+	struct r600_context *rctx = r600_context(ctx);
+
+	return r600_context_state(rctx, pipe_sampler_type, state);
+}
+
+static void r600_sampler_view_destroy(struct pipe_context *ctx,
+				      struct pipe_sampler_view *state)
+{
+	struct r600_context_state *rstate = (struct r600_context_state *)state;
+
+	r600_context_state_decref(rstate);
+}
+
+static struct pipe_sampler_view *r600_create_sampler_view(struct pipe_context *ctx,
+							struct pipe_resource *texture,
+							const struct pipe_sampler_view *state)
+{
+	struct r600_context *rctx = r600_context(ctx);
+	struct r600_context_state *rstate;
+
+	rstate = r600_context_state(rctx, pipe_sampler_type, state);
+	pipe_reference(NULL, &texture->reference);
+	rstate->state.sampler_view.texture = texture;
+	rstate->state.sampler_view.reference.count = 1;
+	rstate->state.sampler_view.context = ctx;
+	return &rstate->state.sampler_view;
+}
+
+static void *r600_create_shader_state(struct pipe_context *ctx,
+					const struct pipe_shader_state *state)
+{
+	struct r600_context *rctx = r600_context(ctx);
+
+	return r600_context_state(rctx, pipe_shader_type, state);
+}
+
+static void *r600_create_vertex_elements(struct pipe_context *ctx,
+				unsigned count,
+				const struct pipe_vertex_element *elements)
+{
+	struct r600_vertex_element *v = CALLOC_STRUCT(r600_vertex_element);
+
+	assert(count < 32);
+	v->count = count;
+	memcpy(v->elements, elements, count * sizeof(struct pipe_vertex_element));
+	v->refcount = 1;
+	return v;
+}
+
+static void r600_bind_state(struct pipe_context *ctx, void *state)
+{
+	struct r600_context *rctx = r600_context(ctx);
+	struct r600_context_state *rstate = (struct r600_context_state *)state;
+
+	if (state == NULL)
+		return;
+	switch (rstate->type) {
+	case pipe_rasterizer_type:
+		rctx->rasterizer = r600_context_state_decref(rctx->rasterizer);
+		rctx->rasterizer = r600_context_state_incref(rstate);
+		break;
+	case pipe_poly_stipple_type:
+		rctx->poly_stipple = r600_context_state_decref(rctx->poly_stipple);
+		rctx->poly_stipple = r600_context_state_incref(rstate);
+		break;
+	case pipe_scissor_type:
+		rctx->scissor = r600_context_state_decref(rctx->scissor);
+		rctx->scissor = r600_context_state_incref(rstate);
+		break;
+	case pipe_clip_type:
+		rctx->clip = r600_context_state_decref(rctx->clip);
+		rctx->clip = r600_context_state_incref(rstate);
+		break;
+	case pipe_depth_type:
+		rctx->depth = r600_context_state_decref(rctx->depth);
+		rctx->depth = r600_context_state_incref(rstate);
+		break;
+	case pipe_stencil_type:
+		rctx->stencil = r600_context_state_decref(rctx->stencil);
+		rctx->stencil = r600_context_state_incref(rstate);
+		break;
+	case pipe_alpha_type:
+		rctx->alpha = r600_context_state_decref(rctx->alpha);
+		rctx->alpha = r600_context_state_incref(rstate);
+		break;
+	case pipe_dsa_type:
+		rctx->dsa = r600_context_state_decref(rctx->dsa);
+		rctx->dsa = r600_context_state_incref(rstate);
+		break;
+	case pipe_blend_type:
+		rctx->blend = r600_context_state_decref(rctx->blend);
+		rctx->blend = r600_context_state_incref(rstate);
+		break;
+	case pipe_framebuffer_type:
+		rctx->framebuffer = r600_context_state_decref(rctx->framebuffer);
+		rctx->framebuffer = r600_context_state_incref(rstate);
+		break;
+	case pipe_stencil_ref_type:
+		rctx->stencil_ref = r600_context_state_decref(rctx->stencil_ref);
+		rctx->stencil_ref = r600_context_state_incref(rstate);
+		break;
+	case pipe_viewport_type:
+		rctx->viewport = r600_context_state_decref(rctx->viewport);
+		rctx->viewport = r600_context_state_incref(rstate);
+		break;
+	case pipe_shader_type:
+	case pipe_sampler_type:
+	case pipe_sampler_view_type:
+	default:
+		R600_ERR("invalid type %d\n", rstate->type);
+		return;
+	}
+}
+
+static void r600_bind_ps_shader(struct pipe_context *ctx, void *state)
+{
+	struct r600_context *rctx = r600_context(ctx);
+	struct r600_context_state *rstate = (struct r600_context_state *)state;
+
+	rctx->ps_shader = r600_context_state_decref(rctx->ps_shader);
+	rctx->ps_shader = r600_context_state_incref(rstate);
+}
+
+static void r600_bind_vs_shader(struct pipe_context *ctx, void *state)
+{
+	struct r600_context *rctx = r600_context(ctx);
+	struct r600_context_state *rstate = (struct r600_context_state *)state;
+
+	rctx->vs_shader = r600_context_state_decref(rctx->vs_shader);
+	rctx->vs_shader = r600_context_state_incref(rstate);
+}
+
+static void r600_delete_vertex_element(struct pipe_context *ctx, void *state)
+{
+	struct r600_vertex_element *v = (struct r600_vertex_element*)state;
+
+	if (v == NULL)
+		return;
+	if (--v->refcount)
+		return;
+	free(v);
+}
+
+static void r600_bind_vertex_elements(struct pipe_context *ctx, void *state)
+{
+	struct r600_context *rctx = r600_context(ctx);
+	struct r600_vertex_element *v = (struct r600_vertex_element*)state;
+
+	r600_delete_vertex_element(ctx, rctx->vertex_elements);
+	rctx->vertex_elements = v;
+	if (v) {
+		v->refcount++;
+	}
+}
+
+static void r600_bind_ps_sampler(struct pipe_context *ctx,
+					unsigned count, void **states)
+{
+	struct r600_context *rctx = r600_context(ctx);
+	struct r600_context_state *rstate;
+	unsigned i;
+
+	for (i = 0; i < rctx->ps_nsampler; i++) {
+		rctx->ps_sampler[i] = r600_context_state_decref(rctx->ps_sampler[i]);
+	}
+	for (i = 0; i < count; i++) {
+		rstate = (struct r600_context_state *)states[i];
+		rctx->ps_sampler[i] = r600_context_state_incref(rstate);
+	}
+	rctx->ps_nsampler = count;
+}
+
+static void r600_bind_vs_sampler(struct pipe_context *ctx,
+					unsigned count, void **states)
+{
+	struct r600_context *rctx = r600_context(ctx);
+	struct r600_context_state *rstate;
+	unsigned i;
+
+	for (i = 0; i < rctx->vs_nsampler; i++) {
+		rctx->vs_sampler[i] = r600_context_state_decref(rctx->vs_sampler[i]);
+	}
+	for (i = 0; i < count; i++) {
+		rstate = (struct r600_context_state *)states[i];
+		rctx->vs_sampler[i] = r600_context_state_incref(rstate);
+	}
+	rctx->vs_nsampler = count;
+}
 
 static void r600_delete_state(struct pipe_context *ctx, void *state)
 {
-	struct radeon_state *rstate = state;
+	struct r600_context_state *rstate = (struct r600_context_state *)state;
 
-	radeon_state_decref(rstate);
+	r600_context_state_decref(rstate);
 }
 
-static void *r600_create_blend_state(struct pipe_context *ctx,
-					const struct pipe_blend_state *state)
+static void r600_set_blend_color(struct pipe_context *ctx,
+					const struct pipe_blend_color *color)
+{
+	struct r600_context *rctx = r600_context(ctx);
+
+	rctx->blend_color = *color;
+}
+
+static void r600_set_clip_state(struct pipe_context *ctx,
+				const struct pipe_clip_state *state)
+{
+}
+
+static void r600_set_constant_buffer(struct pipe_context *ctx,
+					uint shader, uint index,
+					struct pipe_resource *buffer)
 {
 	struct r600_screen *rscreen = r600_screen(ctx->screen);
+	struct r600_context *rctx = r600_context(ctx);
+	unsigned nconstant = 0, i, type, id;
+	struct radeon_state *rstate;
+	struct pipe_transfer *transfer;
+	u32 *ptr;
+
+	switch (shader) {
+	case PIPE_SHADER_VERTEX:
+		id = R600_VS_CONSTANT;
+		type = R600_VS_CONSTANT_TYPE;
+		break;
+	case PIPE_SHADER_FRAGMENT:
+		id = R600_PS_CONSTANT;
+		type = R600_PS_CONSTANT_TYPE;
+		break;
+	default:
+		R600_ERR("unsupported %d\n", shader);
+		return;
+	}
+	if (buffer && buffer->width0 > 0) {
+		nconstant = buffer->width0 / 16;
+		ptr = pipe_buffer_map(ctx, buffer, PIPE_TRANSFER_READ, &transfer);
+		if (ptr == NULL)
+			return;
+		for (i = 0; i < nconstant; i++) {
+			rstate = radeon_state(rscreen->rw, type, id + i);
+			if (rstate == NULL)
+				return;
+			rstate->states[R600_PS_CONSTANT__SQ_ALU_CONSTANT0_0] = ptr[i * 4 + 0];
+			rstate->states[R600_PS_CONSTANT__SQ_ALU_CONSTANT1_0] = ptr[i * 4 + 1];
+			rstate->states[R600_PS_CONSTANT__SQ_ALU_CONSTANT2_0] = ptr[i * 4 + 2];
+			rstate->states[R600_PS_CONSTANT__SQ_ALU_CONSTANT3_0] = ptr[i * 4 + 3];
+			if (radeon_state_pm4(rstate))
+				return;
+			if (radeon_draw_set_new(rctx->draw, rstate))
+				return;
+		}
+		pipe_buffer_unmap(ctx, buffer, transfer);
+	}
+}
+
+static void r600_set_ps_sampler_view(struct pipe_context *ctx,
+					unsigned count,
+					struct pipe_sampler_view **views)
+{
+	struct r600_context *rctx = r600_context(ctx);
+	struct r600_context_state *rstate;
+	unsigned i;
+
+	for (i = 0; i < rctx->ps_nsampler_view; i++) {
+		rctx->ps_sampler_view[i] = r600_context_state_decref(rctx->ps_sampler_view[i]);
+	}
+	for (i = 0; i < count; i++) {
+		rstate = (struct r600_context_state *)views[i];
+		rctx->ps_sampler_view[i] = r600_context_state_incref(rstate);
+	}
+	rctx->ps_nsampler_view = count;
+}
+
+static void r600_set_vs_sampler_view(struct pipe_context *ctx,
+					unsigned count,
+					struct pipe_sampler_view **views)
+{
+	struct r600_context *rctx = r600_context(ctx);
+	struct r600_context_state *rstate;
+	unsigned i;
+
+	for (i = 0; i < rctx->vs_nsampler_view; i++) {
+		rctx->vs_sampler_view[i] = r600_context_state_decref(rctx->vs_sampler_view[i]);
+	}
+	for (i = 0; i < count; i++) {
+		rstate = (struct r600_context_state *)views[i];
+		rctx->vs_sampler_view[i] = r600_context_state_incref(rstate);
+	}
+	rctx->vs_nsampler_view = count;
+}
+
+static void r600_set_framebuffer_state(struct pipe_context *ctx,
+					const struct pipe_framebuffer_state *state)
+{
+	struct r600_context *rctx = r600_context(ctx);
+	struct r600_context_state *rstate;
+
+	rstate = r600_context_state(rctx, pipe_framebuffer_type, state);
+	r600_bind_state(ctx, rstate);
+}
+
+static void r600_set_polygon_stipple(struct pipe_context *ctx,
+					 const struct pipe_poly_stipple *state)
+{
+}
+
+static void r600_set_sample_mask(struct pipe_context *pipe, unsigned sample_mask)
+{
+}
+
+static void r600_set_scissor_state(struct pipe_context *ctx,
+					const struct pipe_scissor_state *state)
+{
+	struct r600_context *rctx = r600_context(ctx);
+	struct r600_context_state *rstate;
+
+	rstate = r600_context_state(rctx, pipe_scissor_type, state);
+	r600_bind_state(ctx, rstate);
+	/* refcount is taken care of this */
+	r600_delete_state(ctx, rstate);
+}
+
+static void r600_set_stencil_ref(struct pipe_context *ctx,
+				const struct pipe_stencil_ref *state)
+{
+	struct r600_context *rctx = r600_context(ctx);
+	struct r600_context_state *rstate;
+
+	rstate = r600_context_state(rctx, pipe_stencil_ref_type, state);
+	r600_bind_state(ctx, rstate);
+	/* refcount is taken care of this */
+	r600_delete_state(ctx, rstate);
+}
+
+static void r600_set_vertex_buffers(struct pipe_context *ctx,
+					unsigned count,
+					const struct pipe_vertex_buffer *buffers)
+{
+	struct r600_context *rctx = r600_context(ctx);
+	unsigned i;
+
+	for (i = 0; i < rctx->nvertex_buffer; i++) {
+		pipe_resource_reference(&rctx->vertex_buffer[i].buffer, NULL);
+	}
+	memcpy(rctx->vertex_buffer, buffers, sizeof(struct pipe_vertex_buffer) * count);
+	for (i = 0; i < count; i++) {
+		rctx->vertex_buffer[i].buffer = NULL;
+		pipe_resource_reference(&rctx->vertex_buffer[i].buffer, buffers[i].buffer);
+	}
+	rctx->nvertex_buffer = count;
+}
+
+static void r600_set_index_buffer(struct pipe_context *ctx,
+				  const struct pipe_index_buffer *ib)
+{
+	struct r600_context *rctx = r600_context(ctx);
+
+	if (ib) {
+		pipe_resource_reference(&rctx->index_buffer.buffer, ib->buffer);
+		memcpy(&rctx->index_buffer, ib, sizeof(rctx->index_buffer));
+	} else {
+		pipe_resource_reference(&rctx->index_buffer.buffer, NULL);
+		memset(&rctx->index_buffer, 0, sizeof(rctx->index_buffer));
+	}
+
+	/* TODO make this more like a state */
+}
+
+static void r600_set_viewport_state(struct pipe_context *ctx,
+					const struct pipe_viewport_state *state)
+{
+	struct r600_context *rctx = r600_context(ctx);
+	struct r600_context_state *rstate;
+
+	rstate = r600_context_state(rctx, pipe_viewport_type, state);
+	r600_bind_state(ctx, rstate);
+	r600_delete_state(ctx, rstate);
+}
+
+void r600_init_state_functions(struct r600_context *rctx)
+{
+	rctx->context.create_blend_state = r600_create_blend_state;
+	rctx->context.create_depth_stencil_alpha_state = r600_create_dsa_state;
+	rctx->context.create_fs_state = r600_create_shader_state;
+	rctx->context.create_rasterizer_state = r600_create_rs_state;
+	rctx->context.create_sampler_state = r600_create_sampler_state;
+	rctx->context.create_sampler_view = r600_create_sampler_view;
+	rctx->context.create_vertex_elements_state = r600_create_vertex_elements;
+	rctx->context.create_vs_state = r600_create_shader_state;
+	rctx->context.bind_blend_state = r600_bind_state;
+	rctx->context.bind_depth_stencil_alpha_state = r600_bind_state;
+	rctx->context.bind_fragment_sampler_states = r600_bind_ps_sampler;
+	rctx->context.bind_fs_state = r600_bind_ps_shader;
+	rctx->context.bind_rasterizer_state = r600_bind_state;
+	rctx->context.bind_vertex_elements_state = r600_bind_vertex_elements;
+	rctx->context.bind_vertex_sampler_states = r600_bind_vs_sampler;
+	rctx->context.bind_vs_state = r600_bind_vs_shader;
+	rctx->context.delete_blend_state = r600_delete_state;
+	rctx->context.delete_depth_stencil_alpha_state = r600_delete_state;
+	rctx->context.delete_fs_state = r600_delete_state;
+	rctx->context.delete_rasterizer_state = r600_delete_state;
+	rctx->context.delete_sampler_state = r600_delete_state;
+	rctx->context.delete_vertex_elements_state = r600_delete_vertex_element;
+	rctx->context.delete_vs_state = r600_delete_state;
+	rctx->context.set_blend_color = r600_set_blend_color;
+	rctx->context.set_clip_state = r600_set_clip_state;
+	rctx->context.set_constant_buffer = r600_set_constant_buffer;
+	rctx->context.set_fragment_sampler_views = r600_set_ps_sampler_view;
+	rctx->context.set_framebuffer_state = r600_set_framebuffer_state;
+	rctx->context.set_polygon_stipple = r600_set_polygon_stipple;
+	rctx->context.set_sample_mask = r600_set_sample_mask;
+	rctx->context.set_scissor_state = r600_set_scissor_state;
+	rctx->context.set_stencil_ref = r600_set_stencil_ref;
+	rctx->context.set_vertex_buffers = r600_set_vertex_buffers;
+	rctx->context.set_index_buffer = r600_set_index_buffer;
+	rctx->context.set_vertex_sampler_views = r600_set_vs_sampler_view;
+	rctx->context.set_viewport_state = r600_set_viewport_state;
+	rctx->context.sampler_view_destroy = r600_sampler_view_destroy;
+}
+
+struct r600_context_state *r600_context_state_incref(struct r600_context_state *rstate)
+{
+	if (rstate == NULL)
+		return NULL;
+	rstate->refcount++;
+	return rstate;
+}
+
+struct r600_context_state *r600_context_state_decref(struct r600_context_state *rstate)
+{
+	unsigned i;
+
+	if (rstate == NULL)
+		return NULL;
+	if (--rstate->refcount)
+		return NULL;
+	switch (rstate->type) {
+	case pipe_sampler_view_type:
+		pipe_resource_reference(&rstate->state.sampler_view.texture, NULL);
+		break;
+	case pipe_framebuffer_type:
+		for (i = 0; i < rstate->state.framebuffer.nr_cbufs; i++) {
+			pipe_surface_reference(&rstate->state.framebuffer.cbufs[i], NULL);
+		}
+		pipe_surface_reference(&rstate->state.framebuffer.zsbuf, NULL);
+		break;
+	case pipe_viewport_type:
+	case pipe_depth_type:
+	case pipe_rasterizer_type:
+	case pipe_poly_stipple_type:
+	case pipe_scissor_type:
+	case pipe_clip_type:
+	case pipe_stencil_type:
+	case pipe_alpha_type:
+	case pipe_dsa_type:
+	case pipe_blend_type:
+	case pipe_stencil_ref_type:
+	case pipe_shader_type:
+	case pipe_sampler_type:
+		break;
+	default:
+		R600_ERR("invalid type %d\n", rstate->type);
+		return NULL;
+	}
+	radeon_state_decref(rstate->rstate);
+	FREE(rstate);
+	return NULL;
+}
+
+struct r600_context_state *r600_context_state(struct r600_context *rctx, unsigned type, const void *state)
+{
+	struct r600_context_state *rstate = CALLOC_STRUCT(r600_context_state);
+	const union pipe_states *states = state;
+	unsigned i;
+	int r;
+
+	if (rstate == NULL)
+		return NULL;
+	rstate->type = type;
+	rstate->refcount = 1;
+
+	switch (rstate->type) {
+	case pipe_sampler_view_type:
+		rstate->state.sampler_view = (*states).sampler_view;
+		rstate->state.sampler_view.texture = NULL;
+		break;
+	case pipe_framebuffer_type:
+		rstate->state.framebuffer = (*states).framebuffer;
+		for (i = 0; i < rstate->state.framebuffer.nr_cbufs; i++) {
+			pipe_surface_reference(&rstate->state.framebuffer.cbufs[i],
+						(*states).framebuffer.cbufs[i]);
+		}
+		pipe_surface_reference(&rstate->state.framebuffer.zsbuf,
+					(*states).framebuffer.zsbuf);
+		break;
+	case pipe_viewport_type:
+		rstate->state.viewport = (*states).viewport;
+		break;
+	case pipe_depth_type:
+		rstate->state.depth = (*states).depth;
+		break;
+	case pipe_rasterizer_type:
+		rstate->state.rasterizer = (*states).rasterizer;
+		break;
+	case pipe_poly_stipple_type:
+		rstate->state.poly_stipple = (*states).poly_stipple;
+		break;
+	case pipe_scissor_type:
+		rstate->state.scissor = (*states).scissor;
+		break;
+	case pipe_clip_type:
+		rstate->state.clip = (*states).clip;
+		break;
+	case pipe_stencil_type:
+		rstate->state.stencil = (*states).stencil;
+		break;
+	case pipe_alpha_type:
+		rstate->state.alpha = (*states).alpha;
+		break;
+	case pipe_dsa_type:
+		rstate->state.dsa = (*states).dsa;
+		break;
+	case pipe_blend_type:
+		rstate->state.blend = (*states).blend;
+		break;
+	case pipe_stencil_ref_type:
+		rstate->state.stencil_ref = (*states).stencil_ref;
+		break;
+	case pipe_shader_type:
+		rstate->state.shader = (*states).shader;
+		r =  r600_pipe_shader_create(&rctx->context, rstate, rstate->state.shader.tokens);
+		if (r) {
+			r600_context_state_decref(rstate);
+			return NULL;
+		}
+		break;
+	case pipe_sampler_type:
+		rstate->state.sampler = (*states).sampler;
+		break;
+	default:
+		R600_ERR("invalid type %d\n", rstate->type);
+		FREE(rstate);
+		return NULL;
+	}
+	return rstate;
+}
+
+static struct radeon_state *r600_blend(struct r600_context *rctx)
+{
+	struct r600_screen *rscreen = rctx->screen;
 	struct radeon_state *rstate;
+	const struct pipe_blend_state *state = &rctx->blend->state.blend;
+	int i;
 
 	rstate = radeon_state(rscreen->rw, R600_BLEND_TYPE, R600_BLEND);
 	if (rstate == NULL)
 		return NULL;
-	rstate->states[R600_BLEND__CB_BLEND_RED] = 0x00000000;
-	rstate->states[R600_BLEND__CB_BLEND_GREEN] = 0x00000000;
-	rstate->states[R600_BLEND__CB_BLEND_BLUE] = 0x00000000;
-	rstate->states[R600_BLEND__CB_BLEND_ALPHA] = 0x00000000;
-	rstate->states[R600_BLEND__CB_BLEND0_CONTROL] = 0x00010001;
+	rstate->states[R600_BLEND__CB_BLEND_RED] = fui(rctx->blend_color.color[0]);
+	rstate->states[R600_BLEND__CB_BLEND_GREEN] = fui(rctx->blend_color.color[1]);
+	rstate->states[R600_BLEND__CB_BLEND_BLUE] = fui(rctx->blend_color.color[2]);
+	rstate->states[R600_BLEND__CB_BLEND_ALPHA] = fui(rctx->blend_color.color[3]);
+	rstate->states[R600_BLEND__CB_BLEND0_CONTROL] = 0x00000000;
 	rstate->states[R600_BLEND__CB_BLEND1_CONTROL] = 0x00000000;
 	rstate->states[R600_BLEND__CB_BLEND2_CONTROL] = 0x00000000;
 	rstate->states[R600_BLEND__CB_BLEND3_CONTROL] = 0x00000000;
@@ -62,6 +631,36 @@ static void *r600_create_blend_state(struct pipe_context *ctx,
 	rstate->states[R600_BLEND__CB_BLEND6_CONTROL] = 0x00000000;
 	rstate->states[R600_BLEND__CB_BLEND7_CONTROL] = 0x00000000;
 	rstate->states[R600_BLEND__CB_BLEND_CONTROL] = 0x00000000;
+
+	for (i = 0; i < 8; i++) {
+		unsigned eqRGB = state->rt[i].rgb_func;
+		unsigned srcRGB = state->rt[i].rgb_src_factor;
+		unsigned dstRGB = state->rt[i].rgb_dst_factor;
+		
+		unsigned eqA = state->rt[i].alpha_func;
+		unsigned srcA = state->rt[i].alpha_src_factor;
+		unsigned dstA = state->rt[i].alpha_dst_factor;
+		uint32_t bc = 0;
+
+		if (!state->rt[i].blend_enable)
+			continue;
+
+		bc |= S_028804_COLOR_COMB_FCN(r600_translate_blend_function(eqRGB));
+		bc |= S_028804_COLOR_SRCBLEND(r600_translate_blend_factor(srcRGB));
+		bc |= S_028804_COLOR_DESTBLEND(r600_translate_blend_factor(dstRGB));
+
+		if (srcA != srcRGB || dstA != dstRGB || eqA != eqRGB) {
+			bc |= S_028804_SEPARATE_ALPHA_BLEND(1);
+			bc |= S_028804_ALPHA_COMB_FCN(r600_translate_blend_function(eqA));
+			bc |= S_028804_ALPHA_SRCBLEND(r600_translate_blend_factor(srcA));
+			bc |= S_028804_ALPHA_DESTBLEND(r600_translate_blend_factor(dstA));
+		}
+
+		rstate->states[R600_BLEND__CB_BLEND0_CONTROL + i] = bc;
+		if (i == 0)
+			rstate->states[R600_BLEND__CB_BLEND_CONTROL] = bc;
+	}
+
 	if (radeon_state_pm4(rstate)) {
 		radeon_state_decref(rstate);
 		return NULL;
@@ -69,38 +668,24 @@ static void *r600_create_blend_state(struct pipe_context *ctx,
 	return rstate;
 }
 
-static void r600_bind_blend_state(struct pipe_context *ctx, void *state)
-{
-	struct r600_context *rctx = r600_context(ctx);
-	radeon_draw_set(rctx->draw, state);
-}
-
-static void r600_set_blend_color(struct pipe_context *ctx,
-					const struct pipe_blend_color *color)
+static struct radeon_state *r600_cb(struct r600_context *rctx, int cb)
 {
-}
-
-static void r600_set_clip_state(struct pipe_context *ctx,
-				const struct pipe_clip_state *state)
-{
-}
-
-static void r600_set_framebuffer_state(struct pipe_context *ctx,
-					const struct pipe_framebuffer_state *state)
-{
-	struct r600_screen *rscreen = r600_screen(ctx->screen);
-	struct r600_context *rctx = r600_context(ctx);
-	struct r600_texture *rtex;
-	struct r600_buffer *rbuffer;
+	struct r600_screen *rscreen = rctx->screen;
+	struct r600_resource_texture *rtex;
+	struct r600_resource *rbuffer;
 	struct radeon_state *rstate;
-	unsigned level = state->cbufs[0]->level;
+	const struct pipe_framebuffer_state *state = &rctx->framebuffer->state.framebuffer;
+	unsigned level = state->cbufs[cb]->level;
 	unsigned pitch, slice;
+	unsigned color_info;
+	unsigned format, swap, ntype;
+	const struct util_format_description *desc;
 
-	rstate = radeon_state(rscreen->rw, R600_CB0_TYPE, R600_CB0);
+	rstate = radeon_state(rscreen->rw, R600_CB0_TYPE + cb, R600_CB0 + cb);
 	if (rstate == NULL)
-		return;
-	rtex = (struct r600_texture*)state->cbufs[0]->texture;
-	rbuffer = (struct r600_buffer*)rtex->buffer;
+		return NULL;
+	rtex = (struct r600_resource_texture*)state->cbufs[cb]->texture;
+	rbuffer = &rtex->resource;
 	rstate->bo[0] = radeon_bo_incref(rscreen->rw, rbuffer->bo);
 	rstate->bo[1] = radeon_bo_incref(rscreen->rw, rbuffer->bo);
 	rstate->bo[2] = radeon_bo_incref(rscreen->rw, rbuffer->bo);
@@ -108,10 +693,25 @@ static void r600_set_framebuffer_state(struct pipe_context *ctx,
 	rstate->placement[2] = RADEON_GEM_DOMAIN_GTT;
 	rstate->placement[4] = RADEON_GEM_DOMAIN_GTT;
 	rstate->nbo = 3;
-	pitch = rtex->pitch[level] / 8 - 1;
-	slice = rtex->pitch[level] * state->cbufs[0]->height / 64 - 1;
+	pitch = (rtex->pitch[level] / rtex->bpt) / 8 - 1;
+	slice = (rtex->pitch[level] / rtex->bpt) * state->cbufs[cb]->height / 64 - 1;
+
+	ntype = 0;
+	desc = util_format_description(rtex->resource.base.b.format);
+	if (desc->colorspace == UTIL_FORMAT_COLORSPACE_SRGB)
+		ntype = V_0280A0_NUMBER_SRGB;
+
+	format = r600_translate_colorformat(rtex->resource.base.b.format);
+	swap = r600_translate_colorswap(rtex->resource.base.b.format);
+
+	color_info = S_0280A0_FORMAT(format) |
+		S_0280A0_COMP_SWAP(swap) |
+		S_0280A0_BLEND_CLAMP(1) |
+		S_0280A0_SOURCE_FORMAT(1) |
+		S_0280A0_NUMBER_TYPE(ntype);
+
 	rstate->states[R600_CB0__CB_COLOR0_BASE] = 0x00000000;
-	rstate->states[R600_CB0__CB_COLOR0_INFO] = 0x08110068;
+	rstate->states[R600_CB0__CB_COLOR0_INFO] = color_info;
 	rstate->states[R600_CB0__CB_COLOR0_SIZE] = S_028060_PITCH_TILE_MAX(pitch) |
 						S_028060_SLICE_TILE_MAX(slice);
 	rstate->states[R600_CB0__CB_COLOR0_VIEW] = 0x00000000;
@@ -120,83 +720,124 @@ static void r600_set_framebuffer_state(struct pipe_context *ctx,
 	rstate->states[R600_CB0__CB_COLOR0_MASK] = 0x00000000;
 	if (radeon_state_pm4(rstate)) {
 		radeon_state_decref(rstate);
-		return;
+		return NULL;
 	}
-	radeon_draw_set_new(rctx->draw, rstate);
-	rctx->db = radeon_state_decref(rctx->db);
-	if(state->zsbuf) {
-		rtex = (struct r600_texture*)state->zsbuf->texture;
-		rbuffer = (struct r600_buffer*)rtex->buffer;
-		rctx->db = radeon_state(rscreen->rw, R600_DB_TYPE, R600_DB);
-		if(rctx->db == NULL)
-		     return;
-		rctx->db->bo[0] = radeon_bo_incref(rscreen->rw, rbuffer->bo);
-		rctx->db->nbo = 1;
-		rctx->db->placement[0] = RADEON_GEM_DOMAIN_VRAM;
-		level = state->zsbuf->level;
-		pitch = rtex->pitch[level] / 8 - 1;
-		slice = rtex->pitch[level] * state->zsbuf->height / 64 - 1;
-
-		rctx->db->states[R600_DB__DB_DEPTH_BASE] = 0x00000000;
-		rctx->db->states[R600_DB__DB_DEPTH_INFO] = 0x00010006;
-		rctx->db->states[R600_DB__DB_DEPTH_VIEW] = 0x00000000;
-		rctx->db->states[R600_DB__DB_PREFETCH_LIMIT] = (state->zsbuf->height / 8) -1;
-		rctx->db->states[R600_DB__DB_DEPTH_SIZE] = S_028000_PITCH_TILE_MAX(pitch) |
-						S_028000_SLICE_TILE_MAX(slice);
-	} else 
-		rctx->db = NULL;
-	rctx->fb_state = *state;
-}
-
-static void *r600_create_fs_state(struct pipe_context *ctx,
-					const struct pipe_shader_state *shader)
-{
-	return r600_pipe_shader_create(ctx, C_PROGRAM_TYPE_FS, shader->tokens);
-}
-
-static void r600_bind_fs_state(struct pipe_context *ctx, void *state)
-{
-	struct r600_context *rctx = r600_context(ctx);
-
-	rctx->ps_shader = state;
+	return rstate;
 }
 
-static void *r600_create_vs_state(struct pipe_context *ctx,
-					const struct pipe_shader_state *shader)
+static struct radeon_state *r600_db(struct r600_context *rctx)
 {
-	return r600_pipe_shader_create(ctx, C_PROGRAM_TYPE_VS, shader->tokens);
-}
+	struct r600_screen *rscreen = rctx->screen;
+	struct r600_resource_texture *rtex;
+	struct r600_resource *rbuffer;
+	struct radeon_state *rstate;
+	const struct pipe_framebuffer_state *state = &rctx->framebuffer->state.framebuffer;
+	unsigned level;
+	unsigned pitch, slice, format;
 
-static void r600_bind_vs_state(struct pipe_context *ctx, void *state)
-{
-	struct r600_context *rctx = r600_context(ctx);
+	if (state->zsbuf == NULL)
+		return NULL;
 
-	rctx->vs_shader = state;
-}
+	rstate = radeon_state(rscreen->rw, R600_DB_TYPE, R600_DB);
+	if (rstate == NULL)
+		return NULL;
 
-static void r600_set_polygon_stipple(struct pipe_context *ctx,
-					 const struct pipe_poly_stipple *state)
-{
+	rtex = (struct r600_resource_texture*)state->zsbuf->texture;
+	rbuffer = &rtex->resource;
+	rstate->bo[0] = radeon_bo_incref(rscreen->rw, rbuffer->bo);
+	rstate->nbo = 1;
+	rstate->placement[0] = RADEON_GEM_DOMAIN_VRAM;
+	level = state->zsbuf->level;
+	pitch = (rtex->pitch[level] / rtex->bpt) / 8 - 1;
+	slice = (rtex->pitch[level] / rtex->bpt) * state->zsbuf->height / 64 - 1;
+	format = r600_translate_dbformat(state->zsbuf->texture->format);
+	rstate->states[R600_DB__DB_DEPTH_BASE] = 0x00000000;
+	rstate->states[R600_DB__DB_DEPTH_INFO] = 0x00010000 |
+					S_028010_FORMAT(format);
+	rstate->states[R600_DB__DB_DEPTH_VIEW] = 0x00000000;
+	rstate->states[R600_DB__DB_PREFETCH_LIMIT] = (state->zsbuf->height / 8) -1;
+	rstate->states[R600_DB__DB_DEPTH_SIZE] = S_028000_PITCH_TILE_MAX(pitch) |
+						S_028000_SLICE_TILE_MAX(slice);
+	if (radeon_state_pm4(rstate)) {
+		radeon_state_decref(rstate);
+		return NULL;
+	}
+	return rstate;
 }
 
-static void *r600_create_rs_state(struct pipe_context *ctx,
-					const struct pipe_rasterizer_state *state)
+static struct radeon_state *r600_rasterizer(struct r600_context *rctx)
 {
-	struct r600_screen *rscreen = r600_screen(ctx->screen);
-	struct r600_context *rctx = r600_context(ctx);
+	const struct pipe_rasterizer_state *state = &rctx->rasterizer->state.rasterizer;
+	const struct pipe_framebuffer_state *fb = &rctx->framebuffer->state.framebuffer;
+	struct r600_screen *rscreen = rctx->screen;
 	struct radeon_state *rstate;
+	float offset_units = 0, offset_scale = 0;
+	char depth = 0;
+	unsigned offset_db_fmt_cntl = 0;
+	unsigned tmp;
+	unsigned prov_vtx = 1;
+	if (fb->zsbuf) {
+		offset_units = state->offset_units;
+		offset_scale = state->offset_scale * 12.0f;
+		switch (fb->zsbuf->texture->format) {
+		case PIPE_FORMAT_Z24X8_UNORM:
+		case PIPE_FORMAT_Z24_UNORM_S8_USCALED:
+			depth = -24;
+			offset_units *= 2.0f;
+			break;
+		case PIPE_FORMAT_Z32_FLOAT:
+			depth = -23;
+			offset_units *= 1.0f;
+			offset_db_fmt_cntl |= S_028DF8_POLY_OFFSET_DB_IS_FLOAT_FMT(1);
+			break;
+		case PIPE_FORMAT_Z16_UNORM:
+			depth = -16;
+			offset_units *= 4.0f;
+			break;
+		default:
+			R600_ERR("unsupported %d\n", fb->zsbuf->texture->format);
+			return NULL;
+		}
+	}
+	offset_db_fmt_cntl |= S_028DF8_POLY_OFFSET_NEG_NUM_DB_BITS(depth);
+
+	if (state->flatshade_first)
+		prov_vtx = 0;
 
 	rctx->flat_shade = state->flatshade;
 	rstate = radeon_state(rscreen->rw, R600_RASTERIZER_TYPE, R600_RASTERIZER);
 	if (rstate == NULL)
 		return NULL;
 	rstate->states[R600_RASTERIZER__SPI_INTERP_CONTROL_0] = 0x00000001;
+	if (state->sprite_coord_enable) {
+		rstate->states[R600_RASTERIZER__SPI_INTERP_CONTROL_0] |=
+				S_0286D4_PNT_SPRITE_ENA(1) |
+				S_0286D4_PNT_SPRITE_OVRD_X(2) |
+				S_0286D4_PNT_SPRITE_OVRD_Y(3) |
+				S_0286D4_PNT_SPRITE_OVRD_Z(0) |
+				S_0286D4_PNT_SPRITE_OVRD_W(1);
+		if (state->sprite_coord_mode != PIPE_SPRITE_COORD_UPPER_LEFT) {
+			rstate->states[R600_RASTERIZER__SPI_INTERP_CONTROL_0] |=
+					S_0286D4_PNT_SPRITE_TOP_1(1);
+		}
+	}
 	rstate->states[R600_RASTERIZER__PA_CL_CLIP_CNTL] = 0x00000000;
-	rstate->states[R600_RASTERIZER__PA_SU_SC_MODE_CNTL] = 0x00080000;
-	rstate->states[R600_RASTERIZER__PA_CL_VS_OUT_CNTL] = 0x00000000;
+	rstate->states[R600_RASTERIZER__PA_SU_SC_MODE_CNTL] =
+		S_028814_PROVOKING_VTX_LAST(prov_vtx) |
+		S_028814_CULL_FRONT((state->cull_face & PIPE_FACE_FRONT) ? 1 : 0) |
+		S_028814_CULL_BACK((state->cull_face & PIPE_FACE_BACK) ? 1 : 0) |
+		S_028814_FACE(!state->front_ccw) |
+		S_028814_POLY_OFFSET_FRONT_ENABLE(state->offset_tri) |
+		S_028814_POLY_OFFSET_BACK_ENABLE(state->offset_tri) |
+		S_028814_POLY_OFFSET_PARA_ENABLE(state->offset_tri);
+	rstate->states[R600_RASTERIZER__PA_CL_VS_OUT_CNTL] =
+			S_02881C_USE_VTX_POINT_SIZE(state->point_size_per_vertex) |
+			S_02881C_VS_OUT_MISC_VEC_ENA(state->point_size_per_vertex);
 	rstate->states[R600_RASTERIZER__PA_CL_NANINF_CNTL] = 0x00000000;
-	rstate->states[R600_RASTERIZER__PA_SU_POINT_SIZE] = 0x00080008;
-	rstate->states[R600_RASTERIZER__PA_SU_POINT_MINMAX] = 0x00000000;
+	/* point size 12.4 fixed point */
+	tmp = (unsigned)(state->point_size * 8.0 / 2.0);
+	rstate->states[R600_RASTERIZER__PA_SU_POINT_SIZE] = S_028A00_HEIGHT(tmp) | S_028A00_WIDTH(tmp);
+	rstate->states[R600_RASTERIZER__PA_SU_POINT_MINMAX] = 0x80000000;
 	rstate->states[R600_RASTERIZER__PA_SU_LINE_CNTL] = 0x00000008;
 	rstate->states[R600_RASTERIZER__PA_SC_LINE_STIPPLE] = 0x00000005;
 	rstate->states[R600_RASTERIZER__PA_SC_MPASS_PS_CNTL] = 0x00000000;
@@ -205,12 +846,12 @@ static void *r600_create_rs_state(struct pipe_context *ctx,
 	rstate->states[R600_RASTERIZER__PA_CL_GB_VERT_DISC_ADJ] = 0x3F800000;
 	rstate->states[R600_RASTERIZER__PA_CL_GB_HORZ_CLIP_ADJ] = 0x3F800000;
 	rstate->states[R600_RASTERIZER__PA_CL_GB_HORZ_DISC_ADJ] = 0x3F800000;
-	rstate->states[R600_RASTERIZER__PA_SU_POLY_OFFSET_DB_FMT_CNTL] = 0x00000000;
+	rstate->states[R600_RASTERIZER__PA_SU_POLY_OFFSET_DB_FMT_CNTL] = offset_db_fmt_cntl;
 	rstate->states[R600_RASTERIZER__PA_SU_POLY_OFFSET_CLAMP] = 0x00000000;
-	rstate->states[R600_RASTERIZER__PA_SU_POLY_OFFSET_FRONT_SCALE] = 0x00000000;
-	rstate->states[R600_RASTERIZER__PA_SU_POLY_OFFSET_FRONT_OFFSET] = 0x00000000;
-	rstate->states[R600_RASTERIZER__PA_SU_POLY_OFFSET_BACK_SCALE] = 0x00000000;
-	rstate->states[R600_RASTERIZER__PA_SU_POLY_OFFSET_BACK_OFFSET] = 0x00000000;
+	rstate->states[R600_RASTERIZER__PA_SU_POLY_OFFSET_FRONT_SCALE] = fui(offset_scale);
+	rstate->states[R600_RASTERIZER__PA_SU_POLY_OFFSET_FRONT_OFFSET] = fui(offset_units);
+	rstate->states[R600_RASTERIZER__PA_SU_POLY_OFFSET_BACK_SCALE] = fui(offset_scale);
+	rstate->states[R600_RASTERIZER__PA_SU_POLY_OFFSET_BACK_OFFSET] = fui(offset_units);
 	if (radeon_state_pm4(rstate)) {
 		radeon_state_decref(rstate);
 		return NULL;
@@ -218,64 +859,31 @@ static void *r600_create_rs_state(struct pipe_context *ctx,
 	return rstate;
 }
 
-static void r600_bind_rs_state(struct pipe_context *ctx, void *state)
+static struct radeon_state *r600_scissor(struct r600_context *rctx)
 {
-	struct r600_context *rctx = r600_context(ctx);
-	radeon_draw_set(rctx->draw, state);
-}
-
-static void *r600_create_sampler_state(struct pipe_context *ctx,
-					const struct pipe_sampler_state *state)
-{
-	return NULL;
-}
-
-static void r600_bind_sampler_states(struct pipe_context *ctx,
-					unsigned count, void **states)
-{
-}
-
-static struct pipe_sampler_view *r600_create_sampler_view(struct pipe_context *ctx,
-							  struct pipe_resource *texture,
-							  const struct pipe_sampler_view *templ)
-{
-	struct pipe_sampler_view *view = CALLOC_STRUCT(pipe_sampler_view);
-
-	*view = *templ;
-	return view;
-}
-
-static void r600_sampler_view_destroy(struct pipe_context *ctx,
-				      struct pipe_sampler_view *view)
-{
-	FREE(view);
-}
-
-static void r600_set_fragment_sampler_views(struct pipe_context *ctx,
-					    unsigned count,
-					    struct pipe_sampler_view **views)
-{
-}
-
-static void r600_set_vertex_sampler_views(struct pipe_context *ctx,
-					  unsigned count,
-					  struct pipe_sampler_view **views)
-{
-}
-
-static void r600_set_scissor_state(struct pipe_context *ctx,
-					const struct pipe_scissor_state *state)
-{
-	struct r600_screen *rscreen = r600_screen(ctx->screen);
-	struct r600_context *rctx = r600_context(ctx);
+	const struct pipe_scissor_state *state = &rctx->scissor->state.scissor;
+	const struct pipe_framebuffer_state *fb = &rctx->framebuffer->state.framebuffer;
+	struct r600_screen *rscreen = rctx->screen;
 	struct radeon_state *rstate;
+	unsigned minx, maxx, miny, maxy;
 	u32 tl, br;
 
-	tl = S_028240_TL_X(state->minx) | S_028240_TL_Y(state->miny) | S_028240_WINDOW_OFFSET_DISABLE(1);
-	br = S_028244_BR_X(state->maxx) | S_028244_BR_Y(state->maxy);
+	if (state == NULL) {
+		minx = 0;
+		miny = 0;
+		maxx = fb->cbufs[0]->width;
+		maxy = fb->cbufs[0]->height;
+	} else {
+		minx = state->minx;
+		miny = state->miny;
+		maxx = state->maxx;
+		maxy = state->maxy;
+	}
+	tl = S_028240_TL_X(minx) | S_028240_TL_Y(miny) | S_028240_WINDOW_OFFSET_DISABLE(1);
+	br = S_028244_BR_X(maxx) | S_028244_BR_Y(maxy);
 	rstate = radeon_state(rscreen->rw, R600_SCISSOR_TYPE, R600_SCISSOR);
 	if (rstate == NULL)
-		return;
+		return NULL;
 	rstate->states[R600_SCISSOR__PA_SC_SCREEN_SCISSOR_TL] = tl;
 	rstate->states[R600_SCISSOR__PA_SC_SCREEN_SCISSOR_BR] = br;
 	rstate->states[R600_SCISSOR__PA_SC_WINDOW_OFFSET] = 0x00000000;
@@ -297,97 +905,104 @@ static void r600_set_scissor_state(struct pipe_context *ctx,
 	rstate->states[R600_SCISSOR__PA_SC_VPORT_SCISSOR_0_BR] = br;
 	if (radeon_state_pm4(rstate)) {
 		radeon_state_decref(rstate);
-		return;
+		return NULL;
 	}
-	radeon_draw_set_new(rctx->draw, rstate);
+	return rstate;
 }
 
-static void r600_set_viewport_state(struct pipe_context *ctx,
-					const struct pipe_viewport_state *state)
+static struct radeon_state *r600_viewport(struct r600_context *rctx)
 {
-	struct r600_screen *rscreen = r600_screen(ctx->screen);
-	struct r600_context *rctx = r600_context(ctx);
+	const struct pipe_viewport_state *state = &rctx->viewport->state.viewport;
+	struct r600_screen *rscreen = rctx->screen;
 	struct radeon_state *rstate;
 
 	rstate = radeon_state(rscreen->rw, R600_VIEWPORT_TYPE, R600_VIEWPORT);
 	if (rstate == NULL)
-		return;
+		return NULL;
 	rstate->states[R600_VIEWPORT__PA_SC_VPORT_ZMIN_0] = 0x00000000;
 	rstate->states[R600_VIEWPORT__PA_SC_VPORT_ZMAX_0] = 0x3F800000;
-	rstate->states[R600_VIEWPORT__PA_CL_VPORT_XSCALE_0] = r600_float_to_u32(state->scale[0]);
-	rstate->states[R600_VIEWPORT__PA_CL_VPORT_YSCALE_0] = r600_float_to_u32(state->scale[1]);
-	rstate->states[R600_VIEWPORT__PA_CL_VPORT_ZSCALE_0] = r600_float_to_u32(state->scale[2]);
-	rstate->states[R600_VIEWPORT__PA_CL_VPORT_XOFFSET_0] = r600_float_to_u32(state->translate[0]);
-	rstate->states[R600_VIEWPORT__PA_CL_VPORT_YOFFSET_0] = r600_float_to_u32(state->translate[1]);
-	rstate->states[R600_VIEWPORT__PA_CL_VPORT_ZOFFSET_0] = r600_float_to_u32(state->translate[2]);
+	rstate->states[R600_VIEWPORT__PA_CL_VPORT_XSCALE_0] = fui(state->scale[0]);
+	rstate->states[R600_VIEWPORT__PA_CL_VPORT_YSCALE_0] = fui(state->scale[1]);
+	rstate->states[R600_VIEWPORT__PA_CL_VPORT_ZSCALE_0] = fui(state->scale[2]);
+	rstate->states[R600_VIEWPORT__PA_CL_VPORT_XOFFSET_0] = fui(state->translate[0]);
+	rstate->states[R600_VIEWPORT__PA_CL_VPORT_YOFFSET_0] = fui(state->translate[1]);
+	rstate->states[R600_VIEWPORT__PA_CL_VPORT_ZOFFSET_0] = fui(state->translate[2]);
 	rstate->states[R600_VIEWPORT__PA_CL_VTE_CNTL] = 0x0000043F;
 	if (radeon_state_pm4(rstate)) {
 		radeon_state_decref(rstate);
-		return;
+		return NULL;
 	}
-	radeon_draw_set_new(rctx->draw, rstate);
-	rctx->viewport = *state;
-}
-
-static void r600_set_vertex_buffers(struct pipe_context *ctx,
-					unsigned count,
-					const struct pipe_vertex_buffer *buffers)
-{
-	struct r600_context *rctx = r600_context(ctx);
-
-	memcpy(rctx->vertex_buffer, buffers, sizeof(struct pipe_vertex_buffer) * count);
-	rctx->nvertex_buffer = count;
+	return rstate;
 }
 
-
-static void *r600_create_vertex_elements_state(struct pipe_context *ctx,
-					       unsigned count,
-					       const struct pipe_vertex_element *elements)
+static struct radeon_state *r600_dsa(struct r600_context *rctx)
 {
-	struct r600_vertex_elements_state *v = CALLOC_STRUCT(r600_vertex_elements_state);
+	const struct pipe_depth_stencil_alpha_state *state = &rctx->dsa->state.dsa;
+	const struct pipe_stencil_ref *stencil_ref = &rctx->stencil_ref->state.stencil_ref;
+	struct r600_screen *rscreen = rctx->screen;
+	unsigned db_depth_control, alpha_test_control, alpha_ref, db_shader_control;
+	unsigned stencil_ref_mask, stencil_ref_mask_bf;
+	struct r600_shader *rshader = &rctx->ps_shader->shader;
+	struct radeon_state *rstate;
+	int i;
 
-	assert(count < 32);
-	v->count = count;
-	memcpy(v->elements, elements, count * sizeof(struct pipe_vertex_element));
-	return v;
-}
+	rstate = radeon_state(rscreen->rw, R600_DSA_TYPE, R600_DSA);
+	if (rstate == NULL)
+		return NULL;
 
-static void r600_bind_vertex_elements_state(struct pipe_context *ctx, void *state)
-{
-	struct r600_context *rctx = r600_context(ctx);
-	struct r600_vertex_elements_state *v = (struct r600_vertex_elements_state*)state;
+	db_shader_control = 0x210;
+	for (i = 0; i < rshader->noutput; i++) {
+		if (rshader->output[i].name == TGSI_SEMANTIC_POSITION)
+			db_shader_control |= 1;
+	}
+	stencil_ref_mask = 0;
+	stencil_ref_mask_bf = 0;
+	db_depth_control = S_028800_Z_ENABLE(state->depth.enabled) |
+		S_028800_Z_WRITE_ENABLE(state->depth.writemask) |
+		S_028800_ZFUNC(state->depth.func);
+	/* set stencil enable */
 
-	rctx->vertex_elements = v;
-}
+	if (state->stencil[0].enabled) {
+		db_depth_control |= S_028800_STENCIL_ENABLE(1);
+		db_depth_control |= S_028800_STENCILFUNC(r600_translate_ds_func(state->stencil[0].func));
+		db_depth_control |= S_028800_STENCILFAIL(r600_translate_stencil_op(state->stencil[0].fail_op));
+		db_depth_control |= S_028800_STENCILZPASS(r600_translate_stencil_op(state->stencil[0].zpass_op));
+		db_depth_control |= S_028800_STENCILZFAIL(r600_translate_stencil_op(state->stencil[0].zfail_op));
 
-static void r600_delete_vertex_elements_state(struct pipe_context *ctx, void *state)
-{
-	FREE(state);
-}
+		stencil_ref_mask = S_028430_STENCILMASK(state->stencil[0].valuemask) |
+			S_028430_STENCILWRITEMASK(state->stencil[0].writemask);
+		stencil_ref_mask |= S_028430_STENCILREF(stencil_ref->ref_value[0]);
+		if (state->stencil[1].enabled) {
+			db_depth_control |= S_028800_BACKFACE_ENABLE(1);
+			db_depth_control |= S_028800_STENCILFUNC_BF(r600_translate_ds_func(state->stencil[1].func));
+			db_depth_control |= S_028800_STENCILFAIL_BF(r600_translate_stencil_op(state->stencil[1].fail_op));
+			db_depth_control |= S_028800_STENCILZPASS_BF(r600_translate_stencil_op(state->stencil[1].zpass_op));
+			db_depth_control |= S_028800_STENCILZFAIL_BF(r600_translate_stencil_op(state->stencil[1].zfail_op));
+			stencil_ref_mask_bf = S_028434_STENCILMASK_BF(state->stencil[1].valuemask) |
+				S_028434_STENCILWRITEMASK_BF(state->stencil[1].writemask);
+			stencil_ref_mask_bf |= S_028430_STENCILREF(stencil_ref->ref_value[1]);
+		}
+	}
 
-static void *r600_create_dsa_state(struct pipe_context *ctx,
-					const struct pipe_depth_stencil_alpha_state *state)
-{
-	struct r600_screen *rscreen = r600_screen(ctx->screen);
-	struct radeon_state *rstate;
-	unsigned db_depth_control;
+	alpha_test_control = 0;
+	alpha_ref = 0;
+	if (state->alpha.enabled) {
+		alpha_test_control = S_028410_ALPHA_FUNC(state->alpha.func);
+		alpha_test_control |= S_028410_ALPHA_TEST_ENABLE(1);
+		alpha_ref = fui(state->alpha.ref_value);
+	}
 
-	rstate = radeon_state(rscreen->rw, R600_DSA_TYPE, R600_DSA);
-	if (rstate == NULL)
-		return NULL;
-	db_depth_control = 0x00700700 | S_028800_Z_ENABLE(state->depth.enabled) | S_028800_Z_WRITE_ENABLE(state->depth.writemask) | S_028800_ZFUNC(state->depth.func);
-	
 	rstate->states[R600_DSA__DB_STENCIL_CLEAR] = 0x00000000;
 	rstate->states[R600_DSA__DB_DEPTH_CLEAR] = 0x3F800000;
-	rstate->states[R600_DSA__SX_ALPHA_TEST_CONTROL] = 0x00000000;
-	rstate->states[R600_DSA__DB_STENCILREFMASK] = 0xFFFFFF00;
-	rstate->states[R600_DSA__DB_STENCILREFMASK_BF] = 0xFFFFFF00;
-	rstate->states[R600_DSA__SX_ALPHA_REF] = 0x00000000;
+	rstate->states[R600_DSA__SX_ALPHA_TEST_CONTROL] = alpha_test_control;
+	rstate->states[R600_DSA__DB_STENCILREFMASK] = stencil_ref_mask;
+	rstate->states[R600_DSA__DB_STENCILREFMASK_BF] = stencil_ref_mask_bf;
+	rstate->states[R600_DSA__SX_ALPHA_REF] = alpha_ref;
 	rstate->states[R600_DSA__SPI_FOG_FUNC_SCALE] = 0x00000000;
 	rstate->states[R600_DSA__SPI_FOG_FUNC_BIAS] = 0x00000000;
 	rstate->states[R600_DSA__SPI_FOG_CNTL] = 0x00000000;
 	rstate->states[R600_DSA__DB_DEPTH_CONTROL] = db_depth_control;
-	rstate->states[R600_DSA__DB_SHADER_CONTROL] = 0x00000210;
+	rstate->states[R600_DSA__DB_SHADER_CONTROL] = db_shader_control;
 	rstate->states[R600_DSA__DB_RENDER_CONTROL] = 0x00000060;
 	rstate->states[R600_DSA__DB_RENDER_OVERRIDE] = 0x0000002A;
 	rstate->states[R600_DSA__DB_SRESULTS_COMPARE_STATE1] = 0x00000000;
@@ -400,105 +1015,390 @@ static void *r600_create_dsa_state(struct pipe_context *ctx,
 	return rstate;
 }
 
-static void r600_bind_dsa_state(struct pipe_context *ctx, void *state)
+static inline unsigned r600_tex_wrap(unsigned wrap)
 {
-	struct r600_context *rctx = r600_context(ctx);
-	radeon_draw_set(rctx->draw, state);
+	switch (wrap) {
+	default:
+	case PIPE_TEX_WRAP_REPEAT:
+		return V_03C000_SQ_TEX_WRAP;
+	case PIPE_TEX_WRAP_CLAMP:
+		return V_03C000_SQ_TEX_CLAMP_LAST_TEXEL;
+	case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+		return V_03C000_SQ_TEX_CLAMP_HALF_BORDER;
+	case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+		return V_03C000_SQ_TEX_CLAMP_BORDER;
+	case PIPE_TEX_WRAP_MIRROR_REPEAT:
+		return V_03C000_SQ_TEX_MIRROR;
+	case PIPE_TEX_WRAP_MIRROR_CLAMP:
+		return V_03C000_SQ_TEX_MIRROR_ONCE_LAST_TEXEL;
+	case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE:
+		return V_03C000_SQ_TEX_MIRROR_ONCE_HALF_BORDER;
+	case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER:
+		return V_03C000_SQ_TEX_MIRROR_ONCE_BORDER;
+	}
 }
 
-static void r600_set_constant_buffer(struct pipe_context *ctx,
-				     uint shader, uint index,
-				     struct pipe_resource *buffer)
+static inline unsigned r600_tex_filter(unsigned filter)
 {
-	struct r600_screen *rscreen = r600_screen(ctx->screen);
-	struct r600_context *rctx = r600_context(ctx);
-	unsigned nconstant = 0, i, type, id;
+	switch (filter) {
+	default:
+	case PIPE_TEX_FILTER_NEAREST:
+		return V_03C000_SQ_TEX_XY_FILTER_POINT;
+	case PIPE_TEX_FILTER_LINEAR:
+		return V_03C000_SQ_TEX_XY_FILTER_BILINEAR;
+	}
+}
+
+static inline unsigned r600_tex_mipfilter(unsigned filter)
+{
+	switch (filter) {
+	case PIPE_TEX_MIPFILTER_NEAREST:
+		return V_03C000_SQ_TEX_Z_FILTER_POINT;
+	case PIPE_TEX_MIPFILTER_LINEAR:
+		return V_03C000_SQ_TEX_Z_FILTER_LINEAR;
+	default:
+	case PIPE_TEX_MIPFILTER_NONE:
+		return V_03C000_SQ_TEX_Z_FILTER_NONE;
+	}
+}
+
+static inline unsigned r600_tex_compare(unsigned compare)
+{
+	switch (compare) {
+	default:
+	case PIPE_FUNC_NEVER:
+		return V_03C000_SQ_TEX_DEPTH_COMPARE_NEVER;
+	case PIPE_FUNC_LESS:
+		return V_03C000_SQ_TEX_DEPTH_COMPARE_LESS;
+	case PIPE_FUNC_EQUAL:
+		return V_03C000_SQ_TEX_DEPTH_COMPARE_EQUAL;
+	case PIPE_FUNC_LEQUAL:
+		return V_03C000_SQ_TEX_DEPTH_COMPARE_LESSEQUAL;
+	case PIPE_FUNC_GREATER:
+		return V_03C000_SQ_TEX_DEPTH_COMPARE_GREATER;
+	case PIPE_FUNC_NOTEQUAL:
+		return V_03C000_SQ_TEX_DEPTH_COMPARE_NOTEQUAL;
+	case PIPE_FUNC_GEQUAL:
+		return V_03C000_SQ_TEX_DEPTH_COMPARE_GREATEREQUAL;
+	case PIPE_FUNC_ALWAYS:
+		return V_03C000_SQ_TEX_DEPTH_COMPARE_ALWAYS;
+	}
+}
+
+static INLINE u32 S_FIXED(float value, u32 frac_bits)
+{
+	return value * (1 << frac_bits);
+}
+
+static struct radeon_state *r600_sampler(struct r600_context *rctx,
+				const struct pipe_sampler_state *state,
+				unsigned id)
+{
+	struct r600_screen *rscreen = rctx->screen;
 	struct radeon_state *rstate;
-	struct pipe_transfer *transfer;
-	u32 *ptr;
 
-	switch (shader) {
-	case PIPE_SHADER_VERTEX:
-		id = R600_VS_CONSTANT;
-		type = R600_VS_CONSTANT_TYPE;
-		break;
-	case PIPE_SHADER_FRAGMENT:
-		id = R600_PS_CONSTANT;
-		type = R600_PS_CONSTANT_TYPE;
-		break;
+	rstate = radeon_state(rscreen->rw, R600_PS_SAMPLER_TYPE, id);
+	if (rstate == NULL)
+		return NULL;
+	rstate->states[R600_PS_SAMPLER__SQ_TEX_SAMPLER_WORD0_0] =
+			S_03C000_CLAMP_X(r600_tex_wrap(state->wrap_s)) |
+			S_03C000_CLAMP_Y(r600_tex_wrap(state->wrap_t)) |
+			S_03C000_CLAMP_Z(r600_tex_wrap(state->wrap_r)) |
+			S_03C000_XY_MAG_FILTER(r600_tex_filter(state->mag_img_filter)) |
+			S_03C000_XY_MIN_FILTER(r600_tex_filter(state->min_img_filter)) |
+			S_03C000_MIP_FILTER(r600_tex_mipfilter(state->min_mip_filter)) |
+			S_03C000_DEPTH_COMPARE_FUNCTION(r600_tex_compare(state->compare_func));
+	/* FIXME LOD it depends on texture base level ... */
+	rstate->states[R600_PS_SAMPLER__SQ_TEX_SAMPLER_WORD1_0] =
+			S_03C004_MIN_LOD(S_FIXED(CLAMP(state->min_lod, 0, 15), 6)) |
+			S_03C004_MAX_LOD(S_FIXED(CLAMP(state->max_lod, 0, 15), 6)) |
+			S_03C004_LOD_BIAS(S_FIXED(CLAMP(state->lod_bias, -16, 16), 6));
+	rstate->states[R600_PS_SAMPLER__SQ_TEX_SAMPLER_WORD2_0] = S_03C008_TYPE(1);
+	if (radeon_state_pm4(rstate)) {
+		radeon_state_decref(rstate);
+		return NULL;
+	}
+	return rstate;
+}
+
+static inline unsigned r600_tex_swizzle(unsigned swizzle)
+{
+	switch (swizzle) {
+	case PIPE_SWIZZLE_RED:
+		return V_038010_SQ_SEL_X;
+	case PIPE_SWIZZLE_GREEN:
+		return V_038010_SQ_SEL_Y;
+	case PIPE_SWIZZLE_BLUE:
+		return V_038010_SQ_SEL_Z;
+	case PIPE_SWIZZLE_ALPHA:
+		return V_038010_SQ_SEL_W;
+	case PIPE_SWIZZLE_ZERO:
+		return V_038010_SQ_SEL_0;
 	default:
-		fprintf(stderr, "%s:%d unsupported %d\n", __func__, __LINE__, shader);
-		return;
+	case PIPE_SWIZZLE_ONE:
+		return V_038010_SQ_SEL_1;
 	}
-	if (buffer && buffer->width0 > 0) {
-		nconstant = buffer->width0 / 16;
-		ptr = pipe_buffer_map(ctx, buffer, PIPE_TRANSFER_READ, &transfer);
-		if (ptr == NULL)
-			return;
-		for (i = 0; i < nconstant; i++) {
-			rstate = radeon_state(rscreen->rw, type, id + i);
-			if (rstate == NULL)
-				return;
-			rstate->states[R600_PS_CONSTANT__SQ_ALU_CONSTANT0_0] = ptr[i * 4 + 0];
-			rstate->states[R600_PS_CONSTANT__SQ_ALU_CONSTANT1_0] = ptr[i * 4 + 1];
-			rstate->states[R600_PS_CONSTANT__SQ_ALU_CONSTANT2_0] = ptr[i * 4 + 2];
-			rstate->states[R600_PS_CONSTANT__SQ_ALU_CONSTANT3_0] = ptr[i * 4 + 3];
-			if (radeon_state_pm4(rstate))
-				return;
-			if (radeon_draw_set_new(rctx->draw, rstate))
-				return;
-		}
-		pipe_buffer_unmap(ctx, buffer, transfer);
+}
+
+static inline unsigned r600_format_type(unsigned format_type)
+{
+	switch (format_type) {
+	default:
+	case UTIL_FORMAT_TYPE_UNSIGNED:
+		return V_038010_SQ_FORMAT_COMP_UNSIGNED;
+	case UTIL_FORMAT_TYPE_SIGNED:
+		return V_038010_SQ_FORMAT_COMP_SIGNED;
+	case UTIL_FORMAT_TYPE_FIXED:
+		return V_038010_SQ_FORMAT_COMP_UNSIGNED_BIASED;
 	}
 }
 
-static void r600_set_stencil_ref(struct pipe_context *ctx,
-				const struct pipe_stencil_ref *sr)
+static inline unsigned r600_tex_dim(unsigned dim)
 {
-	struct r600_context *rctx = r600_context(ctx);
-	rctx->stencil_ref = *sr;
+	switch (dim) {
+	default:
+	case PIPE_TEXTURE_1D:
+		return V_038000_SQ_TEX_DIM_1D;
+	case PIPE_TEXTURE_2D:
+		return V_038000_SQ_TEX_DIM_2D;
+	case PIPE_TEXTURE_3D:
+		return V_038000_SQ_TEX_DIM_3D;
+	case PIPE_TEXTURE_CUBE:
+		return V_038000_SQ_TEX_DIM_CUBEMAP;
+	}
 }
 
-static void r600_set_sample_mask(struct pipe_context *pipe, unsigned sample_mask)
+static struct radeon_state *r600_resource(struct r600_context *rctx,
+					const struct pipe_sampler_view *view,
+					unsigned id)
 {
+	struct r600_screen *rscreen = rctx->screen;
+	const struct util_format_description *desc;
+	struct r600_resource_texture *tmp;
+	struct r600_resource *rbuffer;
+	struct radeon_state *rstate;
+	unsigned format;
+	uint32_t word4 = 0, yuv_format = 0;
+	unsigned char swizzle[4];
+
+	swizzle[0] = view->swizzle_r;
+	swizzle[1] = view->swizzle_g;
+	swizzle[2] = view->swizzle_b;
+	swizzle[3] = view->swizzle_a;
+	format = r600_translate_texformat(view->texture->format,
+					  swizzle,
+					  &word4, &yuv_format);
+	if (format == ~0)
+		return NULL;
+	desc = util_format_description(view->texture->format);
+	if (desc == NULL) {
+		R600_ERR("unknow format %d\n", view->texture->format);
+		return NULL;
+	}
+	rstate = radeon_state(rscreen->rw, R600_PS_RESOURCE_TYPE, id);
+	if (rstate == NULL) {
+		return NULL;
+	}
+	tmp = (struct r600_resource_texture*)view->texture;
+	rbuffer = &tmp->resource;
+	rstate->bo[0] = radeon_bo_incref(rscreen->rw, rbuffer->bo);
+	rstate->bo[1] = radeon_bo_incref(rscreen->rw, rbuffer->bo);
+	rstate->nbo = 2;
+	rstate->placement[0] = RADEON_GEM_DOMAIN_GTT;
+	rstate->placement[1] = RADEON_GEM_DOMAIN_GTT;
+	rstate->placement[2] = RADEON_GEM_DOMAIN_GTT;
+	rstate->placement[3] = RADEON_GEM_DOMAIN_GTT;
+
+	/* FIXME properly handle first level != 0 */
+	rstate->states[R600_PS_RESOURCE__RESOURCE0_WORD0] =
+			S_038000_DIM(r600_tex_dim(view->texture->target)) |
+			S_038000_PITCH(((tmp->pitch[0] / tmp->bpt) / 8) - 1) |
+			S_038000_TEX_WIDTH(view->texture->width0 - 1);
+	rstate->states[R600_PS_RESOURCE__RESOURCE0_WORD1] =
+			S_038004_TEX_HEIGHT(view->texture->height0 - 1) |
+			S_038004_TEX_DEPTH(view->texture->depth0 - 1) |
+			S_038004_DATA_FORMAT(format);
+	rstate->states[R600_PS_RESOURCE__RESOURCE0_WORD2] = 0;
+	rstate->states[R600_PS_RESOURCE__RESOURCE0_WORD3] = tmp->offset[1] >> 8;
+	rstate->states[R600_PS_RESOURCE__RESOURCE0_WORD4] =
+		        word4 | 
+			S_038010_NUM_FORMAT_ALL(V_038010_SQ_NUM_FORMAT_NORM) |
+			S_038010_SRF_MODE_ALL(V_038010_SFR_MODE_NO_ZERO) |
+			S_038010_REQUEST_SIZE(1) |
+			S_038010_BASE_LEVEL(view->first_level);
+	rstate->states[R600_PS_RESOURCE__RESOURCE0_WORD5] =
+			S_038014_LAST_LEVEL(view->last_level) |
+			S_038014_BASE_ARRAY(0) |
+			S_038014_LAST_ARRAY(0);
+	rstate->states[R600_PS_RESOURCE__RESOURCE0_WORD6] =
+			S_038018_TYPE(V_038010_SQ_TEX_VTX_VALID_TEXTURE);
+	if (radeon_state_pm4(rstate)) {
+		radeon_state_decref(rstate);
+		return NULL;
+	}
+	return rstate;
 }
 
-void r600_init_state_functions(struct r600_context *rctx)
+static struct radeon_state *r600_cb_cntl(struct r600_context *rctx)
 {
-	rctx->context.set_sample_mask = r600_set_sample_mask;
-	rctx->context.create_blend_state = r600_create_blend_state;
-	rctx->context.bind_blend_state = r600_bind_blend_state;
-	rctx->context.delete_blend_state = r600_delete_state;
-	rctx->context.set_blend_color = r600_set_blend_color;
-	rctx->context.set_clip_state = r600_set_clip_state;
-	rctx->context.set_constant_buffer = r600_set_constant_buffer;
-	rctx->context.create_depth_stencil_alpha_state = r600_create_dsa_state;
-	rctx->context.bind_depth_stencil_alpha_state = r600_bind_dsa_state;
-	rctx->context.delete_depth_stencil_alpha_state = r600_delete_state;
-	rctx->context.set_framebuffer_state = r600_set_framebuffer_state;
-	rctx->context.create_fs_state = r600_create_fs_state;
-	rctx->context.bind_fs_state = r600_bind_fs_state;
-	rctx->context.delete_fs_state = r600_delete_state;
-	rctx->context.set_polygon_stipple = r600_set_polygon_stipple;
-	rctx->context.create_rasterizer_state = r600_create_rs_state;
-	rctx->context.bind_rasterizer_state = r600_bind_rs_state;
-	rctx->context.delete_rasterizer_state = r600_delete_state;
-	rctx->context.create_sampler_state = r600_create_sampler_state;
-	rctx->context.bind_fragment_sampler_states = r600_bind_sampler_states;
-	rctx->context.bind_vertex_sampler_states = r600_bind_sampler_states;
-	rctx->context.delete_sampler_state = r600_delete_state;
-	rctx->context.create_sampler_view = r600_create_sampler_view;
-	rctx->context.sampler_view_destroy = r600_sampler_view_destroy;
-	rctx->context.set_fragment_sampler_views = r600_set_fragment_sampler_views;
-	rctx->context.set_vertex_sampler_views = r600_set_vertex_sampler_views;
-	rctx->context.set_scissor_state = r600_set_scissor_state;
-	rctx->context.set_viewport_state = r600_set_viewport_state;
-	rctx->context.set_vertex_buffers = r600_set_vertex_buffers;
-	rctx->context.create_vertex_elements_state = r600_create_vertex_elements_state;
-	rctx->context.bind_vertex_elements_state = r600_bind_vertex_elements_state;
-	rctx->context.delete_vertex_elements_state = r600_delete_vertex_elements_state;
-	rctx->context.create_vs_state = r600_create_vs_state;
-	rctx->context.bind_vs_state = r600_bind_vs_state;
-	rctx->context.delete_vs_state = r600_delete_state;
-	rctx->context.set_stencil_ref = r600_set_stencil_ref;
+	struct r600_screen *rscreen = rctx->screen;
+	struct radeon_state *rstate;
+	const struct pipe_blend_state *pbs = &rctx->blend->state.blend;
+	int nr_cbufs = rctx->framebuffer->state.framebuffer.nr_cbufs;
+	uint32_t color_control, target_mask, shader_mask;
+	int i;
+
+	target_mask = 0;
+	shader_mask = 0;
+	color_control = S_028808_PER_MRT_BLEND(1);
+
+	for (i = 0; i < nr_cbufs; i++) {
+		shader_mask |= 0xf << (i * 4);
+	}
+
+	if (pbs->logicop_enable) {
+		color_control |= (pbs->logicop_func) << 16;
+	} else {
+		color_control |= (0xcc << 16);
+	}
+
+	if (pbs->independent_blend_enable) {
+		for (i = 0; i < nr_cbufs; i++) {
+			if (pbs->rt[i].blend_enable) {
+				color_control |= S_028808_TARGET_BLEND_ENABLE(1 << i);
+			}
+			target_mask |= (pbs->rt[i].colormask << (4 * i));
+		}
+	} else {
+		for (i = 0; i < nr_cbufs; i++) {
+			if (pbs->rt[0].blend_enable) {
+				color_control |= S_028808_TARGET_BLEND_ENABLE(1 << i);
+			}
+			target_mask |= (pbs->rt[0].colormask << (4 * i));
+		}
+	}
+	rstate = radeon_state(rscreen->rw, R600_CB_CNTL_TYPE, R600_CB_CNTL);
+	rstate->states[R600_CB_CNTL__CB_SHADER_MASK] = shader_mask;
+	rstate->states[R600_CB_CNTL__CB_TARGET_MASK] = target_mask;
+	rstate->states[R600_CB_CNTL__CB_COLOR_CONTROL] = color_control;
+	rstate->states[R600_CB_CNTL__PA_SC_AA_CONFIG] = 0x00000000;
+	rstate->states[R600_CB_CNTL__PA_SC_AA_SAMPLE_LOCS_MCTX] = 0x00000000;
+	rstate->states[R600_CB_CNTL__PA_SC_AA_SAMPLE_LOCS_8S_WD1_MCTX] = 0x00000000;
+	rstate->states[R600_CB_CNTL__CB_CLRCMP_CONTROL] = 0x01000000;
+	rstate->states[R600_CB_CNTL__CB_CLRCMP_SRC] = 0x00000000;
+	rstate->states[R600_CB_CNTL__CB_CLRCMP_DST] = 0x000000FF;
+	rstate->states[R600_CB_CNTL__CB_CLRCMP_MSK] = 0xFFFFFFFF;
+	rstate->states[R600_CB_CNTL__PA_SC_AA_MASK] = 0xFFFFFFFF;
+	if (radeon_state_pm4(rstate)) {
+		radeon_state_decref(rstate);
+		return NULL;
+	}
+	return rstate;
+}
+
+int r600_context_hw_states(struct r600_context *rctx)
+{
+	unsigned i;
+	int r;
+	int nr_cbufs = rctx->framebuffer->state.framebuffer.nr_cbufs;
+
+	/* free previous TODO determine what need to be updated, what
+	 * doesn't
+	 */
+	//radeon_state_decref(rctx->hw_states.config);
+	rctx->hw_states.cb_cntl = radeon_state_decref(rctx->hw_states.cb_cntl);
+	rctx->hw_states.db = radeon_state_decref(rctx->hw_states.db);
+	rctx->hw_states.rasterizer = radeon_state_decref(rctx->hw_states.rasterizer);
+	rctx->hw_states.scissor = radeon_state_decref(rctx->hw_states.scissor);
+	rctx->hw_states.dsa = radeon_state_decref(rctx->hw_states.dsa);
+	rctx->hw_states.blend = radeon_state_decref(rctx->hw_states.blend);
+	rctx->hw_states.viewport = radeon_state_decref(rctx->hw_states.viewport);
+	for (i = 0; i < 8; i++) {
+		rctx->hw_states.cb[i] = radeon_state_decref(rctx->hw_states.cb[i]);
+	}
+	for (i = 0; i < rctx->hw_states.ps_nresource; i++) {
+		radeon_state_decref(rctx->hw_states.ps_resource[i]);
+		rctx->hw_states.ps_resource[i] = NULL;
+	}
+	rctx->hw_states.ps_nresource = 0;
+	for (i = 0; i < rctx->hw_states.ps_nsampler; i++) {
+		radeon_state_decref(rctx->hw_states.ps_sampler[i]);
+		rctx->hw_states.ps_sampler[i] = NULL;
+	}
+	rctx->hw_states.ps_nsampler = 0;
+
+	/* build new states */
+	rctx->hw_states.rasterizer = r600_rasterizer(rctx);
+	rctx->hw_states.scissor = r600_scissor(rctx);
+	rctx->hw_states.dsa = r600_dsa(rctx);
+	rctx->hw_states.blend = r600_blend(rctx);
+	rctx->hw_states.viewport = r600_viewport(rctx);
+	for (i = 0; i < nr_cbufs; i++) {
+		rctx->hw_states.cb[i] = r600_cb(rctx, i);
+	}
+	rctx->hw_states.db = r600_db(rctx);
+	rctx->hw_states.cb_cntl = r600_cb_cntl(rctx);
+
+	for (i = 0; i < rctx->ps_nsampler; i++) {
+		if (rctx->ps_sampler[i]) {
+			rctx->hw_states.ps_sampler[i] = r600_sampler(rctx,
+							&rctx->ps_sampler[i]->state.sampler,
+							R600_PS_SAMPLER + i);
+		}
+	}
+	rctx->hw_states.ps_nsampler = rctx->ps_nsampler;
+	for (i = 0; i < rctx->ps_nsampler_view; i++) {
+		if (rctx->ps_sampler_view[i]) {
+			rctx->hw_states.ps_resource[i] = r600_resource(rctx,
+							&rctx->ps_sampler_view[i]->state.sampler_view,
+							R600_PS_RESOURCE + i);
+		}
+	}
+	rctx->hw_states.ps_nresource = rctx->ps_nsampler_view;
+
+	/* bind states */
+	r = radeon_draw_set(rctx->draw, rctx->hw_states.db);
+	if (r)
+		return r;
+	r = radeon_draw_set(rctx->draw, rctx->hw_states.rasterizer);
+	if (r)
+		return r;
+	r = radeon_draw_set(rctx->draw, rctx->hw_states.scissor);
+	if (r)
+		return r;
+	r = radeon_draw_set(rctx->draw, rctx->hw_states.dsa);
+	if (r)
+		return r;
+	r = radeon_draw_set(rctx->draw, rctx->hw_states.blend);
+	if (r)
+		return r;
+	r = radeon_draw_set(rctx->draw, rctx->hw_states.viewport);
+	if (r)
+		return r;
+	for (i = 0; i < nr_cbufs; i++) {
+		r = radeon_draw_set(rctx->draw, rctx->hw_states.cb[i]);
+		if (r)
+			return r;
+	}
+	r = radeon_draw_set(rctx->draw, rctx->hw_states.config);
+	if (r)
+		return r;
+	r = radeon_draw_set(rctx->draw, rctx->hw_states.cb_cntl);
+	if (r)
+		return r;
+	for (i = 0; i < rctx->hw_states.ps_nresource; i++) {
+		if (rctx->hw_states.ps_resource[i]) {
+			r = radeon_draw_set(rctx->draw, rctx->hw_states.ps_resource[i]);
+			if (r)
+				return r;
+		}
+	}
+	for (i = 0; i < rctx->hw_states.ps_nsampler; i++) {
+		if (rctx->hw_states.ps_sampler[i]) {
+			r = radeon_draw_set(rctx->draw, rctx->hw_states.ps_sampler[i]);
+			if (r)
+				return r;
+		}
+	}
+	return 0;
 }
diff --git a/src/gallium/drivers/r600/r600_state_inlines.h b/src/gallium/drivers/r600/r600_state_inlines.h
new file mode 100644
index 00000000000..f93c20da35e
--- /dev/null
+++ b/src/gallium/drivers/r600/r600_state_inlines.h
@@ -0,0 +1,311 @@
+/*
+ * Copyright 2010 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+#ifndef R600_STATE_INLINES_H
+#define R600_STATE_INLINES_H
+
+#include "util/u_format.h"
+#include "r600d.h"
+
+static INLINE uint32_t r600_translate_blend_function(int blend_func)
+{
+	switch (blend_func) {
+	case PIPE_BLEND_ADD:
+		return V_028804_COMB_DST_PLUS_SRC;
+	case PIPE_BLEND_SUBTRACT:
+		return V_028804_COMB_SRC_MINUS_DST;
+	case PIPE_BLEND_REVERSE_SUBTRACT:
+		return V_028804_COMB_DST_MINUS_SRC;
+	case PIPE_BLEND_MIN:
+		return V_028804_COMB_MIN_DST_SRC;
+	case PIPE_BLEND_MAX:
+		return V_028804_COMB_MAX_DST_SRC;
+	default:
+		R600_ERR("Unknown blend function %d\n", blend_func);
+		assert(0);
+		break;
+	}
+	return 0;
+}
+
+static INLINE uint32_t r600_translate_blend_factor(int blend_fact)
+{
+	switch (blend_fact) {
+	case PIPE_BLENDFACTOR_ONE:
+		return V_028804_BLEND_ONE;
+	case PIPE_BLENDFACTOR_SRC_COLOR:
+		return V_028804_BLEND_SRC_COLOR;
+	case PIPE_BLENDFACTOR_SRC_ALPHA:
+		return V_028804_BLEND_SRC_ALPHA;
+	case PIPE_BLENDFACTOR_DST_ALPHA:
+		return V_028804_BLEND_DST_ALPHA;
+	case PIPE_BLENDFACTOR_DST_COLOR:
+		return V_028804_BLEND_DST_COLOR;
+	case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE:
+		return V_028804_BLEND_SRC_ALPHA_SATURATE;
+	case PIPE_BLENDFACTOR_CONST_COLOR:
+		return V_028804_BLEND_CONST_COLOR;
+	case PIPE_BLENDFACTOR_CONST_ALPHA:
+		return V_028804_BLEND_CONST_ALPHA;
+	case PIPE_BLENDFACTOR_ZERO:
+		return V_028804_BLEND_ZERO;
+	case PIPE_BLENDFACTOR_INV_SRC_COLOR:
+		return V_028804_BLEND_ONE_MINUS_SRC_COLOR;
+	case PIPE_BLENDFACTOR_INV_SRC_ALPHA:
+		return V_028804_BLEND_ONE_MINUS_SRC_ALPHA;
+	case PIPE_BLENDFACTOR_INV_DST_ALPHA:
+		return V_028804_BLEND_ONE_MINUS_DST_ALPHA;
+	case PIPE_BLENDFACTOR_INV_DST_COLOR:
+		return V_028804_BLEND_ONE_MINUS_DST_COLOR;
+	case PIPE_BLENDFACTOR_INV_CONST_COLOR:
+		return V_028804_BLEND_ONE_MINUS_CONST_COLOR;
+	case PIPE_BLENDFACTOR_INV_CONST_ALPHA:
+		return V_028804_BLEND_ONE_MINUS_CONST_ALPHA;
+	case PIPE_BLENDFACTOR_SRC1_COLOR:
+		return V_028804_BLEND_SRC1_COLOR;
+	case PIPE_BLENDFACTOR_SRC1_ALPHA:
+		return V_028804_BLEND_SRC1_ALPHA;
+	case PIPE_BLENDFACTOR_INV_SRC1_COLOR:
+		return V_028804_BLEND_INV_SRC1_COLOR;
+	case PIPE_BLENDFACTOR_INV_SRC1_ALPHA:
+		return V_028804_BLEND_INV_SRC1_ALPHA;
+	default:
+		R600_ERR("Bad blend factor %d not supported!\n", blend_fact);
+		assert(0);
+		break;
+	}
+	return 0;
+}
+
+static INLINE uint32_t r600_translate_stencil_op(int s_op)
+{
+	switch (s_op) {
+	case PIPE_STENCIL_OP_KEEP:
+		return V_028800_STENCIL_KEEP;
+	case PIPE_STENCIL_OP_ZERO:
+		return V_028800_STENCIL_ZERO;
+	case PIPE_STENCIL_OP_REPLACE:
+		return V_028800_STENCIL_REPLACE;
+	case PIPE_STENCIL_OP_INCR:
+		return V_028800_STENCIL_INCR;
+	case PIPE_STENCIL_OP_DECR:
+		return V_028800_STENCIL_DECR;
+	case PIPE_STENCIL_OP_INCR_WRAP:
+		return V_028800_STENCIL_INCR_WRAP;
+	case PIPE_STENCIL_OP_DECR_WRAP:
+		return V_028800_STENCIL_DECR_WRAP;
+	case PIPE_STENCIL_OP_INVERT:
+		return V_028800_STENCIL_INVERT;
+	default:
+		R600_ERR("Unknown stencil op %d", s_op);
+		assert(0);
+		break;
+	}
+	return 0;
+}
+
+/* translates straight */
+static INLINE uint32_t r600_translate_ds_func(int func)
+{
+	return func;
+}
+
+static uint32_t r600_translate_dbformat(enum pipe_format format)
+{
+	switch (format) {
+	case PIPE_FORMAT_Z16_UNORM:
+		return V_028010_DEPTH_16;
+	case PIPE_FORMAT_Z24X8_UNORM:
+		return V_028010_DEPTH_X8_24;
+	case PIPE_FORMAT_Z24_UNORM_S8_USCALED:
+		return V_028010_DEPTH_8_24;
+	default:
+		return ~0;
+	}
+}
+
+static uint32_t r600_translate_colorswap(enum pipe_format format)
+{
+	switch (format) {
+		/* 8-bit buffers. */
+	case PIPE_FORMAT_A8_UNORM:
+	case PIPE_FORMAT_I8_UNORM:
+	case PIPE_FORMAT_L8_UNORM:
+	case PIPE_FORMAT_R8_UNORM:
+	case PIPE_FORMAT_R8_SNORM:
+		return V_0280A0_SWAP_STD;
+
+		/* 16-bit buffers. */
+	case PIPE_FORMAT_B5G6R5_UNORM:
+		return V_0280A0_SWAP_STD_REV;
+
+	case PIPE_FORMAT_B5G5R5A1_UNORM:
+	case PIPE_FORMAT_B5G5R5X1_UNORM:
+		return V_0280A0_SWAP_ALT;
+
+	case PIPE_FORMAT_B4G4R4A4_UNORM:
+	case PIPE_FORMAT_B4G4R4X4_UNORM:
+		return V_0280A0_SWAP_ALT;
+		/* 32-bit buffers. */
+
+	case PIPE_FORMAT_A8B8G8R8_SRGB:
+		return V_0280A0_SWAP_STD_REV;
+	case PIPE_FORMAT_B8G8R8A8_SRGB:
+		return V_0280A0_SWAP_ALT;
+
+	case PIPE_FORMAT_B8G8R8A8_UNORM:
+	case PIPE_FORMAT_B8G8R8X8_UNORM:
+		return V_0280A0_SWAP_ALT;
+
+	case PIPE_FORMAT_A8R8G8B8_UNORM:
+	case PIPE_FORMAT_X8R8G8B8_UNORM:
+		return V_0280A0_SWAP_ALT_REV;
+	case PIPE_FORMAT_R8G8B8A8_SNORM:
+	case PIPE_FORMAT_R8G8B8X8_UNORM:
+		return V_0280A0_SWAP_STD;
+
+	case PIPE_FORMAT_A8B8G8R8_UNORM:
+	case PIPE_FORMAT_X8B8G8R8_UNORM:
+		//        case PIPE_FORMAT_R8SG8SB8UX8U_NORM:
+		return V_0280A0_SWAP_STD_REV;
+
+	case PIPE_FORMAT_Z24X8_UNORM:
+	case PIPE_FORMAT_Z24_UNORM_S8_USCALED:
+		return V_0280A0_SWAP_STD;
+
+	case PIPE_FORMAT_R10G10B10A2_UNORM:
+	case PIPE_FORMAT_R10G10B10X2_SNORM:
+	case PIPE_FORMAT_B10G10R10A2_UNORM:
+	case PIPE_FORMAT_R10SG10SB10SA2U_NORM:
+		return V_0280A0_SWAP_STD_REV;
+
+		/* 64-bit buffers. */
+	case PIPE_FORMAT_R16G16B16A16_UNORM:
+	case PIPE_FORMAT_R16G16B16A16_SNORM:
+		//		return V_0280A0_COLOR_16_16_16_16;
+	case PIPE_FORMAT_R16G16B16A16_FLOAT:
+		//		return V_0280A0_COLOR_16_16_16_16_FLOAT;
+
+		/* 128-bit buffers. */
+	case PIPE_FORMAT_R32G32B32A32_FLOAT:
+		//		return V_0280A0_COLOR_32_32_32_32_FLOAT;
+		return 0;
+	default:
+		R600_ERR("unsupported colorswap format %d\n", format);
+		return ~0;
+	}
+	return ~0;
+}
+
+static INLINE uint32_t r600_translate_colorformat(enum pipe_format format)
+{
+	switch (format) {
+		/* 8-bit buffers. */
+	case PIPE_FORMAT_A8_UNORM:
+	case PIPE_FORMAT_I8_UNORM:
+	case PIPE_FORMAT_L8_UNORM:
+	case PIPE_FORMAT_R8_UNORM:
+	case PIPE_FORMAT_R8_SNORM:
+		return V_0280A0_COLOR_8;
+
+		/* 16-bit buffers. */
+	case PIPE_FORMAT_B5G6R5_UNORM:
+		return V_0280A0_COLOR_5_6_5;
+
+	case PIPE_FORMAT_B5G5R5A1_UNORM:
+	case PIPE_FORMAT_B5G5R5X1_UNORM:
+		return V_0280A0_COLOR_1_5_5_5;
+
+	case PIPE_FORMAT_B4G4R4A4_UNORM:
+	case PIPE_FORMAT_B4G4R4X4_UNORM:
+		return V_0280A0_COLOR_4_4_4_4;
+
+		/* 32-bit buffers. */
+	case PIPE_FORMAT_A8B8G8R8_SRGB:
+	case PIPE_FORMAT_A8B8G8R8_UNORM:
+	case PIPE_FORMAT_A8R8G8B8_UNORM:
+	case PIPE_FORMAT_B8G8R8A8_SRGB:
+	case PIPE_FORMAT_B8G8R8A8_UNORM:
+	case PIPE_FORMAT_B8G8R8X8_UNORM:
+	case PIPE_FORMAT_R8G8B8A8_SNORM:
+	case PIPE_FORMAT_R8G8B8A8_UNORM:
+	case PIPE_FORMAT_R8G8B8X8_UNORM:
+	case PIPE_FORMAT_R8SG8SB8UX8U_NORM:
+	case PIPE_FORMAT_X8B8G8R8_UNORM:
+	case PIPE_FORMAT_X8R8G8B8_UNORM:
+		return V_0280A0_COLOR_8_8_8_8;
+
+	case PIPE_FORMAT_R10G10B10A2_UNORM:
+	case PIPE_FORMAT_R10G10B10X2_SNORM:
+	case PIPE_FORMAT_B10G10R10A2_UNORM:
+	case PIPE_FORMAT_R10SG10SB10SA2U_NORM:
+		return V_0280A0_COLOR_10_10_10_2;
+
+	case PIPE_FORMAT_Z24X8_UNORM:
+	case PIPE_FORMAT_Z24_UNORM_S8_USCALED:
+		return V_0280A0_COLOR_24_8;
+
+		/* 64-bit buffers. */
+	case PIPE_FORMAT_R16G16B16A16_UNORM:
+	case PIPE_FORMAT_R16G16B16A16_SNORM:
+		return V_0280A0_COLOR_16_16_16_16;
+	case PIPE_FORMAT_R16G16B16A16_FLOAT:
+		return V_0280A0_COLOR_16_16_16_16_FLOAT;
+	case PIPE_FORMAT_R32G32_FLOAT:
+		return V_0280A0_COLOR_32_32_FLOAT;
+
+		/* 128-bit buffers. */
+	case PIPE_FORMAT_R32G32B32_FLOAT:
+	case PIPE_FORMAT_R32G32B32A32_FLOAT:
+		return V_0280A0_COLOR_32_32_32_32_FLOAT;
+
+		/* YUV buffers. */
+	case PIPE_FORMAT_UYVY:
+	case PIPE_FORMAT_YUYV:
+	default:
+		R600_ERR("unsupported color format %d\n", format);
+		return ~0; /* Unsupported. */
+	}
+}
+
+static INLINE boolean r600_is_sampler_format_supported(enum pipe_format format)
+{
+	return r600_translate_texformat(format, NULL, NULL, NULL) != ~0;
+}
+
+static INLINE boolean r600_is_colorbuffer_format_supported(enum pipe_format format)
+{
+	return r600_translate_colorformat(format) != ~0 &&
+		r600_translate_colorswap(format) != ~0;
+}
+
+static INLINE boolean r600_is_zs_format_supported(enum pipe_format format)
+{
+	return r600_translate_dbformat(format) != ~0;
+}
+
+static INLINE boolean r600_is_vertex_format_supported(enum pipe_format format)
+{
+	return r600_translate_colorformat(format) != ~0;
+}
+
+#endif
diff --git a/src/gallium/drivers/r600/r600_texture.c b/src/gallium/drivers/r600/r600_texture.c
index 903cfad80a8..30d79ebdd6f 100644
--- a/src/gallium/drivers/r600/r600_texture.c
+++ b/src/gallium/drivers/r600/r600_texture.c
@@ -31,15 +31,19 @@
 #include <util/u_memory.h>
 #include "state_tracker/drm_driver.h"
 #include "r600_screen.h"
-#include "r600_texture.h"
+#include "r600_context.h"
+#include "r600_resource.h"
+#include "r600d.h"
 
 extern struct u_resource_vtbl r600_texture_vtbl;
 
-unsigned long r600_texture_get_offset(struct r600_texture *rtex, unsigned level, unsigned zslice, unsigned face)
+static unsigned long r600_texture_get_offset(struct r600_resource_texture *rtex,
+					unsigned level, unsigned zslice,
+					unsigned face)
 {
 	unsigned long offset = rtex->offset[level];
 
-	switch (rtex->b.b.target) {
+	switch (rtex->resource.base.b.target) {
 	case PIPE_TEXTURE_3D:
 		assert(face == 0);
 		return offset + zslice * rtex->layer_size[level];
@@ -52,67 +56,68 @@ unsigned long r600_texture_get_offset(struct r600_texture *rtex, unsigned level,
 	}
 }
 
-static void r600_setup_miptree(struct r600_screen *rscreen, struct r600_texture *rtex)
+static void r600_setup_miptree(struct r600_screen *rscreen, struct r600_resource_texture *rtex)
 {
-	struct pipe_resource *ptex = &rtex->b.b;
-	unsigned long w, h, stride, size, layer_size, i, offset;
+	struct pipe_resource *ptex = &rtex->resource.base.b;
+	unsigned long w, h, pitch, size, layer_size, i, offset;
 
+	rtex->bpt = util_format_get_blocksize(ptex->format);
 	for (i = 0, offset = 0; i <= ptex->last_level; i++) {
 		w = u_minify(ptex->width0, i);
 		h = u_minify(ptex->height0, i);
-		stride = align(util_format_get_stride(ptex->format, w), 32);
-		layer_size = stride * h;
+		pitch = util_format_get_stride(ptex->format, align(w, 64));
+		layer_size = pitch * h;
 		if (ptex->target == PIPE_TEXTURE_CUBE)
 			size = layer_size * 6;
 		else
 			size = layer_size * u_minify(ptex->depth0, i);
 		rtex->offset[i] = offset;
 		rtex->layer_size[i] = layer_size;
-		rtex->pitch[i] = stride / util_format_get_blocksize(ptex->format);
-		rtex->stride[i] = stride;
-		offset += align(size, 32);
+		rtex->pitch[i] = pitch;
+		offset += size;
 	}
 	rtex->size = offset;
 }
 
 struct pipe_resource *r600_texture_create(struct pipe_screen *screen,
-					  const struct pipe_resource *templ)
+						const struct pipe_resource *templ)
 {
-	struct r600_texture *rtex = CALLOC_STRUCT(r600_texture);
+	struct r600_resource_texture *rtex;
+	struct r600_resource *resource;
 	struct r600_screen *rscreen = r600_screen(screen);
-	struct pipe_resource templ_buf;
 
+	rtex = CALLOC_STRUCT(r600_resource_texture);
 	if (!rtex) {
 		return NULL;
 	}
-	rtex->b.b = *templ;
-	rtex->b.vtbl = &r600_texture_vtbl;
-	pipe_reference_init(&rtex->b.b.reference, 1);
-	rtex->b.b.screen = screen;
+	resource = &rtex->resource;
+	resource->base.b = *templ;
+	resource->base.vtbl = &r600_texture_vtbl;
+	pipe_reference_init(&resource->base.b.reference, 1);
+	resource->base.b.screen = screen;
 	r600_setup_miptree(rscreen, rtex);
 
-	memset(&templ_buf, 0, sizeof(struct pipe_resource));
-	templ_buf.target = PIPE_BUFFER;
-	templ_buf.format = PIPE_FORMAT_R8_UNORM;
-	templ_buf.usage = templ->usage;
-	templ_buf.bind = templ->bind;
-	templ_buf.width0 = rtex->size;
-	templ_buf.height0 = 1;
-	templ_buf.depth0 = 1;
-
-	rtex->buffer = screen->resource_create(screen, &templ_buf);
-	if (!rtex->buffer) {
+	/* FIXME alignment 4096 enought ? too much ? */
+	resource->domain = r600_domain_from_usage(resource->base.b.bind);
+	resource->bo = radeon_bo(rscreen->rw, 0, rtex->size, 4096, NULL);
+	if (resource->bo == NULL) {
 		FREE(rtex);
 		return NULL;
 	}
-	return &rtex->b.b;
+
+	return &resource->base.b;
 }
 
 static void r600_texture_destroy(struct pipe_screen *screen,
 				 struct pipe_resource *ptex)
 {
-	struct r600_texture *rtex = (struct r600_texture*)ptex;
+	struct r600_resource_texture *rtex = (struct r600_resource_texture*)ptex;
+	struct r600_resource *resource = &rtex->resource;
+	struct r600_screen *rscreen = r600_screen(screen);
 
+	if (resource->bo) {
+		radeon_bo_decref(rscreen->rw, resource->bo);
+	}
 	FREE(rtex);
 }
 
@@ -121,7 +126,7 @@ static struct pipe_surface *r600_get_tex_surface(struct pipe_screen *screen,
 						unsigned face, unsigned level,
 						unsigned zslice, unsigned flags)
 {
-	struct r600_texture *rtex = (struct r600_texture*)texture;
+	struct r600_resource_texture *rtex = (struct r600_resource_texture*)texture;
 	struct pipe_surface *surface = CALLOC_STRUCT(pipe_surface);
 	unsigned long offset;
 
@@ -149,71 +154,115 @@ static void r600_tex_surface_destroy(struct pipe_surface *surface)
 }
 
 struct pipe_resource *r600_texture_from_handle(struct pipe_screen *screen,
-					       const struct pipe_resource *base,
+					       const struct pipe_resource *templ,
 					       struct winsys_handle *whandle)
 {
-	struct pipe_resource *buffer;
-	struct r600_texture *rtex;
+	struct radeon *rw = (struct radeon*)screen->winsys;
+	struct r600_resource_texture *rtex;
+	struct r600_resource *resource;
+	struct radeon_bo *bo = NULL;
 
-	buffer = r600_buffer_from_handle(screen, whandle);
-	if (buffer == NULL) {
+	bo = radeon_bo(rw, whandle->handle, 0, 0, NULL);
+	if (bo == NULL) {
 		return NULL;
 	}
 
 	/* Support only 2D textures without mipmaps */
-	if (base->target != PIPE_TEXTURE_2D || base->depth0 != 1 || base->last_level != 0)
+	if (templ->target != PIPE_TEXTURE_2D || templ->depth0 != 1 || templ->last_level != 0)
 		return NULL;
 
-	rtex = CALLOC_STRUCT(r600_texture);
+	rtex = CALLOC_STRUCT(r600_resource_texture);
 	if (rtex == NULL)
 		return NULL;
 
-	/* one ref already taken */
-	rtex->buffer = buffer;
-
-	rtex->b.b = *base;
-	rtex->b.vtbl = &r600_texture_vtbl;
-	pipe_reference_init(&rtex->b.b.reference, 1);
-	rtex->b.b.screen = screen;
-	rtex->stride_override = whandle->stride;
-	rtex->pitch[0] = whandle->stride / util_format_get_blocksize(base->format);
-	rtex->stride[0] = whandle->stride;
+	resource = &rtex->resource;
+	resource->base.b = *templ;
+	resource->base.vtbl = &r600_texture_vtbl;
+	pipe_reference_init(&resource->base.b.reference, 1);
+	resource->base.b.screen = screen;
+	resource->bo = bo;
+	rtex->pitch_override = whandle->stride;
+	rtex->bpt = util_format_get_blocksize(templ->format);
+	rtex->pitch[0] = whandle->stride;
 	rtex->offset[0] = 0;
-	rtex->size = align(rtex->stride[0] * base->height0, 32);
+	rtex->size = align(rtex->pitch[0] * templ->height0, 64);
 
-	return &rtex->b.b;
+	return &resource->base.b;
 }
 
-static boolean r600_texture_get_handle(struct pipe_screen* screen,
-				       struct pipe_resource *texture,
-				       struct winsys_handle *whandle)
+static unsigned int r600_texture_is_referenced(struct pipe_context *context,
+						struct pipe_resource *texture,
+						unsigned face, unsigned level)
 {
-	struct r600_screen *rscreen = r600_screen(screen);
-	struct r600_texture* rtex = (struct r600_texture*)texture;
+	/* FIXME */
+	return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE;
+}
 
-	if (!rtex) {
-		return FALSE;
-	}
+struct pipe_transfer* r600_texture_get_transfer(struct pipe_context *ctx,
+						struct pipe_resource *texture,
+						struct pipe_subresource sr,
+						unsigned usage,
+						const struct pipe_box *box)
+{
+	struct r600_resource_texture *rtex = (struct r600_resource_texture*)texture;
+	struct r600_transfer *trans;
 
-	whandle->stride = rtex->stride[0];
+	trans = CALLOC_STRUCT(r600_transfer);
+	if (trans == NULL)
+		return NULL;
+	pipe_resource_reference(&trans->transfer.resource, texture);
+	trans->transfer.sr = sr;
+	trans->transfer.usage = usage;
+	trans->transfer.box = *box;
+	trans->transfer.stride = rtex->pitch[sr.level];
+	trans->offset = r600_texture_get_offset(rtex, sr.level, box->z, sr.face);
+	return &trans->transfer;
+}
 
-	r600_buffer_get_handle(rscreen->rw, rtex->buffer, whandle);
+void r600_texture_transfer_destroy(struct pipe_context *ctx,
+				   struct pipe_transfer *trans)
+{
+	pipe_resource_reference(&trans->resource, NULL);
+	FREE(trans);
+}
 
-	return TRUE;
+void* r600_texture_transfer_map(struct pipe_context *ctx,
+				struct pipe_transfer* transfer)
+{
+	struct r600_transfer *rtransfer = (struct r600_transfer*)transfer;
+	struct r600_resource *resource;
+	enum pipe_format format = transfer->resource->format;
+	struct r600_screen *rscreen = r600_screen(ctx->screen);
+	char *map;
+
+	r600_flush(ctx, 0, NULL);
+
+	resource = (struct r600_resource *)transfer->resource;
+	if (radeon_bo_map(rscreen->rw, resource->bo)) {
+		return NULL;
+	}
+	radeon_bo_wait(rscreen->rw, resource->bo);
+
+	map = resource->bo->data;
+
+	return map + rtransfer->offset +
+		transfer->box.y / util_format_get_blockheight(format) * transfer->stride +
+		transfer->box.x / util_format_get_blockwidth(format) * util_format_get_blocksize(format);
 }
 
-static unsigned int r600_texture_is_referenced(struct pipe_context *context,
-						struct pipe_resource *texture,
-						unsigned face, unsigned level)
+void r600_texture_transfer_unmap(struct pipe_context *ctx,
+				 struct pipe_transfer* transfer)
 {
-	struct r600_texture *rtex = (struct r600_texture*)texture;
+	struct r600_screen *rscreen = r600_screen(ctx->screen);
+	struct r600_resource *resource;
 
-	return r600_buffer_is_referenced_by_cs(context, rtex->buffer, face, level);
+	resource = (struct r600_resource *)transfer->resource;
+	radeon_bo_unmap(rscreen->rw, resource->bo);
 }
 
 struct u_resource_vtbl r600_texture_vtbl =
 {
-	r600_texture_get_handle,	/* get_handle */
+	u_default_resource_get_handle,	/* get_handle */
 	r600_texture_destroy,		/* resource_destroy */
 	r600_texture_is_referenced,	/* is_resource_referenced */
 	r600_texture_get_transfer,	/* get_transfer */
@@ -229,3 +278,250 @@ void r600_init_screen_texture_functions(struct pipe_screen *screen)
 	screen->get_tex_surface = r600_get_tex_surface;
 	screen->tex_surface_destroy = r600_tex_surface_destroy;
 }
+
+static unsigned r600_get_swizzle_combined(const unsigned char *swizzle_format,
+					  const unsigned char *swizzle_view)
+{
+    unsigned i;
+    unsigned char swizzle[4];
+    unsigned result = 0;
+    const uint32_t swizzle_shift[4] = {
+	    16, 19, 22, 25,
+    };
+    const uint32_t swizzle_bit[4] = {
+	    0, 1, 2, 3,
+    };
+
+    if (swizzle_view) {
+        /* Combine two sets of swizzles. */
+        for (i = 0; i < 4; i++) {
+            swizzle[i] = swizzle_view[i] <= UTIL_FORMAT_SWIZZLE_W ?
+                         swizzle_format[swizzle_view[i]] : swizzle_view[i];
+        }
+    } else {
+        memcpy(swizzle, swizzle_format, 4);
+    }
+
+    /* Get swizzle. */
+    for (i = 0; i < 4; i++) {
+        switch (swizzle[i]) {
+            case UTIL_FORMAT_SWIZZLE_Y:
+                result |= swizzle_bit[1] << swizzle_shift[i];
+                break;
+            case UTIL_FORMAT_SWIZZLE_Z:
+                result |= swizzle_bit[2] << swizzle_shift[i];
+                break;
+            case UTIL_FORMAT_SWIZZLE_W:
+                result |= swizzle_bit[3] << swizzle_shift[i];
+                break;
+            case UTIL_FORMAT_SWIZZLE_0:
+                result |= V_038010_SQ_SEL_0 << swizzle_shift[i];
+                break;
+            case UTIL_FORMAT_SWIZZLE_1:
+                result |= V_038010_SQ_SEL_1 << swizzle_shift[i];
+                break;
+            default: /* UTIL_FORMAT_SWIZZLE_X */
+                result |= swizzle_bit[0] << swizzle_shift[i];
+        }
+    }
+    return result;
+}
+
+/* texture format translate */
+uint32_t r600_translate_texformat(enum pipe_format format,
+				  const unsigned char *swizzle_view, 
+				  uint32_t *word4_p, uint32_t *yuv_format_p)
+{
+	uint32_t result = 0, word4 = 0, yuv_format = 0;
+	const struct util_format_description *desc;
+	boolean uniform = TRUE;
+	int i;
+	const uint32_t sign_bit[4] = {
+		S_038010_FORMAT_COMP_X(V_038010_SQ_FORMAT_COMP_SIGNED),
+		S_038010_FORMAT_COMP_Y(V_038010_SQ_FORMAT_COMP_SIGNED),
+		S_038010_FORMAT_COMP_Z(V_038010_SQ_FORMAT_COMP_SIGNED),
+		S_038010_FORMAT_COMP_W(V_038010_SQ_FORMAT_COMP_SIGNED)
+	};
+	desc = util_format_description(format);
+
+	/* Colorspace (return non-RGB formats directly). */
+	switch (desc->colorspace) {
+		/* Depth stencil formats */
+	case UTIL_FORMAT_COLORSPACE_ZS:
+		switch (format) {
+		case PIPE_FORMAT_Z16_UNORM:
+			result = V_028010_DEPTH_16;
+			goto out_word4;
+		case PIPE_FORMAT_Z24X8_UNORM:
+			result = V_028010_DEPTH_X8_24;
+			goto out_word4;
+		case PIPE_FORMAT_Z24_UNORM_S8_USCALED:
+			result = V_028010_DEPTH_8_24;
+			goto out_word4;
+		default:
+			goto out_unknown;
+		}
+
+	case UTIL_FORMAT_COLORSPACE_YUV:
+		yuv_format |= (1 << 30);
+		switch (format) {
+                case PIPE_FORMAT_UYVY:
+                case PIPE_FORMAT_YUYV:
+		default:
+			break;
+		}
+		goto out_unknown; /* TODO */
+		
+	case UTIL_FORMAT_COLORSPACE_SRGB:
+		word4 |= S_038010_FORCE_DEGAMMA(1);
+		if (format == PIPE_FORMAT_L8A8_SRGB || format == PIPE_FORMAT_L8_SRGB)
+			goto out_unknown; /* fails for some reason - TODO */
+		break;
+
+	default:
+		break;
+	}
+	
+	word4 |= r600_get_swizzle_combined(desc->swizzle, swizzle_view);
+
+	/* S3TC formats. TODO */
+	if (desc->layout == UTIL_FORMAT_LAYOUT_S3TC) {
+		goto out_unknown;
+	}
+
+
+	for (i = 0; i < desc->nr_channels; i++) {	
+		if (desc->channel[i].type == UTIL_FORMAT_TYPE_SIGNED) {
+			word4 |= sign_bit[i];
+		}
+	}
+
+	/* R8G8Bx_SNORM - TODO CxV8U8 */
+
+	/* RGTC - TODO */
+
+	/* See whether the components are of the same size. */
+	for (i = 1; i < desc->nr_channels; i++) {
+		uniform = uniform && desc->channel[0].size == desc->channel[i].size;
+	}
+	
+	/* Non-uniform formats. */
+	if (!uniform) {
+		switch(desc->nr_channels) {
+		case 3:
+			if (desc->channel[0].size == 5 &&
+			    desc->channel[1].size == 6 &&
+			    desc->channel[2].size == 5) {
+				result |= V_0280A0_COLOR_5_6_5;
+				goto out_word4;
+			}
+			goto out_unknown;
+		case 4:
+			if (desc->channel[0].size == 5 &&
+			    desc->channel[1].size == 5 &&
+			    desc->channel[2].size == 5 &&
+			    desc->channel[3].size == 1) {
+				result |= V_0280A0_COLOR_1_5_5_5;
+				goto out_word4;
+			}
+			if (desc->channel[0].size == 10 &&
+			    desc->channel[1].size == 10 &&
+			    desc->channel[2].size == 10 &&
+			    desc->channel[3].size == 2) {
+				result |= V_0280A0_COLOR_10_10_10_2;
+				goto out_word4;
+			}
+			goto out_unknown;
+		}
+		goto out_unknown;
+	}
+
+	/* uniform formats */
+	switch (desc->channel[0].type) {
+	case UTIL_FORMAT_TYPE_UNSIGNED:
+	case UTIL_FORMAT_TYPE_SIGNED:
+		if (!desc->channel[0].normalized &&
+		    desc->colorspace != UTIL_FORMAT_COLORSPACE_SRGB) {
+			goto out_unknown;
+		}
+
+		switch (desc->channel[0].size) {
+		case 4:
+			switch (desc->nr_channels) {
+			case 2:
+				result |= V_0280A0_COLOR_4_4;
+				goto out_word4;
+			case 4:
+				result |= V_0280A0_COLOR_4_4_4_4;
+				goto out_word4;
+			}
+			goto out_unknown;
+		case 8:
+			switch (desc->nr_channels) {
+			case 1:
+				result |= V_0280A0_COLOR_8;
+				goto out_word4;
+			case 2:
+				result |= V_0280A0_COLOR_8_8;
+				goto out_word4;
+			case 4:
+				result |= V_0280A0_COLOR_8_8_8_8;
+				goto out_word4;
+			}
+			goto out_unknown;
+		case 16:
+			switch (desc->nr_channels) {
+			case 1:
+				result |= V_0280A0_COLOR_16;
+				goto out_word4;
+			case 2:
+				result |= V_0280A0_COLOR_16_16;
+				goto out_word4;
+			case 4:
+				result |= V_0280A0_COLOR_16_16_16_16;
+				goto out_word4;
+			}
+		}
+		goto out_unknown;
+
+	case UTIL_FORMAT_TYPE_FLOAT:
+		switch (desc->channel[0].size) {
+		case 16:
+			switch (desc->nr_channels) {
+			case 1:
+				result |= V_0280A0_COLOR_16_FLOAT;
+				goto out_word4;
+			case 2:
+				result |= V_0280A0_COLOR_16_16_FLOAT;
+				goto out_word4;
+			case 4:
+				result |= V_0280A0_COLOR_16_16_16_16_FLOAT;
+				goto out_word4;
+			}
+			goto out_unknown;
+		case 32:
+			switch (desc->nr_channels) {
+			case 1:
+				result |= V_0280A0_COLOR_32_FLOAT;
+				goto out_word4;
+			case 2:
+				result |= V_0280A0_COLOR_32_32_FLOAT;
+				goto out_word4;
+			case 4:
+				result |= V_0280A0_COLOR_32_32_32_32_FLOAT;
+				goto out_word4;
+			}
+		}
+		
+	}
+out_word4:
+	if (word4_p)
+		*word4_p = word4;
+	if (yuv_format_p)
+		*yuv_format_p = yuv_format;
+//	fprintf(stderr,"returning %08x %08x %08x\n", result, word4, yuv_format);
+	return result;
+out_unknown:
+//	R600_ERR("Unable to handle texformat %d %s\n", format, util_format_name(format));
+	return ~0;
+}
diff --git a/src/gallium/drivers/r600/r600_texture.h b/src/gallium/drivers/r600/r600_texture.h
deleted file mode 100644
index 9bc08d6b040..00000000000
--- a/src/gallium/drivers/r600/r600_texture.h
+++ /dev/null
@@ -1,53 +0,0 @@
-/*
- * Copyright 2010 Jerome Glisse <[email protected]>
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * on the rights to use, copy, modify, merge, publish, distribute, sub
- * license, and/or sell copies of the Software, and to permit persons to whom
- * the Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
- * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
- * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
- * USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-#ifndef R600_TEXTURE_H
-#define R600_TEXTURE_H
-
-#include <pipe/p_state.h>
-
-struct r600_texture {
-	struct u_resource		b;
-	unsigned long			offset[PIPE_MAX_TEXTURE_LEVELS];
-	unsigned long			pitch[PIPE_MAX_TEXTURE_LEVELS];
-	unsigned long			stride[PIPE_MAX_TEXTURE_LEVELS];
-	unsigned long			layer_size[PIPE_MAX_TEXTURE_LEVELS];
-	unsigned long			stride_override;
-	unsigned long			size;
-	struct pipe_resource		*buffer;
-};
-
-struct pipe_resource *r600_texture_create(struct pipe_screen *screen,
-					  const struct pipe_resource *templ);
-unsigned long r600_texture_get_offset(struct r600_texture *rtex, unsigned level, unsigned zslice, unsigned face);
-struct pipe_resource *r600_texture_from_handle(struct pipe_screen *screen,
-					       const struct pipe_resource *base,
-					       struct winsys_handle *whandle);
-void r600_init_screen_texture_functions(struct pipe_screen *screen);
-
-/* This should be implemented by winsys. */
-boolean r600_buffer_get_handle(struct radeon *rw,
-			       struct pipe_resource *buf,
-			       struct winsys_handle *whandle);
-
-
-#endif
diff --git a/src/gallium/drivers/r600/r600d.h b/src/gallium/drivers/r600/r600d.h
index 44834984c68..53388f822ea 100644
--- a/src/gallium/drivers/r600/r600d.h
+++ b/src/gallium/drivers/r600/r600d.h
@@ -26,6 +26,8 @@
 #ifndef R600D_H
 #define R600D_H
 
+#define R600_TEXEL_PITCH_ALIGNMENT_MASK        0x7
+
 #define PKT3_NOP                               0x10
 #define PKT3_INDIRECT_BUFFER_END               0x17
 #define PKT3_SET_PREDICATION                   0x20
@@ -207,12 +209,24 @@
 #define   S_0280A0_NUMBER_TYPE(x)                      (((x) & 0x7) << 12)
 #define   G_0280A0_NUMBER_TYPE(x)                      (((x) >> 12) & 0x7)
 #define   C_0280A0_NUMBER_TYPE                         0xFFFF8FFF
+#define     V_0280A0_NUMBER_UNORM                      0x00000000
+#define     V_0280A0_NUMBER_SNORM                      0x00000001
+#define     V_0280A0_NUMBER_USCALED                    0x00000002
+#define     V_0280A0_NUMBER_SSCALED                    0x00000003
+#define     V_0280A0_NUMBER_UINT                       0x00000004
+#define     V_0280A0_NUMBER_SINT                       0x00000005
+#define     V_0280A0_NUMBER_SRGB                       0x00000006
+#define     V_0280A0_NUMBER_FLOAT                      0x00000007
 #define   S_0280A0_READ_SIZE(x)                        (((x) & 0x1) << 15)
 #define   G_0280A0_READ_SIZE(x)                        (((x) >> 15) & 0x1)
 #define   C_0280A0_READ_SIZE                           0xFFFF7FFF
 #define   S_0280A0_COMP_SWAP(x)                        (((x) & 0x3) << 16)
 #define   G_0280A0_COMP_SWAP(x)                        (((x) >> 16) & 0x3)
 #define   C_0280A0_COMP_SWAP                           0xFFFCFFFF
+#define     V_0280A0_SWAP_STD                          0x00000000
+#define     V_0280A0_SWAP_ALT                          0x00000001
+#define     V_0280A0_SWAP_STD_REV                      0x00000002
+#define     V_0280A0_SWAP_ALT_REV                      0x00000003
 #define   S_0280A0_TILE_MODE(x)                        (((x) & 0x3) << 18)
 #define   G_0280A0_TILE_MODE(x)                        (((x) >> 18) & 0x3)
 #define   C_0280A0_TILE_MODE                           0xFFF3FFFF
@@ -247,6 +261,16 @@
 #define   S_028060_SLICE_TILE_MAX(x)                   (((x) & 0xFFFFF) << 10)
 #define   G_028060_SLICE_TILE_MAX(x)                   (((x) >> 10) & 0xFFFFF)
 #define   C_028060_SLICE_TILE_MAX                      0xC00003FF
+#define R_028410_SX_ALPHA_TEST_CONTROL               0x028410
+#define   S_028410_ALPHA_FUNC(x)                       (((x) & 0x7) << 0)
+#define   G_028410_ALPHA_FUNC(x)                       (((x) >> 0) & 0x7)
+#define   C_028410_ALPHA_FUNC                          0xFFFFFFF8
+#define   S_028410_ALPHA_TEST_ENABLE(x)                (((x) & 0x1) << 3)
+#define   G_028410_ALPHA_TEST_ENABLE(x)                (((x) >> 3) & 0x1)
+#define   C_028410_ALPHA_TEST_ENABLE                   0xFFFFFFF7
+#define   S_028410_ALPHA_TEST_BYPASS(x)                (((x) & 0x1) << 8)
+#define   G_028410_ALPHA_TEST_BYPASS(x)                (((x) >> 8) & 0x1)
+#define   C_028410_ALPHA_TEST_BYPASS                   0xFFFFFEFF
 #define R_028800_DB_DEPTH_CONTROL                    0x028800
 #define   S_028800_STENCIL_ENABLE(x)                   (((x) & 0x1) << 0)
 #define   G_028800_STENCIL_ENABLE(x)                   (((x) >> 0) & 0x1)
@@ -266,9 +290,25 @@
 #define   S_028800_STENCILFUNC(x)                      (((x) & 0x7) << 8)
 #define   G_028800_STENCILFUNC(x)                      (((x) >> 8) & 0x7)
 #define   C_028800_STENCILFUNC                         0xFFFFF8FF
+#define     V_028800_STENCILFUNC_NEVER                 0x00000000
+#define     V_028800_STENCILFUNC_LESS                  0x00000001
+#define     V_028800_STENCILFUNC_EQUAL                 0x00000002
+#define     V_028800_STENCILFUNC_LEQUAL                0x00000003
+#define     V_028800_STENCILFUNC_GREATER               0x00000004
+#define     V_028800_STENCILFUNC_NOTEQUAL              0x00000005
+#define     V_028800_STENCILFUNC_GEQUAL                0x00000006
+#define     V_028800_STENCILFUNC_ALWAYS                0x00000007
 #define   S_028800_STENCILFAIL(x)                      (((x) & 0x7) << 11)
 #define   G_028800_STENCILFAIL(x)                      (((x) >> 11) & 0x7)
 #define   C_028800_STENCILFAIL                         0xFFFFC7FF
+#define     V_028800_STENCIL_KEEP                      0x00000000
+#define     V_028800_STENCIL_ZERO                      0x00000001
+#define     V_028800_STENCIL_REPLACE                   0x00000002
+#define     V_028800_STENCIL_INCR                      0x00000003
+#define     V_028800_STENCIL_DECR                      0x00000004
+#define     V_028800_STENCIL_INVERT                    0x00000005
+#define     V_028800_STENCIL_INCR_WRAP                 0x00000006
+#define     V_028800_STENCIL_DECR_WRAP                 0x00000007
 #define   S_028800_STENCILZPASS(x)                     (((x) & 0x7) << 14)
 #define   G_028800_STENCILZPASS(x)                     (((x) >> 14) & 0x7)
 #define   C_028800_STENCILZPASS                        0xFFFE3FFF
@@ -287,6 +327,86 @@
 #define   S_028800_STENCILZFAIL_BF(x)                  (((x) & 0x7) << 29)
 #define   G_028800_STENCILZFAIL_BF(x)                  (((x) >> 29) & 0x7)
 #define   C_028800_STENCILZFAIL_BF                     0x1FFFFFFF
+#define R_028808_CB_COLOR_CONTROL                    0x028808
+#define   S_028808_FOG_ENABLE(x)                       (((x) & 0x1) << 0)
+#define   G_028808_FOG_ENABLE(x)                       (((x) >> 0) & 0x1)
+#define   C_028808_FOG_ENABLE                          0xFFFFFFFE
+#define   S_028808_MULTIWRITE_ENABLE(x)                (((x) & 0x1) << 1)
+#define   G_028808_MULTIWRITE_ENABLE(x)                (((x) >> 1) & 0x1)
+#define   C_028808_MULTIWRITE_ENABLE                   0xFFFFFFFD
+#define   S_028808_DITHER_ENABLE(x)                    (((x) & 0x1) << 2)
+#define   G_028808_DITHER_ENABLE(x)                    (((x) >> 2) & 0x1)
+#define   C_028808_DITHER_ENABLE                       0xFFFFFFFB
+#define   S_028808_DEGAMMA_ENABLE(x)                   (((x) & 0x1) << 3)
+#define   G_028808_DEGAMMA_ENABLE(x)                   (((x) >> 3) & 0x1)
+#define   C_028808_DEGAMMA_ENABLE                      0xFFFFFFF7
+#define   S_028808_SPECIAL_OP(x)                       (((x) & 0x7) << 4)
+#define   G_028808_SPECIAL_OP(x)                       (((x) >> 4) & 0x7)
+#define   C_028808_SPECIAL_OP                          0xFFFFFF8F
+#define   S_028808_PER_MRT_BLEND(x)                    (((x) & 0x1) << 7)
+#define   G_028808_PER_MRT_BLEND(x)                    (((x) >> 7) & 0x1)
+#define   C_028808_PER_MRT_BLEND                       0xFFFFFF7F
+#define   S_028808_TARGET_BLEND_ENABLE(x)              (((x) & 0xFF) << 8)
+#define   G_028808_TARGET_BLEND_ENABLE(x)              (((x) >> 8) & 0xFF)
+#define   C_028808_TARGET_BLEND_ENABLE                 0xFFFF00FF
+#define   S_028808_ROP3(x)                             (((x) & 0xFF) << 16)
+#define   G_028808_ROP3(x)                             (((x) >> 16) & 0xFF)
+#define   C_028808_ROP3                                0xFF00FFFF
+#define R_028810_PA_CL_CLIP_CNTL                     0x028810
+#define   S_028810_UCP_ENA_0(x)                        (((x) & 0x1) << 0)
+#define   G_028810_UCP_ENA_0(x)                        (((x) >> 0) & 0x1)
+#define   C_028810_UCP_ENA_0                           0xFFFFFFFE
+#define   S_028810_UCP_ENA_1(x)                        (((x) & 0x1) << 1)
+#define   G_028810_UCP_ENA_1(x)                        (((x) >> 1) & 0x1)
+#define   C_028810_UCP_ENA_1                           0xFFFFFFFD
+#define   S_028810_UCP_ENA_2(x)                        (((x) & 0x1) << 2)
+#define   G_028810_UCP_ENA_2(x)                        (((x) >> 2) & 0x1)
+#define   C_028810_UCP_ENA_2                           0xFFFFFFFB
+#define   S_028810_UCP_ENA_3(x)                        (((x) & 0x1) << 3)
+#define   G_028810_UCP_ENA_3(x)                        (((x) >> 3) & 0x1)
+#define   C_028810_UCP_ENA_3                           0xFFFFFFF7
+#define   S_028810_UCP_ENA_4(x)                        (((x) & 0x1) << 4)
+#define   G_028810_UCP_ENA_4(x)                        (((x) >> 4) & 0x1)
+#define   C_028810_UCP_ENA_4                           0xFFFFFFEF
+#define   S_028810_UCP_ENA_5(x)                        (((x) & 0x1) << 5)
+#define   G_028810_UCP_ENA_5(x)                        (((x) >> 5) & 0x1)
+#define   C_028810_UCP_ENA_5                           0xFFFFFFDF
+#define   S_028810_PS_UCP_Y_SCALE_NEG(x)               (((x) & 0x1) << 13)
+#define   G_028810_PS_UCP_Y_SCALE_NEG(x)               (((x) >> 13) & 0x1)
+#define   C_028810_PS_UCP_Y_SCALE_NEG                  0xFFFFDFFF
+#define   S_028810_PS_UCP_MODE(x)                      (((x) & 0x3) << 14)
+#define   G_028810_PS_UCP_MODE(x)                      (((x) >> 14) & 0x3)
+#define   C_028810_PS_UCP_MODE                         0xFFFF3FFF
+#define   S_028810_CLIP_DISABLE(x)                     (((x) & 0x1) << 16)
+#define   G_028810_CLIP_DISABLE(x)                     (((x) >> 16) & 0x1)
+#define   C_028810_CLIP_DISABLE                        0xFFFEFFFF
+#define   S_028810_UCP_CULL_ONLY_ENA(x)                (((x) & 0x1) << 17)
+#define   G_028810_UCP_CULL_ONLY_ENA(x)                (((x) >> 17) & 0x1)
+#define   C_028810_UCP_CULL_ONLY_ENA                   0xFFFDFFFF
+#define   S_028810_BOUNDARY_EDGE_FLAG_ENA(x)           (((x) & 0x1) << 18)
+#define   G_028810_BOUNDARY_EDGE_FLAG_ENA(x)           (((x) >> 18) & 0x1)
+#define   C_028810_BOUNDARY_EDGE_FLAG_ENA              0xFFFBFFFF
+#define   S_028810_DX_CLIP_SPACE_DEF(x)                (((x) & 0x1) << 19)
+#define   G_028810_DX_CLIP_SPACE_DEF(x)                (((x) >> 19) & 0x1)
+#define   C_028810_DX_CLIP_SPACE_DEF                   0xFFF7FFFF
+#define   S_028810_DIS_CLIP_ERR_DETECT(x)              (((x) & 0x1) << 20)
+#define   G_028810_DIS_CLIP_ERR_DETECT(x)              (((x) >> 20) & 0x1)
+#define   C_028810_DIS_CLIP_ERR_DETECT                 0xFFEFFFFF
+#define   S_028810_VTX_KILL_OR(x)                      (((x) & 0x1) << 21)
+#define   G_028810_VTX_KILL_OR(x)                      (((x) >> 21) & 0x1)
+#define   C_028810_VTX_KILL_OR                         0xFFDFFFFF
+#define   S_028810_DX_LINEAR_ATTR_CLIP_ENA(x)          (((x) & 0x1) << 24)
+#define   G_028810_DX_LINEAR_ATTR_CLIP_ENA(x)          (((x) >> 24) & 0x1)
+#define   C_028810_DX_LINEAR_ATTR_CLIP_ENA             0xFEFFFFFF
+#define   S_028810_VTE_VPORT_PROVOKE_DISABLE(x)        (((x) & 0x1) << 25)
+#define   G_028810_VTE_VPORT_PROVOKE_DISABLE(x)        (((x) >> 25) & 0x1)
+#define   C_028810_VTE_VPORT_PROVOKE_DISABLE           0xFDFFFFFF
+#define   S_028810_ZCLIP_NEAR_DISABLE(x)               (((x) & 0x1) << 26)
+#define   G_028810_ZCLIP_NEAR_DISABLE(x)               (((x) >> 26) & 0x1)
+#define   C_028810_ZCLIP_NEAR_DISABLE                  0xFBFFFFFF
+#define   S_028810_ZCLIP_FAR_DISABLE(x)                (((x) & 0x1) << 27)
+#define   G_028810_ZCLIP_FAR_DISABLE(x)                (((x) >> 27) & 0x1)
+#define   C_028810_ZCLIP_FAR_DISABLE                   0xF7FFFFFF
 #define R_028010_DB_DEPTH_INFO                       0x028010
 #define   S_028010_FORMAT(x)                           (((x) & 0x7) << 0)
 #define   G_028010_FORMAT(x)                           (((x) >> 0) & 0x7)
@@ -314,6 +434,117 @@
 #define   S_028010_ZRANGE_PRECISION(x)                 (((x) & 0x1) << 31)
 #define   G_028010_ZRANGE_PRECISION(x)                 (((x) >> 31) & 0x1)
 #define   C_028010_ZRANGE_PRECISION                    0x7FFFFFFF
+#define R_028430_DB_STENCILREFMASK                   0x028430
+#define   S_028430_STENCILREF(x)                       (((x) & 0xFF) << 0)
+#define   G_028430_STENCILREF(x)                       (((x) >> 0) & 0xFF)
+#define   C_028430_STENCILREF                          0xFFFFFF00
+#define   S_028430_STENCILMASK(x)                      (((x) & 0xFF) << 8)
+#define   G_028430_STENCILMASK(x)                      (((x) >> 8) & 0xFF)
+#define   C_028430_STENCILMASK                         0xFFFF00FF
+#define   S_028430_STENCILWRITEMASK(x)                 (((x) & 0xFF) << 16)
+#define   G_028430_STENCILWRITEMASK(x)                 (((x) >> 16) & 0xFF)
+#define   C_028430_STENCILWRITEMASK                    0xFF00FFFF
+#define R_028434_DB_STENCILREFMASK_BF                0x028434
+#define   S_028434_STENCILREF_BF(x)                    (((x) & 0xFF) << 0)
+#define   G_028434_STENCILREF_BF(x)                    (((x) >> 0) & 0xFF)
+#define   C_028434_STENCILREF_BF                       0xFFFFFF00
+#define   S_028434_STENCILMASK_BF(x)                   (((x) & 0xFF) << 8)
+#define   G_028434_STENCILMASK_BF(x)                   (((x) >> 8) & 0xFF)
+#define   C_028434_STENCILMASK_BF                      0xFFFF00FF
+#define   S_028434_STENCILWRITEMASK_BF(x)              (((x) & 0xFF) << 16)
+#define   G_028434_STENCILWRITEMASK_BF(x)              (((x) >> 16) & 0xFF)
+#define   C_028434_STENCILWRITEMASK_BF                 0xFF00FFFF
+#define R_028804_CB_BLEND_CONTROL                    0x028804
+#define   S_028804_COLOR_SRCBLEND(x)                   (((x) & 0x1F) << 0)
+#define   G_028804_COLOR_SRCBLEND(x)                   (((x) >> 0) & 0x1F)
+#define   C_028804_COLOR_SRCBLEND                      0xFFFFFFE0
+#define     V_028804_BLEND_ZERO                        0x00000000
+#define     V_028804_BLEND_ONE                         0x00000001
+#define     V_028804_BLEND_SRC_COLOR                   0x00000002
+#define     V_028804_BLEND_ONE_MINUS_SRC_COLOR         0x00000003
+#define     V_028804_BLEND_SRC_ALPHA                   0x00000004
+#define     V_028804_BLEND_ONE_MINUS_SRC_ALPHA         0x00000005
+#define     V_028804_BLEND_DST_ALPHA                   0x00000006
+#define     V_028804_BLEND_ONE_MINUS_DST_ALPHA         0x00000007
+#define     V_028804_BLEND_DST_COLOR                   0x00000008
+#define     V_028804_BLEND_ONE_MINUS_DST_COLOR         0x00000009
+#define     V_028804_BLEND_SRC_ALPHA_SATURATE          0x0000000A
+#define     V_028804_BLEND_BOTH_SRC_ALPHA              0x0000000B
+#define     V_028804_BLEND_BOTH_INV_SRC_ALPHA          0x0000000C
+#define     V_028804_BLEND_CONST_COLOR                 0x0000000D
+#define     V_028804_BLEND_ONE_MINUS_CONST_COLOR       0x0000000E
+#define     V_028804_BLEND_SRC1_COLOR                  0x0000000F
+#define     V_028804_BLEND_INV_SRC1_COLOR              0x00000010
+#define     V_028804_BLEND_SRC1_ALPHA                  0x00000011
+#define     V_028804_BLEND_INV_SRC1_ALPHA              0x00000012
+#define     V_028804_BLEND_CONST_ALPHA                 0x00000013
+#define     V_028804_BLEND_ONE_MINUS_CONST_ALPHA       0x00000014
+#define   S_028804_COLOR_COMB_FCN(x)                   (((x) & 0x7) << 5)
+#define   G_028804_COLOR_COMB_FCN(x)                   (((x) >> 5) & 0x7)
+#define   C_028804_COLOR_COMB_FCN                      0xFFFFFF1F
+#define     V_028804_COMB_DST_PLUS_SRC                 0x00000000
+#define     V_028804_COMB_SRC_MINUS_DST                0x00000001
+#define     V_028804_COMB_MIN_DST_SRC                  0x00000002
+#define     V_028804_COMB_MAX_DST_SRC                  0x00000003
+#define     V_028804_COMB_DST_MINUS_SRC                0x00000004
+#define   S_028804_COLOR_DESTBLEND(x)                  (((x) & 0x1F) << 8)
+#define   G_028804_COLOR_DESTBLEND(x)                  (((x) >> 8) & 0x1F)
+#define   C_028804_COLOR_DESTBLEND                     0xFFFFE0FF
+#define   S_028804_OPACITY_WEIGHT(x)                   (((x) & 0x1) << 13)
+#define   G_028804_OPACITY_WEIGHT(x)                   (((x) >> 13) & 0x1)
+#define   C_028804_OPACITY_WEIGHT                      0xFFFFDFFF
+#define   S_028804_ALPHA_SRCBLEND(x)                   (((x) & 0x1F) << 16)
+#define   G_028804_ALPHA_SRCBLEND(x)                   (((x) >> 16) & 0x1F)
+#define   C_028804_ALPHA_SRCBLEND                      0xFFE0FFFF
+#define   S_028804_ALPHA_COMB_FCN(x)                   (((x) & 0x7) << 21)
+#define   G_028804_ALPHA_COMB_FCN(x)                   (((x) >> 21) & 0x7)
+#define   C_028804_ALPHA_COMB_FCN                      0xFF1FFFFF
+#define   S_028804_ALPHA_DESTBLEND(x)                  (((x) & 0x1F) << 24)
+#define   G_028804_ALPHA_DESTBLEND(x)                  (((x) >> 24) & 0x1F)
+#define   C_028804_ALPHA_DESTBLEND                     0xE0FFFFFF
+#define   S_028804_SEPARATE_ALPHA_BLEND(x)             (((x) & 0x1) << 29)
+#define   G_028804_SEPARATE_ALPHA_BLEND(x)             (((x) >> 29) & 0x1)
+#define   C_028804_SEPARATE_ALPHA_BLEND                0xDFFFFFFF
+#define R_028814_PA_SU_SC_MODE_CNTL                  0x028814
+#define   S_028814_CULL_FRONT(x)                       (((x) & 0x1) << 0)
+#define   G_028814_CULL_FRONT(x)                       (((x) >> 0) & 0x1)
+#define   C_028814_CULL_FRONT                          0xFFFFFFFE
+#define   S_028814_CULL_BACK(x)                        (((x) & 0x1) << 1)
+#define   G_028814_CULL_BACK(x)                        (((x) >> 1) & 0x1)
+#define   C_028814_CULL_BACK                           0xFFFFFFFD
+#define   S_028814_FACE(x)                             (((x) & 0x1) << 2)
+#define   G_028814_FACE(x)                             (((x) >> 2) & 0x1)
+#define   C_028814_FACE                                0xFFFFFFFB
+#define   S_028814_POLY_MODE(x)                        (((x) & 0x3) << 3)
+#define   G_028814_POLY_MODE(x)                        (((x) >> 3) & 0x3)
+#define   C_028814_POLY_MODE                           0xFFFFFFE7
+#define   S_028814_POLYMODE_FRONT_PTYPE(x)             (((x) & 0x7) << 5)
+#define   G_028814_POLYMODE_FRONT_PTYPE(x)             (((x) >> 5) & 0x7)
+#define   C_028814_POLYMODE_FRONT_PTYPE                0xFFFFFF1F
+#define   S_028814_POLYMODE_BACK_PTYPE(x)              (((x) & 0x7) << 8)
+#define   G_028814_POLYMODE_BACK_PTYPE(x)              (((x) >> 8) & 0x7)
+#define   C_028814_POLYMODE_BACK_PTYPE                 0xFFFFF8FF
+#define   S_028814_POLY_OFFSET_FRONT_ENABLE(x)         (((x) & 0x1) << 11)
+#define   G_028814_POLY_OFFSET_FRONT_ENABLE(x)         (((x) >> 11) & 0x1)
+#define   C_028814_POLY_OFFSET_FRONT_ENABLE            0xFFFFF7FF
+#define   S_028814_POLY_OFFSET_BACK_ENABLE(x)          (((x) & 0x1) << 12)
+#define   G_028814_POLY_OFFSET_BACK_ENABLE(x)          (((x) >> 12) & 0x1)
+#define   C_028814_POLY_OFFSET_BACK_ENABLE             0xFFFFEFFF
+#define   S_028814_POLY_OFFSET_PARA_ENABLE(x)          (((x) & 0x1) << 13)
+#define   G_028814_POLY_OFFSET_PARA_ENABLE(x)          (((x) >> 13) & 0x1)
+#define   C_028814_POLY_OFFSET_PARA_ENABLE             0xFFFFDFFF
+#define   S_028814_VTX_WINDOW_OFFSET_ENABLE(x)         (((x) & 0x1) << 16)
+#define   G_028814_VTX_WINDOW_OFFSET_ENABLE(x)         (((x) >> 16) & 0x1)
+#define   C_028814_VTX_WINDOW_OFFSET_ENABLE            0xFFFEFFFF
+#define   S_028814_PROVOKING_VTX_LAST(x)               (((x) & 0x1) << 19)
+#define   G_028814_PROVOKING_VTX_LAST(x)               (((x) >> 19) & 0x1)
+#define   C_028814_PROVOKING_VTX_LAST                  0xFFF7FFFF
+#define   S_028814_PERSP_CORR_DIS(x)                   (((x) & 0x1) << 20)
+#define   G_028814_PERSP_CORR_DIS(x)                   (((x) >> 20) & 0x1)
+#define   C_028814_PERSP_CORR_DIS                      0xFFEFFFFF
+#define   S_028814_MULTI_PRIM_IB_ENA(x)                (((x) & 0x1) << 21)
+#define   G_028814_MULTI_PRIM_IB_ENA(x)                (((x) >> 21) & 0x1)
+#define   C_028814_MULTI_PRIM_IB_ENA                   0xFFDFFFFF
 #define R_028000_DB_DEPTH_SIZE                       0x028000
 #define   S_028000_PITCH_TILE_MAX(x)                   (((x) & 0x3FF) << 0)
 #define   G_028000_PITCH_TILE_MAX(x)                   (((x) >> 0) & 0x3FF)
@@ -400,6 +631,36 @@
 #define   S_028D10_IGNORE_SC_ZRANGE(x)                 (((x) & 0x1) << 17)
 #define   G_028D10_IGNORE_SC_ZRANGE(x)                 (((x) >> 17) & 0x1)
 #define   C_028D10_IGNORE_SC_ZRANGE                    0xFFFDFFFF
+#define R_028DF8_PA_SU_POLY_OFFSET_DB_FMT_CNTL       0x028DF8
+#define   S_028DF8_POLY_OFFSET_NEG_NUM_DB_BITS(x)      (((x) & 0xFF) << 0)
+#define   G_028DF8_POLY_OFFSET_NEG_NUM_DB_BITS(x)      (((x) >> 0) & 0xFF)
+#define   C_028DF8_POLY_OFFSET_NEG_NUM_DB_BITS         0xFFFFFF00
+#define   S_028DF8_POLY_OFFSET_DB_IS_FLOAT_FMT(x)      (((x) & 0x1) << 8)
+#define   G_028DF8_POLY_OFFSET_DB_IS_FLOAT_FMT(x)      (((x) >> 8) & 0x1)
+#define   C_028DF8_POLY_OFFSET_DB_IS_FLOAT_FMT         0xFFFFFEFF
+#define R_028E00_PA_SU_POLY_OFFSET_FRONT_SCALE       0x028E00
+#define   S_028E00_SCALE(x)                            (((x) & 0xFFFFFFFF) << 0)
+#define   G_028E00_SCALE(x)                            (((x) >> 0) & 0xFFFFFFFF)
+#define   C_028E00_SCALE                               0x00000000
+#define R_028E04_PA_SU_POLY_OFFSET_FRONT_OFFSET      0x028E04
+#define   S_028E04_OFFSET(x)                           (((x) & 0xFFFFFFFF) << 0)
+#define   G_028E04_OFFSET(x)                           (((x) >> 0) & 0xFFFFFFFF)
+#define   C_028E04_OFFSET                              0x00000000
+#define R_028E08_PA_SU_POLY_OFFSET_BACK_SCALE        0x028E08
+#define   S_028E08_SCALE(x)                            (((x) & 0xFFFFFFFF) << 0)
+#define   G_028E08_SCALE(x)                            (((x) >> 0) & 0xFFFFFFFF)
+#define   C_028E08_SCALE                               0x00000000
+#define R_028E0C_PA_SU_POLY_OFFSET_BACK_OFFSET       0x028E0C
+#define   S_028E0C_OFFSET(x)                           (((x) & 0xFFFFFFFF) << 0)
+#define   G_028E0C_OFFSET(x)                           (((x) >> 0) & 0xFFFFFFFF)
+#define   C_028E0C_OFFSET                              0x00000000
+#define R_028A00_PA_SU_POINT_SIZE                    0x028A00
+#define   S_028A00_HEIGHT(x)                           (((x) & 0xFFFF) << 0)
+#define   G_028A00_HEIGHT(x)                           (((x) >> 0) & 0xFFFF)
+#define   C_028A00_HEIGHT                              0xFFFF0000
+#define   S_028A00_WIDTH(x)                            (((x) & 0xFFFF) << 16)
+#define   G_028A00_WIDTH(x)                            (((x) >> 16) & 0xFFFF)
+#define   C_028A00_WIDTH                               0x0000FFFF
 #define R_028A40_VGT_GS_MODE                         0x028A40
 #define   S_028A40_MODE(x)                             (((x) & 0x3) << 0)
 #define   G_028A40_MODE(x)                             (((x) >> 0) & 0x3)
@@ -574,6 +835,132 @@
 #define   S_0287F0_USE_OPAQUE(x)                       (((x) & 0x1) << 6)
 #define   G_0287F0_USE_OPAQUE(x)                       (((x) >> 6) & 0x1)
 #define   C_0287F0_USE_OPAQUE                          0xFFFFFFBF
+#define R_038000_SQ_TEX_RESOURCE_WORD0_0             0x038000
+#define   S_038000_DIM(x)                              (((x) & 0x7) << 0)
+#define   G_038000_DIM(x)                              (((x) >> 0) & 0x7)
+#define   C_038000_DIM                                 0xFFFFFFF8
+#define     V_038000_SQ_TEX_DIM_1D                     0x00000000
+#define     V_038000_SQ_TEX_DIM_2D                     0x00000001
+#define     V_038000_SQ_TEX_DIM_3D                     0x00000002
+#define     V_038000_SQ_TEX_DIM_CUBEMAP                0x00000003
+#define     V_038000_SQ_TEX_DIM_1D_ARRAY               0x00000004
+#define     V_038000_SQ_TEX_DIM_2D_ARRAY               0x00000005
+#define     V_038000_SQ_TEX_DIM_2D_MSAA                0x00000006
+#define     V_038000_SQ_TEX_DIM_2D_ARRAY_MSAA          0x00000007
+#define   S_038000_TILE_MODE(x)                        (((x) & 0xF) << 3)
+#define   G_038000_TILE_MODE(x)                        (((x) >> 3) & 0xF)
+#define   C_038000_TILE_MODE                           0xFFFFFF87
+#define   S_038000_TILE_TYPE(x)                        (((x) & 0x1) << 7)
+#define   G_038000_TILE_TYPE(x)                        (((x) >> 7) & 0x1)
+#define   C_038000_TILE_TYPE                           0xFFFFFF7F
+#define   S_038000_PITCH(x)                            (((x) & 0x7FF) << 8)
+#define   G_038000_PITCH(x)                            (((x) >> 8) & 0x7FF)
+#define   C_038000_PITCH                               0xFFF800FF
+#define   S_038000_TEX_WIDTH(x)                        (((x) & 0x1FFF) << 19)
+#define   G_038000_TEX_WIDTH(x)                        (((x) >> 19) & 0x1FFF)
+#define   C_038000_TEX_WIDTH                           0x0007FFFF
+#define R_038004_SQ_TEX_RESOURCE_WORD1_0             0x038004
+#define   S_038004_TEX_HEIGHT(x)                       (((x) & 0x1FFF) << 0)
+#define   G_038004_TEX_HEIGHT(x)                       (((x) >> 0) & 0x1FFF)
+#define   C_038004_TEX_HEIGHT                          0xFFFFE000
+#define   S_038004_TEX_DEPTH(x)                        (((x) & 0x1FFF) << 13)
+#define   G_038004_TEX_DEPTH(x)                        (((x) >> 13) & 0x1FFF)
+#define   C_038004_TEX_DEPTH                           0xFC001FFF
+#define   S_038004_DATA_FORMAT(x)                      (((x) & 0x3F) << 26)
+#define   G_038004_DATA_FORMAT(x)                      (((x) >> 26) & 0x3F)
+#define   C_038004_DATA_FORMAT                         0x03FFFFFF
+#define R_038008_SQ_TEX_RESOURCE_WORD2_0             0x038008
+#define   S_038008_BASE_ADDRESS(x)                     (((x) & 0xFFFFFFFF) << 0)
+#define   G_038008_BASE_ADDRESS(x)                     (((x) >> 0) & 0xFFFFFFFF)
+#define   C_038008_BASE_ADDRESS                        0x00000000
+#define R_03800C_SQ_TEX_RESOURCE_WORD3_0             0x03800C
+#define   S_03800C_MIP_ADDRESS(x)                      (((x) & 0xFFFFFFFF) << 0)
+#define   G_03800C_MIP_ADDRESS(x)                      (((x) >> 0) & 0xFFFFFFFF)
+#define   C_03800C_MIP_ADDRESS                         0x00000000
+#define R_038010_SQ_TEX_RESOURCE_WORD4_0             0x038010
+#define   S_038010_FORMAT_COMP_X(x)                    (((x) & 0x3) << 0)
+#define   G_038010_FORMAT_COMP_X(x)                    (((x) >> 0) & 0x3)
+#define   C_038010_FORMAT_COMP_X                       0xFFFFFFFC
+#define     V_038010_SQ_FORMAT_COMP_UNSIGNED           0x00000000
+#define     V_038010_SQ_FORMAT_COMP_SIGNED             0x00000001
+#define     V_038010_SQ_FORMAT_COMP_UNSIGNED_BIASED    0x00000002
+#define   S_038010_FORMAT_COMP_Y(x)                    (((x) & 0x3) << 2)
+#define   G_038010_FORMAT_COMP_Y(x)                    (((x) >> 2) & 0x3)
+#define   C_038010_FORMAT_COMP_Y                       0xFFFFFFF3
+#define   S_038010_FORMAT_COMP_Z(x)                    (((x) & 0x3) << 4)
+#define   G_038010_FORMAT_COMP_Z(x)                    (((x) >> 4) & 0x3)
+#define   C_038010_FORMAT_COMP_Z                       0xFFFFFFCF
+#define   S_038010_FORMAT_COMP_W(x)                    (((x) & 0x3) << 6)
+#define   G_038010_FORMAT_COMP_W(x)                    (((x) >> 6) & 0x3)
+#define   C_038010_FORMAT_COMP_W                       0xFFFFFF3F
+#define   S_038010_NUM_FORMAT_ALL(x)                   (((x) & 0x3) << 8)
+#define   G_038010_NUM_FORMAT_ALL(x)                   (((x) >> 8) & 0x3)
+#define   C_038010_NUM_FORMAT_ALL                      0xFFFFFCFF
+#define     V_038010_SQ_NUM_FORMAT_NORM                0x00000000
+#define     V_038010_SQ_NUM_FORMAT_INT                 0x00000001
+#define     V_038010_SQ_NUM_FORMAT_SCALED              0x00000002
+#define   S_038010_SRF_MODE_ALL(x)                     (((x) & 0x1) << 10)
+#define   G_038010_SRF_MODE_ALL(x)                     (((x) >> 10) & 0x1)
+#define   C_038010_SRF_MODE_ALL                        0xFFFFFBFF
+#define     V_038010_SFR_MODE_ZERO_CLAMP_MINUS_ONE     0x00000000
+#define     V_038010_SFR_MODE_NO_ZERO                  0x00000001
+#define   S_038010_FORCE_DEGAMMA(x)                    (((x) & 0x1) << 11)
+#define   G_038010_FORCE_DEGAMMA(x)                    (((x) >> 11) & 0x1)
+#define   C_038010_FORCE_DEGAMMA                       0xFFFFF7FF
+#define   S_038010_ENDIAN_SWAP(x)                      (((x) & 0x3) << 12)
+#define   G_038010_ENDIAN_SWAP(x)                      (((x) >> 12) & 0x3)
+#define   C_038010_ENDIAN_SWAP                         0xFFFFCFFF
+#define   S_038010_REQUEST_SIZE(x)                     (((x) & 0x3) << 14)
+#define   G_038010_REQUEST_SIZE(x)                     (((x) >> 14) & 0x3)
+#define   C_038010_REQUEST_SIZE                        0xFFFF3FFF
+#define   S_038010_DST_SEL_X(x)                        (((x) & 0x7) << 16)
+#define   G_038010_DST_SEL_X(x)                        (((x) >> 16) & 0x7)
+#define   C_038010_DST_SEL_X                           0xFFF8FFFF
+#define     V_038010_SQ_SEL_X                          0x00000000
+#define     V_038010_SQ_SEL_Y                          0x00000001
+#define     V_038010_SQ_SEL_Z                          0x00000002
+#define     V_038010_SQ_SEL_W                          0x00000003
+#define     V_038010_SQ_SEL_0                          0x00000004
+#define     V_038010_SQ_SEL_1                          0x00000005
+#define   S_038010_DST_SEL_Y(x)                        (((x) & 0x7) << 19)
+#define   G_038010_DST_SEL_Y(x)                        (((x) >> 19) & 0x7)
+#define   C_038010_DST_SEL_Y                           0xFFC7FFFF
+#define   S_038010_DST_SEL_Z(x)                        (((x) & 0x7) << 22)
+#define   G_038010_DST_SEL_Z(x)                        (((x) >> 22) & 0x7)
+#define   C_038010_DST_SEL_Z                           0xFE3FFFFF
+#define   S_038010_DST_SEL_W(x)                        (((x) & 0x7) << 25)
+#define   G_038010_DST_SEL_W(x)                        (((x) >> 25) & 0x7)
+#define   C_038010_DST_SEL_W                           0xF1FFFFFF
+#define   S_038010_BASE_LEVEL(x)                       (((x) & 0xF) << 28)
+#define   G_038010_BASE_LEVEL(x)                       (((x) >> 28) & 0xF)
+#define   C_038010_BASE_LEVEL                          0x0FFFFFFF
+#define R_038014_SQ_TEX_RESOURCE_WORD5_0             0x038014
+#define   S_038014_LAST_LEVEL(x)                       (((x) & 0xF) << 0)
+#define   G_038014_LAST_LEVEL(x)                       (((x) >> 0) & 0xF)
+#define   C_038014_LAST_LEVEL                          0xFFFFFFF0
+#define   S_038014_BASE_ARRAY(x)                       (((x) & 0x1FFF) << 4)
+#define   G_038014_BASE_ARRAY(x)                       (((x) >> 4) & 0x1FFF)
+#define   C_038014_BASE_ARRAY                          0xFFFE000F
+#define   S_038014_LAST_ARRAY(x)                       (((x) & 0x1FFF) << 17)
+#define   G_038014_LAST_ARRAY(x)                       (((x) >> 17) & 0x1FFF)
+#define   C_038014_LAST_ARRAY                          0xC001FFFF
+#define R_038018_SQ_TEX_RESOURCE_WORD6_0             0x038018
+#define   S_038018_MPEG_CLAMP(x)                       (((x) & 0x3) << 0)
+#define   G_038018_MPEG_CLAMP(x)                       (((x) >> 0) & 0x3)
+#define   C_038018_MPEG_CLAMP                          0xFFFFFFFC
+#define   S_038018_PERF_MODULATION(x)                  (((x) & 0x7) << 5)
+#define   G_038018_PERF_MODULATION(x)                  (((x) >> 5) & 0x7)
+#define   C_038018_PERF_MODULATION                     0xFFFFFF1F
+#define   S_038018_INTERLACED(x)                       (((x) & 0x1) << 8)
+#define   G_038018_INTERLACED(x)                       (((x) >> 8) & 0x1)
+#define   C_038018_INTERLACED                          0xFFFFFEFF
+#define   S_038018_TYPE(x)                             (((x) & 0x3) << 30)
+#define   G_038018_TYPE(x)                             (((x) >> 30) & 0x3)
+#define   C_038018_TYPE                                0x3FFFFFFF
+#define     V_038010_SQ_TEX_VTX_INVALID_TEXTURE        0x00000000
+#define     V_038010_SQ_TEX_VTX_INVALID_BUFFER         0x00000001
+#define     V_038010_SQ_TEX_VTX_VALID_TEXTURE          0x00000002
+#define     V_038010_SQ_TEX_VTX_VALID_BUFFER           0x00000003
 #define R_038008_SQ_VTX_CONSTANT_WORD2_0             0x038008
 #define   S_038008_BASE_ADDRESS_HI(x)                  (((x) & 0xFF) << 0)
 #define   G_038008_BASE_ADDRESS_HI(x)                  (((x) >> 0) & 0xFF)
@@ -633,6 +1020,113 @@
 #define   S_038008_ENDIAN_SWAP(x)                      (((x) & 0x3) << 30)
 #define   G_038008_ENDIAN_SWAP(x)                      (((x) >> 30) & 0x3)
 #define   C_038008_ENDIAN_SWAP                         0x3FFFFFFF
+#define R_03C000_SQ_TEX_SAMPLER_WORD0_0              0x03C000
+#define   S_03C000_CLAMP_X(x)                          (((x) & 0x7) << 0)
+#define   G_03C000_CLAMP_X(x)                          (((x) >> 0) & 0x7)
+#define   C_03C000_CLAMP_X                             0xFFFFFFF8
+#define     V_03C000_SQ_TEX_WRAP                       0x00000000
+#define     V_03C000_SQ_TEX_MIRROR                     0x00000001
+#define     V_03C000_SQ_TEX_CLAMP_LAST_TEXEL           0x00000002
+#define     V_03C000_SQ_TEX_MIRROR_ONCE_LAST_TEXEL     0x00000003
+#define     V_03C000_SQ_TEX_CLAMP_HALF_BORDER          0x00000004
+#define     V_03C000_SQ_TEX_MIRROR_ONCE_HALF_BORDER    0x00000005
+#define     V_03C000_SQ_TEX_CLAMP_BORDER               0x00000006
+#define     V_03C000_SQ_TEX_MIRROR_ONCE_BORDER         0x00000007
+#define   S_03C000_CLAMP_Y(x)                          (((x) & 0x7) << 3)
+#define   G_03C000_CLAMP_Y(x)                          (((x) >> 3) & 0x7)
+#define   C_03C000_CLAMP_Y                             0xFFFFFFC7
+#define   S_03C000_CLAMP_Z(x)                          (((x) & 0x7) << 6)
+#define   G_03C000_CLAMP_Z(x)                          (((x) >> 6) & 0x7)
+#define   C_03C000_CLAMP_Z                             0xFFFFFE3F
+#define   S_03C000_XY_MAG_FILTER(x)                    (((x) & 0x7) << 9)
+#define   G_03C000_XY_MAG_FILTER(x)                    (((x) >> 9) & 0x7)
+#define   C_03C000_XY_MAG_FILTER                       0xFFFFF1FF
+#define     V_03C000_SQ_TEX_XY_FILTER_POINT            0x00000000
+#define     V_03C000_SQ_TEX_XY_FILTER_BILINEAR         0x00000001
+#define     V_03C000_SQ_TEX_XY_FILTER_BICUBIC          0x00000002
+#define   S_03C000_XY_MIN_FILTER(x)                    (((x) & 0x7) << 12)
+#define   G_03C000_XY_MIN_FILTER(x)                    (((x) >> 12) & 0x7)
+#define   C_03C000_XY_MIN_FILTER                       0xFFFF8FFF
+#define   S_03C000_Z_FILTER(x)                         (((x) & 0x3) << 15)
+#define   G_03C000_Z_FILTER(x)                         (((x) >> 15) & 0x3)
+#define   C_03C000_Z_FILTER                            0xFFFE7FFF
+#define     V_03C000_SQ_TEX_Z_FILTER_NONE              0x00000000
+#define     V_03C000_SQ_TEX_Z_FILTER_POINT             0x00000001
+#define     V_03C000_SQ_TEX_Z_FILTER_LINEAR            0x00000002
+#define   S_03C000_MIP_FILTER(x)                       (((x) & 0x3) << 17)
+#define   G_03C000_MIP_FILTER(x)                       (((x) >> 17) & 0x3)
+#define   C_03C000_MIP_FILTER                          0xFFF9FFFF
+#define   S_03C000_BORDER_COLOR_TYPE(x)                (((x) & 0x3) << 22)
+#define   G_03C000_BORDER_COLOR_TYPE(x)                (((x) >> 22) & 0x3)
+#define   C_03C000_BORDER_COLOR_TYPE                   0xFF3FFFFF
+#define     V_03C000_SQ_TEX_BORDER_COLOR_TRANS_BLACK   0x00000000
+#define     V_03C000_SQ_TEX_BORDER_COLOR_OPAQUE_BLACK  0x00000001
+#define     V_03C000_SQ_TEX_BORDER_COLOR_OPAQUE_WHITE  0x00000002
+#define     V_03C000_SQ_TEX_BORDER_COLOR_REGISTER      0x00000003
+#define   S_03C000_POINT_SAMPLING_CLAMP(x)             (((x) & 0x1) << 24)
+#define   G_03C000_POINT_SAMPLING_CLAMP(x)             (((x) >> 24) & 0x1)
+#define   C_03C000_POINT_SAMPLING_CLAMP                0xFEFFFFFF
+#define   S_03C000_TEX_ARRAY_OVERRIDE(x)               (((x) & 0x1) << 25)
+#define   G_03C000_TEX_ARRAY_OVERRIDE(x)               (((x) >> 25) & 0x1)
+#define   C_03C000_TEX_ARRAY_OVERRIDE                  0xFDFFFFFF
+#define   S_03C000_DEPTH_COMPARE_FUNCTION(x)           (((x) & 0x7) << 26)
+#define   G_03C000_DEPTH_COMPARE_FUNCTION(x)           (((x) >> 26) & 0x7)
+#define   C_03C000_DEPTH_COMPARE_FUNCTION              0xE3FFFFFF
+#define     V_03C000_SQ_TEX_DEPTH_COMPARE_NEVER        0x00000000
+#define     V_03C000_SQ_TEX_DEPTH_COMPARE_LESS         0x00000001
+#define     V_03C000_SQ_TEX_DEPTH_COMPARE_EQUAL        0x00000002
+#define     V_03C000_SQ_TEX_DEPTH_COMPARE_LESSEQUAL    0x00000003
+#define     V_03C000_SQ_TEX_DEPTH_COMPARE_GREATER      0x00000004
+#define     V_03C000_SQ_TEX_DEPTH_COMPARE_NOTEQUAL     0x00000005
+#define     V_03C000_SQ_TEX_DEPTH_COMPARE_GREATEREQUAL 0x00000006
+#define     V_03C000_SQ_TEX_DEPTH_COMPARE_ALWAYS       0x00000007
+#define   S_03C000_CHROMA_KEY(x)                       (((x) & 0x3) << 29)
+#define   G_03C000_CHROMA_KEY(x)                       (((x) >> 29) & 0x3)
+#define   C_03C000_CHROMA_KEY                          0x9FFFFFFF
+#define     V_03C000_SQ_TEX_CHROMA_KEY_DISABLE         0x00000000
+#define     V_03C000_SQ_TEX_CHROMA_KEY_KILL            0x00000001
+#define     V_03C000_SQ_TEX_CHROMA_KEY_BLEND           0x00000002
+#define   S_03C000_LOD_USES_MINOR_AXIS(x)              (((x) & 0x1) << 31)
+#define   G_03C000_LOD_USES_MINOR_AXIS(x)              (((x) >> 31) & 0x1)
+#define   C_03C000_LOD_USES_MINOR_AXIS                 0x7FFFFFFF
+#define R_03C004_SQ_TEX_SAMPLER_WORD1_0              0x03C004
+#define   S_03C004_MIN_LOD(x)                          (((x) & 0x3FF) << 0)
+#define   G_03C004_MIN_LOD(x)                          (((x) >> 0) & 0x3FF)
+#define   C_03C004_MIN_LOD                             0xFFFFFC00
+#define   S_03C004_MAX_LOD(x)                          (((x) & 0x3FF) << 10)
+#define   G_03C004_MAX_LOD(x)                          (((x) >> 10) & 0x3FF)
+#define   C_03C004_MAX_LOD                             0xFFF003FF
+#define   S_03C004_LOD_BIAS(x)                         (((x) & 0xFFF) << 20)
+#define   G_03C004_LOD_BIAS(x)                         (((x) >> 20) & 0xFFF)
+#define   C_03C004_LOD_BIAS                            0x000FFFFF
+#define R_03C008_SQ_TEX_SAMPLER_WORD2_0              0x03C008
+#define   S_03C008_LOD_BIAS_SEC(x)                     (((x) & 0xFFF) << 0)
+#define   G_03C008_LOD_BIAS_SEC(x)                     (((x) >> 0) & 0xFFF)
+#define   C_03C008_LOD_BIAS_SEC                        0xFFFFF000
+#define   S_03C008_MC_COORD_TRUNCATE(x)                (((x) & 0x1) << 12)
+#define   G_03C008_MC_COORD_TRUNCATE(x)                (((x) >> 12) & 0x1)
+#define   C_03C008_MC_COORD_TRUNCATE                   0xFFFFEFFF
+#define   S_03C008_FORCE_DEGAMMA(x)                    (((x) & 0x1) << 13)
+#define   G_03C008_FORCE_DEGAMMA(x)                    (((x) >> 13) & 0x1)
+#define   C_03C008_FORCE_DEGAMMA                       0xFFFFDFFF
+#define   S_03C008_HIGH_PRECISION_FILTER(x)            (((x) & 0x1) << 14)
+#define   G_03C008_HIGH_PRECISION_FILTER(x)            (((x) >> 14) & 0x1)
+#define   C_03C008_HIGH_PRECISION_FILTER               0xFFFFBFFF
+#define   S_03C008_PERF_MIP(x)                         (((x) & 0x7) << 15)
+#define   G_03C008_PERF_MIP(x)                         (((x) >> 15) & 0x7)
+#define   C_03C008_PERF_MIP                            0xFFFC7FFF
+#define   S_03C008_PERF_Z(x)                           (((x) & 0x3) << 18)
+#define   G_03C008_PERF_Z(x)                           (((x) >> 18) & 0x3)
+#define   C_03C008_PERF_Z                              0xFFF3FFFF
+#define   S_03C008_FETCH_4(x)                          (((x) & 0x1) << 26)
+#define   G_03C008_FETCH_4(x)                          (((x) >> 26) & 0x1)
+#define   C_03C008_FETCH_4                             0xFBFFFFFF
+#define   S_03C008_SAMPLE_IS_PCF(x)                    (((x) & 0x1) << 27)
+#define   G_03C008_SAMPLE_IS_PCF(x)                    (((x) >> 27) & 0x1)
+#define   C_03C008_SAMPLE_IS_PCF                       0xF7FFFFFF
+#define   S_03C008_TYPE(x)                             (((x) & 0x1) << 31)
+#define   G_03C008_TYPE(x)                             (((x) >> 31) & 0x1)
+#define   C_03C008_TYPE                                0x7FFFFFFF
 #define R_008958_VGT_PRIMITIVE_TYPE                  0x008958
 #define   S_008958_PRIM_TYPE(x)                        (((x) & 0x3F) << 0)
 #define   G_008958_PRIM_TYPE(x)                        (((x) >> 0) & 0x3F)
@@ -666,6 +1160,79 @@
 #define     V_008958_DI_PT_2D_FILL_RECT_LIST           0x0000001A
 #define     V_008958_DI_PT_2D_LINE_STRIP               0x0000001B
 #define     V_008958_DI_PT_2D_TRI_STRIP                0x0000001C
+#define R_02881C_PA_CL_VS_OUT_CNTL                   0x02881C
+#define   S_02881C_CLIP_DIST_ENA_0(x)                  (((x) & 0x1) << 0)
+#define   G_02881C_CLIP_DIST_ENA_0(x)                  (((x) >> 0) & 0x1)
+#define   C_02881C_CLIP_DIST_ENA_0                     0xFFFFFFFE
+#define   S_02881C_CLIP_DIST_ENA_1(x)                  (((x) & 0x1) << 1)
+#define   G_02881C_CLIP_DIST_ENA_1(x)                  (((x) >> 1) & 0x1)
+#define   C_02881C_CLIP_DIST_ENA_1                     0xFFFFFFFD
+#define   S_02881C_CLIP_DIST_ENA_2(x)                  (((x) & 0x1) << 2)
+#define   G_02881C_CLIP_DIST_ENA_2(x)                  (((x) >> 2) & 0x1)
+#define   C_02881C_CLIP_DIST_ENA_2                     0xFFFFFFFB
+#define   S_02881C_CLIP_DIST_ENA_3(x)                  (((x) & 0x1) << 3)
+#define   G_02881C_CLIP_DIST_ENA_3(x)                  (((x) >> 3) & 0x1)
+#define   C_02881C_CLIP_DIST_ENA_3                     0xFFFFFFF7
+#define   S_02881C_CLIP_DIST_ENA_4(x)                  (((x) & 0x1) << 4)
+#define   G_02881C_CLIP_DIST_ENA_4(x)                  (((x) >> 4) & 0x1)
+#define   C_02881C_CLIP_DIST_ENA_4                     0xFFFFFFEF
+#define   S_02881C_CLIP_DIST_ENA_5(x)                  (((x) & 0x1) << 5)
+#define   G_02881C_CLIP_DIST_ENA_5(x)                  (((x) >> 5) & 0x1)
+#define   C_02881C_CLIP_DIST_ENA_5                     0xFFFFFFDF
+#define   S_02881C_CLIP_DIST_ENA_6(x)                  (((x) & 0x1) << 6)
+#define   G_02881C_CLIP_DIST_ENA_6(x)                  (((x) >> 6) & 0x1)
+#define   C_02881C_CLIP_DIST_ENA_6                     0xFFFFFFBF
+#define   S_02881C_CLIP_DIST_ENA_7(x)                  (((x) & 0x1) << 7)
+#define   G_02881C_CLIP_DIST_ENA_7(x)                  (((x) >> 7) & 0x1)
+#define   C_02881C_CLIP_DIST_ENA_7                     0xFFFFFF7F
+#define   S_02881C_CULL_DIST_ENA_0(x)                  (((x) & 0x1) << 8)
+#define   G_02881C_CULL_DIST_ENA_0(x)                  (((x) >> 8) & 0x1)
+#define   C_02881C_CULL_DIST_ENA_0                     0xFFFFFEFF
+#define   S_02881C_CULL_DIST_ENA_1(x)                  (((x) & 0x1) << 9)
+#define   G_02881C_CULL_DIST_ENA_1(x)                  (((x) >> 9) & 0x1)
+#define   C_02881C_CULL_DIST_ENA_1                     0xFFFFFDFF
+#define   S_02881C_CULL_DIST_ENA_2(x)                  (((x) & 0x1) << 10)
+#define   G_02881C_CULL_DIST_ENA_2(x)                  (((x) >> 10) & 0x1)
+#define   C_02881C_CULL_DIST_ENA_2                     0xFFFFFBFF
+#define   S_02881C_CULL_DIST_ENA_3(x)                  (((x) & 0x1) << 11)
+#define   G_02881C_CULL_DIST_ENA_3(x)                  (((x) >> 11) & 0x1)
+#define   C_02881C_CULL_DIST_ENA_3                     0xFFFFF7FF
+#define   S_02881C_CULL_DIST_ENA_4(x)                  (((x) & 0x1) << 12)
+#define   G_02881C_CULL_DIST_ENA_4(x)                  (((x) >> 12) & 0x1)
+#define   C_02881C_CULL_DIST_ENA_4                     0xFFFFEFFF
+#define   S_02881C_CULL_DIST_ENA_5(x)                  (((x) & 0x1) << 13)
+#define   G_02881C_CULL_DIST_ENA_5(x)                  (((x) >> 13) & 0x1)
+#define   C_02881C_CULL_DIST_ENA_5                     0xFFFFDFFF
+#define   S_02881C_CULL_DIST_ENA_6(x)                  (((x) & 0x1) << 14)
+#define   G_02881C_CULL_DIST_ENA_6(x)                  (((x) >> 14) & 0x1)
+#define   C_02881C_CULL_DIST_ENA_6                     0xFFFFBFFF
+#define   S_02881C_CULL_DIST_ENA_7(x)                  (((x) & 0x1) << 15)
+#define   G_02881C_CULL_DIST_ENA_7(x)                  (((x) >> 15) & 0x1)
+#define   C_02881C_CULL_DIST_ENA_7                     0xFFFF7FFF
+#define   S_02881C_USE_VTX_POINT_SIZE(x)               (((x) & 0x1) << 16)
+#define   G_02881C_USE_VTX_POINT_SIZE(x)               (((x) >> 16) & 0x1)
+#define   C_02881C_USE_VTX_POINT_SIZE                  0xFFFEFFFF
+#define   S_02881C_USE_VTX_EDGE_FLAG(x)                (((x) & 0x1) << 17)
+#define   G_02881C_USE_VTX_EDGE_FLAG(x)                (((x) >> 17) & 0x1)
+#define   C_02881C_USE_VTX_EDGE_FLAG                   0xFFFDFFFF
+#define   S_02881C_USE_VTX_RENDER_TARGET_INDX(x)       (((x) & 0x1) << 18)
+#define   G_02881C_USE_VTX_RENDER_TARGET_INDX(x)       (((x) >> 18) & 0x1)
+#define   C_02881C_USE_VTX_RENDER_TARGET_INDX          0xFFFBFFFF
+#define   S_02881C_USE_VTX_VIEWPORT_INDX(x)            (((x) & 0x1) << 19)
+#define   G_02881C_USE_VTX_VIEWPORT_INDX(x)            (((x) >> 19) & 0x1)
+#define   C_02881C_USE_VTX_VIEWPORT_INDX               0xFFF7FFFF
+#define   S_02881C_USE_VTX_KILL_FLAG(x)                (((x) & 0x1) << 20)
+#define   G_02881C_USE_VTX_KILL_FLAG(x)                (((x) >> 20) & 0x1)
+#define   C_02881C_USE_VTX_KILL_FLAG                   0xFFEFFFFF
+#define   S_02881C_VS_OUT_MISC_VEC_ENA(x)              (((x) & 0x1) << 21)
+#define   G_02881C_VS_OUT_MISC_VEC_ENA(x)              (((x) >> 21) & 0x1)
+#define   C_02881C_VS_OUT_MISC_VEC_ENA                 0xFFDFFFFF
+#define   S_02881C_VS_OUT_CCDIST0_VEC_ENA(x)           (((x) & 0x1) << 22)
+#define   G_02881C_VS_OUT_CCDIST0_VEC_ENA(x)           (((x) >> 22) & 0x1)
+#define   C_02881C_VS_OUT_CCDIST0_VEC_ENA              0xFFBFFFFF
+#define   S_02881C_VS_OUT_CCDIST1_VEC_ENA(x)           (((x) & 0x1) << 23)
+#define   G_02881C_VS_OUT_CCDIST1_VEC_ENA(x)           (((x) >> 23) & 0x1)
+#define   C_02881C_VS_OUT_CCDIST1_VEC_ENA              0xFF7FFFFF
 #define R_028868_SQ_PGM_RESOURCES_VS                 0x028868
 #define   S_028868_NUM_GPRS(x)                         (((x) & 0xFF) << 0)
 #define   G_028868_NUM_GPRS(x)                         (((x) >> 0) & 0xFF)
diff --git a/src/gallium/drivers/r600/r700_asm.c b/src/gallium/drivers/r600/r700_asm.c
new file mode 100644
index 00000000000..1ebe20d6ab0
--- /dev/null
+++ b/src/gallium/drivers/r600/r700_asm.c
@@ -0,0 +1,69 @@
+/*
+ * Copyright 2010 Jerome Glisse <[email protected]>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+#include "r600_asm.h"
+#include "r600_context.h"
+#include "util/u_memory.h"
+#include "r700_sq.h"
+#include <stdio.h>
+
+int r700_bc_alu_build(struct r600_bc *bc, struct r600_bc_alu *alu, unsigned id)
+{
+	unsigned i;
+
+	/* don't replace gpr by pv or ps for destination register */
+	if (alu->is_op3) {
+		bc->bytecode[id++] = S_SQ_ALU_WORD0_SRC0_SEL(alu->src[0].sel) |
+					S_SQ_ALU_WORD0_SRC0_CHAN(alu->src[0].chan) |
+					S_SQ_ALU_WORD0_SRC1_SEL(alu->src[1].sel) |
+					S_SQ_ALU_WORD0_SRC1_CHAN(alu->src[1].chan) |
+					S_SQ_ALU_WORD0_LAST(alu->last);
+		bc->bytecode[id++] = S_SQ_ALU_WORD1_DST_GPR(alu->dst.sel) |
+					S_SQ_ALU_WORD1_DST_CHAN(alu->dst.chan) |
+					S_SQ_ALU_WORD1_OP3_SRC2_SEL(alu->src[2].sel) |
+					S_SQ_ALU_WORD1_OP3_SRC2_CHAN(alu->src[2].chan) |
+					S_SQ_ALU_WORD1_OP3_SRC2_NEG(alu->src[2].neg) |
+					S_SQ_ALU_WORD1_OP3_ALU_INST(alu->inst) |
+					S_SQ_ALU_WORD1_BANK_SWIZZLE(0);
+	} else {
+		bc->bytecode[id++] = S_SQ_ALU_WORD0_SRC0_SEL(alu->src[0].sel) |
+					S_SQ_ALU_WORD0_SRC0_CHAN(alu->src[0].chan) |
+					S_SQ_ALU_WORD0_SRC0_NEG(alu->src[0].neg) |
+					S_SQ_ALU_WORD0_SRC1_SEL(alu->src[1].sel) |
+					S_SQ_ALU_WORD0_SRC1_CHAN(alu->src[1].chan) |
+					S_SQ_ALU_WORD0_SRC1_NEG(alu->src[1].neg) |
+					S_SQ_ALU_WORD0_LAST(alu->last);
+		bc->bytecode[id++] = S_SQ_ALU_WORD1_DST_GPR(alu->dst.sel) |
+					S_SQ_ALU_WORD1_DST_CHAN(alu->dst.chan) |
+					S_SQ_ALU_WORD1_OP2_SRC0_ABS(alu->src[0].abs) |
+					S_SQ_ALU_WORD1_OP2_SRC1_ABS(alu->src[1].abs) |
+					S_SQ_ALU_WORD1_OP2_WRITE_MASK(alu->dst.write) |
+					S_SQ_ALU_WORD1_OP2_ALU_INST(alu->inst) |
+					S_SQ_ALU_WORD1_BANK_SWIZZLE(0);
+	}
+	if (alu->last) {
+		for (i = 0; i < alu->nliteral; i++) {
+			bc->bytecode[id++] = alu->value[i];
+		}
+	}
+	return 0;
+}
diff --git a/src/gallium/drivers/r600/radeon.h b/src/gallium/drivers/r600/radeon.h
index 3a8405f9b40..8f00a4895a0 100644
--- a/src/gallium/drivers/r600/radeon.h
+++ b/src/gallium/drivers/r600/radeon.h
@@ -157,11 +157,42 @@ int radeon_ctx_submit(struct radeon_ctx *ctx);
 void radeon_ctx_dump_bof(struct radeon_ctx *ctx, const char *file);
 
 /*
+ * radeon context functions
+ */
+#pragma pack(1)
+struct radeon_cs_reloc {
+	uint32_t	handle;
+	uint32_t	read_domain;
+	uint32_t	write_domain;
+	uint32_t	flags;
+};
+#pragma pack()
+
+struct radeon_ctx {
+	int				refcount;
+	struct radeon			*radeon;
+	u32				*pm4;
+	u32				cpm4;
+	u32				draw_cpm4;
+	unsigned			id;
+	unsigned			next_id;
+	unsigned			nreloc;
+	struct radeon_cs_reloc		*reloc;
+	unsigned			nbo;
+	struct radeon_bo		**bo;
+	unsigned			ndraw;
+	struct radeon_draw		*cdraw;
+	struct radeon_draw		**draw;
+	unsigned			nstate;
+	struct radeon_state		**state;
+};
+
+/*
  * R600/R700
  */
 
-#define R600_NSTATE				1273
-#define R600_NTYPE				25
+#define R600_NSTATE				1280
+#define R600_NTYPE				32
 
 #define R600_CONFIG				0
 #define R600_CONFIG_TYPE				0
@@ -207,12 +238,26 @@ void radeon_ctx_dump_bof(struct radeon_ctx *ctx, const char *file);
 #define R600_GS_SAMPLER_BORDER_TYPE				20
 #define R600_CB0				1269
 #define R600_CB0_TYPE				21
-#define R600_DB				1270
-#define R600_DB_TYPE				22
-#define R600_VGT				1271
-#define R600_VGT_TYPE				23
-#define R600_DRAW				1272
-#define R600_DRAW_TYPE				24
+#define R600_CB1				1270
+#define R600_CB1_TYPE				22
+#define R600_CB2				1271
+#define R600_CB2_TYPE				23
+#define R600_CB3				1272
+#define R600_CB3_TYPE				24
+#define R600_CB4				1273
+#define R600_CB4_TYPE				25
+#define R600_CB5				1274
+#define R600_CB5_TYPE				26
+#define R600_CB6				1275
+#define R600_CB6_TYPE				27
+#define R600_CB7				1276
+#define R600_CB7_TYPE				28
+#define R600_DB				1277
+#define R600_DB_TYPE				29
+#define R600_VGT				1278
+#define R600_VGT_TYPE				30
+#define R600_DRAW				1279
+#define R600_DRAW_TYPE				31
 /* R600_CONFIG */
 #define R600_CONFIG__SQ_CONFIG			0
 #define R600_CONFIG__SQ_GPR_RESOURCE_MGMT_1			1
diff --git a/src/gallium/drivers/rbug/rbug_context.c b/src/gallium/drivers/rbug/rbug_context.c
index e0dd5cf8c2b..3ffda87520f 100644
--- a/src/gallium/drivers/rbug/rbug_context.c
+++ b/src/gallium/drivers/rbug/rbug_context.c
@@ -103,10 +103,7 @@ rbug_draw_block_locked(struct rbug_context *rb_pipe, int flag)
 }
 
 static void
-rbug_draw_arrays(struct pipe_context *_pipe,
-                 unsigned prim,
-                 unsigned start,
-                 unsigned count)
+rbug_draw_vbo(struct pipe_context *_pipe, const struct pipe_draw_info *info)
 {
    struct rbug_context *rb_pipe = rbug_context(_pipe);
    struct pipe_context *pipe = rb_pipe->pipe;
@@ -114,72 +111,7 @@ rbug_draw_arrays(struct pipe_context *_pipe,
    pipe_mutex_lock(rb_pipe->draw_mutex);
    rbug_draw_block_locked(rb_pipe, RBUG_BLOCK_BEFORE);
 
-   pipe->draw_arrays(pipe,
-                     prim,
-                     start,
-                     count);
-
-   rbug_draw_block_locked(rb_pipe, RBUG_BLOCK_AFTER);
-   pipe_mutex_unlock(rb_pipe->draw_mutex);
-}
-
-static void
-rbug_draw_elements(struct pipe_context *_pipe,
-                   struct pipe_resource *_indexResource,
-                   unsigned indexSize,
-                   int indexBias,
-                   unsigned prim,
-                   unsigned start,
-                   unsigned count)
-{
-   struct rbug_context *rb_pipe = rbug_context(_pipe);
-   struct rbug_resource *rb_resource = rbug_resource(_indexResource);
-   struct pipe_context *pipe = rb_pipe->pipe;
-   struct pipe_resource *indexResource = rb_resource->resource;
-
-   pipe_mutex_lock(rb_pipe->draw_mutex);
-   rbug_draw_block_locked(rb_pipe, RBUG_BLOCK_BEFORE);
-
-   pipe->draw_elements(pipe,
-                       indexResource,
-                       indexSize,
-                       indexBias,
-                       prim,
-                       start,
-                       count);
-
-   rbug_draw_block_locked(rb_pipe, RBUG_BLOCK_AFTER);
-   pipe_mutex_unlock(rb_pipe->draw_mutex);
-}
-
-static void
-rbug_draw_range_elements(struct pipe_context *_pipe,
-                         struct pipe_resource *_indexResource,
-                         unsigned indexSize,
-                         int indexBias,
-                         unsigned minIndex,
-                         unsigned maxIndex,
-                         unsigned mode,
-                         unsigned start,
-                         unsigned count)
-{
-   struct rbug_context *rb_pipe = rbug_context(_pipe);
-   struct rbug_resource *rb_resource = rbug_resource(_indexResource);
-   struct pipe_context *pipe = rb_pipe->pipe;
-   struct pipe_resource *indexResource = rb_resource->resource;
-
-   pipe_mutex_lock(rb_pipe->draw_mutex);
-   rbug_draw_block_locked(rb_pipe, RBUG_BLOCK_BEFORE);
-
-   pipe->draw_range_elements(pipe,
-                             indexResource,
-                             indexSize,
-                             indexBias,
-                             minIndex,
-                             maxIndex,
-                             mode,
-                             start,
-                             count);
+   pipe->draw_vbo(pipe, info);
 
    rbug_draw_block_locked(rb_pipe, RBUG_BLOCK_AFTER);
    pipe_mutex_unlock(rb_pipe->draw_mutex);
@@ -745,6 +677,23 @@ rbug_set_vertex_buffers(struct pipe_context *_pipe,
 }
 
 static void
+rbug_set_index_buffer(struct pipe_context *_pipe,
+                      const struct pipe_index_buffer *_ib)
+{
+   struct rbug_context *rb_pipe = rbug_context(_pipe);
+   struct pipe_context *pipe = rb_pipe->pipe;
+   struct pipe_index_buffer unwrapped_ib, *ib = NULL;
+
+   if (_ib) {
+      unwrapped_ib = *_ib;
+      unwrapped_ib.buffer = rbug_resource_unwrap(_ib->buffer);
+      ib = &unwrapped_ib;
+   }
+
+   pipe->set_index_buffer(pipe, ib);
+}
+
+static void
 rbug_set_sample_mask(struct pipe_context *_pipe,
                      unsigned sample_mask)
 {
@@ -1040,9 +989,7 @@ rbug_context_create(struct pipe_screen *_screen, struct pipe_context *pipe)
    rb_pipe->base.draw = NULL;
 
    rb_pipe->base.destroy = rbug_destroy;
-   rb_pipe->base.draw_arrays = rbug_draw_arrays;
-   rb_pipe->base.draw_elements = rbug_draw_elements;
-   rb_pipe->base.draw_range_elements = rbug_draw_range_elements;
+   rb_pipe->base.draw_vbo = rbug_draw_vbo;
    rb_pipe->base.create_query = rbug_create_query;
    rb_pipe->base.destroy_query = rbug_destroy_query;
    rb_pipe->base.begin_query = rbug_begin_query;
@@ -1084,6 +1031,7 @@ rbug_context_create(struct pipe_screen *_screen, struct pipe_context *pipe)
    rb_pipe->base.set_fragment_sampler_views = rbug_set_fragment_sampler_views;
    rb_pipe->base.set_vertex_sampler_views = rbug_set_vertex_sampler_views;
    rb_pipe->base.set_vertex_buffers = rbug_set_vertex_buffers;
+   rb_pipe->base.set_index_buffer = rbug_set_index_buffer;
    rb_pipe->base.set_sample_mask = rbug_set_sample_mask;
    rb_pipe->base.resource_copy_region = rbug_resource_copy_region;
    rb_pipe->base.clear = rbug_clear;
diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c
index 12ef98aac75..a7c9959b3e1 100644
--- a/src/gallium/drivers/softpipe/sp_context.c
+++ b/src/gallium/drivers/softpipe/sp_context.c
@@ -282,12 +282,9 @@ softpipe_create_context( struct pipe_screen *screen,
    softpipe->pipe.set_viewport_state = softpipe_set_viewport_state;
    softpipe->pipe.set_stream_output_buffers = softpipe_set_stream_output_buffers;
    softpipe->pipe.set_vertex_buffers = softpipe_set_vertex_buffers;
+   softpipe->pipe.set_index_buffer = softpipe_set_index_buffer;
 
-   softpipe->pipe.draw_arrays = softpipe_draw_arrays;
-   softpipe->pipe.draw_elements = softpipe_draw_elements;
-   softpipe->pipe.draw_range_elements = softpipe_draw_range_elements;
-   softpipe->pipe.draw_arrays_instanced = softpipe_draw_arrays_instanced;
-   softpipe->pipe.draw_elements_instanced = softpipe_draw_elements_instanced;
+   softpipe->pipe.draw_vbo = softpipe_draw_vbo;
    softpipe->pipe.draw_stream_output = softpipe_draw_stream_output;
 
    softpipe->pipe.clear = softpipe_clear;
diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h
index 53115a827d0..9361a3df09e 100644
--- a/src/gallium/drivers/softpipe/sp_context.h
+++ b/src/gallium/drivers/softpipe/sp_context.h
@@ -82,6 +82,7 @@ struct softpipe_context {
    struct pipe_sampler_view *geometry_sampler_views[PIPE_MAX_GEOMETRY_SAMPLERS];
    struct pipe_viewport_state viewport;
    struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS];
+   struct pipe_index_buffer index_buffer;
    struct {
       struct softpipe_resource *buffer[PIPE_MAX_SO_BUFFERS];
       int offset[PIPE_MAX_SO_BUFFERS];
@@ -111,6 +112,7 @@ struct softpipe_context {
 
    /** Mapped constant buffers */
    const void *mapped_constants[PIPE_SHADER_TYPES][PIPE_MAX_CONSTANT_BUFFERS];
+   unsigned const_buffer_size[PIPE_SHADER_TYPES][PIPE_MAX_CONSTANT_BUFFERS];
 
    /** Vertex format */
    struct vertex_info vertex_info;
diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c
index 9e727c93811..386c8acb8ce 100644
--- a/src/gallium/drivers/softpipe/sp_draw_arrays.c
+++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c
@@ -111,27 +111,19 @@ softpipe_draw_stream_output(struct pipe_context *pipe, unsigned mode)
  * When the min/max element indexes aren't known, minIndex should be 0
  * and maxIndex should be ~0.
  */
-static void
-softpipe_draw_range_elements_instanced(struct pipe_context *pipe,
-                                       struct pipe_resource *indexBuffer,
-                                       unsigned indexSize,
-                                       int indexBias,
-                                       unsigned minIndex,
-                                       unsigned maxIndex,
-                                       unsigned mode,
-                                       unsigned start,
-                                       unsigned count,
-                                       unsigned startInstance,
-                                       unsigned instanceCount)
+void
+softpipe_draw_vbo(struct pipe_context *pipe,
+                  const struct pipe_draw_info *info)
 {
    struct softpipe_context *sp = softpipe_context(pipe);
    struct draw_context *draw = sp->draw;
+   void *mapped_indices = NULL;
    unsigned i;
 
    if (!softpipe_check_render_cond(sp))
       return;
 
-   sp->reduced_api_prim = u_reduced_prim(mode);
+   sp->reduced_api_prim = u_reduced_prim(info->mode);
 
    if (sp->dirty) {
       softpipe_update_derived(sp);
@@ -146,31 +138,27 @@ softpipe_draw_range_elements_instanced(struct pipe_context *pipe,
    }
 
    /* Map index buffer, if present */
-   if (indexBuffer) {
-      void *mapped_indexes = softpipe_resource(indexBuffer)->data;
-      draw_set_mapped_element_buffer_range(draw,
-                                           indexSize,
-                                           indexBias,
-                                           minIndex,
-                                           maxIndex,
-                                           mapped_indexes);
-   } else {
-      /* no index/element buffer */
-      draw_set_mapped_element_buffer_range(draw,
-                                           0, 0,
-                                           start,
-                                           start + count - 1,
-                                           NULL);
+   if (info->indexed && sp->index_buffer.buffer) {
+      char *indices = (char *) softpipe_resource(sp->index_buffer.buffer)->data;
+      mapped_indices = (void *) (indices + sp->index_buffer.offset);
    }
 
+   draw_set_mapped_element_buffer_range(draw, (mapped_indices) ?
+                                        sp->index_buffer.index_size : 0,
+                                        info->index_bias,
+                                        info->min_index,
+                                        info->max_index,
+                                        mapped_indices);
+
    /* draw! */
-   draw_arrays_instanced(draw, mode, start, count, startInstance, instanceCount);
+   draw_arrays_instanced(draw, info->mode, info->start, info->count,
+         info->start_instance, info->instance_count);
 
    /* unmap vertex/index buffers - will cause draw module to flush */
    for (i = 0; i < sp->num_vertex_buffers; i++) {
       draw_set_mapped_vertex_buffer(draw, i, NULL);
    }
-   if (indexBuffer) {
+   if (mapped_indices) {
       draw_set_mapped_element_buffer(draw, 0, 0, NULL);
    }
 
@@ -184,109 +172,3 @@ softpipe_draw_range_elements_instanced(struct pipe_context *pipe,
    /* Note: leave drawing surfaces mapped */
    sp->dirty_render_cache = TRUE;
 }
-
-
-void
-softpipe_draw_range_elements(struct pipe_context *pipe,
-                             struct pipe_resource *indexBuffer,
-                             unsigned indexSize,
-                             int indexBias,
-                             unsigned min_index,
-                             unsigned max_index,
-                             unsigned mode, unsigned start, unsigned count)
-{
-   softpipe_draw_range_elements_instanced(pipe,
-                                          indexBuffer,
-                                          indexSize,
-                                          indexBias,
-                                          min_index,
-                                          max_index,
-                                          mode,
-                                          start,
-                                          count,
-                                          0,
-                                          1);
-}
-
-
-void
-softpipe_draw_elements(struct pipe_context *pipe,
-                       struct pipe_resource *indexBuffer,
-                       unsigned indexSize, int indexBias,
-                       unsigned mode, unsigned start, unsigned count)
-{
-   softpipe_draw_range_elements_instanced(pipe,
-                                          indexBuffer,
-                                          indexSize,
-                                          indexBias,
-                                          0,
-                                          0xffffffff,
-                                          mode,
-                                          start,
-                                          count,
-                                          0,
-                                          1);
-}
-
-void
-softpipe_draw_arrays_instanced(struct pipe_context *pipe,
-                               unsigned mode,
-                               unsigned start,
-                               unsigned count,
-                               unsigned startInstance,
-                               unsigned instanceCount)
-{
-   softpipe_draw_range_elements_instanced(pipe,
-                                          NULL,
-                                          0,
-                                          0,
-                                          0,
-                                          0xffffffff,
-                                          mode,
-                                          start,
-                                          count,
-                                          startInstance,
-                                          instanceCount);
-}
-
-void
-softpipe_draw_elements_instanced(struct pipe_context *pipe,
-                                 struct pipe_resource *indexBuffer,
-                                 unsigned indexSize,
-                                 int indexBias,
-                                 unsigned mode,
-                                 unsigned start,
-                                 unsigned count,
-                                 unsigned startInstance,
-                                 unsigned instanceCount)
-{
-   softpipe_draw_range_elements_instanced(pipe,
-                                          indexBuffer,
-                                          indexSize,
-                                          indexBias,
-                                          0,
-                                          0xffffffff,
-                                          mode,
-                                          start,
-                                          count,
-                                          startInstance,
-                                          instanceCount);
-}
-
-void
-softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode,
-                     unsigned start, unsigned count)
-{
-   softpipe_draw_range_elements_instanced(pipe,
-                                          NULL,
-                                          0,
-                                          0,
-                                          0,
-                                          0xffffffff,
-                                          mode,
-                                          start,
-                                          count,
-                                          0,
-                                          1);
-}
-
diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c
index d240bcbf3bb..90f4787d599 100644
--- a/src/gallium/drivers/softpipe/sp_quad_fs.c
+++ b/src/gallium/drivers/softpipe/sp_quad_fs.c
@@ -111,9 +111,10 @@ shade_quads(struct quad_stage *qs,
    struct tgsi_exec_machine *machine = softpipe->fs_machine;
    unsigned i, nr_quads = 0;
 
-   for (i = 0; i < PIPE_MAX_CONSTANT_BUFFERS; i++) {
-      machine->Consts[i] = softpipe->mapped_constants[PIPE_SHADER_FRAGMENT][i];
-   }
+   tgsi_exec_set_constant_buffers(machine, PIPE_MAX_CONSTANT_BUFFERS,
+                         softpipe->mapped_constants[PIPE_SHADER_FRAGMENT],
+                         softpipe->const_buffer_size[PIPE_SHADER_FRAGMENT]);
+
    machine->InterpCoefs = quads[0]->coef;
 
    for (i = 0; i < nr; i++) {
diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h
index 7d6b86dce04..39d204de8a9 100644
--- a/src/gallium/drivers/softpipe/sp_state.h
+++ b/src/gallium/drivers/softpipe/sp_state.h
@@ -221,44 +221,16 @@ void softpipe_set_vertex_buffers(struct pipe_context *,
                                  unsigned count,
                                  const struct pipe_vertex_buffer *);
 
-
-void softpipe_update_derived( struct softpipe_context *softpipe );
+void softpipe_set_index_buffer(struct pipe_context *,
+                               const struct pipe_index_buffer *);
 
 
-void softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode,
-                          unsigned start, unsigned count);
-
-void softpipe_draw_elements(struct pipe_context *pipe,
-                            struct pipe_resource *indexBuffer,
-                            unsigned indexSize, int indexBias,
-                            unsigned mode, unsigned start, unsigned count);
-void
-softpipe_draw_range_elements(struct pipe_context *pipe,
-                             struct pipe_resource *indexBuffer,
-                             unsigned indexSize,
-                             int indexBias,
-                             unsigned min_index,
-                             unsigned max_index,
-                             unsigned mode, unsigned start, unsigned count);
+void softpipe_update_derived( struct softpipe_context *softpipe );
 
-void
-softpipe_draw_arrays_instanced(struct pipe_context *pipe,
-                               unsigned mode,
-                               unsigned start,
-                               unsigned count,
-                               unsigned startInstance,
-                               unsigned instanceCount);
 
 void
-softpipe_draw_elements_instanced(struct pipe_context *pipe,
-                                 struct pipe_resource *indexBuffer,
-                                 unsigned indexSize,
-                                 int indexBias,
-                                 unsigned mode,
-                                 unsigned start,
-                                 unsigned count,
-                                 unsigned startInstance,
-                                 unsigned instanceCount);
+softpipe_draw_vbo(struct pipe_context *pipe,
+                  const struct pipe_draw_info *info);
 
 void softpipe_draw_stream_output(struct pipe_context *pipe, unsigned mode);
 
diff --git a/src/gallium/drivers/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c
index 3fbf1f25781..ded242d3dc5 100644
--- a/src/gallium/drivers/softpipe/sp_state_fs.c
+++ b/src/gallium/drivers/softpipe/sp_state_fs.c
@@ -195,6 +195,8 @@ softpipe_set_constant_buffer(struct pipe_context *pipe,
    }
 
    softpipe->mapped_constants[shader][index] = data;
+   softpipe->const_buffer_size[shader][index] = size;
+
    softpipe->dirty |= SP_NEW_CONSTANTS;
 }
 
diff --git a/src/gallium/drivers/softpipe/sp_state_vertex.c b/src/gallium/drivers/softpipe/sp_state_vertex.c
index 462f4d2655e..880a7c7cd26 100644
--- a/src/gallium/drivers/softpipe/sp_state_vertex.c
+++ b/src/gallium/drivers/softpipe/sp_state_vertex.c
@@ -88,3 +88,17 @@ softpipe_set_vertex_buffers(struct pipe_context *pipe,
 
    draw_set_vertex_buffers(softpipe->draw, count, buffers);
 }
+
+void
+softpipe_set_index_buffer(struct pipe_context *pipe,
+                          const struct pipe_index_buffer *ib)
+{
+   struct softpipe_context *softpipe = softpipe_context(pipe);
+
+   if (ib)
+      memcpy(&softpipe->index_buffer, ib, sizeof(softpipe->index_buffer));
+   else
+      memset(&softpipe->index_buffer, 0, sizeof(softpipe->index_buffer));
+
+   /* TODO make this more like a state */
+}
diff --git a/src/gallium/drivers/svga/svga_context.h b/src/gallium/drivers/svga/svga_context.h
index 9a46de643fd..67a7614c8af 100644
--- a/src/gallium/drivers/svga/svga_context.h
+++ b/src/gallium/drivers/svga/svga_context.h
@@ -190,6 +190,7 @@ struct svga_state
    struct svga_vertex_shader *vs;
 
    struct pipe_vertex_buffer vb[PIPE_MAX_ATTRIBS];
+   struct pipe_index_buffer ib;
    struct pipe_resource *cb[PIPE_SHADER_TYPES];
 
    struct pipe_framebuffer_state framebuffer;
diff --git a/src/gallium/drivers/svga/svga_pipe_draw.c b/src/gallium/drivers/svga/svga_pipe_draw.c
index 58e930d9835..de08bc5e562 100644
--- a/src/gallium/drivers/svga/svga_pipe_draw.c
+++ b/src/gallium/drivers/svga/svga_pipe_draw.c
@@ -227,31 +227,30 @@ svga_draw_range_elements( struct pipe_context *pipe,
 
 
 static void
-svga_draw_elements( struct pipe_context *pipe,
-                    struct pipe_resource *index_buffer,
-                    unsigned index_size, int index_bias,
-                    unsigned prim, unsigned start, unsigned count)
+svga_draw_vbo(struct pipe_context *pipe, const struct pipe_draw_info *info)
 {
-   svga_draw_range_elements( pipe, index_buffer,
-                             index_size, index_bias,
-                             0, 0xffffffff,
-                             prim, start, count );
-}
+   struct svga_context *svga = svga_context(pipe);
 
-static void
-svga_draw_arrays( struct pipe_context *pipe,
-                  unsigned prim, unsigned start, unsigned count)
-{
-   svga_draw_range_elements(pipe, NULL, 0, 0,
-                            start, start + count - 1, 
-                            prim, 
-                            start, count);
+   if (info->indexed && svga->curr.ib.buffer) {
+      unsigned offset;
+
+      assert(svga->curr.ib.offset % svga->curr.ib.index_size == 0);
+      offset = svga->curr.ib.offset / svga->curr.ib.index_size;
+
+      svga_draw_range_elements(pipe, svga->curr.ib.buffer,
+                               svga->curr.ib.index_size, info->index_bias,
+                               info->min_index, info->max_index,
+                               info->mode, info->start + offset, info->count);
+   }
+   else {
+      svga_draw_range_elements(pipe, NULL, 0, 0,
+                               info->min_index, info->max_index,
+                               info->mode, info->start, info->count);
+   }
 }
 
 
 void svga_init_draw_functions( struct svga_context *svga )
 {
-   svga->pipe.draw_arrays = svga_draw_arrays;
-   svga->pipe.draw_elements = svga_draw_elements;
-   svga->pipe.draw_range_elements = svga_draw_range_elements;
+   svga->pipe.draw_vbo = svga_draw_vbo;
 }
diff --git a/src/gallium/drivers/svga/svga_pipe_vertex.c b/src/gallium/drivers/svga/svga_pipe_vertex.c
index 23808ad08e0..86c79459f3e 100644
--- a/src/gallium/drivers/svga/svga_pipe_vertex.c
+++ b/src/gallium/drivers/svga/svga_pipe_vertex.c
@@ -66,6 +66,24 @@ static void svga_set_vertex_buffers(struct pipe_context *pipe,
 }
 
 
+static void svga_set_index_buffer(struct pipe_context *pipe,
+                                  const struct pipe_index_buffer *ib)
+{
+   struct svga_context *svga = svga_context(pipe);
+
+   if (ib) {
+      pipe_resource_reference(&svga->curr.ib.buffer, ib->buffer);
+      memcpy(&svga->curr.ib, ib, sizeof(svga->curr.ib));
+   }
+   else {
+      pipe_resource_reference(&svga->curr.ib.buffer, NULL);
+      memset(&svga->curr.ib, 0, sizeof(svga->curr.ib));
+   }
+
+   /* TODO make this more like a state */
+}
+
+
 static void *
 svga_create_vertex_elements_state(struct pipe_context *pipe,
                                   unsigned count,
@@ -109,6 +127,7 @@ void svga_cleanup_vertex_state( struct svga_context *svga )
 void svga_init_vertex_functions( struct svga_context *svga )
 {
    svga->pipe.set_vertex_buffers = svga_set_vertex_buffers;
+   svga->pipe.set_index_buffer = svga_set_index_buffer;
    svga->pipe.create_vertex_elements_state = svga_create_vertex_elements_state;
    svga->pipe.bind_vertex_elements_state = svga_bind_vertex_elements_state;
    svga->pipe.delete_vertex_elements_state = svga_delete_vertex_elements_state;
diff --git a/src/gallium/drivers/trace/tr_context.c b/src/gallium/drivers/trace/tr_context.c
index 55dd6cf8837..84e5a6a8242 100644
--- a/src/gallium/drivers/trace/tr_context.c
+++ b/src/gallium/drivers/trace/tr_context.c
@@ -83,85 +83,26 @@ trace_surface_unwrap(struct trace_context *tr_ctx,
 
 
 static INLINE void
-trace_context_draw_arrays(struct pipe_context *_pipe,
-                          unsigned mode, unsigned start, unsigned count)
+trace_context_draw_vbo(struct pipe_context *_pipe,
+                       const struct pipe_draw_info *info)
 {
    struct trace_context *tr_ctx = trace_context(_pipe);
    struct pipe_context *pipe = tr_ctx->pipe;
 
-   trace_dump_call_begin("pipe_context", "draw_arrays");
+   trace_dump_call_begin("pipe_context", "draw_vbo");
 
-   trace_dump_arg(ptr, pipe);
-   trace_dump_arg(uint, mode);
-   trace_dump_arg(uint, start);
-   trace_dump_arg(uint, count);
-
-   pipe->draw_arrays(pipe, mode, start, count);
-
-   trace_dump_call_end();
-}
-
-
-static INLINE void
-trace_context_draw_elements(struct pipe_context *_pipe,
-                            struct pipe_resource *_indexBuffer,
-                            unsigned indexSize, int indexBias,
-                            unsigned mode, unsigned start, unsigned count)
-{
-   struct trace_context *tr_ctx = trace_context(_pipe);
-   struct trace_resource *tr_buf = trace_resource(_indexBuffer);
-   struct pipe_context *pipe = tr_ctx->pipe;
-   struct pipe_resource *indexBuffer = tr_buf->resource;
-
-   trace_dump_call_begin("pipe_context", "draw_elements");
-
-   trace_dump_arg(ptr, pipe);
-   trace_dump_arg(ptr, indexBuffer);
-   trace_dump_arg(uint, indexSize);
-   trace_dump_arg(int, indexBias);
-   trace_dump_arg(uint, mode);
-   trace_dump_arg(uint, start);
-   trace_dump_arg(uint, count);
-
-   pipe->draw_elements(pipe, indexBuffer, indexSize, indexBias,
-                       mode, start, count);
-
-   trace_dump_call_end();
-}
-
-
-static INLINE void
-trace_context_draw_range_elements(struct pipe_context *_pipe,
-                                  struct pipe_resource *_indexBuffer,
-                                  unsigned indexSize,
-                                  int indexBias,
-                                  unsigned minIndex,
-                                  unsigned maxIndex,
-                                  unsigned mode,
-                                  unsigned start,
-                                  unsigned count)
-{
-   struct trace_context *tr_ctx = trace_context(_pipe);
-   struct trace_resource *tr_buf = trace_resource(_indexBuffer);
-   struct pipe_context *pipe = tr_ctx->pipe;
-   struct pipe_resource *indexBuffer = tr_buf->resource;
-
-   trace_dump_call_begin("pipe_context", "draw_range_elements");
+   trace_dump_arg(ptr,  pipe);
+   trace_dump_arg(bool, info->indexed);
+   trace_dump_arg(uint, info->mode);
+   trace_dump_arg(uint, info->start);
+   trace_dump_arg(uint, info->count);
+   trace_dump_arg(uint, info->start_instance);
+   trace_dump_arg(uint, info->instance_count);
+   trace_dump_arg(int,  info->index_bias);
+   trace_dump_arg(uint, info->min_index);
+   trace_dump_arg(uint, info->max_index);
 
-   trace_dump_arg(ptr, pipe);
-   trace_dump_arg(ptr, indexBuffer);
-   trace_dump_arg(uint, indexSize);
-   trace_dump_arg(int, indexBias);
-   trace_dump_arg(uint, minIndex);
-   trace_dump_arg(uint, maxIndex);
-   trace_dump_arg(uint, mode);
-   trace_dump_arg(uint, start);
-   trace_dump_arg(uint, count);
-
-   pipe->draw_range_elements(pipe,
-                             indexBuffer, indexSize, indexBias,
-                             minIndex, maxIndex,
-                             mode, start, count);
+   pipe->draw_vbo(pipe, info);
 
    trace_dump_call_end();
 }
@@ -1045,6 +986,30 @@ trace_context_set_vertex_buffers(struct pipe_context *_pipe,
 
 
 static INLINE void
+trace_context_set_index_buffer(struct pipe_context *_pipe,
+                               const struct pipe_index_buffer *_ib)
+{
+   struct trace_context *tr_ctx = trace_context(_pipe);
+   struct pipe_context *pipe = tr_ctx->pipe;
+   struct pipe_index_buffer unwrapped_ib, *ib = NULL;
+
+   if (_ib) {
+      unwrapped_ib = *_ib;
+      unwrapped_ib.buffer = trace_resource_unwrap(tr_ctx, _ib->buffer);
+      ib = &unwrapped_ib;
+   }
+
+   trace_dump_call_begin("pipe_context", "set_index_buffer");
+
+   trace_dump_arg(ptr, pipe);
+   trace_dump_arg(index_buffer, ib);
+
+   pipe->set_index_buffer(pipe, ib);
+
+   trace_dump_call_end();
+}
+
+static INLINE void
 trace_context_resource_copy_region(struct pipe_context *_pipe,
                                    struct pipe_resource *dst,
                                    struct pipe_subresource subdst,
@@ -1433,9 +1398,7 @@ trace_context_create(struct trace_screen *tr_scr,
    tr_ctx->base.screen = &tr_scr->base;
 
    tr_ctx->base.destroy = trace_context_destroy;
-   tr_ctx->base.draw_arrays = trace_context_draw_arrays;
-   tr_ctx->base.draw_elements = trace_context_draw_elements;
-   tr_ctx->base.draw_range_elements = trace_context_draw_range_elements;
+   tr_ctx->base.draw_vbo = trace_context_draw_vbo;
    tr_ctx->base.create_query = trace_context_create_query;
    tr_ctx->base.destroy_query = trace_context_destroy_query;
    tr_ctx->base.begin_query = trace_context_begin_query;
@@ -1477,6 +1440,7 @@ trace_context_create(struct trace_screen *tr_scr,
    tr_ctx->base.create_sampler_view = trace_create_sampler_view;
    tr_ctx->base.sampler_view_destroy = trace_sampler_view_destroy;
    tr_ctx->base.set_vertex_buffers = trace_context_set_vertex_buffers;
+   tr_ctx->base.set_index_buffer = trace_context_set_index_buffer;
    tr_ctx->base.resource_copy_region = trace_context_resource_copy_region;
    tr_ctx->base.clear = trace_context_clear;
    tr_ctx->base.clear_render_target = trace_context_clear_render_target;
diff --git a/src/gallium/drivers/trace/tr_dump_state.c b/src/gallium/drivers/trace/tr_dump_state.c
index 1727c2a0206..bd9a9bfaf16 100644
--- a/src/gallium/drivers/trace/tr_dump_state.c
+++ b/src/gallium/drivers/trace/tr_dump_state.c
@@ -533,6 +533,26 @@ void trace_dump_vertex_buffer(const struct pipe_vertex_buffer *state)
 }
 
 
+void trace_dump_index_buffer(const struct pipe_index_buffer *state)
+{
+   if (!trace_dumping_enabled_locked())
+      return;
+
+   if(!state) {
+      trace_dump_null();
+      return;
+   }
+
+   trace_dump_struct_begin("pipe_index_buffer");
+
+   trace_dump_member(uint, state, index_size);
+   trace_dump_member(uint, state, offset);
+   trace_dump_member(resource_ptr, state, buffer);
+
+   trace_dump_struct_end();
+}
+
+
 void trace_dump_vertex_element(const struct pipe_vertex_element *state)
 {
    if (!trace_dumping_enabled_locked())
diff --git a/src/gallium/drivers/trace/tr_dump_state.h b/src/gallium/drivers/trace/tr_dump_state.h
index e614e8355e3..2e70f4e1c74 100644
--- a/src/gallium/drivers/trace/tr_dump_state.h
+++ b/src/gallium/drivers/trace/tr_dump_state.h
@@ -75,6 +75,8 @@ void trace_dump_transfer(const struct pipe_transfer *state);
 
 void trace_dump_vertex_buffer(const struct pipe_vertex_buffer *state);
 
+void trace_dump_index_buffer(const struct pipe_index_buffer *state);
+
 void trace_dump_vertex_element(const struct pipe_vertex_element *state);
 
 
diff --git a/src/gallium/include/pipe/p_compiler.h b/src/gallium/include/pipe/p_compiler.h
index 0358c14e24b..1fa3ec8300a 100644
--- a/src/gallium/include/pipe/p_compiler.h
+++ b/src/gallium/include/pipe/p_compiler.h
@@ -102,6 +102,16 @@ typedef unsigned char boolean;
 #  endif
 #endif
 
+/* Forced function inlining */
+#ifndef ALWAYS_INLINE
+#  ifdef __GNUC__
+#    define ALWAYS_INLINE inline __attribute__((always_inline))
+#  elif defined(_MSC_VER)
+#    define ALWAYS_INLINE __forceinline
+#  else
+#    define ALWAYS_INLINE INLINE
+#  endif
+#endif
 
 /* Function visibility */
 #ifndef PUBLIC
diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h
index 7ec3d63a3fd..0579962ec69 100644
--- a/src/gallium/include/pipe/p_context.h
+++ b/src/gallium/include/pipe/p_context.h
@@ -61,46 +61,8 @@ struct pipe_context {
     * VBO drawing
     */
    /*@{*/
-   void (*draw_arrays)( struct pipe_context *pipe,
-                        unsigned mode, unsigned start, unsigned count);
-
-   void (*draw_elements)( struct pipe_context *pipe,
-                          struct pipe_resource *indexBuffer,
-                          unsigned indexSize,
-                          int indexBias,
-                          unsigned mode, unsigned start, unsigned count);
-
-   void (*draw_arrays_instanced)(struct pipe_context *pipe,
-                                 unsigned mode,
-                                 unsigned start,
-                                 unsigned count,
-                                 unsigned startInstance,
-                                 unsigned instanceCount);
-
-   void (*draw_elements_instanced)(struct pipe_context *pipe,
-                                   struct pipe_resource *indexBuffer,
-                                   unsigned indexSize,
-                                   int indexBias,
-                                   unsigned mode,
-                                   unsigned start,
-                                   unsigned count,
-                                   unsigned startInstance,
-                                   unsigned instanceCount);
-
-   /* XXX: this is (probably) a temporary entrypoint, as the range
-    * information should be available from the vertex_buffer state.
-    * Using this to quickly evaluate a specialized path in the draw
-    * module.
-    */
-   void (*draw_range_elements)( struct pipe_context *pipe,
-                                struct pipe_resource *indexBuffer,
-                                unsigned indexSize,
-                                int indexBias,
-                                unsigned minIndex,
-                                unsigned maxIndex,
-                                unsigned mode, 
-                                unsigned start, 
-                                unsigned count);
+   void (*draw_vbo)( struct pipe_context *pipe,
+                     const struct pipe_draw_info *info );
 
    /**
     * Draw the stream output buffer at index 0
@@ -249,6 +211,9 @@ struct pipe_context {
                                unsigned num_buffers,
                                const struct pipe_vertex_buffer * );
 
+   void (*set_index_buffer)( struct pipe_context *pipe,
+                             const struct pipe_index_buffer * );
+
    void (*set_stream_output_buffers)(struct pipe_context *,
                                      struct pipe_resource **buffers,
                                      int *offsets, /*array of offsets
diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h
index 301fe2b74f0..0f1a44cde42 100644
--- a/src/gallium/include/pipe/p_state.h
+++ b/src/gallium/include/pipe/p_state.h
@@ -426,6 +426,41 @@ struct pipe_vertex_element
 };
 
 
+/**
+ * An index buffer.  When an index buffer is bound, all indices to vertices
+ * will be looked up in the buffer.
+ */
+struct pipe_index_buffer
+{
+   unsigned index_size;  /**< size of an index, in bytes */
+   unsigned offset;  /**< offset to start of data in buffer, in bytes */
+   struct pipe_resource *buffer; /**< the actual buffer */
+};
+
+
+/**
+ * Information to describe a draw_vbo call.
+ */
+struct pipe_draw_info
+{
+   boolean indexed;  /**< use index buffer */
+
+   unsigned mode;  /**< the mode of the primitive */
+   unsigned start;  /**< the index of the first vertex */
+   unsigned count;  /**< number of vertices */
+
+   unsigned start_instance; /**< first instance id */
+   unsigned instance_count; /**< number of instances */
+
+   /**
+    * For indexed drawing, these fields apply after index lookup.
+    */
+   int index_bias; /**< a bias to be added to each index */
+   unsigned min_index; /**< the min index */
+   unsigned max_index; /**< the max index */
+};
+
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/src/gallium/state_trackers/dri/common/dri1_helper.c b/src/gallium/state_trackers/dri/common/dri1_helper.c
deleted file mode 100644
index ad6c7d37504..00000000000
--- a/src/gallium/state_trackers/dri/common/dri1_helper.c
+++ /dev/null
@@ -1,129 +0,0 @@
-/**************************************************************************
- *
- * Copyright 2009, VMware, Inc.
- * All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- *
- **************************************************************************/
-/*
- * Management of pipe objects (surface / pipe / fences) used by DRI1 and DRISW.
- *
- * Author: Keith Whitwell <[email protected]>
- * Author: Jakob Bornecrantz <[email protected]>
- */
-
-#include "util/u_inlines.h"
-#include "pipe/p_context.h"
-
-#include "dri_screen.h"
-#include "dri_context.h"
-#include "dri_drawable.h"
-#include "dri1_helper.h"
-
-struct pipe_fence_handle *
-dri1_swap_fences_pop_front(struct dri_drawable *draw)
-{
-   struct pipe_screen *screen = dri_screen(draw->sPriv)->base.screen;
-   struct pipe_fence_handle *fence = NULL;
-
-   if (draw->cur_fences >= draw->desired_fences) {
-      screen->fence_reference(screen, &fence, draw->swap_fences[draw->tail]);
-      screen->fence_reference(screen, &draw->swap_fences[draw->tail++], NULL);
-      --draw->cur_fences;
-      draw->tail &= DRI_SWAP_FENCES_MASK;
-   }
-   return fence;
-}
-
-void
-dri1_swap_fences_push_back(struct dri_drawable *draw,
-                           struct pipe_fence_handle *fence)
-{
-   struct pipe_screen *screen = dri_screen(draw->sPriv)->base.screen;
-
-   if (!fence)
-      return;
-
-   if (draw->cur_fences < DRI_SWAP_FENCES_MAX) {
-      draw->cur_fences++;
-      screen->fence_reference(screen, &draw->swap_fences[draw->head++],
-			      fence);
-      draw->head &= DRI_SWAP_FENCES_MASK;
-   }
-}
-
-void
-dri1_swap_fences_clear(struct dri_drawable *drawable)
-{
-   struct pipe_screen *screen = dri_screen(drawable->sPriv)->base.screen;
-   struct pipe_fence_handle *fence;
-
-   while (drawable->cur_fences) {
-      fence = dri1_swap_fences_pop_front(drawable);
-      screen->fence_reference(screen, &fence, NULL);
-   }
-}
-
-struct pipe_surface *
-dri1_get_pipe_surface(struct dri_drawable *drawable, struct pipe_resource *ptex)
-{
-   struct pipe_screen *pipe_screen = dri_screen(drawable->sPriv)->base.screen;
-   struct pipe_surface *psurf = drawable->dri1_surface;
-
-   if (!psurf || psurf->texture != ptex) {
-      pipe_surface_reference(&drawable->dri1_surface, NULL);
-
-      drawable->dri1_surface = pipe_screen->get_tex_surface(pipe_screen,
-            ptex, 0, 0, 0, 0/* no bind flag???*/);
-
-      psurf = drawable->dri1_surface;
-   }
-
-   return psurf;
-}
-
-void
-dri1_destroy_pipe_surface(struct dri_drawable *drawable)
-{
-   pipe_surface_reference(&drawable->dri1_surface, NULL);
-}
-
-struct pipe_context *
-dri1_get_pipe_context(struct dri_screen *screen)
-{
-   struct pipe_context *pipe = screen->dri1_pipe;
-
-   if (!pipe) {
-      screen->dri1_pipe =
-         screen->base.screen->context_create(screen->base.screen, NULL);
-      pipe = screen->dri1_pipe;
-   }
-
-   return pipe;
-}
-
-void
-dri1_destroy_pipe_context(struct dri_screen *screen)
-{
-   if (screen->dri1_pipe)
-      screen->dri1_pipe->destroy(screen->dri1_pipe);
-}
diff --git a/src/gallium/state_trackers/dri/common/dri1_helper.h b/src/gallium/state_trackers/dri/common/dri1_helper.h
deleted file mode 100644
index c98adf2df22..00000000000
--- a/src/gallium/state_trackers/dri/common/dri1_helper.h
+++ /dev/null
@@ -1,61 +0,0 @@
-/**************************************************************************
- *
- * Copyright 2009, VMware, Inc.
- * All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- *
- **************************************************************************/
-/*
- * Author: Keith Whitwell <[email protected]>
- * Author: Jakob Bornecrantz <[email protected]>
- */
-
-#ifndef DRI1_HELPER_H
-#define DRI1_HELPER_H
-
-#include "dri_screen.h"
-#include "dri_context.h"
-#include "dri_drawable.h"
-
-struct pipe_fence_handle *
-dri1_swap_fences_pop_front(struct dri_drawable *draw);
-
-void
-dri1_swap_fences_push_back(struct dri_drawable *draw,
-                           struct pipe_fence_handle *fence);
-
-void
-dri1_swap_fences_clear(struct dri_drawable *drawable);
-
-struct pipe_surface *
-dri1_get_pipe_surface(struct dri_drawable *drawable, struct pipe_resource *ptex);
-
-void
-dri1_destroy_pipe_surface(struct dri_drawable *drawable);
-
-struct pipe_context *
-dri1_get_pipe_context(struct dri_screen *screen);
-
-void
-dri1_destroy_pipe_context(struct dri_screen *screen);
-
-#endif /* DRI1_HELPER_H */
diff --git a/src/gallium/state_trackers/dri/common/dri_context.h b/src/gallium/state_trackers/dri/common/dri_context.h
index b29e8533837..692c49d7cd5 100644
--- a/src/gallium/state_trackers/dri/common/dri_context.h
+++ b/src/gallium/state_trackers/dri/common/dri_context.h
@@ -60,6 +60,9 @@ struct dri_context
 
    /* gallium */
    struct st_context_iface *st;
+
+   /* hooks filled in by dri2 & drisw */
+   __DRIimage * (*lookup_egl_image)(struct dri_context *ctx, void *handle);
 };
 
 static INLINE struct dri_context *
diff --git a/src/gallium/state_trackers/dri/common/dri_drawable.c b/src/gallium/state_trackers/dri/common/dri_drawable.c
index 2bc0faffeff..1bdfdccf439 100644
--- a/src/gallium/state_trackers/dri/common/dri_drawable.c
+++ b/src/gallium/state_trackers/dri/common/dri_drawable.c
@@ -32,7 +32,6 @@
 #include "dri_screen.h"
 #include "dri_context.h"
 #include "dri_drawable.h"
-#include "dri1_helper.h"
 
 #include "pipe/p_screen.h"
 #include "util/u_format.h"
@@ -68,10 +67,10 @@ dri_st_framebuffer_validate(struct st_framebuffer_iface *stfbi,
    new_stamp = (drawable->texture_stamp != drawable->dPriv->lastStamp);
 
    if (new_stamp || new_mask || screen->broken_invalidate) {
-      if (new_stamp && screen->update_drawable_info)
-         screen->update_drawable_info(drawable);
+      if (new_stamp && drawable->update_drawable_info)
+         drawable->update_drawable_info(drawable);
 
-      screen->allocate_textures(drawable, statts, count);
+      drawable->allocate_textures(drawable, statts, count);
 
       /* add existing textures */
       for (i = 0; i < ST_ATTACHMENT_COUNT; i++) {
@@ -100,10 +99,9 @@ dri_st_framebuffer_flush_front(struct st_framebuffer_iface *stfbi,
 {
    struct dri_drawable *drawable =
       (struct dri_drawable *) stfbi->st_manager_private;
-   struct dri_screen *screen = dri_screen(drawable->sPriv);
 
    /* XXX remove this and just set the correct one on the framebuffer */
-   screen->flush_frontbuffer(drawable, statt);
+   drawable->flush_frontbuffer(drawable, statt);
 
    return TRUE;
 }
@@ -138,8 +136,6 @@ dri_create_buffer(__DRIscreen * sPriv,
    drawable->dPriv = dPriv;
    dPriv->driverPrivate = (void *)drawable;
 
-   drawable->desired_fences = 2;
-
    return GL_TRUE;
 fail:
    FREE(drawable);
@@ -152,23 +148,19 @@ dri_destroy_buffer(__DRIdrawable * dPriv)
    struct dri_drawable *drawable = dri_drawable(dPriv);
    int i;
 
-   dri1_swap_fences_clear(drawable);
-
-   dri1_destroy_pipe_surface(drawable);
+   pipe_surface_reference(&drawable->drisw_surface, NULL);
 
    for (i = 0; i < ST_ATTACHMENT_COUNT; i++)
       pipe_resource_reference(&drawable->textures[i], NULL);
 
-   drawable->desired_fences = 0;
-
    FREE(drawable);
 }
 
 /**
  * Validate the texture at an attachment.  Allocate the texture if it does not
- * exist.
+ * exist.  Used by the TFP extension.
  */
-void
+static void
 dri_drawable_validate_att(struct dri_drawable *drawable,
                           enum st_attachment_type statt)
 {
@@ -189,11 +181,61 @@ dri_drawable_validate_att(struct dri_drawable *drawable,
 
    drawable->texture_stamp = drawable->dPriv->lastStamp - 1;
 
-   /* this calles into the manager */
    drawable->base.validate(&drawable->base, statts, count, NULL);
 }
 
 /**
+ * These are used for GLX_EXT_texture_from_pixmap
+ */
+static void
+dri_set_tex_buffer2(__DRIcontext *pDRICtx, GLint target,
+                    GLint format, __DRIdrawable *dPriv)
+{
+   struct dri_context *ctx = dri_context(pDRICtx);
+   struct dri_drawable *drawable = dri_drawable(dPriv);
+   struct pipe_resource *pt;
+
+   dri_drawable_validate_att(drawable, ST_ATTACHMENT_FRONT_LEFT);
+
+   pt = drawable->textures[ST_ATTACHMENT_FRONT_LEFT];
+
+   if (pt) {
+      enum pipe_format internal_format = pt->format;
+
+      if (format == __DRI_TEXTURE_FORMAT_RGB)  {
+         /* only need to cover the formats recognized by dri_fill_st_visual */
+         switch (internal_format) {
+         case PIPE_FORMAT_B8G8R8A8_UNORM:
+            internal_format = PIPE_FORMAT_B8G8R8X8_UNORM;
+            break;
+         case PIPE_FORMAT_A8R8G8B8_UNORM:
+            internal_format = PIPE_FORMAT_X8R8G8B8_UNORM;
+            break;
+         default:
+            break;
+         }
+      }
+
+      ctx->st->teximage(ctx->st,
+            (target == GL_TEXTURE_2D) ? ST_TEXTURE_2D : ST_TEXTURE_RECT,
+            0, internal_format, pt, FALSE);
+   }
+}
+
+static void
+dri_set_tex_buffer(__DRIcontext *pDRICtx, GLint target,
+                   __DRIdrawable *dPriv)
+{
+   dri_set_tex_buffer2(pDRICtx, target, __DRI_TEXTURE_FORMAT_RGBA, dPriv);
+}
+
+const __DRItexBufferExtension driTexBufferExtension = {
+    { __DRI_TEX_BUFFER, __DRI_TEX_BUFFER_VERSION },
+   dri_set_tex_buffer,
+   dri_set_tex_buffer2,
+};
+
+/**
  * Get the format and binding of an attachment.
  */
 void
diff --git a/src/gallium/state_trackers/dri/common/dri_drawable.h b/src/gallium/state_trackers/dri/common/dri_drawable.h
index 5fd650ac88e..74e662d36c4 100644
--- a/src/gallium/state_trackers/dri/common/dri_drawable.h
+++ b/src/gallium/state_trackers/dri/common/dri_drawable.h
@@ -33,13 +33,9 @@
 #include "state_tracker/st_api.h"
 
 struct pipe_surface;
-struct pipe_fence_handle;
 struct st_framebuffer;
 struct dri_context;
 
-#define DRI_SWAP_FENCES_MAX  8
-#define DRI_SWAP_FENCES_MASK 7
-
 struct dri_drawable
 {
    struct st_framebuffer_iface base;
@@ -57,14 +53,18 @@ struct dri_drawable
    struct pipe_resource *textures[ST_ATTACHMENT_COUNT];
    unsigned int texture_mask, texture_stamp;
 
-   struct pipe_fence_handle *swap_fences[DRI_SWAP_FENCES_MAX];
-   unsigned int head;
-   unsigned int tail;
-   unsigned int desired_fences;
-   unsigned int cur_fences;
+   /* used only by DRISW */
+   struct pipe_surface *drisw_surface;
+
+   /* hooks filled in by dri2 & drisw */
+   void (*allocate_textures)(struct dri_drawable *drawable,
+                             const enum st_attachment_type *statts,
+                             unsigned count);
 
-   /* used only by DRI1 */
-   struct pipe_surface *dri1_surface;
+   void (*update_drawable_info)(struct dri_drawable *drawable);
+
+   void (*flush_frontbuffer)(struct dri_drawable *drawable,
+                             enum st_attachment_type statt);
 };
 
 static INLINE struct dri_drawable *
@@ -89,9 +89,7 @@ dri_drawable_get_format(struct dri_drawable *drawable,
                         enum pipe_format *format,
                         unsigned *bind);
 
-void
-dri_drawable_validate_att(struct dri_drawable *drawable,
-                          enum st_attachment_type statt);
+extern const __DRItexBufferExtension driTexBufferExtension;
 
 #endif
 
diff --git a/src/gallium/state_trackers/dri/common/dri_screen.c b/src/gallium/state_trackers/dri/common/dri_screen.c
index 25cad8d46c4..6ad2c7da4d6 100644
--- a/src/gallium/state_trackers/dri/common/dri_screen.c
+++ b/src/gallium/state_trackers/dri/common/dri_screen.c
@@ -30,20 +30,10 @@
  */
 
 #include "utils.h"
-#ifndef __NOT_HAVE_DRM_H
-#include "vblank.h"
-#endif
 #include "xmlpool.h"
 
 #include "dri_screen.h"
 #include "dri_context.h"
-#include "dri_drawable.h"
-#include "dri1_helper.h"
-#ifndef __NOT_HAVE_DRM_H
-#include "dri2.h"
-#else
-#include "drisw.h"
-#endif
 
 #include "util/u_inlines.h"
 #include "pipe/p_screen.h"
@@ -303,11 +293,10 @@ dri_get_egl_image(struct st_manager *smapi,
 {
    struct dri_context *ctx =
       (struct dri_context *)stctxi->st_manager_private;
-   struct dri_screen *screen = dri_screen(ctx->sPriv);
    __DRIimage *img = NULL;
 
-   if (screen->lookup_egl_image) {
-      img = screen->lookup_egl_image(ctx, egl_image);
+   if (ctx->lookup_egl_image) {
+      img = ctx->lookup_egl_image(ctx, egl_image);
    }
 
    if (!img)
@@ -355,8 +344,6 @@ dri_destroy_option_cache(struct dri_screen * screen)
 void
 dri_destroy_screen_helper(struct dri_screen * screen)
 {
-   dri1_destroy_pipe_context(screen);
-
    if (screen->st_api && screen->st_api->destroy)
       screen->st_api->destroy(screen->st_api);
 
@@ -366,7 +353,7 @@ dri_destroy_screen_helper(struct dri_screen * screen)
    dri_destroy_option_cache(screen);
 }
 
-static void
+void
 dri_destroy_screen(__DRIscreen * sPriv)
 {
    struct dri_screen *screen = dri_screen(sPriv);
@@ -402,38 +389,4 @@ dri_init_screen_helper(struct dri_screen *screen,
    return dri_fill_in_modes(screen, pixel_bits);
 }
 
-/**
- * DRI driver virtual function table.
- *
- * DRI versions differ in their implementation of init_screen and swap_buffers.
- */
-const struct __DriverAPIRec driDriverAPI = {
-   .DestroyScreen = dri_destroy_screen,
-   .CreateContext = dri_create_context,
-   .DestroyContext = dri_destroy_context,
-   .CreateBuffer = dri_create_buffer,
-   .DestroyBuffer = dri_destroy_buffer,
-   .MakeCurrent = dri_make_current,
-   .UnbindContext = dri_unbind_context,
-
-#ifndef __NOT_HAVE_DRM_H
-
-   .GetSwapInfo = NULL,
-   .GetDrawableMSC = NULL,
-   .WaitForMSC = NULL,
-   .InitScreen2 = dri2_init_screen,
-
-   .InitScreen = NULL,
-   .SwapBuffers = NULL,
-   .CopySubBuffer = NULL,
-
-#else
-
-   .InitScreen = drisw_init_screen,
-   .SwapBuffers = drisw_swap_buffers,
-
-#endif
-
-};
-
 /* vim: set sw=3 ts=8 sts=3 expandtab: */
diff --git a/src/gallium/state_trackers/dri/common/dri_screen.h b/src/gallium/state_trackers/dri/common/dri_screen.h
index 087ae8d2a4a..53ccce145ba 100644
--- a/src/gallium/state_trackers/dri/common/dri_screen.h
+++ b/src/gallium/state_trackers/dri/common/dri_screen.h
@@ -64,22 +64,10 @@ struct dri_screen
    int fd;
    drmLock *drmLock;
 
-   /* hooks filled in by dri1, dri2 & drisw */
-   __DRIimage * (*lookup_egl_image)(struct dri_context *ctx, void *handle);
-   void (*allocate_textures)(struct dri_drawable *drawable,
-                             const enum st_attachment_type *statts,
-                             unsigned count);
-   void (*update_drawable_info)(struct dri_drawable *drawable);
-   void (*flush_frontbuffer)(struct dri_drawable *drawable,
-                             enum st_attachment_type statt);
-
    /* gallium */
    boolean d_depth_bits_last;
    boolean sd_depth_bits_last;
    boolean auto_fake_front;
-
-   /* used only by DRI1 */
-   struct pipe_context *dri1_pipe;
 };
 
 /** cast wrapper */
@@ -132,6 +120,9 @@ dri_init_screen_helper(struct dri_screen *screen,
 void
 dri_destroy_screen_helper(struct dri_screen * screen);
 
+void
+dri_destroy_screen(__DRIscreen * sPriv);
+
 #endif
 
 /* vim: set sw=3 ts=8 sts=3 expandtab: */
diff --git a/src/gallium/state_trackers/dri/drm/Makefile b/src/gallium/state_trackers/dri/drm/Makefile
index 94fa61fec73..c717b2bdeb5 100644
--- a/src/gallium/state_trackers/dri/drm/Makefile
+++ b/src/gallium/state_trackers/dri/drm/Makefile
@@ -17,7 +17,6 @@ C_SOURCES = \
 	dri_context.c \
 	dri_screen.c \
 	dri_drawable.c \
-	dri1_helper.c \
 	dri2.c
 
 #	$(TOP)/src/mesa/drivers/dri/common/utils.c \
diff --git a/src/gallium/state_trackers/dri/drm/SConscript b/src/gallium/state_trackers/dri/drm/SConscript
index 0c279d22366..2a0af65f9bd 100644
--- a/src/gallium/state_trackers/dri/drm/SConscript
+++ b/src/gallium/state_trackers/dri/drm/SConscript
@@ -21,7 +21,6 @@ if env['dri']:
 	source = [ 'dri_context.c',
 		'dri_drawable.c',
 		'dri_screen.c',
-		'dri1_helper.c',
 		'dri2.c',
 		]
     )
diff --git a/src/gallium/state_trackers/dri/drm/dri1_helper.c b/src/gallium/state_trackers/dri/drm/dri1_helper.c
deleted file mode 120000
index c45ebf5c102..00000000000
--- a/src/gallium/state_trackers/dri/drm/dri1_helper.c
+++ /dev/null
@@ -1 +0,0 @@
-../common/dri1_helper.c
\ No newline at end of file
diff --git a/src/gallium/state_trackers/dri/drm/dri2.c b/src/gallium/state_trackers/dri/drm/dri2.c
index 5c6573fa69b..47005c17e2b 100644
--- a/src/gallium/state_trackers/dri/drm/dri2.c
+++ b/src/gallium/state_trackers/dri/drm/dri2.c
@@ -38,9 +38,6 @@
 #include "dri_screen.h"
 #include "dri_context.h"
 #include "dri_drawable.h"
-#include "dri2.h"
-
-#include "GL/internal/dri_interface.h"
 
 /**
  * DRI2 flush extension.
@@ -70,86 +67,6 @@ static const __DRI2flushExtension dri2FlushExtension = {
 };
 
 /**
- * These are used for GLX_EXT_texture_from_pixmap
- */
-static void
-dri2_set_tex_buffer2(__DRIcontext *pDRICtx, GLint target,
-                     GLint format, __DRIdrawable *dPriv)
-{
-   struct dri_context *ctx = dri_context(pDRICtx);
-   struct dri_drawable *drawable = dri_drawable(dPriv);
-   struct pipe_resource *pt;
-
-   dri_drawable_validate_att(drawable, ST_ATTACHMENT_FRONT_LEFT);
-
-   pt = drawable->textures[ST_ATTACHMENT_FRONT_LEFT];
-
-   if (pt) {
-      enum pipe_format internal_format = pt->format;
-
-      if (format == __DRI_TEXTURE_FORMAT_RGB)  {
-         /* only need to cover the formats recognized by dri_fill_st_visual */
-         switch (internal_format) {
-         case PIPE_FORMAT_B8G8R8A8_UNORM:
-            internal_format = PIPE_FORMAT_B8G8R8X8_UNORM;
-            break;
-         case PIPE_FORMAT_A8R8G8B8_UNORM:
-            internal_format = PIPE_FORMAT_X8R8G8B8_UNORM;
-            break;
-         default:
-            break;
-         }
-      }
-
-      ctx->st->teximage(ctx->st,
-            (target == GL_TEXTURE_2D) ? ST_TEXTURE_2D : ST_TEXTURE_RECT,
-            0, internal_format, pt, FALSE);
-   }
-}
-
-static void
-dri2_set_tex_buffer(__DRIcontext *pDRICtx, GLint target,
-                    __DRIdrawable *dPriv)
-{
-   dri2_set_tex_buffer2(pDRICtx, target, __DRI_TEXTURE_FORMAT_RGBA, dPriv);
-}
-
-static const __DRItexBufferExtension dri2TexBufferExtension = {
-    { __DRI_TEX_BUFFER, __DRI_TEX_BUFFER_VERSION },
-   dri2_set_tex_buffer,
-   dri2_set_tex_buffer2,
-};
-
-/**
- * Get the format and binding of an attachment.
- */
-static INLINE void
-dri2_drawable_get_format(struct dri_drawable *drawable,
-                         enum st_attachment_type statt,
-                         enum pipe_format *format,
-                         unsigned *bind)
-{
-   switch (statt) {
-   case ST_ATTACHMENT_FRONT_LEFT:
-   case ST_ATTACHMENT_BACK_LEFT:
-   case ST_ATTACHMENT_FRONT_RIGHT:
-   case ST_ATTACHMENT_BACK_RIGHT:
-      *format = drawable->stvis.color_format;
-      *bind = PIPE_BIND_RENDER_TARGET | PIPE_BIND_SAMPLER_VIEW;
-      break;
-   case ST_ATTACHMENT_DEPTH_STENCIL:
-      *format = drawable->stvis.depth_stencil_format;
-      *bind = PIPE_BIND_DEPTH_STENCIL; /* XXX sampler? */
-      break;
-   default:
-      *format = PIPE_FORMAT_NONE;
-      *bind = 0;
-      break;
-   }
-}
-
-
-/**
  * Retrieve __DRIbuffer from the DRI loader.
  */
 static __DRIbuffer *
@@ -179,7 +96,7 @@ dri2_drawable_get_buffers(struct dri_drawable *drawable,
       unsigned bind;
       int att, bpp;
 
-      dri2_drawable_get_format(drawable, statts[i], &format, &bind);
+      dri_drawable_get_format(drawable, statts[i], &format, &bind);
       if (format == PIPE_FORMAT_NONE)
          continue;
 
@@ -321,7 +238,7 @@ dri2_drawable_process_buffers(struct dri_drawable *drawable,
          break;
       }
 
-      dri2_drawable_get_format(drawable, statt, &format, &bind);
+      dri_drawable_get_format(drawable, statt, &format, &bind);
       if (statt == ST_ATTACHMENT_INVALID || format == PIPE_FORMAT_NONE)
          continue;
 
@@ -354,7 +271,8 @@ dri2_allocate_textures(struct dri_drawable *drawable,
    unsigned num_buffers = count;
 
    buffers = dri2_drawable_get_buffers(drawable, statts, &num_buffers);
-   dri2_drawable_process_buffers(drawable, buffers, num_buffers);
+   if (buffers)
+      dri2_drawable_process_buffers(drawable, buffers, num_buffers);
 }
 
 static void
@@ -485,7 +403,7 @@ static const __DRIextension *dri_screen_extensions[] = {
    &driCopySubBufferExtension.base,
    &driSwapControlExtension.base,
    &driMediaStreamCounterExtension.base,
-   &dri2TexBufferExtension.base,
+   &driTexBufferExtension.base,
    &dri2FlushExtension.base,
    &dri2ImageExtension.base,
    &dri2ConfigQueryExtension.base,
@@ -497,7 +415,7 @@ static const __DRIextension *dri_screen_extensions[] = {
  *
  * Returns the __GLcontextModes supported by this driver.
  */
-const __DRIconfig **
+static const __DRIconfig **
 dri2_init_screen(__DRIscreen * sPriv)
 {
    const __DRIconfig **configs;
@@ -510,9 +428,6 @@ dri2_init_screen(__DRIscreen * sPriv)
 
    screen->sPriv = sPriv;
    screen->fd = sPriv->fd;
-   screen->lookup_egl_image = dri2_lookup_egl_image;
-   screen->allocate_textures = dri2_allocate_textures;
-   screen->flush_frontbuffer = dri2_flush_frontbuffer;
 
    sPriv->private = (void *)screen;
    sPriv->extensions = dri_screen_extensions;
@@ -534,6 +449,64 @@ fail:
    return NULL;
 }
 
+static boolean
+dri2_create_context(gl_api api, const __GLcontextModes * visual,
+                    __DRIcontext * cPriv, void *sharedContextPrivate)
+{
+   struct dri_context *ctx = NULL;
+
+   if (!dri_create_context(api, visual, cPriv, sharedContextPrivate))
+      return FALSE;
+
+   ctx = cPriv->driverPrivate;
+
+   ctx->lookup_egl_image = dri2_lookup_egl_image;
+
+   return TRUE;
+}
+
+static boolean
+dri2_create_buffer(__DRIscreen * sPriv,
+                   __DRIdrawable * dPriv,
+                   const __GLcontextModes * visual, boolean isPixmap)
+{
+   struct dri_drawable *drawable = NULL;
+
+   if (!dri_create_buffer(sPriv, dPriv, visual, isPixmap))
+      return FALSE;
+
+   drawable = dPriv->driverPrivate;
+
+   drawable->allocate_textures = dri2_allocate_textures;
+   drawable->flush_frontbuffer = dri2_flush_frontbuffer;
+
+   return TRUE;
+}
+
+/**
+ * DRI driver virtual function table.
+ *
+ * DRI versions differ in their implementation of init_screen and swap_buffers.
+ */
+const struct __DriverAPIRec driDriverAPI = {
+   .InitScreen = NULL,
+   .InitScreen2 = dri2_init_screen,
+   .DestroyScreen = dri_destroy_screen,
+   .CreateContext = dri2_create_context,
+   .DestroyContext = dri_destroy_context,
+   .CreateBuffer = dri2_create_buffer,
+   .DestroyBuffer = dri_destroy_buffer,
+   .MakeCurrent = dri_make_current,
+   .UnbindContext = dri_unbind_context,
+
+   .GetSwapInfo = NULL,
+   .GetDrawableMSC = NULL,
+   .WaitForMSC = NULL,
+
+   .SwapBuffers = NULL,
+   .CopySubBuffer = NULL,
+};
+
 /* This is the table of extensions that the loader will dlsym() for. */
 PUBLIC const __DRIextension *__driDriverExtensions[] = {
     &driCoreExtension.base,
diff --git a/src/gallium/state_trackers/dri/sw/Makefile b/src/gallium/state_trackers/dri/sw/Makefile
index a1dadeba5e6..33bc0ed9c94 100644
--- a/src/gallium/state_trackers/dri/sw/Makefile
+++ b/src/gallium/state_trackers/dri/sw/Makefile
@@ -20,7 +20,6 @@ C_SOURCES = \
 	dri_context.c \
 	dri_screen.c \
 	dri_drawable.c \
-	dri1_helper.c \
 	drisw.c
 
 include ../../../Makefile.template
diff --git a/src/gallium/state_trackers/dri/sw/SConscript b/src/gallium/state_trackers/dri/sw/SConscript
index 0c5194d6edc..d2eb66668e0 100644
--- a/src/gallium/state_trackers/dri/sw/SConscript
+++ b/src/gallium/state_trackers/dri/sw/SConscript
@@ -21,7 +21,6 @@ if env['dri']:
 	source = [ 'dri_context.c',
 		'dri_drawable.c',
 		'dri_screen.c',
-		'dri1_helper.c',
 		'drisw.c',
 		]
     )
diff --git a/src/gallium/state_trackers/dri/sw/dri1_helper.c b/src/gallium/state_trackers/dri/sw/dri1_helper.c
deleted file mode 120000
index c45ebf5c102..00000000000
--- a/src/gallium/state_trackers/dri/sw/dri1_helper.c
+++ /dev/null
@@ -1 +0,0 @@
-../common/dri1_helper.c
\ No newline at end of file
diff --git a/src/gallium/state_trackers/dri/sw/drisw.c b/src/gallium/state_trackers/dri/sw/drisw.c
index 23e99aa0add..249ccd7fcf6 100644
--- a/src/gallium/state_trackers/dri/sw/drisw.c
+++ b/src/gallium/state_trackers/dri/sw/drisw.c
@@ -43,9 +43,9 @@
 #include "dri_screen.h"
 #include "dri_context.h"
 #include "dri_drawable.h"
-#include "dri1_helper.h"
-#include "drisw.h"
 
+DEBUG_GET_ONCE_BOOL_OPTION(swrast_no_present, "SWRAST_NO_PRESENT", FALSE);
+static boolean swrast_no_present = FALSE;
 
 static INLINE void
 get_drawable_info(__DRIdrawable *dPriv, int *w, int *h)
@@ -87,6 +87,24 @@ drisw_put_image(struct dri_drawable *drawable,
    put_image(dPriv, data, width, height);
 }
 
+static struct pipe_surface *
+drisw_get_pipe_surface(struct dri_drawable *drawable, struct pipe_resource *ptex)
+{
+   struct pipe_screen *pipe_screen = dri_screen(drawable->sPriv)->base.screen;
+   struct pipe_surface *psurf = drawable->drisw_surface;
+
+   if (!psurf || psurf->texture != ptex) {
+      pipe_surface_reference(&drawable->drisw_surface, NULL);
+
+      drawable->drisw_surface = pipe_screen->get_tex_surface(pipe_screen,
+            ptex, 0, 0, 0, 0/* no bind flag???*/);
+
+      psurf = drawable->drisw_surface;
+   }
+
+   return psurf;
+}
+
 static INLINE void
 drisw_present_texture(__DRIdrawable *dPriv,
                       struct pipe_resource *ptex)
@@ -95,7 +113,10 @@ drisw_present_texture(__DRIdrawable *dPriv,
    struct dri_screen *screen = dri_screen(drawable->sPriv);
    struct pipe_surface *psurf;
 
-   psurf = dri1_get_pipe_surface(drawable, ptex);
+   if (swrast_no_present)
+      return;
+
+   psurf = drisw_get_pipe_surface(drawable, ptex);
    if (!psurf)
       return;
 
@@ -128,7 +149,7 @@ drisw_copy_to_front(__DRIdrawable * dPriv,
  * Backend functions for st_framebuffer interface and swap_buffers.
  */
 
-void
+static void
 drisw_swap_buffers(__DRIdrawable *dPriv)
 {
    struct dri_context *ctx = dri_get_current(dPriv->driScreenPriv);
@@ -170,10 +191,6 @@ drisw_flush_frontbuffer(struct dri_drawable *drawable,
  * During fixed-size operation, the function keeps allocating new attachments
  * as they are requested. Unused attachments are not removed, not until the
  * framebuffer is resized or destroyed.
- *
- * It should be possible for DRI1 and DRISW to share this function, but it
- * seems a better seperation and safer for each DRI version to provide its own
- * function.
  */
 static void
 drisw_allocate_textures(struct dri_drawable *drawable,
@@ -184,7 +201,7 @@ drisw_allocate_textures(struct dri_drawable *drawable,
    struct pipe_resource templ;
    unsigned width, height;
    boolean resized;
-   int i;
+   unsigned i;
 
    width  = drawable->dPriv->w;
    height = drawable->dPriv->h;
@@ -205,7 +222,7 @@ drisw_allocate_textures(struct dri_drawable *drawable,
    templ.depth0 = 1;
    templ.last_level = 0;
 
-   for (i = 0; i < ST_ATTACHMENT_COUNT; i++) {
+   for (i = 0; i < count; i++) {
       enum pipe_format format;
       unsigned bind;
 
@@ -215,7 +232,8 @@ drisw_allocate_textures(struct dri_drawable *drawable,
 
       dri_drawable_get_format(drawable, statts[i], &format, &bind);
 
-      if (statts[i] != ST_ATTACHMENT_DEPTH_STENCIL)
+      /* if we don't do any present, no need for display targets */
+      if (statts[i] != ST_ATTACHMENT_DEPTH_STENCIL && !swrast_no_present)
          bind |= PIPE_BIND_DISPLAY_TARGET;
 
       if (format == PIPE_FORMAT_NONE)
@@ -244,7 +262,7 @@ static struct drisw_loader_funcs drisw_lf = {
    .put_image = drisw_put_image
 };
 
-const __DRIconfig **
+static const __DRIconfig **
 drisw_init_screen(__DRIscreen * sPriv)
 {
    const __DRIconfig **configs;
@@ -257,9 +275,8 @@ drisw_init_screen(__DRIscreen * sPriv)
 
    screen->sPriv = sPriv;
    screen->fd = -1;
-   screen->allocate_textures = drisw_allocate_textures;
-   screen->update_drawable_info = drisw_update_drawable_info;
-   screen->flush_frontbuffer = drisw_flush_frontbuffer;
+
+   swrast_no_present = debug_get_option_swrast_no_present();
 
    sPriv->private = (void *)screen;
    sPriv->extensions = drisw_screen_extensions;
@@ -278,6 +295,43 @@ fail:
    return NULL;
 }
 
+static boolean
+drisw_create_buffer(__DRIscreen * sPriv,
+                    __DRIdrawable * dPriv,
+                    const __GLcontextModes * visual, boolean isPixmap)
+{
+   struct dri_drawable *drawable = NULL;
+
+   if (!dri_create_buffer(sPriv, dPriv, visual, isPixmap))
+      return FALSE;
+
+   drawable = dPriv->driverPrivate;
+
+   drawable->allocate_textures = drisw_allocate_textures;
+   drawable->update_drawable_info = drisw_update_drawable_info;
+   drawable->flush_frontbuffer = drisw_flush_frontbuffer;
+
+   return TRUE;
+}
+
+/**
+ * DRI driver virtual function table.
+ *
+ * DRI versions differ in their implementation of init_screen and swap_buffers.
+ */
+const struct __DriverAPIRec driDriverAPI = {
+   .InitScreen = drisw_init_screen,
+   .DestroyScreen = dri_destroy_screen,
+   .CreateContext = dri_create_context,
+   .DestroyContext = dri_destroy_context,
+   .CreateBuffer = drisw_create_buffer,
+   .DestroyBuffer = dri_destroy_buffer,
+   .MakeCurrent = dri_make_current,
+   .UnbindContext = dri_unbind_context,
+
+   .SwapBuffers = drisw_swap_buffers,
+};
+
 /* This is the table of extensions that the loader will dlsym() for. */
 PUBLIC const __DRIextension *__driDriverExtensions[] = {
     &driCoreExtension.base,
diff --git a/src/gallium/state_trackers/dri/sw/drisw.h b/src/gallium/state_trackers/dri/sw/drisw.h
deleted file mode 100644
index 6c6c891f356..00000000000
--- a/src/gallium/state_trackers/dri/sw/drisw.h
+++ /dev/null
@@ -1,43 +0,0 @@
-/**************************************************************************
- *
- * Copyright 2009, VMware, Inc.
- * All Rights Reserved.
- * Copyright 2010 George Sapountzis <[email protected]>
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- *
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- *
- **************************************************************************/
-
-#ifndef DRISW_H
-#define DRISW_H
-
-#include "dri_context.h"
-#include "dri_drawable.h"
-
-#include "state_tracker/st_api.h"
-#include "dri_wrapper.h"
-
-const __DRIconfig **
-drisw_init_screen(__DRIscreen * sPriv);
-
-void drisw_swap_buffers(__DRIdrawable * dPriv);
-
-#endif /* DRISW_H */
diff --git a/src/gallium/state_trackers/egl/common/egl_g3d.c b/src/gallium/state_trackers/egl/common/egl_g3d.c
index b6321e6b437..56d575ffe08 100644
--- a/src/gallium/state_trackers/egl/common/egl_g3d.c
+++ b/src/gallium/state_trackers/egl/common/egl_g3d.c
@@ -434,9 +434,11 @@ egl_g3d_free_config(void *conf)
 static void
 egl_g3d_free_screen(void *scr)
 {
+#ifdef EGL_MESA_screen_surface
    struct egl_g3d_screen *gscr = egl_g3d_screen((_EGLScreen *) scr);
    FREE(gscr->native_modes);
    FREE(gscr);
+#endif
 }
 
 static EGLBoolean
diff --git a/src/gallium/state_trackers/egl/common/egl_g3d.h b/src/gallium/state_trackers/egl/common/egl_g3d.h
index ed2b0409bb9..f33dc91cf90 100644
--- a/src/gallium/state_trackers/egl/common/egl_g3d.h
+++ b/src/gallium/state_trackers/egl/common/egl_g3d.h
@@ -95,15 +95,19 @@ struct egl_g3d_image {
    unsigned zslice;
 };
 
+/* standard typecasts */
+_EGL_DRIVER_STANDARD_TYPECASTS(egl_g3d)
+_EGL_DRIVER_TYPECAST(egl_g3d_image, _EGLImage, obj)
+
+#ifdef EGL_MESA_screen_surface
+
 struct egl_g3d_screen {
    _EGLScreen base;
    const struct native_connector *native;
    const struct native_mode **native_modes;
 };
-
-/* standard typecasts */
-_EGL_DRIVER_STANDARD_TYPECASTS(egl_g3d)
 _EGL_DRIVER_TYPECAST(egl_g3d_screen, _EGLScreen, obj)
-_EGL_DRIVER_TYPECAST(egl_g3d_image, _EGLImage, obj)
+
+#endif /* EGL_MESA_screen_surface */
 
 #endif /* _EGL_G3D_H_ */
diff --git a/src/gallium/state_trackers/egl/common/egl_g3d_image.c b/src/gallium/state_trackers/egl/common/egl_g3d_image.c
index b1fe30a776c..1e13cfcf7e9 100644
--- a/src/gallium/state_trackers/egl/common/egl_g3d_image.c
+++ b/src/gallium/state_trackers/egl/common/egl_g3d_image.c
@@ -78,7 +78,7 @@ egl_g3d_create_image(_EGLDriver *drv, _EGLDisplay *dpy, _EGLContext *ctx,
 
    gimg = CALLOC_STRUCT(egl_g3d_image);
    if (!gimg) {
-      _eglError(EGL_BAD_ALLOC, "eglCreatePbufferSurface");
+      _eglError(EGL_BAD_ALLOC, "eglCreateEGLImageKHR");
       return NULL;
    }
 
diff --git a/src/gallium/state_trackers/egl/x11/glxinit.c b/src/gallium/state_trackers/egl/x11/glxinit.c
index 809a0987e55..57c6aaff864 100644
--- a/src/gallium/state_trackers/egl/x11/glxinit.c
+++ b/src/gallium/state_trackers/egl/x11/glxinit.c
@@ -10,10 +10,16 @@
 #include <assert.h>
 #include <X11/Xlib.h>
 #include <X11/Xproto.h>
+#include <X11/Xlibint.h>
 #include <X11/extensions/Xext.h>
 #include <X11/extensions/extutil.h>
 #include <sys/time.h>
 
+#include "GL/glxproto.h"
+#include "GL/glxtokens.h"
+#include "GL/gl.h" /* for GL types needed by __GLcontextModes */
+#include "GL/internal/glcore.h"  /* for __GLcontextModes */
+
 #include "glxinit.h"
 
 #ifdef GLX_DIRECT_RENDERING
@@ -55,9 +61,9 @@ static /* const */ XExtensionHooks __glXExtensionHooks = {
   NULL,                   /* error_string */
 };
 
-XEXT_GENERATE_FIND_DISPLAY(__glXFindDisplay, __glXExtensionInfo,
-                           __glXExtensionName, &__glXExtensionHooks,
-                           __GLX_NUMBER_EVENTS, NULL)
+static XEXT_GENERATE_FIND_DISPLAY(__glXFindDisplay, __glXExtensionInfo,
+				  __glXExtensionName, &__glXExtensionHooks,
+				  __GLX_NUMBER_EVENTS, NULL)
 
 static GLint
 _gl_convert_from_x_visual_type(int visualType)
@@ -73,6 +79,17 @@ _gl_convert_from_x_visual_type(int visualType)
       ? glx_visual_types[visualType] : GLX_NONE;
 }
 
+static void
+_gl_context_modes_destroy(__GLcontextModes * modes)
+{
+   while (modes != NULL) {
+      __GLcontextModes *const next = modes->next;
+
+      Xfree(modes);
+      modes = next;
+   }
+}
+
 static __GLcontextModes *
 _gl_context_modes_create(unsigned count, size_t minimum_size)
 {
@@ -116,18 +133,7 @@ _gl_context_modes_create(unsigned count, size_t minimum_size)
    return base;
 }
 
-_X_HIDDEN void
-_gl_context_modes_destroy(__GLcontextModes * modes)
-{
-   while (modes != NULL) {
-      __GLcontextModes *const next = modes->next;
-
-      Xfree(modes);
-      modes = next;
-   }
-}
-
-_X_HIDDEN char *
+static char *
 __glXQueryServerString(Display * dpy, int opcode, CARD32 screen, CARD32 name)
 {
    xGLXGenericGetStringReq *req;
@@ -194,10 +200,6 @@ FreeScreenConfigs(__GLXdisplayPrivate * priv)
          _gl_context_modes_destroy(psc->configs);
          psc->configs = NULL;   /* NOTE: just for paranoia */
       }
-      if (psc->visuals) {
-         _gl_context_modes_destroy(psc->visuals);
-         psc->visuals = NULL;   /* NOTE: just for paranoia */
-      }
       Xfree((char *) psc->serverGLXexts);
    }
    XFree((char *) priv->screenConfigs);
@@ -215,14 +217,8 @@ __glXFreeDisplayPrivate(XExtData * extension)
 
    priv = (__GLXdisplayPrivate *) extension->private_data;
    FreeScreenConfigs(priv);
-   if (priv->serverGLXvendor) {
-      Xfree((char *) priv->serverGLXvendor);
-      priv->serverGLXvendor = 0x0;      /* to protect against double free's */
-   }
-   if (priv->serverGLXversion) {
+   if (priv->serverGLXversion)
       Xfree((char *) priv->serverGLXversion);
-      priv->serverGLXversion = 0x0;     /* to protect against double free's */
-   }
 
    Xfree((char *) priv);
    return 0;
@@ -234,6 +230,10 @@ __glXFreeDisplayPrivate(XExtData * extension)
 ** Query the version of the GLX extension.  This procedure works even if
 ** the client extension is not completely set up.
 */
+
+#define GLX_MAJOR_VERSION 1       /* current version numbers */
+#define GLX_MINOR_VERSION 4
+
 static Bool
 QueryVersion(Display * dpy, int opcode, int *major, int *minor)
 {
@@ -263,7 +263,13 @@ QueryVersion(Display * dpy, int opcode, int *major, int *minor)
    return GL_TRUE;
 }
 
-_X_HIDDEN void
+#define __GLX_MIN_CONFIG_PROPS	18
+#define __GLX_MAX_CONFIG_PROPS	500
+#define __GLX_EXT_CONFIG_PROPS 	10
+#define __GLX_TOTAL_CONFIG       (__GLX_MIN_CONFIG_PROPS +      \
+                                    2 * __GLX_EXT_CONFIG_PROPS)
+
+static void
 __glXInitializeVisualConfigFromTags(__GLcontextModes * config, int count,
                                     const INT32 * bp, Bool tagged_only,
                                     Bool fbconfig_style_tags)
@@ -506,35 +512,6 @@ createConfigsFromProperties(Display * dpy, int nvisuals, int nprops,
 }
 
 static GLboolean
-getVisualConfigs(__GLXscreenConfigs *psc,
-                 __GLXdisplayPrivate *priv, int screen)
-{
-   xGLXGetVisualConfigsReq *req;
-   xGLXGetVisualConfigsReply reply;
-   Display *dpy = priv->dpy;
-
-   LockDisplay(dpy);
-
-   psc->visuals = NULL;
-   GetReq(GLXGetVisualConfigs, req);
-   req->reqType = priv->majorOpcode;
-   req->glxCode = X_GLXGetVisualConfigs;
-   req->screen = screen;
-
-   if (!_XReply(dpy, (xReply *) & reply, 0, False))
-      goto out;
-
-   psc->visuals = createConfigsFromProperties(dpy,
-                                              reply.numVisuals,
-                                              reply.numProps,
-                                              screen, GL_FALSE);
-
- out:
-   UnlockDisplay(dpy);
-   return psc->visuals != NULL;
-}
-
-static GLboolean
 getFBConfigs(__GLXscreenConfigs *psc, __GLXdisplayPrivate *priv, int screen)
 {
    xGLXGetFBConfigsReq *fb_req;
@@ -581,32 +558,6 @@ getFBConfigs(__GLXscreenConfigs *psc, __GLXdisplayPrivate *priv, int screen)
    return psc->configs != NULL;
 }
 
-_X_HIDDEN Bool
-glx_screen_init(__GLXscreenConfigs *psc,
-               int screen, __GLXdisplayPrivate * priv)
-{
-   /* Initialize per screen dynamic client GLX extensions */
-   psc->ext_list_first_time = GL_TRUE;
-   psc->scr = screen;
-   psc->dpy = priv->dpy;
-
-   getVisualConfigs(psc, priv, screen);
-   getFBConfigs(psc, priv, screen);
-
-   return GL_TRUE;
-}
-
-static __GLXscreenConfigs *
-createIndirectScreen()
-{
-   __GLXscreenConfigs *psc;
-
-   psc = Xmalloc(sizeof *psc);
-   memset(psc, 0, sizeof *psc);
-
-   return psc;
-}
-
 static GLboolean
 AllocAndFetchScreenConfigs(Display * dpy, __GLXdisplayPrivate * priv)
 {
@@ -630,10 +581,10 @@ AllocAndFetchScreenConfigs(Display * dpy, __GLXdisplayPrivate * priv)
    }
 
    for (i = 0; i < screens; i++) {
-      psc = createIndirectScreen();
+      psc = Xcalloc(1, sizeof *psc);
       if (!psc)
          return GL_FALSE;
-      glx_screen_init(psc, i, priv);
+      getFBConfigs(psc, priv, i);
       priv->screenConfigs[i] = psc;
    }
 
@@ -682,13 +633,8 @@ __glXInitialize(Display * dpy)
     ** structures from the server.
     */
    dpyPriv->majorOpcode = info->codes->major_opcode;
-   dpyPriv->majorVersion = major;
-   dpyPriv->minorVersion = minor;
    dpyPriv->dpy = dpy;
 
-   dpyPriv->serverGLXvendor = NULL;
-   dpyPriv->serverGLXversion = NULL;
-
    if (!AllocAndFetchScreenConfigs(dpy, dpyPriv)) {
       Xfree(dpyPriv);
       Xfree(private);
diff --git a/src/gallium/state_trackers/egl/x11/glxinit.h b/src/gallium/state_trackers/egl/x11/glxinit.h
index 1cc7c460fe2..4078aef2fee 100644
--- a/src/gallium/state_trackers/egl/x11/glxinit.h
+++ b/src/gallium/state_trackers/egl/x11/glxinit.h
@@ -2,10 +2,21 @@
 #define GLXINIT_INCLUDED
 
 #include <X11/Xlib.h>
-#include "glxclient.h"
+#include <GL/gl.h>
 
-/* this is used by DRI loaders */
-extern void
-_gl_context_modes_destroy(__GLcontextModes * modes);
+typedef struct {
+   __GLcontextModes *configs;
+   char *serverGLXexts;
+} __GLXscreenConfigs;
+
+typedef struct {
+   Display *dpy;
+   __GLXscreenConfigs **screenConfigs;
+   char *serverGLXversion;
+   int majorOpcode;
+   struct x11_screen *xscr;
+} __GLXdisplayPrivate;
+
+extern __GLXdisplayPrivate *__glXInitialize(Display * dpy);
 
 #endif /* GLXINIT_INCLUDED */
diff --git a/src/gallium/state_trackers/egl/x11/x11_screen.c b/src/gallium/state_trackers/egl/x11/x11_screen.c
index bc6482ab15d..c07ebb7ef6c 100644
--- a/src/gallium/state_trackers/egl/x11/x11_screen.c
+++ b/src/gallium/state_trackers/egl/x11/x11_screen.c
@@ -39,11 +39,6 @@
 #include "glxinit.h"
 
 struct x11_screen {
-#ifdef GLX_DIRECT_RENDERING
-   /* dummy base class */
-   struct __GLXDRIdisplayRec base;
-#endif
-
    Display *dpy;
    int number;
 
@@ -108,7 +103,7 @@ x11_screen_destroy(struct x11_screen *xscr)
 #ifdef GLX_DIRECT_RENDERING
    /* xscr->glx_dpy will be destroyed with the X display */
    if (xscr->glx_dpy)
-      xscr->glx_dpy->dri2Display = NULL;
+      xscr->glx_dpy->xscr = NULL;
 #endif
 
    if (xscr->visuals)
@@ -231,17 +226,6 @@ x11_screen_get_glx_configs(struct x11_screen *xscr)
 }
 
 /**
- * Return the GLX visuals.
- */
-const __GLcontextModes *
-x11_screen_get_glx_visuals(struct x11_screen *xscr)
-{
-   return (x11_screen_init_glx(xscr))
-      ? xscr->glx_dpy->screenConfigs[xscr->number]->visuals
-      : NULL;
-}
-
-/**
  * Probe the screen for the DRI2 driver name.
  */
 const char *
@@ -306,14 +290,14 @@ x11_screen_enable_dri2(struct x11_screen *xscr,
          close(fd);
          return -1;
       }
-      if (xscr->glx_dpy->dri2Display) {
+      if (xscr->glx_dpy->xscr) {
          _eglLog(_EGL_WARNING,
                "display is already managed by another x11 screen");
          close(fd);
          return -1;
       }
 
-      xscr->glx_dpy->dri2Display = (__GLXDRIdisplay *) xscr;
+      xscr->glx_dpy->xscr = xscr;
       xscr->dri_invalidate_buffers = invalidate_buffers;
       xscr->dri_user_data = user_data;
 
@@ -428,6 +412,9 @@ x11_context_modes_count(const __GLcontextModes *modes)
    return count;
 }
 
+extern void
+dri2InvalidateBuffers(Display *dpy, XID drawable);
+
 /**
  * This is called from src/glx/dri2.c.
  */
@@ -437,8 +424,8 @@ dri2InvalidateBuffers(Display *dpy, XID drawable)
    __GLXdisplayPrivate *priv = __glXInitialize(dpy);
    struct x11_screen *xscr = NULL;
 
-   if (priv && priv->dri2Display)
-      xscr = (struct x11_screen *) priv->dri2Display;
+   if (priv && priv->xscr)
+      xscr = priv->xscr;
    if (!xscr || !xscr->dri_invalidate_buffers)
       return;
 
diff --git a/src/gallium/state_trackers/python/p_context.i b/src/gallium/state_trackers/python/p_context.i
index cf0144b5dc6..40c4603fb9d 100644
--- a/src/gallium/state_trackers/python/p_context.i
+++ b/src/gallium/state_trackers/python/p_context.i
@@ -289,6 +289,20 @@ struct st_context {
                                       $self->vertex_buffers);
    }
 
+   void set_index_buffer(unsigned index_size,
+                         unsigned offset,
+                         struct pipe_resource *buffer)
+   {
+      struct pipe_index_buffer ib;
+
+      memset(&ib, 0, sizeof(ib));
+      ib.index_size = index_size;
+      ib.offset = offset;
+      ib.buffer = buffer;
+
+      $self->pipe->set_index_buffer($self->pipe, &ib);
+   }
+
    void set_vertex_element(unsigned index,
                            const struct pipe_vertex_element *element) 
    {
@@ -308,29 +322,12 @@ struct st_context {
     */
    
    void draw_arrays(unsigned mode, unsigned start, unsigned count) {
-      $self->pipe->draw_arrays($self->pipe, mode, start, count);
-   }
-
-   void draw_elements( struct pipe_resource *indexBuffer,
-                       unsigned indexSize, int indexBias,
-                       unsigned mode, unsigned start, unsigned count) 
-   {
-      $self->pipe->draw_elements($self->pipe, 
-                                 indexBuffer, 
-                                 indexSize, 
-                                 indexBias,
-                                 mode, start, count);
+      util_draw_arrays($self->pipe, mode, start, count);
    }
 
-   void draw_range_elements( struct pipe_resource *indexBuffer,
-                             unsigned indexSize, int indexBias,
-                             unsigned minIndex, unsigned maxIndex,
-                             unsigned mode, unsigned start, unsigned count)
+   void draw_vbo(const struct pipe_draw_info *info)
    {
-      $self->pipe->draw_range_elements($self->pipe, 
-                                       indexBuffer, indexSize, indexBias,
-                                       minIndex, maxIndex,
-                                       mode, start, count);
+      $self->pipe->draw_vbo($self->pipe, info);
    }
 
    void draw_vertices(unsigned prim,
@@ -382,7 +379,7 @@ struct st_context {
       pipe->set_vertex_buffers(pipe, 1, &vbuffer);
 
       /* draw */
-      pipe->draw_arrays(pipe, prim, 0, num_verts);
+      util_draw_arrays(pipe, prim, 0, num_verts);
 
       cso_restore_vertex_elements($self->cso);
 
diff --git a/src/gallium/state_trackers/vega/polygon.c b/src/gallium/state_trackers/vega/polygon.c
index e9c8f031373..bc94170eb96 100644
--- a/src/gallium/state_trackers/vega/polygon.c
+++ b/src/gallium/state_trackers/vega/polygon.c
@@ -301,8 +301,7 @@ static void draw_polygon(struct vg_context *ctx,
    cso_set_vertex_elements(ctx->cso_context, 1, &velement);
 
    /* draw */
-   pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLE_FAN, 
-                     0, poly->num_verts);
+   util_draw_arrays(pipe, PIPE_PRIM_TRIANGLE_FAN, 0, (uint) poly->num_verts);
 }
 
 void polygon_fill(struct polygon *poly, struct vg_context *ctx)
diff --git a/src/gallium/state_trackers/xorg/xorg_driver.c b/src/gallium/state_trackers/xorg/xorg_driver.c
index e993ccc9bf0..e10ff2f9508 100644
--- a/src/gallium/state_trackers/xorg/xorg_driver.c
+++ b/src/gallium/state_trackers/xorg/xorg_driver.c
@@ -472,7 +472,6 @@ drv_pre_init(ScrnInfoPtr pScrn, int flags)
 	    max_height = max < max_height ? max : max_height;
 	}
 
-	drmModeFreeResources(res);
 	xf86CrtcSetSizeRange(pScrn, res->min_width,
 			     res->min_height, max_width, max_height);
 	xf86DrvMsg(pScrn->scrnIndex, X_PROBED,
@@ -481,6 +480,7 @@ drv_pre_init(ScrnInfoPtr pScrn, int flags)
 	xf86DrvMsg(pScrn->scrnIndex, X_PROBED,
 		   "Min height %d, Max Height %d.\n",
 		   res->min_height, max_height);
+	drmModeFreeResources(res);
     }
 
 
diff --git a/src/gallium/targets/Makefile.xorg b/src/gallium/targets/Makefile.xorg
index c2d00649789..762c905985e 100644
--- a/src/gallium/targets/Makefile.xorg
+++ b/src/gallium/targets/Makefile.xorg
@@ -42,7 +42,7 @@ endif
 default: depend $(TOP)/$(LIB_DIR)/gallium $(LIBNAME) $(LIBNAME_STAGING)
 
 $(LIBNAME): $(OBJECTS) Makefile ../Makefile.xorg $(LIBS) $(DRIVER_PIPES)
-	$(MKLIB) -noprefix -o $@ $(OBJECTS) $(DRIVER_PIPES) $(GALLIUM_AUXILIARIES) $(DRIVER_LINKS)
+	$(MKLIB) -noprefix -o $@ $(LDFLAGS) $(OBJECTS) $(DRIVER_PIPES) $(GALLIUM_AUXILIARIES) $(DRIVER_LINKS)
 
 depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(SYMLINKS) $(GENERATED_SOURCES)
 	rm -f depend
diff --git a/src/gallium/targets/dri-i915/Makefile b/src/gallium/targets/dri-i915/Makefile
index e88c3c9f66a..9c10d71a4a6 100644
--- a/src/gallium/targets/dri-i915/Makefile
+++ b/src/gallium/targets/dri-i915/Makefile
@@ -22,7 +22,7 @@ DRIVER_DEFINES = \
 	-DGALLIUM_RBUG -DGALLIUM_TRACE -DGALLIUM_GALAHAD -DGALLIUM_SOFTPIPE
 
 ifeq ($(MESA_LLVM),1)
-DRIVER_DEFINS += -DGALLIUM_LLVMPIPE
+DRIVER_DEFINES += -DGALLIUM_LLVMPIPE
 endif
 
 include ../Makefile.dri
diff --git a/src/gallium/targets/dri-i965/Makefile b/src/gallium/targets/dri-i965/Makefile
index 3679c075b24..4b50d04255f 100644
--- a/src/gallium/targets/dri-i965/Makefile
+++ b/src/gallium/targets/dri-i965/Makefile
@@ -6,10 +6,11 @@ LIBNAME = i965_dri.so
 PIPE_DRIVERS = \
 	$(TOP)/src/gallium/state_trackers/dri/drm/libdridrm.a \
 	$(TOP)/src/gallium/winsys/i965/drm/libi965drm.a \
-	$(TOP)/src/gallium/drivers/trace/libtrace.a \
-	$(TOP)/src/gallium/drivers/rbug/librbug.a \
 	$(TOP)/src/gallium/winsys/sw/wrapper/libwsw.a \
 	$(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \
+	$(TOP)/src/gallium/drivers/galahad/libgalahad.a \
+	$(TOP)/src/gallium/drivers/trace/libtrace.a \
+	$(TOP)/src/gallium/drivers/rbug/librbug.a \
 	$(TOP)/src/gallium/drivers/i965/libi965.a
 
 C_SOURCES = \
@@ -18,7 +19,11 @@ C_SOURCES = \
 	$(DRIVER_SOURCES)
 
 DRIVER_DEFINES = \
-	-DGALLIUM_SOFTPIPE -DGALLIUM_RBUG -DGALLIUM_TRACE
+	-DGALLIUM_RBUG -DGALLIUM_TRACE -DGALLIUM_GALAHAD -DGALLIUM_SOFTPIPE
+
+ifeq ($(MESA_LLVM),1)
+DRIVER_DEFINES += -DGALLIUM_LLVMPIPE
+endif
 
 include ../Makefile.dri
 
diff --git a/src/gallium/targets/dri-r600/SConscript b/src/gallium/targets/dri-r600/SConscript
index 97c5df01fe2..64d6d2a7f6f 100644
--- a/src/gallium/targets/dri-r600/SConscript
+++ b/src/gallium/targets/dri-r600/SConscript
@@ -12,7 +12,7 @@ env.Append(CPPDEFINES = ['GALLIUM_RBUG', 'GALLIUM_TRACE'])
 
 env.Prepend(LIBS = [
     st_dri,
-    r600drm,
+    r600winsys,
     r600,
     trace,
     rbug,
diff --git a/src/gallium/targets/egl/st_GL.c b/src/gallium/targets/egl/st_GL.c
index 676300b0cc1..17b7bf9d481 100644
--- a/src/gallium/targets/egl/st_GL.c
+++ b/src/gallium/targets/egl/st_GL.c
@@ -1,4 +1,5 @@
 #include "state_tracker/st_gl_api.h"
+#include "state_tracker/st_api.h"
 
 #if FEATURE_GL
 PUBLIC struct st_api *
diff --git a/src/gallium/targets/egl/st_GLESv1_CM.c b/src/gallium/targets/egl/st_GLESv1_CM.c
index 0c8de8992f3..c1652d5131a 100644
--- a/src/gallium/targets/egl/st_GLESv1_CM.c
+++ b/src/gallium/targets/egl/st_GLESv1_CM.c
@@ -1,3 +1,4 @@
+#include "state_tracker/st_api.h"
 #include "state_tracker/st_gl_api.h"
 
 PUBLIC struct st_api *
diff --git a/src/gallium/targets/egl/st_GLESv2.c b/src/gallium/targets/egl/st_GLESv2.c
index 87b3e65e239..9c269890089 100644
--- a/src/gallium/targets/egl/st_GLESv2.c
+++ b/src/gallium/targets/egl/st_GLESv2.c
@@ -1,3 +1,4 @@
+#include "state_tracker/st_api.h"
 #include "state_tracker/st_gl_api.h"
 
 PUBLIC struct st_api *
diff --git a/src/gallium/tests/graw/SConscript b/src/gallium/tests/graw/SConscript
index 61121732e38..860a17e13e7 100644
--- a/src/gallium/tests/graw/SConscript
+++ b/src/gallium/tests/graw/SConscript
@@ -11,6 +11,12 @@ env = env.Clone()
 env.Prepend(LIBPATH = [graw.dir])
 env.Prepend(LIBS = ['graw'] + gallium)
 
+if platform in ('freebsd8', 'sunos5'):
+    env.Append(LIBS = ['m'])
+
+if platform == 'freebsd8':
+    env.Append(LIBS = ['pthread'])
+
 progs = [
     'clear',
     'tri',
diff --git a/src/gallium/tests/graw/fs-test.c b/src/gallium/tests/graw/fs-test.c
index dea087357d6..53fbb744d86 100644
--- a/src/gallium/tests/graw/fs-test.c
+++ b/src/gallium/tests/graw/fs-test.c
@@ -13,6 +13,7 @@
 #include "util/u_debug.h"       /* debug_dump_surface_bmp() */
 #include "util/u_inlines.h"
 #include "util/u_memory.h"      /* Offset() */
+#include "util/u_draw_quad.h"
 #include "util/u_box.h"    
 
 static const char *filename = NULL;
@@ -275,7 +276,7 @@ static void draw( void )
    float clear_color[4] = {.1,.3,.5,0};
 
    ctx->clear(ctx, PIPE_CLEAR_COLOR, clear_color, 0, 0);
-   ctx->draw_arrays(ctx, PIPE_PRIM_TRIANGLES, 0, 3);
+   util_draw_arrays(ctx, PIPE_PRIM_TRIANGLES, 0, 3);
    ctx->flush(ctx, PIPE_FLUSH_RENDER_CACHE, NULL);
 
 #if 0
diff --git a/src/gallium/tests/graw/gs-test.c b/src/gallium/tests/graw/gs-test.c
index 3087d446fca..62714900bd9 100644
--- a/src/gallium/tests/graw/gs-test.c
+++ b/src/gallium/tests/graw/gs-test.c
@@ -13,6 +13,7 @@
 #include "util/u_debug.h"       /* debug_dump_surface_bmp() */
 #include "util/u_inlines.h"
 #include "util/u_memory.h"      /* Offset() */
+#include "util/u_draw_quad.h"
 #include "util/u_box.h"    
 
 static const char *filename = NULL;
@@ -336,9 +337,9 @@ static void draw( void )
 
    ctx->clear(ctx, PIPE_CLEAR_COLOR, clear_color, 0, 0);
    if (draw_strip)
-      ctx->draw_arrays(ctx, PIPE_PRIM_TRIANGLE_STRIP, 0, 4);
+      util_draw_arrays(ctx, PIPE_PRIM_TRIANGLE_STRIP, 0, 4);
    else
-      ctx->draw_arrays(ctx, PIPE_PRIM_TRIANGLES, 0, 3);
+      util_draw_arrays(ctx, PIPE_PRIM_TRIANGLES, 0, 3);
 
    ctx->flush(ctx, PIPE_FLUSH_RENDER_CACHE, NULL);
 
diff --git a/src/gallium/tests/graw/quad-tex.c b/src/gallium/tests/graw/quad-tex.c
index 6a0a2ba2950..c50ef12ab5a 100644
--- a/src/gallium/tests/graw/quad-tex.c
+++ b/src/gallium/tests/graw/quad-tex.c
@@ -12,6 +12,7 @@
 #include "util/u_debug.h"       /* debug_dump_surface_bmp() */
 #include "util/u_inlines.h"
 #include "util/u_memory.h"      /* Offset() */
+#include "util/u_draw_quad.h"
 #include "util/u_box.h"    
 
 enum pipe_format formats[] = {
@@ -146,7 +147,7 @@ static void draw( void )
    float clear_color[4] = {.5,.5,.5,1};
 
    ctx->clear(ctx, PIPE_CLEAR_COLOR, clear_color, 0, 0);
-   ctx->draw_arrays(ctx, PIPE_PRIM_QUADS, 0, 4);
+   util_draw_arrays(ctx, PIPE_PRIM_QUADS, 0, 4);
    ctx->flush(ctx, PIPE_FLUSH_RENDER_CACHE, NULL);
 
 #if 0
diff --git a/src/gallium/tests/graw/tri-gs.c b/src/gallium/tests/graw/tri-gs.c
index d187505f8d4..152ae408eb0 100644
--- a/src/gallium/tests/graw/tri-gs.c
+++ b/src/gallium/tests/graw/tri-gs.c
@@ -10,6 +10,7 @@
 
 #include "util/u_debug.h"       /* debug_dump_surface_bmp() */
 #include "util/u_memory.h"      /* Offset() */
+#include "util/u_draw_quad.h"
 
 enum pipe_format formats[] = {
    PIPE_FORMAT_R8G8B8A8_UNORM,
@@ -161,7 +162,7 @@ static void draw( void )
    float clear_color[4] = {1,0,1,1};
 
    ctx->clear(ctx, PIPE_CLEAR_COLOR, clear_color, 0, 0);
-   ctx->draw_arrays(ctx, PIPE_PRIM_TRIANGLES, 0, 3);
+   util_draw_arrays(ctx, PIPE_PRIM_TRIANGLES, 0, 3);
    ctx->flush(ctx, PIPE_FLUSH_RENDER_CACHE, NULL);
 
    screen->flush_frontbuffer(screen, surf, window);
diff --git a/src/gallium/tests/graw/tri-instanced.c b/src/gallium/tests/graw/tri-instanced.c
index 30e205f1434..8859f745fdb 100644
--- a/src/gallium/tests/graw/tri-instanced.c
+++ b/src/gallium/tests/graw/tri-instanced.c
@@ -13,6 +13,7 @@
 
 #include "util/u_debug.h"       /* debug_dump_surface_bmp() */
 #include "util/u_memory.h"      /* Offset() */
+#include "util/u_draw_quad.h"
 
 
 enum pipe_format formats[] = {
@@ -27,7 +28,6 @@ static const int HEIGHT = 300;
 static struct pipe_screen *screen = NULL;
 static struct pipe_context *ctx = NULL;
 static struct pipe_surface *surf = NULL;
-static struct pipe_resource *indexBuffer = NULL;
 static void *window = NULL;
 
 struct vertex {
@@ -105,6 +105,7 @@ static void set_vertices( void )
 {
    struct pipe_vertex_element ve[3];
    struct pipe_vertex_buffer vbuf[2];
+   struct pipe_index_buffer ibuf;
    void *handle;
 
    memset(ve, 0, sizeof ve);
@@ -151,11 +152,14 @@ static void set_vertices( void )
    ctx->set_vertex_buffers(ctx, 2, vbuf);
 
    /* index data */
-   indexBuffer = screen->user_buffer_create(screen,
+   ibuf.buffer = screen->user_buffer_create(screen,
                                             indices,
                                             sizeof(indices),
                                             PIPE_BIND_VERTEX_BUFFER);
+   ibuf.offset = 0;
+   ibuf.index_size = 2;
 
+   ctx->set_index_buffer(ctx, &ibuf);
 
 }
 
@@ -195,18 +199,19 @@ static void set_fragment_shader( void )
 static void draw( void )
 {
    float clear_color[4] = {1,0,1,1};
+   struct pipe_draw_info info;
 
    ctx->clear(ctx, PIPE_CLEAR_COLOR, clear_color, 0, 0);
 
+   util_draw_init_info(&info);
+   info.indexed = (draw_elements != 0);
+   info.mode = PIPE_PRIM_TRIANGLES;
+   info.start = 0;
+   info.count = 3;
    /* draw NUM_INST triangles */
-   if (draw_elements)
-      ctx->draw_elements_instanced(ctx, indexBuffer, 2,
-                                   0, /* indexBias */
-                                   PIPE_PRIM_TRIANGLES,
-                                   0, 3, /* start, count */
-                                   0, NUM_INST); /* startInst, instCount */
-   else
-      ctx->draw_arrays_instanced(ctx, PIPE_PRIM_TRIANGLES, 0, 3, 0, NUM_INST);
+   info.instance_count = NUM_INST;
+
+   ctx->draw_vbo(ctx, &info);
 
    ctx->flush(ctx, PIPE_FLUSH_RENDER_CACHE, NULL);
 
diff --git a/src/gallium/tests/graw/tri.c b/src/gallium/tests/graw/tri.c
index 80377f526d6..4dbd2c062a5 100644
--- a/src/gallium/tests/graw/tri.c
+++ b/src/gallium/tests/graw/tri.c
@@ -10,6 +10,7 @@
 
 #include "util/u_debug.h"       /* debug_dump_surface_bmp() */
 #include "util/u_memory.h"      /* Offset() */
+#include "util/u_draw_quad.h"
 
 enum pipe_format formats[] = {
    PIPE_FORMAT_R8G8B8A8_UNORM,
@@ -134,7 +135,7 @@ static void draw( void )
    float clear_color[4] = {1,0,1,1};
 
    ctx->clear(ctx, PIPE_CLEAR_COLOR, clear_color, 0, 0);
-   ctx->draw_arrays(ctx, PIPE_PRIM_TRIANGLES, 0, 3);
+   util_draw_arrays(ctx, PIPE_PRIM_TRIANGLES, 0, 3);
    ctx->flush(ctx, PIPE_FLUSH_RENDER_CACHE, NULL);
 
 #if 0
diff --git a/src/gallium/tests/graw/vs-test.c b/src/gallium/tests/graw/vs-test.c
index 7f93db42c0d..e1cd814bf72 100644
--- a/src/gallium/tests/graw/vs-test.c
+++ b/src/gallium/tests/graw/vs-test.c
@@ -14,6 +14,7 @@
 #include "util/u_debug.h"       /* debug_dump_surface_bmp() */
 #include "util/u_inlines.h"
 #include "util/u_memory.h"      /* Offset() */
+#include "util/u_draw_quad.h"
 #include "util/u_box.h"    
 
 static const char *filename = NULL;
@@ -226,7 +227,7 @@ static void draw( void )
    float clear_color[4] = {.1,.3,.5,0};
 
    ctx->clear(ctx, PIPE_CLEAR_COLOR, clear_color, 0, 0);
-   ctx->draw_arrays(ctx, PIPE_PRIM_POINTS, 0, Elements(vertices));
+   util_draw_arrays(ctx, PIPE_PRIM_POINTS, 0, Elements(vertices));
    ctx->flush(ctx, PIPE_FLUSH_RENDER_CACHE, NULL);
 
 #if 0
diff --git a/src/gallium/tests/unit/Makefile b/src/gallium/tests/unit/Makefile
index f65958dadd5..345bd1f6941 100644
--- a/src/gallium/tests/unit/Makefile
+++ b/src/gallium/tests/unit/Makefile
@@ -22,7 +22,8 @@ SOURCES = \
 	pipe_barrier_test.c \
 	u_cache_test.c \
 	u_half_test.c \
-	u_format_test.c
+	u_format_test.c \
+	translate_test.c
 
 
 OBJECTS = $(SOURCES:.c=.o)
diff --git a/src/gallium/tests/unit/SConscript b/src/gallium/tests/unit/SConscript
index 8a9f3504c75..edc68e34d9e 100644
--- a/src/gallium/tests/unit/SConscript
+++ b/src/gallium/tests/unit/SConscript
@@ -4,11 +4,18 @@ env = env.Clone()
 
 env.Prepend(LIBS = [gallium])
 
+if platform in ('freebsd8', 'sunos5'):
+    env.Append(LIBS = ['m'])
+
+if platform == 'freebsd8':
+    env.Append(LIBS = ['pthread'])
+
 progs = [
     'pipe_barrier_test',
     'u_cache_test',
     'u_format_test',
-    'u_half_test'
+    'u_half_test',
+    'translate_test'
 ]
 
 for prog in progs:
diff --git a/src/gallium/tests/unit/translate_test.c b/src/gallium/tests/unit/translate_test.c
new file mode 100644
index 00000000000..d0946a91a26
--- /dev/null
+++ b/src/gallium/tests/unit/translate_test.c
@@ -0,0 +1,310 @@
+/**************************************************************************
+ *
+ * Copyright © 2010 Luca Barbieri
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include <stdio.h>
+#include <translate/translate.h>
+#include <util/u_memory.h>
+#include <util/u_format.h>
+#include <util/u_cpu_detect.h>
+#include <rtasm/rtasm_cpu.h>
+
+/* don't use this for serious use */
+static double rand_double()
+{
+   const double rm = (double)RAND_MAX + 1;
+   double div = 1;
+   double v = 0;
+   unsigned i;
+   for(i = 0; i < 4; ++i)
+   {
+      div *= rm;
+      v += (double)rand() / div;
+   }
+   return v;
+}
+
+int main(int argc, char** argv)
+{
+   struct translate *(*create_fn)(const struct translate_key *key) = 0;
+
+   struct translate_key key;
+   unsigned output_format;
+   unsigned input_format;
+   unsigned buffer_size = 4096;
+   unsigned char* buffer[5];
+   unsigned char* byte_buffer;
+   float* float_buffer;
+   double* double_buffer;
+   unsigned count = 4;
+   unsigned i, j, k;
+   unsigned passed = 0;
+   unsigned total = 0;
+   const float error = 0.03125;
+
+   create_fn = 0;
+
+   util_cpu_detect();
+
+   if(argc <= 1)
+   {}
+   else if (!strcmp(argv[1], "generic"))
+      create_fn = translate_generic_create;
+   else if (!strcmp(argv[1], "x86"))
+      create_fn = translate_sse2_create;
+   else if (!strcmp(argv[1], "nosse"))
+   {
+      util_cpu_caps.has_sse = 0;
+      util_cpu_caps.has_sse2 = 0;
+      util_cpu_caps.has_sse3 = 0;
+      util_cpu_caps.has_sse4_1 = 0;
+      create_fn = translate_sse2_create;
+   }
+   else if (!strcmp(argv[1], "sse"))
+   {
+      if(!util_cpu_caps.has_sse || !rtasm_cpu_has_sse())
+      {
+         printf("Error: CPU doesn't support SSE (test with qemu)\n");
+         return 2;
+      }
+      util_cpu_caps.has_sse2 = 0;
+      util_cpu_caps.has_sse3 = 0;
+      util_cpu_caps.has_sse4_1 = 0;
+      create_fn = translate_sse2_create;
+   }
+   else if (!strcmp(argv[1], "sse2"))
+   {
+      if(!util_cpu_caps.has_sse2 || !rtasm_cpu_has_sse())
+      {
+         printf("Error: CPU doesn't support SSE2 (test with qemu)\n");
+         return 2;
+      }
+      util_cpu_caps.has_sse3 = 0;
+      util_cpu_caps.has_sse4_1 = 0;
+      create_fn = translate_sse2_create;
+   }
+   else if (!strcmp(argv[1], "sse3"))
+   {
+      if(!util_cpu_caps.has_sse3 || !rtasm_cpu_has_sse())
+      {
+         printf("Error: CPU doesn't support SSE3 (test with qemu)\n");
+         return 2;
+      }
+      util_cpu_caps.has_sse4_1 = 0;
+      create_fn = translate_sse2_create;
+   }
+   else if (!strcmp(argv[1], "sse4.1"))
+   {
+      if(!util_cpu_caps.has_sse4_1 || !rtasm_cpu_has_sse())
+      {
+         printf("Error: CPU doesn't support SSE4.1 (test with qemu)\n");
+         return 2;
+      }
+      create_fn = translate_sse2_create;
+   }
+
+   if (!create_fn)
+   {
+      printf("Usage: ./translate_test [generic|x86|nosse|sse|sse2|sse3|sse4.1]\n");
+      return 2;
+   }
+
+   for (i = 1; i < Elements(buffer); ++i)
+      buffer[i] = align_malloc(buffer_size, 4096);
+
+   byte_buffer = align_malloc(buffer_size, 4096);
+   float_buffer = align_malloc(buffer_size, 4096);
+   double_buffer = align_malloc(buffer_size, 4096);
+
+   key.nr_elements = 1;
+   key.element[0].input_buffer = 0;
+   key.element[0].input_offset = 0;
+   key.element[0].output_offset = 0;
+   key.element[0].type = TRANSLATE_ELEMENT_NORMAL;
+   key.element[0].instance_divisor = 0;
+
+   srand(4359025);
+
+   /* avoid negative values that work badly when converted to unsigned format*/
+   for (i = 0; i < buffer_size; ++i)
+      byte_buffer[i] = rand() & 0x7f7f7f7f;
+
+   for (i = 0; i < buffer_size / sizeof(float); ++i)
+      float_buffer[i] = (float)rand_double();
+
+   for (i = 0; i < buffer_size / sizeof(double); ++i)
+      double_buffer[i] = rand_double();
+
+   for (output_format = 1; output_format < PIPE_FORMAT_COUNT; ++output_format)
+   {
+      const struct util_format_description* output_format_desc = util_format_description(output_format);
+      unsigned output_format_size;
+      unsigned output_normalized = 0;
+
+      if (!output_format_desc
+            || !output_format_desc->fetch_rgba_float
+            || !output_format_desc->pack_rgba_float
+            || output_format_desc->colorspace != UTIL_FORMAT_COLORSPACE_RGB
+            || output_format_desc->layout != UTIL_FORMAT_LAYOUT_PLAIN
+            || !translate_is_output_format_supported(output_format))
+         continue;
+
+      for(i = 0; i < output_format_desc->nr_channels; ++i)
+      {
+         if(output_format_desc->channel[i].type != UTIL_FORMAT_TYPE_FLOAT)
+            output_normalized |= (1 << output_format_desc->channel[i].normalized);
+      }
+
+      output_format_size = util_format_get_stride(output_format, 1);
+
+      for (input_format = 1; input_format < PIPE_FORMAT_COUNT; ++input_format)
+      {
+         const struct util_format_description* input_format_desc = util_format_description(input_format);
+         unsigned input_format_size;
+         struct translate* translate[2];
+         unsigned fail = 0;
+         unsigned used_generic = 0;
+         unsigned input_normalized = 0;
+         boolean input_is_float = FALSE;
+
+         if (!input_format_desc
+               || !input_format_desc->fetch_rgba_float
+               || !input_format_desc->pack_rgba_float
+               || input_format_desc->colorspace != UTIL_FORMAT_COLORSPACE_RGB
+               || input_format_desc->layout != UTIL_FORMAT_LAYOUT_PLAIN
+               || !translate_is_output_format_supported(input_format))
+            continue;
+
+         input_format_size = util_format_get_stride(input_format, 1);
+
+         for(i = 0; i < input_format_desc->nr_channels; ++i)
+         {
+            if(input_format_desc->channel[i].type == UTIL_FORMAT_TYPE_FLOAT)
+            {
+               input_is_float = 1;
+               input_normalized |= 1 << 1;
+            }
+            else
+               input_normalized |= (1 << input_format_desc->channel[i].normalized);
+         }
+
+         if(((input_normalized | output_normalized) == 3)
+               || ((input_normalized & 1) && (output_normalized & 1)
+                     && input_format_size * output_format_desc->nr_channels > output_format_size * input_format_desc->nr_channels))
+            continue;
+
+         key.element[0].input_format = input_format;
+         key.element[0].output_format = output_format;
+         key.output_stride = output_format_size;
+         translate[0] = create_fn(&key);
+         if (!translate[0])
+            continue;
+
+         key.element[0].input_format = output_format;
+         key.element[0].output_format = input_format;
+         key.output_stride = input_format_size;
+         translate[1] = create_fn(&key);
+         if(!translate[1])
+         {
+            used_generic = 1;
+            translate[1] = translate_generic_create(&key);
+            if(!translate[1])
+               continue;
+         }
+
+         for(i = 1; i < 5; ++i)
+            memset(buffer[i], 0xcd - (0x22 * i), 4096);
+
+         if(input_is_float && input_format_desc->channel[0].size == 32)
+            buffer[0] = (unsigned char*)float_buffer;
+         else if(input_is_float && input_format_desc->channel[0].size == 64)
+            buffer[0] = (unsigned char*)double_buffer;
+         else if(input_is_float)
+            abort();
+         else
+            buffer[0] = byte_buffer;
+
+         translate[0]->set_buffer(translate[0], 0, buffer[0], input_format_size, ~0);
+         translate[0]->run(translate[0], 0, count, 0, buffer[1]);
+         translate[1]->set_buffer(translate[1], 0, buffer[1], output_format_size, ~0);
+         translate[1]->run(translate[1], 0, count, 0, buffer[2]);
+         translate[0]->set_buffer(translate[0], 0, buffer[2], input_format_size, ~0);
+         translate[0]->run(translate[0], 0, count, 0, buffer[3]);
+         translate[1]->set_buffer(translate[1], 0, buffer[3], output_format_size, ~0);
+         translate[1]->run(translate[1], 0, count, 0, buffer[4]);
+
+         for (i = 0; i < count; ++i)
+         {
+            float a[4];
+            float b[4];
+            input_format_desc->fetch_rgba_float(a, buffer[2] + i * input_format_size, 0, 0);
+            input_format_desc->fetch_rgba_float(b, buffer[4] + i * input_format_size, 0, 0);
+
+            for (j = 0; j < count; ++j)
+            {
+               float d = a[j] - b[j];
+               if (d > error || d < -error)
+               {
+                  fail = 1;
+                  break;
+               }
+            }
+         }
+
+         printf("%s%s: %s -> %s -> %s -> %s -> %s\n",
+               fail ? "FAIL" : "PASS",
+               used_generic ? "[GENERIC]" : "",
+               input_format_desc->name, output_format_desc->name, input_format_desc->name, output_format_desc->name, input_format_desc->name);
+
+         if (1)
+         {
+            for (i = 0; i < Elements(buffer); ++i)
+            {
+               unsigned format_size = (i & 1) ? output_format_size : input_format_size;
+               printf("%c ", (i == 2 || i == 4) ? '*' : ' ');
+               for (j = 0; j < count; ++j)
+               {
+                  for (k = 0; k < format_size; ++k)
+                  {
+                     printf("%02x", buffer[i][j * format_size + k]);
+                  }
+                  printf(" ");
+               }
+               printf("\n");
+            }
+         }
+
+         if (!fail)
+            ++passed;
+         ++total;
+
+         if(translate[1])
+            translate[1]->release(translate[1]);
+         translate[0]->release(translate[0]);
+      }
+   }
+
+   printf("%u/%u tests passed for translate_%s\n", passed, total, argv[1]);
+   return passed != total;
+}
diff --git a/src/gallium/winsys/r600/drm/SConscript b/src/gallium/winsys/r600/drm/SConscript
new file mode 100644
index 00000000000..2f20d9f8957
--- /dev/null
+++ b/src/gallium/winsys/r600/drm/SConscript
@@ -0,0 +1,25 @@
+Import('*')
+
+env = env.Clone()
+
+r600_sources = [
+    'bof.c',
+    'r600_state.c',
+    'radeon_ctx.c',
+    'radeon_draw.c',
+    'radeon_state.c',
+    'radeon_bo.c',
+    'radeon_pciid.c',
+    'radeon.c',
+    'r600_drm.c'
+]
+
+env.ParseConfig('pkg-config --cflags libdrm_radeon')
+env.Append(CPPPATH = '#/src/gallium/drivers/r600')
+
+r600winsys = env.ConvenienceLibrary(
+    target ='r600winsys',
+    source = r600_sources,
+)
+
+Export('r600winsys')
diff --git a/src/gallium/winsys/r600/drm/r600_drm.c b/src/gallium/winsys/r600/drm/r600_drm.c
index 3d87a994c15..c76e7f5fa51 100644
--- a/src/gallium/winsys/r600/drm/r600_drm.c
+++ b/src/gallium/winsys/r600/drm/r600_drm.c
@@ -30,8 +30,7 @@
 #include "util/u_debug.h"
 #include "radeon_priv.h"
 #include "r600_screen.h"
-#include "r600_texture.h"
-#include "r600_public.h"
+#include "r600_resource.h"
 #include "r600_drm_public.h"
 #include "state_tracker/drm_driver.h"
 
@@ -45,7 +44,7 @@ boolean r600_buffer_get_handle(struct radeon *rw,
 			       struct winsys_handle *whandle)
 {
 	struct drm_gem_flink flink;
-	struct r600_buffer* rbuffer = (struct r600_buffer*)buf;
+	struct r600_resource* rbuffer = (struct r600_resource*)buf;
 
 	if (whandle->type == DRM_API_HANDLE_TYPE_SHARED) {
 		if (!rbuffer->flink) {
diff --git a/src/gallium/winsys/r600/drm/r600_states.h b/src/gallium/winsys/r600/drm/r600_states.h
index 5896df21b21..e40c77d8f6c 100644
--- a/src/gallium/winsys/r600/drm/r600_states.h
+++ b/src/gallium/winsys/r600/drm/r600_states.h
@@ -372,6 +372,76 @@ static const struct radeon_register R600_CB0_names[] = {
 	{0x00028100, 0, 0, "CB_COLOR0_MASK"},
 };
 
+static const struct radeon_register R600_CB1_names[] = {
+	{0x00028044, 1, 0, "CB_COLOR1_BASE"},
+	{0x000280A4, 0, 0, "CB_COLOR1_INFO"},
+	{0x00028064, 0, 0, "CB_COLOR1_SIZE"},
+	{0x00028084, 0, 0, "CB_COLOR1_VIEW"},
+	{0x000280E4, 1, 1, "CB_COLOR1_FRAG"},
+	{0x000280C4, 1, 2, "CB_COLOR1_TILE"},
+	{0x00028104, 0, 0, "CB_COLOR1_MASK"},
+};
+
+static const struct radeon_register R600_CB2_names[] = {
+	{0x00028048, 1, 0, "CB_COLOR2_BASE"},
+	{0x000280A8, 0, 0, "CB_COLOR2_INFO"},
+	{0x00028068, 0, 0, "CB_COLOR2_SIZE"},
+	{0x00028088, 0, 0, "CB_COLOR2_VIEW"},
+	{0x000280E8, 1, 1, "CB_COLOR2_FRAG"},
+	{0x000280C8, 1, 2, "CB_COLOR2_TILE"},
+	{0x00028108, 0, 0, "CB_COLOR2_MASK"},
+};
+
+static const struct radeon_register R600_CB3_names[] = {
+	{0x0002804C, 1, 0, "CB_COLOR3_BASE"},
+	{0x000280AC, 0, 0, "CB_COLOR3_INFO"},
+	{0x0002806C, 0, 0, "CB_COLOR3_SIZE"},
+	{0x0002808C, 0, 0, "CB_COLOR3_VIEW"},
+	{0x000280EC, 1, 1, "CB_COLOR3_FRAG"},
+	{0x000280CC, 1, 2, "CB_COLOR3_TILE"},
+	{0x0002810C, 0, 0, "CB_COLOR3_MASK"},
+};
+
+static const struct radeon_register R600_CB4_names[] = {
+	{0x00028050, 1, 0, "CB_COLOR4_BASE"},
+	{0x000280B0, 0, 0, "CB_COLOR4_INFO"},
+	{0x00028070, 0, 0, "CB_COLOR4_SIZE"},
+	{0x00028090, 0, 0, "CB_COLOR4_VIEW"},
+	{0x000280F0, 1, 1, "CB_COLOR4_FRAG"},
+	{0x000280D0, 1, 2, "CB_COLOR4_TILE"},
+	{0x00028110, 0, 0, "CB_COLOR4_MASK"},
+};
+
+static const struct radeon_register R600_CB5_names[] = {
+	{0x00028054, 1, 0, "CB_COLOR5_BASE"},
+	{0x000280B4, 0, 0, "CB_COLOR5_INFO"},
+	{0x00028074, 0, 0, "CB_COLOR5_SIZE"},
+	{0x00028094, 0, 0, "CB_COLOR5_VIEW"},
+	{0x000280F4, 1, 1, "CB_COLOR5_FRAG"},
+	{0x000280D4, 1, 2, "CB_COLOR5_TILE"},
+	{0x00028114, 0, 0, "CB_COLOR5_MASK"},
+};
+
+static const struct radeon_register R600_CB6_names[] = {
+	{0x00028058, 1, 0, "CB_COLOR6_BASE"},
+	{0x000280B8, 0, 0, "CB_COLOR6_INFO"},
+	{0x00028078, 0, 0, "CB_COLOR6_SIZE"},
+	{0x00028098, 0, 0, "CB_COLOR6_VIEW"},
+	{0x000280F8, 1, 1, "CB_COLOR6_FRAG"},
+	{0x000280D8, 1, 2, "CB_COLOR6_TILE"},
+	{0x00028118, 0, 0, "CB_COLOR6_MASK"},
+};
+
+static const struct radeon_register R600_CB7_names[] = {
+	{0x0002805C, 1, 0, "CB_COLOR7_BASE"},
+	{0x000280BC, 0, 0, "CB_COLOR7_INFO"},
+	{0x0002807C, 0, 0, "CB_COLOR7_SIZE"},
+	{0x0002809C, 0, 0, "CB_COLOR7_VIEW"},
+	{0x000280FC, 1, 1, "CB_COLOR7_FRAG"},
+	{0x000280DC, 1, 2, "CB_COLOR7_TILE"},
+	{0x0002811C, 0, 0, "CB_COLOR7_MASK"},
+};
+
 static const struct radeon_register R600_DB_names[] = {
 	{0x0002800C, 1, 0, "DB_DEPTH_BASE"},
 	{0x00028000, 0, 0, "DB_DEPTH_SIZE"},
@@ -425,9 +495,16 @@ static struct radeon_type R600_types[] = {
 	{ 128, 1233, 0x0000A600, 0x0000A720, 0x0010, 0, "R600_VS_SAMPLER_BORDER", 4, r600_state_pm4_generic, R600_VS_SAMPLER_BORDER_names},
 	{ 128, 1251, 0x0000A800, 0x0000A920, 0x0010, 0, "R600_GS_SAMPLER_BORDER", 4, r600_state_pm4_generic, R600_GS_SAMPLER_BORDER_names},
 	{ 128, 1269, 0x00000000, 0x00000000, 0x0000, 0, "R600_CB0", 7, r600_state_pm4_cb0, R600_CB0_names},
-	{ 128, 1270, 0x00000000, 0x00000000, 0x0000, 0, "R600_DB", 6, r600_state_pm4_db, R600_DB_names},
-	{ 128, 1271, 0x00000000, 0x00000000, 0x0000, 0, "R600_VGT", 11, r600_state_pm4_vgt, R600_VGT_names},
-	{ 128, 1272, 0x00000000, 0x00000000, 0x0000, 0, "R600_DRAW", 4, r600_state_pm4_draw, R600_DRAW_names},
+	{ 128, 1270, 0x00000000, 0x00000000, 0x0000, 0, "R600_CB1", 7, r600_state_pm4_cb0, R600_CB1_names},
+	{ 128, 1271, 0x00000000, 0x00000000, 0x0000, 0, "R600_CB2", 7, r600_state_pm4_cb0, R600_CB2_names},
+	{ 128, 1272, 0x00000000, 0x00000000, 0x0000, 0, "R600_CB3", 7, r600_state_pm4_cb0, R600_CB3_names},
+	{ 128, 1273, 0x00000000, 0x00000000, 0x0000, 0, "R600_CB4", 7, r600_state_pm4_cb0, R600_CB4_names},
+	{ 128, 1274, 0x00000000, 0x00000000, 0x0000, 0, "R600_CB5", 7, r600_state_pm4_cb0, R600_CB5_names},
+	{ 128, 1275, 0x00000000, 0x00000000, 0x0000, 0, "R600_CB6", 7, r600_state_pm4_cb0, R600_CB6_names},
+	{ 128, 1276, 0x00000000, 0x00000000, 0x0000, 0, "R600_CB7", 7, r600_state_pm4_cb0, R600_CB7_names},
+	{ 128, 1277, 0x00000000, 0x00000000, 0x0000, 0, "R600_DB", 6, r600_state_pm4_db, R600_DB_names},
+	{ 128, 1278, 0x00000000, 0x00000000, 0x0000, 0, "R600_VGT", 11, r600_state_pm4_vgt, R600_VGT_names},
+	{ 128, 1279, 0x00000000, 0x00000000, 0x0000, 0, "R600_DRAW", 4, r600_state_pm4_draw, R600_DRAW_names},
 };
 
 static struct radeon_type R700_types[] = {
@@ -453,9 +530,16 @@ static struct radeon_type R700_types[] = {
 	{ 128, 1233, 0x0000A600, 0x0000A720, 0x0010, 0, "R600_VS_SAMPLER_BORDER", 4, r600_state_pm4_generic, R600_VS_SAMPLER_BORDER_names},
 	{ 128, 1251, 0x0000A800, 0x0000A920, 0x0010, 0, "R600_GS_SAMPLER_BORDER", 4, r600_state_pm4_generic, R600_GS_SAMPLER_BORDER_names},
 	{ 128, 1269, 0x00000000, 0x00000000, 0x0000, 0, "R600_CB0", 7, r700_state_pm4_cb0, R600_CB0_names},
-	{ 128, 1270, 0x00000000, 0x00000000, 0x0000, 0, "R600_DB", 6, r700_state_pm4_db, R600_DB_names},
-	{ 128, 1271, 0x00000000, 0x00000000, 0x0000, 0, "R600_VGT", 11, r600_state_pm4_vgt, R600_VGT_names},
-	{ 128, 1272, 0x00000000, 0x00000000, 0x0000, 0, "R600_DRAW", 4, r600_state_pm4_draw, R600_DRAW_names},
+	{ 128, 1270, 0x00000000, 0x00000000, 0x0000, 0, "R600_CB1", 7, r600_state_pm4_cb0, R600_CB1_names},
+	{ 128, 1271, 0x00000000, 0x00000000, 0x0000, 0, "R600_CB2", 7, r600_state_pm4_cb0, R600_CB2_names},
+	{ 128, 1272, 0x00000000, 0x00000000, 0x0000, 0, "R600_CB3", 7, r600_state_pm4_cb0, R600_CB3_names},
+	{ 128, 1273, 0x00000000, 0x00000000, 0x0000, 0, "R600_CB4", 7, r600_state_pm4_cb0, R600_CB4_names},
+	{ 128, 1274, 0x00000000, 0x00000000, 0x0000, 0, "R600_CB5", 7, r600_state_pm4_cb0, R600_CB5_names},
+	{ 128, 1275, 0x00000000, 0x00000000, 0x0000, 0, "R600_CB6", 7, r600_state_pm4_cb0, R600_CB6_names},
+	{ 128, 1276, 0x00000000, 0x00000000, 0x0000, 0, "R600_CB7", 7, r600_state_pm4_cb0, R600_CB7_names},
+	{ 128, 1277, 0x00000000, 0x00000000, 0x0000, 0, "R600_DB", 6, r700_state_pm4_db, R600_DB_names},
+	{ 128, 1278, 0x00000000, 0x00000000, 0x0000, 0, "R600_VGT", 11, r600_state_pm4_vgt, R600_VGT_names},
+	{ 128, 1279, 0x00000000, 0x00000000, 0x0000, 0, "R600_DRAW", 4, r600_state_pm4_draw, R600_DRAW_names},
 };
 
 #endif
diff --git a/src/gallium/winsys/r600/drm/radeon.c b/src/gallium/winsys/r600/drm/radeon.c
index 7e656698064..80b0a1d3972 100644
--- a/src/gallium/winsys/r600/drm/radeon.c
+++ b/src/gallium/winsys/r600/drm/radeon.c
@@ -23,7 +23,6 @@
 #include "xf86drm.h"
 #include "radeon_priv.h"
 #include "radeon_drm.h"
-#include "r600d.h"
 
 enum radeon_family radeon_get_family(struct radeon *radeon)
 {
diff --git a/src/gallium/winsys/r600/drm/radeon_ctx.c b/src/gallium/winsys/r600/drm/radeon_ctx.c
index 6b0eba0b289..45b706bb0f9 100644
--- a/src/gallium/winsys/r600/drm/radeon_ctx.c
+++ b/src/gallium/winsys/r600/drm/radeon_ctx.c
@@ -112,6 +112,7 @@ struct radeon_ctx *radeon_ctx_decref(struct radeon_ctx *ctx)
 		ctx->bo[i] = radeon_bo_decref(ctx->radeon, ctx->bo[i]);
 	}
 	ctx->radeon = radeon_decref(ctx->radeon);
+	free(ctx->state);
 	free(ctx->draw);
 	free(ctx->bo);
 	free(ctx->pm4);
@@ -151,6 +152,8 @@ int radeon_ctx_submit(struct radeon_ctx *ctx)
 	uint64_t chunk_array[2];
 	int r = 0;
 
+	if (!ctx->cpm4)
+		return 0;
 #if 0
 	for (r = 0; r < ctx->cpm4; r++) {
 		fprintf(stderr, "0x%08X\n", ctx->pm4[r]);
diff --git a/src/gallium/winsys/r600/drm/radeon_priv.h b/src/gallium/winsys/r600/drm/radeon_priv.h
index b91421f4389..96c0d060f7e 100644
--- a/src/gallium/winsys/r600/drm/radeon_priv.h
+++ b/src/gallium/winsys/r600/drm/radeon_priv.h
@@ -68,36 +68,6 @@ extern int radeon_is_family_compatible(unsigned family1, unsigned family2);
 extern int radeon_reg_id(struct radeon *radeon, unsigned offset, unsigned *typeid, unsigned *stateid, unsigned *id);
 extern unsigned radeon_type_from_id(struct radeon *radeon, unsigned id);
 
-/*
- * radeon context functions
- */
-#pragma pack(1)
-struct radeon_cs_reloc {
-	uint32_t	handle;
-	uint32_t	read_domain;
-	uint32_t	write_domain;
-	uint32_t	flags;
-};
-#pragma pack()
-
-struct radeon_ctx {
-	int				refcount;
-	struct radeon			*radeon;
-	u32				*pm4;
-	u32				cpm4;
-	u32				draw_cpm4;
-	unsigned			id;
-	unsigned			next_id;
-	unsigned			nreloc;
-	struct radeon_cs_reloc		*reloc;
-	unsigned			nbo;
-	struct radeon_bo		**bo;
-	unsigned			ndraw;
-	struct radeon_draw		*cdraw;
-	struct radeon_draw		**draw;
-	unsigned			nstate;
-	struct radeon_state		**state;
-};
 
 int radeon_ctx_set_bo_new(struct radeon_ctx *ctx, struct radeon_bo *bo);
 struct radeon_bo *radeon_ctx_get_bo(struct radeon_ctx *ctx, unsigned reloc);
diff --git a/src/gallium/winsys/radeon/drm/radeon_drm.c b/src/gallium/winsys/radeon/drm/radeon_drm.c
index e9a276362f3..86d4f949697 100644
--- a/src/gallium/winsys/radeon/drm/radeon_drm.c
+++ b/src/gallium/winsys/radeon/drm/radeon_drm.c
@@ -39,7 +39,6 @@
 #include "util/u_memory.h"
 
 #include "xf86drm.h"
-#include <sys/ioctl.h>
 
 static struct radeon_libdrm_winsys *
 radeon_winsys_create(int fd)
@@ -55,6 +54,31 @@ radeon_winsys_create(int fd)
     return rws;
 }
 
+/* Enable/disable Hyper-Z access. Return TRUE on success. */
+static boolean radeon_set_hyperz_access(int fd, boolean enable)
+{
+#ifndef RADEON_INFO_WANT_HYPERZ
+#define RADEON_INFO_WANT_HYPERZ 7
+#endif
+
+    struct drm_radeon_info info = {0};
+    unsigned value = enable ? 1 : 0;
+
+    if (!debug_get_bool_option("RADEON_HYPERZ", FALSE))
+        return FALSE;
+
+    info.value = (unsigned long)&value;
+    info.request = RADEON_INFO_WANT_HYPERZ;
+
+    if (drmCommandWriteRead(fd, DRM_RADEON_INFO, &info, sizeof(info)) != 0)
+        return FALSE;
+
+    if (enable && !value)
+        return FALSE;
+
+    return TRUE;
+}
+
 /* Helper function to do the ioctls needed for setup and init. */
 static void do_ioctls(int fd, struct radeon_libdrm_winsys* winsys)
 {
@@ -103,6 +127,10 @@ static void do_ioctls(int fd, struct radeon_libdrm_winsys* winsys)
     winsys->drm_2_3_0 = version->version_major > 2 ||
                         version->version_minor >= 3;
 
+    winsys->drm_2_6_0 = version->version_major > 2 ||
+                        (version->version_major == 2 &&
+                         version->version_minor >= 6);
+
     info.request = RADEON_INFO_DEVICE_ID;
     retval = drmCommandWriteRead(fd, DRM_RADEON_INFO, &info, sizeof(info));
     if (retval) {
@@ -130,6 +158,8 @@ static void do_ioctls(int fd, struct radeon_libdrm_winsys* winsys)
     }
     winsys->z_pipes = target;
 
+    winsys->hyperz = radeon_set_hyperz_access(fd, TRUE);
+
     retval = drmCommandWriteRead(fd, DRM_RADEON_GEM_INFO,
             &gem_info, sizeof(gem_info));
     if (retval) {
@@ -142,12 +172,14 @@ static void do_ioctls(int fd, struct radeon_libdrm_winsys* winsys)
 
     debug_printf("radeon: Successfully grabbed chipset info from kernel!\n"
                  "radeon: DRM version: %d.%d.%d ID: 0x%04x GB: %d Z: %d\n"
-                 "radeon: GART size: %d MB VRAM size: %d MB\n",
+                 "radeon: GART size: %d MB VRAM size: %d MB\n"
+                 "radeon: HyperZ: %s\n",
                  version->version_major, version->version_minor,
                  version->version_patchlevel, winsys->pci_id,
                  winsys->gb_pipes, winsys->z_pipes,
                  winsys->gart_size / 1024 / 1024,
-                 winsys->vram_size / 1024 / 1024);
+                 winsys->vram_size / 1024 / 1024,
+                 winsys->hyperz ? "YES" : "NO");
 
     drmFreeVersion(version);
 }
diff --git a/src/gallium/winsys/radeon/drm/radeon_drm_buffer.c b/src/gallium/winsys/radeon/drm/radeon_drm_buffer.c
index 5ea5912089e..017eac8464e 100644
--- a/src/gallium/winsys/radeon/drm/radeon_drm_buffer.c
+++ b/src/gallium/winsys/radeon/drm/radeon_drm_buffer.c
@@ -189,7 +189,7 @@ struct pb_buffer *radeon_drm_bufmgr_create_buffer_from_handle(struct pb_manager
     pipe_reference_init(&buf->base.base.reference, 1);
     buf->base.base.alignment = 0;
     buf->base.base.usage = PB_USAGE_GPU_WRITE | PB_USAGE_GPU_READ;
-    buf->base.base.size = 0;
+    buf->base.base.size = bo->size;
     buf->base.vtbl = &radeon_drm_buffer_vtbl;
     buf->mgr = mgr;
 
diff --git a/src/gallium/winsys/radeon/drm/radeon_r300.c b/src/gallium/winsys/radeon/drm/radeon_r300.c
index effa27f5c72..58400986421 100644
--- a/src/gallium/winsys/radeon/drm/radeon_r300.c
+++ b/src/gallium/winsys/radeon/drm/radeon_r300.c
@@ -109,14 +109,19 @@ static void radeon_r300_winsys_buffer_reference(struct r300_winsys_screen *rws,
 
 static struct r300_winsys_buffer *radeon_r300_winsys_buffer_from_handle(struct r300_winsys_screen *rws,
                                                                         struct winsys_handle *whandle,
-                                                                        unsigned *stride)
+                                                                        unsigned *stride,
+                                                                        unsigned *size)
 {
     struct radeon_libdrm_winsys *ws = radeon_libdrm_winsys(rws);
     struct pb_buffer *_buf;
 
-    *stride = whandle->stride;
-
     _buf = radeon_drm_bufmgr_create_buffer_from_handle(ws->kman, whandle->handle);
+
+    if (stride)
+        *stride = whandle->stride;
+    if (size)
+        *size = _buf->base.size;
+
     return radeon_libdrm_winsys_buffer(_buf);
 }
 
@@ -206,6 +211,10 @@ static uint32_t radeon_get_value(struct r300_winsys_screen *rws,
         return ws->squaretiling;
     case R300_VID_DRM_2_3_0:
         return ws->drm_2_3_0;
+    case R300_VID_DRM_2_6_0:
+        return ws->drm_2_6_0;
+    case R300_CAN_HYPERZ:
+        return ws->hyperz;
     }
     return 0;
 }
diff --git a/src/gallium/winsys/radeon/drm/radeon_winsys.h b/src/gallium/winsys/radeon/drm/radeon_winsys.h
index 533b7b2e2d2..6f4aa4bce30 100644
--- a/src/gallium/winsys/radeon/drm/radeon_winsys.h
+++ b/src/gallium/winsys/radeon/drm/radeon_winsys.h
@@ -65,6 +65,15 @@ struct radeon_libdrm_winsys {
      */
     boolean drm_2_3_0;
 
+    /* DRM 2.6.0
+     *   - Hyper-Z
+     *   - GB_Z_PEQ_CONFIG allowed on rv350->r4xx, we should initialize it
+     */
+    boolean drm_2_6_0;
+
+    /* hyperz user */
+    boolean hyperz;
+
     /* DRM FD */
     int fd;
 
diff --git a/src/gallium/winsys/svga/drm/vmw_screen_dri.c b/src/gallium/winsys/svga/drm/vmw_screen_dri.c
index 1b0d10f60d6..7bd4407e9f1 100644
--- a/src/gallium/winsys/svga/drm/vmw_screen_dri.c
+++ b/src/gallium/winsys/svga/drm/vmw_screen_dri.c
@@ -32,8 +32,6 @@
 
 #include "vmw_screen.h"
 #include "vmw_surface.h"
-#include "vmw_fence.h"
-#include "vmw_context.h"
 #include "svga_drm_public.h"
 
 #include "state_tracker/drm_driver.h"
diff --git a/src/glsl/apps/compile.c b/src/glsl/apps/compile.c
index 3aa4fd4d53e..5114fc9d0be 100644
--- a/src/glsl/apps/compile.c
+++ b/src/glsl/apps/compile.c
@@ -30,6 +30,7 @@
 #include <string.h>
 #include <assert.h>
 #include "../pp/sl_pp_public.h"
+#include "../pp/sl_pp_purify.h"
 #include "../cl/sl_cl_parse.h"
 
 
diff --git a/src/glsl/apps/process.c b/src/glsl/apps/process.c
index caf72a71cf1..6d5ce6eea3f 100644
--- a/src/glsl/apps/process.c
+++ b/src/glsl/apps/process.c
@@ -30,6 +30,8 @@
 #include <string.h>
 #include <assert.h>
 #include "../pp/sl_pp_public.h"
+#include "../pp/sl_pp_purify.h"
+#include "../pp/sl_pp_token.h"
 
 
 int
diff --git a/src/glsl/apps/purify.c b/src/glsl/apps/purify.c
index 0f09b157efd..e3fca59ab45 100644
--- a/src/glsl/apps/purify.c
+++ b/src/glsl/apps/purify.c
@@ -30,6 +30,7 @@
 #include <stdlib.h>
 #include <string.h>
 #include "../pp/sl_pp_public.h"
+#include "../pp/sl_pp_purify.h"
 
 
 int
diff --git a/src/glsl/apps/tokenise.c b/src/glsl/apps/tokenise.c
index f89f47d0611..3d68334bed3 100644
--- a/src/glsl/apps/tokenise.c
+++ b/src/glsl/apps/tokenise.c
@@ -30,6 +30,8 @@
 #include <string.h>
 #include <assert.h>
 #include "../pp/sl_pp_public.h"
+#include "../pp/sl_pp_purify.h"
+#include "../pp/sl_pp_token.h"
 
 
 int
diff --git a/src/glsl/apps/version.c b/src/glsl/apps/version.c
index fa5c226da83..8506f35ba11 100644
--- a/src/glsl/apps/version.c
+++ b/src/glsl/apps/version.c
@@ -30,6 +30,7 @@
 #include <string.h>
 #include <assert.h>
 #include "../pp/sl_pp_public.h"
+#include "../pp/sl_pp_purify.h"
 
 
 int
diff --git a/src/glsl/cl/sl_cl_parse.c b/src/glsl/cl/sl_cl_parse.c
index 09456f5219a..c1bc6031ce6 100644
--- a/src/glsl/cl/sl_cl_parse.c
+++ b/src/glsl/cl/sl_cl_parse.c
@@ -29,6 +29,7 @@
 #include <stdlib.h>
 #include <string.h>
 #include "../pp/sl_pp_public.h"
+#include "../pp/sl_pp_token.h"
 #include "sl_cl_parse.h"
 
 
diff --git a/src/glsl/cl/sl_cl_parse.h b/src/glsl/cl/sl_cl_parse.h
index dd5791d5901..a954d439276 100644
--- a/src/glsl/cl/sl_cl_parse.h
+++ b/src/glsl/cl/sl_cl_parse.h
@@ -28,6 +28,8 @@
 #ifndef SL_CL_PARSE_H
 #define SL_CL_PARSE_H
 
+struct sl_pp_context;
+
 int
 sl_cl_compile(struct sl_pp_context *context,
               unsigned int shader_type,
diff --git a/src/glsl/pp/sl_pp_context.c b/src/glsl/pp/sl_pp_context.c
index 74a9bdddfdc..b8e1e99fc86 100644
--- a/src/glsl/pp/sl_pp_context.c
+++ b/src/glsl/pp/sl_pp_context.c
@@ -27,6 +27,7 @@
 
 #include <stdlib.h>
 #include <string.h>
+#include "sl_pp_macro.h"
 #include "sl_pp_public.h"
 #include "sl_pp_context.h"
 
diff --git a/src/glsl/pp/sl_pp_context.h b/src/glsl/pp/sl_pp_context.h
index 8abb9708b85..e6244f62575 100644
--- a/src/glsl/pp/sl_pp_context.h
+++ b/src/glsl/pp/sl_pp_context.h
@@ -29,7 +29,6 @@
 #define SL_PP_CONTEXT_H
 
 #include "sl_pp_dict.h"
-#include "sl_pp_macro.h"
 #include "sl_pp_process.h"
 #include "sl_pp_purify.h"
 #include "sl_pp_token_util.h"
diff --git a/src/glsl/pp/sl_pp_define.c b/src/glsl/pp/sl_pp_define.c
index 808a6a0d4f1..370e6aa6606 100644
--- a/src/glsl/pp/sl_pp_define.c
+++ b/src/glsl/pp/sl_pp_define.c
@@ -28,8 +28,10 @@
 #include <stdlib.h>
 #include <string.h>
 #include "sl_pp_context.h"
+#include "sl_pp_macro.h"
 #include "sl_pp_process.h"
 #include "sl_pp_public.h"
+#include "sl_pp_token.h"
 
 
 static void
diff --git a/src/glsl/pp/sl_pp_error.c b/src/glsl/pp/sl_pp_error.c
index b628e37ce83..482b67fcafb 100644
--- a/src/glsl/pp/sl_pp_error.c
+++ b/src/glsl/pp/sl_pp_error.c
@@ -30,6 +30,7 @@
 #include "sl_pp_context.h"
 #include "sl_pp_process.h"
 #include "sl_pp_public.h"
+#include "sl_pp_token.h"
 
 
 void
diff --git a/src/glsl/pp/sl_pp_expression.c b/src/glsl/pp/sl_pp_expression.c
index ec904787dd7..c3f48356b09 100644
--- a/src/glsl/pp/sl_pp_expression.c
+++ b/src/glsl/pp/sl_pp_expression.c
@@ -27,8 +27,10 @@
 
 #include <stdlib.h>
 #include <string.h>
+#include "sl_pp_context.h"
 #include "sl_pp_expression.h"
 #include "sl_pp_public.h"
+#include "sl_pp_token.h"
 
 
 struct parse_context {
diff --git a/src/glsl/pp/sl_pp_expression.h b/src/glsl/pp/sl_pp_expression.h
index 377d5b4cbd9..522263bb259 100644
--- a/src/glsl/pp/sl_pp_expression.h
+++ b/src/glsl/pp/sl_pp_expression.h
@@ -28,8 +28,8 @@
 #ifndef SL_PP_EXPRESSION_H
 #define SL_PP_EXPRESSION_H
 
-#include "sl_pp_context.h"
-#include "sl_pp_token.h"
+struct sl_pp_context;
+struct sl_pp_token_info;
 
 
 int
diff --git a/src/glsl/pp/sl_pp_extension.c b/src/glsl/pp/sl_pp_extension.c
index d119677c268..00dbdcf22bc 100644
--- a/src/glsl/pp/sl_pp_extension.c
+++ b/src/glsl/pp/sl_pp_extension.c
@@ -25,11 +25,13 @@
  * 
  **************************************************************************/
 
+#include <assert.h>
 #include <stdlib.h>
 #include <string.h>
 #include "sl_pp_context.h"
 #include "sl_pp_process.h"
 #include "sl_pp_public.h"
+#include "sl_pp_token.h"
 
 
 /**
diff --git a/src/glsl/pp/sl_pp_if.c b/src/glsl/pp/sl_pp_if.c
index 25cb7a3ca11..6b7a1590b42 100644
--- a/src/glsl/pp/sl_pp_if.c
+++ b/src/glsl/pp/sl_pp_if.c
@@ -27,8 +27,11 @@
 
 #include <stdlib.h>
 #include <string.h>
+#include "sl_pp_context.h"
 #include "sl_pp_expression.h"
+#include "sl_pp_macro.h"
 #include "sl_pp_process.h"
+#include "sl_pp_token.h"
 
 
 static int
diff --git a/src/glsl/pp/sl_pp_line.c b/src/glsl/pp/sl_pp_line.c
index 6f7e9eb562c..51581c7bb59 100644
--- a/src/glsl/pp/sl_pp_line.c
+++ b/src/glsl/pp/sl_pp_line.c
@@ -28,8 +28,10 @@
 #include <stdlib.h>
 #include <string.h>
 #include "sl_pp_context.h"
+#include "sl_pp_macro.h"
 #include "sl_pp_public.h"
 #include "sl_pp_process.h"
+#include "sl_pp_token.h"
 
 
 int
diff --git a/src/glsl/pp/sl_pp_macro.c b/src/glsl/pp/sl_pp_macro.c
index 9f520b8fc53..2cf9ea342b2 100644
--- a/src/glsl/pp/sl_pp_macro.c
+++ b/src/glsl/pp/sl_pp_macro.c
@@ -32,6 +32,7 @@
 #include "sl_pp_public.h"
 #include "sl_pp_macro.h"
 #include "sl_pp_process.h"
+#include "sl_pp_token.h"
 
 
 static void
diff --git a/src/glsl/pp/sl_pp_macro.h b/src/glsl/pp/sl_pp_macro.h
index 1d210681091..6e65a0a588d 100644
--- a/src/glsl/pp/sl_pp_macro.h
+++ b/src/glsl/pp/sl_pp_macro.h
@@ -28,9 +28,6 @@
 #ifndef SL_PP_MACRO_H
 #define SL_PP_MACRO_H
 
-#include "sl_pp_token.h"
-
-
 struct sl_pp_context;
 struct sl_pp_process_state;
 struct sl_pp_token_buffer;
diff --git a/src/glsl/pp/sl_pp_pragma.c b/src/glsl/pp/sl_pp_pragma.c
index caf4c63f657..6789704db0c 100644
--- a/src/glsl/pp/sl_pp_pragma.c
+++ b/src/glsl/pp/sl_pp_pragma.c
@@ -29,6 +29,7 @@
 #include <string.h>
 #include "sl_pp_context.h"
 #include "sl_pp_process.h"
+#include "sl_pp_token.h"
 
 
 int
diff --git a/src/glsl/pp/sl_pp_process.c b/src/glsl/pp/sl_pp_process.c
index 315ad9bf1cd..2f12393237c 100644
--- a/src/glsl/pp/sl_pp_process.c
+++ b/src/glsl/pp/sl_pp_process.c
@@ -25,11 +25,14 @@
  * 
  **************************************************************************/
 
+#include <assert.h>
 #include <stdlib.h>
 #include <string.h>
 #include "sl_pp_context.h"
+#include "sl_pp_macro.h"
 #include "sl_pp_process.h"
 #include "sl_pp_public.h"
+#include "sl_pp_token.h"
 
 
 int
diff --git a/src/glsl/pp/sl_pp_process.h b/src/glsl/pp/sl_pp_process.h
index fe6ff0d4648..04e9be43989 100644
--- a/src/glsl/pp/sl_pp_process.h
+++ b/src/glsl/pp/sl_pp_process.h
@@ -28,11 +28,8 @@
 #ifndef SL_PP_PROCESS_H
 #define SL_PP_PROCESS_H
 
-#include "sl_pp_macro.h"
-#include "sl_pp_token.h"
-
-
 struct sl_pp_context;
+struct sl_pp_token_buffer;
 
 struct sl_pp_process_state {
    struct sl_pp_token_info *out;
diff --git a/src/glsl/pp/sl_pp_public.h b/src/glsl/pp/sl_pp_public.h
index ca6f722543d..66ced6cf589 100644
--- a/src/glsl/pp/sl_pp_public.h
+++ b/src/glsl/pp/sl_pp_public.h
@@ -28,13 +28,9 @@
 #ifndef SL_PP_PUBLIC_H
 #define SL_PP_PUBLIC_H
 
-
 struct sl_pp_context;
-
-
-#include "sl_pp_purify.h"
-#include "sl_pp_token.h"
-
+struct sl_pp_purify_options;
+struct sl_pp_token_info;
 
 struct sl_pp_context *
 sl_pp_context_create(const char *input,
diff --git a/src/glsl/pp/sl_pp_token_util.c b/src/glsl/pp/sl_pp_token_util.c
index c85263d9a11..43be49670b0 100644
--- a/src/glsl/pp/sl_pp_token_util.c
+++ b/src/glsl/pp/sl_pp_token_util.c
@@ -28,6 +28,7 @@
 #include <assert.h>
 #include <stdlib.h>
 #include "sl_pp_token_util.h"
+#include "sl_pp_token.h"
 
 
 int
diff --git a/src/glsl/pp/sl_pp_token_util.h b/src/glsl/pp/sl_pp_token_util.h
index 2a668ad0a84..35d72101ca5 100644
--- a/src/glsl/pp/sl_pp_token_util.h
+++ b/src/glsl/pp/sl_pp_token_util.h
@@ -28,11 +28,6 @@
 #ifndef SL_PP_TOKEN_UTIL_H
 #define SL_PP_TOKEN_UTIL_H
 
-#include <assert.h>
-#include <stdlib.h>
-#include "sl_pp_token.h"
-
-
 struct sl_pp_context;
 
 /*
diff --git a/src/glsl/pp/sl_pp_version.c b/src/glsl/pp/sl_pp_version.c
index 3c995b77501..6735c05e8ae 100644
--- a/src/glsl/pp/sl_pp_version.c
+++ b/src/glsl/pp/sl_pp_version.c
@@ -29,6 +29,7 @@
 #include <string.h>
 #include "sl_pp_public.h"
 #include "sl_pp_context.h"
+#include "sl_pp_token.h"
 
 
 int
diff --git a/src/glut/glx/glut_dstr.c b/src/glut/glx/glut_dstr.c
index 2513af45394..319930c4b11 100644
--- a/src/glut/glx/glut_dstr.c
+++ b/src/glut/glx/glut_dstr.c
@@ -232,7 +232,7 @@ loadVisuals(int *nitems_return)
   XVisualInfo *vinfo, **vlist, template;
   FrameBufferMode *fbmodes, *mode;
   int n, i, j, rc, glcapable;
-#if defined(GLX_VERSION_1_1) && defined(GLX_SGIS_multisample)
+#if defined(GLX_VERSION_1_1) && (defined(GLX_SGIS_multisample) || defined(GLX_ARB_multisample))
   int multisample;
 #endif
 #if defined(GLX_VERSION_1_1) && defined(GLX_EXT_visual_info)
@@ -275,8 +275,9 @@ loadVisuals(int *nitems_return)
     }
   }
 
-#if defined(GLX_VERSION_1_1) && defined(GLX_SGIS_multisample)
-  multisample = __glutIsSupportedByGLX("GLX_SGIS_multisample");
+#if defined(GLX_VERSION_1_1) && (defined(GLX_SGIS_multisample) || defined(GLX_ARB_multisample))
+  multisample = __glutIsSupportedByGLX("GLX_SGIS_multisample") ||
+                __glutIsSupportedByGLX("GLX_ARB_multisample");
 #endif
 #if defined(GLX_VERSION_1_1) && defined(GLX_EXT_visual_info)
   visual_info = __glutIsSupportedByGLX("GLX_EXT_visual_info");
@@ -572,7 +573,7 @@ loadVisuals(int *nitems_return)
 #else
                 mode->cap[TRANSPARENT] = 0;
 #endif
-#if defined(GLX_VERSION_1_1) && defined(GLX_SGIS_multisample)
+#if defined(GLX_VERSION_1_1) && (defined(GLX_SGIS_multisample) || defined(GLX_ARB_multisample))
                 if (multisample) {
                   rc = __glut_glXGetFBConfigAttribSGIX(__glutDisplay,
 		    fbc, GLX_SAMPLES_SGIS, &mode->cap[SAMPLES]);
@@ -1250,8 +1251,9 @@ parseModeString(char *mode, int *ncriteria, Bool * allowDoubleAsSingle,
     word = strtok(NULL, " \t");
   }
 
-#if defined(GLX_VERSION_1_1) && defined(GLX_SGIS_multisample)
-  if (__glutIsSupportedByGLX("GLX_SGIS_multisample")) {
+#if defined(GLX_VERSION_1_1) && (defined(GLX_SGIS_multisample) || defined(GLX_ARB_multisample))
+  if (__glutIsSupportedByGLX("GLX_SGIS_multisample") ||
+      __glutIsSupportedByGLX("GLX_ARB_multisample")) {
     if (!(mask & (1 << SAMPLES))) {
       criteria[n].capability = SAMPLES;
       criteria[n].comparison = EQ;
diff --git a/src/glut/glx/glut_overlay.c b/src/glut/glx/glut_overlay.c
index 32434650ebe..ce8e68d164a 100644
--- a/src/glut/glx/glut_overlay.c
+++ b/src/glut/glx/glut_overlay.c
@@ -81,10 +81,11 @@ checkOverlayAcceptability(XVisualInfo * vi, unsigned int mode)
   if (GLUT_WIND_HAS_STENCIL(mode) && (value <= 0))
     return 1;
 
-#if defined(GLX_VERSION_1_1) && defined(GLX_SGIS_multisample)
+#if defined(GLX_VERSION_1_1) && (defined(GLX_SGIS_multisample) || defined(GLX_ARB_multisample))
   /* XXX Multisampled overlay color index??  Pretty unlikely. */
   /* Look for multisampling if requested. */
-  if (__glutIsSupportedByGLX("GLX_SGIS_multisample"))
+  if (__glutIsSupportedByGLX("GLX_SGIS_multisample") ||
+      __glutIsSupportedByGLX("GLX_ARB_multisample"))
     glXGetConfig(__glutDisplay, vi, GLX_SAMPLES_SGIS, &value);
   else
     value = 0;
diff --git a/src/glx/Makefile b/src/glx/Makefile
index 48d901fb894..9a22d0c547a 100644
--- a/src/glx/Makefile
+++ b/src/glx/Makefile
@@ -5,14 +5,15 @@ EXTRA_DEFINES = -DXF86VIDMODE -D_REENTRANT \
                 -DDEFAULT_DRIVER_DIR=\"$(DRI_DRIVER_SEARCH_DIR)\"
 
 SOURCES = \
-	  glcontextmodes.c \
 	  clientattrib.c \
 	  compsize.c \
 	  eval.c \
+	  glxconfig.c \
 	  glxcmds.c \
 	  glxcurrent.c \
 	  glxext.c \
 	  glxextensions.c \
+	  indirect_glx.c \
 	  indirect.c \
 	  indirect_init.c \
 	  indirect_size.c \
@@ -37,7 +38,9 @@ SOURCES = \
 	  XF86dri.c \
 	  glxhash.c \
 	  dri2_glx.c \
-	  dri2.c
+	  dri2.c \
+	  applegl_glx.c
+
 
 GLAPI_LIB = $(TOP)/src/mapi/glapi/libglapi.a
 
diff --git a/src/glx/XF86dri.c b/src/glx/XF86dri.c
index 36b43f0f84a..5c181d6db9e 100644
--- a/src/glx/XF86dri.c
+++ b/src/glx/XF86dri.c
@@ -39,21 +39,11 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
 
 #include <X11/Xlibint.h>
+#include <X11/Xfuncproto.h>
 #include <X11/extensions/Xext.h>
 #include <X11/extensions/extutil.h>
 #include "xf86dristr.h"
 
-
-#if defined(__GNUC__)
-#  define PUBLIC __attribute__((visibility("default")))
-#  define USED __attribute__((used))
-#else
-#  define PUBLIC
-#  define USED
-#endif
-
-
-
 static XExtensionInfo _xf86dri_info_data;
 static XExtensionInfo *xf86dri_info = &_xf86dri_info_data;
 static char xf86dri_extension_name[] = XF86DRINAME;
diff --git a/src/glx/apple/apple_glx_drawable.h b/src/glx/apple/apple_glx_drawable.h
index e49eae355e9..af79e0590e1 100644
--- a/src/glx/apple/apple_glx_drawable.h
+++ b/src/glx/apple/apple_glx_drawable.h
@@ -211,7 +211,7 @@ bool apple_glx_pbuffer_get_event_mask(GLXDrawable d, unsigned long *mask);
 
 /* Pixmaps */
 
-/* mode is a __GLcontextModes * */
+/* mode is a struct glx_config * */
 /* Returns true if an error occurred. */
 bool apple_glx_pixmap_create(Display * dpy, int screen, Pixmap pixmap,
                              const void *mode);
diff --git a/src/glx/apple/apple_glx_pbuffer.c b/src/glx/apple/apple_glx_pbuffer.c
index 1466fea4874..a5ef59ccf82 100644
--- a/src/glx/apple/apple_glx_pbuffer.c
+++ b/src/glx/apple/apple_glx_pbuffer.c
@@ -128,7 +128,7 @@ apple_glx_pbuffer_create(Display * dpy, GLXFBConfig config,
    Window root;
    int screen;
    Pixmap xid;
-   __GLcontextModes *modes = (__GLcontextModes *) config;
+   struct glx_config *modes = (__GLcontextModes *) config;
 
    root = DefaultRootWindow(dpy);
    screen = DefaultScreen(dpy);
diff --git a/src/glx/apple/apple_glx_pixmap.c b/src/glx/apple/apple_glx_pixmap.c
index af1791afb70..fec05e02de6 100644
--- a/src/glx/apple/apple_glx_pixmap.c
+++ b/src/glx/apple/apple_glx_pixmap.c
@@ -125,7 +125,7 @@ apple_glx_pixmap_create(Display * dpy, int screen, Pixmap pixmap,
    bool double_buffered;
    bool uses_stereo;
    CGLError error;
-   const __GLcontextModes *cmodes = mode;
+   const struct glx_config *cmodes = mode;
 
    if (apple_glx_drawable_create(dpy, screen, pixmap, &d, &callbacks))
       return true;
diff --git a/src/glx/apple/apple_visual.c b/src/glx/apple/apple_visual.c
index da5aa05fd50..7a9525e8170 100644
--- a/src/glx/apple/apple_visual.c
+++ b/src/glx/apple/apple_visual.c
@@ -61,7 +61,7 @@ apple_visual_create_pfobj(CGLPixelFormatObj * pfobj, const void *mode,
                           bool offscreen)
 {
    CGLPixelFormatAttribute attr[MAX_ATTR];
-   const __GLcontextModes *c = mode;
+   const struct glx_config *c = mode;
    int numattr = 0;
    GLint vsref = 0;
    CGLError error = 0;
diff --git a/src/glx/apple/apple_visual.h b/src/glx/apple/apple_visual.h
index ebfafa340bf..2cf9a9e2513 100644
--- a/src/glx/apple/apple_visual.h
+++ b/src/glx/apple/apple_visual.h
@@ -33,7 +33,7 @@
 #include <stdbool.h>
 #include <OpenGL/CGLTypes.h>
 
-/* mode is expected to be of type __GLcontextModes. */
+/* mode is expected to be of type struct glx_config. */
 void apple_visual_create_pfobj(CGLPixelFormatObj * pfobj, const void *mode,
                                bool * double_buffered, bool * uses_stereo,
                                bool offscreen);
diff --git a/src/glx/apple/glx_empty.c b/src/glx/apple/glx_empty.c
index 1569679c713..5967ecabc1a 100644
--- a/src/glx/apple/glx_empty.c
+++ b/src/glx/apple/glx_empty.c
@@ -152,7 +152,7 @@ glXReleaseBuffersMESA(Display * dpy, GLXDrawable d)
 }
 
 
-PUBLIC GLXPixmap
+_X_EXPORT GLXPixmap
 glXCreateGLXPixmapMESA(Display * dpy, XVisualInfo * visual,
                        Pixmap pixmap, Colormap cmap)
 {
@@ -180,7 +180,7 @@ glXCopySubBufferMESA(Display * dpy, GLXDrawable drawable,
 }
 
 
-PUBLIC int
+_X_EXPORT int
 glXQueryGLXPbufferSGIX(Display * dpy, GLXDrawable drawable,
                        int attribute, unsigned int *value)
 {
@@ -191,7 +191,7 @@ glXQueryGLXPbufferSGIX(Display * dpy, GLXDrawable drawable,
    return 0;
 }
 
-PUBLIC GLXDrawable
+_X_EXPORT GLXDrawable
 glXCreateGLXPbufferSGIX(Display * dpy, GLXFBConfig config,
                         unsigned int width, unsigned int height,
                         int *attrib_list)
@@ -206,7 +206,7 @@ glXCreateGLXPbufferSGIX(Display * dpy, GLXFBConfig config,
 
 #if 0
 /* GLX_SGIX_fbconfig */
-PUBLIC int
+_X_EXPORT int
 glXGetFBConfigAttribSGIX(Display * dpy, void *config, int a, int *b)
 {
    (void) dpy;
@@ -216,7 +216,7 @@ glXGetFBConfigAttribSGIX(Display * dpy, void *config, int a, int *b)
    return 0;
 }
 
-PUBLIC void *
+_X_EXPORT void *
 glXChooseFBConfigSGIX(Display * dpy, int a, int *b, int *c)
 {
    (void) dpy;
@@ -226,7 +226,7 @@ glXChooseFBConfigSGIX(Display * dpy, int a, int *b, int *c)
    return NULL;
 }
 
-PUBLIC GLXPixmap
+_X_EXPORT GLXPixmap
 glXCreateGLXPixmapWithConfigSGIX(Display * dpy, void *config, Pixmap p)
 {
    (void) dpy;
@@ -235,7 +235,7 @@ glXCreateGLXPixmapWithConfigSGIX(Display * dpy, void *config, Pixmap p)
    return None;
 }
 
-PUBLIC GLXContext
+_X_EXPORT GLXContext
 glXCreateContextWithConfigSGIX(Display * dpy, void *config, int a,
                                GLXContext b, Bool c)
 {
@@ -247,7 +247,7 @@ glXCreateContextWithConfigSGIX(Display * dpy, void *config, int a,
    return NULL;
 }
 
-PUBLIC XVisualInfo *
+_X_EXPORT XVisualInfo *
 glXGetVisualFromFBConfigSGIX(Display * dpy, void *config)
 {
    (void) dpy;
@@ -255,7 +255,7 @@ glXGetVisualFromFBConfigSGIX(Display * dpy, void *config)
    return NULL;
 }
 
-PUBLIC void *
+_X_EXPORT void *
 glXGetFBConfigFromVisualSGIX(Display * dpy, XVisualInfo * visinfo)
 {
    (void) dpy;
@@ -265,17 +265,17 @@ glXGetFBConfigFromVisualSGIX(Display * dpy, XVisualInfo * visinfo)
 #endif
 
 
-PUBLIC
+_X_EXPORT
 GLX_ALIAS_VOID(glXDestroyGLXPbufferSGIX,
                (Display * dpy, GLXDrawable pbuf),
                (dpy, pbuf), glXDestroyPbuffer)
 
-     PUBLIC GLX_ALIAS_VOID(glXSelectEventSGIX,
+     _X_EXPORT GLX_ALIAS_VOID(glXSelectEventSGIX,
                            (Display * dpy, GLXDrawable drawable,
                             unsigned long mask), (dpy, drawable, mask),
                            glXSelectEvent)
 
-     PUBLIC GLX_ALIAS_VOID(glXGetSelectedEventSGIX,
+     _X_EXPORT GLX_ALIAS_VOID(glXGetSelectedEventSGIX,
                            (Display * dpy, GLXDrawable drawable,
                             unsigned long *mask), (dpy, drawable, mask),
                            glXGetSelectedEvent)
diff --git a/src/glx/apple/glxreply.c b/src/glx/apple/glxreply.c
index 7280bc9715a..f17ebe6b406 100644
--- a/src/glx/apple/glxreply.c
+++ b/src/glx/apple/glxreply.c
@@ -55,7 +55,7 @@ __glXReadReply(Display * dpy, size_t size, void *dest,
 }
 
 void
-__glXReadPixelReply(Display * dpy, __GLXcontext * gc, unsigned max_dim,
+__glXReadPixelReply(Display * dpy, struct glx_context * gc, unsigned max_dim,
                     GLint width, GLint height, GLint depth, GLenum format,
                     GLenum type, void *dest, GLboolean dimensions_in_reply)
 {
@@ -101,7 +101,7 @@ __glXReadPixelReply(Display * dpy, __GLXcontext * gc, unsigned max_dim,
 
 #if 0
 GLubyte *
-__glXSetupSingleRequest(__GLXcontext * gc, GLint sop, GLint cmdlen)
+__glXSetupSingleRequest(struct glx_context * gc, GLint sop, GLint cmdlen)
 {
    xGLXSingleReq *req;
    Display *const dpy = gc->currentDpy;
@@ -117,7 +117,7 @@ __glXSetupSingleRequest(__GLXcontext * gc, GLint sop, GLint cmdlen)
 #endif
 
 GLubyte *
-__glXSetupVendorRequest(__GLXcontext * gc, GLint code, GLint vop,
+__glXSetupVendorRequest(struct glx_context * gc, GLint code, GLint vop,
                         GLint cmdlen)
 {
    xGLXVendorPrivateReq *req;
diff --git a/src/glx/applegl_glx.c b/src/glx/applegl_glx.c
new file mode 100644
index 00000000000..3da1f19323f
--- /dev/null
+++ b/src/glx/applegl_glx.c
@@ -0,0 +1,155 @@
+/*
+ * Copyright © 2010 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Soft-
+ * ware"), to deal in the Software without restriction, including without
+ * limitation the rights to use, copy, modify, merge, publish, distribute,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, provided that the above copyright
+ * notice(s) and this permission notice appear in all copies of the Soft-
+ * ware and that both the above copyright notice(s) and this permission
+ * notice appear in supporting documentation.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABIL-
+ * ITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT OF THIRD PARTY
+ * RIGHTS. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR HOLDERS INCLUDED IN
+ * THIS NOTICE BE LIABLE FOR ANY CLAIM, OR ANY SPECIAL INDIRECT OR CONSE-
+ * QUENTIAL DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE,
+ * DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER
+ * TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFOR-
+ * MANCE OF THIS SOFTWARE.
+ *
+ * Except as contained in this notice, the name of a copyright holder shall
+ * not be used in advertising or otherwise to promote the sale, use or
+ * other dealings in this Software without prior written authorization of
+ * the copyright holder.
+ *
+ * Authors:
+ *   Kristian Høgsberg ([email protected])
+ */
+
+#if defined(GLX_USE_APPLEGL)
+
+static void
+applegl_destroy_context(struct glx_context *gc)
+{
+   apple_glx_destroy_context(&gc->driContext, gc->currentDpy);
+}
+
+static int
+applegl_bind_context(struct glx_context *gc, struct glx_context *old,
+		     GLXDrawable draw, GLXDrawable read)
+{
+   bool error = apple_glx_make_current_context(dpy,
+					       (oldGC && oldGC != &dummyContext) ? oldGC->driContext : NUL~
+					       gc ? gc->driContext : NULL, draw);
+
+   apple_glx_diagnostic("%s: error %s\n", __func__, error ? "YES" : "NO");
+   if (error)
+      return GLXBadContext;
+
+   return Success;
+}
+
+static void
+applegl_unbind_context(struct glx_context *gc, struct glx_context *new)
+{
+}
+
+static void
+applegl_wait_gl(struct glx_context *gc)
+{
+   glFinish();
+}
+
+static void
+applegl_wait_x(struct glx_context *gc)
+{
+   apple_glx_waitx(gc->dpy, gc->driContext);
+}
+
+static const struct glx_context_vtable applegl_context_vtable = {
+   applegl_destroy_context,
+   applegl_bind_context,
+   applegl_unbind_context,
+   applegl_wait_gl,
+   applegl_wait_x,
+   DRI_glXUseXFont,
+   NULL, /* bind_tex_image, */
+   NULL, /* release_tex_image, */
+};
+
+static struct glx_context *
+applegl_create_context(struct glx_screen *psc,
+		       struct glx_config *mode,
+		       struct glx_context *shareList, int renderType)
+{
+   struct glx_context *gc;
+   int errorcode;
+   bool x11error;
+
+   /* TODO: Integrate this with apple_glx_create_context and make
+    * struct apple_glx_context inherit from struct glx_context. */
+
+   gc = Xmalloc(sizeof *gc);
+   if (pcp == NULL)
+      return NULL;
+
+   memset(gc, 0, sizeof *gc);
+   if (!glx_context_init(&gc->base, &psc->base, mode)) {
+      Xfree(gc);
+      return NULL;
+   }
+
+   gc->vtable = &applegl_context_vtable;
+   gc->driContext = NULL;
+   gc->do_destroy = False;
+
+   /* TODO: darwin: Integrate with above to do indirect */
+   if(apple_glx_create_context(&gc->driContext, dpy, screen, fbconfig, 
+			       shareList ? shareList->driContext : NULL,
+			       &errorcode, &x11error)) {
+      __glXSendError(dpy, errorcode, 0, X_GLXCreateContext, x11error);
+      gc->vtable->destroy(gc);
+      return NULL;
+   }
+
+   gc->currentContextTag = -1;
+   gc->mode = fbconfig;
+   gc->isDirect = allowDirect;
+   gc->xid = 1; /* Just something not None, so we know when to destroy
+		 * it in MakeContextCurrent. */
+
+   return gc;
+}
+
+struct glx_screen_vtable appegl_screen_vtable = {
+   applegl_create_context
+};
+
+_X_HIDDEN struct glx_screen *
+applegl_create_screen(int screen, struct glx_display * priv)
+{
+   struct glx_screen *psc;
+
+   psc = Xmalloc(sizeof *psc);
+   if (psc == NULL)
+      return NULL;
+
+   memset(psc, 0, sizeof *psc);
+   glx_screen_init(psc, screen, priv);
+   psc->vtable = &applegl_screen_vtable;
+
+   return psc;
+}
+
+_X_HIDDEN int
+applegl_create_display(struct glx_display *display)
+{
+   /* create applegl display and stuff in display->appleglDisplay */
+   apple_init_glx(display);
+}
+
+#endif
diff --git a/src/glx/clientattrib.c b/src/glx/clientattrib.c
index b26c17938db..7792fa31f9d 100644
--- a/src/glx/clientattrib.c
+++ b/src/glx/clientattrib.c
@@ -39,7 +39,7 @@
 static void
 do_enable_disable(GLenum array, GLboolean val)
 {
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    __GLXattribute *state = (__GLXattribute *) (gc->client_state_private);
    unsigned index = 0;
 
@@ -69,7 +69,7 @@ __indirect_glDisableClientState(GLenum array)
 void
 __indirect_glPushClientAttrib(GLuint mask)
 {
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    __GLXattribute *state = (__GLXattribute *) (gc->client_state_private);
    __GLXattribute **spp = gc->attributes.stackPointer, *sp;
 
@@ -97,7 +97,7 @@ __indirect_glPushClientAttrib(GLuint mask)
 void
 __indirect_glPopClientAttrib(void)
 {
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    __GLXattribute *state = (__GLXattribute *) (gc->client_state_private);
    __GLXattribute **spp = gc->attributes.stackPointer, *sp;
    GLuint mask;
@@ -127,7 +127,7 @@ __indirect_glPopClientAttrib(void)
 #endif
 
 void
-__glFreeAttributeState(__GLXcontext * gc)
+__glFreeAttributeState(struct glx_context * gc)
 {
    __GLXattribute *sp, **spp;
 
diff --git a/src/glx/dri2.c b/src/glx/dri2.c
index d53431c19a6..d70ec5a3ecf 100644
--- a/src/glx/dri2.c
+++ b/src/glx/dri2.c
@@ -88,7 +88,6 @@ static Bool
 DRI2WireToEvent(Display *dpy, XEvent *event, xEvent *wire)
 {
    XExtDisplayInfo *info = DRI2FindDisplay(dpy);
-   __GLXdisplayPrivate *glx_dpy = __glXInitialize(dpy);
 
    XextCheckExtension(dpy, info, dri2ExtensionName, False);
 
@@ -100,6 +99,7 @@ DRI2WireToEvent(Display *dpy, XEvent *event, xEvent *wire)
       GLXBufferSwapComplete *aevent = (GLXBufferSwapComplete *)event;
       xDRI2BufferSwapComplete *awire = (xDRI2BufferSwapComplete *)wire;
       __GLXDRIdrawable *pdraw;
+      struct glx_display *glx_dpy = __glXInitialize(dpy);
 
       /* Ignore swap events if we're not looking for them */
       pdraw = dri2GetGlxDrawableFromXDrawableId(dpy, awire->drawable);
diff --git a/src/glx/dri2_glx.c b/src/glx/dri2_glx.c
index 2f12387860e..ff48c79c272 100644
--- a/src/glx/dri2_glx.c
+++ b/src/glx/dri2_glx.c
@@ -76,7 +76,7 @@ struct dri2_display
 };
 
 struct dri2_screen {
-   __GLXscreenConfigs base;
+   struct glx_screen base;
 
    __DRIscreen *driScreen;
    __GLXDRIscreen vtable;
@@ -94,8 +94,7 @@ struct dri2_screen {
 
 struct dri2_context
 {
-   __GLXcontext base;
-   __GLXDRIcontext dri_vtable;
+   struct glx_context base;
    __DRIcontext *driContext;
 };
 
@@ -114,31 +113,47 @@ struct dri2_drawable
 static const struct glx_context_vtable dri2_context_vtable;
 
 static void
-dri2DestroyContext(__GLXcontext *context)
+dri2_destroy_context(struct glx_context *context)
 {
    struct dri2_context *pcp = (struct dri2_context *) context;
    struct dri2_screen *psc = (struct dri2_screen *) context->psc;
 
+   if (context->xid)
+      glx_send_destroy_context(psc->base.dpy, context->xid);
+
+   if (context->extensions)
+      XFree((char *) context->extensions);
+
+   GarbageCollectDRIDrawables(context->psc);
+
    (*psc->core->destroyContext) (pcp->driContext);
 
    Xfree(pcp);
 }
 
 static Bool
-dri2BindContext(__GLXcontext *context,
-		__GLXDRIdrawable *draw, __GLXDRIdrawable *read)
+dri2_bind_context(struct glx_context *context, struct glx_context *old,
+		  GLXDrawable draw, GLXDrawable read)
 {
    struct dri2_context *pcp = (struct dri2_context *) context;
    struct dri2_screen *psc = (struct dri2_screen *) pcp->base.psc;
-   struct dri2_drawable *pdr = (struct dri2_drawable *) draw;
-   struct dri2_drawable *prd = (struct dri2_drawable *) read;
+   struct dri2_drawable *pdraw, *pread;
+
+   pdraw = (struct dri2_drawable *) driFetchDrawable(context, draw);
+   pread = (struct dri2_drawable *) driFetchDrawable(context, read);
+
+   if (pdraw == NULL || pread == NULL)
+      return GLXBadDrawable;
 
-   return (*psc->core->bindContext) (pcp->driContext,
-				     pdr->driDrawable, prd->driDrawable);
+   if ((*psc->core->bindContext) (pcp->driContext,
+				  pdraw->driDrawable, pread->driDrawable))
+      return Success;
+
+   return GLXBadContext;
 }
 
 static void
-dri2UnbindContext(__GLXcontext *context)
+dri2_unbind_context(struct glx_context *context, struct glx_context *new)
 {
    struct dri2_context *pcp = (struct dri2_context *) context;
    struct dri2_screen *psc = (struct dri2_screen *) pcp->base.psc;
@@ -146,18 +161,18 @@ dri2UnbindContext(__GLXcontext *context)
    (*psc->core->unbindContext) (pcp->driContext);
 }
 
-static __GLXcontext *
-dri2CreateContext(__GLXscreenConfigs *base,
-                  const __GLcontextModes * mode,
-                  GLXContext shareList, int renderType)
+static struct glx_context *
+dri2_create_context(struct glx_screen *base,
+		    struct glx_config *config_base,
+		    struct glx_context *shareList, int renderType)
 {
    struct dri2_context *pcp, *pcp_shared;
    struct dri2_screen *psc = (struct dri2_screen *) base;
-   __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) mode;
+   __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) config_base;
    __DRIcontext *shared = NULL;
 
    if (shareList) {
-      pcp_shared = (struct dri2_context *) shareList->driContext;
+      pcp_shared = (struct dri2_context *) shareList;
       shared = pcp_shared->driContext;
    }
 
@@ -166,7 +181,7 @@ dri2CreateContext(__GLXscreenConfigs *base,
       return NULL;
 
    memset(pcp, 0, sizeof *pcp);
-   if (!glx_context_init(&pcp->base, &psc->base, mode)) {
+   if (!glx_context_init(&pcp->base, &psc->base, &config->base)) {
       Xfree(pcp);
       return NULL;
    }
@@ -181,10 +196,6 @@ dri2CreateContext(__GLXscreenConfigs *base,
    }
 
    pcp->base.vtable = &dri2_context_vtable;
-   pcp->base.driContext = &pcp->dri_vtable;
-   pcp->dri_vtable.destroyContext = dri2DestroyContext;
-   pcp->dri_vtable.bindContext = dri2BindContext;
-   pcp->dri_vtable.unbindContext = dri2UnbindContext;
 
    return &pcp->base;
 }
@@ -194,7 +205,7 @@ dri2DestroyDrawable(__GLXDRIdrawable *base)
 {
    struct dri2_screen *psc = (struct dri2_screen *) base->psc;
    struct dri2_drawable *pdraw = (struct dri2_drawable *) base;
-   __GLXdisplayPrivate *dpyPriv = psc->base.display;
+   struct glx_display *dpyPriv = psc->base.display;
    struct dri2_display *pdp = (struct dri2_display *)dpyPriv->dri2Display;
 
    __glxHashDelete(pdp->dri2Hash, pdraw->base.xDrawable);
@@ -204,13 +215,13 @@ dri2DestroyDrawable(__GLXDRIdrawable *base)
 }
 
 static __GLXDRIdrawable *
-dri2CreateDrawable(__GLXscreenConfigs *base, XID xDrawable,
-		   GLXDrawable drawable, const __GLcontextModes * modes)
+dri2CreateDrawable(struct glx_screen *base, XID xDrawable,
+		   GLXDrawable drawable, struct glx_config *config_base)
 {
    struct dri2_drawable *pdraw;
    struct dri2_screen *psc = (struct dri2_screen *) base;
-   __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) modes;
-   __GLXdisplayPrivate *dpyPriv;
+   __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) config_base;
+   struct glx_display *dpyPriv;
    struct dri2_display *pdp;
    GLint vblank_mode = DRI_CONF_VBLANK_DEF_INTERVAL_1;
 
@@ -218,6 +229,7 @@ dri2CreateDrawable(__GLXscreenConfigs *base, XID xDrawable,
    if (!pdraw)
       return NULL;
 
+   memset(pdraw, 0, sizeof *pdraw);
    pdraw->base.destroyDrawable = dri2DestroyDrawable;
    pdraw->base.xDrawable = xDrawable;
    pdraw->base.drawable = drawable;
@@ -280,7 +292,7 @@ dri2CreateDrawable(__GLXscreenConfigs *base, XID xDrawable,
 #ifdef X_DRI2GetMSC
 
 static int
-dri2DrawableGetMSC(__GLXscreenConfigs *psc, __GLXDRIdrawable *pdraw,
+dri2DrawableGetMSC(struct glx_screen *psc, __GLXDRIdrawable *pdraw,
 		   int64_t *ust, int64_t *msc, int64_t *sbc)
 {
    CARD64 dri2_ust, dri2_msc, dri2_sbc;
@@ -394,7 +406,7 @@ dri2_copy_drawable(struct dri2_drawable *priv, int dest, int src)
 }
 
 static void
-dri2_wait_x(__GLXcontext *gc)
+dri2_wait_x(struct glx_context *gc)
 {
    struct dri2_drawable *priv = (struct dri2_drawable *)
       GetGLXDRIDrawable(gc->currentDpy, gc->currentDrawable);
@@ -406,7 +418,7 @@ dri2_wait_x(__GLXcontext *gc)
 }
 
 static void
-dri2_wait_gl(__GLXcontext *gc)
+dri2_wait_gl(struct glx_context *gc)
 {
    struct dri2_drawable *priv = (struct dri2_drawable *)
       GetGLXDRIDrawable(gc->currentDpy, gc->currentDrawable);
@@ -421,19 +433,20 @@ static void
 dri2FlushFrontBuffer(__DRIdrawable *driDrawable, void *loaderPrivate)
 {
    struct dri2_drawable *pdraw = loaderPrivate;
-   __GLXdisplayPrivate *priv = __glXInitialize(pdraw->base.psc->dpy);
+   struct glx_display *priv = __glXInitialize(pdraw->base.psc->dpy);
    struct dri2_display *pdp = (struct dri2_display *)priv->dri2Display;
+   struct glx_context *gc = __glXGetCurrentContext();
 
    /* Old servers don't send invalidate events */
    if (!pdp->invalidateAvailable)
-       dri2InvalidateBuffers(priv->dpy, pdraw->base.drawable);
+       dri2InvalidateBuffers(priv->dpy, pdraw->base.xDrawable);
 
-   dri2_copy_drawable(pdraw, DRI2BufferFrontLeft, DRI2BufferFakeFrontLeft);
+   dri2_wait_gl(gc);
 }
 
 
 static void
-dri2DestroyScreen(__GLXscreenConfigs *base)
+dri2DestroyScreen(struct glx_screen *base)
 {
    struct dri2_screen *psc = (struct dri2_screen *) base;
 
@@ -481,11 +494,11 @@ dri2SwapBuffers(__GLXDRIdrawable *pdraw, int64_t target_msc, int64_t divisor,
 		int64_t remainder)
 {
     struct dri2_drawable *priv = (struct dri2_drawable *) pdraw;
-    __GLXdisplayPrivate *dpyPriv = __glXInitialize(priv->base.psc->dpy);
+    struct glx_display *dpyPriv = __glXInitialize(priv->base.psc->dpy);
     struct dri2_screen *psc = (struct dri2_screen *) priv->base.psc;
     struct dri2_display *pdp =
 	(struct dri2_display *)dpyPriv->dri2Display;
-    CARD64 ret;
+    CARD64 ret = 0;
 
 #ifdef __DRI2_FLUSH
     if (psc->f)
@@ -494,7 +507,7 @@ dri2SwapBuffers(__GLXDRIdrawable *pdraw, int64_t target_msc, int64_t divisor,
 
     /* Old servers don't send invalidate events */
     if (!pdp->invalidateAvailable)
-       dri2InvalidateBuffers(dpyPriv->dpy, pdraw->drawable);
+       dri2InvalidateBuffers(dpyPriv->dpy, pdraw->xDrawable);
 
     /* Old servers can't handle swapbuffers */
     if (!pdp->swapAvailable) {
@@ -637,16 +650,17 @@ dri2_bind_tex_image(Display * dpy,
 		    GLXDrawable drawable,
 		    int buffer, const int *attrib_list)
 {
-   GLXContext gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
+   struct dri2_context *pcp = (struct dri2_context *) gc;
    __GLXDRIdrawable *base = GetGLXDRIDrawable(dpy, drawable);
-   __GLXdisplayPrivate *dpyPriv = __glXInitialize(dpy);
+   struct glx_display *dpyPriv = __glXInitialize(dpy);
    struct dri2_drawable *pdraw = (struct dri2_drawable *) base;
    struct dri2_display *pdp =
       (struct dri2_display *) dpyPriv->dri2Display;
-   struct dri2_screen *psc = (struct dri2_screen *) base->psc;
-   struct dri2_context *pcp = (struct dri2_context *) gc->driContext;
+   struct dri2_screen *psc;
 
    if (pdraw != NULL) {
+      psc = (struct dri2_screen *) base->psc;
 
 #if __DRI2_FLUSH_VERSION >= 3
       if (!pdp->invalidateAvailable && psc->f)
@@ -674,6 +688,9 @@ dri2_release_tex_image(Display * dpy, GLXDrawable drawable, int buffer)
 }
 
 static const struct glx_context_vtable dri2_context_vtable = {
+   dri2_destroy_context,
+   dri2_bind_context,
+   dri2_unbind_context,
    dri2_wait_gl,
    dri2_wait_x,
    DRI_glXUseXFont,
@@ -710,9 +727,12 @@ dri2BindExtensions(struct dri2_screen *psc, const __DRIextension **extensions)
    }
 }
 
+static const struct glx_screen_vtable dri2_screen_vtable = {
+   dri2_create_context
+};
 
-static __GLXscreenConfigs *
-dri2CreateScreen(int screen, __GLXdisplayPrivate * priv)
+static struct glx_screen *
+dri2CreateScreen(int screen, struct glx_display * priv)
 {
    const __DRIconfig **driver_configs;
    const __DRIextension **extensions;
@@ -803,10 +823,10 @@ dri2CreateScreen(int screen, __GLXdisplayPrivate * priv)
 
    psc->driver_configs = driver_configs;
 
+   psc->base.vtable = &dri2_screen_vtable;
    psp = &psc->vtable;
    psc->base.driScreen = psp;
    psp->destroyScreen = dri2DestroyScreen;
-   psp->createContext = dri2CreateContext;
    psp->createDrawable = dri2CreateDrawable;
    psp->swapBuffers = dri2SwapBuffers;
    psp->getDrawableMSC = NULL;
@@ -863,7 +883,7 @@ dri2DestroyDisplay(__GLXDRIdisplay * dpy)
 _X_HIDDEN __GLXDRIdrawable *
 dri2GetGlxDrawableFromXDrawableId(Display *dpy, XID id)
 {
-   __GLXdisplayPrivate *d = __glXInitialize(dpy);
+   struct glx_display *d = __glXInitialize(dpy);
    struct dri2_display *pdp = (struct dri2_display *) d->dri2Display;
    __GLXDRIdrawable *pdraw;
 
diff --git a/src/glx/dri_common.c b/src/glx/dri_common.c
index 9b7da3e7df5..a7fb4c64244 100644
--- a/src/glx/dri_common.c
+++ b/src/glx/dri_common.c
@@ -39,7 +39,6 @@
 #include <dlfcn.h>
 #include <stdarg.h>
 #include "glxclient.h"
-#include "glcontextmodes.h"
 #include "dri_common.h"
 
 #ifndef RTLD_NOW
@@ -176,7 +175,7 @@ _X_HIDDEN const __DRIsystemTimeExtension systemTimeExtension = {
 };
 
 #define __ATTRIB(attrib, field) \
-    { attrib, offsetof(__GLcontextModes, field) }
+    { attrib, offsetof(struct glx_config, field) }
 
 static const struct
 {
@@ -225,10 +224,8 @@ __ATTRIB(__DRI_ATTRIB_BIND_TO_TEXTURE_RGB, bindToTextureRgb),
                      bindToMipmapTexture),
       __ATTRIB(__DRI_ATTRIB_YINVERTED, yInverted),};
 
-#define ARRAY_SIZE(a) (sizeof (a) / sizeof ((a)[0]))
-
 static int
-scalarEqual(__GLcontextModes * mode, unsigned int attrib, unsigned int value)
+scalarEqual(struct glx_config *mode, unsigned int attrib, unsigned int value)
 {
    unsigned int glxValue;
    int i;
@@ -243,8 +240,8 @@ scalarEqual(__GLcontextModes * mode, unsigned int attrib, unsigned int value)
 }
 
 static int
-driConfigEqual(const __DRIcoreExtension * core,
-               __GLcontextModes * modes, const __DRIconfig * driConfig)
+driConfigEqual(const __DRIcoreExtension *core,
+               struct glx_config *config, const __DRIconfig *driConfig)
 {
    unsigned int attrib, value, glxValue;
    int i;
@@ -260,7 +257,7 @@ driConfigEqual(const __DRIcoreExtension * core,
          else if (value & __DRI_ATTRIB_COLOR_INDEX_BIT) {
             glxValue |= GLX_COLOR_INDEX_BIT;
          }
-         if (glxValue != modes->renderType)
+         if (glxValue != config->renderType)
             return GL_FALSE;
          break;
 
@@ -271,7 +268,7 @@ driConfigEqual(const __DRIcoreExtension * core,
             glxValue = GLX_SLOW_CONFIG;
          else
             glxValue = GLX_NONE;
-         if (glxValue != modes->visualRating)
+         if (glxValue != config->visualRating)
             return GL_FALSE;
          break;
 
@@ -283,13 +280,13 @@ driConfigEqual(const __DRIcoreExtension * core,
             glxValue |= GLX_TEXTURE_2D_BIT_EXT;
          if (value & __DRI_ATTRIB_TEXTURE_RECTANGLE_BIT)
             glxValue |= GLX_TEXTURE_RECTANGLE_BIT_EXT;
-         if (modes->bindToTextureTargets != GLX_DONT_CARE &&
-             glxValue != modes->bindToTextureTargets)
+         if (config->bindToTextureTargets != GLX_DONT_CARE &&
+             glxValue != config->bindToTextureTargets)
             return GL_FALSE;
          break;
 
       default:
-         if (!scalarEqual(modes, attrib, value))
+         if (!scalarEqual(config, attrib, value))
             return GL_FALSE;
       }
    }
@@ -297,41 +294,41 @@ driConfigEqual(const __DRIcoreExtension * core,
    return GL_TRUE;
 }
 
-static __GLcontextModes *
+static struct glx_config *
 createDriMode(const __DRIcoreExtension * core,
-              __GLcontextModes * modes, const __DRIconfig ** driConfigs)
+	      struct glx_config *config, const __DRIconfig **driConfigs)
 {
-   __GLXDRIconfigPrivate *config;
+   __GLXDRIconfigPrivate *driConfig;
    int i;
 
    for (i = 0; driConfigs[i]; i++) {
-      if (driConfigEqual(core, modes, driConfigs[i]))
+      if (driConfigEqual(core, config, driConfigs[i]))
          break;
    }
 
    if (driConfigs[i] == NULL)
       return NULL;
 
-   config = Xmalloc(sizeof *config);
-   if (config == NULL)
+   driConfig = Xmalloc(sizeof *driConfig);
+   if (driConfig == NULL)
       return NULL;
 
-   config->modes = *modes;
-   config->driConfig = driConfigs[i];
+   driConfig->base = *config;
+   driConfig->driConfig = driConfigs[i];
 
-   return &config->modes;
+   return &driConfig->base;
 }
 
-_X_HIDDEN __GLcontextModes *
+_X_HIDDEN struct glx_config *
 driConvertConfigs(const __DRIcoreExtension * core,
-                  __GLcontextModes * modes, const __DRIconfig ** configs)
+                  struct glx_config *configs, const __DRIconfig **driConfigs)
 {
-   __GLcontextModes head, *tail, *m;
+   struct glx_config head, *tail, *m;
 
    tail = &head;
    head.next = NULL;
-   for (m = modes; m; m = m->next) {
-      tail->next = createDriMode(core, m, configs);
+   for (m = configs; m; m = m->next) {
+      tail->next = createDriMode(core, m, driConfigs);
       if (tail->next == NULL) {
          /* no matching dri config for m */
          continue;
@@ -341,7 +338,7 @@ driConvertConfigs(const __DRIcoreExtension * core,
       tail = tail->next;
    }
 
-   _gl_context_modes_destroy(modes);
+   glx_config_destroy_list(configs);
 
    return head.next;
 }
@@ -356,4 +353,31 @@ driDestroyConfigs(const __DRIconfig **configs)
    free(configs);
 }
 
+_X_HIDDEN __GLXDRIdrawable *
+driFetchDrawable(struct glx_context *gc, GLXDrawable glxDrawable)
+{
+   struct glx_display *const priv = __glXInitialize(gc->psc->dpy);
+   __GLXDRIdrawable *pdraw;
+   struct glx_screen *psc;
+
+   if (priv == NULL)
+      return NULL;
+
+   psc = priv->screens[gc->screen];
+   if (priv->drawHash == NULL)
+      return NULL;
+
+   if (__glxHashLookup(priv->drawHash, glxDrawable, (void *) &pdraw) == 0)
+      return pdraw;
+
+   pdraw = psc->driScreen->createDrawable(psc, glxDrawable,
+                                          glxDrawable, gc->config);
+   if (__glxHashInsert(priv->drawHash, glxDrawable, pdraw)) {
+      (*pdraw->destroyDrawable) (pdraw);
+      return NULL;
+   }
+
+   return pdraw;
+}
+
 #endif /* GLX_DIRECT_RENDERING */
diff --git a/src/glx/dri_common.h b/src/glx/dri_common.h
index f3da50ecf09..846a905a880 100644
--- a/src/glx/dri_common.h
+++ b/src/glx/dri_common.h
@@ -42,16 +42,19 @@ typedef struct __GLXDRIconfigPrivateRec __GLXDRIconfigPrivate;
 
 struct __GLXDRIconfigPrivateRec
 {
-   __GLcontextModes modes;
+   struct glx_config base;
    const __DRIconfig *driConfig;
 };
 
-extern __GLcontextModes *driConvertConfigs(const __DRIcoreExtension * core,
-                                           __GLcontextModes * modes,
+extern struct glx_config *driConvertConfigs(const __DRIcoreExtension * core,
+                                           struct glx_config * modes,
                                            const __DRIconfig ** configs);
 
 extern void driDestroyConfigs(const __DRIconfig **configs);
 
+extern __GLXDRIdrawable *
+driFetchDrawable(struct glx_context *gc, GLXDrawable glxDrawable);
+
 extern const __DRIsystemTimeExtension systemTimeExtension;
 
 extern void InfoMessageF(const char *f, ...);
diff --git a/src/glx/dri_glx.c b/src/glx/dri_glx.c
index 352d833fd7d..43a2aa495a7 100644
--- a/src/glx/dri_glx.c
+++ b/src/glx/dri_glx.c
@@ -61,7 +61,7 @@ struct dri_display
 
 struct dri_screen
 {
-   __GLXscreenConfigs base;
+   struct glx_screen base;
 
    __DRIscreen *driScreen;
    __GLXDRIscreen vtable;
@@ -78,11 +78,9 @@ struct dri_screen
 
 struct dri_context
 {
-   __GLXcontext base;
-   __GLXDRIcontext dri_vtable;
+   struct glx_context base;
    __DRIcontext *driContext;
    XID hwContextID;
-   __GLXscreenConfigs *psc;
 };
 
 struct dri_drawable
@@ -92,13 +90,7 @@ struct dri_drawable
    __DRIdrawable *driDrawable;
 };
 
-static const struct glx_context_vtable dri_context_vtable = {
-   NULL,
-   NULL,
-   DRI_glXUseXFont,
-   NULL,
-   NULL,
-};
+static const struct glx_context_vtable dri_context_vtable;
 
 /*
  * Given a display pointer and screen number, determine the name of
@@ -158,7 +150,7 @@ driGetDriverName(Display * dpy, int scrNum, char **driverName)
  * The returned char pointer points to a static array that will be
  * overwritten by subsequent calls.
  */
-PUBLIC const char *
+_X_EXPORT const char *
 glXGetScreenDriver(Display * dpy, int scrNum)
 {
    static char ret[32];
@@ -188,7 +180,7 @@ glXGetScreenDriver(Display * dpy, int scrNum)
  *
  * Note: The driver remains opened after this function returns.
  */
-PUBLIC const char *
+_X_EXPORT const char *
 glXGetDriverConfig(const char *driverName)
 {
    void *handle = driOpenDriver(driverName);
@@ -233,7 +225,7 @@ __glXReportDamage(__DRIdrawable * driDraw,
    int i;
    int x_off, y_off;
    __GLXDRIdrawable *glxDraw = loaderPrivate;
-   __GLXscreenConfigs *psc = glxDraw->psc;
+   struct glx_screen *psc = glxDraw->psc;
    Display *dpy = psc->dpy;
    Drawable drawable;
 
@@ -285,7 +277,7 @@ __glXDRIGetDrawableInfo(__DRIdrawable * drawable,
                         void *loaderPrivate)
 {
    __GLXDRIdrawable *glxDraw = loaderPrivate;
-   __GLXscreenConfigs *psc = glxDraw->psc;
+   struct glx_screen *psc = glxDraw->psc;
    Display *dpy = psc->dpy;
 
    return XF86DRIGetDrawableInfo(dpy, psc->scr, glxDraw->drawable,
@@ -344,7 +336,7 @@ CallCreateNewScreen(Display *dpy, int scrn, struct dri_screen *psc,
    drm_handle_t hFB;
    int junk;
    const __DRIconfig **driver_configs;
-   __GLcontextModes *visual;
+   struct glx_config *visual;
 
    /* DRI protocol version. */
    dri_version.major = driDpy->driMajor;
@@ -506,55 +498,82 @@ CallCreateNewScreen(Display *dpy, int scrn, struct dri_screen *psc,
 }
 
 static void
-driDestroyContext(__GLXcontext * context)
+dri_destroy_context(struct glx_context * context)
 {
    struct dri_context *pcp = (struct dri_context *) context;
    struct dri_screen *psc = (struct dri_screen *) context->psc;
 
+   if (context->xid)
+      glx_send_destroy_context(psc->base.dpy, context->xid);
+
+   if (context->extensions)
+      XFree((char *) context->extensions);
+
+   GarbageCollectDRIDrawables(context->psc);
+
    (*psc->core->destroyContext) (pcp->driContext);
 
    XF86DRIDestroyContext(psc->base.dpy, psc->base.scr, pcp->hwContextID);
    Xfree(pcp);
 }
 
-static Bool
-driBindContext(__GLXcontext *context,
-	       __GLXDRIdrawable *draw, __GLXDRIdrawable *read)
+static int
+dri_bind_context(struct glx_context *context, struct glx_context *old,
+		 GLXDrawable draw, GLXDrawable read)
 {
    struct dri_context *pcp = (struct dri_context *) context;
-   struct dri_screen *psc = (struct dri_screen *) pcp->psc;
-   struct dri_drawable *pdr = (struct dri_drawable *) draw;
-   struct dri_drawable *prd = (struct dri_drawable *) read;
+   struct dri_screen *psc = (struct dri_screen *) pcp->base.psc;
+   struct dri_drawable *pdraw, *pread;
+
+   pdraw = (struct dri_drawable *) driFetchDrawable(context, draw);
+   pread = (struct dri_drawable *) driFetchDrawable(context, read);
 
-   return (*psc->core->bindContext) (pcp->driContext,
-				     pdr->driDrawable, prd->driDrawable);
+   if (pdraw == NULL || pread == NULL)
+      return GLXBadDrawable;
+
+   if ((*psc->core->bindContext) (pcp->driContext,
+				  pdraw->driDrawable, pread->driDrawable))
+      return Success;
+
+   return GLXBadContext;
 }
 
 static void
-driUnbindContext(__GLXcontext * context)
+dri_unbind_context(struct glx_context *context, struct glx_context *new)
 {
    struct dri_context *pcp = (struct dri_context *) context;
-   struct dri_screen *psc = (struct dri_screen *) pcp->psc;
+   struct dri_screen *psc = (struct dri_screen *) pcp->base.psc;
 
    (*psc->core->unbindContext) (pcp->driContext);
 }
 
-static __GLXcontext *
-driCreateContext(__GLXscreenConfigs *base,
-                 const __GLcontextModes * mode,
-                 GLXContext shareList, int renderType)
+static const struct glx_context_vtable dri_context_vtable = {
+   dri_destroy_context,
+   dri_bind_context,
+   dri_unbind_context,
+   NULL,
+   NULL,
+   DRI_glXUseXFont,
+   NULL,
+   NULL,
+};
+
+static struct glx_context *
+dri_create_context(struct glx_screen *base,
+		   struct glx_config *config_base,
+		   struct glx_context *shareList, int renderType)
 {
    struct dri_context *pcp, *pcp_shared;
    struct dri_screen *psc = (struct dri_screen *) base;
    drm_context_t hwContext;
    __DRIcontext *shared = NULL;
-   __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) mode;
+   __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) config_base;
 
    if (!psc->base.driScreen)
       return NULL;
 
    if (shareList) {
-      pcp_shared = (struct dri_context *) shareList->driContext;
+      pcp_shared = (struct dri_context *) shareList;
       shared = pcp_shared->driContext;
    }
 
@@ -563,13 +582,13 @@ driCreateContext(__GLXscreenConfigs *base,
       return NULL;
 
    memset(pcp, 0, sizeof *pcp);
-   if (!glx_context_init(&pcp->base, &psc->base, mode)) {
+   if (!glx_context_init(&pcp->base, &psc->base, &config->base)) {
       Xfree(pcp);
       return NULL;
    }
 
    if (!XF86DRICreateContextWithConfig(psc->base.dpy, psc->base.scr,
-                                       mode->visualID,
+                                       config->base.visualID,
                                        &pcp->hwContextID, &hwContext)) {
       Xfree(pcp);
       return NULL;
@@ -586,10 +605,6 @@ driCreateContext(__GLXscreenConfigs *base,
    }
 
    pcp->base.vtable = &dri_context_vtable;
-   pcp->base.driContext = &pcp->dri_vtable;
-   pcp->dri_vtable.destroyContext = driDestroyContext;
-   pcp->dri_vtable.bindContext = driBindContext;
-   pcp->dri_vtable.unbindContext = driUnbindContext;
 
    return &pcp->base;
 }
@@ -606,13 +621,13 @@ driDestroyDrawable(__GLXDRIdrawable * pdraw)
 }
 
 static __GLXDRIdrawable *
-driCreateDrawable(__GLXscreenConfigs *base,
+driCreateDrawable(struct glx_screen *base,
                   XID xDrawable,
-                  GLXDrawable drawable, const __GLcontextModes * modes)
+                  GLXDrawable drawable, struct glx_config *config_base)
 {
    drm_drawable_t hwDrawable;
    void *empty_attribute_list = NULL;
-   __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) modes;
+   __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) config_base;
    struct dri_screen *psc = (struct dri_screen *) base;
    struct dri_drawable *pdp;
 
@@ -624,6 +639,7 @@ driCreateDrawable(__GLXscreenConfigs *base,
    if (!pdp)
       return NULL;
 
+   memset(pdp, 0, sizeof *pdp);
    pdp->base.drawable = drawable;
    pdp->base.psc = &psc->base;
 
@@ -675,7 +691,7 @@ driCopySubBuffer(__GLXDRIdrawable * pdraw,
 }
 
 static void
-driDestroyScreen(__GLXscreenConfigs *base)
+driDestroyScreen(struct glx_screen *base)
 {
    struct dri_screen *psc = (struct dri_screen *) base;
 
@@ -691,7 +707,7 @@ driDestroyScreen(__GLXscreenConfigs *base)
 #ifdef __DRI_SWAP_BUFFER_COUNTER
 
 static int
-driDrawableGetMSC(__GLXscreenConfigs *base, __GLXDRIdrawable *pdraw,
+driDrawableGetMSC(struct glx_screen *base, __GLXDRIdrawable *pdraw,
 		   int64_t *ust, int64_t *msc, int64_t *sbc)
 {
    struct dri_screen *psc = (struct dri_screen *) base;
@@ -746,17 +762,12 @@ driWaitForSBC(__GLXDRIdrawable *pdraw, int64_t target_sbc, int64_t *ust,
 static int
 driSetSwapInterval(__GLXDRIdrawable *pdraw, int interval)
 {
-   GLXContext gc = __glXGetCurrentContext();
    struct dri_drawable *pdp = (struct dri_drawable *) pdraw;
-   struct dri_screen *psc;
-
-   if (gc->driContext) {
-      psc = (struct dri_screen *) pdraw->psc;
+   struct dri_screen *psc = (struct dri_screen *) pdraw->psc;
 
-      if (psc->swapControl != NULL && pdraw != NULL) {
-	 psc->swapControl->setSwapInterval(pdp->driDrawable, interval);
-	 return 0;
-      }
+   if (psc->swapControl != NULL && pdraw != NULL) {
+      psc->swapControl->setSwapInterval(pdp->driDrawable, interval);
+      return 0;
    }
 
    return GLX_BAD_CONTEXT;
@@ -765,17 +776,11 @@ driSetSwapInterval(__GLXDRIdrawable *pdraw, int interval)
 static int
 driGetSwapInterval(__GLXDRIdrawable *pdraw)
 {
-   GLXContext gc = __glXGetCurrentContext();
    struct dri_drawable *pdp = (struct dri_drawable *) pdraw;
-   struct dri_screen *psc;
-
-   if (gc != NULL && gc->driContext) {
-      psc = (struct dri_screen *) pdraw->psc;
+   struct dri_screen *psc = (struct dri_screen *) pdraw->psc;
 
-      if (psc->swapControl != NULL && pdraw != NULL) {
-	 return psc->swapControl->getSwapInterval(pdp->driDrawable);
-      }
-   }
+   if (psc->swapControl != NULL && pdraw != NULL)
+      return psc->swapControl->getSwapInterval(pdp->driDrawable);
 
    return 0;
 }
@@ -812,8 +817,12 @@ driBindExtensions(struct dri_screen *psc, const __DRIextension **extensions)
    }
 }
 
-static __GLXscreenConfigs *
-driCreateScreen(int screen, __GLXdisplayPrivate *priv)
+static const struct glx_screen_vtable dri_screen_vtable = {
+   dri_create_context
+};
+
+static struct glx_screen *
+driCreateScreen(int screen, struct glx_display *priv)
 {
    struct dri_display *pdp;
    __GLXDRIscreen *psp;
@@ -873,13 +882,13 @@ driCreateScreen(int screen, __GLXdisplayPrivate *priv)
    extensions = psc->core->getExtensions(psc->driScreen);
    driBindExtensions(psc, extensions);
 
+   psc->base.vtable = &dri_screen_vtable;
    psp = &psc->vtable;
    psc->base.driScreen = psp;
    if (psc->driCopySubBuffer)
       psp->copySubBuffer = driCopySubBuffer;
 
    psp->destroyScreen = driDestroyScreen;
-   psp->createContext = driCreateContext;
    psp->createDrawable = driCreateDrawable;
    psp->swapBuffers = driSwapBuffers;
 
diff --git a/src/glx/drisw_glx.c b/src/glx/drisw_glx.c
index 0ad73914575..c5b179157b3 100644
--- a/src/glx/drisw_glx.c
+++ b/src/glx/drisw_glx.c
@@ -35,15 +35,14 @@ struct drisw_display
 
 struct drisw_context
 {
-   __GLXcontext base;
-   __GLXDRIcontext dri_vtable;
+   struct glx_context base;
    __DRIcontext *driContext;
-   __GLXscreenConfigs *psc;
+
 };
 
 struct drisw_screen
 {
-   __GLXscreenConfigs base;
+   struct glx_screen base;
 
    __DRIscreen *driScreen;
    __GLXDRIscreen vtable;
@@ -240,39 +239,58 @@ static const __DRIextension *loader_extensions[] = {
  */
 
 static void
-driDestroyContext(__GLXcontext *context)
+drisw_destroy_context(struct glx_context *context)
 {
    struct drisw_context *pcp = (struct drisw_context *) context;
    struct drisw_screen *psc = (struct drisw_screen *) context->psc;
 
+   if (context->xid)
+      glx_send_destroy_context(psc->base.dpy, context->xid);
+
+   if (context->extensions)
+      XFree((char *) context->extensions);
+
+   GarbageCollectDRIDrawables(context->psc);
+
    (*psc->core->destroyContext) (pcp->driContext);
 
    Xfree(pcp);
 }
 
-static Bool
-driBindContext(__GLXcontext * context,
-	       __GLXDRIdrawable * draw, __GLXDRIdrawable * read)
+static int
+drisw_bind_context(struct glx_context *context, struct glx_context *old,
+		   GLXDrawable draw, GLXDrawable read)
 {
    struct drisw_context *pcp = (struct drisw_context *) context;
-   struct drisw_screen *psc = (struct drisw_screen *) pcp->psc;
-   struct drisw_drawable *pdr = (struct drisw_drawable *) draw;
-   struct drisw_drawable *prd = (struct drisw_drawable *) read;
+   struct drisw_screen *psc = (struct drisw_screen *) pcp->base.psc;
+   struct drisw_drawable *pdraw, *pread;
+
+   pdraw = (struct drisw_drawable *) driFetchDrawable(context, draw);
+   pread = (struct drisw_drawable *) driFetchDrawable(context, read);
+
+   if (pdraw == NULL || pread == NULL)
+      return GLXBadDrawable;
 
-   return (*psc->core->bindContext) (pcp->driContext,
-				     pdr->driDrawable, prd->driDrawable);
+   if ((*psc->core->bindContext) (pcp->driContext,
+				  pdraw->driDrawable, pread->driDrawable))
+      return Success;
+
+   return GLXBadContext;
 }
 
 static void
-driUnbindContext(__GLXcontext * context)
+drisw_unbind_context(struct glx_context *context, struct glx_context *new)
 {
    struct drisw_context *pcp = (struct drisw_context *) context;
-   struct drisw_screen *psc = (struct drisw_screen *) pcp->psc;
+   struct drisw_screen *psc = (struct drisw_screen *) pcp->base.psc;
 
    (*psc->core->unbindContext) (pcp->driContext);
 }
 
 static const struct glx_context_vtable drisw_context_vtable = {
+   drisw_destroy_context,
+   drisw_bind_context,
+   drisw_unbind_context,
    NULL,
    NULL,
    DRI_glXUseXFont,
@@ -280,13 +298,13 @@ static const struct glx_context_vtable drisw_context_vtable = {
    NULL,
 };
 
-static __GLXcontext *
-driCreateContext(__GLXscreenConfigs *base,
-		 const __GLcontextModes *mode,
-		 GLXContext shareList, int renderType)
+static struct glx_context *
+drisw_create_context(struct glx_screen *base,
+		     struct glx_config *config_base,
+		     struct glx_context *shareList, int renderType)
 {
    struct drisw_context *pcp, *pcp_shared;
-   __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) mode;
+   __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) config_base;
    struct drisw_screen *psc = (struct drisw_screen *) base;
    __DRIcontext *shared = NULL;
 
@@ -294,7 +312,7 @@ driCreateContext(__GLXscreenConfigs *base,
       return NULL;
 
    if (shareList) {
-      pcp_shared = (struct drisw_context *) shareList->driContext;
+      pcp_shared = (struct drisw_context *) shareList;
       shared = pcp_shared->driContext;
    }
 
@@ -303,7 +321,7 @@ driCreateContext(__GLXscreenConfigs *base,
       return NULL;
 
    memset(pcp, 0, sizeof *pcp);
-   if (!glx_context_init(&pcp->base, &psc->base, mode)) {
+   if (!glx_context_init(&pcp->base, &psc->base, &config->base)) {
       Xfree(pcp);
       return NULL;
    }
@@ -317,10 +335,6 @@ driCreateContext(__GLXscreenConfigs *base,
    }
 
    pcp->base.vtable = &drisw_context_vtable;
-   pcp->base.driContext = &pcp->dri_vtable;
-   pcp->dri_vtable.destroyContext = driDestroyContext;
-   pcp->dri_vtable.bindContext = driBindContext;
-   pcp->dri_vtable.unbindContext = driUnbindContext;
 
    return &pcp->base;
 }
@@ -338,8 +352,8 @@ driDestroyDrawable(__GLXDRIdrawable * pdraw)
 }
 
 static __GLXDRIdrawable *
-driCreateDrawable(__GLXscreenConfigs *base, XID xDrawable,
-		  GLXDrawable drawable, const __GLcontextModes * modes)
+driCreateDrawable(struct glx_screen *base, XID xDrawable,
+		  GLXDrawable drawable, struct glx_config *modes)
 {
    struct drisw_drawable *pdp;
    __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) modes;
@@ -355,6 +369,7 @@ driCreateDrawable(__GLXscreenConfigs *base, XID xDrawable,
    if (!pdp)
       return NULL;
 
+   memset(pdp, 0, sizeof *pdp);
    pdp->base.xDrawable = xDrawable;
    pdp->base.drawable = drawable;
    pdp->base.psc = &psc->base;
@@ -393,7 +408,7 @@ driSwapBuffers(__GLXDRIdrawable * pdraw,
 }
 
 static void
-driDestroyScreen(__GLXscreenConfigs *base)
+driDestroyScreen(struct glx_screen *base)
 {
    struct drisw_screen *psc = (struct drisw_screen *) base;
 
@@ -419,8 +434,12 @@ driOpenSwrast(void)
    return driver;
 }
 
-static __GLXscreenConfigs *
-driCreateScreen(int screen, __GLXdisplayPrivate *priv)
+static const struct glx_screen_vtable drisw_screen_vtable = {
+   drisw_create_context
+};
+
+static struct glx_screen *
+driCreateScreen(int screen, struct glx_display *priv)
 {
    __GLXDRIscreen *psp;
    const __DRIconfig **driver_configs;
@@ -466,8 +485,6 @@ driCreateScreen(int screen, __GLXdisplayPrivate *priv)
       goto handle_error;
    }
 
-   extensions = psc->core->getExtensions(psc->driScreen);
-
    psc->base.configs =
       driConvertConfigs(psc->core, psc->base.configs, driver_configs);
    psc->base.visuals =
@@ -475,20 +492,19 @@ driCreateScreen(int screen, __GLXdisplayPrivate *priv)
 
    psc->driver_configs = driver_configs;
 
+   psc->base.vtable = &drisw_screen_vtable;
    psp = &psc->vtable;
    psc->base.driScreen = psp;
    psp->destroyScreen = driDestroyScreen;
-   psp->createContext = driCreateContext;
    psp->createDrawable = driCreateDrawable;
    psp->swapBuffers = driSwapBuffers;
 
    return &psc->base;
 
  handle_error:
-   Xfree(psc);
-
    if (psc->driver)
       dlclose(psc->driver);
+   Xfree(psc);
 
    ErrorMessageF("reverting to indirect rendering\n");
 
diff --git a/src/glx/glcontextmodes.c b/src/glx/glcontextmodes.c
deleted file mode 100644
index cdc16f8330a..00000000000
--- a/src/glx/glcontextmodes.c
+++ /dev/null
@@ -1,518 +0,0 @@
-/*
- * (C) Copyright IBM Corporation 2003
- * All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * on the rights to use, copy, modify, merge, publish, distribute, sub
- * license, and/or sell copies of the Software, and to permit persons to whom
- * the Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
- * VA LINUX SYSTEM, IBM AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
- * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
- * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
- * USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-
-/**
- * \file glcontextmodes.c
- * Utility routines for working with \c __GLcontextModes structures.  At
- * some point most or all of these functions will be moved to the Mesa
- * code base.
- *
- * \author Ian Romanick <[email protected]>
- */
-
-#include <GL/glx.h>
-#include "GL/glxint.h"
-#include <stdlib.h>
-#include <string.h>
-
-#include "glcontextmodes.h"
-
-#define NUM_VISUAL_TYPES   6
-
-/**
- * Convert an X visual type to a GLX visual type.
- * 
- * \param visualType X visual type (i.e., \c TrueColor, \c StaticGray, etc.)
- *        to be converted.
- * \return If \c visualType is a valid X visual type, a GLX visual type will
- *         be returned.  Otherwise \c GLX_NONE will be returned.
- */
-GLint
-_gl_convert_from_x_visual_type(int visualType)
-{
-   static const int glx_visual_types[NUM_VISUAL_TYPES] = {
-      GLX_STATIC_GRAY, GLX_GRAY_SCALE,
-      GLX_STATIC_COLOR, GLX_PSEUDO_COLOR,
-      GLX_TRUE_COLOR, GLX_DIRECT_COLOR
-   };
-
-   return ((unsigned) visualType < NUM_VISUAL_TYPES)
-      ? glx_visual_types[visualType] : GLX_NONE;
-}
-
-
-/**
- * Convert a GLX visual type to an X visual type.
- * 
- * \param visualType GLX visual type (i.e., \c GLX_TRUE_COLOR, 
- *                   \c GLX_STATIC_GRAY, etc.) to be converted.
- * \return If \c visualType is a valid GLX visual type, an X visual type will
- *         be returned.  Otherwise -1 will be returned.
- */
-GLint
-_gl_convert_to_x_visual_type(int visualType)
-{
-   static const int x_visual_types[NUM_VISUAL_TYPES] = {
-      TrueColor, DirectColor,
-      PseudoColor, StaticColor,
-      GrayScale, StaticGray
-   };
-
-   return ((unsigned) (visualType - GLX_TRUE_COLOR) < NUM_VISUAL_TYPES)
-      ? x_visual_types[visualType - GLX_TRUE_COLOR] : -1;
-}
-
-
-/**
- * Copy a GLX visual config structure to a GL context mode structure.  All
- * of the fields in \c config are copied to \c mode.  Additional fields in
- * \c mode that can be derrived from the fields of \c config (i.e.,
- * \c haveDepthBuffer) are also filled in.  The remaining fields in \c mode
- * that cannot be derived are set to default values.
- * 
- * \param mode   Destination GL context mode.
- * \param config Source GLX visual config.
- * 
- * \note
- * The \c fbconfigID and \c visualID fields of the \c __GLcontextModes
- * structure will be set to the \c vid of the \c __GLXvisualConfig structure.
- */
-void
-_gl_copy_visual_to_context_mode(__GLcontextModes * mode,
-                                const __GLXvisualConfig * config)
-{
-   __GLcontextModes *const next = mode->next;
-
-   (void) memset(mode, 0, sizeof(__GLcontextModes));
-   mode->next = next;
-
-   mode->visualID = config->vid;
-   mode->visualType = _gl_convert_from_x_visual_type(config->class);
-   mode->xRenderable = GL_TRUE;
-   mode->fbconfigID = config->vid;
-   mode->drawableType = GLX_WINDOW_BIT | GLX_PIXMAP_BIT;
-
-   mode->rgbMode = (config->rgba != 0);
-   mode->renderType = (mode->rgbMode) ? GLX_RGBA_BIT : GLX_COLOR_INDEX_BIT;
-
-   mode->colorIndexMode = !(mode->rgbMode);
-   mode->doubleBufferMode = (config->doubleBuffer != 0);
-   mode->stereoMode = (config->stereo != 0);
-
-   mode->haveAccumBuffer = ((config->accumRedSize +
-                             config->accumGreenSize +
-                             config->accumBlueSize +
-                             config->accumAlphaSize) > 0);
-   mode->haveDepthBuffer = (config->depthSize > 0);
-   mode->haveStencilBuffer = (config->stencilSize > 0);
-
-   mode->redBits = config->redSize;
-   mode->greenBits = config->greenSize;
-   mode->blueBits = config->blueSize;
-   mode->alphaBits = config->alphaSize;
-   mode->redMask = config->redMask;
-   mode->greenMask = config->greenMask;
-   mode->blueMask = config->blueMask;
-   mode->alphaMask = config->alphaMask;
-   mode->rgbBits = mode->rgbMode ? config->bufferSize : 0;
-   mode->indexBits = mode->colorIndexMode ? config->bufferSize : 0;
-
-   mode->accumRedBits = config->accumRedSize;
-   mode->accumGreenBits = config->accumGreenSize;
-   mode->accumBlueBits = config->accumBlueSize;
-   mode->accumAlphaBits = config->accumAlphaSize;
-   mode->depthBits = config->depthSize;
-   mode->stencilBits = config->stencilSize;
-
-   mode->numAuxBuffers = config->auxBuffers;
-   mode->level = config->level;
-
-   mode->visualRating = config->visualRating;
-   mode->transparentPixel = config->transparentPixel;
-   mode->transparentRed = config->transparentRed;
-   mode->transparentGreen = config->transparentGreen;
-   mode->transparentBlue = config->transparentBlue;
-   mode->transparentAlpha = config->transparentAlpha;
-   mode->transparentIndex = config->transparentIndex;
-   mode->samples = config->multiSampleSize;
-   mode->sampleBuffers = config->nMultiSampleBuffers;
-   /* mode->visualSelectGroup = config->visualSelectGroup; ? */
-
-   mode->swapMethod = GLX_SWAP_UNDEFINED_OML;
-
-   mode->bindToTextureRgb = (mode->rgbMode) ? GL_TRUE : GL_FALSE;
-   mode->bindToTextureRgba = (mode->rgbMode && mode->alphaBits) ?
-      GL_TRUE : GL_FALSE;
-   mode->bindToMipmapTexture = mode->rgbMode ? GL_TRUE : GL_FALSE;
-   mode->bindToTextureTargets = mode->rgbMode ?
-      GLX_TEXTURE_1D_BIT_EXT | GLX_TEXTURE_2D_BIT_EXT |
-      GLX_TEXTURE_RECTANGLE_BIT_EXT : 0;
-   mode->yInverted = GL_FALSE;
-}
-
-
-/**
- * Get data from a GL context mode.
- * 
- * \param mode         GL context mode whose data is to be returned.
- * \param attribute    Attribute of \c mode that is to be returned.
- * \param value_return Location to store the data member of \c mode.
- * \return  If \c attribute is a valid attribute of \c mode, zero is
- *          returned.  Otherwise \c GLX_BAD_ATTRIBUTE is returned.
- */
-int
-_gl_get_context_mode_data(const __GLcontextModes * mode, int attribute,
-                          int *value_return)
-{
-   switch (attribute) {
-   case GLX_USE_GL:
-      *value_return = GL_TRUE;
-      return 0;
-   case GLX_BUFFER_SIZE:
-      *value_return = mode->rgbBits;
-      return 0;
-   case GLX_RGBA:
-      *value_return = mode->rgbMode;
-      return 0;
-   case GLX_RED_SIZE:
-      *value_return = mode->redBits;
-      return 0;
-   case GLX_GREEN_SIZE:
-      *value_return = mode->greenBits;
-      return 0;
-   case GLX_BLUE_SIZE:
-      *value_return = mode->blueBits;
-      return 0;
-   case GLX_ALPHA_SIZE:
-      *value_return = mode->alphaBits;
-      return 0;
-   case GLX_DOUBLEBUFFER:
-      *value_return = mode->doubleBufferMode;
-      return 0;
-   case GLX_STEREO:
-      *value_return = mode->stereoMode;
-      return 0;
-   case GLX_AUX_BUFFERS:
-      *value_return = mode->numAuxBuffers;
-      return 0;
-   case GLX_DEPTH_SIZE:
-      *value_return = mode->depthBits;
-      return 0;
-   case GLX_STENCIL_SIZE:
-      *value_return = mode->stencilBits;
-      return 0;
-   case GLX_ACCUM_RED_SIZE:
-      *value_return = mode->accumRedBits;
-      return 0;
-   case GLX_ACCUM_GREEN_SIZE:
-      *value_return = mode->accumGreenBits;
-      return 0;
-   case GLX_ACCUM_BLUE_SIZE:
-      *value_return = mode->accumBlueBits;
-      return 0;
-   case GLX_ACCUM_ALPHA_SIZE:
-      *value_return = mode->accumAlphaBits;
-      return 0;
-   case GLX_LEVEL:
-      *value_return = mode->level;
-      return 0;
-#ifndef GLX_USE_APPLEGL               /* This isn't supported by CGL. */
-   case GLX_TRANSPARENT_TYPE_EXT:
-      *value_return = mode->transparentPixel;
-      return 0;
-#endif
-   case GLX_TRANSPARENT_RED_VALUE:
-      *value_return = mode->transparentRed;
-      return 0;
-   case GLX_TRANSPARENT_GREEN_VALUE:
-      *value_return = mode->transparentGreen;
-      return 0;
-   case GLX_TRANSPARENT_BLUE_VALUE:
-      *value_return = mode->transparentBlue;
-      return 0;
-   case GLX_TRANSPARENT_ALPHA_VALUE:
-      *value_return = mode->transparentAlpha;
-      return 0;
-   case GLX_TRANSPARENT_INDEX_VALUE:
-      *value_return = mode->transparentIndex;
-      return 0;
-   case GLX_X_VISUAL_TYPE:
-      *value_return = mode->visualType;
-      return 0;
-   case GLX_CONFIG_CAVEAT:
-      *value_return = mode->visualRating;
-      return 0;
-   case GLX_VISUAL_ID:
-      *value_return = mode->visualID;
-      return 0;
-   case GLX_DRAWABLE_TYPE:
-      *value_return = mode->drawableType;
-      return 0;
-   case GLX_RENDER_TYPE:
-      *value_return = mode->renderType;
-      return 0;
-   case GLX_X_RENDERABLE:
-      *value_return = mode->xRenderable;
-      return 0;
-   case GLX_FBCONFIG_ID:
-      *value_return = mode->fbconfigID;
-      return 0;
-   case GLX_MAX_PBUFFER_WIDTH:
-      *value_return = mode->maxPbufferWidth;
-      return 0;
-   case GLX_MAX_PBUFFER_HEIGHT:
-      *value_return = mode->maxPbufferHeight;
-      return 0;
-   case GLX_MAX_PBUFFER_PIXELS:
-      *value_return = mode->maxPbufferPixels;
-      return 0;
-#ifndef GLX_USE_APPLEGL               /* These aren't supported by CGL. */
-   case GLX_OPTIMAL_PBUFFER_WIDTH_SGIX:
-      *value_return = mode->optimalPbufferWidth;
-      return 0;
-   case GLX_OPTIMAL_PBUFFER_HEIGHT_SGIX:
-      *value_return = mode->optimalPbufferHeight;
-      return 0;
-   case GLX_SWAP_METHOD_OML:
-      *value_return = mode->swapMethod;
-      return 0;
-#endif
-   case GLX_SAMPLE_BUFFERS_SGIS:
-      *value_return = mode->sampleBuffers;
-      return 0;
-   case GLX_SAMPLES_SGIS:
-      *value_return = mode->samples;
-      return 0;
-   case GLX_BIND_TO_TEXTURE_RGB_EXT:
-      *value_return = mode->bindToTextureRgb;
-      return 0;
-   case GLX_BIND_TO_TEXTURE_RGBA_EXT:
-      *value_return = mode->bindToTextureRgba;
-      return 0;
-   case GLX_BIND_TO_MIPMAP_TEXTURE_EXT:
-      *value_return = mode->bindToMipmapTexture == GL_TRUE ? GL_TRUE :
-         GL_FALSE;
-      return 0;
-   case GLX_BIND_TO_TEXTURE_TARGETS_EXT:
-      *value_return = mode->bindToTextureTargets;
-      return 0;
-   case GLX_Y_INVERTED_EXT:
-      *value_return = mode->yInverted;
-      return 0;
-
-      /* Applications are NOT allowed to query GLX_VISUAL_SELECT_GROUP_SGIX.
-       * It is ONLY for communication between the GLX client and the GLX
-       * server.
-       */
-   case GLX_VISUAL_SELECT_GROUP_SGIX:
-   default:
-      return GLX_BAD_ATTRIBUTE;
-   }
-}
-
-
-/**
- * Allocate a linked list of \c __GLcontextModes structures.  The fields of
- * each structure will be initialized to "reasonable" default values.  In
- * most cases this is the default value defined by table 3.4 of the GLX
- * 1.3 specification.  This means that most values are either initialized to
- * zero or \c GLX_DONT_CARE (which is -1).  As support for additional
- * extensions is added, the new values will be initialized to appropriate
- * values from the extension specification.
- * 
- * \param count         Number of structures to allocate.
- * \param minimum_size  Minimum size of a structure to allocate.  This allows
- *                      for differences in the version of the
- *                      \c __GLcontextModes stucture used in libGL and in a
- *                      DRI-based driver.
- * \returns A pointer to the first element in a linked list of \c count
- *          stuctures on success, or \c NULL on failure.
- * 
- * \warning Use of \c minimum_size does \b not guarantee binary compatibility.
- *          The fundamental assumption is that if the \c minimum_size
- *          specified by the driver and the size of the \c __GLcontextModes
- *          structure in libGL is the same, then the meaning of each byte in
- *          the structure is the same in both places.  \b Be \b careful!
- *          Basically this means that fields have to be added in libGL and
- *          then propagated to drivers.  Drivers should \b never arbitrarilly
- *          extend the \c __GLcontextModes data-structure.
- */
-__GLcontextModes *
-_gl_context_modes_create(unsigned count, size_t minimum_size)
-{
-   const size_t size = (minimum_size > sizeof(__GLcontextModes))
-      ? minimum_size : sizeof(__GLcontextModes);
-   __GLcontextModes *base = NULL;
-   __GLcontextModes **next;
-   unsigned i;
-
-   next = &base;
-   for (i = 0; i < count; i++) {
-      *next = (__GLcontextModes *) malloc(size);
-      if (*next == NULL) {
-         _gl_context_modes_destroy(base);
-         base = NULL;
-         break;
-      }
-
-      (void) memset(*next, 0, size);
-      (*next)->visualID = GLX_DONT_CARE;
-      (*next)->visualType = GLX_DONT_CARE;
-      (*next)->visualRating = GLX_NONE;
-      (*next)->transparentPixel = GLX_NONE;
-      (*next)->transparentRed = GLX_DONT_CARE;
-      (*next)->transparentGreen = GLX_DONT_CARE;
-      (*next)->transparentBlue = GLX_DONT_CARE;
-      (*next)->transparentAlpha = GLX_DONT_CARE;
-      (*next)->transparentIndex = GLX_DONT_CARE;
-      (*next)->xRenderable = GLX_DONT_CARE;
-      (*next)->fbconfigID = GLX_DONT_CARE;
-      (*next)->swapMethod = GLX_SWAP_UNDEFINED_OML;
-      (*next)->bindToTextureRgb = GLX_DONT_CARE;
-      (*next)->bindToTextureRgba = GLX_DONT_CARE;
-      (*next)->bindToMipmapTexture = GLX_DONT_CARE;
-      (*next)->bindToTextureTargets = GLX_DONT_CARE;
-      (*next)->yInverted = GLX_DONT_CARE;
-
-      next = &((*next)->next);
-   }
-
-   return base;
-}
-
-
-/**
- * Destroy a linked list of \c __GLcontextModes structures created by
- * \c _gl_context_modes_create.
- * 
- * \param modes  Linked list of structures to be destroyed.  All structres
- *               in the list will be freed.
- */
-void
-_gl_context_modes_destroy(__GLcontextModes * modes)
-{
-   while (modes != NULL) {
-      __GLcontextModes *const next = modes->next;
-
-      free(modes);
-      modes = next;
-   }
-}
-
-
-/**
- * Find a context mode matching a Visual ID.
- *
- * \param modes  List list of context-mode structures to be searched.
- * \param vid    Visual ID to be found.
- * \returns A pointer to a context-mode in \c modes if \c vid was found in
- *          the list, or \c NULL if it was not.
- */
-
-__GLcontextModes *
-_gl_context_modes_find_visual(__GLcontextModes * modes, int vid)
-{
-   __GLcontextModes *m;
-
-   for (m = modes; m != NULL; m = m->next)
-      if (m->visualID == vid)
-         return m;
-
-   return NULL;
-}
-
-__GLcontextModes *
-_gl_context_modes_find_fbconfig(__GLcontextModes * modes, int fbid)
-{
-   __GLcontextModes *m;
-
-   for (m = modes; m != NULL; m = m->next)
-      if (m->fbconfigID == fbid)
-         return m;
-
-   return NULL;
-}
-
-/**
- * Determine if two context-modes are the same.  This is intended to be used
- * by libGL implementations to compare to sets of driver generated FBconfigs.
- * 
- * \param a  Context-mode to be compared.
- * \param b  Context-mode to be compared.
- * \returns \c GL_TRUE if the two context-modes are the same.  \c GL_FALSE is
- *          returned otherwise.
- */
-GLboolean
-_gl_context_modes_are_same(const __GLcontextModes * a,
-                           const __GLcontextModes * b)
-{
-   return ((a->rgbMode == b->rgbMode) &&
-           (a->floatMode == b->floatMode) &&
-           (a->colorIndexMode == b->colorIndexMode) &&
-           (a->doubleBufferMode == b->doubleBufferMode) &&
-           (a->stereoMode == b->stereoMode) &&
-           (a->redBits == b->redBits) &&
-           (a->greenBits == b->greenBits) &&
-           (a->blueBits == b->blueBits) && (a->alphaBits == b->alphaBits) &&
-#if 0                           /* For some reason these don't get set on the client-side in libGL. */
-           (a->redMask == b->redMask) &&
-           (a->greenMask == b->greenMask) &&
-           (a->blueMask == b->blueMask) && (a->alphaMask == b->alphaMask) &&
-#endif
-           (a->rgbBits == b->rgbBits) &&
-           (a->indexBits == b->indexBits) &&
-           (a->accumRedBits == b->accumRedBits) &&
-           (a->accumGreenBits == b->accumGreenBits) &&
-           (a->accumBlueBits == b->accumBlueBits) &&
-           (a->accumAlphaBits == b->accumAlphaBits) &&
-           (a->depthBits == b->depthBits) &&
-           (a->stencilBits == b->stencilBits) &&
-           (a->numAuxBuffers == b->numAuxBuffers) &&
-           (a->level == b->level) &&
-           (a->pixmapMode == b->pixmapMode) &&
-           (a->visualRating == b->visualRating) &&
-           (a->transparentPixel == b->transparentPixel) &&
-           ((a->transparentPixel != GLX_TRANSPARENT_RGB) ||
-            ((a->transparentRed == b->transparentRed) &&
-             (a->transparentGreen == b->transparentGreen) &&
-             (a->transparentBlue == b->transparentBlue) &&
-             (a->transparentAlpha == b->transparentAlpha))) &&
-           ((a->transparentPixel != GLX_TRANSPARENT_INDEX) ||
-            (a->transparentIndex == b->transparentIndex)) &&
-           (a->sampleBuffers == b->sampleBuffers) &&
-           (a->samples == b->samples) &&
-           ((a->drawableType & b->drawableType) != 0) &&
-           (a->renderType == b->renderType) &&
-           (a->maxPbufferWidth == b->maxPbufferWidth) &&
-           (a->maxPbufferHeight == b->maxPbufferHeight) &&
-           (a->maxPbufferPixels == b->maxPbufferPixels) &&
-           (a->optimalPbufferWidth == b->optimalPbufferWidth) &&
-           (a->optimalPbufferHeight == b->optimalPbufferHeight) &&
-           (a->swapMethod == b->swapMethod) &&
-           (a->bindToTextureRgb == b->bindToTextureRgb) &&
-           (a->bindToTextureRgba == b->bindToTextureRgba) &&
-           (a->bindToMipmapTexture == b->bindToMipmapTexture) &&
-           (a->bindToTextureTargets == b->bindToTextureTargets) &&
-           (a->yInverted == b->yInverted));
-}
diff --git a/src/glx/glcontextmodes.h b/src/glx/glcontextmodes.h
deleted file mode 100644
index ecf0280cba3..00000000000
--- a/src/glx/glcontextmodes.h
+++ /dev/null
@@ -1,52 +0,0 @@
-/*
- * (C) Copyright IBM Corporation 2003
- * All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * on the rights to use, copy, modify, merge, publish, distribute, sub
- * license, and/or sell copies of the Software, and to permit persons to whom
- * the Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
- * VA LINUX SYSTEM, IBM AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
- * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
- * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
- * USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-
-/**
- * \file glcontextmodes.h
- * \author Ian Romanick <[email protected]>
- */
-
-#ifndef GLCONTEXTMODES_H
-#define GLCONTEXTMODES_H
-
-#include "GL/internal/glcore.h"
-
-extern GLint _gl_convert_from_x_visual_type(int visualType);
-extern GLint _gl_convert_to_x_visual_type(int visualType);
-extern void _gl_copy_visual_to_context_mode(__GLcontextModes * mode,
-                                            const __GLXvisualConfig * config);
-extern int _gl_get_context_mode_data(const __GLcontextModes * mode,
-                                     int attribute, int *value_return);
-
-extern __GLcontextModes *_gl_context_modes_create(unsigned count,
-                                                  size_t minimum_size);
-extern void _gl_context_modes_destroy(__GLcontextModes * modes);
-extern __GLcontextModes *_gl_context_modes_find_visual(__GLcontextModes *
-                                                       modes, int vid);
-extern __GLcontextModes *_gl_context_modes_find_fbconfig(__GLcontextModes *
-                                                         modes, int fbid);
-extern GLboolean _gl_context_modes_are_same(const __GLcontextModes * a,
-                                            const __GLcontextModes * b);
-
-#endif /* GLCONTEXTMODES_H */
diff --git a/src/glx/glx_pbuffer.c b/src/glx/glx_pbuffer.c
index 02809aacfa5..15bfb159191 100644
--- a/src/glx/glx_pbuffer.c
+++ b/src/glx/glx_pbuffer.c
@@ -57,7 +57,7 @@
 static void
 warn_GLX_1_3(Display * dpy, const char *function_name)
 {
-   __GLXdisplayPrivate *priv = __glXInitialize(dpy);
+   struct glx_display *priv = __glXInitialize(dpy);
 
    if (priv->minorVersion < 3) {
       fprintf(stderr,
@@ -85,7 +85,7 @@ static void
 ChangeDrawableAttribute(Display * dpy, GLXDrawable drawable,
                         const CARD32 * attribs, size_t num_attribs)
 {
-   __GLXdisplayPrivate *priv = __glXInitialize(dpy);
+   struct glx_display *priv = __glXInitialize(dpy);
    __GLXDRIdrawable *pdraw;
    CARD32 *output;
    CARD8 opcode;
@@ -183,20 +183,20 @@ determineTextureFormat(const int *attribs, int numAttribs)
 }
 
 static void
-CreateDRIDrawable(Display *dpy, const __GLcontextModes *fbconfig,
+CreateDRIDrawable(Display *dpy, struct glx_config *config,
 		  XID drawable, XID glxdrawable,
 		  const int *attrib_list, size_t num_attribs)
 {
-   __GLXdisplayPrivate *const priv = __glXInitialize(dpy);
+   struct glx_display *const priv = __glXInitialize(dpy);
    __GLXDRIdrawable *pdraw;
-   __GLXscreenConfigs *psc;
+   struct glx_screen *psc;
 
-   psc = priv->screenConfigs[fbconfig->screen];
+   psc = priv->screens[config->screen];
    if (psc->driScreen == NULL)
       return;
 
    pdraw = psc->driScreen->createDrawable(psc, drawable,
-					  glxdrawable, fbconfig);
+					  glxdrawable, config);
    if (pdraw == NULL) {
       fprintf(stderr, "failed to create drawable\n");
       return;
@@ -214,7 +214,7 @@ CreateDRIDrawable(Display *dpy, const __GLcontextModes *fbconfig,
 static void
 DestroyDRIDrawable(Display *dpy, GLXDrawable drawable, int destroy_xdrawable)
 {
-   __GLXdisplayPrivate *const priv = __glXInitialize(dpy);
+   struct glx_display *const priv = __glXInitialize(dpy);
    __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable);
 
    if (pdraw != NULL) {
@@ -228,7 +228,7 @@ DestroyDRIDrawable(Display *dpy, GLXDrawable drawable, int destroy_xdrawable)
 #else
 
 static void
-CreateDRIDrawable(Display *dpy, const __GLcontextModes * fbconfig,
+CreateDRIDrawable(Display *dpy, const struct glx_config * fbconfig,
 		  XID drawable, XID glxdrawable,
 		  const int *attrib_list, size_t num_attribs)
 {
@@ -263,7 +263,7 @@ static int
 GetDrawableAttribute(Display * dpy, GLXDrawable drawable,
                      int attribute, unsigned int *value)
 {
-   __GLXdisplayPrivate *priv;
+   struct glx_display *priv;
    xGLXGetDrawableAttributesReply reply;
    CARD32 *data;
    CARD8 opcode;
@@ -367,7 +367,7 @@ GetDrawableAttribute(Display * dpy, GLXDrawable drawable,
  * This function needs to be modified to work with direct-rendering drivers.
  */
 static GLXDrawable
-CreateDrawable(Display * dpy, const __GLcontextModes * fbconfig,
+CreateDrawable(Display *dpy, struct glx_config *config,
                Drawable drawable, const int *attrib_list, CARD8 glxCode)
 {
    xGLXCreateWindowReq *req;
@@ -391,11 +391,11 @@ CreateDrawable(Display * dpy, const __GLcontextModes * fbconfig,
 
    req->reqType = opcode;
    req->glxCode = glxCode;
-   req->screen = (CARD32) fbconfig->screen;
-   req->fbconfig = fbconfig->fbconfigID;
-   req->window = (CARD32) drawable;
-   req->glxwindow = (GLXWindow) XAllocID(dpy);
-   req->numAttribs = (CARD32) i;
+   req->screen = config->screen;
+   req->fbconfig = config->fbconfigID;
+   req->window = drawable;
+   req->glxwindow = XAllocID(dpy);
+   req->numAttribs = i;
 
    if (attrib_list)
       memcpy(data, attrib_list, 8 * i);
@@ -403,9 +403,9 @@ CreateDrawable(Display * dpy, const __GLcontextModes * fbconfig,
    UnlockDisplay(dpy);
    SyncHandle();
 
-   CreateDRIDrawable(dpy, fbconfig, drawable, req->glxwindow, attrib_list, i);
+   CreateDRIDrawable(dpy, config, drawable, req->glxwindow, attrib_list, i);
 
-   return (GLXDrawable) req->glxwindow;
+   return req->glxwindow;
 }
 
 
@@ -457,11 +457,11 @@ DestroyDrawable(Display * dpy, GLXDrawable drawable, CARD32 glxCode)
  * This function needs to be modified to work with direct-rendering drivers.
  */
 static GLXDrawable
-CreatePbuffer(Display * dpy, const __GLcontextModes * fbconfig,
+CreatePbuffer(Display * dpy, struct glx_config *config,
               unsigned int width, unsigned int height,
               const int *attrib_list, GLboolean size_in_attribs)
 {
-   __GLXdisplayPrivate *priv = __glXInitialize(dpy);
+   struct glx_display *priv = __glXInitialize(dpy);
    GLXDrawable id = 0;
    CARD32 *data;
    CARD8 opcode;
@@ -490,10 +490,10 @@ CreatePbuffer(Display * dpy, const __GLcontextModes * fbconfig,
 
       req->reqType = opcode;
       req->glxCode = X_GLXCreatePbuffer;
-      req->screen = (CARD32) fbconfig->screen;
-      req->fbconfig = fbconfig->fbconfigID;
-      req->pbuffer = (GLXPbuffer) id;
-      req->numAttribs = (CARD32) (i + extra);
+      req->screen = config->screen;
+      req->fbconfig = config->fbconfigID;
+      req->pbuffer = id;
+      req->numAttribs = i + extra;
 
       if (!size_in_attribs) {
          data[(2 * i) + 0] = GLX_PBUFFER_WIDTH;
@@ -513,11 +513,11 @@ CreatePbuffer(Display * dpy, const __GLcontextModes * fbconfig,
       vpreq->glxCode = X_GLXVendorPrivate;
       vpreq->vendorCode = X_GLXvop_CreateGLXPbufferSGIX;
 
-      data[0] = (CARD32) fbconfig->screen;
-      data[1] = (CARD32) fbconfig->fbconfigID;
-      data[2] = (CARD32) id;
-      data[3] = (CARD32) width;
-      data[4] = (CARD32) height;
+      data[0] = config->screen;
+      data[1] = config->fbconfigID;
+      data[2] = id;
+      data[3] = width;
+      data[4] = height;
       data += 5;
    }
 
@@ -526,10 +526,10 @@ CreatePbuffer(Display * dpy, const __GLcontextModes * fbconfig,
    UnlockDisplay(dpy);
    SyncHandle();
 
-   pixmap = XCreatePixmap(dpy, RootWindow(dpy, fbconfig->screen),
-			  width, height, fbconfig->rgbBits);
+   pixmap = XCreatePixmap(dpy, RootWindow(dpy, config->screen),
+			  width, height, config->rgbBits);
 
-   CreateDRIDrawable(dpy, fbconfig, pixmap, id, attrib_list, i);
+   CreateDRIDrawable(dpy, config, pixmap, id, attrib_list, i);
 
    return id;
 }
@@ -547,7 +547,7 @@ CreatePbuffer(Display * dpy, const __GLcontextModes * fbconfig,
 static void
 DestroyPbuffer(Display * dpy, GLXDrawable drawable)
 {
-   __GLXdisplayPrivate *priv = __glXInitialize(dpy);
+   struct glx_display *priv = __glXInitialize(dpy);
    CARD8 opcode;
 
    if ((dpy == NULL) || (drawable == 0)) {
@@ -593,12 +593,12 @@ DestroyPbuffer(Display * dpy, GLXDrawable drawable)
 /**
  * Create a new pbuffer.
  */
-PUBLIC GLXPbufferSGIX
+_X_EXPORT GLXPbufferSGIX
 glXCreateGLXPbufferSGIX(Display * dpy, GLXFBConfigSGIX config,
                         unsigned int width, unsigned int height,
                         int *attrib_list)
 {
-   return (GLXPbufferSGIX) CreatePbuffer(dpy, (__GLcontextModes *) config,
+   return (GLXPbufferSGIX) CreatePbuffer(dpy, (struct glx_config *) config,
                                          width, height,
                                          attrib_list, GL_FALSE);
 }
@@ -608,7 +608,7 @@ glXCreateGLXPbufferSGIX(Display * dpy, GLXFBConfigSGIX config,
 /**
  * Create a new pbuffer.
  */
-PUBLIC GLXPbuffer
+_X_EXPORT GLXPbuffer
 glXCreatePbuffer(Display * dpy, GLXFBConfig config, const int *attrib_list)
 {
    int i, width, height;
@@ -674,7 +674,7 @@ glXCreatePbuffer(Display * dpy, GLXFBConfig config, const int *attrib_list)
       }
    }
 
-   return (GLXPbuffer) CreatePbuffer(dpy, (__GLcontextModes *) config,
+   return (GLXPbuffer) CreatePbuffer(dpy, (struct glx_config *) config,
                                      width, height, attrib_list, GL_TRUE);
 #endif
 }
@@ -683,7 +683,7 @@ glXCreatePbuffer(Display * dpy, GLXFBConfig config, const int *attrib_list)
 /**
  * Destroy an existing pbuffer.
  */
-PUBLIC void
+_X_EXPORT void
 glXDestroyPbuffer(Display * dpy, GLXPbuffer pbuf)
 {
 #ifdef GLX_USE_APPLEGL
@@ -699,7 +699,7 @@ glXDestroyPbuffer(Display * dpy, GLXPbuffer pbuf)
 /**
  * Query an attribute of a drawable.
  */
-PUBLIC void
+_X_EXPORT void
 glXQueryDrawable(Display * dpy, GLXDrawable drawable,
                  int attribute, unsigned int *value)
 {
@@ -748,7 +748,7 @@ glXQueryDrawable(Display * dpy, GLXDrawable drawable,
 /**
  * Query an attribute of a pbuffer.
  */
-PUBLIC int
+_X_EXPORT int
 glXQueryGLXPbufferSGIX(Display * dpy, GLXPbufferSGIX drawable,
                        int attribute, unsigned int *value)
 {
@@ -759,7 +759,7 @@ glXQueryGLXPbufferSGIX(Display * dpy, GLXPbufferSGIX drawable,
 /**
  * Select the event mask for a drawable.
  */
-PUBLIC void
+_X_EXPORT void
 glXSelectEvent(Display * dpy, GLXDrawable drawable, unsigned long mask)
 {
 #ifdef GLX_USE_APPLEGL
@@ -792,7 +792,7 @@ glXSelectEvent(Display * dpy, GLXDrawable drawable, unsigned long mask)
 /**
  * Get the selected event mask for a drawable.
  */
-PUBLIC void
+_X_EXPORT void
 glXGetSelectedEvent(Display * dpy, GLXDrawable drawable, unsigned long *mask)
 {
 #ifdef GLX_USE_APPLEGL
@@ -829,27 +829,27 @@ glXGetSelectedEvent(Display * dpy, GLXDrawable drawable, unsigned long *mask)
 }
 
 
-PUBLIC GLXPixmap
+_X_EXPORT GLXPixmap
 glXCreatePixmap(Display * dpy, GLXFBConfig config, Pixmap pixmap,
                 const int *attrib_list)
 {
    WARN_ONCE_GLX_1_3(dpy, __func__);
 
 #ifdef GLX_USE_APPLEGL
-   const __GLcontextModes *modes = (const __GLcontextModes *) config;
+   const struct glx_config *modes = (const __GLcontextModes *) config;
 
    if (apple_glx_pixmap_create(dpy, modes->screen, pixmap, modes))
       return None;
 
    return pixmap;
 #else
-   return CreateDrawable(dpy, (__GLcontextModes *) config,
+   return CreateDrawable(dpy, (struct glx_config *) config,
                          (Drawable) pixmap, attrib_list, X_GLXCreatePixmap);
 #endif
 }
 
 
-PUBLIC GLXWindow
+_X_EXPORT GLXWindow
 glXCreateWindow(Display * dpy, GLXFBConfig config, Window win,
                 const int *attrib_list)
 {
@@ -878,13 +878,13 @@ glXCreateWindow(Display * dpy, GLXFBConfig config, Window win,
 
    return win;
 #else
-   return CreateDrawable(dpy, (__GLcontextModes *) config,
+   return CreateDrawable(dpy, (struct glx_config *) config,
                          (Drawable) win, attrib_list, X_GLXCreateWindow);
 #endif
 }
 
 
-PUBLIC void
+_X_EXPORT void
 glXDestroyPixmap(Display * dpy, GLXPixmap pixmap)
 {
    WARN_ONCE_GLX_1_3(dpy, __func__);
@@ -897,7 +897,7 @@ glXDestroyPixmap(Display * dpy, GLXPixmap pixmap)
 }
 
 
-PUBLIC void
+_X_EXPORT void
 glXDestroyWindow(Display * dpy, GLXWindow win)
 {
    WARN_ONCE_GLX_1_3(dpy, __func__);
@@ -907,17 +907,17 @@ glXDestroyWindow(Display * dpy, GLXWindow win)
 }
 
 #ifndef GLX_USE_APPLEGL
-PUBLIC
+_X_EXPORT
 GLX_ALIAS_VOID(glXDestroyGLXPbufferSGIX,
                (Display * dpy, GLXPbufferSGIX pbuf),
                (dpy, pbuf), glXDestroyPbuffer)
 
-PUBLIC
+_X_EXPORT
 GLX_ALIAS_VOID(glXSelectEventSGIX,
                (Display * dpy, GLXDrawable drawable,
                 unsigned long mask), (dpy, drawable, mask), glXSelectEvent)
 
-PUBLIC
+_X_EXPORT
 GLX_ALIAS_VOID(glXGetSelectedEventSGIX,
                (Display * dpy, GLXDrawable drawable,
                 unsigned long *mask), (dpy, drawable, mask),
diff --git a/src/glx/glxclient.h b/src/glx/glxclient.h
index 8112c01e7cc..81c9a266696 100644
--- a/src/glx/glxclient.h
+++ b/src/glx/glxclient.h
@@ -39,6 +39,7 @@
 #define _GLX_client_h_
 #include <X11/Xproto.h>
 #include <X11/Xlibint.h>
+#include <X11/Xfuncproto.h>
 #include <X11/extensions/extutil.h>
 #define GLX_GLXEXT_PROTOTYPES
 #include <GL/glx.h>
@@ -49,10 +50,9 @@
 #ifdef WIN32
 #include <stdint.h>
 #endif
-#include "GL/glxint.h"
 #include "GL/glxproto.h"
-#include "GL/internal/glcore.h"
 #include "glapi/glapitable.h"
+#include "glxconfig.h"
 #include "glxhash.h"
 #if defined( PTHREADS )
 # include <pthread.h>
@@ -60,42 +60,21 @@
 
 #include "glxextensions.h"
 
-
-/* If we build the library with gcc's -fvisibility=hidden flag, we'll
- * use the PUBLIC macro to mark functions that are to be exported.
- *
- * We also need to define a USED attribute, so the optimizer doesn't
- * inline a static function that we later use in an alias. - ajax
- */
-#if defined(__GNUC__)
-#  define PUBLIC __attribute__((visibility("default")))
-#  define USED __attribute__((used))
-#else
-#  define PUBLIC
-#  define USED
-#endif
-
-
+#define ARRAY_SIZE(a) (sizeof (a) / sizeof ((a)[0]))
 
 #define GLX_MAJOR_VERSION 1       /* current version numbers */
 #define GLX_MINOR_VERSION 4
 
 #define __GLX_MAX_TEXTURE_UNITS 32
 
-typedef struct __GLXscreenConfigsRec __GLXscreenConfigs;
-typedef struct __GLXcontextRec __GLXcontext;
-typedef struct __GLXdrawableRec __GLXdrawable;
-typedef struct __GLXdisplayPrivateRec __GLXdisplayPrivate;
-typedef struct _glapi_table __GLapi;
+struct glx_display;
+struct glx_context;
 
 /************************************************************************/
 
 #ifdef GLX_DIRECT_RENDERING
 
-#define containerOf(ptr, type, member)              \
-    (type *)( (char *)ptr - offsetof(type,member) )
-
-extern void DRI_glXUseXFont(GLXContext CC,
+extern void DRI_glXUseXFont(struct glx_context *ctx,
 			    Font font, int first, int count, int listbase);
 
 #endif
@@ -109,7 +88,6 @@ extern void DRI_glXUseXFont(GLXContext CC,
 typedef struct __GLXDRIdisplayRec __GLXDRIdisplay;
 typedef struct __GLXDRIscreenRec __GLXDRIscreen;
 typedef struct __GLXDRIdrawableRec __GLXDRIdrawable;
-typedef struct __GLXDRIcontextRec __GLXDRIcontext;
 
 #include "glxextensions.h"
 
@@ -120,27 +98,28 @@ struct __GLXDRIdisplayRec
      */
    void (*destroyDisplay) (__GLXDRIdisplay * display);
 
-   __GLXscreenConfigs *(*createScreen)(int screen, __GLXdisplayPrivate * priv);
+   struct glx_screen *(*createScreen)(int screen, struct glx_display * priv);
 };
 
 struct __GLXDRIscreenRec {
 
-   void (*destroyScreen)(__GLXscreenConfigs *psc);
+   void (*destroyScreen)(struct glx_screen *psc);
 
-   __GLXcontext *(*createContext)(__GLXscreenConfigs *psc,
-				  const __GLcontextModes *mode,
-				  GLXContext shareList, int renderType);
+   struct glx_context *(*createContext)(struct glx_screen *psc,
+					struct glx_config *config,
+					struct glx_context *shareList,
+					int renderType);
 
-   __GLXDRIdrawable *(*createDrawable)(__GLXscreenConfigs *psc,
+   __GLXDRIdrawable *(*createDrawable)(struct glx_screen *psc,
 				       XID drawable,
 				       GLXDrawable glxDrawable,
-				       const __GLcontextModes *modes);
+				       struct glx_config *config);
 
    int64_t (*swapBuffers)(__GLXDRIdrawable *pdraw, int64_t target_msc,
 			  int64_t divisor, int64_t remainder);
    void (*copySubBuffer)(__GLXDRIdrawable *pdraw,
 			 int x, int y, int width, int height);
-   int (*getDrawableMSC)(__GLXscreenConfigs *psc, __GLXDRIdrawable *pdraw,
+   int (*getDrawableMSC)(struct glx_screen *psc, __GLXDRIdrawable *pdraw,
 			 int64_t *ust, int64_t *msc, int64_t *sbc);
    int (*waitForMSC)(__GLXDRIdrawable *pdraw, int64_t target_msc,
 		     int64_t divisor, int64_t remainder, int64_t *ust,
@@ -151,21 +130,13 @@ struct __GLXDRIscreenRec {
    int (*getSwapInterval)(__GLXDRIdrawable *pdraw);
 };
 
-struct __GLXDRIcontextRec
-{
-   void (*destroyContext) (__GLXcontext *context);
-   Bool(*bindContext) (__GLXcontext *context, __GLXDRIdrawable *pdraw,
-		       __GLXDRIdrawable *pread);
-   void (*unbindContext) (__GLXcontext *context);
-};
-
 struct __GLXDRIdrawableRec
 {
    void (*destroyDrawable) (__GLXDRIdrawable * drawable);
 
    XID xDrawable;
    XID drawable;
-   __GLXscreenConfigs *psc;
+   struct glx_screen *psc;
    GLenum textureTarget;
    GLenum textureFormat;        /* EXT_texture_from_pixmap support */
    unsigned long eventMask;
@@ -241,9 +212,13 @@ typedef struct __GLXattributeMachineRec
 } __GLXattributeMachine;
 
 struct glx_context_vtable {
-   void (*wait_gl)(__GLXcontext *ctx);
-   void (*wait_x)(__GLXcontext *ctx);
-   void (*use_x_font)(__GLXcontext *ctx,
+   void (*destroy)(struct glx_context *ctx);
+   int (*bind)(struct glx_context *context, struct glx_context *old,
+	       GLXDrawable draw, GLXDrawable read);
+   void (*unbind)(struct glx_context *context, struct glx_context *new);
+   void (*wait_gl)(struct glx_context *ctx);
+   void (*wait_x)(struct glx_context *ctx);
+   void (*use_x_font)(struct glx_context *ctx,
 		      Font font, int first, int count, int listBase);
    void (*bind_tex_image)(Display * dpy,
 			  GLXDrawable drawable,
@@ -252,11 +227,14 @@ struct glx_context_vtable {
    
 };
 
+extern void
+glx_send_destroy_context(Display *dpy, XID xid);
+
 /**
  * GLX state that needs to be kept on the client.  One of these records
  * exist for each context that has been made current by this client.
  */
-struct __GLXcontextRec
+struct glx_context
 {
     /**
      * \name Drawing command buffer.
@@ -299,7 +277,7 @@ struct __GLXcontextRec
      * Screen number.
      */
    GLint screen;
-   __GLXscreenConfigs *psc;
+   struct glx_screen *psc;
 
     /**
      * \c GL_TRUE if the context was created with ImportContext, which
@@ -342,7 +320,7 @@ struct __GLXcontextRec
      * Fill newImage with the unpacked form of \c oldImage getting it
      * ready for transport to the server.
      */
-   void (*fillImage) (__GLXcontext *, GLint, GLint, GLint, GLint, GLenum,
+   void (*fillImage) (struct glx_context *, GLint, GLint, GLint, GLint, GLenum,
                       GLenum, const GLvoid *, GLubyte *, GLubyte *);
 
     /**
@@ -401,18 +379,9 @@ struct __GLXcontextRec
    GLint majorOpcode;
 
     /**
-     * Pointer to the mode used to create this context.
+     * Pointer to the config used to create this context.
      */
-   const __GLcontextModes *mode;
-
-#ifdef GLX_DIRECT_RENDERING
-#ifdef GLX_USE_APPLEGL
-   void *driContext;
-   Bool do_destroy;
-#else
-   __GLXDRIcontext *driContext;
-#endif
-#endif
+   struct glx_config *config;
 
     /**
      * The current read-drawable for this context.  Will be None if this
@@ -459,15 +428,15 @@ struct __GLXcontextRec
 };
 
 extern Bool
-glx_context_init(__GLXcontext *gc,
-		 __GLXscreenConfigs *psc, const __GLcontextModes *fbconfig);
+glx_context_init(struct glx_context *gc,
+		 struct glx_screen *psc, struct glx_config *fbconfig);
 
 #define __glXSetError(gc,code)  \
    if (!(gc)->error) {          \
       (gc)->error = code;       \
    }
 
-extern void __glFreeAttributeState(__GLXcontext *);
+extern void __glFreeAttributeState(struct glx_context *);
 
 /************************************************************************/
 
@@ -501,8 +470,17 @@ extern void __glFreeAttributeState(__GLXcontext *);
  * One of these records exists per screen of the display.  It contains
  * a pointer to the config data for that screen (if the screen supports GL).
  */
-struct __GLXscreenConfigsRec
+struct glx_screen_vtable {
+   struct glx_context *(*create_context)(struct glx_screen *psc,
+					 struct glx_config *config,
+					 struct glx_context *shareList,
+					 int renderType);
+};
+
+struct glx_screen
 {
+   const struct glx_screen_vtable *vtable;
+
     /**
      * GLX extension string reported by the X-server.
      */
@@ -514,7 +492,7 @@ struct __GLXscreenConfigsRec
      */
    char *effectiveGLXexts;
 
-   __GLXdisplayPrivate *display;
+   struct glx_display *display;
 
 #if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
     /**
@@ -529,13 +507,13 @@ struct __GLXscreenConfigsRec
     /**
      * Linked list of glx visuals and  fbconfigs for this screen.
      */
-   __GLcontextModes *visuals, *configs;
+   struct glx_config *visuals, *configs;
 
     /**
      * Per-screen dynamic GLX extension tracking.  The \c direct_support
      * field only contains enough bits for 64 extensions.  Should libGL
      * ever need to track more than 64 GLX extensions, we can safely grow
-     * this field.  The \c __GLXscreenConfigs structure is not used outside
+     * this field.  The \c struct glx_screen structure is not used outside
      * libGL.
      */
    /*@{ */
@@ -549,11 +527,11 @@ struct __GLXscreenConfigsRec
  * Per display private data.  One of these records exists for each display
  * that is using the OpenGL (GLX) extension.
  */
-struct __GLXdisplayPrivateRec
+struct glx_display
 {
    /* The extension protocol codes */
    XExtCodes *codes;
-   struct __GLXdisplayPrivateRec *next;
+   struct glx_display *next;
 
     /**
      * Back pointer to the display
@@ -591,7 +569,7 @@ struct __GLXdisplayPrivateRec
      * Also, per screen data which now includes the server \c GLX_EXTENSION
      * string.
      */
-   __GLXscreenConfigs **screenConfigs;
+   struct glx_screen **screens;
 
 #if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
    __glxHashTable *drawHash;
@@ -606,25 +584,25 @@ struct __GLXdisplayPrivateRec
 };
 
 extern int
-glx_screen_init(__GLXscreenConfigs *psc,
-		int screen, __GLXdisplayPrivate * priv);
+glx_screen_init(struct glx_screen *psc,
+		int screen, struct glx_display * priv);
 
 #if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
 extern __GLXDRIdrawable *
 dri2GetGlxDrawableFromXDrawableId(Display *dpy, XID id);
 #endif
 
-extern GLubyte *__glXFlushRenderBuffer(__GLXcontext *, GLubyte *);
+extern GLubyte *__glXFlushRenderBuffer(struct glx_context *, GLubyte *);
 
-extern void __glXSendLargeChunk(__GLXcontext * gc, GLint requestNumber,
+extern void __glXSendLargeChunk(struct glx_context * gc, GLint requestNumber,
                                 GLint totalRequests,
                                 const GLvoid * data, GLint dataLen);
 
-extern void __glXSendLargeCommand(__GLXcontext *, const GLvoid *, GLint,
+extern void __glXSendLargeCommand(struct glx_context *, const GLvoid *, GLint,
                                   const GLvoid *, GLint);
 
 /* Initialize the GLX extension for dpy */
-extern __GLXdisplayPrivate *__glXInitialize(Display *);
+extern struct glx_display *__glXInitialize(Display *);
 
 extern void __glXPreferEGL(int state);
 
@@ -635,7 +613,7 @@ extern int __glXDebug;
 /* This is per-thread storage in an MT environment */
 #if defined( PTHREADS )
 
-extern void __glXSetCurrentContext(__GLXcontext * c);
+extern void __glXSetCurrentContext(struct glx_context * c);
 
 # if defined( GLX_USE_TLS )
 
@@ -646,13 +624,13 @@ extern __thread void *__glX_tls_Context
 
 # else
 
-extern __GLXcontext *__glXGetCurrentContext(void);
+extern struct glx_context *__glXGetCurrentContext(void);
 
 # endif /* defined( GLX_USE_TLS ) */
 
 #else
 
-extern __GLXcontext *__glXcurrentContext;
+extern struct glx_context *__glXcurrentContext;
 #define __glXGetCurrentContext() __glXcurrentContext
 #define __glXSetCurrentContext(gc) __glXcurrentContext = gc
 
@@ -660,8 +638,6 @@ extern __GLXcontext *__glXcurrentContext;
 
 extern void __glXSetCurrentContextNull(void);
 
-extern void __glXFreeContext(__GLXcontext *);
-
 
 /*
 ** Global lock for all threads in this address space using the GLX
@@ -690,7 +666,7 @@ extern CARD8 __glXSetupForCommand(Display * dpy);
 extern const GLuint __glXDefaultPixelStore[9];
 
 /* Send an image to the server using RenderLarge. */
-extern void __glXSendLargeImage(__GLXcontext * gc, GLint compsize, GLint dim,
+extern void __glXSendLargeImage(struct glx_context * gc, GLint compsize, GLint dim,
                                 GLint width, GLint height, GLint depth,
                                 GLenum format, GLenum type,
                                 const GLvoid * src, GLubyte * pc,
@@ -714,7 +690,7 @@ extern GLint __glBytesPerElement(GLenum type);
 ** updated to contain the modes needed by the server to decode the
 ** sent data.
 */
-extern void __glFillImage(__GLXcontext *, GLint, GLint, GLint, GLint, GLenum,
+extern void __glFillImage(struct glx_context *, GLint, GLint, GLint, GLint, GLenum,
                           GLenum, const GLvoid *, GLubyte *, GLubyte *);
 
 /* Copy map data with a stride into a packed buffer */
@@ -729,15 +705,15 @@ extern void __glFillMap2d(GLint, GLint, GLint, GLint, GLint,
 ** Empty an image out of the reply buffer into the clients memory applying
 ** the pack modes to pack back into the clients requested format.
 */
-extern void __glEmptyImage(__GLXcontext *, GLint, GLint, GLint, GLint, GLenum,
+extern void __glEmptyImage(struct glx_context *, GLint, GLint, GLint, GLint, GLenum,
                            GLenum, const GLubyte *, GLvoid *);
 
 
 /*
 ** Allocate and Initialize Vertex Array client state, and free.
 */
-extern void __glXInitVertexArrayState(__GLXcontext *);
-extern void __glXFreeVertexArrayState(__GLXcontext *);
+extern void __glXInitVertexArrayState(struct glx_context *);
+extern void __glXFreeVertexArrayState(struct glx_context *);
 
 /*
 ** Inform the Server of the major and minor numbers and of the client
@@ -758,7 +734,7 @@ extern void _XSend(Display *, const void *, long);
 #endif
 
 
-extern void __glXInitializeVisualConfigFromTags(__GLcontextModes * config,
+extern void __glXInitializeVisualConfigFromTags(struct glx_config * config,
                                                 int count, const INT32 * bp,
                                                 Bool tagged_only,
                                                 Bool fbconfig_style_tags);
@@ -790,9 +766,21 @@ __glxGetMscRate(__GLXDRIdrawable *glxDraw,
  * glx_info->codes->first_event */
 XExtDisplayInfo *__glXFindDisplay (Display *dpy);
 
+extern void
+GarbageCollectDRIDrawables(struct glx_screen *psc);
+
 extern __GLXDRIdrawable *
 GetGLXDRIDrawable(Display *dpy, GLXDrawable drawable);
 
 #endif
 
+extern struct glx_context dummyContext;
+
+extern struct glx_screen *
+indirect_create_screen(int screen, struct glx_display * priv);
+extern struct glx_context *
+indirect_create_context(struct glx_screen *psc,
+			struct glx_config *mode,
+			struct glx_context *shareList, int renderType);
+
 #endif /* !__GLX_client_h__ */
diff --git a/src/glx/glxcmds.c b/src/glx/glxcmds.c
index 8ee9a999e4a..d4aa8b58668 100644
--- a/src/glx/glxcmds.c
+++ b/src/glx/glxcmds.c
@@ -36,22 +36,18 @@
 #include "glxclient.h"
 #include "glapi.h"
 #include "glxextensions.h"
-#include "glcontextmodes.h"
 
 #ifdef GLX_DIRECT_RENDERING
 #ifdef GLX_USE_APPLEGL
 #include "apple_glx_context.h"
 #include "apple_glx.h"
 #include "glx_error.h"
-#define GC_IS_DIRECT(gc) ((gc)->isDirect)
 #else
 #include <sys/time.h>
 #include <X11/extensions/xf86vmode.h>
 #include "xf86dri.h"
-#define GC_IS_DIRECT(gc) ((gc)->driContext != NULL)
 #endif
 #else
-#define GC_IS_DIRECT(gc) (0)
 #endif
 
 #if defined(USE_XCB)
@@ -62,7 +58,6 @@
 
 static const char __glXGLXClientVendorName[] = "Mesa Project and SGI";
 static const char __glXGLXClientVersion[] = "1.4";
-static const struct glx_context_vtable indirect_context_vtable;
 
 #if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
 
@@ -85,12 +80,12 @@ windowExistsErrorHandler(Display * dpy, XErrorEvent * xerr)
  * \param dpy    Display to destroy drawables for
  * \param screen Screen number to destroy drawables for
  */
-static void
-GarbageCollectDRIDrawables(__GLXscreenConfigs * sc)
+_X_HIDDEN void
+GarbageCollectDRIDrawables(struct glx_screen * sc)
 {
    XID draw;
    __GLXDRIdrawable *pdraw;
-   __GLXdisplayPrivate *priv = sc->display;
+   struct glx_display *priv = sc->display;
    XWindowAttributes xwa;
    int (*oldXErrorHandler) (Display *, XErrorEvent *);
 
@@ -128,7 +123,7 @@ GarbageCollectDRIDrawables(__GLXscreenConfigs * sc)
 _X_HIDDEN __GLXDRIdrawable *
 GetGLXDRIDrawable(Display * dpy, GLXDrawable drawable)
 {
-   __GLXdisplayPrivate *priv = __glXInitialize(dpy);
+   struct glx_display *priv = __glXInitialize(dpy);
    __GLXDRIdrawable *pdraw;
 
    if (priv == NULL)
@@ -157,20 +152,20 @@ GetGLXDRIDrawable(Display * dpy, GLXDrawable drawable)
  *       number range for \c dpy?
  */
 
-static __GLXscreenConfigs *
+static struct glx_screen *
 GetGLXScreenConfigs(Display * dpy, int scrn)
 {
-   __GLXdisplayPrivate *const priv = __glXInitialize(dpy);
+   struct glx_display *const priv = __glXInitialize(dpy);
 
    return (priv
-           && priv->screenConfigs !=
-           NULL) ? priv->screenConfigs[scrn] : NULL;
+           && priv->screens !=
+           NULL) ? priv->screens[scrn] : NULL;
 }
 
 
 static int
-GetGLXPrivScreenConfig(Display * dpy, int scrn, __GLXdisplayPrivate ** ppriv,
-                       __GLXscreenConfigs ** ppsc)
+GetGLXPrivScreenConfig(Display * dpy, int scrn, struct glx_display ** ppriv,
+                       struct glx_screen ** ppsc)
 {
    /* Initialize the extension, if needed .  This has the added value
     * of initializing/allocating the display private
@@ -191,7 +186,7 @@ GetGLXPrivScreenConfig(Display * dpy, int scrn, __GLXdisplayPrivate ** ppriv,
    }
 
    /* Check to see if the GL is supported on this screen */
-   *ppsc = (*ppriv)->screenConfigs[scrn];
+   *ppsc = (*ppriv)->screens[scrn];
    if ((*ppsc)->configs == NULL) {
       /* No support for GL on this screen regardless of visual */
       return GLX_BAD_VISUAL;
@@ -208,148 +203,34 @@ GetGLXPrivScreenConfig(Display * dpy, int scrn, __GLXdisplayPrivate ** ppriv,
  * \param config  Application supplied \c GLXFBConfig.
  *
  * \returns If the \c GLXFBConfig is valid, the a pointer to the matching
- *          \c __GLcontextModes structure is returned.  Otherwise, \c NULL
+ *          \c struct glx_config structure is returned.  Otherwise, \c NULL
  *          is returned.
  */
-static __GLcontextModes *
-ValidateGLXFBConfig(Display * dpy, GLXFBConfig config)
+static struct glx_config *
+ValidateGLXFBConfig(Display * dpy, GLXFBConfig fbconfig)
 {
-   __GLXdisplayPrivate *const priv = __glXInitialize(dpy);
-   const unsigned num_screens = ScreenCount(dpy);
+   struct glx_display *const priv = __glXInitialize(dpy);
+   int num_screens = ScreenCount(dpy);
    unsigned i;
-   const __GLcontextModes *modes;
-
+   struct glx_config *config;
 
    if (priv != NULL) {
       for (i = 0; i < num_screens; i++) {
-         for (modes = priv->screenConfigs[i]->configs; modes != NULL;
-              modes = modes->next) {
-            if (modes == (__GLcontextModes *) config) {
-               return (__GLcontextModes *) config;
-            }
-         }
+	 for (config = priv->screens[i]->configs; config != NULL;
+	      config = config->next) {
+	    if (config == (struct glx_config *) fbconfig) {
+	       return config;
+	    }
+	 }
       }
    }
 
    return NULL;
 }
 
-
-/**
- * \todo It should be possible to move the allocate of \c client_state_private
- * later in the function for direct-rendering contexts.  Direct-rendering
- * contexts don't need to track client state, so they don't need that memory
- * at all.
- *
- * \todo Eliminate \c __glXInitVertexArrayState.  Replace it with a new
- * function called \c __glXAllocateClientState that allocates the memory and
- * does all the initialization (including the pixel pack / unpack).
- */
-static GLXContext
-AllocateGLXContext(Display * dpy)
-{
-   GLXContext gc;
-   int bufSize;
-   CARD8 opcode;
-   __GLXattribute *state;
-
-   if (!dpy)
-      return NULL;
-
-   opcode = __glXSetupForCommand(dpy);
-   if (!opcode) {
-      return NULL;
-   }
-
-   /* Allocate our context record */
-   gc = (GLXContext) Xmalloc(sizeof(struct __GLXcontextRec));
-   if (!gc) {
-      /* Out of memory */
-      return NULL;
-   }
-   memset(gc, 0, sizeof(struct __GLXcontextRec));
-
-   gc->vtable = &indirect_context_vtable;
-   state = Xmalloc(sizeof(struct __GLXattributeRec));
-   if (state == NULL) {
-      /* Out of memory */
-      Xfree(gc);
-      return NULL;
-   }
-   gc->client_state_private = state;
-   memset(gc->client_state_private, 0, sizeof(struct __GLXattributeRec));
-   state->NoDrawArraysProtocol = (getenv("LIBGL_NO_DRAWARRAYS") != NULL);
-
-   /*
-    ** Create a temporary buffer to hold GLX rendering commands.  The size
-    ** of the buffer is selected so that the maximum number of GLX rendering
-    ** commands can fit in a single X packet and still have room in the X
-    ** packet for the GLXRenderReq header.
-    */
-
-   bufSize = (XMaxRequestSize(dpy) * 4) - sz_xGLXRenderReq;
-   gc->buf = (GLubyte *) Xmalloc(bufSize);
-   if (!gc->buf) {
-      Xfree(gc->client_state_private);
-      Xfree(gc);
-      return NULL;
-   }
-   gc->bufSize = bufSize;
-
-   /* Fill in the new context */
-   gc->renderMode = GL_RENDER;
-
-   state->storePack.alignment = 4;
-   state->storeUnpack.alignment = 4;
-
-   gc->attributes.stackPointer = &gc->attributes.stack[0];
-
-   /*
-    ** PERFORMANCE NOTE: A mode dependent fill image can speed things up.
-    ** Other code uses the fastImageUnpack bit, but it is never set
-    ** to GL_TRUE.
-    */
-   gc->fastImageUnpack = GL_FALSE;
-   gc->fillImage = __glFillImage;
-   gc->pc = gc->buf;
-   gc->bufEnd = gc->buf + bufSize;
-   gc->isDirect = GL_FALSE;
-   if (__glXDebug) {
-      /*
-       ** Set limit register so that there will be one command per packet
-       */
-      gc->limit = gc->buf;
-   }
-   else {
-      gc->limit = gc->buf + bufSize - __GLX_BUFFER_LIMIT_SIZE;
-   }
-   gc->majorOpcode = opcode;
-
-   /*
-    ** Constrain the maximum drawing command size allowed to be
-    ** transfered using the X_GLXRender protocol request.  First
-    ** constrain by a software limit, then constrain by the protocl
-    ** limit.
-    */
-   if (bufSize > __GLX_RENDER_CMD_SIZE_LIMIT) {
-      bufSize = __GLX_RENDER_CMD_SIZE_LIMIT;
-   }
-   if (bufSize > __GLX_MAX_RENDER_CMD_SIZE) {
-      bufSize = __GLX_MAX_RENDER_CMD_SIZE;
-   }
-   gc->maxSmallRenderCommandSize = bufSize;
-   
-#ifdef GLX_USE_APPLEGL
-   gc->driContext = NULL;
-   gc->do_destroy = False;   
-#endif
-
-   return gc;
-}
-
 _X_HIDDEN Bool
-glx_context_init(__GLXcontext *gc,
-		 __GLXscreenConfigs *psc, const __GLcontextModes *fbconfig)
+glx_context_init(struct glx_context *gc,
+		 struct glx_screen *psc, struct glx_config *config)
 {
    gc->majorOpcode = __glXSetupForCommand(psc->display->dpy);
    if (!gc->majorOpcode)
@@ -357,8 +238,9 @@ glx_context_init(__GLXcontext *gc,
 
    gc->screen = psc->scr;
    gc->psc = psc;
-   gc->mode = fbconfig;
+   gc->config = config;
    gc->isDirect = GL_TRUE;
+   gc->currentContextTag = -1;
 
    return GL_TRUE;
 }
@@ -375,34 +257,25 @@ glx_context_init(__GLXcontext *gc,
 
 static GLXContext
 CreateContext(Display * dpy, int generic_id,
-              const __GLcontextModes * const fbconfig,
-              GLXContext shareList,
+              struct glx_config *config,
+              GLXContext shareList_user,
               Bool allowDirect,
 	      unsigned code, int renderType, int screen)
 {
-   GLXContext gc = NULL;
-   __GLXscreenConfigs *const psc = GetGLXScreenConfigs(dpy, screen);
-#if defined(GLX_DIRECT_RENDERING) && defined(GLX_USE_APPLEGL)
-   int errorcode;
-   bool x11error;
-#endif
-    
+   struct glx_context *gc = NULL;
+   struct glx_screen *const psc = GetGLXScreenConfigs(dpy, screen);
+   struct glx_context *shareList = (struct glx_context *) shareList_user;
    if (dpy == NULL)
       return NULL;
 
    if (generic_id == None)
       return NULL;
 
-#ifndef GLX_USE_APPLEGL  /* TODO: darwin indirect */
-#ifdef GLX_DIRECT_RENDERING
-   if (allowDirect && psc->driScreen) {
-      gc = psc->driScreen->createContext(psc, fbconfig,
-					 shareList, renderType);
-   }
-#endif
-
+   gc = NULL;
+   if (allowDirect && psc->vtable->create_context)
+      gc = psc->vtable->create_context(psc, config, shareList, renderType);
    if (!gc)
-      gc = AllocateGLXContext(dpy);
+      gc = indirect_create_context(psc, config, shareList, renderType);
    if (!gc)
       return NULL;
 
@@ -419,7 +292,7 @@ CreateContext(Display * dpy, int generic_id,
       req->visual = generic_id;
       req->screen = screen;
       req->shareList = shareList ? shareList->xid : None;
-      req->isDirect = GC_IS_DIRECT(gc);
+      req->isDirect = gc->isDirect;
       break;
    }
 
@@ -435,7 +308,7 @@ CreateContext(Display * dpy, int generic_id,
       req->screen = screen;
       req->renderType = renderType;
       req->shareList = shareList ? shareList->xid : None;
-      req->isDirect = GC_IS_DIRECT(gc);
+      req->isDirect = gc->isDirect;
       break;
    }
 
@@ -456,7 +329,7 @@ CreateContext(Display * dpy, int generic_id,
       req->screen = screen;
       req->renderType = renderType;
       req->shareList = shareList ? shareList->xid : None;
-      req->isDirect = GC_IS_DIRECT(gc);
+      req->isDirect = gc->isDirect;
       break;
    }
 
@@ -469,41 +342,25 @@ CreateContext(Display * dpy, int generic_id,
 
    UnlockDisplay(dpy);
    SyncHandle();
-#endif
 
    gc->imported = GL_FALSE;
    gc->renderType = renderType;
 
-   /* TODO: darwin: Integrate with above to do indirect */
-#ifdef GLX_USE_APPLEGL
-   if(apple_glx_create_context(&gc->driContext, dpy, screen, fbconfig, 
-                               shareList ? shareList->driContext : NULL,
-                               &errorcode, &x11error)) {
-      __glXSendError(dpy, errorcode, 0, X_GLXCreateContext, x11error);
-      __glXFreeContext(gc);
-      return NULL;
-   }
-   
-   gc->currentContextTag = -1;
-   gc->mode = fbconfig;
-   gc->isDirect = allowDirect;
-#endif
-
-   return gc;
+   return (GLXContext) gc;
 }
 
-PUBLIC GLXContext
+_X_EXPORT GLXContext
 glXCreateContext(Display * dpy, XVisualInfo * vis,
                  GLXContext shareList, Bool allowDirect)
 {
-   const __GLcontextModes *mode = NULL;
+   struct glx_config *config = NULL;
    int renderType = 0;
 
 #if defined(GLX_DIRECT_RENDERING) || defined(GLX_USE_APPLEGL)
-   __GLXscreenConfigs *const psc = GetGLXScreenConfigs(dpy, vis->screen);
+   struct glx_screen *const psc = GetGLXScreenConfigs(dpy, vis->screen);
 
-   mode = _gl_context_modes_find_visual(psc->visuals, vis->visualid);
-   if (mode == NULL) {
+   config = glx_config_find_visual(psc->visuals, vis->visualid);
+   if (config == NULL) {
       xError error;
 
       error.errorCode = BadValue;
@@ -516,115 +373,59 @@ glXCreateContext(Display * dpy, XVisualInfo * vis,
       return None;
    }
 
-   renderType = mode->rgbMode ? GLX_RGBA_TYPE : GLX_COLOR_INDEX_TYPE;
+   renderType = config->rgbMode ? GLX_RGBA_TYPE : GLX_COLOR_INDEX_TYPE;
 #endif
 
-   return CreateContext(dpy, vis->visualid, mode, shareList, allowDirect,
+   return CreateContext(dpy, vis->visualid, config, shareList, allowDirect,
                         X_GLXCreateContext, renderType, vis->screen);
 }
 
 _X_HIDDEN void
-__glXFreeContext(__GLXcontext * gc)
+glx_send_destroy_context(Display *dpy, XID xid)
 {
-   if (gc->vendor)
-      XFree((char *) gc->vendor);
-   if (gc->renderer)
-      XFree((char *) gc->renderer);
-   if (gc->version)
-      XFree((char *) gc->version);
-   if (gc->extensions)
-      XFree((char *) gc->extensions);
-   __glFreeAttributeState(gc);
-   XFree((char *) gc->buf);
-   Xfree((char *) gc->client_state_private);
-   XFree((char *) gc);
+   CARD8 opcode = __glXSetupForCommand(dpy);
+   xGLXDestroyContextReq *req;
 
+   LockDisplay(dpy);
+   GetReq(GLXDestroyContext, req);
+   req->reqType = opcode;
+   req->glxCode = X_GLXDestroyContext;
+   req->context = xid;
+   UnlockDisplay(dpy);
+   SyncHandle();
 }
 
 /*
 ** Destroy the named context
 */
 static void
-DestroyContext(Display * dpy, GLXContext gc)
+DestroyContext(Display * dpy, GLXContext ctx)
 {
-#ifndef GLX_USE_APPLEGL /* TODO: darwin: indirect */
-   xGLXDestroyContextReq *req;
-   GLXContextID xid;
-   CARD8 opcode;
-   GLboolean imported;
+   struct glx_context *gc = (struct glx_context *) ctx;
 
-   opcode = __glXSetupForCommand(dpy);
-   if (!opcode || !gc) {
+   if (!gc)
       return;
-   }
 
    __glXLock();
-   xid = gc->xid;
-   imported = gc->imported;
-   gc->xid = None;
-
    if (gc->currentDpy) {
       /* This context is bound to some thread.  According to the man page,
        * we should not actually delete the context until it's unbound.
        * Note that we set gc->xid = None above.  In MakeContextCurrent()
        * we check for that and delete the context there.
        */
+      if (!gc->imported)
+	 glx_send_destroy_context(dpy, gc->xid);
+      gc->xid = None;
       __glXUnlock();
       return;
    }
+   __glXUnlock();
 
-#if defined(GLX_DIRECT_RENDERING)
-   /* Destroy the direct rendering context */
-   if (gc->driContext) {
-      GarbageCollectDRIDrawables(gc->psc);
-      if (gc->extensions)
-	 XFree((char *) gc->extensions);
-      (*gc->driContext->destroyContext) (gc);
-   }
-   else
-#endif
-   {
-      __glXFreeVertexArrayState(gc);
-      __glXFreeContext(gc);
-   }
-
-   if (!imported) {
-      /*
-       ** This dpy also created the server side part of the context.
-       ** Send the glXDestroyContext request.
-       */
-      LockDisplay(dpy);
-      GetReq(GLXDestroyContext, req);
-      req->reqType = opcode;
-      req->glxCode = X_GLXDestroyContext;
-      req->context = xid;
-      UnlockDisplay(dpy);
-      SyncHandle();
-   }
-
-#else
-
-   __glXLock();
-   if (gc->currentDpy) {
-      /* 
-       * Set the Bool that indicates that we should destroy this GLX context
-       * when the context is no longer current.
-       */
-      gc->do_destroy = True;
-      /* Have to free later cuz it's in use now */
-      __glXUnlock();
-   }
-   else {
-      /* Destroy the handle if not current to anybody */
-      __glXUnlock();
-      if(gc->driContext)
-	 apple_glx_destroy_context(&gc->driContext, dpy);
-      __glXFreeContext(gc);
-   }
-#endif
+   if (gc->vtable->destroy)
+      gc->vtable->destroy(gc);
 }
 
-PUBLIC void
+_X_EXPORT void
 glXDestroyContext(Display * dpy, GLXContext gc)
 {
    DestroyContext(dpy, gc);
@@ -633,10 +434,10 @@ glXDestroyContext(Display * dpy, GLXContext gc)
 /*
 ** Return the major and minor version #s for the GLX extension
 */
-PUBLIC Bool
+_X_EXPORT Bool
 glXQueryVersion(Display * dpy, int *major, int *minor)
 {
-   __GLXdisplayPrivate *priv;
+   struct glx_display *priv;
 
    /* Init the extension.  This fetches the major and minor version. */
    priv = __glXInitialize(dpy);
@@ -653,7 +454,7 @@ glXQueryVersion(Display * dpy, int *major, int *minor)
 /*
 ** Query the existance of the GLX extension
 */
-PUBLIC Bool
+_X_EXPORT Bool
 glXQueryExtension(Display * dpy, int *errorBase, int *eventBase)
 {
    int major_op, erb, evb;
@@ -669,121 +470,36 @@ glXQueryExtension(Display * dpy, int *errorBase, int *eventBase)
    return rv;
 }
 
-static void
-indirect_wait_gl(__GLXcontext *gc)
-{
-   xGLXWaitGLReq *req;
-   Display *dpy = gc->currentDpy;
-
-   /* Flush any pending commands out */
-   __glXFlushRenderBuffer(gc, gc->pc);
-
-   /* Send the glXWaitGL request */
-   LockDisplay(dpy);
-   GetReq(GLXWaitGL, req);
-   req->reqType = gc->majorOpcode;
-   req->glxCode = X_GLXWaitGL;
-   req->contextTag = gc->currentContextTag;
-   UnlockDisplay(dpy);
-   SyncHandle();
-}
-
 /*
 ** Put a barrier in the token stream that forces the GL to finish its
 ** work before X can proceed.
 */
-PUBLIC void
+_X_EXPORT void
 glXWaitGL(void)
 {
-   GLXContext gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
 
-   if (gc && gc->vtable->use_x_font)
+   if (gc && gc->vtable->wait_gl)
       gc->vtable->wait_gl(gc);
 }
 
-static void
-indirect_wait_x(__GLXcontext *gc)
-{
-   xGLXWaitXReq *req;
-   Display *dpy = gc->currentDpy;
-
-   /* Flush any pending commands out */
-   __glXFlushRenderBuffer(gc, gc->pc);
-
-   LockDisplay(dpy);
-   GetReq(GLXWaitX, req);
-   req->reqType = gc->majorOpcode;
-   req->glxCode = X_GLXWaitX;
-   req->contextTag = gc->currentContextTag;
-   UnlockDisplay(dpy);
-   SyncHandle();
-}
-
 /*
 ** Put a barrier in the token stream that forces X to finish its
 ** work before GL can proceed.
 */
-PUBLIC void
+_X_EXPORT void
 glXWaitX(void)
 {
-   GLXContext gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
 
-   if (gc && gc->vtable->use_x_font)
+   if (gc && gc->vtable->wait_x)
       gc->vtable->wait_x(gc);
 }
 
-static void
-indirect_use_x_font(__GLXcontext *gc,
-		    Font font, int first, int count, int listBase)
-{
-   xGLXUseXFontReq *req;
-   Display *dpy = gc->currentDpy;
-
-   /* Flush any pending commands out */
-   __glXFlushRenderBuffer(gc, gc->pc);
-
-   /* Send the glXUseFont request */
-   LockDisplay(dpy);
-   GetReq(GLXUseXFont, req);
-   req->reqType = gc->majorOpcode;
-   req->glxCode = X_GLXUseXFont;
-   req->contextTag = gc->currentContextTag;
-   req->font = font;
-   req->first = first;
-   req->count = count;
-   req->listBase = listBase;
-   UnlockDisplay(dpy);
-   SyncHandle();
-}
-
-#ifdef GLX_USE_APPLEGL
-
-static void
-applegl_wait_gl(__GLXcontext *gc)
-{
-   glFinish();
-}
-
-static void
-applegl_wait_x(__GLXcontext *gc)
-{
-   apple_glx_waitx(gc->dpy, gc->driContext);
-}
-
-static const struct glx_context_vtable applegl_context_vtable = {
-   applegl_wait_gl,
-   applegl_wait_x,
-   DRI_glXUseXFont,
-   NULL, /* bind_tex_image, */
-   NULL, /* release_tex_image, */
-};
-
-#endif
-
-PUBLIC void
+_X_EXPORT void
 glXUseXFont(Font font, int first, int count, int listBase)
 {
-   GLXContext gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
 
    if (gc && gc->vtable->use_x_font)
       gc->vtable->use_x_font(gc, font, first, count, listBase);
@@ -795,12 +511,14 @@ glXUseXFont(Font font, int first, int count, int listBase)
 ** Copy the source context to the destination context using the
 ** attribute "mask".
 */
-PUBLIC void
-glXCopyContext(Display * dpy, GLXContext source,
-               GLXContext dest, unsigned long mask)
+_X_EXPORT void
+glXCopyContext(Display * dpy, GLXContext source_user,
+	       GLXContext dest_user, unsigned long mask)
 {
+   struct glx_context *source = (struct glx_context *) source_user;
+   struct glx_context *dest = (struct glx_context *) dest_user;
 #ifdef GLX_USE_APPLEGL
-   GLXContext gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    int errorcode;
    bool x11error;
 
@@ -811,7 +529,7 @@ glXCopyContext(Display * dpy, GLXContext source,
    
 #else
    xGLXCopyContextReq *req;
-   GLXContext gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    GLXContextTag tag;
    CARD8 opcode;
 
@@ -821,7 +539,7 @@ glXCopyContext(Display * dpy, GLXContext source,
    }
 
 #if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
-   if (gc->driContext) {
+   if (gc->isDirect) {
       /* NOT_DONE: This does not work yet */
    }
 #endif
@@ -906,13 +624,15 @@ __glXIsDirect(Display * dpy, GLXContextID contextID)
  * \c GLX_DIRECT_RENDERING is not defined?  Do we really need to bother with
  * the GLX protocol here at all?
  */
-PUBLIC Bool
-glXIsDirect(Display * dpy, GLXContext gc)
+_X_EXPORT Bool
+glXIsDirect(Display * dpy, GLXContext gc_user)
 {
+   struct glx_context *gc = (struct glx_context *) gc_user;
+
    if (!gc) {
       return GL_FALSE;
    }
-   else if (GC_IS_DIRECT(gc)) {
+   else if (gc->isDirect) {
       return GL_TRUE;
    }
 #ifdef GLX_USE_APPLEGL  /* TODO: indirect on darwin */
@@ -922,17 +642,17 @@ glXIsDirect(Display * dpy, GLXContext gc)
 #endif
 }
 
-PUBLIC GLXPixmap
+_X_EXPORT GLXPixmap
 glXCreateGLXPixmap(Display * dpy, XVisualInfo * vis, Pixmap pixmap)
 {
 #ifdef GLX_USE_APPLEGL
    int screen = vis->screen;
-   __GLXscreenConfigs *const psc = GetGLXScreenConfigs(dpy, screen);
-   const __GLcontextModes *modes;
+   struct glx_screen *const psc = GetGLXScreenConfigs(dpy, screen);
+   const struct glx_config *config;
 
-   modes = _gl_context_modes_find_visual(psc->visuals, vis->visualid);
+   config = _gl_context_modes_find_visual(psc->visuals, vis->visualid);
    
-   if(apple_glx_pixmap_create(dpy, vis->screen, pixmap, modes))
+   if(apple_glx_pixmap_create(dpy, vis->screen, pixmap, config))
       return None;
    
    return pixmap;
@@ -963,16 +683,16 @@ glXCreateGLXPixmap(Display * dpy, XVisualInfo * vis, Pixmap pixmap)
       /* FIXME: Maybe delay __DRIdrawable creation until the drawable
        * is actually bound to a context... */
 
-      __GLXdisplayPrivate *const priv = __glXInitialize(dpy);
+      struct glx_display *const priv = __glXInitialize(dpy);
       __GLXDRIdrawable *pdraw;
-      __GLXscreenConfigs *psc;
-      __GLcontextModes *modes;
+      struct glx_screen *psc;
+      struct glx_config *config;
 
-      psc = priv->screenConfigs[vis->screen];
+      psc = priv->screens[vis->screen];
       if (psc->driScreen == NULL)
          break;
-      modes = _gl_context_modes_find_visual(psc->visuals, vis->visualid);
-      pdraw = psc->driScreen->createDrawable(psc, pixmap, req->glxpixmap, modes);
+      config = glx_config_find_visual(psc->visuals, vis->visualid);
+      pdraw = psc->driScreen->createDrawable(psc, pixmap, req->glxpixmap, config);
       if (pdraw == NULL) {
          fprintf(stderr, "failed to create pixmap\n");
          break;
@@ -992,7 +712,7 @@ glXCreateGLXPixmap(Display * dpy, XVisualInfo * vis, Pixmap pixmap)
 /*
 ** Destroy the named pixmap
 */
-PUBLIC void
+_X_EXPORT void
 glXDestroyGLXPixmap(Display * dpy, GLXPixmap glxpixmap)
 {
 #ifdef GLX_USE_APPLEGL
@@ -1018,7 +738,7 @@ glXDestroyGLXPixmap(Display * dpy, GLXPixmap glxpixmap)
 
 #if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
    {
-      __GLXdisplayPrivate *const priv = __glXInitialize(dpy);
+      struct glx_display *const priv = __glXInitialize(dpy);
       __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, glxpixmap);
 
       if (pdraw != NULL) {
@@ -1030,7 +750,7 @@ glXDestroyGLXPixmap(Display * dpy, GLXPixmap glxpixmap)
 #endif /* GLX_USE_APPLEGL */
 }
 
-PUBLIC void
+_X_EXPORT void
 glXSwapBuffers(Display * dpy, GLXDrawable drawable)
 {
 #ifdef GLX_USE_APPLEGL
@@ -1041,7 +761,7 @@ glXSwapBuffers(Display * dpy, GLXDrawable drawable)
       __glXSendError(dpy, GLXBadCurrentWindow, 0, X_GLXSwapBuffers, false);
    }
 #else
-   GLXContext gc;
+   struct glx_context *gc;
    GLXContextTag tag;
    CARD8 opcode;
 #ifdef USE_XCB
@@ -1103,22 +823,22 @@ glXSwapBuffers(Display * dpy, GLXDrawable drawable)
 ** Return configuration information for the given display, screen and
 ** visual combination.
 */
-PUBLIC int
+_X_EXPORT int
 glXGetConfig(Display * dpy, XVisualInfo * vis, int attribute,
              int *value_return)
 {
-   __GLXdisplayPrivate *priv;
-   __GLXscreenConfigs *psc;
-   __GLcontextModes *modes;
+   struct glx_display *priv;
+   struct glx_screen *psc;
+   struct glx_config *config;
    int status;
 
    status = GetGLXPrivScreenConfig(dpy, vis->screen, &priv, &psc);
    if (status == Success) {
-      modes = _gl_context_modes_find_visual(psc->visuals, vis->visualid);
+      config = glx_config_find_visual(psc->visuals, vis->visualid);
 
       /* Lookup attribute after first finding a match on the visual */
-      if (modes != NULL) {
-         return _gl_get_context_mode_data(modes, attribute, value_return);
+      if (config != NULL) {
+	 return glx_config_get(config, attribute, value_return);
       }
 
       status = GLX_BAD_VISUAL;
@@ -1139,10 +859,10 @@ glXGetConfig(Display * dpy, XVisualInfo * vis, int attribute,
 /************************************************************************/
 
 static void
-init_fbconfig_for_chooser(__GLcontextModes * config,
+init_fbconfig_for_chooser(struct glx_config * config,
                           GLboolean fbconfig_style_tags)
 {
-   memset(config, 0, sizeof(__GLcontextModes));
+   memset(config, 0, sizeof(struct glx_config));
    config->visualID = (XID) GLX_DONT_CARE;
    config->visualType = GLX_DONT_CARE;
 
@@ -1208,8 +928,8 @@ init_fbconfig_for_chooser(__GLcontextModes * config,
  * \param b  Server specified config to test against \c a.
  */
 static Bool
-fbconfigs_compatible(const __GLcontextModes * const a,
-                     const __GLcontextModes * const b)
+fbconfigs_compatible(const struct glx_config * const a,
+                     const struct glx_config * const b)
 {
    MATCH_DONT_CARE(doubleBufferMode);
    MATCH_DONT_CARE(visualType);
@@ -1328,8 +1048,7 @@ fbconfigs_compatible(const __GLcontextModes * const a,
  * \sa qsort, glXChooseVisual, glXChooseFBConfig, glXChooseFBConfigSGIX
  */
 static int
-fbconfig_compare(const __GLcontextModes * const *const a,
-                 const __GLcontextModes * const *const b)
+fbconfig_compare(struct glx_config **a, struct glx_config **b)
 {
    /* The order of these comparisons must NOT change.  It is defined by
     * the GLX 1.3 spec and ARB_multisample.
@@ -1414,10 +1133,10 @@ fbconfig_compare(const __GLcontextModes * const *const a,
  * \sa glXChooseVisual, glXChooseFBConfig, glXChooseFBConfigSGIX
  */
 static int
-choose_visual(__GLcontextModes ** configs, int num_configs,
+choose_visual(struct glx_config ** configs, int num_configs,
               const int *attribList, GLboolean fbconfig_style_tags)
 {
-   __GLcontextModes test_config;
+   struct glx_config test_config;
    int base;
    int i;
 
@@ -1453,7 +1172,7 @@ choose_visual(__GLcontextModes ** configs, int num_configs,
     * specifications.
     */
 
-   qsort(configs, base, sizeof(__GLcontextModes *),
+   qsort(configs, base, sizeof(struct glx_config *),
          (int (*)(const void *, const void *)) fbconfig_compare);
    return base;
 }
@@ -1465,15 +1184,15 @@ choose_visual(__GLcontextModes ** configs, int num_configs,
 ** Return the visual that best matches the template.  Return None if no
 ** visual matches the template.
 */
-PUBLIC XVisualInfo *
+_X_EXPORT XVisualInfo *
 glXChooseVisual(Display * dpy, int screen, int *attribList)
 {
    XVisualInfo *visualList = NULL;
-   __GLXdisplayPrivate *priv;
-   __GLXscreenConfigs *psc;
-   __GLcontextModes test_config;
-   __GLcontextModes *modes;
-   const __GLcontextModes *best_config = NULL;
+   struct glx_display *priv;
+   struct glx_screen *psc;
+   struct glx_config test_config;
+   struct glx_config *config;
+   struct glx_config *best_config = NULL;
 
    /*
     ** Get a list of all visuals, return if list is empty
@@ -1500,26 +1219,23 @@ glXChooseVisual(Display * dpy, int screen, int *attribList)
     ** Otherwise, create an XVisualInfo list with just the selected X visual
     ** and return this.
     */
-   for (modes = psc->visuals; modes != NULL; modes = modes->next) {
-      if (fbconfigs_compatible(&test_config, modes)
-          && ((best_config == NULL)
-              ||
-              (fbconfig_compare
-               ((const __GLcontextModes * const *const) &modes,
-                &best_config) < 0))) {
+   for (config = psc->visuals; config != NULL; config = config->next) {
+      if (fbconfigs_compatible(&test_config, config)
+          && ((best_config == NULL) ||
+              (fbconfig_compare (&config, &best_config) < 0))) {
          XVisualInfo visualTemplate;
          XVisualInfo *newList;
          int i;
 
          visualTemplate.screen = screen;
-         visualTemplate.visualid = modes->visualID;
+         visualTemplate.visualid = config->visualID;
          newList = XGetVisualInfo(dpy, VisualScreenMask | VisualIDMask,
                                   &visualTemplate, &i);
 
          if (newList) {
             Xfree(visualList);
             visualList = newList;
-            best_config = modes;
+            best_config = config;
          }
       }
    }
@@ -1534,11 +1250,11 @@ glXChooseVisual(Display * dpy, int screen, int *attribList)
 }
 
 
-PUBLIC const char *
+_X_EXPORT const char *
 glXQueryExtensionsString(Display * dpy, int screen)
 {
-   __GLXscreenConfigs *psc;
-   __GLXdisplayPrivate *priv;
+   struct glx_screen *psc;
+   struct glx_display *priv;
 
    if (GetGLXPrivScreenConfig(dpy, screen, &priv, &psc) != Success) {
       return NULL;
@@ -1563,7 +1279,7 @@ glXQueryExtensionsString(Display * dpy, int screen)
    return psc->effectiveGLXexts;
 }
 
-PUBLIC const char *
+_X_EXPORT const char *
 glXGetClientString(Display * dpy, int name)
 {
    (void) dpy;
@@ -1580,11 +1296,11 @@ glXGetClientString(Display * dpy, int name)
    }
 }
 
-PUBLIC const char *
+_X_EXPORT const char *
 glXQueryServerString(Display * dpy, int screen, int name)
 {
-   __GLXscreenConfigs *psc;
-   __GLXdisplayPrivate *priv;
+   struct glx_screen *psc;
+   struct glx_display *priv;
    const char **str;
 
 
@@ -1650,175 +1366,141 @@ __glXClientInfo(Display * dpy, int opcode)
 ** EXT_import_context
 */
 
-PUBLIC Display *
+_X_EXPORT Display *
 glXGetCurrentDisplay(void)
 {
-   GLXContext gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    if (NULL == gc)
       return NULL;
    return gc->currentDpy;
 }
 
-PUBLIC
+_X_EXPORT
 GLX_ALIAS(Display *, glXGetCurrentDisplayEXT, (void), (),
           glXGetCurrentDisplay)
 
 #ifndef GLX_USE_APPLEGL
-/**
- * Used internally by libGL to send \c xGLXQueryContextinfoExtReq requests
- * to the X-server.
- *
- * \param dpy  Display where \c ctx was created.
- * \param ctx  Context to query.
- * \returns  \c Success on success.  \c GLX_BAD_CONTEXT if \c ctx is invalid,
- *           or zero if the request failed due to internal problems (i.e.,
- *           unable to allocate temporary memory, etc.)
- *
- * \note
- * This function dynamically determines whether to use the EXT_import_context
- * version of the protocol or the GLX 1.3 version of the protocol.
- */
-static int __glXQueryContextInfo(Display * dpy, GLXContext ctx)
+_X_EXPORT GLXContext
+glXImportContextEXT(Display *dpy, GLXContextID contextID)
 {
-   __GLXdisplayPrivate *priv = __glXInitialize(dpy);
+   struct glx_display *priv = __glXInitialize(dpy);
+   struct glx_screen *psc;
    xGLXQueryContextReply reply;
    CARD8 opcode;
-   GLuint numValues;
-   int retval;
+   struct glx_context *ctx;
+   int propList[__GLX_MAX_CONTEXT_PROPS * 2], *pProp, nPropListBytes;
+   int i, renderType;
+   XID share;
+   struct glx_config *mode;
+
+   if (contextID == None || __glXIsDirect(dpy, contextID))
+      return NULL;
 
-   if (ctx == NULL) {
-      return GLX_BAD_CONTEXT;
-   }
    opcode = __glXSetupForCommand(dpy);
-   if (!opcode) {
+   if (!opcode)
       return 0;
-   }
 
    /* Send the glXQueryContextInfoEXT request */
    LockDisplay(dpy);
 
-   if ((priv->majorVersion > 1) || (priv->minorVersion >= 3)) {
+   if (priv->majorVersion > 1 || priv->minorVersion >= 3) {
       xGLXQueryContextReq *req;
 
       GetReq(GLXQueryContext, req);
 
       req->reqType = opcode;
       req->glxCode = X_GLXQueryContext;
-      req->context = (unsigned int) (ctx->xid);
+      req->context = contextID;
    }
    else {
       xGLXVendorPrivateReq *vpreq;
       xGLXQueryContextInfoEXTReq *req;
 
       GetReqExtra(GLXVendorPrivate,
-                  sz_xGLXQueryContextInfoEXTReq - sz_xGLXVendorPrivateReq,
-                  vpreq);
+		  sz_xGLXQueryContextInfoEXTReq - sz_xGLXVendorPrivateReq,
+		  vpreq);
       req = (xGLXQueryContextInfoEXTReq *) vpreq;
       req->reqType = opcode;
       req->glxCode = X_GLXVendorPrivateWithReply;
       req->vendorCode = X_GLXvop_QueryContextInfoEXT;
-      req->context = (unsigned int) (ctx->xid);
+      req->context = contextID;
    }
 
    _XReply(dpy, (xReply *) & reply, 0, False);
 
-   numValues = reply.n;
-   if (numValues == 0)
-      retval = Success;
-   else if (numValues > __GLX_MAX_CONTEXT_PROPS)
-      retval = 0;
-   else {
-      int *propList, *pProp;
-      int nPropListBytes;
-
-      nPropListBytes = numValues << 3;
-      propList = (int *) Xmalloc(nPropListBytes);
-      if (NULL == propList) {
-         retval = 0;
-      }
-      else {
-	 unsigned i;
-
-         _XRead(dpy, (char *) propList, nPropListBytes);
-
-	 /* Look up screen first so we can look up visuals/fbconfigs later */
-         pProp = propList;
-         for (i = 0; i < numValues; i++, pProp += 2)
-	     if (pProp[0] == GLX_SCREEN) {
-		 ctx->screen = pProp[1];
-		 ctx->psc = GetGLXScreenConfigs(dpy, ctx->screen);
-	     }
-
-         pProp = propList;
-         for (i = 0; i < numValues; i++) {
-            switch (*pProp++) {
-            case GLX_SHARE_CONTEXT_EXT:
-               ctx->share_xid = *pProp++;
-               break;
-            case GLX_VISUAL_ID_EXT:
-               ctx->mode =
-                  _gl_context_modes_find_visual(ctx->psc->visuals, *pProp++);
-               break;
-            case GLX_FBCONFIG_ID:
-               ctx->mode =
-                  _gl_context_modes_find_fbconfig(ctx->psc->configs,
-                                                  *pProp++);
-               break;
-            case GLX_RENDER_TYPE:
-               ctx->renderType = *pProp++;
-               break;
-            case GLX_SCREEN:
-            default:
-               pProp++;
-               continue;
-            }
-         }
-         Xfree((char *) propList);
-         retval = Success;
-      }
-   }
+   if (reply.n <= __GLX_MAX_CONTEXT_PROPS)
+      nPropListBytes = reply.n * 2 * sizeof propList[0];
+   else
+      nPropListBytes = 0;
+   _XRead(dpy, (char *) propList, nPropListBytes);
    UnlockDisplay(dpy);
    SyncHandle();
-   return retval;
+
+   /* Look up screen first so we can look up visuals/fbconfigs later */
+   psc = NULL;
+   for (i = 0, pProp = propList; i < reply.n; i++, pProp += 2)
+      if (pProp[0] == GLX_SCREEN)
+	 psc = GetGLXScreenConfigs(dpy, pProp[1]);
+   if (psc == NULL)
+      return NULL;
+
+   share = None;
+   mode = NULL;
+   renderType = 0;
+   pProp = propList;
+
+   for (i = 0, pProp = propList; i < reply.n; i++, pProp += 2)
+      switch (pProp[0]) {
+      case GLX_SHARE_CONTEXT_EXT:
+	 share = pProp[1];
+	 break;
+      case GLX_VISUAL_ID_EXT:
+	 mode = glx_config_find_visual(psc->visuals, pProp[1]);
+	 break;
+      case GLX_FBCONFIG_ID:
+	 mode = glx_config_find_fbconfig(psc->configs, pProp[1]);
+	 break;
+      case GLX_RENDER_TYPE:
+	 renderType = pProp[1];
+	 break;
+      }
+
+   if (mode == NULL)
+      return NULL;
+
+   ctx = indirect_create_context(psc, mode, NULL, renderType);
+   if (ctx == NULL)
+      return NULL;
+
+   ctx->xid = contextID;
+   ctx->imported = GL_TRUE;
+   ctx->share_xid = share;
+
+   return (GLXContext) ctx;
 }
 
 #endif
 
-PUBLIC int
-glXQueryContext(Display * dpy, GLXContext ctx, int attribute, int *value)
+_X_EXPORT int
+glXQueryContext(Display * dpy, GLXContext ctx_user, int attribute, int *value)
 {
-#ifndef GLX_USE_APPLEGL
-   int retVal;
-
-   /* get the information from the server if we don't have it already */
-#if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
-   if (!ctx->driContext && (ctx->mode == NULL)) {
-#else
-   if (ctx->mode == NULL) {
-#endif
-      retVal = __glXQueryContextInfo(dpy, ctx);
-      if (Success != retVal)
-         return retVal;
-   }
-#endif
+   struct glx_context *ctx = (struct glx_context *) ctx_user;
 
    switch (attribute) {
-#ifndef GLX_USE_APPLEGL
       case GLX_SHARE_CONTEXT_EXT:
-      *value = (int) (ctx->share_xid);
+      *value = ctx->share_xid;
       break;
    case GLX_VISUAL_ID_EXT:
-      *value = ctx->mode ? ctx->mode->visualID : None;
+      *value = ctx->config ? ctx->config->visualID : None;
       break;
-#endif
    case GLX_SCREEN:
-      *value = (int) (ctx->screen);
+      *value = ctx->screen;
       break;
    case GLX_FBCONFIG_ID:
-      *value = ctx->mode ? ctx->mode->fbconfigID : None;
+      *value = ctx->config ? ctx->config->fbconfigID : None;
       break;
    case GLX_RENDER_TYPE:
-      *value = (int) (ctx->renderType);
+      *value = ctx->renderType;
       break;
    default:
       return GLX_BAD_ATTRIBUTE;
@@ -1826,66 +1508,34 @@ glXQueryContext(Display * dpy, GLXContext ctx, int attribute, int *value)
    return Success;
 }
 
-PUBLIC
+_X_EXPORT
 GLX_ALIAS(int, glXQueryContextInfoEXT,
           (Display * dpy, GLXContext ctx, int attribute, int *value),
           (dpy, ctx, attribute, value), glXQueryContext)
 
-PUBLIC GLXContextID glXGetContextIDEXT(const GLXContext ctx)
+_X_EXPORT GLXContextID glXGetContextIDEXT(const GLXContext ctx_user)
 {
-   return ctx->xid;
-}
+   struct glx_context *ctx = (struct glx_context *) ctx_user;
 
-PUBLIC GLXContext
-glXImportContextEXT(Display * dpy, GLXContextID contextID)
-{
-#ifdef GLX_USE_APPLEGL
-   return NULL;
-#else
-   GLXContext ctx;
-
-   if (contextID == None) {
-      return NULL;
-   }
-   if (__glXIsDirect(dpy, contextID)) {
-      return NULL;
-   }
-
-   ctx = AllocateGLXContext(dpy);
-   if (NULL != ctx) {
-      ctx->xid = contextID;
-      ctx->imported = GL_TRUE;
-
-      if (Success != __glXQueryContextInfo(dpy, ctx)) {
-	 __glXFreeContext(ctx);
-	 ctx = NULL;
-      }
-   }
-   return ctx;
-#endif
+   return ctx->xid;
 }
 
-PUBLIC void
+_X_EXPORT void
 glXFreeContextEXT(Display * dpy, GLXContext ctx)
 {
    DestroyContext(dpy, ctx);
 }
 
 
-
-/*
- * GLX 1.3 functions - these are just stubs for now!
- */
-
-PUBLIC GLXFBConfig *
+_X_EXPORT GLXFBConfig *
 glXChooseFBConfig(Display * dpy, int screen,
                   const int *attribList, int *nitems)
 {
-   __GLcontextModes **config_list;
+   struct glx_config **config_list;
    int list_size;
 
 
-   config_list = (__GLcontextModes **)
+   config_list = (struct glx_config **)
       glXGetFBConfigs(dpy, screen, &list_size);
 
    if ((config_list != NULL) && (list_size > 0) && (attribList != NULL)) {
@@ -1901,92 +1551,92 @@ glXChooseFBConfig(Display * dpy, int screen,
 }
 
 
-PUBLIC GLXContext
-glXCreateNewContext(Display * dpy, GLXFBConfig config,
+_X_EXPORT GLXContext
+glXCreateNewContext(Display * dpy, GLXFBConfig fbconfig,
                     int renderType, GLXContext shareList, Bool allowDirect)
 {
-   const __GLcontextModes *const fbconfig =
-      (const __GLcontextModes *const) config;
+   struct glx_config *config = (struct glx_config *) fbconfig;
 
-   return CreateContext(dpy, fbconfig->fbconfigID, fbconfig, shareList,
-                        allowDirect, X_GLXCreateNewContext, renderType,
-			fbconfig->screen);
+   return CreateContext(dpy, config->fbconfigID, config, shareList,
+			allowDirect, X_GLXCreateNewContext, renderType,
+			config->screen);
 }
 
 
-PUBLIC GLXDrawable
+_X_EXPORT GLXDrawable
 glXGetCurrentReadDrawable(void)
 {
-   GLXContext gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
+
    return gc->currentReadable;
 }
 
 
-PUBLIC GLXFBConfig *
+_X_EXPORT GLXFBConfig *
 glXGetFBConfigs(Display * dpy, int screen, int *nelements)
 {
-   __GLXdisplayPrivate *priv = __glXInitialize(dpy);
-   __GLcontextModes **config = NULL;
+   struct glx_display *priv = __glXInitialize(dpy);
+   struct glx_config **config_list = NULL;
+   struct glx_config *config;
+   unsigned num_configs = 0;
    int i;
 
    *nelements = 0;
-   if (priv && (priv->screenConfigs != NULL)
+   if (priv && (priv->screens != NULL)
        && (screen >= 0) && (screen <= ScreenCount(dpy))
-       && (priv->screenConfigs[screen]->configs != NULL)
-       && (priv->screenConfigs[screen]->configs->fbconfigID
+       && (priv->screens[screen]->configs != NULL)
+       && (priv->screens[screen]->configs->fbconfigID
 	   != (int) GLX_DONT_CARE)) {
-      unsigned num_configs = 0;
-      __GLcontextModes *modes;
-
 
-      for (modes = priv->screenConfigs[screen]->configs; modes != NULL;
-           modes = modes->next) {
-         if (modes->fbconfigID != (int) GLX_DONT_CARE) {
+      for (config = priv->screens[screen]->configs; config != NULL;
+           config = config->next) {
+         if (config->fbconfigID != (int) GLX_DONT_CARE) {
             num_configs++;
          }
       }
 
-      config = (__GLcontextModes **) Xmalloc(sizeof(__GLcontextModes *)
-                                             * num_configs);
-      if (config != NULL) {
+      config_list = Xmalloc(num_configs * sizeof *config_list);
+      if (config_list != NULL) {
          *nelements = num_configs;
          i = 0;
-         for (modes = priv->screenConfigs[screen]->configs; modes != NULL;
-              modes = modes->next) {
-            if (modes->fbconfigID != (int) GLX_DONT_CARE) {
-               config[i] = modes;
+         for (config = priv->screens[screen]->configs; config != NULL;
+              config = config->next) {
+            if (config->fbconfigID != (int) GLX_DONT_CARE) {
+               config_list[i] = config;
                i++;
             }
          }
       }
    }
-   return (GLXFBConfig *) config;
+
+   return (GLXFBConfig *) config_list;
 }
 
 
-PUBLIC int
-glXGetFBConfigAttrib(Display * dpy, GLXFBConfig config,
+_X_EXPORT int
+glXGetFBConfigAttrib(Display * dpy, GLXFBConfig fbconfig,
                      int attribute, int *value)
 {
-   __GLcontextModes *const modes = ValidateGLXFBConfig(dpy, config);
+   struct glx_config *config = ValidateGLXFBConfig(dpy, fbconfig);
 
-   return (modes != NULL)
-      ? _gl_get_context_mode_data(modes, attribute, value)
-      : GLXBadFBConfig;
+   if (config == NULL)
+      return GLXBadFBConfig;
+
+   return glx_config_get(config, attribute, value);
 }
 
 
-PUBLIC XVisualInfo *
-glXGetVisualFromFBConfig(Display * dpy, GLXFBConfig config)
+_X_EXPORT XVisualInfo *
+glXGetVisualFromFBConfig(Display * dpy, GLXFBConfig fbconfig)
 {
    XVisualInfo visualTemplate;
-   __GLcontextModes *fbconfig = (__GLcontextModes *) config;
+   struct glx_config *config = (struct glx_config *) fbconfig;
    int count;
 
    /*
     ** Get a list of all visuals, return if list is empty
     */
-   visualTemplate.visualid = fbconfig->visualID;
+   visualTemplate.visualid = config->visualID;
    return XGetVisualInfo(dpy, VisualIDMask, &visualTemplate, &count);
 }
 
@@ -1998,8 +1648,8 @@ static int
 __glXSwapIntervalSGI(int interval)
 {
    xGLXVendorPrivateReq *req;
-   GLXContext gc = __glXGetCurrentContext();
-   __GLXscreenConfigs *psc;
+   struct glx_context *gc = __glXGetCurrentContext();
+   struct glx_screen *psc;
    Display *dpy;
    CARD32 *interval_ptr;
    CARD8 opcode;
@@ -2015,7 +1665,7 @@ __glXSwapIntervalSGI(int interval)
    psc = GetGLXScreenConfigs( gc->currentDpy, gc->screen);
 
 #ifdef GLX_DIRECT_RENDERING
-   if (gc->driContext && psc->driScreen && psc->driScreen->setSwapInterval) {
+   if (gc->isDirect && psc->driScreen && psc->driScreen->setSwapInterval) {
       __GLXDRIdrawable *pdraw =
 	 GetGLXDRIDrawable(gc->currentDpy, gc->currentDrawable);
       psc->driScreen->setSwapInterval(pdraw, interval);
@@ -2055,10 +1705,10 @@ static int
 __glXSwapIntervalMESA(unsigned int interval)
 {
 #ifdef GLX_DIRECT_RENDERING
-   GLXContext gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
 
-   if (gc != NULL && gc->driContext) {
-      __GLXscreenConfigs *psc;
+   if (gc != NULL && gc->isDirect) {
+      struct glx_screen *psc;
 
       psc = GetGLXScreenConfigs( gc->currentDpy, gc->screen);
       if (psc->driScreen && psc->driScreen->setSwapInterval) {
@@ -2077,10 +1727,10 @@ static int
 __glXGetSwapIntervalMESA(void)
 {
 #ifdef GLX_DIRECT_RENDERING
-   GLXContext gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
 
-   if (gc != NULL && gc->driContext) {
-      __GLXscreenConfigs *psc;
+   if (gc != NULL && gc->isDirect) {
+      struct glx_screen *psc;
 
       psc = GetGLXScreenConfigs( gc->currentDpy, gc->screen);
       if (psc->driScreen && psc->driScreen->getSwapInterval) {
@@ -2103,8 +1753,8 @@ __glXGetVideoSyncSGI(unsigned int *count)
 {
    int64_t ust, msc, sbc;
    int ret;
-   GLXContext gc = __glXGetCurrentContext();
-   __GLXscreenConfigs *psc;
+   struct glx_context *gc = __glXGetCurrentContext();
+   struct glx_screen *psc;
 #ifdef GLX_DIRECT_RENDERING
    __GLXDRIdrawable *pdraw;
 #endif
@@ -2113,7 +1763,7 @@ __glXGetVideoSyncSGI(unsigned int *count)
       return GLX_BAD_CONTEXT;
 
 #ifdef GLX_DIRECT_RENDERING
-   if (!gc->driContext)
+   if (!gc->isDirect)
       return GLX_BAD_CONTEXT;
 #endif
 
@@ -2140,8 +1790,8 @@ __glXGetVideoSyncSGI(unsigned int *count)
 static int
 __glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count)
 {
-   GLXContext gc = __glXGetCurrentContext();
-   __GLXscreenConfigs *psc;
+   struct glx_context *gc = __glXGetCurrentContext();
+   struct glx_screen *psc;
 #ifdef GLX_DIRECT_RENDERING
    __GLXDRIdrawable *pdraw;
 #endif
@@ -2155,7 +1805,7 @@ __glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count)
       return GLX_BAD_CONTEXT;
 
 #ifdef GLX_DIRECT_RENDERING
-   if (!gc->driContext)
+   if (!gc->isDirect)
       return GLX_BAD_CONTEXT;
 #endif
 
@@ -2184,23 +1834,23 @@ __glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count)
 ** GLX_functions table.
 */
 
-PUBLIC
+_X_EXPORT
 GLX_ALIAS(int, glXGetFBConfigAttribSGIX,
           (Display * dpy, GLXFBConfigSGIX config, int attribute, int *value),
           (dpy, config, attribute, value), glXGetFBConfigAttrib)
 
-PUBLIC GLX_ALIAS(GLXFBConfigSGIX *, glXChooseFBConfigSGIX,
+_X_EXPORT GLX_ALIAS(GLXFBConfigSGIX *, glXChooseFBConfigSGIX,
                  (Display * dpy, int screen, int *attrib_list,
                   int *nelements), (dpy, screen, attrib_list, nelements),
                  glXChooseFBConfig)
 
-PUBLIC GLX_ALIAS(XVisualInfo *, glXGetVisualFromFBConfigSGIX,
+_X_EXPORT GLX_ALIAS(XVisualInfo *, glXGetVisualFromFBConfigSGIX,
                  (Display * dpy, GLXFBConfigSGIX config),
                  (dpy, config), glXGetVisualFromFBConfig)
 
-PUBLIC GLXPixmap
+_X_EXPORT GLXPixmap
 glXCreateGLXPixmapWithConfigSGIX(Display * dpy,
-                                 GLXFBConfigSGIX config,
+                                 GLXFBConfigSGIX fbconfig,
                                  Pixmap pixmap)
 {
 #ifndef GLX_USE_APPLEGL
@@ -2208,21 +1858,21 @@ glXCreateGLXPixmapWithConfigSGIX(Display * dpy,
    xGLXCreateGLXPixmapWithConfigSGIXReq *req;
    GLXPixmap xid = None;
    CARD8 opcode;
-   __GLXscreenConfigs *psc;
+   struct glx_screen *psc;
 #endif
-   const __GLcontextModes *const fbconfig = (__GLcontextModes *) config;
+   struct glx_config *config = (struct glx_config *) fbconfig;
 
 
    if ((dpy == NULL) || (config == NULL)) {
       return None;
    }
 #ifdef GLX_USE_APPLEGL
-   if(apple_glx_pixmap_create(dpy, fbconfig->screen, pixmap, fbconfig))
+   if(apple_glx_pixmap_create(dpy, config->screen, pixmap, config))
       return None;
    return pixmap;
 #else
 
-   psc = GetGLXScreenConfigs(dpy, fbconfig->screen);
+   psc = GetGLXScreenConfigs(dpy, config->screen);
    if ((psc != NULL)
        && __glXExtensionBitIsEnabled(psc, SGIX_fbconfig_bit)) {
       opcode = __glXSetupForCommand(dpy);
@@ -2239,8 +1889,8 @@ glXCreateGLXPixmapWithConfigSGIX(Display * dpy,
       req->reqType = opcode;
       req->glxCode = X_GLXVendorPrivateWithReply;
       req->vendorCode = X_GLXvop_CreateGLXPixmapWithConfigSGIX;
-      req->screen = fbconfig->screen;
-      req->fbconfig = fbconfig->fbconfigID;
+      req->screen = config->screen;
+      req->fbconfig = config->fbconfigID;
       req->pixmap = pixmap;
       req->glxpixmap = xid = XAllocID(dpy);
       UnlockDisplay(dpy);
@@ -2251,44 +1901,44 @@ glXCreateGLXPixmapWithConfigSGIX(Display * dpy,
 #endif
 }
 
-PUBLIC GLXContext
+_X_EXPORT GLXContext
 glXCreateContextWithConfigSGIX(Display * dpy,
-                               GLXFBConfigSGIX config, int renderType,
+                               GLXFBConfigSGIX fbconfig, int renderType,
                                GLXContext shareList, Bool allowDirect)
 {
    GLXContext gc = NULL;
-   const __GLcontextModes *const fbconfig = (__GLcontextModes *) config;
-   __GLXscreenConfigs *psc;
+   struct glx_config *config = (struct glx_config *) fbconfig;
+   struct glx_screen *psc;
 
 
    if ((dpy == NULL) || (config == NULL)) {
       return None;
    }
 
-   psc = GetGLXScreenConfigs(dpy, fbconfig->screen);
+   psc = GetGLXScreenConfigs(dpy, config->screen);
    if ((psc != NULL)
        && __glXExtensionBitIsEnabled(psc, SGIX_fbconfig_bit)) {
-      gc = CreateContext(dpy, fbconfig->fbconfigID, fbconfig, shareList,
+      gc = CreateContext(dpy, config->fbconfigID, config, shareList,
                          allowDirect,
 			 X_GLXvop_CreateContextWithConfigSGIX, renderType,
-			 fbconfig->screen);
+			 config->screen);
    }
 
    return gc;
 }
 
 
-PUBLIC GLXFBConfigSGIX
+_X_EXPORT GLXFBConfigSGIX
 glXGetFBConfigFromVisualSGIX(Display * dpy, XVisualInfo * vis)
 {
-   __GLXdisplayPrivate *priv;
-   __GLXscreenConfigs *psc = NULL;
+   struct glx_display *priv;
+   struct glx_screen *psc = NULL;
 
    if ((GetGLXPrivScreenConfig(dpy, vis->screen, &priv, &psc) != Success)
        && __glXExtensionBitIsEnabled(psc, SGIX_fbconfig_bit)
        && (psc->configs->fbconfigID != (int) GLX_DONT_CARE)) {
-      return (GLXFBConfigSGIX) _gl_context_modes_find_visual(psc->configs,
-                                                             vis->visualid);
+      return (GLXFBConfigSGIX) glx_config_find_visual(psc->configs,
+						      vis->visualid);
    }
 
    return NULL;
@@ -2336,12 +1986,12 @@ static Bool
 __glXGetSyncValuesOML(Display * dpy, GLXDrawable drawable,
                       int64_t * ust, int64_t * msc, int64_t * sbc)
 {
-   __GLXdisplayPrivate * const priv = __glXInitialize(dpy);
+   struct glx_display * const priv = __glXInitialize(dpy);
    int ret;
 #ifdef GLX_DIRECT_RENDERING
    __GLXDRIdrawable *pdraw;
 #endif
-   __GLXscreenConfigs *psc;
+   struct glx_screen *psc;
 
    if (!priv)
       return False;
@@ -2364,7 +2014,7 @@ __glxGetMscRate(__GLXDRIdrawable *glxDraw,
 		int32_t * numerator, int32_t * denominator)
 {
 #ifdef XF86VIDMODE
-   __GLXscreenConfigs *psc;
+   struct glx_screen *psc;
    XF86VidModeModeLine mode_line;
    int dot_clock;
    int i;
@@ -2466,17 +2116,17 @@ static int64_t
 __glXSwapBuffersMscOML(Display * dpy, GLXDrawable drawable,
                        int64_t target_msc, int64_t divisor, int64_t remainder)
 {
-   GLXContext gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
 #ifdef GLX_DIRECT_RENDERING
    __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable);
-   __GLXscreenConfigs *psc = pdraw ? pdraw->psc : NULL;
+   struct glx_screen *psc = pdraw ? pdraw->psc : NULL;
 #endif
 
    if (!gc) /* no GLX for this */
       return -1;
 
 #ifdef GLX_DIRECT_RENDERING
-   if (!pdraw || !gc->driContext)
+   if (!pdraw || !gc->isDirect)
       return -1;
 #endif
 
@@ -2512,7 +2162,7 @@ __glXWaitForMscOML(Display * dpy, GLXDrawable drawable,
 #ifdef GLX_DIRECT_RENDERING
    __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable);
 #endif
-   __GLXscreenConfigs *psc = pdraw ? pdraw->psc : NULL;
+   struct glx_screen *psc = pdraw ? pdraw->psc : NULL;
    int ret;
 
 
@@ -2544,7 +2194,7 @@ __glXWaitForSbcOML(Display * dpy, GLXDrawable drawable,
 #ifdef GLX_DIRECT_RENDERING
    __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable);
 #endif
-   __GLXscreenConfigs *psc = pdraw ? pdraw->psc : NULL;
+   struct glx_screen *psc = pdraw ? pdraw->psc : NULL;
    int ret;
 
    /* The OML_sync_control spec says this should "generate a GLX_BAD_VALUE
@@ -2602,7 +2252,7 @@ __glXReleaseBuffersMESA(Display * dpy, GLXDrawable d)
 }
 
 
-PUBLIC GLXPixmap
+_X_EXPORT GLXPixmap
 glXCreateGLXPixmapMESA(Display * dpy, XVisualInfo * visual,
                        Pixmap pixmap, Colormap cmap)
 {
@@ -2625,16 +2275,16 @@ __glXCopySubBufferMESA(Display * dpy, GLXDrawable drawable,
                        int x, int y, int width, int height)
 {
    xGLXVendorPrivateReq *req;
-   GLXContext gc;
+   struct glx_context *gc;
    GLXContextTag tag;
    CARD32 *drawable_ptr;
    INT32 *x_ptr, *y_ptr, *w_ptr, *h_ptr;
    CARD8 opcode;
 
-#ifdef __DRI_COPY_SUB_BUFFER
+#if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
    __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable);
    if (pdraw != NULL) {
-      __GLXscreenConfigs *psc = pdraw->psc;
+      struct glx_screen *psc = pdraw->psc;
       if (psc->driScreen->copySubBuffer != NULL) {
          glFlush();
          (*psc->driScreen->copySubBuffer) (pdraw, x, y, width, height);
@@ -2685,107 +2335,12 @@ __glXCopySubBufferMESA(Display * dpy, GLXDrawable drawable,
    SyncHandle();
 }
 
-
-/**
- * GLX_EXT_texture_from_pixmap
- */
-static void
-indirect_bind_tex_image(Display * dpy,
-			GLXDrawable drawable,
-			int buffer, const int *attrib_list)
-{
-   xGLXVendorPrivateReq *req;
-   GLXContext gc = __glXGetCurrentContext();
-   CARD32 *drawable_ptr;
-   INT32 *buffer_ptr;
-   CARD32 *num_attrib_ptr;
-   CARD32 *attrib_ptr;
-   CARD8 opcode;
-   unsigned int i;
-
-   i = 0;
-   if (attrib_list) {
-      while (attrib_list[i * 2] != None)
-         i++;
-   }
-
-   opcode = __glXSetupForCommand(dpy);
-   if (!opcode)
-      return;
-
-   LockDisplay(dpy);
-   GetReqExtra(GLXVendorPrivate, 12 + 8 * i, req);
-   req->reqType = opcode;
-   req->glxCode = X_GLXVendorPrivate;
-   req->vendorCode = X_GLXvop_BindTexImageEXT;
-   req->contextTag = gc->currentContextTag;
-
-   drawable_ptr = (CARD32 *) (req + 1);
-   buffer_ptr = (INT32 *) (drawable_ptr + 1);
-   num_attrib_ptr = (CARD32 *) (buffer_ptr + 1);
-   attrib_ptr = (CARD32 *) (num_attrib_ptr + 1);
-
-   *drawable_ptr = drawable;
-   *buffer_ptr = buffer;
-   *num_attrib_ptr = (CARD32) i;
-
-   i = 0;
-   if (attrib_list) {
-      while (attrib_list[i * 2] != None) {
-         *attrib_ptr++ = (CARD32) attrib_list[i * 2 + 0];
-         *attrib_ptr++ = (CARD32) attrib_list[i * 2 + 1];
-         i++;
-      }
-   }
-
-   UnlockDisplay(dpy);
-   SyncHandle();
-}
-
-static void
-indirect_release_tex_image(Display * dpy, GLXDrawable drawable, int buffer)
-{
-   xGLXVendorPrivateReq *req;
-   GLXContext gc = __glXGetCurrentContext();
-   CARD32 *drawable_ptr;
-   INT32 *buffer_ptr;
-   CARD8 opcode;
-
-   opcode = __glXSetupForCommand(dpy);
-   if (!opcode)
-      return;
-
-   LockDisplay(dpy);
-   GetReqExtra(GLXVendorPrivate, sizeof(CARD32) + sizeof(INT32), req);
-   req->reqType = opcode;
-   req->glxCode = X_GLXVendorPrivate;
-   req->vendorCode = X_GLXvop_ReleaseTexImageEXT;
-   req->contextTag = gc->currentContextTag;
-
-   drawable_ptr = (CARD32 *) (req + 1);
-   buffer_ptr = (INT32 *) (drawable_ptr + 1);
-
-   *drawable_ptr = drawable;
-   *buffer_ptr = buffer;
-
-   UnlockDisplay(dpy);
-   SyncHandle();
-}
-
-static const struct glx_context_vtable indirect_context_vtable = {
-   indirect_wait_gl,
-   indirect_wait_x,
-   indirect_use_x_font,
-   indirect_bind_tex_image,
-   indirect_release_tex_image,
-};
-
 /*@{*/
 static void
 __glXBindTexImageEXT(Display * dpy,
                      GLXDrawable drawable, int buffer, const int *attrib_list)
 {
-   GLXContext gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
 
    if (gc == NULL || gc->vtable->bind_tex_image == NULL)
       return;
@@ -2796,7 +2351,7 @@ __glXBindTexImageEXT(Display * dpy,
 static void
 __glXReleaseTexImageEXT(Display * dpy, GLXDrawable drawable, int buffer)
 {
-   GLXContext gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
 
    if (gc == NULL || gc->vtable->release_tex_image == NULL)
       return;
@@ -2995,7 +2550,7 @@ get_glx_proc_address(const char *funcName)
  *
  * \sa glXGetProcAddress
  */
-PUBLIC void (*glXGetProcAddressARB(const GLubyte * procName)) (void)
+_X_EXPORT void (*glXGetProcAddressARB(const GLubyte * procName)) (void)
 {
    typedef void (*gl_function) (void);
    gl_function f;
@@ -3028,7 +2583,7 @@ PUBLIC void (*glXGetProcAddressARB(const GLubyte * procName)) (void)
  *
  * \sa glXGetProcAddressARB
  */
-PUBLIC void (*glXGetProcAddress(const GLubyte * procName)) (void)
+_X_EXPORT void (*glXGetProcAddress(const GLubyte * procName)) (void)
 #if defined(__GNUC__) && !defined(GLX_ALIAS_UNSUPPORTED)
    __attribute__ ((alias("glXGetProcAddressARB")));
 #else
diff --git a/src/glx/glxconfig.c b/src/glx/glxconfig.c
new file mode 100644
index 00000000000..1d9678f48cd
--- /dev/null
+++ b/src/glx/glxconfig.c
@@ -0,0 +1,304 @@
+/*
+ * (C) Copyright IBM Corporation 2003
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEM, IBM AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file glxconfig.c
+ * Utility routines for working with \c struct glx_config structures.  At
+ * some point most or all of these functions will be moved to the Mesa
+ * code base.
+ *
+ * \author Ian Romanick <[email protected]>
+ */
+
+#include <GL/glx.h>
+#include "GL/glxint.h"
+#include <stdlib.h>
+#include <string.h>
+
+#include "glxconfig.h"
+
+#define NUM_VISUAL_TYPES   6
+
+/**
+ * Get data from a GLX config
+ * 
+ * \param mode         GL context mode whose data is to be returned.
+ * \param attribute    Attribute of \c mode that is to be returned.
+ * \param value_return Location to store the data member of \c mode.
+ * \return  If \c attribute is a valid attribute of \c mode, zero is
+ *          returned.  Otherwise \c GLX_BAD_ATTRIBUTE is returned.
+ */
+_X_HIDDEN int
+glx_config_get(struct glx_config * mode, int attribute, int *value_return)
+{
+   switch (attribute) {
+   case GLX_USE_GL:
+      *value_return = GL_TRUE;
+      return 0;
+   case GLX_BUFFER_SIZE:
+      *value_return = mode->rgbBits;
+      return 0;
+   case GLX_RGBA:
+      *value_return = mode->rgbMode;
+      return 0;
+   case GLX_RED_SIZE:
+      *value_return = mode->redBits;
+      return 0;
+   case GLX_GREEN_SIZE:
+      *value_return = mode->greenBits;
+      return 0;
+   case GLX_BLUE_SIZE:
+      *value_return = mode->blueBits;
+      return 0;
+   case GLX_ALPHA_SIZE:
+      *value_return = mode->alphaBits;
+      return 0;
+   case GLX_DOUBLEBUFFER:
+      *value_return = mode->doubleBufferMode;
+      return 0;
+   case GLX_STEREO:
+      *value_return = mode->stereoMode;
+      return 0;
+   case GLX_AUX_BUFFERS:
+      *value_return = mode->numAuxBuffers;
+      return 0;
+   case GLX_DEPTH_SIZE:
+      *value_return = mode->depthBits;
+      return 0;
+   case GLX_STENCIL_SIZE:
+      *value_return = mode->stencilBits;
+      return 0;
+   case GLX_ACCUM_RED_SIZE:
+      *value_return = mode->accumRedBits;
+      return 0;
+   case GLX_ACCUM_GREEN_SIZE:
+      *value_return = mode->accumGreenBits;
+      return 0;
+   case GLX_ACCUM_BLUE_SIZE:
+      *value_return = mode->accumBlueBits;
+      return 0;
+   case GLX_ACCUM_ALPHA_SIZE:
+      *value_return = mode->accumAlphaBits;
+      return 0;
+   case GLX_LEVEL:
+      *value_return = mode->level;
+      return 0;
+#ifndef GLX_USE_APPLEGL               /* This isn't supported by CGL. */
+   case GLX_TRANSPARENT_TYPE_EXT:
+      *value_return = mode->transparentPixel;
+      return 0;
+#endif
+   case GLX_TRANSPARENT_RED_VALUE:
+      *value_return = mode->transparentRed;
+      return 0;
+   case GLX_TRANSPARENT_GREEN_VALUE:
+      *value_return = mode->transparentGreen;
+      return 0;
+   case GLX_TRANSPARENT_BLUE_VALUE:
+      *value_return = mode->transparentBlue;
+      return 0;
+   case GLX_TRANSPARENT_ALPHA_VALUE:
+      *value_return = mode->transparentAlpha;
+      return 0;
+   case GLX_TRANSPARENT_INDEX_VALUE:
+      *value_return = mode->transparentIndex;
+      return 0;
+   case GLX_X_VISUAL_TYPE:
+      *value_return = mode->visualType;
+      return 0;
+   case GLX_CONFIG_CAVEAT:
+      *value_return = mode->visualRating;
+      return 0;
+   case GLX_VISUAL_ID:
+      *value_return = mode->visualID;
+      return 0;
+   case GLX_DRAWABLE_TYPE:
+      *value_return = mode->drawableType;
+      return 0;
+   case GLX_RENDER_TYPE:
+      *value_return = mode->renderType;
+      return 0;
+   case GLX_X_RENDERABLE:
+      *value_return = mode->xRenderable;
+      return 0;
+   case GLX_FBCONFIG_ID:
+      *value_return = mode->fbconfigID;
+      return 0;
+   case GLX_MAX_PBUFFER_WIDTH:
+      *value_return = mode->maxPbufferWidth;
+      return 0;
+   case GLX_MAX_PBUFFER_HEIGHT:
+      *value_return = mode->maxPbufferHeight;
+      return 0;
+   case GLX_MAX_PBUFFER_PIXELS:
+      *value_return = mode->maxPbufferPixels;
+      return 0;
+#ifndef GLX_USE_APPLEGL               /* These aren't supported by CGL. */
+   case GLX_OPTIMAL_PBUFFER_WIDTH_SGIX:
+      *value_return = mode->optimalPbufferWidth;
+      return 0;
+   case GLX_OPTIMAL_PBUFFER_HEIGHT_SGIX:
+      *value_return = mode->optimalPbufferHeight;
+      return 0;
+   case GLX_SWAP_METHOD_OML:
+      *value_return = mode->swapMethod;
+      return 0;
+#endif
+   case GLX_SAMPLE_BUFFERS_SGIS:
+      *value_return = mode->sampleBuffers;
+      return 0;
+   case GLX_SAMPLES_SGIS:
+      *value_return = mode->samples;
+      return 0;
+   case GLX_BIND_TO_TEXTURE_RGB_EXT:
+      *value_return = mode->bindToTextureRgb;
+      return 0;
+   case GLX_BIND_TO_TEXTURE_RGBA_EXT:
+      *value_return = mode->bindToTextureRgba;
+      return 0;
+   case GLX_BIND_TO_MIPMAP_TEXTURE_EXT:
+      *value_return = mode->bindToMipmapTexture == GL_TRUE ? GL_TRUE :
+         GL_FALSE;
+      return 0;
+   case GLX_BIND_TO_TEXTURE_TARGETS_EXT:
+      *value_return = mode->bindToTextureTargets;
+      return 0;
+   case GLX_Y_INVERTED_EXT:
+      *value_return = mode->yInverted;
+      return 0;
+
+      /* Applications are NOT allowed to query GLX_VISUAL_SELECT_GROUP_SGIX.
+       * It is ONLY for communication between the GLX client and the GLX
+       * server.
+       */
+   case GLX_VISUAL_SELECT_GROUP_SGIX:
+   default:
+      return GLX_BAD_ATTRIBUTE;
+   }
+}
+
+
+/**
+ * Allocate a linked list of \c struct glx_config structures.  The fields of
+ * each structure will be initialized to "reasonable" default values.  In
+ * most cases this is the default value defined by table 3.4 of the GLX
+ * 1.3 specification.  This means that most values are either initialized to
+ * zero or \c GLX_DONT_CARE (which is -1).  As support for additional
+ * extensions is added, the new values will be initialized to appropriate
+ * values from the extension specification.
+ * 
+ * \param count         Number of structures to allocate.
+ * \param minimum_size  Minimum size of a structure to allocate.  This allows
+ *                      for differences in the version of the
+ *                      \c struct glx_config stucture used in libGL and in a
+ *                      DRI-based driver.
+ * \returns A pointer to the first element in a linked list of \c count
+ *          stuctures on success, or \c NULL on failure.
+ */
+_X_HIDDEN struct glx_config *
+glx_config_create_list(unsigned count)
+{
+   const size_t size = sizeof(struct glx_config);
+   struct glx_config *base = NULL;
+   struct glx_config **next;
+   unsigned i;
+
+   next = &base;
+   for (i = 0; i < count; i++) {
+      *next = (struct glx_config *) malloc(size);
+      if (*next == NULL) {
+	 glx_config_destroy_list(base);
+	 base = NULL;
+	 break;
+      }
+
+      (void) memset(*next, 0, size);
+      (*next)->visualID = GLX_DONT_CARE;
+      (*next)->visualType = GLX_DONT_CARE;
+      (*next)->visualRating = GLX_NONE;
+      (*next)->transparentPixel = GLX_NONE;
+      (*next)->transparentRed = GLX_DONT_CARE;
+      (*next)->transparentGreen = GLX_DONT_CARE;
+      (*next)->transparentBlue = GLX_DONT_CARE;
+      (*next)->transparentAlpha = GLX_DONT_CARE;
+      (*next)->transparentIndex = GLX_DONT_CARE;
+      (*next)->xRenderable = GLX_DONT_CARE;
+      (*next)->fbconfigID = GLX_DONT_CARE;
+      (*next)->swapMethod = GLX_SWAP_UNDEFINED_OML;
+      (*next)->bindToTextureRgb = GLX_DONT_CARE;
+      (*next)->bindToTextureRgba = GLX_DONT_CARE;
+      (*next)->bindToMipmapTexture = GLX_DONT_CARE;
+      (*next)->bindToTextureTargets = GLX_DONT_CARE;
+      (*next)->yInverted = GLX_DONT_CARE;
+
+      next = &((*next)->next);
+   }
+
+   return base;
+}
+
+_X_HIDDEN void
+glx_config_destroy_list(struct glx_config *configs)
+{
+   while (configs != NULL) {
+      struct glx_config *const next = configs->next;
+
+      free(configs);
+      configs = next;
+   }
+}
+
+
+/**
+ * Find a context mode matching a Visual ID.
+ *
+ * \param modes  List list of context-mode structures to be searched.
+ * \param vid    Visual ID to be found.
+ * \returns A pointer to a context-mode in \c modes if \c vid was found in
+ *          the list, or \c NULL if it was not.
+ */
+
+_X_HIDDEN struct glx_config *
+glx_config_find_visual(struct glx_config *configs, int vid)
+{
+   struct glx_config *c;
+
+   for (c = configs; c != NULL; c = c->next)
+      if (c->visualID == vid)
+	 return c;
+
+   return NULL;
+}
+
+_X_HIDDEN struct glx_config *
+glx_config_find_fbconfig(struct glx_config *configs, int fbid)
+{
+   struct glx_config *c;
+
+   for (c = configs; c != NULL; c = c->next)
+      if (c->fbconfigID == fbid)
+	 return c;
+
+   return NULL;
+}
diff --git a/src/glx/glxconfig.h b/src/glx/glxconfig.h
new file mode 100644
index 00000000000..f7ee958e4be
--- /dev/null
+++ b/src/glx/glxconfig.h
@@ -0,0 +1,133 @@
+/*
+ * (C) Copyright IBM Corporation 2003
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * VA LINUX SYSTEM, IBM AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * \file glcontextmodes.h
+ * \author Ian Romanick <[email protected]>
+ */
+
+#ifndef GLCONTEXTMODES_H
+#define GLCONTEXTMODES_H
+
+struct glx_config {
+    struct glx_config * next;
+
+    GLboolean rgbMode;
+    GLboolean floatMode;
+    GLboolean colorIndexMode;
+    GLuint doubleBufferMode;
+    GLuint stereoMode;
+
+    GLint redBits, greenBits, blueBits, alphaBits;	/* bits per comp */
+    GLuint redMask, greenMask, blueMask, alphaMask;
+    GLint rgbBits;		/* total bits for rgb */
+    GLint indexBits;		/* total bits for colorindex */
+
+    GLint accumRedBits, accumGreenBits, accumBlueBits, accumAlphaBits;
+    GLint depthBits;
+    GLint stencilBits;
+
+    GLint numAuxBuffers;
+
+    GLint level;
+
+    GLint pixmapMode;
+
+    /* GLX */
+    GLint visualID;
+    GLint visualType;     /**< One of the GLX X visual types. (i.e., 
+			   * \c GLX_TRUE_COLOR, etc.)
+			   */
+
+    /* EXT_visual_rating / GLX 1.2 */
+    GLint visualRating;
+
+    /* EXT_visual_info / GLX 1.2 */
+    GLint transparentPixel;
+				/*    colors are floats scaled to ints */
+    GLint transparentRed, transparentGreen, transparentBlue, transparentAlpha;
+    GLint transparentIndex;
+
+    /* ARB_multisample / SGIS_multisample */
+    GLint sampleBuffers;
+    GLint samples;
+
+    /* SGIX_fbconfig / GLX 1.3 */
+    GLint drawableType;
+    GLint renderType;
+    GLint xRenderable;
+    GLint fbconfigID;
+
+    /* SGIX_pbuffer / GLX 1.3 */
+    GLint maxPbufferWidth;
+    GLint maxPbufferHeight;
+    GLint maxPbufferPixels;
+    GLint optimalPbufferWidth;   /* Only for SGIX_pbuffer. */
+    GLint optimalPbufferHeight;  /* Only for SGIX_pbuffer. */
+
+    /* SGIX_visual_select_group */
+    GLint visualSelectGroup;
+
+    /* OML_swap_method */
+    GLint swapMethod;
+
+    GLint screen;
+
+    /* EXT_texture_from_pixmap */
+    GLint bindToTextureRgb;
+    GLint bindToTextureRgba;
+    GLint bindToMipmapTexture;
+    GLint bindToTextureTargets;
+    GLint yInverted;
+};
+
+#define __GLX_MIN_CONFIG_PROPS	18
+#define __GLX_MAX_CONFIG_PROPS	500
+#define __GLX_EXT_CONFIG_PROPS	10
+
+/*
+** Since we send all non-core visual properties as token, value pairs,
+** we require 2 words across the wire. In order to maintain backwards
+** compatibility, we need to send the total number of words that the
+** VisualConfigs are sent back in so old libraries can simply "ignore"
+** the new properties.
+*/
+#define __GLX_TOTAL_CONFIG \
+   (__GLX_MIN_CONFIG_PROPS + 2 * __GLX_EXT_CONFIG_PROPS)
+
+extern GLint _gl_convert_from_x_visual_type(int visualType);
+
+extern int
+glx_config_get(struct glx_config * mode, int attribute, int *value_return);
+extern struct glx_config *
+glx_config_create_list(unsigned count);
+extern void
+glx_config_destroy_list(struct glx_config *configs);
+extern struct glx_config *
+glx_config_find_visual(struct glx_config *configs, int vid);
+extern struct glx_config *
+glx_config_find_fbconfig(struct glx_config *configs, int fbid);
+
+#endif /* GLCONTEXTMODES_H */
+
diff --git a/src/glx/glxcurrent.c b/src/glx/glxcurrent.c
index 0bf61779c4a..e2569974c2f 100644
--- a/src/glx/glxcurrent.c
+++ b/src/glx/glxcurrent.c
@@ -45,7 +45,6 @@
 #include "apple_glx_context.h"
 #else
 #include "glapi.h"
-#include "indirect_init.h"
 #endif
 
 /*
@@ -61,7 +60,7 @@ static GLubyte dummyBuffer[__GLX_BUFFER_LIMIT_SIZE];
 ** gl and glx entry points are designed to operate as nop's when using
 ** the dummy context structure.
 */
-static __GLXcontext dummyContext = {
+struct glx_context dummyContext = {
    &dummyBuffer[0],
    &dummyBuffer[0],
    &dummyBuffer[0],
@@ -69,14 +68,6 @@ static __GLXcontext dummyContext = {
    sizeof(dummyBuffer),
 };
 
-
-#ifndef GLX_USE_APPLEGL
-/*
-** All indirect rendering contexts will share the same indirect dispatch table.
-*/
-static __GLapi *IndirectAPI = NULL;
-#endif
-
 /*
  * Current context management and locking
  */
@@ -98,7 +89,7 @@ __thread void *__glX_tls_Context __attribute__ ((tls_model("initial-exec")))
    = &dummyContext;
 
 _X_HIDDEN void
-__glXSetCurrentContext(__GLXcontext * c)
+__glXSetCurrentContext(struct glx_context * c)
 {
    __glX_tls_Context = (c != NULL) ? c : &dummyContext;
 }
@@ -133,13 +124,13 @@ init_thread_data(void)
 }
 
 _X_HIDDEN void
-__glXSetCurrentContext(__GLXcontext * c)
+__glXSetCurrentContext(struct glx_context * c)
 {
    pthread_once(&once_control, init_thread_data);
    pthread_setspecific(ContextTSD, c);
 }
 
-_X_HIDDEN __GLXcontext *
+_X_HIDDEN struct glx_context *
 __glXGetCurrentContext(void)
 {
    void *v;
@@ -147,7 +138,7 @@ __glXGetCurrentContext(void)
    pthread_once(&once_control, init_thread_data);
 
    v = pthread_getspecific(ContextTSD);
-   return (v == NULL) ? &dummyContext : (__GLXcontext *) v;
+   return (v == NULL) ? &dummyContext : (struct glx_context *) v;
 }
 
 # endif /* defined( GLX_USE_TLS ) */
@@ -159,7 +150,7 @@ __glXGetCurrentContext(void)
 #else
 
 /* not thread safe */
-_X_HIDDEN __GLXcontext *__glXcurrentContext = &dummyContext;
+_X_HIDDEN struct glx_context *__glXcurrentContext = &dummyContext;
 
 #endif
 
@@ -176,145 +167,28 @@ __glXSetCurrentContextNull(void)
 #endif
 }
 
-
-/************************************************************************/
-
-PUBLIC GLXContext
+_X_EXPORT GLXContext
 glXGetCurrentContext(void)
 {
-   GLXContext cx = __glXGetCurrentContext();
+   struct glx_context *cx = __glXGetCurrentContext();
 
    if (cx == &dummyContext) {
       return NULL;
    }
    else {
-      return cx;
+      return (GLXContext) cx;
    }
 }
 
-PUBLIC GLXDrawable
+_X_EXPORT GLXDrawable
 glXGetCurrentDrawable(void)
 {
-   GLXContext gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    return gc->currentDrawable;
 }
 
-
-#ifndef GLX_USE_APPLEGL
-/************************************************************************/
-
-/**
- * Sends a GLX protocol message to the specified display to make the context
- * and the drawables current.
- *
- * \param dpy     Display to send the message to.
- * \param opcode  Major opcode value for the display.
- * \param gc_id   Context tag for the context to be made current.
- * \param draw    Drawable ID for the "draw" drawable.
- * \param read    Drawable ID for the "read" drawable.
- * \param reply   Space to store the X-server's reply.
- *
- * \warning
- * This function assumes that \c dpy is locked with \c LockDisplay on entry.
- */
-static Bool
-SendMakeCurrentRequest(Display * dpy, CARD8 opcode,
-                       GLXContextID gc_id, GLXContextTag gc_tag,
-                       GLXDrawable draw, GLXDrawable read,
-                       xGLXMakeCurrentReply * reply)
-{
-   Bool ret;
-
-
-   LockDisplay(dpy);
-
-   if (draw == read) {
-      xGLXMakeCurrentReq *req;
-
-      GetReq(GLXMakeCurrent, req);
-      req->reqType = opcode;
-      req->glxCode = X_GLXMakeCurrent;
-      req->drawable = draw;
-      req->context = gc_id;
-      req->oldContextTag = gc_tag;
-   }
-   else {
-      __GLXdisplayPrivate *priv = __glXInitialize(dpy);
-
-      /* If the server can support the GLX 1.3 version, we should
-       * perfer that.  Not only that, some servers support GLX 1.3 but
-       * not the SGI extension.
-       */
-
-      if ((priv->majorVersion > 1) || (priv->minorVersion >= 3)) {
-         xGLXMakeContextCurrentReq *req;
-
-         GetReq(GLXMakeContextCurrent, req);
-         req->reqType = opcode;
-         req->glxCode = X_GLXMakeContextCurrent;
-         req->drawable = draw;
-         req->readdrawable = read;
-         req->context = gc_id;
-         req->oldContextTag = gc_tag;
-      }
-      else {
-         xGLXVendorPrivateWithReplyReq *vpreq;
-         xGLXMakeCurrentReadSGIReq *req;
-
-         GetReqExtra(GLXVendorPrivateWithReply,
-                     sz_xGLXMakeCurrentReadSGIReq -
-                     sz_xGLXVendorPrivateWithReplyReq, vpreq);
-         req = (xGLXMakeCurrentReadSGIReq *) vpreq;
-         req->reqType = opcode;
-         req->glxCode = X_GLXVendorPrivateWithReply;
-         req->vendorCode = X_GLXvop_MakeCurrentReadSGI;
-         req->drawable = draw;
-         req->readable = read;
-         req->context = gc_id;
-         req->oldContextTag = gc_tag;
-      }
-   }
-
-   ret = _XReply(dpy, (xReply *) reply, 0, False);
-
-   UnlockDisplay(dpy);
-   SyncHandle();
-
-   return ret;
-}
-
-
-#if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
-static __GLXDRIdrawable *
-FetchDRIDrawable(Display * dpy, GLXDrawable glxDrawable, GLXContext gc)
-{
-   __GLXdisplayPrivate *const priv = __glXInitialize(dpy);
-   __GLXDRIdrawable *pdraw;
-   __GLXscreenConfigs *psc;
-
-   if (priv == NULL)
-      return NULL;
-
-   psc = priv->screenConfigs[gc->screen];
-   if (priv->drawHash == NULL)
-      return NULL;
-
-   if (__glxHashLookup(priv->drawHash, glxDrawable, (void *) &pdraw) == 0)
-      return pdraw;
-
-   pdraw = psc->driScreen->createDrawable(psc, glxDrawable,
-                                          glxDrawable, gc->mode);
-   if (__glxHashInsert(priv->drawHash, glxDrawable, pdraw)) {
-      (*pdraw->destroyDrawable) (pdraw);
-      return NULL;
-   }
-
-   return pdraw;
-}
-#endif /* GLX_DIRECT_RENDERING */
-
 static void
-__glXGenerateError(Display * dpy, GLXContext gc, XID resource,
+__glXGenerateError(Display * dpy, struct glx_context *gc, XID resource,
                    BYTE errorCode, CARD16 minorCode)
 {
    xError error;
@@ -328,8 +202,6 @@ __glXGenerateError(Display * dpy, GLXContext gc, XID resource,
    _XError(dpy, &error);
 }
 
-#endif /* GLX_USE_APPLEGL */
-
 /**
  * Make a particular context current.
  *
@@ -337,28 +209,11 @@ __glXGenerateError(Display * dpy, GLXContext gc, XID resource,
  */
 static Bool
 MakeContextCurrent(Display * dpy, GLXDrawable draw,
-                   GLXDrawable read, GLXContext gc)
+                   GLXDrawable read, GLXContext gc_user)
 {
-   const GLXContext oldGC = __glXGetCurrentContext();
-#ifdef GLX_USE_APPLEGL
-   bool error = apple_glx_make_current_context(dpy, 
-                   (oldGC && oldGC != &dummyContext) ? oldGC->driContext : NULL, 
-                   gc ? gc->driContext : NULL, draw);
-   
-   apple_glx_diagnostic("%s: error %s\n", __func__, error ? "YES" : "NO");
-   if(error)
-      return GL_FALSE;
-#else
-   xGLXMakeCurrentReply reply;
-   const CARD8 opcode = __glXSetupForCommand(dpy);
-   const CARD8 oldOpcode = ((gc == oldGC) || (oldGC == &dummyContext))
-      ? opcode : __glXSetupForCommand(oldGC->currentDpy);
-   Bool bindReturnValue;
-   __GLXattribute *state;
-
-   if (!opcode || !oldOpcode) {
-      return GL_FALSE;
-   }
+   struct glx_context *gc = (struct glx_context *) gc_user;
+   struct glx_context *oldGC = __glXGetCurrentContext();
+   int ret = Success;
 
    /* Make sure that the new context has a nonzero ID.  In the request,
     * a zero context ID is used only to mean that we bind to no current
@@ -386,169 +241,52 @@ MakeContextCurrent(Display * dpy, GLXDrawable draw,
       return False;
    }
 
-#if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
-   /* Bind the direct rendering context to the drawable */
-   if (gc && gc->driContext) {
-      __GLXDRIdrawable *pdraw = FetchDRIDrawable(dpy, draw, gc);
-      __GLXDRIdrawable *pread = FetchDRIDrawable(dpy, read, gc);
-
-      if ((pdraw == NULL) || (pread == NULL)) {
-         __glXGenerateError(dpy, gc, (pdraw == NULL) ? draw : read,
-                            GLXBadDrawable, X_GLXMakeContextCurrent);
-         return False;
-      }
-
-      bindReturnValue =
-         (gc->driContext->bindContext) (gc, pdraw, pread);
-   }
-   else if (!gc && oldGC && oldGC->driContext) {
-      bindReturnValue = True;
-   }
-   else
-#endif
-   {
-      /* Send a glXMakeCurrent request to bind the new context. */
-      bindReturnValue =
-         SendMakeCurrentRequest(dpy, opcode, gc ? gc->xid : None,
-                                ((dpy != oldGC->currentDpy)
-                                 || oldGC->isDirect)
-                                ? None : oldGC->currentContextTag, draw, read,
-                                &reply);
+   if (oldGC != &dummyContext && oldGC != gc) {
+      oldGC->vtable->unbind(oldGC, gc);
+      oldGC->currentDpy = 0;
+      oldGC->currentDrawable = None;
+      oldGC->currentReadable = None;
+      oldGC->thread_id = 0;
+      if (oldGC->xid == None)
+	 /* We are switching away from a context that was
+	  * previously destroyed, so we need to free the memory
+	  * for the old handle.
+	  */
+	 oldGC->vtable->destroy(oldGC);
    }
 
-
-   if (!bindReturnValue) {
-      return False;
+   if (gc) {
+      ret = gc->vtable->bind(gc, oldGC, draw, read);
+      gc->currentDpy = dpy;
+      gc->currentDrawable = draw;
+      gc->currentReadable = read;
+      gc->thread_id = _glthread_GetID();
+      __glXSetCurrentContext(gc);
+   } else {
+      __glXSetCurrentContextNull();
    }
 
-#if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
-   if ((dpy != oldGC->currentDpy || (gc && gc->driContext)) &&
-       !oldGC->isDirect && oldGC != &dummyContext) {
-#else
-   if ((dpy != oldGC->currentDpy) && oldGC != &dummyContext) {
-#endif
-      xGLXMakeCurrentReply dummy_reply;
-
-      /* We are either switching from one dpy to another and have to
-       * send a request to the previous dpy to unbind the previous
-       * context, or we are switching away from a indirect context to
-       * a direct context and have to send a request to the dpy to
-       * unbind the previous context.
-       */
-      (void) SendMakeCurrentRequest(oldGC->currentDpy, oldOpcode, None,
-                                    oldGC->currentContextTag, None, None,
-                                    &dummy_reply);
-   }
-#if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
-   else if (oldGC->driContext && oldGC != gc) {
-      oldGC->driContext->unbindContext(oldGC);
-   }
-#endif
-
-#endif /* GLX_USE_APPLEGL */
-
-   /* Update our notion of what is current */
-   __glXLock();
-   if (gc == oldGC) {
-      /* Even though the contexts are the same the drawable might have
-       * changed.  Note that gc cannot be the dummy, and that oldGC
-       * cannot be NULL, therefore if they are the same, gc is not
-       * NULL and not the dummy.
-       */
-      if(gc) {
-        gc->currentDrawable = draw;
-        gc->currentReadable = read;
-      }
+   if (ret) {
+      __glXGenerateError(dpy, gc, None, ret, X_GLXMakeContextCurrent);
+      return GL_FALSE;
    }
-   else {
-      if (oldGC != &dummyContext) {
-         /* Old current context is no longer current to anybody */
-         oldGC->currentDpy = 0;
-         oldGC->currentDrawable = None;
-         oldGC->currentReadable = None;
-         oldGC->currentContextTag = 0;
-         oldGC->thread_id = 0;
-#ifdef GLX_USE_APPLEGL
-         
-         /*
-          * At this point we should check if the context has been
-          * through glXDestroyContext, and redestroy it if so.
-          */
-         if(oldGC->do_destroy) {
-            __glXUnlock();
-            /* glXDestroyContext uses the same global lock. */
-            glXDestroyContext(dpy, oldGC);
-            __glXLock();
-#else
-         if (oldGC->xid == None) {
-            /* We are switching away from a context that was
-             * previously destroyed, so we need to free the memory
-             * for the old handle.
-             */
-#if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
-            /* Destroy the old direct rendering context */
-            if (oldGC->driContext) {
-               oldGC->driContext->destroyContext(oldGC);
-               oldGC->driContext = NULL;
-            }
-#endif
-            __glXFreeContext(oldGC);
-#endif /* GLX_USE_APPLEGL */
-         }
-      }
-      if (gc) {
-         __glXSetCurrentContext(gc);
-
-         gc->currentDpy = dpy;
-         gc->currentDrawable = draw;
-         gc->currentReadable = read;
-#ifndef GLX_USE_APPLEGL
-         gc->thread_id = _glthread_GetID();
 
-#if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
-         if (!gc->driContext) {
-#endif
-            if (!IndirectAPI)
-               IndirectAPI = __glXNewIndirectAPI();
-            _glapi_set_dispatch(IndirectAPI);
-
-            state = (__GLXattribute *) (gc->client_state_private);
-
-            gc->currentContextTag = reply.contextTag;
-            if (state->array_state == NULL) {
-               (void) glGetString(GL_EXTENSIONS);
-               (void) glGetString(GL_VERSION);
-               __glXInitVertexArrayState(gc);
-            }
-#if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
-         }
-         else {
-            gc->currentContextTag = -1;
-         }
-#endif
-#endif /* GLX_USE_APPLEGL */
-      }
-      else {
-         __glXSetCurrentContextNull();
-      }
-   }
-   __glXUnlock();
    return GL_TRUE;
 }
 
 
-PUBLIC Bool
+_X_EXPORT Bool
 glXMakeCurrent(Display * dpy, GLXDrawable draw, GLXContext gc)
 {
    return MakeContextCurrent(dpy, draw, draw, gc);
 }
 
-PUBLIC
+_X_EXPORT
 GLX_ALIAS(Bool, glXMakeCurrentReadSGI,
           (Display * dpy, GLXDrawable d, GLXDrawable r, GLXContext ctx),
           (dpy, d, r, ctx), MakeContextCurrent)
 
-PUBLIC
+_X_EXPORT
 GLX_ALIAS(Bool, glXMakeContextCurrent,
           (Display * dpy, GLXDrawable d, GLXDrawable r,
            GLXContext ctx), (dpy, d, r, ctx), MakeContextCurrent)
diff --git a/src/glx/glxext.c b/src/glx/glxext.c
index 88e74c2a386..9e42d83c4d5 100644
--- a/src/glx/glxext.c
+++ b/src/glx/glxext.c
@@ -47,7 +47,6 @@
 #include "apple_visual.h"
 #endif
 #include "glxextensions.h"
-#include "glcontextmodes.h"
 
 #ifdef USE_XCB
 #include <X11/Xlib-xcb.h>
@@ -57,7 +56,7 @@
 
 
 #ifdef DEBUG
-void __glXDumpDrawBuffer(__GLXcontext * ctx);
+void __glXDumpDrawBuffer(struct glx_context * ctx);
 #endif
 
 /*
@@ -69,7 +68,7 @@ _X_HIDDEN int __glXDebug = 0;
 /* Extension required boiler plate */
 
 static const char __glXExtensionName[] = GLX_EXTENSION_NAME;
-static __GLXdisplayPrivate *glx_displays;
+  static struct glx_display *glx_displays;
 
 static /* const */ char *error_list[] = {
    "GLXBadContext",
@@ -96,13 +95,6 @@ static
 XEXT_GENERATE_ERROR_STRING(__glXErrorString, __glXExtensionName,
                            __GLX_NUMBER_ERRORS, error_list)
 
-static int
-__glXCloseDisplay(Display * dpy, XExtCodes * codes);
-static Bool
-__glXWireToEvent(Display *dpy, XEvent *event, xEvent *wire);
-static Status
-__glXEventToWire(Display *dpy, XEvent *event, xEvent *wire);
-
 /*
  * GLX events are a bit funky.  We don't stuff the X event code into
  * our user exposed (via XNextEvent) structure.  Instead we use the GLX
@@ -115,7 +107,7 @@ __glXEventToWire(Display *dpy, XEvent *event, xEvent *wire);
 static Bool
 __glXWireToEvent(Display *dpy, XEvent *event, xEvent *wire)
 {
-   __GLXdisplayPrivate *glx_dpy = __glXInitialize(dpy);
+     struct glx_display *glx_dpy = __glXInitialize(dpy);
 
    if (glx_dpy == NULL)
       return False;
@@ -175,7 +167,7 @@ __glXWireToEvent(Display *dpy, XEvent *event, xEvent *wire)
 static Status
 __glXEventToWire(Display *dpy, XEvent *event, xEvent *wire)
 {
-   __GLXdisplayPrivate *glx_dpy = __glXInitialize(dpy);
+     struct glx_display *glx_dpy = __glXInitialize(dpy);
 
    if (glx_dpy == NULL)
       return False;
@@ -205,31 +197,30 @@ __glXEventToWire(Display *dpy, XEvent *event, xEvent *wire)
 ** __glXScreenConfigs.
 */
 static void
-FreeScreenConfigs(__GLXdisplayPrivate * priv)
+FreeScreenConfigs(struct glx_display * priv)
 {
-   __GLXscreenConfigs *psc;
+   struct glx_screen *psc;
    GLint i, screens;
 
    /* Free screen configuration information */
    screens = ScreenCount(priv->dpy);
    for (i = 0; i < screens; i++) {
-      psc = priv->screenConfigs[i];
+      psc = priv->screens[i];
       if (psc->configs) {
-         _gl_context_modes_destroy(psc->configs);
+	 glx_config_destroy_list(psc->configs);
          if (psc->effectiveGLXexts)
             Xfree(psc->effectiveGLXexts);
          psc->configs = NULL;   /* NOTE: just for paranoia */
       }
       if (psc->visuals) {
-         _gl_context_modes_destroy(psc->visuals);
-         psc->visuals = NULL;   /* NOTE: just for paranoia */
+	 glx_config_destroy_list(psc->visuals);
+	 psc->visuals = NULL;   /* NOTE: just for paranoia */
       }
       Xfree((char *) psc->serverGLXexts);
 
 #if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
       if (psc->driScreen) {
          psc->driScreen->destroyScreen(psc);
-         psc->driScreen = NULL;
       } else {
 	 Xfree(psc);
       }
@@ -237,34 +228,19 @@ FreeScreenConfigs(__GLXdisplayPrivate * priv)
       Xfree(psc);
 #endif
    }
-   XFree((char *) priv->screenConfigs);
-   priv->screenConfigs = NULL;
+   XFree((char *) priv->screens);
+   priv->screens = NULL;
 }
 
-/*
-** Release the private memory referred to in a display private
-** structure.  The caller will free the extension structure.
-*/
-static int
-__glXCloseDisplay(Display * dpy, XExtCodes * codes)
+static void
+glx_display_free(struct glx_display *priv)
 {
-   __GLXdisplayPrivate *priv, **prev;
-   GLXContext gc;
-
-   _XLockMutex(_Xglobal_lock);
-   prev = &glx_displays;
-   for (priv = glx_displays; priv; prev = &priv->next, priv = priv->next) {
-      if (priv->dpy == dpy) {
-	 (*prev)->next = priv->next;
-	 break;
-      }
-   }
-   _XUnlockMutex(_Xglobal_lock);
+   struct glx_context *gc;
 
    gc = __glXGetCurrentContext();
-   if (dpy == gc->currentDpy) {
+   if (priv->dpy == gc->currentDpy) {
+      gc->vtable->destroy(gc);
       __glXSetCurrentContextNull();
-      __glXFreeContext(gc);
    }
 
    FreeScreenConfigs(priv);
@@ -291,6 +267,24 @@ __glXCloseDisplay(Display * dpy, XExtCodes * codes)
 #endif
 
    Xfree((char *) priv);
+}
+
+static int
+__glXCloseDisplay(Display * dpy, XExtCodes * codes)
+{
+   struct glx_display *priv, **prev;
+
+   _XLockMutex(_Xglobal_lock);
+   prev = &glx_displays;
+   for (priv = glx_displays; priv; prev = &priv->next, priv = priv->next) {
+      if (priv->dpy == dpy) {
+	 (*prev) = priv->next;
+	 break;
+      }
+   }
+   _XUnlockMutex(_Xglobal_lock);
+
+   glx_display_free(priv);
 
    return 1;
 }
@@ -357,12 +351,27 @@ enum {
 };
 
 
+static GLint
+convert_from_x_visual_type(int visualType)
+{
+   static const int glx_visual_types[] = {
+      GLX_STATIC_GRAY, GLX_GRAY_SCALE,
+      GLX_STATIC_COLOR, GLX_PSEUDO_COLOR,
+      GLX_TRUE_COLOR, GLX_DIRECT_COLOR
+   };
+
+   if (visualType < ARRAY_SIZE(glx_visual_types))
+      return glx_visual_types[visualType];
+
+   return GLX_NONE;
+}
+
 /*
  * getVisualConfigs uses the !tagged_only path.
  * getFBConfigs uses the tagged_only path.
  */
 _X_HIDDEN void
-__glXInitializeVisualConfigFromTags(__GLcontextModes * config, int count,
+__glXInitializeVisualConfigFromTags(struct glx_config * config, int count,
                                     const INT32 * bp, Bool tagged_only,
                                     Bool fbconfig_style_tags)
 {
@@ -372,7 +381,7 @@ __glXInitializeVisualConfigFromTags(__GLcontextModes * config, int count,
       /* Copy in the first set of properties */
       config->visualID = *bp++;
 
-      config->visualType = _gl_convert_from_x_visual_type(*bp++);
+      config->visualType = convert_from_x_visual_type(*bp++);
 
       config->rgbMode = *bp++;
 
@@ -581,22 +590,15 @@ __glXInitializeVisualConfigFromTags(__GLcontextModes * config, int count,
 
    config->renderType =
       (config->rgbMode) ? GLX_RGBA_BIT : GLX_COLOR_INDEX_BIT;
-
-   config->haveAccumBuffer = ((config->accumRedBits +
-                               config->accumGreenBits +
-                               config->accumBlueBits +
-                               config->accumAlphaBits) > 0);
-   config->haveDepthBuffer = (config->depthBits > 0);
-   config->haveStencilBuffer = (config->stencilBits > 0);
 }
 
-static __GLcontextModes *
+static struct glx_config *
 createConfigsFromProperties(Display * dpy, int nvisuals, int nprops,
                             int screen, GLboolean tagged_only)
 {
    INT32 buf[__GLX_TOTAL_CONFIG], *props;
    unsigned prop_size;
-   __GLcontextModes *modes, *m;
+   struct glx_config *modes, *m;
    int i;
 
    if (nprops == 0)
@@ -609,7 +611,7 @@ createConfigsFromProperties(Display * dpy, int nvisuals, int nprops,
       return NULL;
 
    /* Allocate memory for our config structure */
-   modes = _gl_context_modes_create(nvisuals, sizeof(__GLcontextModes));
+   modes = glx_config_create_list(nvisuals);
    if (!modes)
       return NULL;
 
@@ -647,8 +649,8 @@ createConfigsFromProperties(Display * dpy, int nvisuals, int nprops,
 }
 
 static GLboolean
-getVisualConfigs(__GLXscreenConfigs *psc,
-		 __GLXdisplayPrivate *priv, int screen)
+getVisualConfigs(struct glx_screen *psc,
+		  struct glx_display *priv, int screen)
 {
    xGLXGetVisualConfigsReq *req;
    xGLXGetVisualConfigsReply reply;
@@ -676,7 +678,7 @@ getVisualConfigs(__GLXscreenConfigs *psc,
 }
 
 static GLboolean
-getFBConfigs(__GLXscreenConfigs *psc, __GLXdisplayPrivate *priv, int screen)
+ getFBConfigs(struct glx_screen *psc, struct glx_display *priv, int screen)
 {
    xGLXGetFBConfigsReq *fb_req;
    xGLXGetFBConfigsSGIXReq *sgi_req;
@@ -723,8 +725,8 @@ getFBConfigs(__GLXscreenConfigs *psc, __GLXdisplayPrivate *priv, int screen)
 }
 
 _X_HIDDEN Bool
-glx_screen_init(__GLXscreenConfigs *psc,
-		int screen, __GLXdisplayPrivate * priv)
+glx_screen_init(struct glx_screen *psc,
+		 int screen, struct glx_display * priv)
 {
    /* Initialize per screen dynamic client GLX extensions */
    psc->ext_list_first_time = GL_TRUE;
@@ -738,33 +740,22 @@ glx_screen_init(__GLXscreenConfigs *psc,
    return GL_TRUE;
 }
 
-static __GLXscreenConfigs *
-createIndirectScreen()
-{
-   __GLXscreenConfigs *psc;
-
-   psc = Xmalloc(sizeof *psc);
-   memset(psc, 0, sizeof *psc);
-
-   return psc;
-}
-
 /*
 ** Allocate the memory for the per screen configs for each screen.
 ** If that works then fetch the per screen configs data.
 */
 static Bool
-AllocAndFetchScreenConfigs(Display * dpy, __GLXdisplayPrivate * priv)
+AllocAndFetchScreenConfigs(Display * dpy, struct glx_display * priv)
 {
-   __GLXscreenConfigs *psc;
+   struct glx_screen *psc;
    GLint i, screens;
 
    /*
     ** First allocate memory for the array of per screen configs.
     */
    screens = ScreenCount(dpy);
-   priv->screenConfigs = Xmalloc(screens * sizeof *priv->screenConfigs);
-   if (!priv->screenConfigs)
+   priv->screens = Xmalloc(screens * sizeof *priv->screens);
+   if (!priv->screens)
       return GL_FALSE;
 
    priv->serverGLXversion =
@@ -775,6 +766,7 @@ AllocAndFetchScreenConfigs(Display * dpy, __GLXdisplayPrivate * priv)
    }
 
    for (i = 0; i < screens; i++, psc++) {
+      psc = NULL;
 #if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
       if (priv->dri2Display)
 	 psc = (*priv->dri2Display->createScreen) (i, priv);
@@ -782,10 +774,14 @@ AllocAndFetchScreenConfigs(Display * dpy, __GLXdisplayPrivate * priv)
 	 psc = (*priv->driDisplay->createScreen) (i, priv);
       if (psc == NULL && priv->driswDisplay)
 	 psc = (*priv->driswDisplay->createScreen) (i, priv);
-      if (psc == NULL)
-	 psc = createIndirectScreen (i, priv);
 #endif
-      priv->screenConfigs[i] = psc;
+#if defined(GLX_USE_APPLEGL)
+      if (psc == NULL && priv->appleglDisplay)
+	 psc = (*priv->appleglDisplay->createScreen) (i, priv);
+#endif
+      if (psc == NULL)
+	 psc = indirect_create_screen(i, priv);
+      priv->screens[i] = psc;
    }
    SyncHandle();
    return GL_TRUE;
@@ -794,10 +790,10 @@ AllocAndFetchScreenConfigs(Display * dpy, __GLXdisplayPrivate * priv)
 /*
 ** Initialize the client side extension code.
 */
-_X_HIDDEN __GLXdisplayPrivate *
+ _X_HIDDEN struct glx_display *
 __glXInitialize(Display * dpy)
 {
-   __GLXdisplayPrivate *dpyPriv;
+   struct glx_display *dpyPriv, *d;
 #if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
    Bool glx_direct, glx_accel;
 #endif
@@ -812,6 +808,9 @@ __glXInitialize(Display * dpy)
       }
    }
 
+   /* Drop the lock while we create the display private. */
+   _XUnlockMutex(_Xglobal_lock);
+
    dpyPriv = Xcalloc(1, sizeof *dpyPriv);
    if (!dpyPriv)
       return NULL;
@@ -864,7 +863,7 @@ __glXInitialize(Display * dpy)
 #endif
 
 #ifdef GLX_USE_APPLEGL
-   if (apple_init_glx(dpy)) {
+   if (!applegl_create_display(dpyPriv)) {
       Xfree(dpyPriv);
       return NULL;
    }
@@ -877,6 +876,18 @@ __glXInitialize(Display * dpy)
    if (dpyPriv->majorVersion == 1 && dpyPriv->minorVersion >= 1)
       __glXClientInfo(dpy, dpyPriv->majorOpcode);
 
+   /* Grab the lock again and add the dispay private, unless somebody
+    * beat us to initializing on this display in the meantime. */
+   _XLockMutex(_Xglobal_lock);
+
+   for (d = glx_displays; d; d = d->next) {
+      if (d->dpy == dpy) {
+	 _XUnlockMutex(_Xglobal_lock);
+	 glx_display_free(dpyPriv);
+	 return d;
+      }
+   }
+
    dpyPriv->next = glx_displays;
    glx_displays = dpyPriv;
 
@@ -892,8 +903,8 @@ __glXInitialize(Display * dpy)
 _X_HIDDEN CARD8
 __glXSetupForCommand(Display * dpy)
 {
-   GLXContext gc;
-   __GLXdisplayPrivate *priv;
+    struct glx_context *gc;
+    struct glx_display *priv;
 
    /* If this thread has a current context, flush its rendering commands */
    gc = __glXGetCurrentContext();
@@ -932,7 +943,7 @@ __glXSetupForCommand(Display * dpy)
  * \c pc parameter.
  */
 _X_HIDDEN GLubyte *
-__glXFlushRenderBuffer(__GLXcontext * ctx, GLubyte * pc)
+__glXFlushRenderBuffer(struct glx_context * ctx, GLubyte * pc)
 {
    Display *const dpy = ctx->currentDpy;
 #ifdef USE_XCB
@@ -983,7 +994,7 @@ __glXFlushRenderBuffer(__GLXcontext * ctx, GLubyte * pc)
  * \param dataLen        Size, in bytes, of the command data.
  */
 _X_HIDDEN void
-__glXSendLargeChunk(__GLXcontext * gc, GLint requestNumber,
+__glXSendLargeChunk(struct glx_context * gc, GLint requestNumber,
                     GLint totalRequests, const GLvoid * data, GLint dataLen)
 {
    Display *dpy = gc->currentDpy;
@@ -1032,7 +1043,7 @@ __glXSendLargeChunk(__GLXcontext * gc, GLint requestNumber,
  * \param dataLen    Size, in bytes, of the command data.
  */
 _X_HIDDEN void
-__glXSendLargeCommand(__GLXcontext * ctx,
+__glXSendLargeCommand(struct glx_context * ctx,
                       const GLvoid * header, GLint headerLen,
                       const GLvoid * data, GLint dataLen)
 {
@@ -1074,7 +1085,7 @@ __glXSendLargeCommand(__GLXcontext * ctx,
 
 #ifdef DEBUG
 _X_HIDDEN void
-__glXDumpDrawBuffer(__GLXcontext * ctx)
+__glXDumpDrawBuffer(struct glx_context * ctx)
 {
    GLubyte *p = ctx->buf;
    GLubyte *end = ctx->pc;
diff --git a/src/glx/glxextensions.c b/src/glx/glxextensions.c
index 4eb6a5536ab..23161ef49a1 100644
--- a/src/glx/glxextensions.c
+++ b/src/glx/glxextensions.c
@@ -312,7 +312,7 @@ static const unsigned gl_minor = 4;
 static const char *__glXGLXClientExtensions = NULL;
 
 static void __glXExtensionsCtr(void);
-static void __glXExtensionsCtrScreen(__GLXscreenConfigs * psc);
+static void __glXExtensionsCtrScreen(struct glx_screen * psc);
 static void __glXProcessServerString(const struct extension_info *ext,
                                      const char *server_string,
                                      unsigned char *server_support);
@@ -396,7 +396,7 @@ __glXProcessServerString(const struct extension_info *ext,
 }
 
 void
-__glXEnableDirectExtension(__GLXscreenConfigs * psc, const char *name)
+__glXEnableDirectExtension(struct glx_screen * psc, const char *name)
 {
    __glXExtensionsCtr();
    __glXExtensionsCtrScreen(psc);
@@ -474,7 +474,7 @@ __glXExtensionsCtr(void)
  */
 
 static void
-__glXExtensionsCtrScreen(__GLXscreenConfigs * psc)
+__glXExtensionsCtrScreen(struct glx_screen * psc)
 {
    if (psc->ext_list_first_time) {
       psc->ext_list_first_time = GL_FALSE;
@@ -494,7 +494,7 @@ __glXExtensionsCtrScreen(__GLXscreenConfigs * psc)
  *          \c NULL, then \c GL_FALSE is returned.
  */
 GLboolean
-__glXExtensionBitIsEnabled(__GLXscreenConfigs * psc, unsigned bit)
+__glXExtensionBitIsEnabled(struct glx_screen * psc, unsigned bit)
 {
    GLboolean enabled = GL_FALSE;
 
@@ -513,7 +513,7 @@ __glXExtensionBitIsEnabled(__GLXscreenConfigs * psc, unsigned bit)
  *
  */
 GLboolean
-__glExtensionBitIsEnabled(const __GLXcontext * gc, unsigned bit)
+__glExtensionBitIsEnabled(struct glx_context *gc, unsigned bit)
 {
    GLboolean enabled = GL_FALSE;
 
@@ -594,7 +594,7 @@ __glXGetClientExtensions(void)
  */
 
 void
-__glXCalculateUsableExtensions(__GLXscreenConfigs * psc,
+__glXCalculateUsableExtensions(struct glx_screen * psc,
                                GLboolean display_is_direct_capable,
                                int minor_version)
 {
@@ -675,7 +675,7 @@ __glXCalculateUsableExtensions(__GLXscreenConfigs * psc,
  */
 
 void
-__glXCalculateUsableGLExtensions(__GLXcontext * gc,
+__glXCalculateUsableGLExtensions(struct glx_context * gc,
                                  const char *server_string,
                                  int major_version, int minor_version)
 {
diff --git a/src/glx/glxextensions.h b/src/glx/glxextensions.h
index 4f1b6619d65..a11fe88ffc9 100644
--- a/src/glx/glxextensions.h
+++ b/src/glx/glxextensions.h
@@ -234,30 +234,29 @@ enum
 
 #define __GL_EXT_BYTES   ((__NUM_GL_EXTS + 7) / 8)
 
-struct __GLXscreenConfigsRec;
-struct __GLXcontextRec;
+struct glx_screen;
+struct glx_context;
 
-extern GLboolean __glXExtensionBitIsEnabled(struct __GLXscreenConfigsRec *psc,
+extern GLboolean __glXExtensionBitIsEnabled(struct glx_screen *psc,
                                             unsigned bit);
 extern const char *__glXGetClientExtensions(void);
-extern void __glXCalculateUsableExtensions(struct __GLXscreenConfigsRec *psc,
+extern void __glXCalculateUsableExtensions(struct glx_screen *psc,
                                            GLboolean
                                            display_is_direct_capable,
                                            int server_minor_version);
 
-extern void __glXCalculateUsableGLExtensions(struct __GLXcontextRec *gc,
+extern void __glXCalculateUsableGLExtensions(struct glx_context *gc,
                                              const char *server_string,
                                              int major_version,
                                              int minor_version);
 extern void __glXGetGLVersion(int *major_version, int *minor_version);
 extern char *__glXGetClientGLExtensionString(void);
 
-extern GLboolean __glExtensionBitIsEnabled(const struct __GLXcontextRec *gc,
+extern GLboolean __glExtensionBitIsEnabled(struct glx_context *gc,
                                            unsigned bit);
 
 extern void
-__glXEnableDirectExtension(struct __GLXscreenConfigsRec *psc,
-                           const char *name);
+__glXEnableDirectExtension(struct glx_screen *psc, const char *name);
 
 /* Source-level backwards compatibility with old drivers. They won't
  * find the respective functions, though. 
diff --git a/src/glx/indirect.c b/src/glx/indirect.c
index 172727860e8..c0fff6c4dcf 100644
--- a/src/glx/indirect.c
+++ b/src/glx/indirect.c
@@ -91,7 +91,7 @@ __glXReadReply(Display * dpy, size_t size, void *dest,
 }
 
 NOINLINE void
-__glXReadPixelReply(Display * dpy, __GLXcontext * gc, unsigned max_dim,
+__glXReadPixelReply(Display * dpy, struct glx_context * gc, unsigned max_dim,
                     GLint width, GLint height, GLint depth, GLenum format,
                     GLenum type, void *dest, GLboolean dimensions_in_reply)
 {
@@ -138,7 +138,7 @@ __glXReadPixelReply(Display * dpy, __GLXcontext * gc, unsigned max_dim,
 #define X_GLXSingle 0
 
 NOINLINE FASTCALL GLubyte *
-__glXSetupSingleRequest(__GLXcontext * gc, GLint sop, GLint cmdlen)
+__glXSetupSingleRequest(struct glx_context * gc, GLint sop, GLint cmdlen)
 {
     xGLXSingleReq *req;
     Display *const dpy = gc->currentDpy;
@@ -153,7 +153,7 @@ __glXSetupSingleRequest(__GLXcontext * gc, GLint sop, GLint cmdlen)
 }
 
 NOINLINE FASTCALL GLubyte *
-__glXSetupVendorRequest(__GLXcontext * gc, GLint code, GLint vop,
+__glXSetupVendorRequest(struct glx_context * gc, GLint code, GLint vop,
                         GLint cmdlen)
 {
     xGLXVendorPrivateReq *req;
@@ -185,7 +185,7 @@ const GLuint __glXDefaultPixelStore[9] = { 0, 0, 0, 0, 0, 0, 0, 0, 1 };
 static FASTCALL NOINLINE void
 generic_3_byte(GLint rop, const void *ptr)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
 
     emit_header(gc->pc, rop, cmdlen);
@@ -199,7 +199,7 @@ generic_3_byte(GLint rop, const void *ptr)
 static FASTCALL NOINLINE void
 generic_4_byte(GLint rop, const void *ptr)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
 
     emit_header(gc->pc, rop, cmdlen);
@@ -213,7 +213,7 @@ generic_4_byte(GLint rop, const void *ptr)
 static FASTCALL NOINLINE void
 generic_6_byte(GLint rop, const void *ptr)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
 
     emit_header(gc->pc, rop, cmdlen);
@@ -227,7 +227,7 @@ generic_6_byte(GLint rop, const void *ptr)
 static FASTCALL NOINLINE void
 generic_8_byte(GLint rop, const void *ptr)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
 
     emit_header(gc->pc, rop, cmdlen);
@@ -241,7 +241,7 @@ generic_8_byte(GLint rop, const void *ptr)
 static FASTCALL NOINLINE void
 generic_12_byte(GLint rop, const void *ptr)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
 
     emit_header(gc->pc, rop, cmdlen);
@@ -255,7 +255,7 @@ generic_12_byte(GLint rop, const void *ptr)
 static FASTCALL NOINLINE void
 generic_16_byte(GLint rop, const void *ptr)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
 
     emit_header(gc->pc, rop, cmdlen);
@@ -269,7 +269,7 @@ generic_16_byte(GLint rop, const void *ptr)
 static FASTCALL NOINLINE void
 generic_24_byte(GLint rop, const void *ptr)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 28;
 
     emit_header(gc->pc, rop, cmdlen);
@@ -283,7 +283,7 @@ generic_24_byte(GLint rop, const void *ptr)
 static FASTCALL NOINLINE void
 generic_32_byte(GLint rop, const void *ptr)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 36;
 
     emit_header(gc->pc, rop, cmdlen);
@@ -298,7 +298,7 @@ generic_32_byte(GLint rop, const void *ptr)
 void
 __indirect_glNewList(GLuint list, GLenum mode)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -324,7 +324,7 @@ __indirect_glNewList(GLuint list, GLenum mode)
 void
 __indirect_glEndList(void)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 0;
@@ -347,7 +347,7 @@ __indirect_glEndList(void)
 void
 __indirect_glCallList(GLuint list)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_CallList, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&list), 4);
@@ -361,7 +361,7 @@ __indirect_glCallList(GLuint list)
 void
 __indirect_glCallLists(GLsizei n, GLenum type, const GLvoid * lists)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize = __glCallLists_size(type);
     const GLuint cmdlen = 12 + __GLX_PAD((compsize * n));
     if (n < 0) {
@@ -399,7 +399,7 @@ __indirect_glCallLists(GLsizei n, GLenum type, const GLvoid * lists)
 void
 __indirect_glDeleteLists(GLuint list, GLsizei range)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -425,7 +425,7 @@ __indirect_glDeleteLists(GLuint list, GLsizei range)
 GLuint
 __indirect_glGenLists(GLsizei range)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     GLuint retval = (GLuint) 0;
 #ifndef USE_XCB
@@ -458,7 +458,7 @@ __indirect_glGenLists(GLsizei range)
 void
 __indirect_glListBase(GLuint base)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_ListBase, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&base), 4);
@@ -472,7 +472,7 @@ __indirect_glListBase(GLuint base)
 void
 __indirect_glBegin(GLenum mode)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_Begin, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&mode), 4);
@@ -488,7 +488,7 @@ __indirect_glBitmap(GLsizei width, GLsizei height, GLfloat xorig,
                     GLfloat yorig, GLfloat xmove, GLfloat ymove,
                     const GLubyte *bitmap)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize =
         (bitmap != NULL) ? __glImageSize(width, height, 1, GL_COLOR_INDEX,
                                          GL_BITMAP, 0) : 0;
@@ -539,7 +539,7 @@ __indirect_glBitmap(GLsizei width, GLsizei height, GLfloat xorig,
 void
 __indirect_glColor3b(GLbyte red, GLbyte green, GLbyte blue)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_Color3bv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 1);
@@ -562,7 +562,7 @@ __indirect_glColor3bv(const GLbyte *v)
 void
 __indirect_glColor3d(GLdouble red, GLdouble green, GLdouble blue)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 28;
     emit_header(gc->pc, X_GLrop_Color3dv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 8);
@@ -585,7 +585,7 @@ __indirect_glColor3dv(const GLdouble * v)
 void
 __indirect_glColor3f(GLfloat red, GLfloat green, GLfloat blue)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_Color3fv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 4);
@@ -608,7 +608,7 @@ __indirect_glColor3fv(const GLfloat * v)
 void
 __indirect_glColor3i(GLint red, GLint green, GLint blue)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_Color3iv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 4);
@@ -631,7 +631,7 @@ __indirect_glColor3iv(const GLint * v)
 void
 __indirect_glColor3s(GLshort red, GLshort green, GLshort blue)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_Color3sv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 2);
@@ -654,7 +654,7 @@ __indirect_glColor3sv(const GLshort * v)
 void
 __indirect_glColor3ub(GLubyte red, GLubyte green, GLubyte blue)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_Color3ubv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 1);
@@ -677,7 +677,7 @@ __indirect_glColor3ubv(const GLubyte *v)
 void
 __indirect_glColor3ui(GLuint red, GLuint green, GLuint blue)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_Color3uiv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 4);
@@ -700,7 +700,7 @@ __indirect_glColor3uiv(const GLuint * v)
 void
 __indirect_glColor3us(GLushort red, GLushort green, GLushort blue)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_Color3usv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 2);
@@ -723,7 +723,7 @@ __indirect_glColor3usv(const GLushort * v)
 void
 __indirect_glColor4b(GLbyte red, GLbyte green, GLbyte blue, GLbyte alpha)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_Color4bv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 1);
@@ -748,7 +748,7 @@ void
 __indirect_glColor4d(GLdouble red, GLdouble green, GLdouble blue,
                      GLdouble alpha)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 36;
     emit_header(gc->pc, X_GLrop_Color4dv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 8);
@@ -772,7 +772,7 @@ __indirect_glColor4dv(const GLdouble * v)
 void
 __indirect_glColor4f(GLfloat red, GLfloat green, GLfloat blue, GLfloat alpha)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_Color4fv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 4);
@@ -796,7 +796,7 @@ __indirect_glColor4fv(const GLfloat * v)
 void
 __indirect_glColor4i(GLint red, GLint green, GLint blue, GLint alpha)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_Color4iv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 4);
@@ -820,7 +820,7 @@ __indirect_glColor4iv(const GLint * v)
 void
 __indirect_glColor4s(GLshort red, GLshort green, GLshort blue, GLshort alpha)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_Color4sv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 2);
@@ -844,7 +844,7 @@ __indirect_glColor4sv(const GLshort * v)
 void
 __indirect_glColor4ub(GLubyte red, GLubyte green, GLubyte blue, GLubyte alpha)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_Color4ubv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 1);
@@ -868,7 +868,7 @@ __indirect_glColor4ubv(const GLubyte *v)
 void
 __indirect_glColor4ui(GLuint red, GLuint green, GLuint blue, GLuint alpha)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_Color4uiv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 4);
@@ -893,7 +893,7 @@ void
 __indirect_glColor4us(GLushort red, GLushort green, GLushort blue,
                       GLushort alpha)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_Color4usv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 2);
@@ -917,7 +917,7 @@ __indirect_glColor4usv(const GLushort * v)
 void
 __indirect_glEdgeFlag(GLboolean flag)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_EdgeFlagv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&flag), 1);
@@ -931,7 +931,7 @@ __indirect_glEdgeFlag(GLboolean flag)
 void
 __indirect_glEdgeFlagv(const GLboolean * flag)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_EdgeFlagv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (flag), 1);
@@ -945,7 +945,7 @@ __indirect_glEdgeFlagv(const GLboolean * flag)
 void
 __indirect_glEnd(void)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 4;
     emit_header(gc->pc, X_GLrop_End, cmdlen);
     gc->pc += cmdlen;
@@ -958,7 +958,7 @@ __indirect_glEnd(void)
 void
 __indirect_glIndexd(GLdouble c)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_Indexdv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&c), 8);
@@ -979,7 +979,7 @@ __indirect_glIndexdv(const GLdouble * c)
 void
 __indirect_glIndexf(GLfloat c)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_Indexfv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&c), 4);
@@ -1000,7 +1000,7 @@ __indirect_glIndexfv(const GLfloat * c)
 void
 __indirect_glIndexi(GLint c)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_Indexiv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&c), 4);
@@ -1021,7 +1021,7 @@ __indirect_glIndexiv(const GLint * c)
 void
 __indirect_glIndexs(GLshort c)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_Indexsv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&c), 2);
@@ -1035,7 +1035,7 @@ __indirect_glIndexs(GLshort c)
 void
 __indirect_glIndexsv(const GLshort * c)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_Indexsv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (c), 2);
@@ -1049,7 +1049,7 @@ __indirect_glIndexsv(const GLshort * c)
 void
 __indirect_glNormal3b(GLbyte nx, GLbyte ny, GLbyte nz)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_Normal3bv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&nx), 1);
@@ -1072,7 +1072,7 @@ __indirect_glNormal3bv(const GLbyte *v)
 void
 __indirect_glNormal3d(GLdouble nx, GLdouble ny, GLdouble nz)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 28;
     emit_header(gc->pc, X_GLrop_Normal3dv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&nx), 8);
@@ -1095,7 +1095,7 @@ __indirect_glNormal3dv(const GLdouble * v)
 void
 __indirect_glNormal3f(GLfloat nx, GLfloat ny, GLfloat nz)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_Normal3fv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&nx), 4);
@@ -1118,7 +1118,7 @@ __indirect_glNormal3fv(const GLfloat * v)
 void
 __indirect_glNormal3i(GLint nx, GLint ny, GLint nz)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_Normal3iv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&nx), 4);
@@ -1141,7 +1141,7 @@ __indirect_glNormal3iv(const GLint * v)
 void
 __indirect_glNormal3s(GLshort nx, GLshort ny, GLshort nz)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_Normal3sv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&nx), 2);
@@ -1164,7 +1164,7 @@ __indirect_glNormal3sv(const GLshort * v)
 void
 __indirect_glRasterPos2d(GLdouble x, GLdouble y)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_RasterPos2dv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 8);
@@ -1186,7 +1186,7 @@ __indirect_glRasterPos2dv(const GLdouble * v)
 void
 __indirect_glRasterPos2f(GLfloat x, GLfloat y)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_RasterPos2fv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 4);
@@ -1208,7 +1208,7 @@ __indirect_glRasterPos2fv(const GLfloat * v)
 void
 __indirect_glRasterPos2i(GLint x, GLint y)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_RasterPos2iv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 4);
@@ -1230,7 +1230,7 @@ __indirect_glRasterPos2iv(const GLint * v)
 void
 __indirect_glRasterPos2s(GLshort x, GLshort y)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_RasterPos2sv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 2);
@@ -1252,7 +1252,7 @@ __indirect_glRasterPos2sv(const GLshort * v)
 void
 __indirect_glRasterPos3d(GLdouble x, GLdouble y, GLdouble z)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 28;
     emit_header(gc->pc, X_GLrop_RasterPos3dv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 8);
@@ -1275,7 +1275,7 @@ __indirect_glRasterPos3dv(const GLdouble * v)
 void
 __indirect_glRasterPos3f(GLfloat x, GLfloat y, GLfloat z)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_RasterPos3fv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 4);
@@ -1298,7 +1298,7 @@ __indirect_glRasterPos3fv(const GLfloat * v)
 void
 __indirect_glRasterPos3i(GLint x, GLint y, GLint z)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_RasterPos3iv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 4);
@@ -1321,7 +1321,7 @@ __indirect_glRasterPos3iv(const GLint * v)
 void
 __indirect_glRasterPos3s(GLshort x, GLshort y, GLshort z)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_RasterPos3sv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 2);
@@ -1344,7 +1344,7 @@ __indirect_glRasterPos3sv(const GLshort * v)
 void
 __indirect_glRasterPos4d(GLdouble x, GLdouble y, GLdouble z, GLdouble w)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 36;
     emit_header(gc->pc, X_GLrop_RasterPos4dv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 8);
@@ -1368,7 +1368,7 @@ __indirect_glRasterPos4dv(const GLdouble * v)
 void
 __indirect_glRasterPos4f(GLfloat x, GLfloat y, GLfloat z, GLfloat w)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_RasterPos4fv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 4);
@@ -1392,7 +1392,7 @@ __indirect_glRasterPos4fv(const GLfloat * v)
 void
 __indirect_glRasterPos4i(GLint x, GLint y, GLint z, GLint w)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_RasterPos4iv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 4);
@@ -1416,7 +1416,7 @@ __indirect_glRasterPos4iv(const GLint * v)
 void
 __indirect_glRasterPos4s(GLshort x, GLshort y, GLshort z, GLshort w)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_RasterPos4sv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 2);
@@ -1440,7 +1440,7 @@ __indirect_glRasterPos4sv(const GLshort * v)
 void
 __indirect_glRectd(GLdouble x1, GLdouble y1, GLdouble x2, GLdouble y2)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 36;
     emit_header(gc->pc, X_GLrop_Rectdv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x1), 8);
@@ -1457,7 +1457,7 @@ __indirect_glRectd(GLdouble x1, GLdouble y1, GLdouble x2, GLdouble y2)
 void
 __indirect_glRectdv(const GLdouble * v1, const GLdouble * v2)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 36;
     emit_header(gc->pc, X_GLrop_Rectdv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (v1), 16);
@@ -1472,7 +1472,7 @@ __indirect_glRectdv(const GLdouble * v1, const GLdouble * v2)
 void
 __indirect_glRectf(GLfloat x1, GLfloat y1, GLfloat x2, GLfloat y2)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_Rectfv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x1), 4);
@@ -1489,7 +1489,7 @@ __indirect_glRectf(GLfloat x1, GLfloat y1, GLfloat x2, GLfloat y2)
 void
 __indirect_glRectfv(const GLfloat * v1, const GLfloat * v2)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_Rectfv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (v1), 8);
@@ -1504,7 +1504,7 @@ __indirect_glRectfv(const GLfloat * v1, const GLfloat * v2)
 void
 __indirect_glRecti(GLint x1, GLint y1, GLint x2, GLint y2)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_Rectiv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x1), 4);
@@ -1521,7 +1521,7 @@ __indirect_glRecti(GLint x1, GLint y1, GLint x2, GLint y2)
 void
 __indirect_glRectiv(const GLint * v1, const GLint * v2)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_Rectiv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (v1), 8);
@@ -1536,7 +1536,7 @@ __indirect_glRectiv(const GLint * v1, const GLint * v2)
 void
 __indirect_glRects(GLshort x1, GLshort y1, GLshort x2, GLshort y2)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_Rectsv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x1), 2);
@@ -1553,7 +1553,7 @@ __indirect_glRects(GLshort x1, GLshort y1, GLshort x2, GLshort y2)
 void
 __indirect_glRectsv(const GLshort * v1, const GLshort * v2)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_Rectsv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (v1), 4);
@@ -1568,7 +1568,7 @@ __indirect_glRectsv(const GLshort * v1, const GLshort * v2)
 void
 __indirect_glTexCoord1d(GLdouble s)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_TexCoord1dv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&s), 8);
@@ -1589,7 +1589,7 @@ __indirect_glTexCoord1dv(const GLdouble * v)
 void
 __indirect_glTexCoord1f(GLfloat s)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_TexCoord1fv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&s), 4);
@@ -1610,7 +1610,7 @@ __indirect_glTexCoord1fv(const GLfloat * v)
 void
 __indirect_glTexCoord1i(GLint s)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_TexCoord1iv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&s), 4);
@@ -1631,7 +1631,7 @@ __indirect_glTexCoord1iv(const GLint * v)
 void
 __indirect_glTexCoord1s(GLshort s)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_TexCoord1sv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&s), 2);
@@ -1645,7 +1645,7 @@ __indirect_glTexCoord1s(GLshort s)
 void
 __indirect_glTexCoord1sv(const GLshort * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_TexCoord1sv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (v), 2);
@@ -1659,7 +1659,7 @@ __indirect_glTexCoord1sv(const GLshort * v)
 void
 __indirect_glTexCoord2d(GLdouble s, GLdouble t)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_TexCoord2dv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&s), 8);
@@ -1681,7 +1681,7 @@ __indirect_glTexCoord2dv(const GLdouble * v)
 void
 __indirect_glTexCoord2f(GLfloat s, GLfloat t)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_TexCoord2fv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&s), 4);
@@ -1703,7 +1703,7 @@ __indirect_glTexCoord2fv(const GLfloat * v)
 void
 __indirect_glTexCoord2i(GLint s, GLint t)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_TexCoord2iv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&s), 4);
@@ -1725,7 +1725,7 @@ __indirect_glTexCoord2iv(const GLint * v)
 void
 __indirect_glTexCoord2s(GLshort s, GLshort t)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_TexCoord2sv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&s), 2);
@@ -1747,7 +1747,7 @@ __indirect_glTexCoord2sv(const GLshort * v)
 void
 __indirect_glTexCoord3d(GLdouble s, GLdouble t, GLdouble r)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 28;
     emit_header(gc->pc, X_GLrop_TexCoord3dv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&s), 8);
@@ -1770,7 +1770,7 @@ __indirect_glTexCoord3dv(const GLdouble * v)
 void
 __indirect_glTexCoord3f(GLfloat s, GLfloat t, GLfloat r)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_TexCoord3fv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&s), 4);
@@ -1793,7 +1793,7 @@ __indirect_glTexCoord3fv(const GLfloat * v)
 void
 __indirect_glTexCoord3i(GLint s, GLint t, GLint r)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_TexCoord3iv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&s), 4);
@@ -1816,7 +1816,7 @@ __indirect_glTexCoord3iv(const GLint * v)
 void
 __indirect_glTexCoord3s(GLshort s, GLshort t, GLshort r)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_TexCoord3sv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&s), 2);
@@ -1839,7 +1839,7 @@ __indirect_glTexCoord3sv(const GLshort * v)
 void
 __indirect_glTexCoord4d(GLdouble s, GLdouble t, GLdouble r, GLdouble q)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 36;
     emit_header(gc->pc, X_GLrop_TexCoord4dv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&s), 8);
@@ -1863,7 +1863,7 @@ __indirect_glTexCoord4dv(const GLdouble * v)
 void
 __indirect_glTexCoord4f(GLfloat s, GLfloat t, GLfloat r, GLfloat q)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_TexCoord4fv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&s), 4);
@@ -1887,7 +1887,7 @@ __indirect_glTexCoord4fv(const GLfloat * v)
 void
 __indirect_glTexCoord4i(GLint s, GLint t, GLint r, GLint q)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_TexCoord4iv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&s), 4);
@@ -1911,7 +1911,7 @@ __indirect_glTexCoord4iv(const GLint * v)
 void
 __indirect_glTexCoord4s(GLshort s, GLshort t, GLshort r, GLshort q)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_TexCoord4sv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&s), 2);
@@ -1935,7 +1935,7 @@ __indirect_glTexCoord4sv(const GLshort * v)
 void
 __indirect_glVertex2d(GLdouble x, GLdouble y)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_Vertex2dv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 8);
@@ -1957,7 +1957,7 @@ __indirect_glVertex2dv(const GLdouble * v)
 void
 __indirect_glVertex2f(GLfloat x, GLfloat y)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_Vertex2fv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 4);
@@ -1979,7 +1979,7 @@ __indirect_glVertex2fv(const GLfloat * v)
 void
 __indirect_glVertex2i(GLint x, GLint y)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_Vertex2iv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 4);
@@ -2001,7 +2001,7 @@ __indirect_glVertex2iv(const GLint * v)
 void
 __indirect_glVertex2s(GLshort x, GLshort y)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_Vertex2sv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 2);
@@ -2023,7 +2023,7 @@ __indirect_glVertex2sv(const GLshort * v)
 void
 __indirect_glVertex3d(GLdouble x, GLdouble y, GLdouble z)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 28;
     emit_header(gc->pc, X_GLrop_Vertex3dv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 8);
@@ -2046,7 +2046,7 @@ __indirect_glVertex3dv(const GLdouble * v)
 void
 __indirect_glVertex3f(GLfloat x, GLfloat y, GLfloat z)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_Vertex3fv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 4);
@@ -2069,7 +2069,7 @@ __indirect_glVertex3fv(const GLfloat * v)
 void
 __indirect_glVertex3i(GLint x, GLint y, GLint z)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_Vertex3iv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 4);
@@ -2092,7 +2092,7 @@ __indirect_glVertex3iv(const GLint * v)
 void
 __indirect_glVertex3s(GLshort x, GLshort y, GLshort z)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_Vertex3sv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 2);
@@ -2115,7 +2115,7 @@ __indirect_glVertex3sv(const GLshort * v)
 void
 __indirect_glVertex4d(GLdouble x, GLdouble y, GLdouble z, GLdouble w)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 36;
     emit_header(gc->pc, X_GLrop_Vertex4dv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 8);
@@ -2139,7 +2139,7 @@ __indirect_glVertex4dv(const GLdouble * v)
 void
 __indirect_glVertex4f(GLfloat x, GLfloat y, GLfloat z, GLfloat w)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_Vertex4fv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 4);
@@ -2163,7 +2163,7 @@ __indirect_glVertex4fv(const GLfloat * v)
 void
 __indirect_glVertex4i(GLint x, GLint y, GLint z, GLint w)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_Vertex4iv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 4);
@@ -2187,7 +2187,7 @@ __indirect_glVertex4iv(const GLint * v)
 void
 __indirect_glVertex4s(GLshort x, GLshort y, GLshort z, GLshort w)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_Vertex4sv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 2);
@@ -2211,7 +2211,7 @@ __indirect_glVertex4sv(const GLshort * v)
 void
 __indirect_glClipPlane(GLenum plane, const GLdouble * equation)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 40;
     emit_header(gc->pc, X_GLrop_ClipPlane, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (equation), 32);
@@ -2226,7 +2226,7 @@ __indirect_glClipPlane(GLenum plane, const GLdouble * equation)
 void
 __indirect_glColorMaterial(GLenum face, GLenum mode)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_ColorMaterial, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&face), 4);
@@ -2241,7 +2241,7 @@ __indirect_glColorMaterial(GLenum face, GLenum mode)
 void
 __indirect_glCullFace(GLenum mode)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_CullFace, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&mode), 4);
@@ -2255,7 +2255,7 @@ __indirect_glCullFace(GLenum mode)
 void
 __indirect_glFogf(GLenum pname, GLfloat param)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_Fogf, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&pname), 4);
@@ -2270,7 +2270,7 @@ __indirect_glFogf(GLenum pname, GLfloat param)
 void
 __indirect_glFogfv(GLenum pname, const GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize = __glFogfv_size(pname);
     const GLuint cmdlen = 8 + __GLX_PAD((compsize * 4));
     emit_header(gc->pc, X_GLrop_Fogfv, cmdlen);
@@ -2286,7 +2286,7 @@ __indirect_glFogfv(GLenum pname, const GLfloat * params)
 void
 __indirect_glFogi(GLenum pname, GLint param)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_Fogi, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&pname), 4);
@@ -2301,7 +2301,7 @@ __indirect_glFogi(GLenum pname, GLint param)
 void
 __indirect_glFogiv(GLenum pname, const GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize = __glFogiv_size(pname);
     const GLuint cmdlen = 8 + __GLX_PAD((compsize * 4));
     emit_header(gc->pc, X_GLrop_Fogiv, cmdlen);
@@ -2317,7 +2317,7 @@ __indirect_glFogiv(GLenum pname, const GLint * params)
 void
 __indirect_glFrontFace(GLenum mode)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_FrontFace, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&mode), 4);
@@ -2331,7 +2331,7 @@ __indirect_glFrontFace(GLenum mode)
 void
 __indirect_glHint(GLenum target, GLenum mode)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_Hint, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -2346,7 +2346,7 @@ __indirect_glHint(GLenum target, GLenum mode)
 void
 __indirect_glLightf(GLenum light, GLenum pname, GLfloat param)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_Lightf, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&light), 4);
@@ -2362,7 +2362,7 @@ __indirect_glLightf(GLenum light, GLenum pname, GLfloat param)
 void
 __indirect_glLightfv(GLenum light, GLenum pname, const GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize = __glLightfv_size(pname);
     const GLuint cmdlen = 12 + __GLX_PAD((compsize * 4));
     emit_header(gc->pc, X_GLrop_Lightfv, cmdlen);
@@ -2379,7 +2379,7 @@ __indirect_glLightfv(GLenum light, GLenum pname, const GLfloat * params)
 void
 __indirect_glLighti(GLenum light, GLenum pname, GLint param)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_Lighti, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&light), 4);
@@ -2395,7 +2395,7 @@ __indirect_glLighti(GLenum light, GLenum pname, GLint param)
 void
 __indirect_glLightiv(GLenum light, GLenum pname, const GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize = __glLightiv_size(pname);
     const GLuint cmdlen = 12 + __GLX_PAD((compsize * 4));
     emit_header(gc->pc, X_GLrop_Lightiv, cmdlen);
@@ -2412,7 +2412,7 @@ __indirect_glLightiv(GLenum light, GLenum pname, const GLint * params)
 void
 __indirect_glLightModelf(GLenum pname, GLfloat param)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_LightModelf, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&pname), 4);
@@ -2427,7 +2427,7 @@ __indirect_glLightModelf(GLenum pname, GLfloat param)
 void
 __indirect_glLightModelfv(GLenum pname, const GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize = __glLightModelfv_size(pname);
     const GLuint cmdlen = 8 + __GLX_PAD((compsize * 4));
     emit_header(gc->pc, X_GLrop_LightModelfv, cmdlen);
@@ -2443,7 +2443,7 @@ __indirect_glLightModelfv(GLenum pname, const GLfloat * params)
 void
 __indirect_glLightModeli(GLenum pname, GLint param)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_LightModeli, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&pname), 4);
@@ -2458,7 +2458,7 @@ __indirect_glLightModeli(GLenum pname, GLint param)
 void
 __indirect_glLightModeliv(GLenum pname, const GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize = __glLightModeliv_size(pname);
     const GLuint cmdlen = 8 + __GLX_PAD((compsize * 4));
     emit_header(gc->pc, X_GLrop_LightModeliv, cmdlen);
@@ -2474,7 +2474,7 @@ __indirect_glLightModeliv(GLenum pname, const GLint * params)
 void
 __indirect_glLineStipple(GLint factor, GLushort pattern)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_LineStipple, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&factor), 4);
@@ -2489,7 +2489,7 @@ __indirect_glLineStipple(GLint factor, GLushort pattern)
 void
 __indirect_glLineWidth(GLfloat width)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_LineWidth, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&width), 4);
@@ -2503,7 +2503,7 @@ __indirect_glLineWidth(GLfloat width)
 void
 __indirect_glMaterialf(GLenum face, GLenum pname, GLfloat param)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_Materialf, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&face), 4);
@@ -2519,7 +2519,7 @@ __indirect_glMaterialf(GLenum face, GLenum pname, GLfloat param)
 void
 __indirect_glMaterialfv(GLenum face, GLenum pname, const GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize = __glMaterialfv_size(pname);
     const GLuint cmdlen = 12 + __GLX_PAD((compsize * 4));
     emit_header(gc->pc, X_GLrop_Materialfv, cmdlen);
@@ -2536,7 +2536,7 @@ __indirect_glMaterialfv(GLenum face, GLenum pname, const GLfloat * params)
 void
 __indirect_glMateriali(GLenum face, GLenum pname, GLint param)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_Materiali, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&face), 4);
@@ -2552,7 +2552,7 @@ __indirect_glMateriali(GLenum face, GLenum pname, GLint param)
 void
 __indirect_glMaterialiv(GLenum face, GLenum pname, const GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize = __glMaterialiv_size(pname);
     const GLuint cmdlen = 12 + __GLX_PAD((compsize * 4));
     emit_header(gc->pc, X_GLrop_Materialiv, cmdlen);
@@ -2569,7 +2569,7 @@ __indirect_glMaterialiv(GLenum face, GLenum pname, const GLint * params)
 void
 __indirect_glPointSize(GLfloat size)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_PointSize, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&size), 4);
@@ -2583,7 +2583,7 @@ __indirect_glPointSize(GLfloat size)
 void
 __indirect_glPolygonMode(GLenum face, GLenum mode)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_PolygonMode, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&face), 4);
@@ -2598,7 +2598,7 @@ __indirect_glPolygonMode(GLenum face, GLenum mode)
 void
 __indirect_glPolygonStipple(const GLubyte *mask)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize =
         (mask != NULL) ? __glImageSize(32, 32, 1, GL_COLOR_INDEX, GL_BITMAP,
                                        0) : 0;
@@ -2621,7 +2621,7 @@ __indirect_glPolygonStipple(const GLubyte *mask)
 void
 __indirect_glScissor(GLint x, GLint y, GLsizei width, GLsizei height)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_Scissor, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 4);
@@ -2638,7 +2638,7 @@ __indirect_glScissor(GLint x, GLint y, GLsizei width, GLsizei height)
 void
 __indirect_glShadeModel(GLenum mode)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_ShadeModel, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&mode), 4);
@@ -2652,7 +2652,7 @@ __indirect_glShadeModel(GLenum mode)
 void
 __indirect_glTexParameterf(GLenum target, GLenum pname, GLfloat param)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_TexParameterf, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -2669,7 +2669,7 @@ void
 __indirect_glTexParameterfv(GLenum target, GLenum pname,
                             const GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize = __glTexParameterfv_size(pname);
     const GLuint cmdlen = 12 + __GLX_PAD((compsize * 4));
     emit_header(gc->pc, X_GLrop_TexParameterfv, cmdlen);
@@ -2686,7 +2686,7 @@ __indirect_glTexParameterfv(GLenum target, GLenum pname,
 void
 __indirect_glTexParameteri(GLenum target, GLenum pname, GLint param)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_TexParameteri, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -2702,7 +2702,7 @@ __indirect_glTexParameteri(GLenum target, GLenum pname, GLint param)
 void
 __indirect_glTexParameteriv(GLenum target, GLenum pname, const GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize = __glTexParameteriv_size(pname);
     const GLuint cmdlen = 12 + __GLX_PAD((compsize * 4));
     emit_header(gc->pc, X_GLrop_TexParameteriv, cmdlen);
@@ -2721,7 +2721,7 @@ __glx_TexImage_1D2D(unsigned opcode, unsigned dim, GLenum target, GLint level,
                     GLint border, GLenum format, GLenum type,
                     const GLvoid * pixels)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize =
         __glImageSize(width, height, 1, format, type, target);
     const GLuint cmdlen = 56 + __GLX_PAD(compsize);
@@ -2795,7 +2795,7 @@ __indirect_glTexImage2D(GLenum target, GLint level, GLint internalformat,
 void
 __indirect_glTexEnvf(GLenum target, GLenum pname, GLfloat param)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_TexEnvf, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -2811,7 +2811,7 @@ __indirect_glTexEnvf(GLenum target, GLenum pname, GLfloat param)
 void
 __indirect_glTexEnvfv(GLenum target, GLenum pname, const GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize = __glTexEnvfv_size(pname);
     const GLuint cmdlen = 12 + __GLX_PAD((compsize * 4));
     emit_header(gc->pc, X_GLrop_TexEnvfv, cmdlen);
@@ -2828,7 +2828,7 @@ __indirect_glTexEnvfv(GLenum target, GLenum pname, const GLfloat * params)
 void
 __indirect_glTexEnvi(GLenum target, GLenum pname, GLint param)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_TexEnvi, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -2844,7 +2844,7 @@ __indirect_glTexEnvi(GLenum target, GLenum pname, GLint param)
 void
 __indirect_glTexEnviv(GLenum target, GLenum pname, const GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize = __glTexEnviv_size(pname);
     const GLuint cmdlen = 12 + __GLX_PAD((compsize * 4));
     emit_header(gc->pc, X_GLrop_TexEnviv, cmdlen);
@@ -2861,7 +2861,7 @@ __indirect_glTexEnviv(GLenum target, GLenum pname, const GLint * params)
 void
 __indirect_glTexGend(GLenum coord, GLenum pname, GLdouble param)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_TexGend, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&param), 8);
@@ -2877,7 +2877,7 @@ __indirect_glTexGend(GLenum coord, GLenum pname, GLdouble param)
 void
 __indirect_glTexGendv(GLenum coord, GLenum pname, const GLdouble * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize = __glTexGendv_size(pname);
     const GLuint cmdlen = 12 + __GLX_PAD((compsize * 8));
     emit_header(gc->pc, X_GLrop_TexGendv, cmdlen);
@@ -2894,7 +2894,7 @@ __indirect_glTexGendv(GLenum coord, GLenum pname, const GLdouble * params)
 void
 __indirect_glTexGenf(GLenum coord, GLenum pname, GLfloat param)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_TexGenf, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&coord), 4);
@@ -2910,7 +2910,7 @@ __indirect_glTexGenf(GLenum coord, GLenum pname, GLfloat param)
 void
 __indirect_glTexGenfv(GLenum coord, GLenum pname, const GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize = __glTexGenfv_size(pname);
     const GLuint cmdlen = 12 + __GLX_PAD((compsize * 4));
     emit_header(gc->pc, X_GLrop_TexGenfv, cmdlen);
@@ -2927,7 +2927,7 @@ __indirect_glTexGenfv(GLenum coord, GLenum pname, const GLfloat * params)
 void
 __indirect_glTexGeni(GLenum coord, GLenum pname, GLint param)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_TexGeni, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&coord), 4);
@@ -2943,7 +2943,7 @@ __indirect_glTexGeni(GLenum coord, GLenum pname, GLint param)
 void
 __indirect_glTexGeniv(GLenum coord, GLenum pname, const GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize = __glTexGeniv_size(pname);
     const GLuint cmdlen = 12 + __GLX_PAD((compsize * 4));
     emit_header(gc->pc, X_GLrop_TexGeniv, cmdlen);
@@ -2960,7 +2960,7 @@ __indirect_glTexGeniv(GLenum coord, GLenum pname, const GLint * params)
 void
 __indirect_glInitNames(void)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 4;
     emit_header(gc->pc, X_GLrop_InitNames, cmdlen);
     gc->pc += cmdlen;
@@ -2973,7 +2973,7 @@ __indirect_glInitNames(void)
 void
 __indirect_glLoadName(GLuint name)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_LoadName, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&name), 4);
@@ -2987,7 +2987,7 @@ __indirect_glLoadName(GLuint name)
 void
 __indirect_glPassThrough(GLfloat token)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_PassThrough, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&token), 4);
@@ -3001,7 +3001,7 @@ __indirect_glPassThrough(GLfloat token)
 void
 __indirect_glPopName(void)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 4;
     emit_header(gc->pc, X_GLrop_PopName, cmdlen);
     gc->pc += cmdlen;
@@ -3014,7 +3014,7 @@ __indirect_glPopName(void)
 void
 __indirect_glPushName(GLuint name)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_PushName, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&name), 4);
@@ -3028,7 +3028,7 @@ __indirect_glPushName(GLuint name)
 void
 __indirect_glDrawBuffer(GLenum mode)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_DrawBuffer, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&mode), 4);
@@ -3042,7 +3042,7 @@ __indirect_glDrawBuffer(GLenum mode)
 void
 __indirect_glClear(GLbitfield mask)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_Clear, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&mask), 4);
@@ -3057,7 +3057,7 @@ void
 __indirect_glClearAccum(GLfloat red, GLfloat green, GLfloat blue,
                         GLfloat alpha)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_ClearAccum, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 4);
@@ -3074,7 +3074,7 @@ __indirect_glClearAccum(GLfloat red, GLfloat green, GLfloat blue,
 void
 __indirect_glClearIndex(GLfloat c)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_ClearIndex, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&c), 4);
@@ -3089,7 +3089,7 @@ void
 __indirect_glClearColor(GLclampf red, GLclampf green, GLclampf blue,
                         GLclampf alpha)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_ClearColor, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 4);
@@ -3106,7 +3106,7 @@ __indirect_glClearColor(GLclampf red, GLclampf green, GLclampf blue,
 void
 __indirect_glClearStencil(GLint s)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_ClearStencil, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&s), 4);
@@ -3120,7 +3120,7 @@ __indirect_glClearStencil(GLint s)
 void
 __indirect_glClearDepth(GLclampd depth)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_ClearDepth, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&depth), 8);
@@ -3134,7 +3134,7 @@ __indirect_glClearDepth(GLclampd depth)
 void
 __indirect_glStencilMask(GLuint mask)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_StencilMask, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&mask), 4);
@@ -3149,7 +3149,7 @@ void
 __indirect_glColorMask(GLboolean red, GLboolean green, GLboolean blue,
                        GLboolean alpha)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_ColorMask, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 1);
@@ -3166,7 +3166,7 @@ __indirect_glColorMask(GLboolean red, GLboolean green, GLboolean blue,
 void
 __indirect_glDepthMask(GLboolean flag)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_DepthMask, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&flag), 1);
@@ -3180,7 +3180,7 @@ __indirect_glDepthMask(GLboolean flag)
 void
 __indirect_glIndexMask(GLuint mask)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_IndexMask, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&mask), 4);
@@ -3194,7 +3194,7 @@ __indirect_glIndexMask(GLuint mask)
 void
 __indirect_glAccum(GLenum op, GLfloat value)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_Accum, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&op), 4);
@@ -3209,7 +3209,7 @@ __indirect_glAccum(GLenum op, GLfloat value)
 void
 __indirect_glPopAttrib(void)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 4;
     emit_header(gc->pc, X_GLrop_PopAttrib, cmdlen);
     gc->pc += cmdlen;
@@ -3222,7 +3222,7 @@ __indirect_glPopAttrib(void)
 void
 __indirect_glPushAttrib(GLbitfield mask)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_PushAttrib, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&mask), 4);
@@ -3236,7 +3236,7 @@ __indirect_glPushAttrib(GLbitfield mask)
 void
 __indirect_glMapGrid1d(GLint un, GLdouble u1, GLdouble u2)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_MapGrid1d, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&u1), 8);
@@ -3252,7 +3252,7 @@ __indirect_glMapGrid1d(GLint un, GLdouble u1, GLdouble u2)
 void
 __indirect_glMapGrid1f(GLint un, GLfloat u1, GLfloat u2)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_MapGrid1f, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&un), 4);
@@ -3269,7 +3269,7 @@ void
 __indirect_glMapGrid2d(GLint un, GLdouble u1, GLdouble u2, GLint vn,
                        GLdouble v1, GLdouble v2)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 44;
     emit_header(gc->pc, X_GLrop_MapGrid2d, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&u1), 8);
@@ -3289,7 +3289,7 @@ void
 __indirect_glMapGrid2f(GLint un, GLfloat u1, GLfloat u2, GLint vn, GLfloat v1,
                        GLfloat v2)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 28;
     emit_header(gc->pc, X_GLrop_MapGrid2f, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&un), 4);
@@ -3308,7 +3308,7 @@ __indirect_glMapGrid2f(GLint un, GLfloat u1, GLfloat u2, GLint vn, GLfloat v1,
 void
 __indirect_glEvalCoord1d(GLdouble u)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_EvalCoord1dv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&u), 8);
@@ -3329,7 +3329,7 @@ __indirect_glEvalCoord1dv(const GLdouble * u)
 void
 __indirect_glEvalCoord1f(GLfloat u)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_EvalCoord1fv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&u), 4);
@@ -3350,7 +3350,7 @@ __indirect_glEvalCoord1fv(const GLfloat * u)
 void
 __indirect_glEvalCoord2d(GLdouble u, GLdouble v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_EvalCoord2dv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&u), 8);
@@ -3372,7 +3372,7 @@ __indirect_glEvalCoord2dv(const GLdouble * u)
 void
 __indirect_glEvalCoord2f(GLfloat u, GLfloat v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_EvalCoord2fv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&u), 4);
@@ -3394,7 +3394,7 @@ __indirect_glEvalCoord2fv(const GLfloat * u)
 void
 __indirect_glEvalMesh1(GLenum mode, GLint i1, GLint i2)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_EvalMesh1, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&mode), 4);
@@ -3410,7 +3410,7 @@ __indirect_glEvalMesh1(GLenum mode, GLint i1, GLint i2)
 void
 __indirect_glEvalPoint1(GLint i)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_EvalPoint1, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&i), 4);
@@ -3424,7 +3424,7 @@ __indirect_glEvalPoint1(GLint i)
 void
 __indirect_glEvalMesh2(GLenum mode, GLint i1, GLint i2, GLint j1, GLint j2)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_EvalMesh2, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&mode), 4);
@@ -3442,7 +3442,7 @@ __indirect_glEvalMesh2(GLenum mode, GLint i1, GLint i2, GLint j1, GLint j2)
 void
 __indirect_glEvalPoint2(GLint i, GLint j)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_EvalPoint2, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&i), 4);
@@ -3457,7 +3457,7 @@ __indirect_glEvalPoint2(GLint i, GLint j)
 void
 __indirect_glAlphaFunc(GLenum func, GLclampf ref)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_AlphaFunc, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&func), 4);
@@ -3472,7 +3472,7 @@ __indirect_glAlphaFunc(GLenum func, GLclampf ref)
 void
 __indirect_glBlendFunc(GLenum sfactor, GLenum dfactor)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_BlendFunc, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&sfactor), 4);
@@ -3487,7 +3487,7 @@ __indirect_glBlendFunc(GLenum sfactor, GLenum dfactor)
 void
 __indirect_glLogicOp(GLenum opcode)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_LogicOp, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&opcode), 4);
@@ -3501,7 +3501,7 @@ __indirect_glLogicOp(GLenum opcode)
 void
 __indirect_glStencilFunc(GLenum func, GLint ref, GLuint mask)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_StencilFunc, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&func), 4);
@@ -3517,7 +3517,7 @@ __indirect_glStencilFunc(GLenum func, GLint ref, GLuint mask)
 void
 __indirect_glStencilOp(GLenum fail, GLenum zfail, GLenum zpass)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_StencilOp, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&fail), 4);
@@ -3533,7 +3533,7 @@ __indirect_glStencilOp(GLenum fail, GLenum zfail, GLenum zpass)
 void
 __indirect_glDepthFunc(GLenum func)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_DepthFunc, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&func), 4);
@@ -3547,7 +3547,7 @@ __indirect_glDepthFunc(GLenum func)
 void
 __indirect_glPixelZoom(GLfloat xfactor, GLfloat yfactor)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_PixelZoom, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&xfactor), 4);
@@ -3562,7 +3562,7 @@ __indirect_glPixelZoom(GLfloat xfactor, GLfloat yfactor)
 void
 __indirect_glPixelTransferf(GLenum pname, GLfloat param)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_PixelTransferf, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&pname), 4);
@@ -3577,7 +3577,7 @@ __indirect_glPixelTransferf(GLenum pname, GLfloat param)
 void
 __indirect_glPixelTransferi(GLenum pname, GLint param)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_PixelTransferi, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&pname), 4);
@@ -3592,7 +3592,7 @@ __indirect_glPixelTransferi(GLenum pname, GLint param)
 void
 __indirect_glPixelMapfv(GLenum map, GLsizei mapsize, const GLfloat * values)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12 + __GLX_PAD((mapsize * 4));
     if (mapsize < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -3629,7 +3629,7 @@ __indirect_glPixelMapfv(GLenum map, GLsizei mapsize, const GLfloat * values)
 void
 __indirect_glPixelMapuiv(GLenum map, GLsizei mapsize, const GLuint * values)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12 + __GLX_PAD((mapsize * 4));
     if (mapsize < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -3666,7 +3666,7 @@ __indirect_glPixelMapuiv(GLenum map, GLsizei mapsize, const GLuint * values)
 void
 __indirect_glPixelMapusv(GLenum map, GLsizei mapsize, const GLushort * values)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12 + __GLX_PAD((mapsize * 2));
     if (mapsize < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -3703,7 +3703,7 @@ __indirect_glPixelMapusv(GLenum map, GLsizei mapsize, const GLushort * values)
 void
 __indirect_glReadBuffer(GLenum mode)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_ReadBuffer, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&mode), 4);
@@ -3718,7 +3718,7 @@ void
 __indirect_glCopyPixels(GLint x, GLint y, GLsizei width, GLsizei height,
                         GLenum type)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_CopyPixels, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 4);
@@ -3737,7 +3737,7 @@ void
 __indirect_glReadPixels(GLint x, GLint y, GLsizei width, GLsizei height,
                         GLenum format, GLenum type, GLvoid * pixels)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const __GLXattribute *const state = gc->client_state_private;
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
@@ -3786,7 +3786,7 @@ void
 __indirect_glDrawPixels(GLsizei width, GLsizei height, GLenum format,
                         GLenum type, const GLvoid * pixels)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize =
         (pixels != NULL) ? __glImageSize(width, height, 1, format, type,
                                          0) : 0;
@@ -3832,7 +3832,7 @@ __indirect_glDrawPixels(GLsizei width, GLsizei height, GLenum format,
 void
 __indirect_glGetClipPlane(GLenum plane, GLdouble * equation)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 4;
@@ -3867,7 +3867,7 @@ __indirect_glGetClipPlane(GLenum plane, GLdouble * equation)
 void
 __indirect_glGetLightfv(GLenum light, GLenum pname, GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -3907,7 +3907,7 @@ __indirect_glGetLightfv(GLenum light, GLenum pname, GLfloat * params)
 void
 __indirect_glGetLightiv(GLenum light, GLenum pname, GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -3947,7 +3947,7 @@ __indirect_glGetLightiv(GLenum light, GLenum pname, GLint * params)
 void
 __indirect_glGetMapdv(GLenum target, GLenum query, GLdouble * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -3985,7 +3985,7 @@ __indirect_glGetMapdv(GLenum target, GLenum query, GLdouble * v)
 void
 __indirect_glGetMapfv(GLenum target, GLenum query, GLfloat * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -4023,7 +4023,7 @@ __indirect_glGetMapfv(GLenum target, GLenum query, GLfloat * v)
 void
 __indirect_glGetMapiv(GLenum target, GLenum query, GLint * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -4061,7 +4061,7 @@ __indirect_glGetMapiv(GLenum target, GLenum query, GLint * v)
 void
 __indirect_glGetMaterialfv(GLenum face, GLenum pname, GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -4101,7 +4101,7 @@ __indirect_glGetMaterialfv(GLenum face, GLenum pname, GLfloat * params)
 void
 __indirect_glGetMaterialiv(GLenum face, GLenum pname, GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -4141,7 +4141,7 @@ __indirect_glGetMaterialiv(GLenum face, GLenum pname, GLint * params)
 void
 __indirect_glGetPixelMapfv(GLenum map, GLfloat * values)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 4;
@@ -4179,7 +4179,7 @@ __indirect_glGetPixelMapfv(GLenum map, GLfloat * values)
 void
 __indirect_glGetPixelMapuiv(GLenum map, GLuint * values)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 4;
@@ -4218,7 +4218,7 @@ __indirect_glGetPixelMapuiv(GLenum map, GLuint * values)
 void
 __indirect_glGetPixelMapusv(GLenum map, GLushort * values)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 4;
@@ -4257,7 +4257,7 @@ __indirect_glGetPixelMapusv(GLenum map, GLushort * values)
 void
 __indirect_glGetPolygonStipple(GLubyte *mask)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 4;
@@ -4294,7 +4294,7 @@ __indirect_glGetPolygonStipple(GLubyte *mask)
 void
 __indirect_glGetTexEnvfv(GLenum target, GLenum pname, GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -4334,7 +4334,7 @@ __indirect_glGetTexEnvfv(GLenum target, GLenum pname, GLfloat * params)
 void
 __indirect_glGetTexEnviv(GLenum target, GLenum pname, GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -4374,7 +4374,7 @@ __indirect_glGetTexEnviv(GLenum target, GLenum pname, GLint * params)
 void
 __indirect_glGetTexGendv(GLenum coord, GLenum pname, GLdouble * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -4414,7 +4414,7 @@ __indirect_glGetTexGendv(GLenum coord, GLenum pname, GLdouble * params)
 void
 __indirect_glGetTexGenfv(GLenum coord, GLenum pname, GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -4454,7 +4454,7 @@ __indirect_glGetTexGenfv(GLenum coord, GLenum pname, GLfloat * params)
 void
 __indirect_glGetTexGeniv(GLenum coord, GLenum pname, GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -4495,7 +4495,7 @@ void
 __indirect_glGetTexImage(GLenum target, GLint level, GLenum format,
                          GLenum type, GLvoid * pixels)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const __GLXattribute *const state = gc->client_state_private;
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
@@ -4542,7 +4542,7 @@ __indirect_glGetTexImage(GLenum target, GLint level, GLenum format,
 void
 __indirect_glGetTexParameterfv(GLenum target, GLenum pname, GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -4583,7 +4583,7 @@ __indirect_glGetTexParameterfv(GLenum target, GLenum pname, GLfloat * params)
 void
 __indirect_glGetTexParameteriv(GLenum target, GLenum pname, GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -4625,7 +4625,7 @@ void
 __indirect_glGetTexLevelParameterfv(GLenum target, GLint level, GLenum pname,
                                     GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 12;
@@ -4668,7 +4668,7 @@ void
 __indirect_glGetTexLevelParameteriv(GLenum target, GLint level, GLenum pname,
                                     GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 12;
@@ -4710,7 +4710,7 @@ __indirect_glGetTexLevelParameteriv(GLenum target, GLint level, GLenum pname,
 GLboolean
 __indirect_glIsList(GLuint list)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     GLboolean retval = (GLboolean) 0;
 #ifndef USE_XCB
@@ -4742,7 +4742,7 @@ __indirect_glIsList(GLuint list)
 void
 __indirect_glDepthRange(GLclampd zNear, GLclampd zFar)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_DepthRange, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&zNear), 8);
@@ -4758,7 +4758,7 @@ void
 __indirect_glFrustum(GLdouble left, GLdouble right, GLdouble bottom,
                      GLdouble top, GLdouble zNear, GLdouble zFar)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 52;
     emit_header(gc->pc, X_GLrop_Frustum, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&left), 8);
@@ -4777,7 +4777,7 @@ __indirect_glFrustum(GLdouble left, GLdouble right, GLdouble bottom,
 void
 __indirect_glLoadIdentity(void)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 4;
     emit_header(gc->pc, X_GLrop_LoadIdentity, cmdlen);
     gc->pc += cmdlen;
@@ -4790,7 +4790,7 @@ __indirect_glLoadIdentity(void)
 void
 __indirect_glLoadMatrixf(const GLfloat * m)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 68;
     emit_header(gc->pc, X_GLrop_LoadMatrixf, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (m), 64);
@@ -4804,7 +4804,7 @@ __indirect_glLoadMatrixf(const GLfloat * m)
 void
 __indirect_glLoadMatrixd(const GLdouble * m)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 132;
     emit_header(gc->pc, X_GLrop_LoadMatrixd, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (m), 128);
@@ -4818,7 +4818,7 @@ __indirect_glLoadMatrixd(const GLdouble * m)
 void
 __indirect_glMatrixMode(GLenum mode)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_MatrixMode, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&mode), 4);
@@ -4832,7 +4832,7 @@ __indirect_glMatrixMode(GLenum mode)
 void
 __indirect_glMultMatrixf(const GLfloat * m)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 68;
     emit_header(gc->pc, X_GLrop_MultMatrixf, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (m), 64);
@@ -4846,7 +4846,7 @@ __indirect_glMultMatrixf(const GLfloat * m)
 void
 __indirect_glMultMatrixd(const GLdouble * m)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 132;
     emit_header(gc->pc, X_GLrop_MultMatrixd, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (m), 128);
@@ -4861,7 +4861,7 @@ void
 __indirect_glOrtho(GLdouble left, GLdouble right, GLdouble bottom,
                    GLdouble top, GLdouble zNear, GLdouble zFar)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 52;
     emit_header(gc->pc, X_GLrop_Ortho, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&left), 8);
@@ -4880,7 +4880,7 @@ __indirect_glOrtho(GLdouble left, GLdouble right, GLdouble bottom,
 void
 __indirect_glPopMatrix(void)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 4;
     emit_header(gc->pc, X_GLrop_PopMatrix, cmdlen);
     gc->pc += cmdlen;
@@ -4893,7 +4893,7 @@ __indirect_glPopMatrix(void)
 void
 __indirect_glPushMatrix(void)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 4;
     emit_header(gc->pc, X_GLrop_PushMatrix, cmdlen);
     gc->pc += cmdlen;
@@ -4906,7 +4906,7 @@ __indirect_glPushMatrix(void)
 void
 __indirect_glRotated(GLdouble angle, GLdouble x, GLdouble y, GLdouble z)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 36;
     emit_header(gc->pc, X_GLrop_Rotated, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&angle), 8);
@@ -4923,7 +4923,7 @@ __indirect_glRotated(GLdouble angle, GLdouble x, GLdouble y, GLdouble z)
 void
 __indirect_glRotatef(GLfloat angle, GLfloat x, GLfloat y, GLfloat z)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_Rotatef, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&angle), 4);
@@ -4940,7 +4940,7 @@ __indirect_glRotatef(GLfloat angle, GLfloat x, GLfloat y, GLfloat z)
 void
 __indirect_glScaled(GLdouble x, GLdouble y, GLdouble z)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 28;
     emit_header(gc->pc, X_GLrop_Scaled, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 8);
@@ -4956,7 +4956,7 @@ __indirect_glScaled(GLdouble x, GLdouble y, GLdouble z)
 void
 __indirect_glScalef(GLfloat x, GLfloat y, GLfloat z)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_Scalef, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 4);
@@ -4972,7 +4972,7 @@ __indirect_glScalef(GLfloat x, GLfloat y, GLfloat z)
 void
 __indirect_glTranslated(GLdouble x, GLdouble y, GLdouble z)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 28;
     emit_header(gc->pc, X_GLrop_Translated, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 8);
@@ -4988,7 +4988,7 @@ __indirect_glTranslated(GLdouble x, GLdouble y, GLdouble z)
 void
 __indirect_glTranslatef(GLfloat x, GLfloat y, GLfloat z)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_Translatef, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 4);
@@ -5004,7 +5004,7 @@ __indirect_glTranslatef(GLfloat x, GLfloat y, GLfloat z)
 void
 __indirect_glViewport(GLint x, GLint y, GLsizei width, GLsizei height)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_Viewport, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 4);
@@ -5021,7 +5021,7 @@ __indirect_glViewport(GLint x, GLint y, GLsizei width, GLsizei height)
 void
 __indirect_glBindTexture(GLenum target, GLuint texture)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_BindTexture, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -5036,7 +5036,7 @@ __indirect_glBindTexture(GLenum target, GLuint texture)
 void
 __indirect_glIndexub(GLubyte c)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_Indexubv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&c), 1);
@@ -5050,7 +5050,7 @@ __indirect_glIndexub(GLubyte c)
 void
 __indirect_glIndexubv(const GLubyte *c)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_Indexubv, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (c), 1);
@@ -5064,7 +5064,7 @@ __indirect_glIndexubv(const GLubyte *c)
 void
 __indirect_glPolygonOffset(GLfloat factor, GLfloat units)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_PolygonOffset, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&factor), 4);
@@ -5080,7 +5080,7 @@ void
 __indirect_glCopyTexImage1D(GLenum target, GLint level, GLenum internalformat,
                             GLint x, GLint y, GLsizei width, GLint border)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 32;
     emit_header(gc->pc, X_GLrop_CopyTexImage1D, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -5102,7 +5102,7 @@ __indirect_glCopyTexImage2D(GLenum target, GLint level, GLenum internalformat,
                             GLint x, GLint y, GLsizei width, GLsizei height,
                             GLint border)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 36;
     emit_header(gc->pc, X_GLrop_CopyTexImage2D, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -5124,7 +5124,7 @@ void
 __indirect_glCopyTexSubImage1D(GLenum target, GLint level, GLint xoffset,
                                GLint x, GLint y, GLsizei width)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 28;
     emit_header(gc->pc, X_GLrop_CopyTexSubImage1D, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -5145,7 +5145,7 @@ __indirect_glCopyTexSubImage2D(GLenum target, GLint level, GLint xoffset,
                                GLint yoffset, GLint x, GLint y, GLsizei width,
                                GLsizei height)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 36;
     emit_header(gc->pc, X_GLrop_CopyTexSubImage2D, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -5166,7 +5166,7 @@ __indirect_glCopyTexSubImage2D(GLenum target, GLint level, GLint xoffset,
 void
 __indirect_glDeleteTextures(GLsizei n, const GLuint * textures)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 4 + __GLX_PAD((n * 4));
@@ -5196,15 +5196,15 @@ __indirect_glDeleteTextures(GLsizei n, const GLuint * textures)
 void
 glDeleteTexturesEXT(GLsizei n, const GLuint * textures)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
 
 #if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
-    if (gc->driContext) {
+    if (gc->isDirect) {
         CALL_DeleteTextures(GET_DISPATCH(), (n, textures));
     } else
 #endif
     {
-        __GLXcontext *const gc = __glXGetCurrentContext();
+        struct glx_context *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 4 + __GLX_PAD((n * 4));
         if (n < 0) {
@@ -5228,7 +5228,7 @@ glDeleteTexturesEXT(GLsizei n, const GLuint * textures)
 void
 __indirect_glGenTextures(GLsizei n, GLuint * textures)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 4;
@@ -5267,15 +5267,15 @@ __indirect_glGenTextures(GLsizei n, GLuint * textures)
 void
 glGenTexturesEXT(GLsizei n, GLuint * textures)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
 
 #if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
-    if (gc->driContext) {
+    if (gc->isDirect) {
         CALL_GenTextures(GET_DISPATCH(), (n, textures));
     } else
 #endif
     {
-        __GLXcontext *const gc = __glXGetCurrentContext();
+        struct glx_context *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 4;
         if (n < 0) {
@@ -5299,7 +5299,7 @@ glGenTexturesEXT(GLsizei n, GLuint * textures)
 GLboolean
 __indirect_glIsTexture(GLuint texture)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     GLboolean retval = (GLboolean) 0;
 #ifndef USE_XCB
@@ -5332,15 +5332,15 @@ __indirect_glIsTexture(GLuint texture)
 GLboolean
 glIsTextureEXT(GLuint texture)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
 
 #if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
-    if (gc->driContext) {
+    if (gc->isDirect) {
         return CALL_IsTexture(GET_DISPATCH(), (texture));
     } else
 #endif
     {
-        __GLXcontext *const gc = __glXGetCurrentContext();
+        struct glx_context *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         GLboolean retval = (GLboolean) 0;
         const GLuint cmdlen = 4;
@@ -5362,7 +5362,7 @@ void
 __indirect_glPrioritizeTextures(GLsizei n, const GLuint * textures,
                                 const GLclampf * priorities)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8 + __GLX_PAD((n * 4)) + __GLX_PAD((n * 4));
     if (n < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -5387,7 +5387,7 @@ __glx_TexSubImage_1D2D(unsigned opcode, unsigned dim, GLenum target,
                        GLsizei width, GLsizei height, GLenum format,
                        GLenum type, const GLvoid * pixels)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize =
         (pixels != NULL) ? __glImageSize(width, height, 1, format, type,
                                          target) : 0;
@@ -5464,7 +5464,7 @@ void
 __indirect_glBlendColor(GLclampf red, GLclampf green, GLclampf blue,
                         GLclampf alpha)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_BlendColor, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 4);
@@ -5481,7 +5481,7 @@ __indirect_glBlendColor(GLclampf red, GLclampf green, GLclampf blue,
 void
 __indirect_glBlendEquation(GLenum mode)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_BlendEquation, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&mode), 4);
@@ -5496,7 +5496,7 @@ void
 __indirect_glColorTable(GLenum target, GLenum internalformat, GLsizei width,
                         GLenum format, GLenum type, const GLvoid * table)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize =
         (table != NULL) ? __glImageSize(width, 1, 1, format, type,
                                         target) : 0;
@@ -5546,7 +5546,7 @@ void
 __indirect_glColorTableParameterfv(GLenum target, GLenum pname,
                                    const GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize = __glColorTableParameterfv_size(pname);
     const GLuint cmdlen = 12 + __GLX_PAD((compsize * 4));
     emit_header(gc->pc, X_GLrop_ColorTableParameterfv, cmdlen);
@@ -5564,7 +5564,7 @@ void
 __indirect_glColorTableParameteriv(GLenum target, GLenum pname,
                                    const GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize = __glColorTableParameteriv_size(pname);
     const GLuint cmdlen = 12 + __GLX_PAD((compsize * 4));
     emit_header(gc->pc, X_GLrop_ColorTableParameteriv, cmdlen);
@@ -5582,7 +5582,7 @@ void
 __indirect_glCopyColorTable(GLenum target, GLenum internalformat, GLint x,
                             GLint y, GLsizei width)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_CopyColorTable, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -5601,7 +5601,7 @@ void
 __indirect_glGetColorTable(GLenum target, GLenum format, GLenum type,
                            GLvoid * table)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const __GLXattribute *const state = gc->client_state_private;
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
@@ -5648,15 +5648,15 @@ __indirect_glGetColorTable(GLenum target, GLenum format, GLenum type,
 void
 glGetColorTableEXT(GLenum target, GLenum format, GLenum type, GLvoid * table)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
 
 #if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
-    if (gc->driContext) {
+    if (gc->isDirect) {
         CALL_GetColorTable(GET_DISPATCH(), (target, format, type, table));
     } else
 #endif
     {
-        __GLXcontext *const gc = __glXGetCurrentContext();
+        struct glx_context *const gc = __glXGetCurrentContext();
         const __GLXattribute *const state = gc->client_state_private;
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 16;
@@ -5683,7 +5683,7 @@ void
 __indirect_glGetColorTableParameterfv(GLenum target, GLenum pname,
                                       GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -5724,16 +5724,16 @@ __indirect_glGetColorTableParameterfv(GLenum target, GLenum pname,
 void
 glGetColorTableParameterfvEXT(GLenum target, GLenum pname, GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
 
 #if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
-    if (gc->driContext) {
+    if (gc->isDirect) {
         CALL_GetColorTableParameterfv(GET_DISPATCH(),
                                       (target, pname, params));
     } else
 #endif
     {
-        __GLXcontext *const gc = __glXGetCurrentContext();
+        struct glx_context *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 8;
         if (__builtin_expect(dpy != NULL, 1)) {
@@ -5756,7 +5756,7 @@ void
 __indirect_glGetColorTableParameteriv(GLenum target, GLenum pname,
                                       GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -5797,16 +5797,16 @@ __indirect_glGetColorTableParameteriv(GLenum target, GLenum pname,
 void
 glGetColorTableParameterivEXT(GLenum target, GLenum pname, GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
 
 #if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
-    if (gc->driContext) {
+    if (gc->isDirect) {
         CALL_GetColorTableParameteriv(GET_DISPATCH(),
                                       (target, pname, params));
     } else
 #endif
     {
-        __GLXcontext *const gc = __glXGetCurrentContext();
+        struct glx_context *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 8;
         if (__builtin_expect(dpy != NULL, 1)) {
@@ -5829,7 +5829,7 @@ void
 __indirect_glColorSubTable(GLenum target, GLsizei start, GLsizei count,
                            GLenum format, GLenum type, const GLvoid * data)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize =
         (data != NULL) ? __glImageSize(count, 1, 1, format, type, target) : 0;
     const GLuint cmdlen = 44 + __GLX_PAD(compsize);
@@ -5877,7 +5877,7 @@ void
 __indirect_glCopyColorSubTable(GLenum target, GLsizei start, GLint x, GLint y,
                                GLsizei width)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_CopyColorSubTable, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -5897,7 +5897,7 @@ __glx_ConvolutionFilter_1D2D(unsigned opcode, unsigned dim, GLenum target,
                              GLsizei height, GLenum format, GLenum type,
                              const GLvoid * image)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize =
         (image != NULL) ? __glImageSize(width, height, 1, format, type,
                                         target) : 0;
@@ -5971,7 +5971,7 @@ void
 __indirect_glConvolutionParameterf(GLenum target, GLenum pname,
                                    GLfloat params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_ConvolutionParameterf, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -5988,7 +5988,7 @@ void
 __indirect_glConvolutionParameterfv(GLenum target, GLenum pname,
                                     const GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize = __glConvolutionParameterfv_size(pname);
     const GLuint cmdlen = 12 + __GLX_PAD((compsize * 4));
     emit_header(gc->pc, X_GLrop_ConvolutionParameterfv, cmdlen);
@@ -6005,7 +6005,7 @@ __indirect_glConvolutionParameterfv(GLenum target, GLenum pname,
 void
 __indirect_glConvolutionParameteri(GLenum target, GLenum pname, GLint params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_ConvolutionParameteri, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -6022,7 +6022,7 @@ void
 __indirect_glConvolutionParameteriv(GLenum target, GLenum pname,
                                     const GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize = __glConvolutionParameteriv_size(pname);
     const GLuint cmdlen = 12 + __GLX_PAD((compsize * 4));
     emit_header(gc->pc, X_GLrop_ConvolutionParameteriv, cmdlen);
@@ -6040,7 +6040,7 @@ void
 __indirect_glCopyConvolutionFilter1D(GLenum target, GLenum internalformat,
                                      GLint x, GLint y, GLsizei width)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_CopyConvolutionFilter1D, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -6060,7 +6060,7 @@ __indirect_glCopyConvolutionFilter2D(GLenum target, GLenum internalformat,
                                      GLint x, GLint y, GLsizei width,
                                      GLsizei height)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 28;
     emit_header(gc->pc, X_GLrop_CopyConvolutionFilter2D, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -6080,7 +6080,7 @@ void
 __indirect_glGetConvolutionFilter(GLenum target, GLenum format, GLenum type,
                                   GLvoid * image)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const __GLXattribute *const state = gc->client_state_private;
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
@@ -6123,16 +6123,16 @@ void
 gl_dispatch_stub_356(GLenum target, GLenum format, GLenum type,
                      GLvoid * image)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
 
 #if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
-    if (gc->driContext) {
+    if (gc->isDirect) {
         CALL_GetConvolutionFilter(GET_DISPATCH(),
                                   (target, format, type, image));
     } else
 #endif
     {
-        __GLXcontext *const gc = __glXGetCurrentContext();
+        struct glx_context *const gc = __glXGetCurrentContext();
         const __GLXattribute *const state = gc->client_state_private;
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 16;
@@ -6160,7 +6160,7 @@ void
 __indirect_glGetConvolutionParameterfv(GLenum target, GLenum pname,
                                        GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -6201,16 +6201,16 @@ __indirect_glGetConvolutionParameterfv(GLenum target, GLenum pname,
 void
 gl_dispatch_stub_357(GLenum target, GLenum pname, GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
 
 #if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
-    if (gc->driContext) {
+    if (gc->isDirect) {
         CALL_GetConvolutionParameterfv(GET_DISPATCH(),
                                        (target, pname, params));
     } else
 #endif
     {
-        __GLXcontext *const gc = __glXGetCurrentContext();
+        struct glx_context *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 8;
         if (__builtin_expect(dpy != NULL, 1)) {
@@ -6233,7 +6233,7 @@ void
 __indirect_glGetConvolutionParameteriv(GLenum target, GLenum pname,
                                        GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -6274,16 +6274,16 @@ __indirect_glGetConvolutionParameteriv(GLenum target, GLenum pname,
 void
 gl_dispatch_stub_358(GLenum target, GLenum pname, GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
 
 #if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
-    if (gc->driContext) {
+    if (gc->isDirect) {
         CALL_GetConvolutionParameteriv(GET_DISPATCH(),
                                        (target, pname, params));
     } else
 #endif
     {
-        __GLXcontext *const gc = __glXGetCurrentContext();
+        struct glx_context *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 8;
         if (__builtin_expect(dpy != NULL, 1)) {
@@ -6306,7 +6306,7 @@ void
 __indirect_glGetHistogram(GLenum target, GLboolean reset, GLenum format,
                           GLenum type, GLvoid * values)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const __GLXattribute *const state = gc->client_state_private;
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
@@ -6354,16 +6354,16 @@ void
 gl_dispatch_stub_361(GLenum target, GLboolean reset, GLenum format,
                      GLenum type, GLvoid * values)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
 
 #if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
-    if (gc->driContext) {
+    if (gc->isDirect) {
         CALL_GetHistogram(GET_DISPATCH(),
                           (target, reset, format, type, values));
     } else
 #endif
     {
-        __GLXcontext *const gc = __glXGetCurrentContext();
+        struct glx_context *const gc = __glXGetCurrentContext();
         const __GLXattribute *const state = gc->client_state_private;
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 16;
@@ -6391,7 +6391,7 @@ void
 __indirect_glGetHistogramParameterfv(GLenum target, GLenum pname,
                                      GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -6431,15 +6431,15 @@ __indirect_glGetHistogramParameterfv(GLenum target, GLenum pname,
 void
 gl_dispatch_stub_362(GLenum target, GLenum pname, GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
 
 #if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
-    if (gc->driContext) {
+    if (gc->isDirect) {
         CALL_GetHistogramParameterfv(GET_DISPATCH(), (target, pname, params));
     } else
 #endif
     {
-        __GLXcontext *const gc = __glXGetCurrentContext();
+        struct glx_context *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 8;
         if (__builtin_expect(dpy != NULL, 1)) {
@@ -6462,7 +6462,7 @@ void
 __indirect_glGetHistogramParameteriv(GLenum target, GLenum pname,
                                      GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -6502,15 +6502,15 @@ __indirect_glGetHistogramParameteriv(GLenum target, GLenum pname,
 void
 gl_dispatch_stub_363(GLenum target, GLenum pname, GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
 
 #if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
-    if (gc->driContext) {
+    if (gc->isDirect) {
         CALL_GetHistogramParameteriv(GET_DISPATCH(), (target, pname, params));
     } else
 #endif
     {
-        __GLXcontext *const gc = __glXGetCurrentContext();
+        struct glx_context *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 8;
         if (__builtin_expect(dpy != NULL, 1)) {
@@ -6533,7 +6533,7 @@ void
 __indirect_glGetMinmax(GLenum target, GLboolean reset, GLenum format,
                        GLenum type, GLvoid * values)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const __GLXattribute *const state = gc->client_state_private;
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
@@ -6577,15 +6577,15 @@ void
 gl_dispatch_stub_364(GLenum target, GLboolean reset, GLenum format,
                      GLenum type, GLvoid * values)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
 
 #if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
-    if (gc->driContext) {
+    if (gc->isDirect) {
         CALL_GetMinmax(GET_DISPATCH(), (target, reset, format, type, values));
     } else
 #endif
     {
-        __GLXcontext *const gc = __glXGetCurrentContext();
+        struct glx_context *const gc = __glXGetCurrentContext();
         const __GLXattribute *const state = gc->client_state_private;
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 16;
@@ -6613,7 +6613,7 @@ void
 __indirect_glGetMinmaxParameterfv(GLenum target, GLenum pname,
                                   GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -6651,15 +6651,15 @@ __indirect_glGetMinmaxParameterfv(GLenum target, GLenum pname,
 void
 gl_dispatch_stub_365(GLenum target, GLenum pname, GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
 
 #if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
-    if (gc->driContext) {
+    if (gc->isDirect) {
         CALL_GetMinmaxParameterfv(GET_DISPATCH(), (target, pname, params));
     } else
 #endif
     {
-        __GLXcontext *const gc = __glXGetCurrentContext();
+        struct glx_context *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 8;
         if (__builtin_expect(dpy != NULL, 1)) {
@@ -6681,7 +6681,7 @@ gl_dispatch_stub_365(GLenum target, GLenum pname, GLfloat * params)
 void
 __indirect_glGetMinmaxParameteriv(GLenum target, GLenum pname, GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -6719,15 +6719,15 @@ __indirect_glGetMinmaxParameteriv(GLenum target, GLenum pname, GLint * params)
 void
 gl_dispatch_stub_366(GLenum target, GLenum pname, GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
 
 #if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
-    if (gc->driContext) {
+    if (gc->isDirect) {
         CALL_GetMinmaxParameteriv(GET_DISPATCH(), (target, pname, params));
     } else
 #endif
     {
-        __GLXcontext *const gc = __glXGetCurrentContext();
+        struct glx_context *const gc = __glXGetCurrentContext();
         Display *const dpy = gc->currentDpy;
         const GLuint cmdlen = 8;
         if (__builtin_expect(dpy != NULL, 1)) {
@@ -6750,7 +6750,7 @@ void
 __indirect_glHistogram(GLenum target, GLsizei width, GLenum internalformat,
                        GLboolean sink)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_Histogram, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -6767,7 +6767,7 @@ __indirect_glHistogram(GLenum target, GLsizei width, GLenum internalformat,
 void
 __indirect_glMinmax(GLenum target, GLenum internalformat, GLboolean sink)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_Minmax, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -6783,7 +6783,7 @@ __indirect_glMinmax(GLenum target, GLenum internalformat, GLboolean sink)
 void
 __indirect_glResetHistogram(GLenum target)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_ResetHistogram, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -6797,7 +6797,7 @@ __indirect_glResetHistogram(GLenum target)
 void
 __indirect_glResetMinmax(GLenum target)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_ResetMinmax, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -6813,7 +6813,7 @@ __glx_TexImage_3D4D(unsigned opcode, unsigned dim, GLenum target, GLint level,
                     GLsizei depth, GLsizei extent, GLint border,
                     GLenum format, GLenum type, const GLvoid * pixels)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize =
         (pixels != NULL) ? __glImageSize(width, height, depth, format, type,
                                          target) : 0;
@@ -6890,7 +6890,7 @@ __glx_TexSubImage_3D4D(unsigned opcode, unsigned dim, GLenum target,
                        GLsizei height, GLsizei depth, GLsizei extent,
                        GLenum format, GLenum type, const GLvoid * pixels)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize =
         (pixels != NULL) ? __glImageSize(width, height, depth, format, type,
                                          target) : 0;
@@ -6968,7 +6968,7 @@ __indirect_glCopyTexSubImage3D(GLenum target, GLint level, GLint xoffset,
                                GLint yoffset, GLint zoffset, GLint x, GLint y,
                                GLsizei width, GLsizei height)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 40;
     emit_header(gc->pc, X_GLrop_CopyTexSubImage3D, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -6990,7 +6990,7 @@ __indirect_glCopyTexSubImage3D(GLenum target, GLint level, GLint xoffset,
 void
 __indirect_glActiveTextureARB(GLenum texture)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_ActiveTextureARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&texture), 4);
@@ -7004,7 +7004,7 @@ __indirect_glActiveTextureARB(GLenum texture)
 void
 __indirect_glMultiTexCoord1dARB(GLenum target, GLdouble s)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_MultiTexCoord1dvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&s), 8);
@@ -7019,7 +7019,7 @@ __indirect_glMultiTexCoord1dARB(GLenum target, GLdouble s)
 void
 __indirect_glMultiTexCoord1dvARB(GLenum target, const GLdouble * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_MultiTexCoord1dvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (v), 8);
@@ -7034,7 +7034,7 @@ __indirect_glMultiTexCoord1dvARB(GLenum target, const GLdouble * v)
 void
 __indirect_glMultiTexCoord1fARB(GLenum target, GLfloat s)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_MultiTexCoord1fvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7049,7 +7049,7 @@ __indirect_glMultiTexCoord1fARB(GLenum target, GLfloat s)
 void
 __indirect_glMultiTexCoord1fvARB(GLenum target, const GLfloat * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_MultiTexCoord1fvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7064,7 +7064,7 @@ __indirect_glMultiTexCoord1fvARB(GLenum target, const GLfloat * v)
 void
 __indirect_glMultiTexCoord1iARB(GLenum target, GLint s)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_MultiTexCoord1ivARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7079,7 +7079,7 @@ __indirect_glMultiTexCoord1iARB(GLenum target, GLint s)
 void
 __indirect_glMultiTexCoord1ivARB(GLenum target, const GLint * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_MultiTexCoord1ivARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7094,7 +7094,7 @@ __indirect_glMultiTexCoord1ivARB(GLenum target, const GLint * v)
 void
 __indirect_glMultiTexCoord1sARB(GLenum target, GLshort s)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_MultiTexCoord1svARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7109,7 +7109,7 @@ __indirect_glMultiTexCoord1sARB(GLenum target, GLshort s)
 void
 __indirect_glMultiTexCoord1svARB(GLenum target, const GLshort * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_MultiTexCoord1svARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7124,7 +7124,7 @@ __indirect_glMultiTexCoord1svARB(GLenum target, const GLshort * v)
 void
 __indirect_glMultiTexCoord2dARB(GLenum target, GLdouble s, GLdouble t)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_MultiTexCoord2dvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&s), 8);
@@ -7140,7 +7140,7 @@ __indirect_glMultiTexCoord2dARB(GLenum target, GLdouble s, GLdouble t)
 void
 __indirect_glMultiTexCoord2dvARB(GLenum target, const GLdouble * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_MultiTexCoord2dvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (v), 16);
@@ -7155,7 +7155,7 @@ __indirect_glMultiTexCoord2dvARB(GLenum target, const GLdouble * v)
 void
 __indirect_glMultiTexCoord2fARB(GLenum target, GLfloat s, GLfloat t)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_MultiTexCoord2fvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7171,7 +7171,7 @@ __indirect_glMultiTexCoord2fARB(GLenum target, GLfloat s, GLfloat t)
 void
 __indirect_glMultiTexCoord2fvARB(GLenum target, const GLfloat * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_MultiTexCoord2fvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7186,7 +7186,7 @@ __indirect_glMultiTexCoord2fvARB(GLenum target, const GLfloat * v)
 void
 __indirect_glMultiTexCoord2iARB(GLenum target, GLint s, GLint t)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_MultiTexCoord2ivARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7202,7 +7202,7 @@ __indirect_glMultiTexCoord2iARB(GLenum target, GLint s, GLint t)
 void
 __indirect_glMultiTexCoord2ivARB(GLenum target, const GLint * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_MultiTexCoord2ivARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7217,7 +7217,7 @@ __indirect_glMultiTexCoord2ivARB(GLenum target, const GLint * v)
 void
 __indirect_glMultiTexCoord2sARB(GLenum target, GLshort s, GLshort t)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_MultiTexCoord2svARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7233,7 +7233,7 @@ __indirect_glMultiTexCoord2sARB(GLenum target, GLshort s, GLshort t)
 void
 __indirect_glMultiTexCoord2svARB(GLenum target, const GLshort * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_MultiTexCoord2svARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7249,7 +7249,7 @@ void
 __indirect_glMultiTexCoord3dARB(GLenum target, GLdouble s, GLdouble t,
                                 GLdouble r)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 32;
     emit_header(gc->pc, X_GLrop_MultiTexCoord3dvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&s), 8);
@@ -7266,7 +7266,7 @@ __indirect_glMultiTexCoord3dARB(GLenum target, GLdouble s, GLdouble t,
 void
 __indirect_glMultiTexCoord3dvARB(GLenum target, const GLdouble * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 32;
     emit_header(gc->pc, X_GLrop_MultiTexCoord3dvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (v), 24);
@@ -7282,7 +7282,7 @@ void
 __indirect_glMultiTexCoord3fARB(GLenum target, GLfloat s, GLfloat t,
                                 GLfloat r)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_MultiTexCoord3fvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7299,7 +7299,7 @@ __indirect_glMultiTexCoord3fARB(GLenum target, GLfloat s, GLfloat t,
 void
 __indirect_glMultiTexCoord3fvARB(GLenum target, const GLfloat * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_MultiTexCoord3fvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7314,7 +7314,7 @@ __indirect_glMultiTexCoord3fvARB(GLenum target, const GLfloat * v)
 void
 __indirect_glMultiTexCoord3iARB(GLenum target, GLint s, GLint t, GLint r)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_MultiTexCoord3ivARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7331,7 +7331,7 @@ __indirect_glMultiTexCoord3iARB(GLenum target, GLint s, GLint t, GLint r)
 void
 __indirect_glMultiTexCoord3ivARB(GLenum target, const GLint * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_MultiTexCoord3ivARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7347,7 +7347,7 @@ void
 __indirect_glMultiTexCoord3sARB(GLenum target, GLshort s, GLshort t,
                                 GLshort r)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_MultiTexCoord3svARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7364,7 +7364,7 @@ __indirect_glMultiTexCoord3sARB(GLenum target, GLshort s, GLshort t,
 void
 __indirect_glMultiTexCoord3svARB(GLenum target, const GLshort * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_MultiTexCoord3svARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7380,7 +7380,7 @@ void
 __indirect_glMultiTexCoord4dARB(GLenum target, GLdouble s, GLdouble t,
                                 GLdouble r, GLdouble q)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 40;
     emit_header(gc->pc, X_GLrop_MultiTexCoord4dvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&s), 8);
@@ -7398,7 +7398,7 @@ __indirect_glMultiTexCoord4dARB(GLenum target, GLdouble s, GLdouble t,
 void
 __indirect_glMultiTexCoord4dvARB(GLenum target, const GLdouble * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 40;
     emit_header(gc->pc, X_GLrop_MultiTexCoord4dvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (v), 32);
@@ -7414,7 +7414,7 @@ void
 __indirect_glMultiTexCoord4fARB(GLenum target, GLfloat s, GLfloat t,
                                 GLfloat r, GLfloat q)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_MultiTexCoord4fvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7432,7 +7432,7 @@ __indirect_glMultiTexCoord4fARB(GLenum target, GLfloat s, GLfloat t,
 void
 __indirect_glMultiTexCoord4fvARB(GLenum target, const GLfloat * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_MultiTexCoord4fvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7448,7 +7448,7 @@ void
 __indirect_glMultiTexCoord4iARB(GLenum target, GLint s, GLint t, GLint r,
                                 GLint q)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_MultiTexCoord4ivARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7466,7 +7466,7 @@ __indirect_glMultiTexCoord4iARB(GLenum target, GLint s, GLint t, GLint r,
 void
 __indirect_glMultiTexCoord4ivARB(GLenum target, const GLint * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_MultiTexCoord4ivARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7482,7 +7482,7 @@ void
 __indirect_glMultiTexCoord4sARB(GLenum target, GLshort s, GLshort t,
                                 GLshort r, GLshort q)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_MultiTexCoord4svARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7500,7 +7500,7 @@ __indirect_glMultiTexCoord4sARB(GLenum target, GLshort s, GLshort t,
 void
 __indirect_glMultiTexCoord4svARB(GLenum target, const GLshort * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_MultiTexCoord4svARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7515,7 +7515,7 @@ __indirect_glMultiTexCoord4svARB(GLenum target, const GLshort * v)
 void
 __indirect_glSampleCoverageARB(GLclampf value, GLboolean invert)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_SampleCoverageARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&value), 4);
@@ -7530,7 +7530,7 @@ __indirect_glSampleCoverageARB(GLclampf value, GLboolean invert)
 void
 __indirect_glGetProgramStringARB(GLenum target, GLenum pname, GLvoid * string)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     const GLuint cmdlen = 8;
     if (__builtin_expect(dpy != NULL, 1)) {
@@ -7550,7 +7550,7 @@ __indirect_glGetProgramStringARB(GLenum target, GLenum pname, GLvoid * string)
 void
 __indirect_glGetProgramivARB(GLenum target, GLenum pname, GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     const GLuint cmdlen = 8;
     if (__builtin_expect(dpy != NULL, 1)) {
@@ -7571,7 +7571,7 @@ void
 __indirect_glProgramEnvParameter4dARB(GLenum target, GLuint index, GLdouble x,
                                       GLdouble y, GLdouble z, GLdouble w)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 44;
     emit_header(gc->pc, X_GLrop_ProgramEnvParameter4dvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7591,7 +7591,7 @@ void
 __indirect_glProgramEnvParameter4dvARB(GLenum target, GLuint index,
                                        const GLdouble * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 44;
     emit_header(gc->pc, X_GLrop_ProgramEnvParameter4dvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7608,7 +7608,7 @@ void
 __indirect_glProgramEnvParameter4fARB(GLenum target, GLuint index, GLfloat x,
                                       GLfloat y, GLfloat z, GLfloat w)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 28;
     emit_header(gc->pc, X_GLrop_ProgramEnvParameter4fvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7628,7 +7628,7 @@ void
 __indirect_glProgramEnvParameter4fvARB(GLenum target, GLuint index,
                                        const GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 28;
     emit_header(gc->pc, X_GLrop_ProgramEnvParameter4fvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7646,7 +7646,7 @@ __indirect_glProgramLocalParameter4dARB(GLenum target, GLuint index,
                                         GLdouble x, GLdouble y, GLdouble z,
                                         GLdouble w)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 44;
     emit_header(gc->pc, X_GLrop_ProgramLocalParameter4dvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7666,7 +7666,7 @@ void
 __indirect_glProgramLocalParameter4dvARB(GLenum target, GLuint index,
                                          const GLdouble * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 44;
     emit_header(gc->pc, X_GLrop_ProgramLocalParameter4dvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7684,7 +7684,7 @@ __indirect_glProgramLocalParameter4fARB(GLenum target, GLuint index,
                                         GLfloat x, GLfloat y, GLfloat z,
                                         GLfloat w)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 28;
     emit_header(gc->pc, X_GLrop_ProgramLocalParameter4fvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7704,7 +7704,7 @@ void
 __indirect_glProgramLocalParameter4fvARB(GLenum target, GLuint index,
                                          const GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 28;
     emit_header(gc->pc, X_GLrop_ProgramLocalParameter4fvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -7721,7 +7721,7 @@ void
 __indirect_glProgramStringARB(GLenum target, GLenum format, GLsizei len,
                               const GLvoid * string)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16 + __GLX_PAD(len);
     if (len < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -7759,7 +7759,7 @@ __indirect_glProgramStringARB(GLenum target, GLenum format, GLsizei len,
 void
 __indirect_glVertexAttrib1dARB(GLuint index, GLdouble x)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_VertexAttrib1dvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -7774,7 +7774,7 @@ __indirect_glVertexAttrib1dARB(GLuint index, GLdouble x)
 void
 __indirect_glVertexAttrib1dvARB(GLuint index, const GLdouble * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_VertexAttrib1dvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -7789,7 +7789,7 @@ __indirect_glVertexAttrib1dvARB(GLuint index, const GLdouble * v)
 void
 __indirect_glVertexAttrib1fARB(GLuint index, GLfloat x)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_VertexAttrib1fvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -7804,7 +7804,7 @@ __indirect_glVertexAttrib1fARB(GLuint index, GLfloat x)
 void
 __indirect_glVertexAttrib1fvARB(GLuint index, const GLfloat * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_VertexAttrib1fvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -7819,7 +7819,7 @@ __indirect_glVertexAttrib1fvARB(GLuint index, const GLfloat * v)
 void
 __indirect_glVertexAttrib1sARB(GLuint index, GLshort x)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_VertexAttrib1svARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -7834,7 +7834,7 @@ __indirect_glVertexAttrib1sARB(GLuint index, GLshort x)
 void
 __indirect_glVertexAttrib1svARB(GLuint index, const GLshort * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_VertexAttrib1svARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -7849,7 +7849,7 @@ __indirect_glVertexAttrib1svARB(GLuint index, const GLshort * v)
 void
 __indirect_glVertexAttrib2dARB(GLuint index, GLdouble x, GLdouble y)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_VertexAttrib2dvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -7865,7 +7865,7 @@ __indirect_glVertexAttrib2dARB(GLuint index, GLdouble x, GLdouble y)
 void
 __indirect_glVertexAttrib2dvARB(GLuint index, const GLdouble * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_VertexAttrib2dvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -7880,7 +7880,7 @@ __indirect_glVertexAttrib2dvARB(GLuint index, const GLdouble * v)
 void
 __indirect_glVertexAttrib2fARB(GLuint index, GLfloat x, GLfloat y)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_VertexAttrib2fvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -7896,7 +7896,7 @@ __indirect_glVertexAttrib2fARB(GLuint index, GLfloat x, GLfloat y)
 void
 __indirect_glVertexAttrib2fvARB(GLuint index, const GLfloat * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_VertexAttrib2fvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -7911,7 +7911,7 @@ __indirect_glVertexAttrib2fvARB(GLuint index, const GLfloat * v)
 void
 __indirect_glVertexAttrib2sARB(GLuint index, GLshort x, GLshort y)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_VertexAttrib2svARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -7927,7 +7927,7 @@ __indirect_glVertexAttrib2sARB(GLuint index, GLshort x, GLshort y)
 void
 __indirect_glVertexAttrib2svARB(GLuint index, const GLshort * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_VertexAttrib2svARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -7943,7 +7943,7 @@ void
 __indirect_glVertexAttrib3dARB(GLuint index, GLdouble x, GLdouble y,
                                GLdouble z)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 32;
     emit_header(gc->pc, X_GLrop_VertexAttrib3dvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -7960,7 +7960,7 @@ __indirect_glVertexAttrib3dARB(GLuint index, GLdouble x, GLdouble y,
 void
 __indirect_glVertexAttrib3dvARB(GLuint index, const GLdouble * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 32;
     emit_header(gc->pc, X_GLrop_VertexAttrib3dvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -7975,7 +7975,7 @@ __indirect_glVertexAttrib3dvARB(GLuint index, const GLdouble * v)
 void
 __indirect_glVertexAttrib3fARB(GLuint index, GLfloat x, GLfloat y, GLfloat z)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_VertexAttrib3fvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -7992,7 +7992,7 @@ __indirect_glVertexAttrib3fARB(GLuint index, GLfloat x, GLfloat y, GLfloat z)
 void
 __indirect_glVertexAttrib3fvARB(GLuint index, const GLfloat * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_VertexAttrib3fvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -8007,7 +8007,7 @@ __indirect_glVertexAttrib3fvARB(GLuint index, const GLfloat * v)
 void
 __indirect_glVertexAttrib3sARB(GLuint index, GLshort x, GLshort y, GLshort z)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_VertexAttrib3svARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -8024,7 +8024,7 @@ __indirect_glVertexAttrib3sARB(GLuint index, GLshort x, GLshort y, GLshort z)
 void
 __indirect_glVertexAttrib3svARB(GLuint index, const GLshort * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_VertexAttrib3svARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -8039,7 +8039,7 @@ __indirect_glVertexAttrib3svARB(GLuint index, const GLshort * v)
 void
 __indirect_glVertexAttrib4NbvARB(GLuint index, const GLbyte *v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_VertexAttrib4NbvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -8054,7 +8054,7 @@ __indirect_glVertexAttrib4NbvARB(GLuint index, const GLbyte *v)
 void
 __indirect_glVertexAttrib4NivARB(GLuint index, const GLint * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_VertexAttrib4NivARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -8069,7 +8069,7 @@ __indirect_glVertexAttrib4NivARB(GLuint index, const GLint * v)
 void
 __indirect_glVertexAttrib4NsvARB(GLuint index, const GLshort * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_VertexAttrib4NsvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -8085,7 +8085,7 @@ void
 __indirect_glVertexAttrib4NubARB(GLuint index, GLubyte x, GLubyte y,
                                  GLubyte z, GLubyte w)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_VertexAttrib4NubvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -8103,7 +8103,7 @@ __indirect_glVertexAttrib4NubARB(GLuint index, GLubyte x, GLubyte y,
 void
 __indirect_glVertexAttrib4NubvARB(GLuint index, const GLubyte *v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_VertexAttrib4NubvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -8118,7 +8118,7 @@ __indirect_glVertexAttrib4NubvARB(GLuint index, const GLubyte *v)
 void
 __indirect_glVertexAttrib4NuivARB(GLuint index, const GLuint * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_VertexAttrib4NuivARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -8133,7 +8133,7 @@ __indirect_glVertexAttrib4NuivARB(GLuint index, const GLuint * v)
 void
 __indirect_glVertexAttrib4NusvARB(GLuint index, const GLushort * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_VertexAttrib4NusvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -8148,7 +8148,7 @@ __indirect_glVertexAttrib4NusvARB(GLuint index, const GLushort * v)
 void
 __indirect_glVertexAttrib4bvARB(GLuint index, const GLbyte *v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_VertexAttrib4bvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -8164,7 +8164,7 @@ void
 __indirect_glVertexAttrib4dARB(GLuint index, GLdouble x, GLdouble y,
                                GLdouble z, GLdouble w)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 40;
     emit_header(gc->pc, X_GLrop_VertexAttrib4dvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -8182,7 +8182,7 @@ __indirect_glVertexAttrib4dARB(GLuint index, GLdouble x, GLdouble y,
 void
 __indirect_glVertexAttrib4dvARB(GLuint index, const GLdouble * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 40;
     emit_header(gc->pc, X_GLrop_VertexAttrib4dvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -8198,7 +8198,7 @@ void
 __indirect_glVertexAttrib4fARB(GLuint index, GLfloat x, GLfloat y, GLfloat z,
                                GLfloat w)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_VertexAttrib4fvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -8216,7 +8216,7 @@ __indirect_glVertexAttrib4fARB(GLuint index, GLfloat x, GLfloat y, GLfloat z,
 void
 __indirect_glVertexAttrib4fvARB(GLuint index, const GLfloat * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_VertexAttrib4fvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -8231,7 +8231,7 @@ __indirect_glVertexAttrib4fvARB(GLuint index, const GLfloat * v)
 void
 __indirect_glVertexAttrib4ivARB(GLuint index, const GLint * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_VertexAttrib4ivARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -8247,7 +8247,7 @@ void
 __indirect_glVertexAttrib4sARB(GLuint index, GLshort x, GLshort y, GLshort z,
                                GLshort w)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_VertexAttrib4svARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -8265,7 +8265,7 @@ __indirect_glVertexAttrib4sARB(GLuint index, GLshort x, GLshort y, GLshort z,
 void
 __indirect_glVertexAttrib4svARB(GLuint index, const GLshort * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_VertexAttrib4svARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -8280,7 +8280,7 @@ __indirect_glVertexAttrib4svARB(GLuint index, const GLshort * v)
 void
 __indirect_glVertexAttrib4ubvARB(GLuint index, const GLubyte *v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_VertexAttrib4ubvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -8295,7 +8295,7 @@ __indirect_glVertexAttrib4ubvARB(GLuint index, const GLubyte *v)
 void
 __indirect_glVertexAttrib4uivARB(GLuint index, const GLuint * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_VertexAttrib4uivARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -8310,7 +8310,7 @@ __indirect_glVertexAttrib4uivARB(GLuint index, const GLuint * v)
 void
 __indirect_glVertexAttrib4usvARB(GLuint index, const GLushort * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_VertexAttrib4usvARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -8325,7 +8325,7 @@ __indirect_glVertexAttrib4usvARB(GLuint index, const GLushort * v)
 void
 __indirect_glBeginQueryARB(GLenum target, GLuint id)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_BeginQueryARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -8340,7 +8340,7 @@ __indirect_glBeginQueryARB(GLenum target, GLuint id)
 void
 __indirect_glDeleteQueriesARB(GLsizei n, const GLuint * ids)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 4 + __GLX_PAD((n * 4));
@@ -8370,7 +8370,7 @@ __indirect_glDeleteQueriesARB(GLsizei n, const GLuint * ids)
 void
 __indirect_glEndQueryARB(GLenum target)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_EndQueryARB, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -8384,7 +8384,7 @@ __indirect_glEndQueryARB(GLenum target)
 void
 __indirect_glGenQueriesARB(GLsizei n, GLuint * ids)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 4;
@@ -8423,7 +8423,7 @@ __indirect_glGenQueriesARB(GLsizei n, GLuint * ids)
 void
 __indirect_glGetQueryObjectivARB(GLuint id, GLenum pname, GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -8461,7 +8461,7 @@ __indirect_glGetQueryObjectivARB(GLuint id, GLenum pname, GLint * params)
 void
 __indirect_glGetQueryObjectuivARB(GLuint id, GLenum pname, GLuint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -8499,7 +8499,7 @@ __indirect_glGetQueryObjectuivARB(GLuint id, GLenum pname, GLuint * params)
 void
 __indirect_glGetQueryivARB(GLenum target, GLenum pname, GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
 #ifndef USE_XCB
     const GLuint cmdlen = 8;
@@ -8540,7 +8540,7 @@ __indirect_glGetQueryivARB(GLenum target, GLenum pname, GLint * params)
 GLboolean
 __indirect_glIsQueryARB(GLuint id)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     GLboolean retval = (GLboolean) 0;
 #ifndef USE_XCB
@@ -8574,7 +8574,7 @@ __indirect_glIsQueryARB(GLuint id)
 void
 __indirect_glDrawBuffersARB(GLsizei n, const GLenum * bufs)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8 + __GLX_PAD((n * 4));
     if (n < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -8610,7 +8610,7 @@ __indirect_glRenderbufferStorageMultisample(GLenum target, GLsizei samples,
                                             GLenum internalformat,
                                             GLsizei width, GLsizei height)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_RenderbufferStorageMultisample, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -8628,7 +8628,7 @@ __indirect_glRenderbufferStorageMultisample(GLenum target, GLsizei samples,
 void
 __indirect_glSampleMaskSGIS(GLclampf value, GLboolean invert)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_SampleMaskSGIS, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&value), 4);
@@ -8643,7 +8643,7 @@ __indirect_glSampleMaskSGIS(GLclampf value, GLboolean invert)
 void
 __indirect_glSamplePatternSGIS(GLenum pattern)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_SamplePatternSGIS, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&pattern), 4);
@@ -8657,7 +8657,7 @@ __indirect_glSamplePatternSGIS(GLenum pattern)
 void
 __indirect_glPointParameterfEXT(GLenum pname, GLfloat param)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_PointParameterfEXT, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&pname), 4);
@@ -8672,7 +8672,7 @@ __indirect_glPointParameterfEXT(GLenum pname, GLfloat param)
 void
 __indirect_glPointParameterfvEXT(GLenum pname, const GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize = __glPointParameterfvEXT_size(pname);
     const GLuint cmdlen = 8 + __GLX_PAD((compsize * 4));
     emit_header(gc->pc, X_GLrop_PointParameterfvEXT, cmdlen);
@@ -8688,7 +8688,7 @@ __indirect_glPointParameterfvEXT(GLenum pname, const GLfloat * params)
 void
 __indirect_glSecondaryColor3bEXT(GLbyte red, GLbyte green, GLbyte blue)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_SecondaryColor3bvEXT, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 1);
@@ -8711,7 +8711,7 @@ __indirect_glSecondaryColor3bvEXT(const GLbyte *v)
 void
 __indirect_glSecondaryColor3dEXT(GLdouble red, GLdouble green, GLdouble blue)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 28;
     emit_header(gc->pc, X_GLrop_SecondaryColor3dvEXT, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 8);
@@ -8734,7 +8734,7 @@ __indirect_glSecondaryColor3dvEXT(const GLdouble * v)
 void
 __indirect_glSecondaryColor3fEXT(GLfloat red, GLfloat green, GLfloat blue)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_SecondaryColor3fvEXT, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 4);
@@ -8757,7 +8757,7 @@ __indirect_glSecondaryColor3fvEXT(const GLfloat * v)
 void
 __indirect_glSecondaryColor3iEXT(GLint red, GLint green, GLint blue)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_SecondaryColor3ivEXT, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 4);
@@ -8780,7 +8780,7 @@ __indirect_glSecondaryColor3ivEXT(const GLint * v)
 void
 __indirect_glSecondaryColor3sEXT(GLshort red, GLshort green, GLshort blue)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_SecondaryColor3svEXT, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 2);
@@ -8803,7 +8803,7 @@ __indirect_glSecondaryColor3svEXT(const GLshort * v)
 void
 __indirect_glSecondaryColor3ubEXT(GLubyte red, GLubyte green, GLubyte blue)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_SecondaryColor3ubvEXT, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 1);
@@ -8826,7 +8826,7 @@ __indirect_glSecondaryColor3ubvEXT(const GLubyte *v)
 void
 __indirect_glSecondaryColor3uiEXT(GLuint red, GLuint green, GLuint blue)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_SecondaryColor3uivEXT, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 4);
@@ -8849,7 +8849,7 @@ __indirect_glSecondaryColor3uivEXT(const GLuint * v)
 void
 __indirect_glSecondaryColor3usEXT(GLushort red, GLushort green, GLushort blue)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_SecondaryColor3usvEXT, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&red), 2);
@@ -8872,7 +8872,7 @@ __indirect_glSecondaryColor3usvEXT(const GLushort * v)
 void
 __indirect_glFogCoorddEXT(GLdouble coord)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_FogCoorddvEXT, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&coord), 8);
@@ -8893,7 +8893,7 @@ __indirect_glFogCoorddvEXT(const GLdouble * coord)
 void
 __indirect_glFogCoordfEXT(GLfloat coord)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_FogCoordfvEXT, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&coord), 4);
@@ -8915,7 +8915,7 @@ void
 __indirect_glBlendFuncSeparateEXT(GLenum sfactorRGB, GLenum dfactorRGB,
                                   GLenum sfactorAlpha, GLenum dfactorAlpha)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_BlendFuncSeparateEXT, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&sfactorRGB), 4);
@@ -8932,7 +8932,7 @@ __indirect_glBlendFuncSeparateEXT(GLenum sfactorRGB, GLenum dfactorRGB,
 void
 __indirect_glWindowPos3fMESA(GLfloat x, GLfloat y, GLfloat z)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_WindowPos3fvMESA, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 4);
@@ -8956,7 +8956,7 @@ GLboolean
 __indirect_glAreProgramsResidentNV(GLsizei n, const GLuint * ids,
                                    GLboolean * residences)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     GLboolean retval = (GLboolean) 0;
     const GLuint cmdlen = 4 + __GLX_PAD((n * 4));
@@ -8981,7 +8981,7 @@ __indirect_glAreProgramsResidentNV(GLsizei n, const GLuint * ids,
 void
 __indirect_glBindProgramNV(GLenum target, GLuint program)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_BindProgramNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -8996,7 +8996,7 @@ __indirect_glBindProgramNV(GLenum target, GLuint program)
 void
 __indirect_glDeleteProgramsNV(GLsizei n, const GLuint * programs)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     const GLuint cmdlen = 4 + __GLX_PAD((n * 4));
     if (n < 0) {
@@ -9020,7 +9020,7 @@ void
 __indirect_glExecuteProgramNV(GLenum target, GLuint id,
                               const GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 28;
     emit_header(gc->pc, X_GLrop_ExecuteProgramNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -9036,7 +9036,7 @@ __indirect_glExecuteProgramNV(GLenum target, GLuint id,
 void
 __indirect_glGenProgramsNV(GLsizei n, GLuint * programs)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     const GLuint cmdlen = 4;
     if (n < 0) {
@@ -9060,7 +9060,7 @@ void
 __indirect_glGetProgramParameterdvNV(GLenum target, GLuint index,
                                      GLenum pname, GLdouble * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     const GLuint cmdlen = 12;
     if (__builtin_expect(dpy != NULL, 1)) {
@@ -9082,7 +9082,7 @@ void
 __indirect_glGetProgramParameterfvNV(GLenum target, GLuint index,
                                      GLenum pname, GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     const GLuint cmdlen = 12;
     if (__builtin_expect(dpy != NULL, 1)) {
@@ -9103,7 +9103,7 @@ __indirect_glGetProgramParameterfvNV(GLenum target, GLuint index,
 void
 __indirect_glGetProgramStringNV(GLuint id, GLenum pname, GLubyte *program)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     const GLuint cmdlen = 8;
     if (__builtin_expect(dpy != NULL, 1)) {
@@ -9123,7 +9123,7 @@ __indirect_glGetProgramStringNV(GLuint id, GLenum pname, GLubyte *program)
 void
 __indirect_glGetProgramivNV(GLuint id, GLenum pname, GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     const GLuint cmdlen = 8;
     if (__builtin_expect(dpy != NULL, 1)) {
@@ -9144,7 +9144,7 @@ void
 __indirect_glGetTrackMatrixivNV(GLenum target, GLuint address, GLenum pname,
                                 GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     const GLuint cmdlen = 12;
     if (__builtin_expect(dpy != NULL, 1)) {
@@ -9166,7 +9166,7 @@ void
 __indirect_glGetVertexAttribdvNV(GLuint index, GLenum pname,
                                  GLdouble * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     const GLuint cmdlen = 8;
     if (__builtin_expect(dpy != NULL, 1)) {
@@ -9186,7 +9186,7 @@ __indirect_glGetVertexAttribdvNV(GLuint index, GLenum pname,
 void
 __indirect_glGetVertexAttribfvNV(GLuint index, GLenum pname, GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     const GLuint cmdlen = 8;
     if (__builtin_expect(dpy != NULL, 1)) {
@@ -9206,7 +9206,7 @@ __indirect_glGetVertexAttribfvNV(GLuint index, GLenum pname, GLfloat * params)
 void
 __indirect_glGetVertexAttribivNV(GLuint index, GLenum pname, GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     const GLuint cmdlen = 8;
     if (__builtin_expect(dpy != NULL, 1)) {
@@ -9226,7 +9226,7 @@ __indirect_glGetVertexAttribivNV(GLuint index, GLenum pname, GLint * params)
 GLboolean
 __indirect_glIsProgramNV(GLuint program)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     GLboolean retval = (GLboolean) 0;
     const GLuint cmdlen = 4;
@@ -9247,7 +9247,7 @@ void
 __indirect_glLoadProgramNV(GLenum target, GLuint id, GLsizei len,
                            const GLubyte *program)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16 + __GLX_PAD(len);
     if (len < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -9271,7 +9271,7 @@ void
 __indirect_glProgramParameters4dvNV(GLenum target, GLuint index, GLuint num,
                                     const GLdouble * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16 + __GLX_PAD((num * 32));
     if (num < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -9295,7 +9295,7 @@ void
 __indirect_glProgramParameters4fvNV(GLenum target, GLuint index, GLuint num,
                                     const GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16 + __GLX_PAD((num * 16));
     if (num < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -9318,7 +9318,7 @@ __indirect_glProgramParameters4fvNV(GLenum target, GLuint index, GLuint num,
 void
 __indirect_glRequestResidentProgramsNV(GLsizei n, const GLuint * ids)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8 + __GLX_PAD((n * 4));
     if (n < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -9340,7 +9340,7 @@ void
 __indirect_glTrackMatrixNV(GLenum target, GLuint address, GLenum matrix,
                            GLenum transform)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_TrackMatrixNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -9357,7 +9357,7 @@ __indirect_glTrackMatrixNV(GLenum target, GLuint address, GLenum matrix,
 void
 __indirect_glVertexAttrib1dNV(GLuint index, GLdouble x)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_VertexAttrib1dvNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -9372,7 +9372,7 @@ __indirect_glVertexAttrib1dNV(GLuint index, GLdouble x)
 void
 __indirect_glVertexAttrib1dvNV(GLuint index, const GLdouble * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_VertexAttrib1dvNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -9387,7 +9387,7 @@ __indirect_glVertexAttrib1dvNV(GLuint index, const GLdouble * v)
 void
 __indirect_glVertexAttrib1fNV(GLuint index, GLfloat x)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_VertexAttrib1fvNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -9402,7 +9402,7 @@ __indirect_glVertexAttrib1fNV(GLuint index, GLfloat x)
 void
 __indirect_glVertexAttrib1fvNV(GLuint index, const GLfloat * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_VertexAttrib1fvNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -9417,7 +9417,7 @@ __indirect_glVertexAttrib1fvNV(GLuint index, const GLfloat * v)
 void
 __indirect_glVertexAttrib1sNV(GLuint index, GLshort x)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_VertexAttrib1svNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -9432,7 +9432,7 @@ __indirect_glVertexAttrib1sNV(GLuint index, GLshort x)
 void
 __indirect_glVertexAttrib1svNV(GLuint index, const GLshort * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_VertexAttrib1svNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -9447,7 +9447,7 @@ __indirect_glVertexAttrib1svNV(GLuint index, const GLshort * v)
 void
 __indirect_glVertexAttrib2dNV(GLuint index, GLdouble x, GLdouble y)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_VertexAttrib2dvNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -9463,7 +9463,7 @@ __indirect_glVertexAttrib2dNV(GLuint index, GLdouble x, GLdouble y)
 void
 __indirect_glVertexAttrib2dvNV(GLuint index, const GLdouble * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_VertexAttrib2dvNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -9478,7 +9478,7 @@ __indirect_glVertexAttrib2dvNV(GLuint index, const GLdouble * v)
 void
 __indirect_glVertexAttrib2fNV(GLuint index, GLfloat x, GLfloat y)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_VertexAttrib2fvNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -9494,7 +9494,7 @@ __indirect_glVertexAttrib2fNV(GLuint index, GLfloat x, GLfloat y)
 void
 __indirect_glVertexAttrib2fvNV(GLuint index, const GLfloat * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_VertexAttrib2fvNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -9509,7 +9509,7 @@ __indirect_glVertexAttrib2fvNV(GLuint index, const GLfloat * v)
 void
 __indirect_glVertexAttrib2sNV(GLuint index, GLshort x, GLshort y)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_VertexAttrib2svNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -9525,7 +9525,7 @@ __indirect_glVertexAttrib2sNV(GLuint index, GLshort x, GLshort y)
 void
 __indirect_glVertexAttrib2svNV(GLuint index, const GLshort * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_VertexAttrib2svNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -9541,7 +9541,7 @@ void
 __indirect_glVertexAttrib3dNV(GLuint index, GLdouble x, GLdouble y,
                               GLdouble z)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 32;
     emit_header(gc->pc, X_GLrop_VertexAttrib3dvNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -9558,7 +9558,7 @@ __indirect_glVertexAttrib3dNV(GLuint index, GLdouble x, GLdouble y,
 void
 __indirect_glVertexAttrib3dvNV(GLuint index, const GLdouble * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 32;
     emit_header(gc->pc, X_GLrop_VertexAttrib3dvNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -9573,7 +9573,7 @@ __indirect_glVertexAttrib3dvNV(GLuint index, const GLdouble * v)
 void
 __indirect_glVertexAttrib3fNV(GLuint index, GLfloat x, GLfloat y, GLfloat z)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_VertexAttrib3fvNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -9590,7 +9590,7 @@ __indirect_glVertexAttrib3fNV(GLuint index, GLfloat x, GLfloat y, GLfloat z)
 void
 __indirect_glVertexAttrib3fvNV(GLuint index, const GLfloat * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_VertexAttrib3fvNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -9605,7 +9605,7 @@ __indirect_glVertexAttrib3fvNV(GLuint index, const GLfloat * v)
 void
 __indirect_glVertexAttrib3sNV(GLuint index, GLshort x, GLshort y, GLshort z)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_VertexAttrib3svNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -9622,7 +9622,7 @@ __indirect_glVertexAttrib3sNV(GLuint index, GLshort x, GLshort y, GLshort z)
 void
 __indirect_glVertexAttrib3svNV(GLuint index, const GLshort * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_VertexAttrib3svNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -9638,7 +9638,7 @@ void
 __indirect_glVertexAttrib4dNV(GLuint index, GLdouble x, GLdouble y,
                               GLdouble z, GLdouble w)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 40;
     emit_header(gc->pc, X_GLrop_VertexAttrib4dvNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -9656,7 +9656,7 @@ __indirect_glVertexAttrib4dNV(GLuint index, GLdouble x, GLdouble y,
 void
 __indirect_glVertexAttrib4dvNV(GLuint index, const GLdouble * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 40;
     emit_header(gc->pc, X_GLrop_VertexAttrib4dvNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -9672,7 +9672,7 @@ void
 __indirect_glVertexAttrib4fNV(GLuint index, GLfloat x, GLfloat y, GLfloat z,
                               GLfloat w)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_VertexAttrib4fvNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -9690,7 +9690,7 @@ __indirect_glVertexAttrib4fNV(GLuint index, GLfloat x, GLfloat y, GLfloat z,
 void
 __indirect_glVertexAttrib4fvNV(GLuint index, const GLfloat * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_VertexAttrib4fvNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -9706,7 +9706,7 @@ void
 __indirect_glVertexAttrib4sNV(GLuint index, GLshort x, GLshort y, GLshort z,
                               GLshort w)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_VertexAttrib4svNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -9724,7 +9724,7 @@ __indirect_glVertexAttrib4sNV(GLuint index, GLshort x, GLshort y, GLshort z,
 void
 __indirect_glVertexAttrib4svNV(GLuint index, const GLshort * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 16;
     emit_header(gc->pc, X_GLrop_VertexAttrib4svNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -9740,7 +9740,7 @@ void
 __indirect_glVertexAttrib4ubNV(GLuint index, GLubyte x, GLubyte y, GLubyte z,
                                GLubyte w)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_VertexAttrib4ubvNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -9758,7 +9758,7 @@ __indirect_glVertexAttrib4ubNV(GLuint index, GLubyte x, GLubyte y, GLubyte z,
 void
 __indirect_glVertexAttrib4ubvNV(GLuint index, const GLubyte *v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_VertexAttrib4ubvNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4);
@@ -9773,7 +9773,7 @@ __indirect_glVertexAttrib4ubvNV(GLuint index, const GLubyte *v)
 void
 __indirect_glVertexAttribs1dvNV(GLuint index, GLsizei n, const GLdouble * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12 + __GLX_PAD((n * 8));
     if (n < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -9795,7 +9795,7 @@ __indirect_glVertexAttribs1dvNV(GLuint index, GLsizei n, const GLdouble * v)
 void
 __indirect_glVertexAttribs1fvNV(GLuint index, GLsizei n, const GLfloat * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12 + __GLX_PAD((n * 4));
     if (n < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -9817,7 +9817,7 @@ __indirect_glVertexAttribs1fvNV(GLuint index, GLsizei n, const GLfloat * v)
 void
 __indirect_glVertexAttribs1svNV(GLuint index, GLsizei n, const GLshort * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12 + __GLX_PAD((n * 2));
     if (n < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -9839,7 +9839,7 @@ __indirect_glVertexAttribs1svNV(GLuint index, GLsizei n, const GLshort * v)
 void
 __indirect_glVertexAttribs2dvNV(GLuint index, GLsizei n, const GLdouble * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12 + __GLX_PAD((n * 16));
     if (n < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -9861,7 +9861,7 @@ __indirect_glVertexAttribs2dvNV(GLuint index, GLsizei n, const GLdouble * v)
 void
 __indirect_glVertexAttribs2fvNV(GLuint index, GLsizei n, const GLfloat * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12 + __GLX_PAD((n * 8));
     if (n < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -9883,7 +9883,7 @@ __indirect_glVertexAttribs2fvNV(GLuint index, GLsizei n, const GLfloat * v)
 void
 __indirect_glVertexAttribs2svNV(GLuint index, GLsizei n, const GLshort * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12 + __GLX_PAD((n * 4));
     if (n < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -9905,7 +9905,7 @@ __indirect_glVertexAttribs2svNV(GLuint index, GLsizei n, const GLshort * v)
 void
 __indirect_glVertexAttribs3dvNV(GLuint index, GLsizei n, const GLdouble * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12 + __GLX_PAD((n * 24));
     if (n < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -9927,7 +9927,7 @@ __indirect_glVertexAttribs3dvNV(GLuint index, GLsizei n, const GLdouble * v)
 void
 __indirect_glVertexAttribs3fvNV(GLuint index, GLsizei n, const GLfloat * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12 + __GLX_PAD((n * 12));
     if (n < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -9949,7 +9949,7 @@ __indirect_glVertexAttribs3fvNV(GLuint index, GLsizei n, const GLfloat * v)
 void
 __indirect_glVertexAttribs3svNV(GLuint index, GLsizei n, const GLshort * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12 + __GLX_PAD((n * 6));
     if (n < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -9971,7 +9971,7 @@ __indirect_glVertexAttribs3svNV(GLuint index, GLsizei n, const GLshort * v)
 void
 __indirect_glVertexAttribs4dvNV(GLuint index, GLsizei n, const GLdouble * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12 + __GLX_PAD((n * 32));
     if (n < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -9993,7 +9993,7 @@ __indirect_glVertexAttribs4dvNV(GLuint index, GLsizei n, const GLdouble * v)
 void
 __indirect_glVertexAttribs4fvNV(GLuint index, GLsizei n, const GLfloat * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12 + __GLX_PAD((n * 16));
     if (n < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -10015,7 +10015,7 @@ __indirect_glVertexAttribs4fvNV(GLuint index, GLsizei n, const GLfloat * v)
 void
 __indirect_glVertexAttribs4svNV(GLuint index, GLsizei n, const GLshort * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12 + __GLX_PAD((n * 8));
     if (n < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -10037,7 +10037,7 @@ __indirect_glVertexAttribs4svNV(GLuint index, GLsizei n, const GLshort * v)
 void
 __indirect_glVertexAttribs4ubvNV(GLuint index, GLsizei n, const GLubyte *v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12 + __GLX_PAD((n * 4));
     if (n < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -10059,7 +10059,7 @@ __indirect_glVertexAttribs4ubvNV(GLuint index, GLsizei n, const GLubyte *v)
 void
 __indirect_glPointParameteriNV(GLenum pname, GLint param)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_PointParameteriNV, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&pname), 4);
@@ -10074,7 +10074,7 @@ __indirect_glPointParameteriNV(GLenum pname, GLint param)
 void
 __indirect_glPointParameterivNV(GLenum pname, const GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint compsize = __glPointParameterivNV_size(pname);
     const GLuint cmdlen = 8 + __GLX_PAD((compsize * 4));
     emit_header(gc->pc, X_GLrop_PointParameterivNV, cmdlen);
@@ -10090,7 +10090,7 @@ __indirect_glPointParameterivNV(GLenum pname, const GLint * params)
 void
 __indirect_glActiveStencilFaceEXT(GLenum face)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_ActiveStencilFaceEXT, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&face), 4);
@@ -10106,7 +10106,7 @@ __indirect_glGetProgramNamedParameterdvNV(GLuint id, GLsizei len,
                                           const GLubyte *name,
                                           GLdouble * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     const GLuint cmdlen = 8 + __GLX_PAD(len);
     if (len < 0) {
@@ -10134,7 +10134,7 @@ __indirect_glGetProgramNamedParameterfvNV(GLuint id, GLsizei len,
                                           const GLubyte *name,
                                           GLfloat * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     const GLuint cmdlen = 8 + __GLX_PAD(len);
     if (len < 0) {
@@ -10162,7 +10162,7 @@ __indirect_glProgramNamedParameter4dNV(GLuint id, GLsizei len,
                                        const GLubyte *name, GLdouble x,
                                        GLdouble y, GLdouble z, GLdouble w)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 44 + __GLX_PAD(len);
     if (len < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -10190,7 +10190,7 @@ __indirect_glProgramNamedParameter4dvNV(GLuint id, GLsizei len,
                                         const GLubyte *name,
                                         const GLdouble * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 44 + __GLX_PAD(len);
     if (len < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -10215,7 +10215,7 @@ __indirect_glProgramNamedParameter4fNV(GLuint id, GLsizei len,
                                        const GLubyte *name, GLfloat x,
                                        GLfloat y, GLfloat z, GLfloat w)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 28 + __GLX_PAD(len);
     if (len < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -10243,7 +10243,7 @@ __indirect_glProgramNamedParameter4fvNV(GLuint id, GLsizei len,
                                         const GLubyte *name,
                                         const GLfloat * v)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 28 + __GLX_PAD(len);
     if (len < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -10266,7 +10266,7 @@ __indirect_glProgramNamedParameter4fvNV(GLuint id, GLsizei len,
 void
 __indirect_glBlendEquationSeparateEXT(GLenum modeRGB, GLenum modeA)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_BlendEquationSeparateEXT, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&modeRGB), 4);
@@ -10281,7 +10281,7 @@ __indirect_glBlendEquationSeparateEXT(GLenum modeRGB, GLenum modeA)
 void
 __indirect_glBindFramebufferEXT(GLenum target, GLuint framebuffer)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_BindFramebufferEXT, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -10296,7 +10296,7 @@ __indirect_glBindFramebufferEXT(GLenum target, GLuint framebuffer)
 void
 __indirect_glBindRenderbufferEXT(GLenum target, GLuint renderbuffer)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 12;
     emit_header(gc->pc, X_GLrop_BindRenderbufferEXT, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -10311,7 +10311,7 @@ __indirect_glBindRenderbufferEXT(GLenum target, GLuint renderbuffer)
 GLenum
 __indirect_glCheckFramebufferStatusEXT(GLenum target)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     GLenum retval = (GLenum) 0;
     const GLuint cmdlen = 4;
@@ -10332,7 +10332,7 @@ __indirect_glCheckFramebufferStatusEXT(GLenum target)
 void
 __indirect_glDeleteFramebuffersEXT(GLsizei n, const GLuint * framebuffers)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8 + __GLX_PAD((n * 4));
     if (n < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -10354,7 +10354,7 @@ __indirect_glDeleteFramebuffersEXT(GLsizei n, const GLuint * framebuffers)
 void
 __indirect_glDeleteRenderbuffersEXT(GLsizei n, const GLuint * renderbuffers)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8 + __GLX_PAD((n * 4));
     if (n < 0) {
         __glXSetError(gc, GL_INVALID_VALUE);
@@ -10378,7 +10378,7 @@ __indirect_glFramebufferRenderbufferEXT(GLenum target, GLenum attachment,
                                         GLenum renderbuffertarget,
                                         GLuint renderbuffer)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_FramebufferRenderbufferEXT, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -10397,7 +10397,7 @@ __indirect_glFramebufferTexture1DEXT(GLenum target, GLenum attachment,
                                      GLenum textarget, GLuint texture,
                                      GLint level)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_FramebufferTexture1DEXT, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -10417,7 +10417,7 @@ __indirect_glFramebufferTexture2DEXT(GLenum target, GLenum attachment,
                                      GLenum textarget, GLuint texture,
                                      GLint level)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_FramebufferTexture2DEXT, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -10437,7 +10437,7 @@ __indirect_glFramebufferTexture3DEXT(GLenum target, GLenum attachment,
                                      GLenum textarget, GLuint texture,
                                      GLint level, GLint zoffset)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 28;
     emit_header(gc->pc, X_GLrop_FramebufferTexture3DEXT, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -10456,7 +10456,7 @@ __indirect_glFramebufferTexture3DEXT(GLenum target, GLenum attachment,
 void
 __indirect_glGenFramebuffersEXT(GLsizei n, GLuint * framebuffers)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     const GLuint cmdlen = 4;
     if (n < 0) {
@@ -10479,7 +10479,7 @@ __indirect_glGenFramebuffersEXT(GLsizei n, GLuint * framebuffers)
 void
 __indirect_glGenRenderbuffersEXT(GLsizei n, GLuint * renderbuffers)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     const GLuint cmdlen = 4;
     if (n < 0) {
@@ -10502,7 +10502,7 @@ __indirect_glGenRenderbuffersEXT(GLsizei n, GLuint * renderbuffers)
 void
 __indirect_glGenerateMipmapEXT(GLenum target)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 8;
     emit_header(gc->pc, X_GLrop_GenerateMipmapEXT, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -10519,7 +10519,7 @@ __indirect_glGetFramebufferAttachmentParameterivEXT(GLenum target,
                                                     GLenum pname,
                                                     GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     const GLuint cmdlen = 12;
     if (__builtin_expect(dpy != NULL, 1)) {
@@ -10542,7 +10542,7 @@ void
 __indirect_glGetRenderbufferParameterivEXT(GLenum target, GLenum pname,
                                            GLint * params)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     const GLuint cmdlen = 8;
     if (__builtin_expect(dpy != NULL, 1)) {
@@ -10563,7 +10563,7 @@ __indirect_glGetRenderbufferParameterivEXT(GLenum target, GLenum pname,
 GLboolean
 __indirect_glIsFramebufferEXT(GLuint framebuffer)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     GLboolean retval = (GLboolean) 0;
     const GLuint cmdlen = 4;
@@ -10583,7 +10583,7 @@ __indirect_glIsFramebufferEXT(GLuint framebuffer)
 GLboolean
 __indirect_glIsRenderbufferEXT(GLuint renderbuffer)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     Display *const dpy = gc->currentDpy;
     GLboolean retval = (GLboolean) 0;
     const GLuint cmdlen = 4;
@@ -10604,7 +10604,7 @@ void
 __indirect_glRenderbufferStorageEXT(GLenum target, GLenum internalformat,
                                     GLsizei width, GLsizei height)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 20;
     emit_header(gc->pc, X_GLrop_RenderbufferStorageEXT, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
@@ -10624,7 +10624,7 @@ __indirect_glBlitFramebufferEXT(GLint srcX0, GLint srcY0, GLint srcX1,
                                 GLint dstX1, GLint dstY1, GLbitfield mask,
                                 GLenum filter)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 44;
     emit_header(gc->pc, X_GLrop_BlitFramebufferEXT, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&srcX0), 4);
@@ -10649,7 +10649,7 @@ __indirect_glFramebufferTextureLayerEXT(GLenum target, GLenum attachment,
                                         GLuint texture, GLint level,
                                         GLint layer)
 {
-    __GLXcontext *const gc = __glXGetCurrentContext();
+    struct glx_context *const gc = __glXGetCurrentContext();
     const GLuint cmdlen = 24;
     emit_header(gc->pc, X_GLrop_FramebufferTextureLayerEXT, cmdlen);
     (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4);
diff --git a/src/glx/indirect.h b/src/glx/indirect.h
index b09b61aae76..36d68b066cc 100644
--- a/src/glx/indirect.h
+++ b/src/glx/indirect.h
@@ -59,15 +59,15 @@ extern HIDDEN NOINLINE CARD32 __glXReadReply( Display *dpy, size_t size,
     void * dest, GLboolean reply_is_always_array );
 
 extern HIDDEN NOINLINE void __glXReadPixelReply( Display *dpy,
-    __GLXcontext * gc, unsigned max_dim, GLint width, GLint height,
+    struct glx_context * gc, unsigned max_dim, GLint width, GLint height,
     GLint depth, GLenum format, GLenum type, void * dest,
     GLboolean dimensions_in_reply );
 
 extern HIDDEN NOINLINE FASTCALL GLubyte * __glXSetupSingleRequest(
-    __GLXcontext * gc, GLint sop, GLint cmdlen );
+    struct glx_context * gc, GLint sop, GLint cmdlen );
 
 extern HIDDEN NOINLINE FASTCALL GLubyte * __glXSetupVendorRequest(
-    __GLXcontext * gc, GLint code, GLint vop, GLint cmdlen );
+    struct glx_context * gc, GLint code, GLint vop, GLint cmdlen );
 
 extern HIDDEN void __indirect_glNewList(GLuint list, GLenum mode);
 extern HIDDEN void __indirect_glEndList(void);
diff --git a/src/glx/indirect_glx.c b/src/glx/indirect_glx.c
new file mode 100644
index 00000000000..f0598409cd6
--- /dev/null
+++ b/src/glx/indirect_glx.c
@@ -0,0 +1,456 @@
+/*
+ * Copyright © 2010 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Soft-
+ * ware"), to deal in the Software without restriction, including without
+ * limitation the rights to use, copy, modify, merge, publish, distribute,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, provided that the above copyright
+ * notice(s) and this permission notice appear in all copies of the Soft-
+ * ware and that both the above copyright notice(s) and this permission
+ * notice appear in supporting documentation.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABIL-
+ * ITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT OF THIRD PARTY
+ * RIGHTS. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR HOLDERS INCLUDED IN
+ * THIS NOTICE BE LIABLE FOR ANY CLAIM, OR ANY SPECIAL INDIRECT OR CONSE-
+ * QUENTIAL DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE,
+ * DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER
+ * TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFOR-
+ * MANCE OF THIS SOFTWARE.
+ *
+ * Except as contained in this notice, the name of a copyright holder shall
+ * not be used in advertising or otherwise to promote the sale, use or
+ * other dealings in this Software without prior written authorization of
+ * the copyright holder.
+ *
+ * Authors:
+ *   Kristian Høgsberg ([email protected])
+ */
+
+#include "glapi.h"
+#include "glxclient.h"
+
+extern struct _glapi_table *__glXNewIndirectAPI(void);
+
+/*
+** All indirect rendering contexts will share the same indirect dispatch table.
+*/
+static struct _glapi_table *IndirectAPI = NULL;
+
+static void
+indirect_destroy_context(struct glx_context *gc)
+{
+   if (!gc->imported && gc->xid)
+      glx_send_destroy_context(gc->psc->dpy, gc->xid);
+
+   __glXFreeVertexArrayState(gc);
+
+   if (gc->vendor)
+      XFree((char *) gc->vendor);
+   if (gc->renderer)
+      XFree((char *) gc->renderer);
+   if (gc->version)
+      XFree((char *) gc->version);
+   if (gc->extensions)
+      XFree((char *) gc->extensions);
+   __glFreeAttributeState(gc);
+   XFree((char *) gc->buf);
+   Xfree((char *) gc->client_state_private);
+   XFree((char *) gc);
+}
+
+static Bool
+SendMakeCurrentRequest(Display * dpy, CARD8 opcode,
+                       GLXContextID gc_id, GLXContextTag gc_tag,
+                       GLXDrawable draw, GLXDrawable read,
+                       xGLXMakeCurrentReply * reply)
+{
+   Bool ret;
+
+   LockDisplay(dpy);
+
+   if (draw == read) {
+      xGLXMakeCurrentReq *req;
+
+      GetReq(GLXMakeCurrent, req);
+      req->reqType = opcode;
+      req->glxCode = X_GLXMakeCurrent;
+      req->drawable = draw;
+      req->context = gc_id;
+      req->oldContextTag = gc_tag;
+   }
+   else {
+      struct glx_display *priv = __glXInitialize(dpy);
+
+      /* If the server can support the GLX 1.3 version, we should
+       * perfer that.  Not only that, some servers support GLX 1.3 but
+       * not the SGI extension.
+       */
+
+      if ((priv->majorVersion > 1) || (priv->minorVersion >= 3)) {
+         xGLXMakeContextCurrentReq *req;
+
+         GetReq(GLXMakeContextCurrent, req);
+         req->reqType = opcode;
+         req->glxCode = X_GLXMakeContextCurrent;
+         req->drawable = draw;
+         req->readdrawable = read;
+         req->context = gc_id;
+         req->oldContextTag = gc_tag;
+      }
+      else {
+         xGLXVendorPrivateWithReplyReq *vpreq;
+         xGLXMakeCurrentReadSGIReq *req;
+
+         GetReqExtra(GLXVendorPrivateWithReply,
+                     sz_xGLXMakeCurrentReadSGIReq -
+                     sz_xGLXVendorPrivateWithReplyReq, vpreq);
+         req = (xGLXMakeCurrentReadSGIReq *) vpreq;
+         req->reqType = opcode;
+         req->glxCode = X_GLXVendorPrivateWithReply;
+         req->vendorCode = X_GLXvop_MakeCurrentReadSGI;
+         req->drawable = draw;
+         req->readable = read;
+         req->context = gc_id;
+         req->oldContextTag = gc_tag;
+      }
+   }
+
+   ret = _XReply(dpy, (xReply *) reply, 0, False);
+
+   UnlockDisplay(dpy);
+   SyncHandle();
+
+   return ret;
+}
+
+static int
+indirect_bind_context(struct glx_context *gc, struct glx_context *old,
+		      GLXDrawable draw, GLXDrawable read)
+{
+   xGLXMakeCurrentReply reply;
+   GLXContextTag tag;
+   __GLXattribute *state;
+   Display *dpy = gc->psc->dpy;
+   int opcode = __glXSetupForCommand(dpy);
+
+   if (old != &dummyContext && !old->isDirect && old->psc->dpy == dpy)
+      tag = old->currentContextTag;
+   else
+      tag = None;
+
+   SendMakeCurrentRequest(dpy, opcode, gc->xid, tag, draw, read, &reply);
+
+   if (!IndirectAPI)
+      IndirectAPI = __glXNewIndirectAPI();
+   _glapi_set_dispatch(IndirectAPI);
+
+   gc->currentContextTag = reply.contextTag;
+   state = gc->client_state_private;
+   if (state->array_state == NULL) {
+      glGetString(GL_EXTENSIONS);
+      glGetString(GL_VERSION);
+      __glXInitVertexArrayState(gc);
+   }
+
+   return Success;
+}
+
+static void
+indirect_unbind_context(struct glx_context *gc, struct glx_context *new)
+{
+   Display *dpy = gc->psc->dpy;
+   int opcode = __glXSetupForCommand(dpy);
+   xGLXMakeCurrentReply reply;
+
+   /* We are either switching to no context, away from a indirect
+    * context to a direct context or from one dpy to another and have
+    * to send a request to the dpy to unbind the previous context.
+    */
+   if (!new || new->isDirect || new->psc->dpy != dpy)
+      SendMakeCurrentRequest(dpy, opcode, None,
+			     gc->currentContextTag, None, None, &reply);
+   gc->currentContextTag = 0;
+}
+
+static void
+indirect_wait_gl(struct glx_context *gc)
+{
+   xGLXWaitGLReq *req;
+   Display *dpy = gc->currentDpy;
+
+   /* Flush any pending commands out */
+   __glXFlushRenderBuffer(gc, gc->pc);
+
+   /* Send the glXWaitGL request */
+   LockDisplay(dpy);
+   GetReq(GLXWaitGL, req);
+   req->reqType = gc->majorOpcode;
+   req->glxCode = X_GLXWaitGL;
+   req->contextTag = gc->currentContextTag;
+   UnlockDisplay(dpy);
+   SyncHandle();
+}
+
+static void
+indirect_wait_x(struct glx_context *gc)
+{
+   xGLXWaitXReq *req;
+   Display *dpy = gc->currentDpy;
+
+   /* Flush any pending commands out */
+   __glXFlushRenderBuffer(gc, gc->pc);
+
+   LockDisplay(dpy);
+   GetReq(GLXWaitX, req);
+   req->reqType = gc->majorOpcode;
+   req->glxCode = X_GLXWaitX;
+   req->contextTag = gc->currentContextTag;
+   UnlockDisplay(dpy);
+   SyncHandle();
+}
+
+static void
+indirect_use_x_font(struct glx_context *gc,
+		    Font font, int first, int count, int listBase)
+{
+   xGLXUseXFontReq *req;
+   Display *dpy = gc->currentDpy;
+
+   /* Flush any pending commands out */
+   __glXFlushRenderBuffer(gc, gc->pc);
+
+   /* Send the glXUseFont request */
+   LockDisplay(dpy);
+   GetReq(GLXUseXFont, req);
+   req->reqType = gc->majorOpcode;
+   req->glxCode = X_GLXUseXFont;
+   req->contextTag = gc->currentContextTag;
+   req->font = font;
+   req->first = first;
+   req->count = count;
+   req->listBase = listBase;
+   UnlockDisplay(dpy);
+   SyncHandle();
+}
+
+static void
+indirect_bind_tex_image(Display * dpy,
+			GLXDrawable drawable,
+			int buffer, const int *attrib_list)
+{
+   xGLXVendorPrivateReq *req;
+   struct glx_context *gc = __glXGetCurrentContext();
+   CARD32 *drawable_ptr;
+   INT32 *buffer_ptr;
+   CARD32 *num_attrib_ptr;
+   CARD32 *attrib_ptr;
+   CARD8 opcode;
+   unsigned int i;
+
+   i = 0;
+   if (attrib_list) {
+      while (attrib_list[i * 2] != None)
+         i++;
+   }
+
+   opcode = __glXSetupForCommand(dpy);
+   if (!opcode)
+      return;
+
+   LockDisplay(dpy);
+   GetReqExtra(GLXVendorPrivate, 12 + 8 * i, req);
+   req->reqType = opcode;
+   req->glxCode = X_GLXVendorPrivate;
+   req->vendorCode = X_GLXvop_BindTexImageEXT;
+   req->contextTag = gc->currentContextTag;
+
+   drawable_ptr = (CARD32 *) (req + 1);
+   buffer_ptr = (INT32 *) (drawable_ptr + 1);
+   num_attrib_ptr = (CARD32 *) (buffer_ptr + 1);
+   attrib_ptr = (CARD32 *) (num_attrib_ptr + 1);
+
+   *drawable_ptr = drawable;
+   *buffer_ptr = buffer;
+   *num_attrib_ptr = (CARD32) i;
+
+   i = 0;
+   if (attrib_list) {
+      while (attrib_list[i * 2] != None) {
+         *attrib_ptr++ = (CARD32) attrib_list[i * 2 + 0];
+         *attrib_ptr++ = (CARD32) attrib_list[i * 2 + 1];
+         i++;
+      }
+   }
+
+   UnlockDisplay(dpy);
+   SyncHandle();
+}
+
+static void
+indirect_release_tex_image(Display * dpy, GLXDrawable drawable, int buffer)
+{
+   xGLXVendorPrivateReq *req;
+   struct glx_context *gc = __glXGetCurrentContext();
+   CARD32 *drawable_ptr;
+   INT32 *buffer_ptr;
+   CARD8 opcode;
+
+   opcode = __glXSetupForCommand(dpy);
+   if (!opcode)
+      return;
+
+   LockDisplay(dpy);
+   GetReqExtra(GLXVendorPrivate, sizeof(CARD32) + sizeof(INT32), req);
+   req->reqType = opcode;
+   req->glxCode = X_GLXVendorPrivate;
+   req->vendorCode = X_GLXvop_ReleaseTexImageEXT;
+   req->contextTag = gc->currentContextTag;
+
+   drawable_ptr = (CARD32 *) (req + 1);
+   buffer_ptr = (INT32 *) (drawable_ptr + 1);
+
+   *drawable_ptr = drawable;
+   *buffer_ptr = buffer;
+
+   UnlockDisplay(dpy);
+   SyncHandle();
+}
+
+static const struct glx_context_vtable indirect_context_vtable = {
+   indirect_destroy_context,
+   indirect_bind_context,
+   indirect_unbind_context,
+   indirect_wait_gl,
+   indirect_wait_x,
+   indirect_use_x_font,
+   indirect_bind_tex_image,
+   indirect_release_tex_image,
+};
+
+/**
+ * \todo Eliminate \c __glXInitVertexArrayState.  Replace it with a new
+ * function called \c __glXAllocateClientState that allocates the memory and
+ * does all the initialization (including the pixel pack / unpack).
+ */
+_X_HIDDEN struct glx_context *
+indirect_create_context(struct glx_screen *psc,
+			struct glx_config *mode,
+			struct glx_context *shareList, int renderType)
+{
+   struct glx_context *gc;
+   int bufSize;
+   CARD8 opcode;
+   __GLXattribute *state;
+
+   opcode = __glXSetupForCommand(psc->dpy);
+   if (!opcode) {
+      return NULL;
+   }
+
+   /* Allocate our context record */
+   gc = Xmalloc(sizeof *gc);
+   if (!gc) {
+      /* Out of memory */
+      return NULL;
+   }
+   memset(gc, 0, sizeof *gc);
+
+   glx_context_init(gc, psc, mode);
+   gc->isDirect = GL_FALSE;
+   gc->vtable = &indirect_context_vtable;
+   state = Xmalloc(sizeof(struct __GLXattributeRec));
+   if (state == NULL) {
+      /* Out of memory */
+      Xfree(gc);
+      return NULL;
+   }
+   gc->client_state_private = state;
+   memset(gc->client_state_private, 0, sizeof(struct __GLXattributeRec));
+   state->NoDrawArraysProtocol = (getenv("LIBGL_NO_DRAWARRAYS") != NULL);
+
+   /*
+    ** Create a temporary buffer to hold GLX rendering commands.  The size
+    ** of the buffer is selected so that the maximum number of GLX rendering
+    ** commands can fit in a single X packet and still have room in the X
+    ** packet for the GLXRenderReq header.
+    */
+
+   bufSize = (XMaxRequestSize(psc->dpy) * 4) - sz_xGLXRenderReq;
+   gc->buf = (GLubyte *) Xmalloc(bufSize);
+   if (!gc->buf) {
+      Xfree(gc->client_state_private);
+      Xfree(gc);
+      return NULL;
+   }
+   gc->bufSize = bufSize;
+
+   /* Fill in the new context */
+   gc->renderMode = GL_RENDER;
+
+   state->storePack.alignment = 4;
+   state->storeUnpack.alignment = 4;
+
+   gc->attributes.stackPointer = &gc->attributes.stack[0];
+
+   /*
+    ** PERFORMANCE NOTE: A mode dependent fill image can speed things up.
+    ** Other code uses the fastImageUnpack bit, but it is never set
+    ** to GL_TRUE.
+    */
+   gc->fastImageUnpack = GL_FALSE;
+   gc->fillImage = __glFillImage;
+   gc->pc = gc->buf;
+   gc->bufEnd = gc->buf + bufSize;
+   gc->isDirect = GL_FALSE;
+   if (__glXDebug) {
+      /*
+       ** Set limit register so that there will be one command per packet
+       */
+      gc->limit = gc->buf;
+   }
+   else {
+      gc->limit = gc->buf + bufSize - __GLX_BUFFER_LIMIT_SIZE;
+   }
+   gc->majorOpcode = opcode;
+
+   /*
+    ** Constrain the maximum drawing command size allowed to be
+    ** transfered using the X_GLXRender protocol request.  First
+    ** constrain by a software limit, then constrain by the protocl
+    ** limit.
+    */
+   if (bufSize > __GLX_RENDER_CMD_SIZE_LIMIT) {
+      bufSize = __GLX_RENDER_CMD_SIZE_LIMIT;
+   }
+   if (bufSize > __GLX_MAX_RENDER_CMD_SIZE) {
+      bufSize = __GLX_MAX_RENDER_CMD_SIZE;
+   }
+   gc->maxSmallRenderCommandSize = bufSize;
+   
+
+   return gc;
+}
+
+struct glx_screen_vtable indirect_screen_vtable = {
+   indirect_create_context
+};
+
+_X_HIDDEN struct glx_screen *
+indirect_create_screen(int screen, struct glx_display * priv)
+{
+   struct glx_screen *psc;
+
+   psc = Xmalloc(sizeof *psc);
+   if (psc == NULL)
+      return NULL;
+
+   memset(psc, 0, sizeof *psc);
+   glx_screen_init(psc, screen, priv);
+   psc->vtable = &indirect_screen_vtable;
+
+   return psc;
+}
diff --git a/src/glx/indirect_init.c b/src/glx/indirect_init.c
index 73ca993027b..ea051883988 100644
--- a/src/glx/indirect_init.c
+++ b/src/glx/indirect_init.c
@@ -53,13 +53,13 @@ static int NoOp(void)
  * Create and initialize a new GL dispatch table.  The table is initialized
  * with GLX indirect rendering protocol functions.
  */
-__GLapi * __glXNewIndirectAPI( void )
+struct _glapi_table *__glXNewIndirectAPI( void )
 {
-    __GLapi *glAPI;
+    struct _glapi_table *glAPI;
     GLuint entries;
 
     entries = _glapi_get_dispatch_table_size();
-    glAPI = (__GLapi *) Xmalloc(entries * sizeof(void *));
+    glAPI = (struct _glapi_table *) Xmalloc(entries * sizeof(void *));
 
     /* first, set all entries to point to no-op functions */
     {
diff --git a/src/glx/indirect_init.h b/src/glx/indirect_init.h
index 72255f13014..2ba01f56c3f 100644
--- a/src/glx/indirect_init.h
+++ b/src/glx/indirect_init.h
@@ -36,6 +36,6 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "glxclient.h"
 
-extern __GLapi *__glXNewIndirectAPI(void);
+extern struct _glapi_table *__glXNewIndirectAPI(void);
 
 #endif /* _INDIRECT_INIT_H_ */
diff --git a/src/glx/indirect_vertex_array.c b/src/glx/indirect_vertex_array.c
index ec0e654ceae..372618de4f2 100644
--- a/src/glx/indirect_vertex_array.c
+++ b/src/glx/indirect_vertex_array.c
@@ -84,9 +84,9 @@ static struct array_state *get_array_entry(const struct array_state_vector
                                            *arrays, GLenum key,
                                            unsigned index);
 static void fill_array_info_cache(struct array_state_vector *arrays);
-static GLboolean validate_mode(__GLXcontext * gc, GLenum mode);
-static GLboolean validate_count(__GLXcontext * gc, GLsizei count);
-static GLboolean validate_type(__GLXcontext * gc, GLenum type);
+static GLboolean validate_mode(struct glx_context * gc, GLenum mode);
+static GLboolean validate_count(struct glx_context * gc, GLsizei count);
+static GLboolean validate_type(struct glx_context * gc, GLenum type);
 
 
 /**
@@ -109,7 +109,7 @@ const GLuint __glXTypeSize_table[16] = {
  * __glXInitVertexArrayState().
  */
 void
-__glXFreeVertexArrayState(__GLXcontext * gc)
+__glXFreeVertexArrayState(struct glx_context * gc)
 {
    __GLXattribute *state = (__GLXattribute *) (gc->client_state_private);
    struct array_state_vector *arrays = state->array_state;
@@ -135,8 +135,8 @@ __glXFreeVertexArrayState(__GLXcontext * gc)
  * \param gc  GLX context whose vertex array state is to be initialized.
  *
  * \warning
- * This function may only be called after __GLXcontext::gl_extension_bits,
- * __GLXcontext::server_minor, and __GLXcontext::server_major have been
+ * This function may only be called after struct glx_context::gl_extension_bits,
+ * struct glx_context::server_minor, and __GLXcontext::server_major have been
  * initialized.  These values are used to determine what vertex arrays are
  * supported.
  *
@@ -144,7 +144,7 @@ __glXFreeVertexArrayState(__GLXcontext * gc)
  * Return values from malloc are not properly tested.
  */
 void
-__glXInitVertexArrayState(__GLXcontext * gc)
+__glXInitVertexArrayState(struct glx_context * gc)
 {
    __GLXattribute *state = (__GLXattribute *) (gc->client_state_private);
    struct array_state_vector *arrays;
@@ -487,7 +487,7 @@ fill_array_info_cache(struct array_state_vector *arrays)
 void
 emit_DrawArrays_none(GLenum mode, GLint first, GLsizei count)
 {
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    const __GLXattribute *state =
       (const __GLXattribute *) (gc->client_state_private);
    struct array_state_vector *arrays = state->array_state;
@@ -549,7 +549,7 @@ emit_DrawArrays_none(GLenum mode, GLint first, GLsizei count)
  * A pointer to the buffer for array data.
  */
 static GLubyte *
-emit_DrawArrays_header_old(__GLXcontext * gc,
+emit_DrawArrays_header_old(struct glx_context * gc,
                            struct array_state_vector *arrays,
                            size_t * elements_per_request,
                            unsigned int *total_requests,
@@ -658,7 +658,7 @@ emit_DrawArrays_header_old(__GLXcontext * gc,
 void
 emit_DrawArrays_old(GLenum mode, GLint first, GLsizei count)
 {
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    const __GLXattribute *state =
       (const __GLXattribute *) (gc->client_state_private);
    struct array_state_vector *arrays = state->array_state;
@@ -720,7 +720,7 @@ void
 emit_DrawElements_none(GLenum mode, GLsizei count, GLenum type,
                        const GLvoid * indices)
 {
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    const __GLXattribute *state =
       (const __GLXattribute *) (gc->client_state_private);
    struct array_state_vector *arrays = state->array_state;
@@ -787,7 +787,7 @@ void
 emit_DrawElements_old(GLenum mode, GLsizei count, GLenum type,
                       const GLvoid * indices)
 {
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    const __GLXattribute *state =
       (const __GLXattribute *) (gc->client_state_private);
    struct array_state_vector *arrays = state->array_state;
@@ -875,7 +875,7 @@ emit_DrawElements_old(GLenum mode, GLsizei count, GLenum type,
  * \c GL_TRUE if the argument is valid, \c GL_FALSE if is not.
  */
 static GLboolean
-validate_mode(__GLXcontext * gc, GLenum mode)
+validate_mode(struct glx_context * gc, GLenum mode)
 {
    switch (mode) {
    case GL_POINTS:
@@ -908,7 +908,7 @@ validate_mode(__GLXcontext * gc, GLenum mode)
  * \c GL_TRUE if the argument is valid, \c GL_FALSE if it is not.
  */
 static GLboolean
-validate_count(__GLXcontext * gc, GLsizei count)
+validate_count(struct glx_context * gc, GLsizei count)
 {
    if (count < 0) {
       __glXSetError(gc, GL_INVALID_VALUE);
@@ -927,7 +927,7 @@ validate_count(__GLXcontext * gc, GLsizei count)
  * \c GL_TRUE if the argument is valid, \c GL_FALSE if it is not.
  */
 static GLboolean
-validate_type(__GLXcontext * gc, GLenum type)
+validate_type(struct glx_context * gc, GLenum type)
 {
    switch (type) {
    case GL_UNSIGNED_INT:
@@ -944,7 +944,7 @@ validate_type(__GLXcontext * gc, GLenum type)
 void
 __indirect_glDrawArrays(GLenum mode, GLint first, GLsizei count)
 {
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    const __GLXattribute *state =
       (const __GLXattribute *) (gc->client_state_private);
    struct array_state_vector *arrays = state->array_state;
@@ -963,7 +963,7 @@ __indirect_glDrawArrays(GLenum mode, GLint first, GLsizei count)
 void
 __indirect_glArrayElement(GLint index)
 {
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    const __GLXattribute *state =
       (const __GLXattribute *) (gc->client_state_private);
    struct array_state_vector *arrays = state->array_state;
@@ -989,7 +989,7 @@ void
 __indirect_glDrawElements(GLenum mode, GLsizei count, GLenum type,
                           const GLvoid * indices)
 {
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    const __GLXattribute *state =
       (const __GLXattribute *) (gc->client_state_private);
    struct array_state_vector *arrays = state->array_state;
@@ -1011,7 +1011,7 @@ __indirect_glDrawRangeElements(GLenum mode, GLuint start, GLuint end,
                                GLsizei count, GLenum type,
                                const GLvoid * indices)
 {
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    const __GLXattribute *state =
       (const __GLXattribute *) (gc->client_state_private);
    struct array_state_vector *arrays = state->array_state;
@@ -1037,7 +1037,7 @@ void
 __indirect_glMultiDrawArraysEXT(GLenum mode, GLint * first, GLsizei * count,
                                 GLsizei primcount)
 {
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    const __GLXattribute *state =
       (const __GLXattribute *) (gc->client_state_private);
    struct array_state_vector *arrays = state->array_state;
@@ -1063,7 +1063,7 @@ __indirect_glMultiDrawElementsEXT(GLenum mode, const GLsizei * count,
                                   GLenum type, const GLvoid ** indices,
                                   GLsizei primcount)
 {
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    const __GLXattribute *state =
       (const __GLXattribute *) (gc->client_state_private);
    struct array_state_vector *arrays = state->array_state;
@@ -1119,7 +1119,7 @@ __indirect_glVertexPointer(GLint size, GLenum type, GLsizei stride,
       0, 0, X_GLrop_Vertex2dv, X_GLrop_Vertex3dv, X_GLrop_Vertex4dv
    };
    uint16_t opcode;
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    __GLXattribute *state = (__GLXattribute *) (gc->client_state_private);
    struct array_state_vector *arrays = state->array_state;
    struct array_state *a;
@@ -1164,7 +1164,7 @@ __indirect_glNormalPointer(GLenum type, GLsizei stride,
                            const GLvoid * pointer)
 {
    uint16_t opcode;
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    __GLXattribute *state = (__GLXattribute *) (gc->client_state_private);
    struct array_state_vector *arrays = state->array_state;
    struct array_state *a;
@@ -1235,7 +1235,7 @@ __indirect_glColorPointer(GLint size, GLenum type, GLsizei stride,
       0, 0, 0, X_GLrop_Color3dv, X_GLrop_Color4dv
    };
    uint16_t opcode;
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    __GLXattribute *state = (__GLXattribute *) (gc->client_state_private);
    struct array_state_vector *arrays = state->array_state;
    struct array_state *a;
@@ -1290,7 +1290,7 @@ void
 __indirect_glIndexPointer(GLenum type, GLsizei stride, const GLvoid * pointer)
 {
    uint16_t opcode;
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    __GLXattribute *state = (__GLXattribute *) (gc->client_state_private);
    struct array_state_vector *arrays = state->array_state;
    struct array_state *a;
@@ -1335,7 +1335,7 @@ __indirect_glIndexPointer(GLenum type, GLsizei stride, const GLvoid * pointer)
 void
 __indirect_glEdgeFlagPointer(GLsizei stride, const GLvoid * pointer)
 {
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    __GLXattribute *state = (__GLXattribute *) (gc->client_state_private);
    struct array_state_vector *arrays = state->array_state;
    struct array_state *a;
@@ -1397,7 +1397,7 @@ __indirect_glTexCoordPointer(GLint size, GLenum type, GLsizei stride,
    };
 
    uint16_t opcode;
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    __GLXattribute *state = (__GLXattribute *) (gc->client_state_private);
    struct array_state_vector *arrays = state->array_state;
    struct array_state *a;
@@ -1470,7 +1470,7 @@ __indirect_glSecondaryColorPointerEXT(GLint size, GLenum type, GLsizei stride,
                                       const GLvoid * pointer)
 {
    uint16_t opcode;
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    __GLXattribute *state = (__GLXattribute *) (gc->client_state_private);
    struct array_state_vector *arrays = state->array_state;
    struct array_state *a;
@@ -1530,7 +1530,7 @@ __indirect_glFogCoordPointerEXT(GLenum type, GLsizei stride,
                                 const GLvoid * pointer)
 {
    uint16_t opcode;
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    __GLXattribute *state = (__GLXattribute *) (gc->client_state_private);
    struct array_state_vector *arrays = state->array_state;
    struct array_state *a;
@@ -1577,7 +1577,7 @@ __indirect_glVertexAttribPointerARB(GLuint index, GLint size,
    static const uint16_t double_ops[5] = { 0, 4197, 4198, 4199, 4200 };
 
    uint16_t opcode;
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    __GLXattribute *state = (__GLXattribute *) (gc->client_state_private);
    struct array_state_vector *arrays = state->array_state;
    struct array_state *a;
@@ -1690,7 +1690,7 @@ __indirect_glVertexAttribPointerNV(GLuint index, GLint size,
                                    GLenum type, GLsizei stride,
                                    const GLvoid * pointer)
 {
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    GLboolean normalized = GL_FALSE;
 
 
@@ -1718,7 +1718,7 @@ __indirect_glVertexAttribPointerNV(GLuint index, GLint size,
 void
 __indirect_glClientActiveTextureARB(GLenum texture)
 {
-   __GLXcontext *const gc = __glXGetCurrentContext();
+   struct glx_context *const gc = __glXGetCurrentContext();
    __GLXattribute *const state =
       (__GLXattribute *) (gc->client_state_private);
    struct array_state_vector *const arrays = state->array_state;
diff --git a/src/glx/indirect_vertex_program.c b/src/glx/indirect_vertex_program.c
index d822a7ee56e..d955fdfa93e 100644
--- a/src/glx/indirect_vertex_program.c
+++ b/src/glx/indirect_vertex_program.c
@@ -37,7 +37,7 @@
 static void
 do_vertex_attrib_enable(GLuint index, GLboolean val)
 {
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    __GLXattribute *state = (__GLXattribute *) (gc->client_state_private);
 
    if (!__glXSetArrayEnable(state, GL_VERTEX_ATTRIB_ARRAY_POINTER_ARB,
@@ -65,7 +65,7 @@ static void
 get_parameter(unsigned opcode, unsigned size, GLenum target, GLuint index,
               void *params)
 {
-   __GLXcontext *const gc = __glXGetCurrentContext();
+   struct glx_context *const gc = __glXGetCurrentContext();
    Display *const dpy = gc->currentDpy;
    const GLuint cmdlen = 12;
 
@@ -122,7 +122,7 @@ void
 __indirect_glGetVertexAttribPointervNV(GLuint index, GLenum pname,
                                        GLvoid ** pointer)
 {
-   __GLXcontext *const gc = __glXGetCurrentContext();
+   struct glx_context *const gc = __glXGetCurrentContext();
    __GLXattribute *state = (__GLXattribute *) (gc->client_state_private);
 
    if (pname != GL_VERTEX_ATTRIB_ARRAY_POINTER_ARB) {
@@ -177,7 +177,7 @@ get_attrib_array_data(__GLXattribute * state, GLuint index, GLenum cap,
 
 
 static void
-get_vertex_attrib(__GLXcontext * gc, unsigned vop,
+get_vertex_attrib(struct glx_context * gc, unsigned vop,
                   GLuint index, GLenum pname, xReply * reply)
 {
    Display *const dpy = gc->currentDpy;
@@ -195,7 +195,7 @@ get_vertex_attrib(__GLXcontext * gc, unsigned vop,
 void
 __indirect_glGetVertexAttribivARB(GLuint index, GLenum pname, GLint * params)
 {
-   __GLXcontext *const gc = __glXGetCurrentContext();
+   struct glx_context *const gc = __glXGetCurrentContext();
    Display *const dpy = gc->currentDpy;
    __GLXattribute *state = (__GLXattribute *) (gc->client_state_private);
    xGLXSingleReply reply;
@@ -229,7 +229,7 @@ void
 __indirect_glGetVertexAttribfvARB(GLuint index, GLenum pname,
                                   GLfloat * params)
 {
-   __GLXcontext *const gc = __glXGetCurrentContext();
+   struct glx_context *const gc = __glXGetCurrentContext();
    Display *const dpy = gc->currentDpy;
    __GLXattribute *state = (__GLXattribute *) (gc->client_state_private);
    xGLXSingleReply reply;
@@ -263,7 +263,7 @@ void
 __indirect_glGetVertexAttribdvARB(GLuint index, GLenum pname,
                                   GLdouble * params)
 {
-   __GLXcontext *const gc = __glXGetCurrentContext();
+   struct glx_context *const gc = __glXGetCurrentContext();
    Display *const dpy = gc->currentDpy;
    __GLXattribute *state = (__GLXattribute *) (gc->client_state_private);
    xGLXSingleReply reply;
diff --git a/src/glx/packrender.h b/src/glx/packrender.h
index 30f6d44bbd2..4266d5cc67b 100644
--- a/src/glx/packrender.h
+++ b/src/glx/packrender.h
@@ -54,7 +54,7 @@
 
 /* Setup for all commands */
 #define __GLX_DECLARE_VARIABLES()               \
-   __GLXcontext *gc;                            \
+   struct glx_context *gc;                            \
    GLubyte *pc, *pixelHeaderPC;                 \
    GLuint compsize, cmdlen
 
diff --git a/src/glx/packsingle.h b/src/glx/packsingle.h
index f33a873f3a7..037265a7671 100644
--- a/src/glx/packsingle.h
+++ b/src/glx/packsingle.h
@@ -49,7 +49,7 @@
 
 /* Declare common variables used during a single command */
 #define __GLX_SINGLE_DECLARE_VARIABLES()         \
-   __GLXcontext *gc = __glXGetCurrentContext();  \
+   struct glx_context *gc = __glXGetCurrentContext();  \
    GLubyte *pc, *pixelHeaderPC;                  \
    GLuint compsize, cmdlen;                      \
    Display *dpy = gc->currentDpy;                \
diff --git a/src/glx/pixel.c b/src/glx/pixel.c
index d36ca31e7dd..d508d620736 100644
--- a/src/glx/pixel.c
+++ b/src/glx/pixel.c
@@ -80,7 +80,7 @@ static const GLubyte HighBitsMask[9] = {
 ** set of pixel modes that are to be done by the server.
 */
 static void
-FillBitmap(__GLXcontext * gc, GLint width, GLint height,
+FillBitmap(struct glx_context * gc, GLint width, GLint height,
            GLenum format, const GLvoid * userdata, GLubyte * destImage)
 {
    const __GLXattribute *state = gc->client_state_private;
@@ -161,7 +161,7 @@ FillBitmap(__GLXcontext * gc, GLint width, GLint height,
 ** ALIGNMENT = 1.
 */
 void
-__glFillImage(__GLXcontext * gc, GLint dim, GLint width, GLint height,
+__glFillImage(struct glx_context * gc, GLint dim, GLint width, GLint height,
               GLint depth, GLenum format, GLenum type,
               const GLvoid * userdata, GLubyte * newimage, GLubyte * modes)
 {
@@ -268,7 +268,7 @@ __glFillImage(__GLXcontext * gc, GLint dim, GLint width, GLint height,
 ** into the clients memory using the pixel store PACK modes.
 */
 static void
-EmptyBitmap(__GLXcontext * gc, GLint width, GLint height,
+EmptyBitmap(struct glx_context * gc, GLint width, GLint height,
             GLenum format, const GLubyte * sourceImage, GLvoid * userdata)
 {
    const __GLXattribute *state = gc->client_state_private;
@@ -388,7 +388,7 @@ EmptyBitmap(__GLXcontext * gc, GLint width, GLint height,
 */
 /* ARGSUSED */
 void
-__glEmptyImage(__GLXcontext * gc, GLint dim, GLint width, GLint height,
+__glEmptyImage(struct glx_context * gc, GLint dim, GLint width, GLint height,
                GLint depth, GLenum format, GLenum type,
                const GLubyte * sourceImage, GLvoid * userdata)
 {
diff --git a/src/glx/pixelstore.c b/src/glx/pixelstore.c
index dc193b9f747..1d776b817e5 100644
--- a/src/glx/pixelstore.c
+++ b/src/glx/pixelstore.c
@@ -46,7 +46,7 @@
  * \sa __indirect_glPixelStorei,  __indirect_glPixelStoref
  */
 static void
-send_PixelStore(__GLXcontext * gc, unsigned sop, GLenum pname,
+send_PixelStore(struct glx_context * gc, unsigned sop, GLenum pname,
                 const void *param)
 {
    Display *const dpy = gc->currentDpy;
@@ -67,7 +67,7 @@ send_PixelStore(__GLXcontext * gc, unsigned sop, GLenum pname,
 void
 __indirect_glPixelStoref(GLenum pname, GLfloat param)
 {
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    __GLXattribute *state = gc->client_state_private;
    Display *dpy = gc->currentDpy;
    GLuint a;
@@ -217,7 +217,7 @@ __indirect_glPixelStoref(GLenum pname, GLfloat param)
 void
 __indirect_glPixelStorei(GLenum pname, GLint param)
 {
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    __GLXattribute *state = gc->client_state_private;
    Display *dpy = gc->currentDpy;
 
diff --git a/src/glx/renderpix.c b/src/glx/renderpix.c
index 9919bbcde3b..8234bbe21f6 100644
--- a/src/glx/renderpix.c
+++ b/src/glx/renderpix.c
@@ -82,7 +82,7 @@
  * broken.
  */
 void
-__glXSendLargeImage(__GLXcontext * gc, GLint compsize, GLint dim,
+__glXSendLargeImage(struct glx_context * gc, GLint compsize, GLint dim,
                     GLint width, GLint height, GLint depth,
                     GLenum format, GLenum type, const GLvoid * src,
                     GLubyte * pc, GLubyte * modes)
diff --git a/src/glx/single2.c b/src/glx/single2.c
index a1461956b99..318d18e5554 100644
--- a/src/glx/single2.c
+++ b/src/glx/single2.c
@@ -156,7 +156,7 @@ __indirect_glGetError(void)
  * On success \c GL_TRUE is returned.  Otherwise, \c GL_FALSE is returned.
  */
 static GLboolean
-get_client_data(__GLXcontext * gc, GLenum cap, GLintptr * data)
+get_client_data(struct glx_context * gc, GLenum cap, GLintptr * data)
 {
    GLboolean retval = GL_TRUE;
    __GLXattribute *state = (__GLXattribute *) (gc->client_state_private);
@@ -646,7 +646,7 @@ version_from_string(const char *ver, int *major_version, int *minor_version)
 const GLubyte *
 __indirect_glGetString(GLenum name)
 {
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    Display *dpy = gc->currentDpy;
    GLubyte *s = NULL;
 
@@ -837,7 +837,7 @@ __indirect_glIsEnabled(GLenum cap)
 void
 __indirect_glGetPointerv(GLenum pname, void **params)
 {
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    __GLXattribute *state = (__GLXattribute *) (gc->client_state_private);
    Display *dpy = gc->currentDpy;
 
@@ -885,7 +885,7 @@ GLboolean
 __indirect_glAreTexturesResident(GLsizei n, const GLuint * textures,
                                  GLboolean * residences)
 {
-   __GLXcontext *const gc = __glXGetCurrentContext();
+   struct glx_context *const gc = __glXGetCurrentContext();
    Display *const dpy = gc->currentDpy;
    GLboolean retval = (GLboolean) 0;
    const GLuint cmdlen = 4 + __GLX_PAD((n * 4));
@@ -941,14 +941,14 @@ GLboolean
 glAreTexturesResidentEXT(GLsizei n, const GLuint * textures,
                          GLboolean * residences)
 {
-   __GLXcontext *const gc = __glXGetCurrentContext();
+   struct glx_context *const gc = __glXGetCurrentContext();
 
    if (gc->isDirect) {
       return CALL_AreTexturesResident(GET_DISPATCH(),
                                       (n, textures, residences));
    }
    else {
-      __GLXcontext *const gc = __glXGetCurrentContext();
+      struct glx_context *const gc = __glXGetCurrentContext();
       Display *const dpy = gc->currentDpy;
       GLboolean retval = (GLboolean) 0;
       const GLuint cmdlen = 4 + __GLX_PAD((n * 4));
diff --git a/src/glx/singlepix.c b/src/glx/singlepix.c
index c4010d79bd0..b61f26b2f3b 100644
--- a/src/glx/singlepix.c
+++ b/src/glx/singlepix.c
@@ -117,10 +117,10 @@ void NAME(_gloffset_GetSeparableFilter) (GLenum target, GLenum format,
                                          GLenum type, GLvoid * row,
                                          GLvoid * column, GLvoid * span)
 {
-   __GLXcontext *const gc = __glXGetCurrentContext();
+   struct glx_context *const gc = __glXGetCurrentContext();
 
 #if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)
-   if (gc->driContext) {
+   if (gc->isDirect) {
       CALL_GetSeparableFilter(GET_DISPATCH(),
                               (target, format, type, row, column, span));
       return;
diff --git a/src/glx/vertarr.c b/src/glx/vertarr.c
index 398cfb1e796..609cafac2b1 100644
--- a/src/glx/vertarr.c
+++ b/src/glx/vertarr.c
@@ -101,7 +101,7 @@ void
 __indirect_glInterleavedArrays(GLenum format, GLsizei stride,
                                const GLvoid * pointer)
 {
-   __GLXcontext *gc = __glXGetCurrentContext();
+   struct glx_context *gc = __glXGetCurrentContext();
    __GLXattribute *state = (__GLXattribute *) (gc->client_state_private);
 
 #define NONE  {0, 0, 0}
diff --git a/src/glx/xfont.c b/src/glx/xfont.c
index db3a5701100..4ca2c8f868f 100644
--- a/src/glx/xfont.c
+++ b/src/glx/xfont.c
@@ -212,7 +212,7 @@ isvalid(XFontStruct * fs, int which)
 }
 
 _X_HIDDEN void
-DRI_glXUseXFont(GLXContext CC, Font font, int first, int count, int listbase)
+DRI_glXUseXFont(struct glx_context *CC, Font font, int first, int count, int listbase)
 {
    Display *dpy;
    Window win;
diff --git a/src/mapi/glapi/gen/glX_proto_recv.py b/src/mapi/glapi/gen/glX_proto_recv.py
index 31745fcef77..887f63191de 100644
--- a/src/mapi/glapi/gen/glX_proto_recv.py
+++ b/src/mapi/glapi/gen/glX_proto_recv.py
@@ -384,9 +384,9 @@ class PrintGlxDispatchFunctions(glX_proto_common.glx_print_proto):
 		print '    int error;'
 
 		if self.do_swap:
-		    print '    __GLXcontext * const cx = __glXForceCurrent(cl, bswap_CARD32( &req->contextTag ), &error);'
+		    print '    struct glx_context * const cx = __glXForceCurrent(cl, bswap_CARD32( &req->contextTag ), &error);'
 		else:
-		    print '    __GLXcontext * const cx = __glXForceCurrent(cl, req->contextTag, &error);'
+		    print '    struct glx_context * const cx = __glXForceCurrent(cl, req->contextTag, &error);'
 
 		print ''
 		if name not in f.glx_vendorpriv_names:
diff --git a/src/mapi/glapi/gen/glX_proto_send.py b/src/mapi/glapi/gen/glX_proto_send.py
index e7322820738..0ca0ff92a6c 100644
--- a/src/mapi/glapi/gen/glX_proto_send.py
+++ b/src/mapi/glapi/gen/glX_proto_send.py
@@ -220,7 +220,7 @@ __glXReadReply( Display *dpy, size_t size, void * dest, GLboolean reply_is_alway
 }
 
 NOINLINE void
-__glXReadPixelReply( Display *dpy, __GLXcontext * gc, unsigned max_dim,
+__glXReadPixelReply( Display *dpy, struct glx_context * gc, unsigned max_dim,
     GLint width, GLint height, GLint depth, GLenum format, GLenum type,
     void * dest, GLboolean dimensions_in_reply )
 {
@@ -264,7 +264,7 @@ __glXReadPixelReply( Display *dpy, __GLXcontext * gc, unsigned max_dim,
 #define X_GLXSingle 0
 
 NOINLINE FASTCALL GLubyte *
-__glXSetupSingleRequest( __GLXcontext * gc, GLint sop, GLint cmdlen )
+__glXSetupSingleRequest( struct glx_context * gc, GLint sop, GLint cmdlen )
 {
     xGLXSingleReq * req;
     Display * const dpy = gc->currentDpy;
@@ -279,7 +279,7 @@ __glXSetupSingleRequest( __GLXcontext * gc, GLint sop, GLint cmdlen )
 }
 
 NOINLINE FASTCALL GLubyte *
-__glXSetupVendorRequest( __GLXcontext * gc, GLint code, GLint vop, GLint cmdlen )
+__glXSetupVendorRequest( struct glx_context * gc, GLint code, GLint vop, GLint cmdlen )
 {
     xGLXVendorPrivateReq * req;
     Display * const dpy = gc->currentDpy;
@@ -371,10 +371,10 @@ const GLuint __glXDefaultPixelStore[9] = { 0, 0, 0, 0, 0, 0, 0, 0, 1 };
 				print '#define %s %d' % (func.opcode_vendor_name(name), func.glx_vendorpriv)
 				print '%s gl%s(%s)' % (func.return_type, func_name, func.get_parameter_string())
 				print '{'
-				print '    __GLXcontext * const gc = __glXGetCurrentContext();'
+				print '    struct glx_context * const gc = __glXGetCurrentContext();'
 				print ''
 				print '#if defined(GLX_DIRECT_RENDERING) && !defined(GLX_USE_APPLEGL)'
-				print '    if (gc->driContext) {'
+				print '    if (gc->isDirect) {'
 				print '    %sCALL_%s(GET_DISPATCH(), (%s));' % (ret_string, func.name, func.get_called_parameter_string())
 				print '    } else'
 				print '#endif'
@@ -408,7 +408,7 @@ const GLuint __glXDefaultPixelStore[9] = { 0, 0, 0, 0, 0, 0, 0, 0, 1 };
 		print """static FASTCALL NOINLINE void
 generic_%u_byte( GLint rop, const void * ptr )
 {
-    __GLXcontext * const gc = __glXGetCurrentContext();
+    struct glx_context * const gc = __glXGetCurrentContext();
     const GLuint cmdlen = %u;
 
     emit_header(gc->pc, rop, cmdlen);
@@ -523,7 +523,7 @@ generic_%u_byte( GLint rop, const void * ptr )
 
 
 	def common_func_print_just_start(self, f, name):
-		print '    __GLXcontext * const gc = __glXGetCurrentContext();'
+		print '    struct glx_context * const gc = __glXGetCurrentContext();'
 
 		# The only reason that single and vendor private commands need
 		# a variable called 'dpy' is becuase they use the SyncHandle
@@ -971,15 +971,15 @@ extern HIDDEN NOINLINE CARD32 __glXReadReply( Display *dpy, size_t size,
     void * dest, GLboolean reply_is_always_array );
 
 extern HIDDEN NOINLINE void __glXReadPixelReply( Display *dpy,
-    __GLXcontext * gc, unsigned max_dim, GLint width, GLint height,
+    struct glx_context * gc, unsigned max_dim, GLint width, GLint height,
     GLint depth, GLenum format, GLenum type, void * dest,
     GLboolean dimensions_in_reply );
 
 extern HIDDEN NOINLINE FASTCALL GLubyte * __glXSetupSingleRequest(
-    __GLXcontext * gc, GLint sop, GLint cmdlen );
+    struct glx_context * gc, GLint sop, GLint cmdlen );
 
 extern HIDDEN NOINLINE FASTCALL GLubyte * __glXSetupVendorRequest(
-    __GLXcontext * gc, GLint code, GLint vop, GLint cmdlen );
+    struct glx_context * gc, GLint code, GLint vop, GLint cmdlen );
 """
 
 
diff --git a/src/mapi/glapi/glapi_nop.c b/src/mapi/glapi/glapi_nop.c
index 4257b1fbce6..9709551ee78 100644
--- a/src/mapi/glapi/glapi_nop.c
+++ b/src/mapi/glapi/glapi_nop.c
@@ -3,7 +3,7 @@
  * Version:  7.8
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
- * Copyright (C) 2010  VMWare, Inc.  All Rights Reserved.
+ * Copyright (C) 2010  VMware, Inc.  All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
diff --git a/src/mesa/Makefile b/src/mesa/Makefile
index 3e0f010671c..7073c92240b 100644
--- a/src/mesa/Makefile
+++ b/src/mesa/Makefile
@@ -20,6 +20,13 @@ MESA_CPPFLAGS := $(API_DEFINES)
 ES1_CPPFLAGS := -DFEATURE_ES1=1
 ES2_CPPFLAGS := -DFEATURE_ES2=1
 
+ifeq ($(MESA_LLVM),1)
+MESA_CPPFLAGS += $(LLVM_CFLAGS)
+ES1_CPPFLAGS += $(LLVM_CFLAGS)
+ES2_CPPFLAGS += $(LLVM_CFLAGS)
+endif
+
+
 include sources.mak
 
 # adjust object dirs
diff --git a/src/mesa/SConscript b/src/mesa/SConscript
index 79e9b4553b7..d31b957234b 100644
--- a/src/mesa/SConscript
+++ b/src/mesa/SConscript
@@ -80,6 +80,7 @@ if env['platform'] != 'winddk':
 		'main/pixelstore.c',
 		'main/points.c',
 		'main/polygon.c',
+		'main/querymatrix.c',
 		'main/queryobj.c',
 		'main/rastpos.c',
 		'main/readpix.c',
diff --git a/src/mesa/drivers/dri/common/dri_metaops.c b/src/mesa/drivers/dri/common/dri_metaops.c
index 86e59a8e51c..a2f404b616f 100644
--- a/src/mesa/drivers/dri/common/dri_metaops.c
+++ b/src/mesa/drivers/dri/common/dri_metaops.c
@@ -29,6 +29,7 @@
 #include "main/arbprogram.h"
 #include "main/arrayobj.h"
 #include "main/bufferobj.h"
+#include "main/context.h"
 #include "main/enable.h"
 #include "main/matrix.h"
 #include "main/texstate.h"
diff --git a/src/mesa/drivers/dri/i810/i810render.c b/src/mesa/drivers/dri/i810/i810render.c
index b543d4f012c..205f0cebc1c 100644
--- a/src/mesa/drivers/dri/i810/i810render.c
+++ b/src/mesa/drivers/dri/i810/i810render.c
@@ -37,6 +37,8 @@
 #include "main/imports.h"
 #include "main/mtypes.h"
 
+#include "math/m_xform.h"
+
 #include "tnl/t_context.h"
 
 #include "i810screen.h"
diff --git a/src/mesa/drivers/dri/i915/Makefile b/src/mesa/drivers/dri/i915/Makefile
index 71ee753748c..65fd658c047 100644
--- a/src/mesa/drivers/dri/i915/Makefile
+++ b/src/mesa/drivers/dri/i915/Makefile
@@ -56,7 +56,7 @@ C_SOURCES = \
 
 ASM_SOURCES = 
 
-DRIVER_DEFINES = -I../intel -I../intel/server -DI915 \
+DRIVER_DEFINES = -I../intel -DI915 \
 	$(shell pkg-config libdrm --atleast-version=2.3.1 \
 				&& echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP")
 
diff --git a/src/mesa/drivers/dri/i915/intel_render.c b/src/mesa/drivers/dri/i915/intel_render.c
index ec209391ab4..add0adacb56 100644
--- a/src/mesa/drivers/dri/i915/intel_render.c
+++ b/src/mesa/drivers/dri/i915/intel_render.c
@@ -37,6 +37,8 @@
 #include "main/mtypes.h"
 #include "main/enums.h"
 
+#include "math/m_xform.h"
+
 #include "tnl/t_context.h"
 #include "tnl/t_vertex.h"
 #include "tnl/t_pipeline.h"
diff --git a/src/mesa/drivers/dri/i965/Makefile b/src/mesa/drivers/dri/i965/Makefile
index 831981558d8..e381a5c714b 100644
--- a/src/mesa/drivers/dri/i965/Makefile
+++ b/src/mesa/drivers/dri/i965/Makefile
@@ -106,7 +106,7 @@ C_SOURCES = \
 
 ASM_SOURCES = 
 
-DRIVER_DEFINES = -I../intel -I../intel/server
+DRIVER_DEFINES = -I../intel
 
 INCLUDES += $(INTEL_CFLAGS)
 DRI_LIB_DEPS += $(INTEL_LIBS)
diff --git a/src/mesa/drivers/dri/i965/brw_clip_util.c b/src/mesa/drivers/dri/i965/brw_clip_util.c
index a74bbc25643..d2ac1235e46 100644
--- a/src/mesa/drivers/dri/i965/brw_clip_util.c
+++ b/src/mesa/drivers/dri/i965/brw_clip_util.c
@@ -192,11 +192,6 @@ void brw_clip_interp_vertex( struct brw_clip_compile *c,
    brw_clip_project_vertex(c, dest_ptr );
 }
 
-
-
-
-#define MAX_MRF 16
-
 void brw_clip_emit_vue(struct brw_clip_compile *c, 
 		       struct brw_indirect vert,
 		       GLboolean allocate,
diff --git a/src/mesa/drivers/dri/i965/brw_defines.h b/src/mesa/drivers/dri/i965/brw_defines.h
index 6b20a2979f8..f7a68cead7c 100644
--- a/src/mesa/drivers/dri/i965/brw_defines.h
+++ b/src/mesa/drivers/dri/i965/brw_defines.h
@@ -604,6 +604,8 @@
 #define BRW_ARF_NOTIFICATION_COUNT    0x90
 #define BRW_ARF_IP                    0xA0
 
+#define BRW_MRF_COMPR4			(1 << 7)
+
 #define BRW_AMASK   0
 #define BRW_IMASK   1
 #define BRW_LMASK   2
diff --git a/src/mesa/drivers/dri/i965/brw_eu.h b/src/mesa/drivers/dri/i965/brw_eu.h
index 31ff86cf731..ffdddd0a388 100644
--- a/src/mesa/drivers/dri/i965/brw_eu.h
+++ b/src/mesa/drivers/dri/i965/brw_eu.h
@@ -984,5 +984,7 @@ void brw_set_src1( struct brw_instruction *insn,
 
 /* brw_optimize.c */
 void brw_optimize(struct brw_compile *p);
+void brw_remove_duplicate_mrf_moves(struct brw_compile *p);
+void brw_remove_grf_to_mrf_moves(struct brw_compile *p);
 
 #endif
diff --git a/src/mesa/drivers/dri/i965/brw_optimize.c b/src/mesa/drivers/dri/i965/brw_optimize.c
index a364b158209..8aa6fb6cc6f 100644
--- a/src/mesa/drivers/dri/i965/brw_optimize.c
+++ b/src/mesa/drivers/dri/i965/brw_optimize.c
@@ -32,6 +32,594 @@
 #include "brw_defines.h"
 #include "brw_eu.h"
 
+static const struct {
+    char    *name;
+    int	    nsrc;
+    int	    ndst;
+    GLboolean is_arith;
+} inst_opcode[128] = {
+    [BRW_OPCODE_MOV] = { .name = "mov", .nsrc = 1, .ndst = 1, .is_arith = 1 },
+    [BRW_OPCODE_FRC] = { .name = "frc", .nsrc = 1, .ndst = 1, .is_arith = 1 },
+    [BRW_OPCODE_RNDU] = { .name = "rndu", .nsrc = 1, .ndst = 1, .is_arith = 1 },
+    [BRW_OPCODE_RNDD] = { .name = "rndd", .nsrc = 1, .ndst = 1, .is_arith = 1 },
+    [BRW_OPCODE_RNDE] = { .name = "rnde", .nsrc = 1, .ndst = 1, .is_arith = 1 },
+    [BRW_OPCODE_RNDZ] = { .name = "rndz", .nsrc = 1, .ndst = 1, .is_arith = 1 },
+    [BRW_OPCODE_NOT] = { .name = "not", .nsrc = 1, .ndst = 1, .is_arith = 1 },
+    [BRW_OPCODE_LZD] = { .name = "lzd", .nsrc = 1, .ndst = 1 },
+
+    [BRW_OPCODE_MUL] = { .name = "mul", .nsrc = 2, .ndst = 1, .is_arith = 1 },
+    [BRW_OPCODE_MAC] = { .name = "mac", .nsrc = 2, .ndst = 1, .is_arith = 1 },
+    [BRW_OPCODE_MACH] = { .name = "mach", .nsrc = 2, .ndst = 1, .is_arith = 1 },
+    [BRW_OPCODE_LINE] = { .name = "line", .nsrc = 2, .ndst = 1, .is_arith = 1 },
+    [BRW_OPCODE_PLN] = { .name = "pln", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_SAD2] = { .name = "sad2", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_SADA2] = { .name = "sada2", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_DP4] = { .name = "dp4", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_DPH] = { .name = "dph", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_DP3] = { .name = "dp3", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_DP2] = { .name = "dp2", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_MATH] = { .name = "math", .nsrc = 2, .ndst = 1 },
+
+    [BRW_OPCODE_AVG] = { .name = "avg", .nsrc = 2, .ndst = 1, .is_arith = 1 },
+    [BRW_OPCODE_ADD] = { .name = "add", .nsrc = 2, .ndst = 1, .is_arith = 1 },
+    [BRW_OPCODE_SEL] = { .name = "sel", .nsrc = 2, .ndst = 1, .is_arith = 1 },
+    [BRW_OPCODE_AND] = { .name = "and", .nsrc = 2, .ndst = 1, .is_arith = 1 },
+    [BRW_OPCODE_OR] = { .name = "or", .nsrc = 2, .ndst = 1, .is_arith = 1 },
+    [BRW_OPCODE_XOR] = { .name = "xor", .nsrc = 2, .ndst = 1, .is_arith = 1 },
+    [BRW_OPCODE_SHR] = { .name = "shr", .nsrc = 2, .ndst = 1, .is_arith = 1 },
+    [BRW_OPCODE_SHL] = { .name = "shl", .nsrc = 2, .ndst = 1, .is_arith = 1 },
+    [BRW_OPCODE_ASR] = { .name = "asr", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_CMP] = { .name = "cmp", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_CMPN] = { .name = "cmpn", .nsrc = 2, .ndst = 1 },
+
+    [BRW_OPCODE_SEND] = { .name = "send", .nsrc = 1, .ndst = 1 },
+    [BRW_OPCODE_NOP] = { .name = "nop", .nsrc = 0, .ndst = 0 },
+    [BRW_OPCODE_JMPI] = { .name = "jmpi", .nsrc = 1, .ndst = 0 },
+    [BRW_OPCODE_IF] = { .name = "if", .nsrc = 2, .ndst = 0 },
+    [BRW_OPCODE_IFF] = { .name = "iff", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_WHILE] = { .name = "while", .nsrc = 2, .ndst = 0 },
+    [BRW_OPCODE_ELSE] = { .name = "else", .nsrc = 2, .ndst = 0 },
+    [BRW_OPCODE_BREAK] = { .name = "break", .nsrc = 2, .ndst = 0 },
+    [BRW_OPCODE_CONTINUE] = { .name = "cont", .nsrc = 1, .ndst = 0 },
+    [BRW_OPCODE_HALT] = { .name = "halt", .nsrc = 1, .ndst = 0 },
+    [BRW_OPCODE_MSAVE] = { .name = "msave", .nsrc = 1, .ndst = 1 },
+    [BRW_OPCODE_PUSH] = { .name = "push", .nsrc = 1, .ndst = 1 },
+    [BRW_OPCODE_MRESTORE] = { .name = "mrest", .nsrc = 1, .ndst = 1 },
+    [BRW_OPCODE_POP] = { .name = "pop", .nsrc = 2, .ndst = 0 },
+    [BRW_OPCODE_WAIT] = { .name = "wait", .nsrc = 1, .ndst = 0 },
+    [BRW_OPCODE_DO] = { .name = "do", .nsrc = 0, .ndst = 0 },
+    [BRW_OPCODE_ENDIF] = { .name = "endif", .nsrc = 2, .ndst = 0 },
+};
+
+static INLINE
+GLboolean brw_is_arithmetic_inst(const struct brw_instruction *inst)
+{
+   return inst_opcode[inst->header.opcode].is_arith;
+}
+
+static const GLuint inst_stride[7] = {
+    [0] = 0,
+    [1] = 1,
+    [2] = 2,
+    [3] = 4,
+    [4] = 8,
+    [5] = 16,
+    [6] = 32
+};
+
+static const GLuint inst_type_size[8] = {
+    [BRW_REGISTER_TYPE_UD] = 4,
+    [BRW_REGISTER_TYPE_D] = 4,
+    [BRW_REGISTER_TYPE_UW] = 2,
+    [BRW_REGISTER_TYPE_W] = 2,
+    [BRW_REGISTER_TYPE_UB] = 1,
+    [BRW_REGISTER_TYPE_B] = 1,
+    [BRW_REGISTER_TYPE_F] = 4
+};
+
+static INLINE GLboolean
+brw_is_grf_written(const struct brw_instruction *inst,
+                   int reg_index, int size,
+                   int gen)
+{
+   if (inst_opcode[inst->header.opcode].ndst == 0)
+      return GL_FALSE;
+
+   if (inst->bits1.da1.dest_address_mode != BRW_ADDRESS_DIRECT)
+      if (inst->bits1.ia1.dest_reg_file == BRW_GENERAL_REGISTER_FILE)
+         return GL_TRUE;
+
+   if (inst->bits1.da1.dest_reg_file != BRW_GENERAL_REGISTER_FILE)
+      return GL_FALSE;
+
+   const int reg_start = reg_index * REG_SIZE;
+   const int reg_end = reg_start + size;
+
+   const int type_size = inst_type_size[inst->bits1.da1.dest_reg_type];
+   const int write_start = inst->bits1.da1.dest_reg_nr*REG_SIZE
+                         + inst->bits1.da1.dest_subreg_nr;
+   int length, write_end;
+
+   /* SEND is specific */
+   if (inst->header.opcode == BRW_OPCODE_SEND) {
+      if (gen >= 5)
+         length = inst->bits3.generic_gen5.response_length*REG_SIZE;
+      else 
+         length = inst->bits3.generic.response_length*REG_SIZE;
+   }
+   else {
+      length = 1 << inst->header.execution_size;
+      length *= type_size;
+      length *= inst->bits1.da1.dest_horiz_stride;
+   }
+
+   /* If the two intervals intersect, we overwrite the register */
+   write_end = write_start + length;
+   const int left = MAX2(write_start, reg_start);
+   const int right = MIN2(write_end, reg_end);
+
+   return left < right;
+}
+
+/* Specific path for message register since we need to handle the compr4 case */
+static INLINE GLboolean
+brw_is_mrf_written(const struct brw_instruction *inst, int reg_index, int size)
+{
+   if (inst_opcode[inst->header.opcode].ndst == 0)
+      return GL_FALSE;
+
+   if (inst->bits1.da1.dest_address_mode != BRW_ADDRESS_DIRECT)
+      if (inst->bits1.ia1.dest_reg_file == BRW_MESSAGE_REGISTER_FILE)
+         return GL_TRUE;
+
+   if (inst->bits1.da1.dest_reg_file != BRW_MESSAGE_REGISTER_FILE)
+      return GL_FALSE;
+
+   const int reg_start = reg_index * REG_SIZE;
+   const int reg_end = reg_start + size;
+
+   const int mrf_index = inst->bits1.da1.dest_reg_nr & 0x0f;
+   const int is_compr4 = inst->bits1.da1.dest_reg_nr & BRW_MRF_COMPR4;
+   const int type_size = inst_type_size[inst->bits1.da1.dest_reg_type];
+
+   /* We use compr4 with a size != 16 elements. Strange, we conservatively
+    * consider that we are writing the register.
+    */
+   if (is_compr4 && inst->header.execution_size != BRW_EXECUTE_16)
+      return GL_TRUE;
+
+   GLboolean is_written = GL_FALSE;
+
+   /* Here we write mrf_{i} and mrf_{i+4}. So we read two times 8 elements */
+   if (is_compr4) {
+      const int length = 8 * type_size * inst->bits1.da1.dest_horiz_stride;
+
+      /* First 8-way register */
+      const int write_start0 = mrf_index*REG_SIZE
+                             + inst->bits1.da1.dest_subreg_nr;
+      const int write_end0 = write_start0 + length;
+
+      /* Second 8-way register */
+      const int write_start1 = (mrf_index+4)*REG_SIZE
+                             + inst->bits1.da1.dest_subreg_nr;
+      const int write_end1 = write_start1 + length;
+
+      /* If the two intervals intersect, we overwrite the register */
+      const int left0 = MAX2(write_start0, reg_start);
+      const int right0 = MIN2(write_end0, reg_end);
+      const int left1 = MAX2(write_start1, reg_start);
+      const int right1 = MIN2(write_end1, reg_end);
+
+      is_written = left0 < right0 || left1 < right1;
+   }
+   else {
+      int length;
+      length = 1 << inst->header.execution_size;
+      length *= type_size;
+      length *= inst->bits1.da1.dest_horiz_stride;
+
+      /* If the two intervals intersect, we write into the register */
+      const int write_start = inst->bits1.da1.dest_reg_nr*REG_SIZE
+                            + inst->bits1.da1.dest_subreg_nr;
+      const int write_end = write_start + length;
+      const int left = MAX2(write_start, reg_start);
+      const int right = MIN2(write_end, reg_end);;
+
+      is_written = left < right;
+   }
+
+   /* SEND may perform an implicit mov to a mrf register */
+   if (is_written == GL_FALSE &&
+       inst->header.opcode == BRW_OPCODE_SEND &&
+       inst->bits1.da1.src0_reg_file != 0) {
+
+      const int mrf_start = inst->header.destreg__conditionalmod;
+      const int write_start = mrf_start * REG_SIZE;
+      const int write_end = write_start + REG_SIZE;
+      const int left = MAX2(write_start, reg_start);
+      const int right = MIN2(write_end, reg_end);;
+      is_written = left < right;
+   }
+
+   return is_written;
+}
+
+static INLINE GLboolean
+brw_is_mrf_read(const struct brw_instruction *inst,
+                int reg_index, int size, int gen)
+{
+   if (inst->header.opcode != BRW_OPCODE_SEND)
+      return GL_FALSE;
+   if (inst->bits2.da1.src0_address_mode != BRW_ADDRESS_DIRECT)
+      return GL_TRUE;
+
+   const int reg_start = reg_index*REG_SIZE;
+   const int reg_end = reg_start + size;
+
+   int length, read_start, read_end;
+   if (gen >= 5)
+      length = inst->bits3.generic_gen5.msg_length*REG_SIZE;
+   else 
+      length = inst->bits3.generic.msg_length*REG_SIZE;
+
+   /* Look if SEND uses an implicit mov. In that case, we read one less register
+    * (but we write it)
+    */
+   if (inst->bits1.da1.src0_reg_file != 0)
+      read_start = inst->header.destreg__conditionalmod;
+   else {
+      length--;
+      read_start = inst->header.destreg__conditionalmod + 1;
+   }
+   read_start *= REG_SIZE;
+   read_end = read_start + length;
+
+   const int left = MAX2(read_start, reg_start);
+   const int right = MIN2(read_end, reg_end);
+
+   return left < right;
+}
+
+static INLINE GLboolean
+brw_is_grf_read(const struct brw_instruction *inst, int reg_index, int size)
+{
+   int i, j;
+   if (inst_opcode[inst->header.opcode].nsrc == 0)
+      return GL_FALSE;
+
+   /* Look at first source. We must take into account register regions to
+    * monitor carefully the read. Note that we are a bit too conservative here
+    * since we do not take into account the fact that some complete registers
+    * may be skipped
+    */
+   if (inst_opcode[inst->header.opcode].nsrc >= 1) {
+
+      if (inst->bits2.da1.src0_address_mode != BRW_ADDRESS_DIRECT)
+         if (inst->bits1.ia1.src0_reg_file == BRW_GENERAL_REGISTER_FILE)
+            return GL_TRUE;
+      if (inst->bits1.da1.src0_reg_file != BRW_GENERAL_REGISTER_FILE)
+         return GL_FALSE;
+
+      const int reg_start = reg_index*REG_SIZE;
+      const int reg_end = reg_start + size;
+
+      /* See if at least one of this element intersects the interval */
+      const int type_size = inst_type_size[inst->bits1.da1.src0_reg_type];
+      const int elem_num = 1 << inst->header.execution_size;
+      const int width = 1 << inst->bits2.da1.src0_width;
+      const int row_num = elem_num >> inst->bits2.da1.src0_width;
+      const int hs = type_size*inst_stride[inst->bits2.da1.src0_horiz_stride];
+      const int vs = type_size*inst_stride[inst->bits2.da1.src0_vert_stride];
+      int row_start = inst->bits2.da1.src0_reg_nr*REG_SIZE
+                    + inst->bits2.da1.src0_subreg_nr;
+      for (j = 0; j < row_num; ++j) {
+         int write_start = row_start;
+         for (i = 0; i < width; ++i) {
+            const int write_end = write_start + type_size;
+            const int left = write_start > reg_start ? write_start : reg_start;
+            const int right = write_end < reg_end ? write_end : reg_end;
+            if (left < right)
+               return GL_TRUE;
+            write_start += hs;
+         }
+         row_start += vs;
+      }
+   }
+
+   /* Second src register */
+   if (inst_opcode[inst->header.opcode].nsrc >= 2) {
+
+      if (inst->bits3.da1.src1_address_mode != BRW_ADDRESS_DIRECT)
+         if (inst->bits1.ia1.src1_reg_file == BRW_GENERAL_REGISTER_FILE)
+            return GL_TRUE;
+      if (inst->bits1.da1.src1_reg_file != BRW_GENERAL_REGISTER_FILE)
+         return GL_FALSE;
+
+      const int reg_start = reg_index*REG_SIZE;
+      const int reg_end = reg_start + size;
+
+      /* See if at least one of this element intersects the interval */
+      const int type_size = inst_type_size[inst->bits1.da1.src1_reg_type];
+      const int elem_num = 1 << inst->header.execution_size;
+      const int width = 1 << inst->bits3.da1.src1_width;
+      const int row_num = elem_num >> inst->bits3.da1.src1_width;
+      const int hs = type_size*inst_stride[inst->bits3.da1.src1_horiz_stride];
+      const int vs = type_size*inst_stride[inst->bits3.da1.src1_vert_stride];
+      int row_start = inst->bits3.da1.src1_reg_nr*REG_SIZE
+                    + inst->bits3.da1.src1_subreg_nr;
+      for (j = 0; j < row_num; ++j) {
+         int write_start = row_start;
+         for (i = 0; i < width; ++i) {
+            const int write_end = write_start + type_size;
+            const int left = write_start > reg_start ? write_start : reg_start;
+            const int right = write_end < reg_end ? write_end : reg_end;
+            if (left < right)
+               return GL_TRUE;
+            write_start += hs;
+         }
+         row_start += vs;
+      }
+   }
+
+   return GL_FALSE;
+}
+
+static INLINE GLboolean
+brw_is_control_done(const struct brw_instruction *mov) {
+   return
+       mov->header.dependency_control != 0 ||
+       mov->header.thread_control != 0 ||
+       mov->header.mask_control != 0 ||
+       mov->header.saturate != 0 ||
+       mov->header.debug_control != 0;
+}
+
+static INLINE GLboolean
+brw_is_predicated(const struct brw_instruction *mov) {
+   return mov->header.predicate_control != 0;
+}
+
+static INLINE GLboolean
+brw_is_grf_to_mrf_mov(const struct brw_instruction *mov,
+                      int *mrf_index,
+                      int *grf_index,
+                      GLboolean *is_compr4)
+{
+   if (brw_is_predicated(mov) ||
+       brw_is_control_done(mov) ||
+       mov->header.debug_control != 0)
+      return GL_FALSE;
+
+   if (mov->bits1.da1.dest_address_mode != BRW_ADDRESS_DIRECT ||
+       mov->bits1.da1.dest_reg_file != BRW_MESSAGE_REGISTER_FILE ||
+       mov->bits1.da1.dest_reg_type != BRW_REGISTER_TYPE_F ||
+       mov->bits1.da1.dest_horiz_stride != BRW_HORIZONTAL_STRIDE_1 ||
+       mov->bits1.da1.dest_subreg_nr != 0)
+      return GL_FALSE;
+
+   if (mov->bits2.da1.src0_address_mode != BRW_ADDRESS_DIRECT ||
+       mov->bits1.da1.src0_reg_file != BRW_GENERAL_REGISTER_FILE ||
+       mov->bits1.da1.src0_reg_type != BRW_REGISTER_TYPE_F ||
+       mov->bits2.da1.src0_width != BRW_WIDTH_8 ||
+       mov->bits2.da1.src0_horiz_stride != BRW_HORIZONTAL_STRIDE_1 ||
+       mov->bits2.da1.src0_vert_stride != BRW_VERTICAL_STRIDE_8 ||
+       mov->bits2.da1.src0_subreg_nr != 0 ||
+       mov->bits2.da1.src0_abs != 0 ||
+       mov->bits2.da1.src0_negate != 0)
+      return GL_FALSE;
+
+   *grf_index = mov->bits2.da1.src0_reg_nr;
+   *mrf_index = mov->bits1.da1.dest_reg_nr & 0x0f;
+   *is_compr4 = (mov->bits1.da1.dest_reg_nr & BRW_MRF_COMPR4) != 0;
+   return GL_TRUE;
+}
+
+static INLINE GLboolean
+brw_is_grf_straight_write(const struct brw_instruction *inst, int grf_index)
+{
+   /* remark: no problem to predicate a SEL instruction */
+   if ((!brw_is_predicated(inst) || inst->header.opcode == BRW_OPCODE_SEL) &&
+       brw_is_control_done(inst) == GL_FALSE &&
+       inst->header.execution_size == 4 &&
+       inst->header.access_mode == BRW_ALIGN_1 &&
+       inst->bits1.da1.dest_address_mode == BRW_ADDRESS_DIRECT &&
+       inst->bits1.da1.dest_reg_file == BRW_GENERAL_REGISTER_FILE &&
+       inst->bits1.da1.dest_reg_type == BRW_REGISTER_TYPE_F &&
+       inst->bits1.da1.dest_horiz_stride == BRW_HORIZONTAL_STRIDE_1 &&
+       inst->bits1.da1.dest_reg_nr == grf_index &&
+       inst->bits1.da1.dest_subreg_nr == 0 &&
+       brw_is_arithmetic_inst(inst))
+      return GL_TRUE;
+
+   return GL_FALSE;
+}
+
+static INLINE GLboolean
+brw_inst_are_equal(const struct brw_instruction *src0,
+                   const struct brw_instruction *src1)
+{
+   const GLuint *field0 = (GLuint *) src0;
+   const GLuint *field1 = (GLuint *) src1;
+   return field0[0] == field1[0] &&
+          field0[1] == field1[1] &&
+          field0[2] == field1[2] &&
+          field0[3] == field1[3];
+}
+
+static INLINE void
+brw_inst_copy(struct brw_instruction *dst,
+              const struct brw_instruction *src)
+{
+   GLuint *field_dst = (GLuint *) dst;
+   const GLuint *field_src = (GLuint *) src;
+   field_dst[0] = field_src[0];
+   field_dst[1] = field_src[1];
+   field_dst[2] = field_src[2];
+   field_dst[3] = field_src[3];
+}
+
+static void brw_remove_inst(struct brw_compile *p, const GLboolean *removeInst)
+{
+   int i, nr_insn = 0, to = 0, from = 0;
+
+   for (from = 0; from < p->nr_insn; ++from) {
+      if (removeInst[from])
+         continue;
+      if(to != from)
+         brw_inst_copy(p->store + to, p->store + from);
+      to++;
+   }
+
+   for (i = 0; i < p->nr_insn; ++i)
+      if (removeInst[i] == GL_FALSE)
+         nr_insn++;
+   p->nr_insn = nr_insn;
+}
+
+/* The gen code emitter generates a lot of duplications in the
+ * grf-to-mrf moves, for example when texture sampling with the same
+ * coordinates from multiple textures..  Here, we monitor same mov
+ * grf-to-mrf instrutions and remove repeated ones where the operands
+ * and dst ahven't changed in between.
+ */
+void brw_remove_duplicate_mrf_moves(struct brw_compile *p)
+{
+   const int gen = p->brw->intel.gen;
+   int i, j;
+
+   GLboolean *removeInst = calloc(sizeof(GLboolean), p->nr_insn);
+   for (i = 0; i < p->nr_insn; i++) {
+      if (removeInst[i])
+         continue;
+
+      const struct brw_instruction *mov = p->store + i;
+      int mrf_index, grf_index;
+      GLboolean is_compr4;
+
+      /* Only consider _straight_ grf-to-mrf moves */
+      if (!brw_is_grf_to_mrf_mov(mov, &mrf_index, &grf_index, &is_compr4))
+         continue;
+
+      const int mrf_index0 = mrf_index;
+      const int mrf_index1 = is_compr4 ? mrf_index0+4 : mrf_index0+1;
+      const int simd16_size = 2 * REG_SIZE;
+
+      for (j = i + 1; j < p->nr_insn; j++) {
+         const struct brw_instruction *inst = p->store + j;
+
+         if (brw_inst_are_equal(mov, inst)) {
+            removeInst[j] = GL_TRUE;
+            continue;
+         }
+
+         if (brw_is_grf_written(inst, grf_index, simd16_size, gen) ||
+             brw_is_mrf_written(inst, mrf_index0, REG_SIZE) ||
+             brw_is_mrf_written(inst, mrf_index1, REG_SIZE))
+            break;
+      }
+   }
+
+   brw_remove_inst(p, removeInst);
+   free(removeInst);
+}
+
+/* Replace moves to MRFs where the value moved is the result of a
+ * normal arithmetic operation with computation right into the MRF.
+ */
+void brw_remove_grf_to_mrf_moves(struct brw_compile *p)
+{
+   int i, j, prev;
+   struct brw_context *brw = p->brw;
+   const int gen = brw->intel.gen;
+   const int simd16_size = 2*REG_SIZE;
+
+   GLboolean *removeInst = calloc(sizeof(GLboolean), p->nr_insn);
+   assert(removeInst);
+
+   for (i = 0; i < p->nr_insn; i++) {
+      if (removeInst[i])
+         continue;
+
+      struct brw_instruction *grf_inst = NULL;
+      const struct brw_instruction *mov = p->store + i;
+      int mrf_index, grf_index;
+      GLboolean is_compr4;
+
+      /* Only consider _straight_ grf-to-mrf moves */
+      if (!brw_is_grf_to_mrf_mov(mov, &mrf_index, &grf_index, &is_compr4))
+         continue;
+
+      /* Using comp4 enables a stride of 4 for this instruction */
+      const int mrf_index0 = mrf_index;
+      const int mrf_index1 = is_compr4 ? mrf_index+4 : mrf_index+1;
+
+      /* Look where the register has been set */
+      prev = i;
+      GLboolean potential_remove = GL_FALSE;
+      while (prev--) {
+
+         /* If _one_ instruction writes the grf, we try to remove the mov */
+         struct brw_instruction *inst = p->store + prev;
+         if (brw_is_grf_straight_write(inst, grf_index)) {
+            potential_remove = GL_TRUE;
+            grf_inst = inst;
+            break;
+         }
+
+      }
+
+      if (potential_remove == GL_FALSE)
+         continue;
+      removeInst[i] = GL_TRUE;
+
+      /* Monitor first the section of code between the grf computation and the
+       * mov. Here we cannot read or write both mrf and grf register
+       */
+      for (j = prev + 1; j < i; ++j) {
+         struct brw_instruction *inst = p->store + j;
+         if (removeInst[j])
+            continue;
+         if (brw_is_grf_written(inst, grf_index, simd16_size, gen)   ||
+             brw_is_grf_read(inst, grf_index, simd16_size)           ||
+             brw_is_mrf_written(inst, mrf_index0, REG_SIZE)   ||
+             brw_is_mrf_written(inst, mrf_index1, REG_SIZE)   ||
+             brw_is_mrf_read(inst, mrf_index0, REG_SIZE, gen) ||
+             brw_is_mrf_read(inst, mrf_index1, REG_SIZE, gen)) {
+            removeInst[i] = GL_FALSE;
+            break;
+         }
+      }
+
+      /* After the mov, we can read or write the mrf. If the grf is overwritten,
+       * we are done
+       */
+      for (j = i + 1; j < p->nr_insn; ++j) {
+         struct brw_instruction *inst = p->store + j;
+         if (removeInst[j])
+            continue;
+
+         if (brw_is_grf_read(inst, grf_index, simd16_size)) {
+            removeInst[i] = GL_FALSE;
+            break;
+         }
+
+         if (brw_is_grf_straight_write(inst, grf_index))
+            break;
+      }
+
+      /* Note that with the top down traversal, we can safely pacth the mov
+       * instruction
+       */
+      if (removeInst[i]) {
+         grf_inst->bits1.da1.dest_reg_file = mov->bits1.da1.dest_reg_file;
+         grf_inst->bits1.da1.dest_reg_nr = mov->bits1.da1.dest_reg_nr;
+      }
+   }
+
+   brw_remove_inst(p, removeInst);
+   free(removeInst);
+}
+
 static GLboolean
 is_single_channel_dp4(struct brw_instruction *insn)
 {
diff --git a/src/mesa/drivers/dri/i965/brw_state.h b/src/mesa/drivers/dri/i965/brw_state.h
index 40eece276b7..af08446f2d8 100644
--- a/src/mesa/drivers/dri/i965/brw_state.h
+++ b/src/mesa/drivers/dri/i965/brw_state.h
@@ -46,68 +46,68 @@ brw_add_validated_bo(struct brw_context *brw, drm_intel_bo *bo)
    }
 };
 
-const struct brw_tracked_state brw_blend_constant_color;
-const struct brw_tracked_state brw_cc_unit;
-const struct brw_tracked_state brw_check_fallback;
-const struct brw_tracked_state brw_clip_prog;
-const struct brw_tracked_state brw_clip_unit;
-const struct brw_tracked_state brw_vs_constants;
-const struct brw_tracked_state brw_wm_constants;
-const struct brw_tracked_state brw_constant_buffer;
-const struct brw_tracked_state brw_curbe_offsets;
-const struct brw_tracked_state brw_invarient_state;
-const struct brw_tracked_state brw_gs_prog;
-const struct brw_tracked_state brw_gs_unit;
-const struct brw_tracked_state brw_line_stipple;
-const struct brw_tracked_state brw_aa_line_parameters;
-const struct brw_tracked_state brw_pipelined_state_pointers;
-const struct brw_tracked_state brw_binding_table_pointers;
-const struct brw_tracked_state brw_depthbuffer;
-const struct brw_tracked_state brw_polygon_stipple_offset;
-const struct brw_tracked_state brw_polygon_stipple;
-const struct brw_tracked_state brw_program_parameters;
-const struct brw_tracked_state brw_recalculate_urb_fence;
-const struct brw_tracked_state brw_sf_prog;
-const struct brw_tracked_state brw_sf_unit;
-const struct brw_tracked_state brw_sf_vp;
-const struct brw_tracked_state brw_state_base_address;
-const struct brw_tracked_state brw_urb_fence;
-const struct brw_tracked_state brw_vertex_state;
-const struct brw_tracked_state brw_vs_surfaces;
-const struct brw_tracked_state brw_vs_prog;
-const struct brw_tracked_state brw_vs_unit;
-const struct brw_tracked_state brw_wm_input_sizes;
-const struct brw_tracked_state brw_wm_prog;
-const struct brw_tracked_state brw_wm_samplers;
-const struct brw_tracked_state brw_wm_constant_surface;
-const struct brw_tracked_state brw_wm_surfaces;
-const struct brw_tracked_state brw_wm_binding_table;
-const struct brw_tracked_state brw_wm_unit;
-
-const struct brw_tracked_state brw_psp_urb_cbs;
-
-const struct brw_tracked_state brw_pipe_control;
-
-const struct brw_tracked_state brw_drawing_rect;
-const struct brw_tracked_state brw_indices;
-const struct brw_tracked_state brw_vertices;
-const struct brw_tracked_state brw_index_buffer;
-const struct brw_tracked_state gen6_binding_table_pointers;
-const struct brw_tracked_state gen6_blend_state;
-const struct brw_tracked_state gen6_cc_state_pointers;
-const struct brw_tracked_state gen6_clip_state;
-const struct brw_tracked_state gen6_clip_vp;
-const struct brw_tracked_state gen6_color_calc_state;
-const struct brw_tracked_state gen6_depth_stencil_state;
-const struct brw_tracked_state gen6_gs_state;
-const struct brw_tracked_state gen6_sampler_state;
-const struct brw_tracked_state gen6_scissor_state;
-const struct brw_tracked_state gen6_sf_state;
-const struct brw_tracked_state gen6_sf_vp;
-const struct brw_tracked_state gen6_urb;
-const struct brw_tracked_state gen6_viewport_state;
-const struct brw_tracked_state gen6_vs_state;
-const struct brw_tracked_state gen6_wm_state;
+extern const struct brw_tracked_state brw_blend_constant_color;
+extern const struct brw_tracked_state brw_cc_unit;
+extern const struct brw_tracked_state brw_check_fallback;
+extern const struct brw_tracked_state brw_clip_prog;
+extern const struct brw_tracked_state brw_clip_unit;
+extern const struct brw_tracked_state brw_vs_constants;
+extern const struct brw_tracked_state brw_wm_constants;
+extern const struct brw_tracked_state brw_constant_buffer;
+extern const struct brw_tracked_state brw_curbe_offsets;
+extern const struct brw_tracked_state brw_invarient_state;
+extern const struct brw_tracked_state brw_gs_prog;
+extern const struct brw_tracked_state brw_gs_unit;
+extern const struct brw_tracked_state brw_line_stipple;
+extern const struct brw_tracked_state brw_aa_line_parameters;
+extern const struct brw_tracked_state brw_pipelined_state_pointers;
+extern const struct brw_tracked_state brw_binding_table_pointers;
+extern const struct brw_tracked_state brw_depthbuffer;
+extern const struct brw_tracked_state brw_polygon_stipple_offset;
+extern const struct brw_tracked_state brw_polygon_stipple;
+extern const struct brw_tracked_state brw_program_parameters;
+extern const struct brw_tracked_state brw_recalculate_urb_fence;
+extern const struct brw_tracked_state brw_sf_prog;
+extern const struct brw_tracked_state brw_sf_unit;
+extern const struct brw_tracked_state brw_sf_vp;
+extern const struct brw_tracked_state brw_state_base_address;
+extern const struct brw_tracked_state brw_urb_fence;
+extern const struct brw_tracked_state brw_vertex_state;
+extern const struct brw_tracked_state brw_vs_surfaces;
+extern const struct brw_tracked_state brw_vs_prog;
+extern const struct brw_tracked_state brw_vs_unit;
+extern const struct brw_tracked_state brw_wm_input_sizes;
+extern const struct brw_tracked_state brw_wm_prog;
+extern const struct brw_tracked_state brw_wm_samplers;
+extern const struct brw_tracked_state brw_wm_constant_surface;
+extern const struct brw_tracked_state brw_wm_surfaces;
+extern const struct brw_tracked_state brw_wm_binding_table;
+extern const struct brw_tracked_state brw_wm_unit;
+
+extern const struct brw_tracked_state brw_psp_urb_cbs;
+
+extern const struct brw_tracked_state brw_pipe_control;
+
+extern const struct brw_tracked_state brw_drawing_rect;
+extern const struct brw_tracked_state brw_indices;
+extern const struct brw_tracked_state brw_vertices;
+extern const struct brw_tracked_state brw_index_buffer;
+extern const struct brw_tracked_state gen6_binding_table_pointers;
+extern const struct brw_tracked_state gen6_blend_state;
+extern const struct brw_tracked_state gen6_cc_state_pointers;
+extern const struct brw_tracked_state gen6_clip_state;
+extern const struct brw_tracked_state gen6_clip_vp;
+extern const struct brw_tracked_state gen6_color_calc_state;
+extern const struct brw_tracked_state gen6_depth_stencil_state;
+extern const struct brw_tracked_state gen6_gs_state;
+extern const struct brw_tracked_state gen6_sampler_state;
+extern const struct brw_tracked_state gen6_scissor_state;
+extern const struct brw_tracked_state gen6_sf_state;
+extern const struct brw_tracked_state gen6_sf_vp;
+extern const struct brw_tracked_state gen6_urb;
+extern const struct brw_tracked_state gen6_viewport_state;
+extern const struct brw_tracked_state gen6_vs_state;
+extern const struct brw_tracked_state gen6_wm_state;
 
 /***********************************************************************
  * brw_state.c
diff --git a/src/mesa/drivers/dri/i965/brw_util.c b/src/mesa/drivers/dri/i965/brw_util.c
index 1db2a210d45..e878da3850d 100644
--- a/src/mesa/drivers/dri/i965/brw_util.c
+++ b/src/mesa/drivers/dri/i965/brw_util.c
@@ -30,6 +30,8 @@
   */
          
 
+#include <assert.h>
+
 #include "main/mtypes.h"
 #include "program/prog_parameter.h"
 #include "brw_util.h"
diff --git a/src/mesa/drivers/dri/i965/brw_vs_emit.c b/src/mesa/drivers/dri/i965/brw_vs_emit.c
index a1bee2e44ab..b6b558e9a69 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_emit.c
@@ -44,6 +44,7 @@ static GLboolean
 brw_vs_arg_can_be_immediate(enum prog_opcode opcode, int arg)
 {
    int opcode_array[] = {
+      [OPCODE_MOV] = 1,
       [OPCODE_ADD] = 2,
       [OPCODE_CMP] = 3,
       [OPCODE_DP3] = 2,
diff --git a/src/mesa/drivers/dri/i965/brw_wm_emit.c b/src/mesa/drivers/dri/i965/brw_wm_emit.c
index 323cfac8fa7..d9fa2e63354 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_emit.c
@@ -1283,7 +1283,7 @@ void emit_fb_write(struct brw_wm_compile *c,
 	  * + 1 for the second half we get destination + 4.
 	  */
 	 brw_MOV(p,
-		 brw_message_reg(nr + channel + (1 << 7)),
+		 brw_message_reg(nr + channel + BRW_MRF_COMPR4),
 		 arg0[channel]);
       } else {
 	 /*  mov (8) m2.0<1>:ud   r28.0<8;8,1>:ud  { Align1 } */
@@ -1712,12 +1712,20 @@ void brw_wm_emit( struct brw_wm_compile *c )
 		      inst->dst[i]->spill_slot);
    }
 
+   /* Only properly tested on ILK */
+   if (p->brw->intel.gen == 5) {
+     brw_remove_duplicate_mrf_moves(p);
+     if (c->dispatch_width == 16)
+	brw_remove_grf_to_mrf_moves(p);
+   }
+
    if (INTEL_DEBUG & DEBUG_WM) {
       int i;
 
-      printf("wm-native:\n");
-      for (i = 0; i < p->nr_insn; i++)
+     printf("wm-native:\n");
+     for (i = 0; i < p->nr_insn; i++)
 	 brw_disasm(stderr, &p->store[i], p->brw->intel.gen);
       printf("\n");
    }
 }
+
diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c
index 5f2035d79c9..e19f44035fd 100644
--- a/src/mesa/drivers/dri/intel/intel_context.c
+++ b/src/mesa/drivers/dri/intel/intel_context.c
@@ -29,6 +29,7 @@
 #include "main/glheader.h"
 #include "main/context.h"
 #include "main/extensions.h"
+#include "main/fbobject.h"
 #include "main/framebuffer.h"
 #include "main/imports.h"
 #include "main/points.h"
@@ -39,8 +40,6 @@
 #include "drivers/common/driverfuncs.h"
 #include "drivers/common/meta.h"
 
-#include "i830_dri.h"
-
 #include "intel_chipset.h"
 #include "intel_buffers.h"
 #include "intel_tex.h"
@@ -420,7 +419,7 @@ intel_prepare_render(struct intel_context *intel)
    __DRIdrawable *drawable;
 
    drawable = driContext->driDrawablePriv;
-   if (drawable->dri2.stamp != driContext->dri2.draw_stamp) {
+   if (drawable && drawable->dri2.stamp != driContext->dri2.draw_stamp) {
       if (drawable->lastStamp != drawable->dri2.stamp)
 	 intel_update_renderbuffers(driContext, drawable);
       intel_draw_buffer(&intel->ctx, intel->ctx.DrawBuffer);
@@ -428,7 +427,7 @@ intel_prepare_render(struct intel_context *intel)
    }
 
    drawable = driContext->driReadablePriv;
-   if (drawable->dri2.stamp != driContext->dri2.read_stamp) {
+   if (drawable && drawable->dri2.stamp != driContext->dri2.read_stamp) {
       if (drawable->lastStamp != drawable->dri2.stamp)
 	 intel_update_renderbuffers(driContext, drawable);
       driContext->dri2.read_stamp = drawable->dri2.stamp;
@@ -613,6 +612,7 @@ intelInitContext(struct intel_context *intel,
    __DRIscreen *sPriv = driContextPriv->driScreenPriv;
    struct intel_screen *intelScreen = sPriv->private;
    int bo_reuse_mode;
+   __GLcontextModes visual;
 
    /* we can't do anything without a connection to the device */
    if (intelScreen->bufmgr == NULL)
@@ -624,6 +624,11 @@ intelInitContext(struct intel_context *intel,
       functions->Viewport = intel_viewport;
    }
 
+   if (mesaVis == NULL) {
+      memset(&visual, 0, sizeof visual);
+      mesaVis = &visual;
+   }
+
    if (!_mesa_initialize_context_for_api(&intel->ctx, api, mesaVis, shareCtx,
 					 functions, (void *) intel)) {
       printf("%s: failed to init mesa context\n", __FUNCTION__);
@@ -890,14 +895,21 @@ intelMakeCurrent(__DRIcontext * driContextPriv,
    }
 
    if (driContextPriv) {
-      struct gl_framebuffer *fb = driDrawPriv->driverPrivate;
-      struct gl_framebuffer *readFb = driReadPriv->driverPrivate;
+      struct gl_framebuffer *fb, *readFb;
+      
+      if (driDrawPriv == NULL && driReadPriv == NULL) {
+	 fb = _mesa_get_incomplete_framebuffer();
+	 readFb = _mesa_get_incomplete_framebuffer();
+      } else {
+	 fb = driDrawPriv->driverPrivate;
+	 readFb = driReadPriv->driverPrivate;
+	 driContextPriv->dri2.draw_stamp = driDrawPriv->dri2.stamp - 1;
+	 driContextPriv->dri2.read_stamp = driReadPriv->dri2.stamp - 1;
+      }
 
-      driContextPriv->dri2.draw_stamp = driDrawPriv->dri2.stamp - 1;
-      driContextPriv->dri2.read_stamp = driReadPriv->dri2.stamp - 1;
       intel_prepare_render(intel);
       _mesa_make_current(&intel->ctx, fb, readFb);
-
+      
       /* We do this in intel_prepare_render() too, but intel->ctx.DrawBuffer
        * is NULL at that point.  We can't call _mesa_makecurrent()
        * first, since we need the buffer size for the initial
diff --git a/src/mesa/drivers/dri/intel/intel_extensions_es2.c b/src/mesa/drivers/dri/intel/intel_extensions_es2.c
index baf8e130010..de34bbb2aec 100644
--- a/src/mesa/drivers/dri/intel/intel_extensions_es2.c
+++ b/src/mesa/drivers/dri/intel/intel_extensions_es2.c
@@ -28,7 +28,6 @@
 #include "main/extensions.h"
 
 #include "intel_extensions.h"
-#include "utils.h"
 
 static const char *es2_extensions[] = {
    /* Used by mesa internally (cf all_mesa_extensions in ../common/utils.c) */
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
index 0e2fe893fed..02c0ffce31d 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
@@ -45,6 +45,7 @@
 #include "main/attrib.h"
 #include "main/enable.h"
 #include "main/viewport.h"
+#include "main/context.h"
 #include "swrast/swrast.h"
 
 #include "intel_screen.h"
diff --git a/src/mesa/drivers/dri/intel/intel_regions.c b/src/mesa/drivers/dri/intel/intel_regions.c
index fe4de189600..680d18ba299 100644
--- a/src/mesa/drivers/dri/intel/intel_regions.c
+++ b/src/mesa/drivers/dri/intel/intel_regions.c
@@ -155,6 +155,9 @@ intel_region_alloc_internal(struct intel_context *intel,
    }
 
    region = calloc(sizeof(*region), 1);
+   if (region == NULL)
+      return region;
+
    region->cpp = cpp;
    region->width = width;
    region->height = height;
@@ -189,6 +192,9 @@ intel_region_alloc(struct intel_context *intel,
 
    region = intel_region_alloc_internal(intel, cpp, width, height,
 					aligned_pitch / cpp, buffer);
+   if (region == NULL)
+      return region;
+
    region->tiling = tiling;
 
    return region;
diff --git a/src/mesa/drivers/dri/intel/intel_tex_copy.c b/src/mesa/drivers/dri/intel/intel_tex_copy.c
index 224b506c05b..6efb2ddc553 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_copy.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_copy.c
@@ -102,7 +102,7 @@ do_copy_texsubimage(struct intel_context *intel,
    GLcontext *ctx = &intel->ctx;
    const struct intel_region *src = get_teximage_source(intel, internalFormat);
 
-   if (!intelImage->mt || !src) {
+   if (!intelImage->mt || !src || !src->buffer) {
       if (INTEL_DEBUG & DEBUG_FALLBACKS)
 	 fprintf(stderr, "%s fail %p %p (0x%08x)\n",
 		 __FUNCTION__, intelImage->mt, src, internalFormat);
diff --git a/src/mesa/drivers/dri/intel/intel_tex_format.c b/src/mesa/drivers/dri/intel/intel_tex_format.c
index 5f813c0efa2..e03b203fb40 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_format.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_format.c
@@ -19,7 +19,6 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat,
                          GLenum format, GLenum type)
 {
    struct intel_context *intel = intel_context(ctx);
-   const GLboolean do32bpt = (intel->ctx.Visual.rgbBits >= 24);
 
 #if 0
    printf("%s intFmt=0x%x format=0x%x type=0x%x\n",
@@ -30,39 +29,28 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat,
    case 4:
    case GL_RGBA:
    case GL_COMPRESSED_RGBA:
-      if (format == GL_BGRA) {
-         if (type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV) {
-            return MESA_FORMAT_ARGB8888;
-         }
-         else if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV) {
-            return MESA_FORMAT_ARGB4444;
-         }
-         else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV) {
-            return MESA_FORMAT_ARGB1555;
-         }
-      }
-      return do32bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB4444;
+      if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV)
+	 return MESA_FORMAT_ARGB4444;
+      else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV)
+	 return MESA_FORMAT_ARGB1555;
+      else
+	 return MESA_FORMAT_ARGB8888;
 
    case 3:
    case GL_RGB:
    case GL_COMPRESSED_RGB:
-      if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) {
-         return MESA_FORMAT_RGB565;
-      }
-      if (do32bpt) {
-	 if (intel->has_xrgb_textures)
-	    return MESA_FORMAT_XRGB8888;
-	 else
-	    return MESA_FORMAT_ARGB8888;
-      } else {
+      if (type == GL_UNSIGNED_SHORT_5_6_5)
 	 return MESA_FORMAT_RGB565;
-      }
+      else if (intel->has_xrgb_textures)
+	 return MESA_FORMAT_XRGB8888;
+      else
+	 return MESA_FORMAT_ARGB8888;
 
    case GL_RGBA8:
    case GL_RGB10_A2:
    case GL_RGBA12:
    case GL_RGBA16:
-      return do32bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB4444;
+      return MESA_FORMAT_ARGB8888;
 
    case GL_RGBA4:
    case GL_RGBA2:
diff --git a/src/mesa/drivers/dri/intel/server/i830_dri.h b/src/mesa/drivers/dri/intel/server/i830_dri.h
deleted file mode 100644
index def049e7a6b..00000000000
--- a/src/mesa/drivers/dri/intel/server/i830_dri.h
+++ /dev/null
@@ -1,62 +0,0 @@
-/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_dri.h,v 1.6 2003/09/28 20:15:59 alanh Exp $ */
-
-#ifndef _I830_DRI_H
-#define _I830_DRI_H
-
-#include "xf86drm.h"
-
-#define I830_MAX_DRAWABLES 256
-
-#define I830_MAJOR_VERSION 1
-#define I830_MINOR_VERSION 9
-#define I830_PATCHLEVEL 0
-
-#define I830_REG_SIZE 0x80000
-
-typedef struct _I830DRIRec {
-   drm_handle_t regs;
-   drmSize regsSize;
-
-   drmSize unused1; /* backbufferSize */
-   drm_handle_t unused2; /* backbuffer */
-
-   drmSize unused3; /* depthbufferSize */
-   drm_handle_t unused4; /* depthbuffer */
-
-   drmSize unused5; /* rotatedSize */
-   drm_handle_t unused6; /* rotatedbuffer */
-
-   drm_handle_t unused7; /* textures */
-   int unused8; /* textureSize */
-
-   drm_handle_t unused9; /* agp_buffers */
-   drmSize unused10; /* agp_buf_size */
-
-   int deviceID;
-   int width;
-   int height;
-   int mem;
-   int cpp;
-   int bitsPerPixel;
-
-   int unused11[8]; /* was front/back/depth/rotated offset/pitch */
-
-   int unused12; /* logTextureGranularity */
-   int unused13; /* textureOffset */
-
-   int irq;
-   int sarea_priv_offset;
-} I830DRIRec, *I830DRIPtr;
-
-typedef struct {
-   /* Nothing here yet */
-   int dummy;
-} I830ConfigPrivRec, *I830ConfigPrivPtr;
-
-typedef struct {
-   /* Nothing here yet */
-   int dummy;
-} I830DRIContextRec, *I830DRIContextPtr;
-
-
-#endif
diff --git a/src/mesa/drivers/dri/intel/server/intel.h b/src/mesa/drivers/dri/intel/server/intel.h
deleted file mode 100644
index 6ea72499c1c..00000000000
--- a/src/mesa/drivers/dri/intel/server/intel.h
+++ /dev/null
@@ -1,331 +0,0 @@
-#ifndef _INTEL_H_
-#define _INTEL_H_
-
-#include "xf86drm.h"		/* drm_handle_t, etc */
-
-/* Intel */
-#ifndef PCI_CHIP_I810
-#define PCI_CHIP_I810              0x7121
-#define PCI_CHIP_I810_DC100        0x7123
-#define PCI_CHIP_I810_E            0x7125
-#define PCI_CHIP_I815              0x1132
-#define PCI_CHIP_I810_BRIDGE       0x7120
-#define PCI_CHIP_I810_DC100_BRIDGE 0x7122
-#define PCI_CHIP_I810_E_BRIDGE     0x7124
-#define PCI_CHIP_I815_BRIDGE       0x1130
-#endif
-
-#define PCI_CHIP_845_G			0x2562
-#define PCI_CHIP_I830_M			0x3577
-
-#ifndef PCI_CHIP_I855_GM
-#define PCI_CHIP_I855_GM	   0x3582
-#define PCI_CHIP_I855_GM_BRIDGE	   0x3580
-#endif
-
-#ifndef PCI_CHIP_I865_G
-#define PCI_CHIP_I865_G		   0x2572
-#define PCI_CHIP_I865_G_BRIDGE	   0x2570
-#endif
-
-#ifndef PCI_CHIP_I915_G
-#define PCI_CHIP_I915_G		   0x2582
-#define PCI_CHIP_I915_G_BRIDGE	   0x2580
-#endif
-
-#ifndef PCI_CHIP_I915_GM
-#define PCI_CHIP_I915_GM	   0x2592
-#define PCI_CHIP_I915_GM_BRIDGE	   0x2590
-#endif
-
-#ifndef PCI_CHIP_E7221_G
-#define PCI_CHIP_E7221_G	   0x258A
-/* Same as I915_G_BRIDGE */
-#define PCI_CHIP_E7221_G_BRIDGE	   0x2580
-#endif
-
-#ifndef PCI_CHIP_I945_G
-#define PCI_CHIP_I945_G        0x2772
-#define PCI_CHIP_I945_G_BRIDGE 0x2770
-#endif
-
-#ifndef PCI_CHIP_I945_GM
-#define PCI_CHIP_I945_GM        0x27A2
-#define PCI_CHIP_I945_GM_BRIDGE 0x27A0
-#endif
-
-#define IS_I810(pI810) (pI810->Chipset == PCI_CHIP_I810 ||	\
-			pI810->Chipset == PCI_CHIP_I810_DC100 || \
-			pI810->Chipset == PCI_CHIP_I810_E)
-#define IS_I815(pI810) (pI810->Chipset == PCI_CHIP_I815)
-#define IS_I830(pI810) (pI810->Chipset == PCI_CHIP_I830_M)
-#define IS_845G(pI810) (pI810->Chipset == PCI_CHIP_845_G)
-#define IS_I85X(pI810)  (pI810->Chipset == PCI_CHIP_I855_GM)
-#define IS_I852(pI810)  (pI810->Chipset == PCI_CHIP_I855_GM && (pI810->variant == I852_GM || pI810->variant == I852_GME))
-#define IS_I855(pI810)  (pI810->Chipset == PCI_CHIP_I855_GM && (pI810->variant == I855_GM || pI810->variant == I855_GME))
-#define IS_I865G(pI810) (pI810->Chipset == PCI_CHIP_I865_G)
-
-#define IS_I915G(pI810) (pI810->Chipset == PCI_CHIP_I915_G || pI810->Chipset == PCI_CHIP_E7221_G)
-#define IS_I915GM(pI810) (pI810->Chipset == PCI_CHIP_I915_GM)
-#define IS_I945G(pI810) (pI810->Chipset == PCI_CHIP_I945_G)
-#define IS_I945GM(pI810) (pI810->Chipset == PCI_CHIP_I945_GM)
-#define IS_I9XX(pI810) (IS_I915G(pI810) || IS_I915GM(pI810) || IS_I945G(pI810) || IS_I945GM(pI810))
-
-#define IS_MOBILE(pI810) (IS_I830(pI810) || IS_I85X(pI810) || IS_I915GM(pI810) || IS_I945GM(pI810))
-
-#define I830_GMCH_CTRL		0x52
-
-#define I830_GMCH_MEM_MASK      0x1
-#define I830_GMCH_MEM_64M       0x1
-#define I830_GMCH_MEM_128M      0
-
-#define I830_GMCH_GMS_MASK			0x70
-#define I830_GMCH_GMS_DISABLED		0x00
-#define I830_GMCH_GMS_LOCAL			0x10
-#define I830_GMCH_GMS_STOLEN_512	0x20
-#define I830_GMCH_GMS_STOLEN_1024	0x30
-#define I830_GMCH_GMS_STOLEN_8192	0x40
-
-#define I855_GMCH_GMS_MASK			(0x7 << 4)
-#define I855_GMCH_GMS_DISABLED			0x00
-#define I855_GMCH_GMS_STOLEN_1M			(0x1 << 4)
-#define I855_GMCH_GMS_STOLEN_4M			(0x2 << 4)
-#define I855_GMCH_GMS_STOLEN_8M			(0x3 << 4)
-#define I855_GMCH_GMS_STOLEN_16M		(0x4 << 4)
-#define I855_GMCH_GMS_STOLEN_32M		(0x5 << 4)
-#define I915G_GMCH_GMS_STOLEN_48M		(0x6 << 4)
-#define I915G_GMCH_GMS_STOLEN_64M		(0x7 << 4)
-
-typedef unsigned char Bool;
-#define TRUE 1
-#define FALSE 0
-
-#define PIPE_NONE	0<<0
-#define PIPE_CRT	1<<0
-#define PIPE_TV		1<<1
-#define PIPE_DFP	1<<2
-#define PIPE_LFP	1<<3
-#define PIPE_CRT2	1<<4
-#define PIPE_TV2	1<<5
-#define PIPE_DFP2	1<<6
-#define PIPE_LFP2	1<<7
-
-typedef struct _I830MemPool *I830MemPoolPtr;
-typedef struct _I830MemRange *I830MemRangePtr;
-typedef struct _I830MemRange {
-   long Start;
-   long End;
-   long Size;
-   unsigned long Physical;
-   unsigned long Offset;		/* Offset of AGP-allocated portion */
-   unsigned long Alignment;
-   drm_handle_t Key;
-   unsigned long Pitch; // add pitch
-   I830MemPoolPtr Pool;
-} I830MemRange;
-
-typedef struct _I830MemPool {
-   I830MemRange Total;
-   I830MemRange Free;
-   I830MemRange Fixed;
-   I830MemRange Allocated;
-} I830MemPool;
-
-typedef struct {
-   int tail_mask;
-   I830MemRange mem;
-   unsigned char *virtual_start;
-   int head;
-   int tail;
-   int space;
-} I830RingBuffer;
-
-typedef struct _I830Rec {
-   unsigned char *MMIOBase;
-   unsigned char *FbBase;
-   int cpp;
-   uint32_t aper_size;
-   unsigned int bios_version;
-
-   /* These are set in PreInit and never changed. */
-   long FbMapSize;
-   long TotalVideoRam;
-   I830MemRange StolenMemory;		/* pre-allocated memory */
-   long BIOSMemorySize;			/* min stolen pool size */
-   int BIOSMemSizeLoc;
-
-   /* These change according to what has been allocated. */
-   long FreeMemory;
-   I830MemRange MemoryAperture;
-   I830MemPool StolenPool;
-   long allocatedMemory;
-
-   /* Regions allocated either from the above pools, or from agpgart. */
-   /* for single and dual head configurations */
-   I830MemRange FrontBuffer;
-   I830MemRange FrontBuffer2;
-   I830MemRange Scratch;
-   I830MemRange Scratch2;
-
-   I830RingBuffer *LpRing;
-
-   I830MemRange BackBuffer;
-   I830MemRange DepthBuffer;
-   I830MemRange TexMem;
-   int TexGranularity;
-   I830MemRange ContextMem;
-   int drmMinor;
-   Bool have3DWindows;
-
-   Bool NeedRingBufferLow;
-   Bool allowPageFlip;
-   Bool disableTiling;
-
-   int Chipset;
-   unsigned long LinearAddr;
-   unsigned long MMIOAddr;
-
-   drmSize           registerSize;     /**< \brief MMIO register map size */
-   drm_handle_t         registerHandle;   /**< \brief MMIO register map handle */
-  //   IOADDRESS ioBase;
-   int               irq;              /**< \brief IRQ number */
-   int GttBound;
-
-   drm_handle_t ring_map;
-   unsigned int Fence[8];
-
-} I830Rec;
-
-/*
- * 12288 is set as the maximum, chosen because it is enough for
- * 1920x1440@32bpp with a 2048 pixel line pitch with some to spare.
- */
-#define I830_MAXIMUM_VBIOS_MEM		12288
-#define I830_DEFAULT_VIDEOMEM_2D	(MB(32) / 1024)
-#define I830_DEFAULT_VIDEOMEM_3D	(MB(64) / 1024)
-
-/* Flags for memory allocation function */
-#define FROM_ANYWHERE			0x00000000
-#define FROM_POOL_ONLY			0x00000001
-#define FROM_NEW_ONLY			0x00000002
-#define FROM_MASK			0x0000000f
-
-#define ALLOCATE_AT_TOP			0x00000010
-#define ALLOCATE_AT_BOTTOM		0x00000020
-#define FORCE_GAPS			0x00000040
-
-#define NEED_PHYSICAL_ADDR		0x00000100
-#define ALIGN_BOTH_ENDS			0x00000200
-#define FORCE_LOW			0x00000400
-
-#define ALLOC_NO_TILING			0x00001000
-#define ALLOC_INITIAL			0x00002000
-
-#define ALLOCATE_DRY_RUN		0x80000000
-
-/* Chipset registers for VIDEO BIOS memory RW access */
-#define _855_DRAM_RW_CONTROL 0x58
-#define _845_DRAM_RW_CONTROL 0x90
-#define DRAM_WRITE    0x33330000
-
-#define KB(x) ((x) * 1024)
-#define MB(x) ((x) * KB(1024))
-
-#define GTT_PAGE_SIZE			KB(4)
-#define ROUND_TO(x, y)			(((x) + (y) - 1) / (y) * (y))
-#define ROUND_DOWN_TO(x, y)		((x) / (y) * (y))
-#define ROUND_TO_PAGE(x)		ROUND_TO((x), GTT_PAGE_SIZE)
-#define ROUND_TO_MB(x)			ROUND_TO((x), MB(1))
-#define PRIMARY_RINGBUFFER_SIZE		KB(128)
-
-
-/* Ring buffer registers, p277, overview p19
- */
-#define LP_RING     0x2030
-#define HP_RING     0x2040
-
-#define RING_TAIL      0x00
-#define TAIL_ADDR           0x000FFFF8
-#define I830_TAIL_MASK	    0x001FFFF8
-
-#define RING_HEAD      0x04
-#define HEAD_WRAP_COUNT     0xFFE00000
-#define HEAD_WRAP_ONE       0x00200000
-#define HEAD_ADDR           0x001FFFFC
-#define I830_HEAD_MASK      0x001FFFFC
-
-#define RING_START     0x08
-#define START_ADDR          0x03FFFFF8
-#define I830_RING_START_MASK	0xFFFFF000
-
-#define RING_LEN       0x0C
-#define RING_NR_PAGES       0x001FF000 
-#define I830_RING_NR_PAGES	0x001FF000
-#define RING_REPORT_MASK    0x00000006
-#define RING_REPORT_64K     0x00000002
-#define RING_REPORT_128K    0x00000004
-#define RING_NO_REPORT      0x00000000
-#define RING_VALID_MASK     0x00000001
-#define RING_VALID          0x00000001
-#define RING_INVALID        0x00000000
-
-
-/* Fence/Tiling ranges [0..7]
- */
-#define FENCE            0x2000
-#define FENCE_NR         8
-
-#define I915G_FENCE_START_MASK	0x0ff00000
-
-#define I830_FENCE_START_MASK	0x07f80000
-
-#define FENCE_START_MASK    0x03F80000
-#define FENCE_X_MAJOR       0x00000000
-#define FENCE_Y_MAJOR       0x00001000
-#define FENCE_SIZE_MASK     0x00000700
-#define FENCE_SIZE_512K     0x00000000
-#define FENCE_SIZE_1M       0x00000100
-#define FENCE_SIZE_2M       0x00000200
-#define FENCE_SIZE_4M       0x00000300
-#define FENCE_SIZE_8M       0x00000400
-#define FENCE_SIZE_16M      0x00000500
-#define FENCE_SIZE_32M      0x00000600
-#define FENCE_SIZE_64M	    0x00000700
-#define I915G_FENCE_SIZE_1M       0x00000000
-#define I915G_FENCE_SIZE_2M       0x00000100
-#define I915G_FENCE_SIZE_4M       0x00000200
-#define I915G_FENCE_SIZE_8M       0x00000300
-#define I915G_FENCE_SIZE_16M      0x00000400
-#define I915G_FENCE_SIZE_32M      0x00000500
-#define I915G_FENCE_SIZE_64M	0x00000600
-#define I915G_FENCE_SIZE_128M	0x00000700
-#define FENCE_PITCH_1       0x00000000
-#define FENCE_PITCH_2       0x00000010
-#define FENCE_PITCH_4       0x00000020
-#define FENCE_PITCH_8       0x00000030
-#define FENCE_PITCH_16      0x00000040
-#define FENCE_PITCH_32      0x00000050
-#define FENCE_PITCH_64	    0x00000060
-#define FENCE_VALID         0x00000001
-
-#include <mmio.h>
-
-#  define MMIO_IN8(base, offset) \
-	*(volatile unsigned char *)(((unsigned char*)(base)) + (offset))
-#  define MMIO_IN32(base, offset) \
-	read_MMIO_LE32(base, offset)
-#  define MMIO_OUT8(base, offset, val) \
-	*(volatile unsigned char *)(((unsigned char*)(base)) + (offset)) = (val)
-#  define MMIO_OUT32(base, offset, val) \
-	*(volatile unsigned int *)(void *)(((unsigned char*)(base)) + (offset)) = CPU_TO_LE32(val)
-
-
-				/* Memory mapped register access macros */
-#define INREG8(addr)        MMIO_IN8(MMIO, addr)
-#define INREG(addr)         MMIO_IN32(MMIO, addr)
-#define OUTREG8(addr, val)  MMIO_OUT8(MMIO, addr, val)
-#define OUTREG(addr, val)   MMIO_OUT32(MMIO, addr, val)
-
-#define DSPABASE		0x70184
-
-#endif
diff --git a/src/mesa/drivers/dri/mach64/mach64_ioctl.h b/src/mesa/drivers/dri/mach64/mach64_ioctl.h
index 1ffda1932f1..9145ee6e6cf 100644
--- a/src/mesa/drivers/dri/mach64/mach64_ioctl.h
+++ b/src/mesa/drivers/dri/mach64/mach64_ioctl.h
@@ -32,6 +32,9 @@
 #ifndef __MACH64_IOCTL_H__
 #define __MACH64_IOCTL_H__
 
+#include <stdio.h>
+#include <stdlib.h>
+
 #include "mach64_dri.h"
 #include "mach64_reg.h"
 #include "mach64_lock.h"
diff --git a/src/mesa/drivers/dri/mga/mgarender.c b/src/mesa/drivers/dri/mga/mgarender.c
index 8b8fc485d31..cc0cea618d1 100644
--- a/src/mesa/drivers/dri/mga/mgarender.c
+++ b/src/mesa/drivers/dri/mga/mgarender.c
@@ -44,6 +44,8 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/imports.h"
 #include "main/mtypes.h"
 
+#include "math/m_xform.h"
+
 #include "tnl/t_context.h"
 
 #include "mgacontext.h"
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_fbo.c b/src/mesa/drivers/dri/nouveau/nouveau_fbo.c
index 8be7edb150b..bd1273beea7 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_fbo.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_fbo.c
@@ -220,7 +220,7 @@ get_tex_format(struct gl_texture_image *ti)
 	case MESA_FORMAT_RGB565:
 		return GL_RGB5;
 	default:
-		assert(0);
+		return GL_NONE;
 	}
 }
 
@@ -231,7 +231,6 @@ nouveau_render_texture(GLcontext *ctx, struct gl_framebuffer *fb,
 	struct gl_renderbuffer *rb = att->Renderbuffer;
 	struct gl_texture_image *ti =
 		att->Texture->Image[att->CubeMapFace][att->TextureLevel];
-	int ret;
 
 	/* Allocate a renderbuffer object for the texture if we
 	 * haven't already done so. */
@@ -244,9 +243,7 @@ nouveau_render_texture(GLcontext *ctx, struct gl_framebuffer *fb,
 	}
 
 	/* Update the renderbuffer fields from the texture. */
-	ret = set_renderbuffer_format(rb, get_tex_format(ti));
-	assert(ret);
-
+	set_renderbuffer_format(rb, get_tex_format(ti));
 	rb->Width = ti->Width;
 	rb->Height = ti->Height;
 	nouveau_surface_ref(&to_nouveau_teximage(ti)->surface,
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_texture.c b/src/mesa/drivers/dri/nouveau/nouveau_texture.c
index dbf9a5cc613..442f4e899ee 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_texture.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_texture.c
@@ -38,6 +38,7 @@
 #include "main/mipmap.h"
 #include "main/texfetch.h"
 #include "main/teximage.h"
+#include "drivers/common/meta.h"
 
 static struct gl_texture_object *
 nouveau_texture_new(GLcontext *ctx, GLuint name, GLenum target)
@@ -182,10 +183,10 @@ teximage_fits(struct gl_texture_object *t, int level)
 	struct nouveau_surface *s = &to_nouveau_texture(t)->surfaces[level];
 	struct gl_texture_image *ti = t->Image[0][level];
 
-	return ti && (t->Target == GL_TEXTURE_RECTANGLE ||
-		      (s->bo && s->width == ti->Width &&
-		       s->height == ti->Height &&
-		       s->format == ti->TexFormat));
+	return ti && to_nouveau_teximage(ti)->surface.bo &&
+		(t->Target == GL_TEXTURE_RECTANGLE ||
+		 (s->bo && s->format == ti->TexFormat &&
+		  s->width == ti->Width && s->height == ti->Height));
 }
 
 static GLboolean
@@ -589,6 +590,53 @@ nouveau_texture_unmap(GLcontext *ctx, struct gl_texture_object *t)
 	}
 }
 
+static void
+store_mipmap(GLcontext *ctx, GLenum target, int first, int last,
+	     struct gl_texture_object *t)
+{
+	struct gl_pixelstore_attrib packing = {
+		.BufferObj = ctx->Shared->NullBufferObj,
+		.Alignment = 1
+	};
+	GLenum format = t->Image[0][first]->TexFormat;
+	unsigned base_format, type, comps;
+	int i;
+
+	base_format = _mesa_get_format_base_format(format);
+	_mesa_format_to_type_and_comps(format, &type, &comps);
+
+	for (i = first; i <= last; i++) {
+		struct gl_texture_image *ti = t->Image[0][i];
+		void *data = ti->Data;
+
+		nouveau_teximage(ctx, 3, target, i, ti->InternalFormat,
+				 ti->Width, ti->Height, ti->Depth,
+				 ti->Border, base_format, type, data,
+				 &packing, t, ti);
+
+		_mesa_free_texmemory(data);
+	}
+}
+
+static void
+nouveau_generate_mipmap(GLcontext *ctx, GLenum target,
+			struct gl_texture_object *t)
+{
+	if (_mesa_meta_check_generate_mipmap_fallback(ctx, target, t)) {
+		struct gl_texture_image *base = t->Image[0][t->BaseLevel];
+
+		nouveau_teximage_map(ctx, base);
+		_mesa_generate_mipmap(ctx, target, t);
+		nouveau_teximage_unmap(ctx, base);
+
+		store_mipmap(ctx, target, t->BaseLevel + 1,
+			     get_last_level(t), t);
+
+	} else {
+		_mesa_meta_GenerateMipmap(ctx, target, t);
+	}
+}
+
 void
 nouveau_texture_functions_init(struct dd_function_table *functions)
 {
@@ -607,4 +655,5 @@ nouveau_texture_functions_init(struct dd_function_table *functions)
 	functions->BindTexture = nouveau_bind_texture;
 	functions->MapTexture = nouveau_texture_map;
 	functions->UnmapTexture = nouveau_texture_unmap;
+	functions->GenerateMipmap = nouveau_generate_mipmap;
 }
diff --git a/src/mesa/drivers/dri/nouveau/nv20_state_fb.c b/src/mesa/drivers/dri/nouveau/nv20_state_fb.c
index 21da4f7af16..95691cad047 100644
--- a/src/mesa/drivers/dri/nouveau/nv20_state_fb.c
+++ b/src/mesa/drivers/dri/nouveau/nv20_state_fb.c
@@ -72,7 +72,7 @@ nv20_emit_framebuffer(GLcontext *ctx, int emit)
 			fb->_ColorDrawBuffers[0])->surface;
 
 		rt_format |= get_rt_format(s->format);
-		zeta_pitch = rt_pitch = s->pitch;
+		rt_pitch = s->pitch;
 
 		nouveau_bo_markl(bctx, kelvin, NV20TCL_COLOR_OFFSET,
 				 s->bo, 0, bo_flags);
@@ -88,6 +88,9 @@ nv20_emit_framebuffer(GLcontext *ctx, int emit)
 
 		nouveau_bo_markl(bctx, kelvin, NV20TCL_ZETA_OFFSET,
 				 s->bo, 0, bo_flags);
+	} else {
+		rt_format |= get_rt_format(MESA_FORMAT_Z24_S8);
+		zeta_pitch = rt_pitch;
 	}
 
 	BEGIN_RING(chan, kelvin, NV20TCL_RT_FORMAT, 2);
diff --git a/src/mesa/drivers/dri/nouveau/nv20_state_tex.c b/src/mesa/drivers/dri/nouveau/nv20_state_tex.c
index e46118e4fce..2d45513bb4c 100644
--- a/src/mesa/drivers/dri/nouveau/nv20_state_tex.c
+++ b/src/mesa/drivers/dri/nouveau/nv20_state_tex.c
@@ -194,7 +194,8 @@ nv20_emit_tex_obj(GLcontext *ctx, int emit)
 		| nvgl_wrap_mode(t->WrapS) << 0;
 
 	tx_filter = nvgl_filter_mode(t->MagFilter) << 24
-		| nvgl_filter_mode(t->MinFilter) << 16;
+		| nvgl_filter_mode(t->MinFilter) << 16
+		| 2 << 12;
 
 	tx_enable = NV20TCL_TX_ENABLE_ENABLE
 		| log2i(t->MaxAnisotropy) << 4;
diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.c b/src/mesa/drivers/dri/r200/r200_swtcl.c
index 262fe3cddee..dbf4ad477db 100644
--- a/src/mesa/drivers/dri/r200/r200_swtcl.c
+++ b/src/mesa/drivers/dri/r200/r200_swtcl.c
@@ -612,6 +612,8 @@ static void r200RasterPrimitive( GLcontext *ctx, GLuint hwprim )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
 
+   radeon_prepare_render(&rmesa->radeon);
+
    if (rmesa->radeon.swtcl.hw_primitive != hwprim) {
       /* need to disable perspective-correct texturing for point sprites */
       if ((hwprim & 0xf) == R200_VF_PRIM_POINT_SPRITES && ctx->Point.PointSprite) {
diff --git a/src/mesa/drivers/dri/r200/r200_tcl.c b/src/mesa/drivers/dri/r200/r200_tcl.c
index d43e14581e9..4ae0f304918 100644
--- a/src/mesa/drivers/dri/r200/r200_tcl.c
+++ b/src/mesa/drivers/dri/r200/r200_tcl.c
@@ -264,6 +264,8 @@ void r200TclPrimitive( GLcontext *ctx,
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
    GLuint newprim = hw_prim | R200_VF_TCL_OUTPUT_VTX_ENABLE;
 
+   radeon_prepare_render(&rmesa->radeon);
+
    if (newprim != rmesa->tcl.hw_primitive ||
        !discrete_prim[hw_prim&0xf]) {
       /* need to disable perspective-correct texturing for point sprites */
diff --git a/src/mesa/drivers/dri/r300/compiler/r3xx_fragprog.c b/src/mesa/drivers/dri/r300/compiler/r3xx_fragprog.c
index a326ee4c4fa..d2fa816894c 100644
--- a/src/mesa/drivers/dri/r300/compiler/r3xx_fragprog.c
+++ b/src/mesa/drivers/dri/r300/compiler/r3xx_fragprog.c
@@ -109,13 +109,13 @@ void r3xx_compile_fragment_program(struct r300_fragment_program_compiler* c)
 	debug_program_log(c, "before compilation");
 
 	if (c->Base.is_r500){
-		r500_transform_unroll_loops(&c->Base, &loop_state);	
-		debug_program_log(c, "after r500 transform loops");
+		rc_unroll_loops(&c->Base, R500_PFS_MAX_INST);
+		debug_program_log(c, "after unroll loops");
 	}
 	else{
-		rc_transform_unroll_loops(&c->Base, &loop_state);
+		rc_transform_loops(&c->Base, &loop_state, -1);
 		debug_program_log(c, "after transform loops");
-		
+
 		rc_emulate_branches(&c->Base);
 		debug_program_log(c, "after emulate branches");
 	}
diff --git a/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog.c b/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog.c
index d347b4df9cd..666c9c2a7a9 100644
--- a/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog.c
+++ b/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog.c
@@ -32,6 +32,11 @@
 #include "radeon_emulate_branches.h"
 #include "radeon_emulate_loops.h"
 
+struct loop {
+	int BgnLoop;
+
+};
+
 /*
  * Take an already-setup and valid source then swizzle it appropriately to
  * obtain a constant ZERO or ONE source.
@@ -332,11 +337,140 @@ static void ei_pow(struct r300_vertex_program_code *vp,
 	inst[3] = t_src_scalar(vp, &vpi->SrcReg[1]);
 }
 
+static void mark_write(void * userdata,	struct rc_instruction * inst,
+		rc_register_file file,	unsigned int index, unsigned int mask)
+{
+	unsigned int * writemasks = userdata;
+
+	if (file != RC_FILE_TEMPORARY)
+		return;
+
+	if (index >= R300_VS_MAX_TEMPS)
+		return;
+
+	writemasks[index] |= mask;
+}
+
+static unsigned long t_pred_src(struct r300_vertex_program_compiler * compiler)
+{
+	return PVS_SRC_OPERAND(compiler->PredicateIndex,
+		t_swizzle(RC_SWIZZLE_ZERO),
+		t_swizzle(RC_SWIZZLE_ZERO),
+		t_swizzle(RC_SWIZZLE_ZERO),
+		t_swizzle(RC_SWIZZLE_W),
+		t_src_class(RC_FILE_TEMPORARY),
+		0);
+}
+
+static unsigned long t_pred_dst(struct r300_vertex_program_compiler * compiler,
+					unsigned int hw_opcode, int is_math)
+{
+	return PVS_OP_DST_OPERAND(hw_opcode,
+	     is_math,
+	     0,
+	     compiler->PredicateIndex,
+	     RC_MASK_W,
+	     t_dst_class(RC_FILE_TEMPORARY));
+
+}
+
+static void ei_if(struct r300_vertex_program_compiler * compiler,
+					struct rc_instruction *rci,
+					unsigned int * inst,
+					unsigned int branch_depth)
+{
+	unsigned int predicate_opcode;
+	int is_math = 0;
+
+	if (!compiler->Base.is_r500) {
+		rc_error(&compiler->Base,"Opcode IF not supported\n");
+		return;
+	}
+
+	/* Reserve a temporary to use as our predicate stack counter, if we
+	 * don't already have one. */
+	if (!compiler->PredicateMask) {
+		unsigned int writemasks[R300_VS_MAX_TEMPS];
+		memset(writemasks, 0, sizeof(writemasks));
+		struct rc_instruction * inst;
+		unsigned int i;
+		for(inst = compiler->Base.Program.Instructions.Next;
+				inst != &compiler->Base.Program.Instructions;
+							inst = inst->Next) {
+			rc_for_all_writes_mask(inst, mark_write, writemasks);
+		}
+		for(i = 0; i < R300_VS_MAX_TEMPS; i++) {
+			unsigned int mask = ~writemasks[i] & RC_MASK_XYZW;
+			/* Only the W component can be used fo the predicate
+			 * stack counter. */
+			if (mask & RC_MASK_W) {
+				compiler->PredicateMask = RC_MASK_W;
+				compiler->PredicateIndex = i;
+				break;
+			}
+		}
+		if (i == R300_VS_MAX_TEMPS) {
+			rc_error(&compiler->Base, "No free temporary to use for"
+					" predicate stack counter.\n");
+			return;
+		}
+	}
+	predicate_opcode =
+			branch_depth ? VE_PRED_SET_NEQ_PUSH : ME_PRED_SET_NEQ;
+
+	rci->U.I.SrcReg[0].Swizzle = RC_MAKE_SWIZZLE_SMEAR(GET_SWZ(rci->U.I.SrcReg[0].Swizzle,0));
+	if (branch_depth == 0) {
+		is_math = 1;
+		predicate_opcode = ME_PRED_SET_NEQ;
+		inst[1] = t_src(compiler->code, &rci->U.I.SrcReg[0]);
+		inst[2] = 0;
+	} else {
+		predicate_opcode = VE_PRED_SET_NEQ_PUSH;
+		inst[1] = t_pred_src(compiler);
+		inst[2] = t_src(compiler->code, &rci->U.I.SrcReg[0]);
+	}
+
+	inst[0] = t_pred_dst(compiler, predicate_opcode, is_math);
+	inst[3] = 0;
+
+}
+
+static void ei_else(struct r300_vertex_program_compiler * compiler,
+							unsigned int * inst)
+{
+	if (!compiler->Base.is_r500) {
+		rc_error(&compiler->Base,"Opcode ELSE not supported\n");
+		return;
+	}
+	inst[0] = t_pred_dst(compiler, ME_PRED_SET_INV, 1);
+	inst[1] = t_pred_src(compiler);
+	inst[2] = 0;
+	inst[3] = 0;
+}
+
+static void ei_endif(struct r300_vertex_program_compiler *compiler,
+							unsigned int * inst)
+{
+	if (!compiler->Base.is_r500) {
+		rc_error(&compiler->Base,"Opcode ENDIF not supported\n");
+		return;
+	}
+	inst[0] = t_pred_dst(compiler, ME_PRED_SET_POP, 1);
+	inst[1] = t_pred_src(compiler);
+	inst[2] = 0;
+	inst[3] = 0;
+}
 
 static void translate_vertex_program(struct r300_vertex_program_compiler * compiler)
 {
 	struct rc_instruction *rci;
 
+	struct loop * loops;
+	int current_loop_depth = 0;
+	int loops_reserved = 0;
+
+	unsigned int branch_depth = 0;
+
 	compiler->code->pos_end = 0;	/* Not supported yet */
 	compiler->code->length = 0;
 
@@ -366,9 +500,12 @@ static void translate_vertex_program(struct r300_vertex_program_compiler * compi
 		case RC_OPCODE_COS: ei_math1(compiler->code, ME_COS, vpi, inst); break;
 		case RC_OPCODE_DP4: ei_vector2(compiler->code, VE_DOT_PRODUCT, vpi, inst); break;
 		case RC_OPCODE_DST: ei_vector2(compiler->code, VE_DISTANCE_VECTOR, vpi, inst); break;
+		case RC_OPCODE_ELSE: ei_else(compiler, inst); break;
+		case RC_OPCODE_ENDIF: ei_endif(compiler, inst); branch_depth--; break;
 		case RC_OPCODE_EX2: ei_math1(compiler->code, ME_EXP_BASE2_FULL_DX, vpi, inst); break;
 		case RC_OPCODE_EXP: ei_math1(compiler->code, ME_EXP_BASE2_DX, vpi, inst); break;
 		case RC_OPCODE_FRC: ei_vector1(compiler->code, VE_FRACTION, vpi, inst); break;
+		case RC_OPCODE_IF: ei_if(compiler, rci, inst, branch_depth); branch_depth++; break;
 		case RC_OPCODE_LG2: ei_math1(compiler->code, ME_LOG_BASE2_FULL_DX, vpi, inst); break;
 		case RC_OPCODE_LIT: ei_lit(compiler->code, vpi, inst); break;
 		case RC_OPCODE_LOG: ei_math1(compiler->code, ME_LOG_BASE2_DX, vpi, inst); break;
@@ -385,11 +522,86 @@ static void translate_vertex_program(struct r300_vertex_program_compiler * compi
 		case RC_OPCODE_SIN: ei_math1(compiler->code, ME_SIN, vpi, inst); break;
 		case RC_OPCODE_SLT: ei_vector2(compiler->code, VE_SET_LESS_THAN, vpi, inst); break;
 		case RC_OPCODE_SNE: ei_vector2(compiler->code, VE_SET_NOT_EQUAL, vpi, inst); break;
+		case RC_OPCODE_BGNLOOP:
+		{
+			struct loop * l;
+
+			if ((!compiler->Base.is_r500
+				&& loops_reserved >= R300_VS_MAX_LOOP_DEPTH)
+				|| loops_reserved >= R500_VS_MAX_FC_DEPTH) {
+				rc_error(&compiler->Base,
+						"Loops are nested too deep.");
+				return;
+			}
+			memory_pool_array_reserve(&compiler->Base.Pool,
+					struct loop, loops, current_loop_depth,
+					loops_reserved, 1);
+			l = &loops[current_loop_depth++];
+			memset(l , 0, sizeof(struct loop));
+			l->BgnLoop = (compiler->code->length / 4);
+			continue;
+		}
+		case RC_OPCODE_ENDLOOP:
+		{
+			struct loop * l = &loops[current_loop_depth - 1];
+			unsigned int act_addr = l->BgnLoop - 1;
+			unsigned int last_addr = (compiler->code->length / 4) - 1;
+			unsigned int ret_addr = l->BgnLoop;
+
+			if (loops_reserved >= R300_VS_MAX_FC_OPS) {
+				rc_error(&compiler->Base,
+					"Too many flow control instructions.");
+				return;
+			}
+			if (compiler->Base.is_r500) {
+				compiler->code->fc_op_addrs.r500
+					[compiler->code->num_fc_ops].lw =
+					R500_PVS_FC_ACT_ADRS(act_addr)
+					| R500_PVS_FC_LOOP_CNT_JMP_INST(0xffff)
+					;
+				compiler->code->fc_op_addrs.r500
+					[compiler->code->num_fc_ops].uw =
+					R500_PVS_FC_LAST_INST(last_addr)
+					| R500_PVS_FC_RTN_INST(ret_addr)
+					;
+			} else {
+				compiler->code->fc_op_addrs.r300
+					[compiler->code->num_fc_ops] =
+					R300_PVS_FC_ACT_ADRS(act_addr)
+					| R300_PVS_FC_LOOP_CNT_JMP_INST(0xff)
+					| R300_PVS_FC_LAST_INST(last_addr)
+					| R300_PVS_FC_RTN_INST(ret_addr)
+					;
+			}
+			compiler->code->fc_loop_index[compiler->code->num_fc_ops] =
+				R300_PVS_FC_LOOP_INIT_VAL(0x0)
+				| R300_PVS_FC_LOOP_STEP_VAL(0x1)
+				;
+			compiler->code->fc_ops |= R300_VAP_PVS_FC_OPC_LOOP(
+						compiler->code->num_fc_ops);
+			compiler->code->num_fc_ops++;
+			current_loop_depth--;
+			continue;
+		}
+
 		default:
 			rc_error(&compiler->Base, "Unknown opcode %s\n", rc_get_opcode_info(vpi->Opcode)->Name);
 			return;
 		}
 
+		/* Non-flow control instructions that are inside an if statement
+		 * need to pay attention to the predicate bit. */
+		if (branch_depth
+			&& vpi->Opcode != RC_OPCODE_IF
+			&& vpi->Opcode != RC_OPCODE_ELSE
+			&& vpi->Opcode != RC_OPCODE_ENDIF) {
+
+			inst[0] |= (PVS_DST_PRED_ENABLE_MASK
+						<< PVS_DST_PRED_ENABLE_SHIFT);
+			inst[0] |= (PVS_DST_PRED_SENSE_MASK
+						<< PVS_DST_PRED_SENSE_SHIFT);
+		}
+
 		compiler->code->length += 4;
 
 		if (compiler->Base.Error)
@@ -406,6 +618,7 @@ struct temporary_allocation {
 static void allocate_temporary_registers(struct r300_vertex_program_compiler * compiler)
 {
 	struct rc_instruction *inst;
+	struct rc_instruction *end_loop = NULL;
 	unsigned int num_orig_temps = 0;
 	char hwtemps[R300_VS_MAX_TEMPS];
 	struct temporary_allocation * ta;
@@ -440,10 +653,35 @@ static void allocate_temporary_registers(struct r300_vertex_program_compiler * c
 	/* Pass 2: Determine original temporary lifetimes */
 	for(inst = compiler->Base.Program.Instructions.Next; inst != &compiler->Base.Program.Instructions; inst = inst->Next) {
 		const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->U.I.Opcode);
+		/* Instructions inside of loops need to use the ENDLOOP
+		 * instruction as their LastRead. */
+		if (!end_loop && inst->U.I.Opcode == RC_OPCODE_BGNLOOP) {
+			int endloops = 1;
+			struct rc_instruction * ptr;
+			for(ptr = inst->Next;
+				ptr != &compiler->Base.Program.Instructions;
+							ptr = ptr->Next){
+				if (ptr->U.I.Opcode == RC_OPCODE_BGNLOOP) {
+					endloops++;
+				} else if (ptr->U.I.Opcode == RC_OPCODE_ENDLOOP) {
+					endloops--;
+					if (endloops <= 0) {
+						end_loop = ptr;
+						break;
+					}
+				}
+			}
+		}
+
+		if (inst == end_loop) {
+			end_loop = NULL;
+			continue;
+		}
 
 		for (i = 0; i < opcode->NumSrcRegs; ++i) {
 			if (inst->U.I.SrcReg[i].File == RC_FILE_TEMPORARY)
-				ta[inst->U.I.SrcReg[i].Index].LastRead = inst;
+				ta[inst->U.I.SrcReg[i].Index].LastRead =
+						end_loop ? end_loop : inst;
 		}
 	}
 
@@ -633,30 +871,24 @@ static struct rc_swizzle_caps r300_vertprog_swizzle_caps = {
 void r3xx_compile_vertex_program(struct r300_vertex_program_compiler* compiler)
 {
 	struct emulate_loop_state loop_state;
-	
+
 	compiler->Base.SwizzleCaps = &r300_vertprog_swizzle_caps;
 
 	addArtificialOutputs(compiler);
 
 	debug_program_log(compiler, "before compilation");
 
-	/* XXX Ideally this should be done only for r3xx, but since
-	 * we don't have branching support for r5xx, we use the emulation
-	 * on all chipsets. */
-	rc_transform_unroll_loops(&compiler->Base, &loop_state);
-	
-	debug_program_log(compiler, "after transform loops");
-	
-	if (compiler->Base.is_r500){
-		rc_emulate_loops(&loop_state, R500_VS_MAX_ALU);
-	} else {
-		rc_emulate_loops(&loop_state, R300_VS_MAX_ALU);
-	}
-	debug_program_log(compiler, "after emulate loops");
+	if (compiler->Base.is_r500)
+		rc_transform_loops(&compiler->Base, &loop_state, R500_VS_MAX_ALU);
+	else
+		rc_transform_loops(&compiler->Base, &loop_state, R300_VS_MAX_ALU);
 
-	rc_emulate_branches(&compiler->Base);
+	debug_program_log(compiler, "after emulate loops");
 
-	debug_program_log(compiler, "after emulate branches");
+	if (!compiler->Base.is_r500) {
+		rc_emulate_branches(&compiler->Base);
+		debug_program_log(compiler, "after emulate branches");
+	}
 
 	if (compiler->Base.is_r500) {
 		struct radeon_program_transformation transformations[] = {
@@ -718,6 +950,6 @@ void r3xx_compile_vertex_program(struct r300_vertex_program_compiler* compiler)
 
 	if (compiler->Base.Debug) {
 		fprintf(stderr, "Final vertex program code:\n");
-		r300_vertex_program_dump(compiler->code);
+		r300_vertex_program_dump(compiler);
 	}
 }
diff --git a/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog_dump.c b/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog_dump.c
index 5800f1a78e1..e6009338e2e 100644
--- a/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog_dump.c
+++ b/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog_dump.c
@@ -20,7 +20,9 @@
  * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
  * USE OR OTHER DEALINGS IN THE SOFTWARE. */
 
+#include "radeon_compiler.h"
 #include "radeon_code.h"
+#include "../r300_reg.h"
 
 #include <stdio.h>
 
@@ -133,6 +135,10 @@ static void r300_vs_op_dump(uint32_t op)
 {
 	fprintf(stderr, " dst: %d%s op: ",
 			(op >> 13) & 0x7f, r300_vs_dst_debug[(op >> 8) & 0x7]);
+	if ((op >> PVS_DST_PRED_ENABLE_SHIFT) & 0x1) {
+		fprintf(stderr, "PRED %u",
+				(op >> PVS_DST_PRED_SENSE_SHIFT) & 0x1);
+	}
 	if (op & 0x80) {
 		if (op & 0x1) {
 			fprintf(stderr, "PVS_MACRO_OP_2CLK_M2X_ADD\n");
@@ -160,8 +166,9 @@ static void r300_vs_src_dump(uint32_t src)
 			r300_vs_swiz_debug[(src >> 22) & 0x7]);
 }
 
-void r300_vertex_program_dump(struct r300_vertex_program_code * vs)
+void r300_vertex_program_dump(struct r300_vertex_program_compiler * c)
 {
+	struct r300_vertex_program_code * vs = c->code;
 	unsigned instrcount = vs->length / 4;
 	unsigned i;
 
@@ -177,4 +184,21 @@ void r300_vertex_program_dump(struct r300_vertex_program_code * vs)
 			r300_vs_src_dump(vs->body.d[offset+1+src]);
 		}
 	}
+
+	fprintf(stderr, "Flow Control Ops: 0x%08x\n",vs->fc_ops);
+	for(i = 0; i < vs->num_fc_ops; i++) {
+		switch((vs->fc_ops >> (i * 2)) & 0x3 ) {
+		case 0: fprintf(stderr, "NOP"); break;
+		case 1: fprintf(stderr, "JUMP"); break;
+		case 2: fprintf(stderr, "LOOP"); break;
+		case 3: fprintf(stderr, "JSR"); break;
+		}
+		if (c->Base.is_r500) {
+			fprintf(stderr,": uw-> 0x%08x lw-> 0x%08x\n",
+				vs->fc_op_addrs.r500[i].uw,
+				vs->fc_op_addrs.r500[i].lw);
+		} else {
+			fprintf(stderr,": 0x%08x\n", vs->fc_op_addrs.r300[i]);
+		}
+	}
 }
diff --git a/src/mesa/drivers/dri/r300/compiler/r500_fragprog.c b/src/mesa/drivers/dri/r300/compiler/r500_fragprog.c
index e6b5522c5b9..80a120497e3 100644
--- a/src/mesa/drivers/dri/r300/compiler/r500_fragprog.c
+++ b/src/mesa/drivers/dri/r300/compiler/r500_fragprog.c
@@ -30,7 +30,6 @@
 #include <stdio.h>
 
 #include "../r300_reg.h"
-#include "radeon_emulate_loops.h"
 
 /**
  * Rewrite IF instructions to use the ALU result special register.
@@ -60,31 +59,6 @@ int r500_transform_IF(
 	return 1;
 }
 
-/**
- * Rewrite loops to make them easier to emit.  This is not a local
- * transformation, because it modifies and reorders an entire block of code.
- */
-void r500_transform_unroll_loops(struct radeon_compiler * c,
-						struct emulate_loop_state *s)
-{
-	int i;
-	
-	rc_transform_unroll_loops(c, s);
-	
-	for( i = s->LoopCount - 1; i >= 0; i-- ){
-		struct rc_instruction * inst_continue;
-		if(!s->Loops[i].EndLoop){
-			continue;
-		}
-		/* Insert a continue instruction at the end of the loop.  This
-		 * is required in order to emit loops correctly. */
-		inst_continue = rc_insert_new_instruction(c,
-						s->Loops[i].EndIf->Prev);
-		inst_continue->U.I.Opcode = RC_OPCODE_CONTINUE;
-	}
-
-}
-
 static int r500_swizzle_is_native(rc_opcode opcode, struct rc_src_register reg)
 {
 	unsigned int relevant;
diff --git a/src/mesa/drivers/dri/r300/compiler/r500_fragprog.h b/src/mesa/drivers/dri/r300/compiler/r500_fragprog.h
index 0d005a794ff..34173351f83 100644
--- a/src/mesa/drivers/dri/r300/compiler/r500_fragprog.h
+++ b/src/mesa/drivers/dri/r300/compiler/r500_fragprog.h
@@ -49,6 +49,4 @@ extern int r500_transform_IF(
 	struct rc_instruction * inst,
 	void* data);
 
-void r500_transform_unroll_loops(struct radeon_compiler * c,
-						struct emulate_loop_state * s);
 #endif
diff --git a/src/mesa/drivers/dri/r300/compiler/r500_fragprog_emit.c b/src/mesa/drivers/dri/r300/compiler/r500_fragprog_emit.c
index 0bd8f0a239f..9b60e30f586 100644
--- a/src/mesa/drivers/dri/r300/compiler/r500_fragprog_emit.c
+++ b/src/mesa/drivers/dri/r300/compiler/r500_fragprog_emit.c
@@ -64,7 +64,16 @@ struct branch_info {
 };
 
 struct loop_info {
-	int LoopStart;
+	int BgnLoop;
+
+	int BranchDepth;
+	int * Brks;
+	int BrkCount;
+	int BrkReserved;
+
+	int * Conts;
+	int ContCount;
+	int ContReserved;
 };
 
 struct emit_state {
@@ -368,6 +377,12 @@ static void emit_flowcontrol(struct emit_state * s, struct rc_instruction * inst
 
 	unsigned int newip = ++s->Code->inst_end;
 
+	/* Currently all loops use the same integer constant to intialize
+	 * the loop variables. */
+	if(!s->Code->int_constants[0]) {
+		s->Code->int_constants[0] = R500_FC_INT_CONST_KR(0xff);
+		s->Code->int_constant_count = 1;
+	}
 	s->Code->inst[newip].inst0 = R500_INST_TYPE_FC | R500_INST_ALU_WAIT;
 
 	switch(inst->U.I.Opcode){
@@ -378,32 +393,77 @@ static void emit_flowcontrol(struct emit_state * s, struct rc_instruction * inst
 			s->Loops, s->CurrentLoopDepth, s->LoopsReserved, 1);
 
 		loop = &s->Loops[s->CurrentLoopDepth++];
-		
-		/* We don't emit an instruction for BGNLOOP, so we need to
-		 * decrement the instruction counter, but first we need to
-		 * set LoopStart to the current value of inst_end, which
-		 * will end up being the first real instruction in the loop.*/
-		loop->LoopStart = s->Code->inst_end--;
+		memset(loop, 0, sizeof(struct loop_info));
+		loop->BranchDepth = s->CurrentBranchDepth;
+		loop->BgnLoop = newip;
+
+		s->Code->inst[newip].inst2 = R500_FC_OP_LOOP
+			| R500_FC_JUMP_FUNC(0x00)
+			| R500_FC_IGNORE_UNCOVERED
+			;
 		break;
-	
 	case RC_OPCODE_BRK:
-		/* Don't emit an instruction for BRK */
-		s->Code->inst_end--;
+		loop = &s->Loops[s->CurrentLoopDepth - 1];
+		memory_pool_array_reserve(&s->C->Pool, int, loop->Brks,
+					loop->BrkCount, loop->BrkReserved, 1);
+
+		loop->Brks[loop->BrkCount++] = newip;
+		s->Code->inst[newip].inst2 = R500_FC_OP_BREAKLOOP
+			| R500_FC_JUMP_FUNC(0xff)
+			| R500_FC_B_OP1_DECR
+			| R500_FC_B_POP_CNT(
+				s->CurrentBranchDepth - loop->BranchDepth)
+			| R500_FC_IGNORE_UNCOVERED
+			;
 		break;
 
-	case RC_OPCODE_CONTINUE:
+	case RC_OPCODE_CONT:
 		loop = &s->Loops[s->CurrentLoopDepth - 1];
-		s->Code->inst[newip].inst2 = R500_FC_OP_JUMP |
-			R500_FC_JUMP_FUNC(0xff);
-		s->Code->inst[newip].inst3 = R500_FC_JUMP_ADDR(loop->LoopStart);
+		memory_pool_array_reserve(&s->C->Pool, int, loop->Conts,
+					loop->ContCount, loop->ContReserved, 1);
+		loop->Conts[loop->ContCount++] = newip;
+		s->Code->inst[newip].inst2 = R500_FC_OP_CONTINUE
+			| R500_FC_JUMP_FUNC(0xff)
+			| R500_FC_B_OP1_DECR
+			| R500_FC_B_POP_CNT(
+				s->CurrentBranchDepth -	loop->BranchDepth)
+			| R500_FC_IGNORE_UNCOVERED
+			;
 		break;
 
 	case RC_OPCODE_ENDLOOP:
-		/* Don't emit an instruction for ENDLOOP */
-		s->Code->inst_end--;
+	{
+		loop = &s->Loops[s->CurrentLoopDepth - 1];
+		/* Emit ENDLOOP */
+		s->Code->inst[newip].inst2 = R500_FC_OP_ENDLOOP
+			| R500_FC_JUMP_FUNC(0xff)
+			| R500_FC_JUMP_ANY
+			| R500_FC_IGNORE_UNCOVERED
+			;
+		/* The constant integer at index 0 is used by all loops. */
+		s->Code->inst[newip].inst3 = R500_FC_INT_ADDR(0)
+			| R500_FC_JUMP_ADDR(loop->BgnLoop + 1)
+			;
+
+		/* Set jump address and int constant for BGNLOOP */
+		s->Code->inst[loop->BgnLoop].inst3 = R500_FC_INT_ADDR(0)
+			| R500_FC_JUMP_ADDR(newip)
+			;
+
+		/* Set jump address for the BRK instructions. */
+		while(loop->BrkCount--) {
+			s->Code->inst[loop->Brks[loop->BrkCount]].inst3 =
+						R500_FC_JUMP_ADDR(newip + 1);
+		}
+
+		/* Set jump address for CONT instructions. */
+		while(loop->ContCount--) {
+			s->Code->inst[loop->Conts[loop->ContCount]].inst3 =
+						R500_FC_JUMP_ADDR(newip);
+		}
 		s->CurrentLoopDepth--;
 		break;
-
+	}
 	case RC_OPCODE_IF:
 		if ( s->CurrentBranchDepth >= MAX_BRANCH_DEPTH_FULL) {
 			rc_error(s->C, "Branch depth exceeds hardware limit");
@@ -442,24 +502,16 @@ static void emit_flowcontrol(struct emit_state * s, struct rc_instruction * inst
 		}
 
 		branch = &s->Branches[s->CurrentBranchDepth - 1];
-		
-		if(inst->Prev->U.I.Opcode == RC_OPCODE_BRK){
-			branch->Endif = --s->Code->inst_end;
-			s->Code->inst[branch->Endif].inst2 |=
-				R500_FC_B_OP0_DECR;
-		}
-		else{
-			branch->Endif = newip;
-		
-			s->Code->inst[branch->Endif].inst2 = R500_FC_OP_JUMP
-				| R500_FC_A_OP_NONE /* no address stack */
-				| R500_FC_JUMP_ANY /* docs says set this, but I don't understand why */
-				| R500_FC_B_OP0_DECR /* decrement branch counter if stay */
-				| R500_FC_B_OP1_NONE /* no branch counter if stay */
-				| R500_FC_B_POP_CNT(1)
+		branch->Endif = newip;
+
+		s->Code->inst[branch->Endif].inst2 = R500_FC_OP_JUMP
+			| R500_FC_A_OP_NONE /* no address stack */
+			| R500_FC_JUMP_ANY /* docs says set this, but I don't understand why */
+			| R500_FC_B_OP0_DECR /* decrement branch counter if stay */
+			| R500_FC_B_OP1_NONE /* no branch counter if stay */
+			| R500_FC_B_POP_CNT(1)
 			;
-			s->Code->inst[branch->Endif].inst3 = R500_FC_JUMP_ADDR(branch->Endif + 1);
-		}
+		s->Code->inst[branch->Endif].inst3 = R500_FC_JUMP_ADDR(branch->Endif + 1);
 		s->Code->inst[branch->If].inst2 = R500_FC_OP_JUMP
 			| R500_FC_A_OP_NONE /* no address stack */
 			| R500_FC_JUMP_FUNC(0x0f) /* jump if ALU result is false */
@@ -544,11 +596,9 @@ void r500BuildFragmentProgramHwCode(struct r300_fragment_program_compiler *compi
 		code->inst[ip].inst0 = R500_INST_TYPE_OUT | R500_INST_TEX_SEM_WAIT;
 	}
 
-	/* Use FULL flow control mode if branches are nested deep enough.
-	 * We don not need to enable FULL flow control mode for loops, becasue
-	 * we aren't using the hardware loop instructions.
-	 */
-	if (s.MaxBranchDepth >= 4) {
+	/* Enable full flow control mode if we are using loops or have if
+	 * statements nested at least four deep. */
+	if (s.MaxBranchDepth >= 4 || s.LoopsReserved > 0) {
 		if (code->max_temp_idx < 1)
 			code->max_temp_idx = 1;
 
diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_code.h b/src/mesa/drivers/dri/r300/compiler/radeon_code.h
index d03689763bc..896246d2035 100644
--- a/src/mesa/drivers/dri/r300/compiler/radeon_code.h
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_code.h
@@ -221,6 +221,9 @@ struct r500_fragment_program_code {
 	int max_temp_idx;
 
 	uint32_t us_fc_ctrl;
+
+	uint32_t int_constants[32];
+	uint32_t int_constant_count;
 };
 
 struct rX00_fragment_program_code {
@@ -240,6 +243,12 @@ struct rX00_fragment_program_code {
 #define R500_VS_MAX_ALU	        1024
 #define R500_VS_MAX_ALU_DWORDS  (R500_VS_MAX_ALU * 4)
 #define R300_VS_MAX_TEMPS	32
+/* This is the max for all chipsets (r300-r500) */
+#define R300_VS_MAX_FC_OPS 16
+/* The r500 maximum depth is not just for loops, but any combination of loops
+ * and subroutine jumps. */
+#define R500_VS_MAX_FC_DEPTH 8
+#define R300_VS_MAX_LOOP_DEPTH 1
 
 #define VSF_MAX_INPUTS 32
 #define VSF_MAX_OUTPUTS 32
@@ -260,9 +269,18 @@ struct r300_vertex_program_code {
 
 	uint32_t InputsRead;
 	uint32_t OutputsWritten;
-};
 
-void r300_vertex_program_dump(struct r300_vertex_program_code * vs);
+	unsigned int num_fc_ops;
+	uint32_t fc_ops;
+	union {
+	        uint32_t r300[R300_VS_MAX_FC_OPS];
+		struct {
+			uint32_t lw;
+			uint32_t uw;
+		} r500[R300_VS_MAX_FC_OPS];
+	} fc_op_addrs;
+	int32_t fc_loop_index[R300_VS_MAX_FC_OPS];
+};
 
 #endif /* RADEON_CODE_H */
 
diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_compiler.c b/src/mesa/drivers/dri/r300/compiler/radeon_compiler.c
index 1c8ba864a41..935dc9b0a80 100644
--- a/src/mesa/drivers/dri/r300/compiler/radeon_compiler.c
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_compiler.c
@@ -307,3 +307,46 @@ void rc_transform_fragment_wpos(struct radeon_compiler * c, unsigned wpos, unsig
 	}
 }
 
+
+/**
+ * The FACE input in hardware contains 1 if it's a back face, 0 otherwise.
+ * Gallium and OpenGL define it the other way around.
+ *
+ * So let's just negate FACE at the beginning of the shader and rewrite the rest
+ * of the shader to read from the newly allocated temporary.
+ */
+void rc_transform_fragment_face(struct radeon_compiler *c, unsigned face)
+{
+	unsigned tempregi = rc_find_free_temporary(c);
+	struct rc_instruction *inst_add;
+	struct rc_instruction *inst;
+
+	/* perspective divide */
+	inst_add = rc_insert_new_instruction(c, &c->Program.Instructions);
+	inst_add->U.I.Opcode = RC_OPCODE_ADD;
+
+	inst_add->U.I.DstReg.File = RC_FILE_TEMPORARY;
+	inst_add->U.I.DstReg.Index = tempregi;
+	inst_add->U.I.DstReg.WriteMask = RC_MASK_X;
+
+	inst_add->U.I.SrcReg[0].File = RC_FILE_NONE;
+	inst_add->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_1111;
+
+	inst_add->U.I.SrcReg[1].File = RC_FILE_INPUT;
+	inst_add->U.I.SrcReg[1].Index = face;
+	inst_add->U.I.SrcReg[1].Swizzle = RC_SWIZZLE_XXXX;
+	inst_add->U.I.SrcReg[1].Negate = RC_MASK_XYZW;
+
+	for (inst = inst_add->Next; inst != &c->Program.Instructions; inst = inst->Next) {
+		const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->U.I.Opcode);
+		unsigned i;
+
+		for(i = 0; i < opcode->NumSrcRegs; i++) {
+			if (inst->U.I.SrcReg[i].File == RC_FILE_INPUT &&
+			    inst->U.I.SrcReg[i].Index == face) {
+				inst->U.I.SrcReg[i].File = RC_FILE_TEMPORARY;
+				inst->U.I.SrcReg[i].Index = tempregi;
+			}
+		}
+	}
+}
diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_compiler.h b/src/mesa/drivers/dri/r300/compiler/radeon_compiler.h
index f15905d79d4..7c42eb3ae57 100644
--- a/src/mesa/drivers/dri/r300/compiler/radeon_compiler.h
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_compiler.h
@@ -81,6 +81,7 @@ void rc_move_output(struct radeon_compiler * c, unsigned output, unsigned new_ou
 void rc_copy_output(struct radeon_compiler * c, unsigned output, unsigned dup_output);
 void rc_transform_fragment_wpos(struct radeon_compiler * c, unsigned wpos, unsigned new_input,
                                 int full_vtransform);
+void rc_transform_fragment_face(struct radeon_compiler *c, unsigned face);
 
 struct r300_fragment_program_compiler {
 	struct radeon_compiler Base;
@@ -110,8 +111,12 @@ struct r300_vertex_program_compiler {
 
 	void * UserData;
 	void (*SetHwInputOutput)(struct r300_vertex_program_compiler * c);
+
+	int PredicateIndex;
+	unsigned int PredicateMask;
 };
 
 void r3xx_compile_vertex_program(struct r300_vertex_program_compiler* c);
+void r300_vertex_program_dump(struct r300_vertex_program_compiler * c);
 
 #endif /* RADEON_COMPILER_H */
diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_dataflow_deadcode.c b/src/mesa/drivers/dri/r300/compiler/radeon_dataflow_deadcode.c
index fbb4235c223..faf531b412e 100644
--- a/src/mesa/drivers/dri/r300/compiler/radeon_dataflow_deadcode.c
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_dataflow_deadcode.c
@@ -43,6 +43,12 @@ struct instruction_state {
 	unsigned char SrcReg[3];
 };
 
+struct loopinfo {
+	struct updatemask_state * Breaks;
+	unsigned int BreakCount;
+	unsigned int BreaksReserved;
+};
+
 struct branchinfo {
 	unsigned int HaveElse:1;
 
@@ -59,6 +65,10 @@ struct deadcode_state {
 	struct branchinfo * BranchStack;
 	unsigned int BranchStackSize;
 	unsigned int BranchStackReserved;
+
+	struct loopinfo * LoopStack;
+	unsigned int LoopStackSize;
+	unsigned int LoopStackReserved;
 };
 
 
@@ -78,6 +88,22 @@ static void or_updatemasks(
 	dst->Address = a->Address | b->Address;
 }
 
+static void push_break(struct deadcode_state *s)
+{
+	struct loopinfo * loop = &s->LoopStack[s->LoopStackSize - 1];
+	memory_pool_array_reserve(&s->C->Pool, struct updatemask_state,
+		loop->Breaks, loop->BreakCount, loop->BreaksReserved, 1);
+
+	memcpy(&loop->Breaks[loop->BreakCount++], &s->R, sizeof(s->R));
+}
+
+static void push_loop(struct deadcode_state * s)
+{
+	memory_pool_array_reserve(&s->C->Pool, struct loopinfo, s->LoopStack,
+			s->LoopStackSize, s->LoopStackReserved, 1);
+	memset(&s->LoopStack[s->LoopStackSize++], 0, sizeof(struct loopinfo));
+}
+
 static void push_branch(struct deadcode_state * s)
 {
 	memory_pool_array_reserve(&s->C->Pool, struct branchinfo, s->BranchStack,
@@ -233,11 +259,22 @@ void rc_dataflow_deadcode(struct radeon_compiler * c, rc_dataflow_mark_outputs_f
 					}
 				}
 			}
+			push_loop(&s);
 			break;
 		}
-		case RC_OPCODE_CONTINUE:
 		case RC_OPCODE_BRK:
+			push_break(&s);
+			break;
 		case RC_OPCODE_BGNLOOP:
+		{
+			unsigned int i;
+			struct loopinfo * loop = &s.LoopStack[s.LoopStackSize-1];
+			for(i = 0; i < loop->BreakCount; i++) {
+				or_updatemasks(&s.R, &s.R, &loop->Breaks[i]);
+			}
+			break;
+		}
+		case RC_OPCODE_CONT:
 			break;
 		case RC_OPCODE_ENDIF:
 			push_branch(&s);
diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_emulate_loops.c b/src/mesa/drivers/dri/r300/compiler/radeon_emulate_loops.c
index 131e9e7436d..32d4b45dd6d 100644
--- a/src/mesa/drivers/dri/r300/compiler/radeon_emulate_loops.c
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_emulate_loops.c
@@ -39,7 +39,6 @@
 #define DBG(...) do { if (VERBOSE) fprintf(stderr, __VA_ARGS__); } while(0)
 
 struct const_value {
-	
 	struct radeon_compiler * C;
 	struct rc_src_register * Src;
 	float Value;
@@ -78,17 +77,17 @@ static int src_reg_is_immediate(struct rc_src_register * src,
 	c->Program.Constants.Constants[src->Index].Type==RC_CONSTANT_IMMEDIATE;
 }
 
-static unsigned int loop_calc_iterations(struct emulate_loop_state *s, 
-			struct loop_info * loop, unsigned int max_instructions)
+static unsigned int loop_max_possible_iterations(struct radeon_compiler *c,
+			struct loop_info * loop, unsigned int prog_inst_limit)
 {
-	unsigned int total_i = rc_recompute_ips(s->C);
+	unsigned int total_i = rc_recompute_ips(c);
 	unsigned int loop_i = (loop->EndLoop->IP - loop->BeginLoop->IP) - 1;
 	/* +1 because the program already has one iteration of the loop. */
-	return 1 + ((max_instructions - total_i) / (s->LoopCount * loop_i));
+	return 1 + ((prog_inst_limit - total_i) / loop_i);
 }
 
-static void loop_unroll(struct emulate_loop_state * s,
-			struct loop_info *loop, unsigned int iterations)
+static void unroll_loop(struct radeon_compiler * c, struct loop_info * loop,
+						unsigned int iterations)
 {
 	unsigned int i;
 	struct rc_instruction * ptr;
@@ -99,7 +98,7 @@ static void loop_unroll(struct emulate_loop_state * s,
 	rc_remove_instruction(loop->EndLoop);
 	for( i = 1; i < iterations; i++){
 		for(ptr = first; ptr != last->Next; ptr = ptr->Next){
-			struct rc_instruction *new = rc_alloc_instruction(s->C);
+			struct rc_instruction *new = rc_alloc_instruction(c);
 			memcpy(new, ptr, sizeof(struct rc_instruction));
 			rc_insert_instruction(append_to, new);
 			append_to = new;
@@ -115,7 +114,7 @@ static void update_const_value(void * data, struct rc_instruction * inst,
 	if(value->Src->File != file ||
 	   value->Src->Index != index ||
 	   !(1 << GET_SWZ(value->Src->Swizzle, 0) & mask)){
-	   	return;
+		return;
 	}
 	switch(inst->U.I.Opcode){
 	case RC_OPCODE_MOV:
@@ -140,7 +139,7 @@ static void get_incr_amount(void * data, struct rc_instruction * inst,
 	if(file != RC_FILE_TEMPORARY ||
 	   count_inst->Index != index ||
 	   (1 << GET_SWZ(count_inst->Swz,0) != mask)){
-	   	return;
+		return;
 	}
 	/* Find the index of the counter register. */
 	opcode = rc_get_opcode_info(inst->U.I.Opcode);
@@ -185,13 +184,16 @@ static void get_incr_amount(void * data, struct rc_instruction * inst,
 		count_inst->Unknown = 1;
 		return;
 	}
-	
 }
 
-static int transform_const_loop(struct emulate_loop_state * s,
-						struct loop_info * loop)
+/**
+ * If prog_inst_limit is -1, then all eligible loops will be unrolled regardless
+ * of how many iterations they have.
+ */
+static int try_unroll_loop(struct radeon_compiler * c, struct loop_info * loop,
+						unsigned int prog_inst_limit)
 {
-	int end_loops = 1;
+	int end_loops;
 	int iterations;
 	struct count_inst count_inst;
 	float limit_value;
@@ -201,12 +203,12 @@ static int transform_const_loop(struct emulate_loop_state * s,
 	struct rc_instruction * inst;
 
 	/* Find the counter and the upper limit */
-	
-	if(src_reg_is_immediate(&loop->Cond->U.I.SrcReg[0], s->C)){
+
+	if(src_reg_is_immediate(&loop->Cond->U.I.SrcReg[0], c)){
 		limit = &loop->Cond->U.I.SrcReg[0];
 		counter = &loop->Cond->U.I.SrcReg[1];
 	}
-	else if(src_reg_is_immediate(&loop->Cond->U.I.SrcReg[1], s->C)){
+	else if(src_reg_is_immediate(&loop->Cond->U.I.SrcReg[1], c)){
 		limit = &loop->Cond->U.I.SrcReg[1];
 		counter = &loop->Cond->U.I.SrcReg[0];
 	}
@@ -214,13 +216,13 @@ static int transform_const_loop(struct emulate_loop_state * s,
 		DBG("No constant limit.\n");
 		return 0;
 	}
-	
+
 	/* Find the initial value of the counter */
 	counter_value.Src = counter;
 	counter_value.Value = 0.0f;
 	counter_value.HasValue = 0;
-	counter_value.C = s->C;
-	for(inst = s->C->Program.Instructions.Next; inst != loop->BeginLoop;
+	counter_value.C = c;
+	for(inst = c->Program.Instructions.Next; inst != loop->BeginLoop;
 							inst = inst->Next){
 		rc_for_all_writes_mask(inst, update_const_value, &counter_value);
 	}
@@ -230,11 +232,12 @@ static int transform_const_loop(struct emulate_loop_state * s,
 	}
 	DBG("Initial counter value is %f\n", counter_value.Value);
 	/* Determine how the counter is modified each loop */
-	count_inst.C = s->C;
+	count_inst.C = c;
 	count_inst.Index = counter->Index;
 	count_inst.Swz = counter->Swizzle;
 	count_inst.Amount = 0.0f;
 	count_inst.Unknown = 0;
+	end_loops = 1;
 	for(inst = loop->BeginLoop->Next; end_loops > 0; inst = inst->Next){
 		switch(inst->U.I.Opcode){
 		/* XXX In the future we might want to try to unroll nested
@@ -246,6 +249,16 @@ static int transform_const_loop(struct emulate_loop_state * s,
 			loop->EndLoop = inst;
 			end_loops--;
 			break;
+		case RC_OPCODE_BRK:
+			/* Don't unroll loops if it has a BRK instruction
+			 * other one used when testing the main conditional
+			 * of the loop. */
+
+			/* Make sure we haven't entered a nested loops. */
+			if(inst != loop->Brk && end_loops == 1) {
+				return 0;
+			}
+			break;
 		/* XXX Check if the counter is modified within an if statement.
 		 */
 		case RC_OPCODE_IF:
@@ -266,17 +279,20 @@ static int transform_const_loop(struct emulate_loop_state * s,
 	/* Calculate the number of iterations of this loop.  Keeping this
 	 * simple, since we only support increment and decrement loops.
 	 */
-	limit_value = get_constant_value(s->C, limit, 0);
+	limit_value = get_constant_value(c, limit, 0);
 	DBG("Limit is %f.\n", limit_value);
+	/* The iteration calculations are opposite of what you would expect.
+	 * In a normal loop, if the condition is met, then loop continues, but
+	 * with our loops, if the condition is met, the is exited. */
 	switch(loop->Cond->U.I.Opcode){
-	case RC_OPCODE_SGT:
-	case RC_OPCODE_SLT:
+	case RC_OPCODE_SGE:
+	case RC_OPCODE_SLE:
 		iterations = (int) ceilf((limit_value - counter_value.Value) /
 							count_inst.Amount);
 		break;
 
-	case RC_OPCODE_SLE:
-	case RC_OPCODE_SGE:
+	case RC_OPCODE_SGT:
+	case RC_OPCODE_SLT:
 		iterations = (int) floorf((limit_value - counter_value.Value) /
 							count_inst.Amount) + 1;
 		break;
@@ -284,77 +300,85 @@ static int transform_const_loop(struct emulate_loop_state * s,
 		return 0;
 	}
 
+	if (prog_inst_limit > 0
+		&& iterations > loop_max_possible_iterations(c, loop,
+							prog_inst_limit)) {
+		return 0;
+	}
+
 	DBG("Loop will have %d iterations.\n", iterations);
-	
+
 	/* Prepare loop for unrolling */
 	rc_remove_instruction(loop->Cond);
 	rc_remove_instruction(loop->If);
 	rc_remove_instruction(loop->Brk);
 	rc_remove_instruction(loop->EndIf);
-	
-	loop_unroll(s, loop, iterations);
+
+	unroll_loop(c, loop, iterations);
 	loop->EndLoop = NULL;
 	return 1;
 }
 
-/** 
- * This function prepares a loop to be unrolled by converting it into an if
- * statement.  Here is an outline of the conversion process:
- * BGNLOOP;                         	-> BGNLOOP;
- * <Additional conditional code>	-> <Additional conditional code>
- * SGE/SLT temp[0], temp[1], temp[2];	-> SLT/SGE temp[0], temp[1], temp[2];
- * IF temp[0];                      	-> IF temp[0];
- * BRK;                             	->
- * ENDIF;                           	-> <Loop Body>
- * <Loop Body>                      	-> ENDIF;
- * ENDLOOP;                         	-> ENDLOOP
- *
+/**
+ * @param c
+ * @param loop
  * @param inst A pointer to a BGNLOOP instruction.
- * @return If the loop can be unrolled, a pointer to the first instruction of
- * 		the unrolled loop.
- * 	   Otherwise, A pointer to the ENDLOOP instruction.
- * 	   Null if there is an error.
+ * @return 1 if all of the members of loop where set.
+ * @return 0 if there was an error and some members of loop are still NULL.
  */
-static struct rc_instruction * transform_loop(struct emulate_loop_state * s,
+static int build_loop_info(struct radeon_compiler * c, struct loop_info * loop,
 						struct rc_instruction * inst)
 {
-	struct loop_info *loop;
 	struct rc_instruction * ptr;
 
-	memory_pool_array_reserve(&s->C->Pool, struct loop_info,
-			s->Loops, s->LoopCount, s->LoopReserved, 1);
-
-	loop = &s->Loops[s->LoopCount++];
-	memset(loop, 0, sizeof(struct loop_info));
 	if(inst->U.I.Opcode != RC_OPCODE_BGNLOOP){
-		rc_error(s->C, "expected BGNLOOP\n", __FUNCTION__);
-		return NULL;
+		rc_error(c, "%s: expected BGNLOOP", __FUNCTION__);
+		return 0;
 	}
+
+	memset(loop, 0, sizeof(struct loop_info));
+
 	loop->BeginLoop = inst;
 
-	for(ptr = loop->BeginLoop->Next; !loop->EndLoop; ptr = ptr->Next){
+	for(ptr = loop->BeginLoop->Next; !loop->EndLoop; ptr = ptr->Next) {
+
+		if (ptr == &c->Program.Instructions) {
+			rc_error(c, "%s: BGNLOOP without an ENDLOOOP.\n",
+								__FUNCTION__);
+			return 0;
+		}
+
 		switch(ptr->U.I.Opcode){
 		case RC_OPCODE_BGNLOOP:
-			/* Nested loop */
-			ptr = transform_loop(s, ptr);
-			if(!ptr){
-				return NULL;
+		{
+			/* Nested loop, skip ahead to the end. */
+			unsigned int loop_depth = 1;
+			for(ptr = ptr->Next; ptr != &c->Program.Instructions;
+							ptr = ptr->Next){
+				if (ptr->U.I.Opcode == RC_OPCODE_BGNLOOP) {
+					loop_depth++;
+				} else if (ptr->U.I.Opcode == RC_OPCODE_ENDLOOP) {
+					if (!--loop_depth) {
+						break;
+					}
+				}
+			}
+			if (ptr == &c->Program.Instructions) {
+				rc_error(c, "%s: BGNLOOP without an ENDLOOOP\n",
+								__FUNCTION__);
+					return 0;
 			}
 			break;
+		}
 		case RC_OPCODE_BRK:
-			loop->Brk = ptr;
-			if(ptr->Next->U.I.Opcode != RC_OPCODE_ENDIF){
-				rc_error(s->C,
-					"%s: expected ENDIF\n",__FUNCTION__);
-				return NULL;
-			}
-			loop->EndIf = ptr->Next;
-			if(ptr->Prev->U.I.Opcode != RC_OPCODE_IF){
-				rc_error(s->C,
-					"%s: expected IF\n", __FUNCTION__);
-				return NULL;
+			if(ptr->Next->U.I.Opcode != RC_OPCODE_ENDIF
+					|| ptr->Prev->U.I.Opcode != RC_OPCODE_IF
+					|| loop->Brk){
+				continue;
 			}
+			loop->Brk = ptr;
 			loop->If = ptr->Prev;
+			loop->EndIf = ptr->Next;
 			switch(loop->If->Prev->U.I.Opcode){
 			case RC_OPCODE_SLT:
 			case RC_OPCODE_SGE:
@@ -364,18 +388,58 @@ static struct rc_instruction * transform_loop(struct emulate_loop_state * s,
 			case RC_OPCODE_SNE:
 				break;
 			default:
-				rc_error(s->C, "%s expected conditional\n",
+				rc_error(c, "%s: expected conditional",
 								__FUNCTION__);
-				return NULL;
+				return 0;
 			}
 			loop->Cond = loop->If->Prev;
-			ptr = loop->EndIf;
 			break;
+
 		case RC_OPCODE_ENDLOOP:
 			loop->EndLoop = ptr;
 			break;
 		}
 	}
+
+	if (loop->BeginLoop && loop->Brk && loop->If && loop->EndIf
+					&& loop->Cond && loop->EndLoop) {
+		return 1;
+	}
+	return 0;
+}
+
+/**
+ * This function prepares a loop to be unrolled by converting it into an if
+ * statement.  Here is an outline of the conversion process:
+ * BGNLOOP;                         	-> BGNLOOP;
+ * <Additional conditional code>	-> <Additional conditional code>
+ * SGE/SLT temp[0], temp[1], temp[2];	-> SLT/SGE temp[0], temp[1], temp[2];
+ * IF temp[0];                      	-> IF temp[0];
+ * BRK;                             	->
+ * ENDIF;                           	-> <Loop Body>
+ * <Loop Body>                      	-> ENDIF;
+ * ENDLOOP;                         	-> ENDLOOP
+ *
+ * @param inst A pointer to a BGNLOOP instruction.
+ * @return 1 for success, 0 for failure
+ */
+static int transform_loop(struct emulate_loop_state * s,
+						struct rc_instruction * inst)
+{
+	struct loop_info * loop;
+
+	memory_pool_array_reserve(&s->C->Pool, struct loop_info,
+			s->Loops, s->LoopCount, s->LoopReserved, 1);
+
+	loop = &s->Loops[s->LoopCount++];
+
+	if (!build_loop_info(s->C, loop, inst))
+		return 0;
+
+	if(try_unroll_loop(s->C, loop, s->prog_inst_limit)){
+		return 1;
+	}
+
 	/* Reverse the conditional instruction */
 	switch(loop->Cond->U.I.Opcode){
 	case RC_OPCODE_SGE:
@@ -398,43 +462,51 @@ static struct rc_instruction * transform_loop(struct emulate_loop_state * s,
 		break;
 	default:
 		rc_error(s->C, "loop->Cond is not a conditional.\n");
-		return NULL;
-	}
-	
-	/* Check if the number of loops is known at compile time. */
-	if(transform_const_loop(s, loop)){
-		return loop->BeginLoop->Next;
+		return 0;
 	}
 
-	/* Prepare the loop to be unrolled */
+	/* Prepare the loop to be emulated */
 	rc_remove_instruction(loop->Brk);
 	rc_remove_instruction(loop->EndIf);
 	rc_insert_instruction(loop->EndLoop->Prev, loop->EndIf);
-	return loop->EndLoop;
+	return 1;
 }
 
-void rc_transform_unroll_loops(struct radeon_compiler *c,
-					struct emulate_loop_state * s)
+void rc_transform_loops(struct radeon_compiler *c,
+			struct emulate_loop_state * s, int prog_inst_limit)
 {
 	struct rc_instruction * ptr;
-	
+
 	memset(s, 0, sizeof(struct emulate_loop_state));
 	s->C = c;
-	ptr = s->C->Program.Instructions.Next;
-	while(ptr != &s->C->Program.Instructions) {
+	s->prog_inst_limit = prog_inst_limit;
+	for(ptr = s->C->Program.Instructions.Next;
+			ptr != &s->C->Program.Instructions; ptr = ptr->Next) {
 		if(ptr->Type == RC_INSTRUCTION_NORMAL &&
 					ptr->U.I.Opcode == RC_OPCODE_BGNLOOP){
-			ptr = transform_loop(s, ptr);
-			if(!ptr){
+			if (!transform_loop(s, ptr))
 				return;
+		}
+	}
+}
+
+void rc_unroll_loops(struct radeon_compiler *c, int prog_inst_limit)
+{
+	struct rc_instruction * inst;
+	struct loop_info loop;
+
+	for(inst = c->Program.Instructions.Next;
+			inst != &c->Program.Instructions; inst = inst->Next) {
+
+		if (inst->U.I.Opcode == RC_OPCODE_BGNLOOP) {
+			if (build_loop_info(c, &loop, inst)) {
+				try_unroll_loop(c, &loop, prog_inst_limit);
 			}
 		}
-		ptr = ptr->Next;
 	}
 }
 
-void rc_emulate_loops(struct emulate_loop_state *s,
-						unsigned int max_instructions)
+void rc_emulate_loops(struct emulate_loop_state *s, int prog_inst_limit)
 {
 	int i;
 	/* Iterate backwards of the list of loops so that loops that nested
@@ -444,8 +516,8 @@ void rc_emulate_loops(struct emulate_loop_state *s,
 		if(!s->Loops[i].EndLoop){
 			continue;
 		}
-		unsigned int iterations = loop_calc_iterations(s, &s->Loops[i],
-							max_instructions);
-		loop_unroll(s, &s->Loops[i], iterations);
+		unsigned int iterations = loop_max_possible_iterations(
+					s->C, &s->Loops[i], prog_inst_limit);
+		unroll_loop(s->C, &s->Loops[i], iterations);
 	}
 }
diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_emulate_loops.h b/src/mesa/drivers/dri/r300/compiler/radeon_emulate_loops.h
index 7748813c4eb..bba1f68e308 100644
--- a/src/mesa/drivers/dri/r300/compiler/radeon_emulate_loops.h
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_emulate_loops.h
@@ -21,12 +21,14 @@ struct emulate_loop_state {
 	struct loop_info * Loops;
 	unsigned int LoopCount;
 	unsigned int LoopReserved;
+	int prog_inst_limit;
 };
 
-void rc_transform_unroll_loops(struct radeon_compiler *c,
-					struct emulate_loop_state * s);
+void rc_transform_loops(struct radeon_compiler *c,
+			struct emulate_loop_state * s, int prog_inst_limit);
 
-void rc_emulate_loops(struct emulate_loop_state *s,
-					unsigned int max_instructions);
+void rc_unroll_loops(struct radeon_compiler * c, int prog_inst_limit);
+
+void rc_emulate_loops(struct emulate_loop_state * s, int prog_inst_limit);
 
 #endif /* RADEON_EMULATE_LOOPS_H */
diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_opcodes.c b/src/mesa/drivers/dri/r300/compiler/radeon_opcodes.c
index 04f234f11d8..2ea830be7f9 100644
--- a/src/mesa/drivers/dri/r300/compiler/radeon_opcodes.c
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_opcodes.c
@@ -386,8 +386,8 @@ struct rc_opcode_info rc_opcodes[MAX_RC_OPCODE] = {
 		.NumSrcRegs = 0,
 	},
 	{
-		.Opcode = RC_OPCODE_CONTINUE,
-		.Name = "CONTINUE",
+		.Opcode = RC_OPCODE_CONT,
+		.Name = "CONT",
 		.IsFlowControl = 1,
 		.NumSrcRegs = 0
 	},
diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_opcodes.h b/src/mesa/drivers/dri/r300/compiler/radeon_opcodes.h
index 8b9fa07dde2..6e18d6eb3f1 100644
--- a/src/mesa/drivers/dri/r300/compiler/radeon_opcodes.h
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_opcodes.h
@@ -187,7 +187,7 @@ typedef enum {
 
 	RC_OPCODE_ENDLOOP,
 
-	RC_OPCODE_CONTINUE,
+	RC_OPCODE_CONT,
 
 	/** special instruction, used in R300-R500 fragment program pair instructions
 	 * indicates that the result of the alpha operation shall be replicated
diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_optimize.c b/src/mesa/drivers/dri/r300/compiler/radeon_optimize.c
index eca06515367..7a3f35950a6 100644
--- a/src/mesa/drivers/dri/r300/compiler/radeon_optimize.c
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_optimize.c
@@ -164,7 +164,8 @@ static void peephole(struct radeon_compiler * c, struct rc_instruction * inst_mo
 	    inst = inst->Next) {
 		/* XXX In the future we might be able to make the optimizer
 		 * smart enough to handle loops. */
-		if(inst->U.I.Opcode == RC_OPCODE_BGNLOOP){
+		if(inst->U.I.Opcode == RC_OPCODE_BGNLOOP
+				|| inst->U.I.Opcode == RC_OPCODE_ENDLOOP){
 			return;
 		}
 		rc_for_all_reads_mask(inst, peephole_scan_read, &s);
diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_pair_regalloc.c b/src/mesa/drivers/dri/r300/compiler/radeon_pair_regalloc.c
index 8a912da4613..ce72cd97ab2 100644
--- a/src/mesa/drivers/dri/r300/compiler/radeon_pair_regalloc.c
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_pair_regalloc.c
@@ -65,6 +65,11 @@ struct regalloc_state {
 
 	struct hardware_register * HwTemporary;
 	unsigned int NumHwTemporaries;
+	/**
+	 * If an instruction is inside of a loop, end_loop will be the
+	 * IP of the ENDLOOP instruction, otherwise end_loop will be 0
+	 */
+	int end_loop;
 };
 
 static void print_live_intervals(struct live_intervals * src)
@@ -178,10 +183,10 @@ static void scan_callback(void * data, struct rc_instruction * inst,
 		else
 			reg->Live.Start = inst->IP;
 		reg->Live.End = inst->IP;
-	} else {
-		if (inst->IP > reg->Live.End)
-			reg->Live.End = inst->IP;
-	}
+	} else if (s->end_loop)
+		reg->Live.End = s->end_loop;
+	else if (inst->IP > reg->Live.End)
+		reg->Live.End = inst->IP;
 }
 
 static void compute_live_intervals(struct regalloc_state * s)
@@ -191,6 +196,31 @@ static void compute_live_intervals(struct regalloc_state * s)
 	for(struct rc_instruction * inst = s->C->Program.Instructions.Next;
 	    inst != &s->C->Program.Instructions;
 	    inst = inst->Next) {
+
+		/* For all instructions inside of a loop, the ENDLOOP
+		 * instruction is used as the end of the live interval. */
+		if (inst->U.I.Opcode == RC_OPCODE_BGNLOOP && !s->end_loop) {
+			int loops = 1;
+			struct rc_instruction * tmp;
+			for(tmp = inst->Next;
+					tmp != &s->C->Program.Instructions;
+					tmp = tmp->Next) {
+				if (tmp->U.I.Opcode == RC_OPCODE_BGNLOOP) {
+					loops++;
+					break;
+				} else if (tmp->U.I.Opcode
+							== RC_OPCODE_ENDLOOP) {
+					if(!--loops) {
+						s->end_loop = tmp->IP;
+						break;
+					}
+				}
+			}
+		}
+
+		if (inst->IP == s->end_loop)
+			s->end_loop = 0;
+
 		rc_for_all_reads_mask(inst, scan_callback, s);
 		rc_for_all_writes_mask(inst, scan_callback, s);
 	}
diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c b/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c
index 3cc28972934..857aae55145 100644
--- a/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c
@@ -988,17 +988,22 @@ void radeonTransformKILP(struct radeon_compiler * c)
 	for (inst = c->Program.Instructions.Next;
 			inst != &c->Program.Instructions; inst = inst->Next) {
 
-		if (inst->U.I.Opcode != RC_OPCODE_KILP
-			|| inst->Prev->U.I.Opcode != RC_OPCODE_IF
-			|| inst->Next->U.I.Opcode != RC_OPCODE_ENDIF) {
+		if (inst->U.I.Opcode != RC_OPCODE_KILP)
 			continue;
-		}
+
 		inst->U.I.Opcode = RC_OPCODE_KIL;
-		inst->U.I.SrcReg[0] = negate(absolute(inst->Prev->U.I.SrcReg[0]));
 
-		/* Remove IF */
-		rc_remove_instruction(inst->Prev);
-		/* Remove ENDIF */
-		rc_remove_instruction(inst->Next);
+		if (inst->Prev->U.I.Opcode != RC_OPCODE_IF
+				|| inst->Next->U.I.Opcode != RC_OPCODE_ENDIF) {
+			inst->U.I.SrcReg[0] = negate(builtin_one);
+		} else {
+
+			inst->U.I.SrcReg[0] =
+				negate(absolute(inst->Prev->U.I.SrcReg[0]));
+			/* Remove IF */
+			rc_remove_instruction(inst->Prev);
+			/* Remove ENDIF */
+			rc_remove_instruction(inst->Next);
+		}
 	}
 }
diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c
index e4b302bbad9..3d2f8928fa6 100644
--- a/src/mesa/drivers/dri/r300/r300_context.c
+++ b/src/mesa/drivers/dri/r300/r300_context.c
@@ -461,7 +461,7 @@ static void r300InitGLExtensions(GLcontext *ctx)
 	if (!r300->radeon.radeonScreen->drmSupportsOcclusionQueries) {
 		_mesa_disable_extension(ctx, "GL_ARB_occlusion_query");
 	}
-	if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV350)
+        if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_R420)
   		_mesa_enable_extension(ctx, "GL_ARB_half_float_vertex");
 
 	if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515)
diff --git a/src/mesa/drivers/dri/r300/r300_reg.h b/src/mesa/drivers/dri/r300/r300_reg.h
index f25264b6f2d..f7705b0f6fe 100644
--- a/src/mesa/drivers/dri/r300/r300_reg.h
+++ b/src/mesa/drivers/dri/r300/r300_reg.h
@@ -441,6 +441,12 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define R300_VAP_GB_HORZ_CLIP_ADJ                   0x2228
 #define R300_VAP_GB_HORZ_DISC_ADJ                   0x222c
 
+#define R300_VAP_PVS_FLOW_CNTL_ADDRS_0      0x2230
+#define R300_PVS_FC_ACT_ADRS(x)             ((x) << 0)
+#define R300_PVS_FC_LOOP_CNT_JMP_INST(x)    ((x) << 8)
+#define R300_PVS_FC_LAST_INST(x)            ((x) << 16)
+#define R300_PVS_FC_RTN_INST(x)             ((x) << 24)
+
 /* gap */
 
 /* Sometimes, END_OF_PKT and 0x2284=0 are the only commands sent between
@@ -459,6 +465,10 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #       define R300_2288_R300                    0x00750000 /* -- nh */
 #       define R300_2288_RV350                   0x0000FFFF /* -- Vladimir */
 
+#define R300_VAP_PVS_FLOW_CNTL_LOOP_INDEX_0 0x2290
+#define R300_PVS_FC_LOOP_INIT_VAL(x)        ((x) << 0)
+#define R300_PVS_FC_LOOP_STEP_VAL(x)        ((x) << 8)
+
 /* gap */
 
 /* Addresses are relative to the vertex program instruction area of the
@@ -489,6 +499,9 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define R300_VAP_PVS_CODE_CNTL_1	    0x22D8
 #       define R300_PVS_LAST_VTX_SRC_INST_SHIFT  0
 #define R300_VAP_PVS_FLOW_CNTL_OPC          0x22DC
+#define R300_VAP_PVS_FC_OPC_JUMP(x)         (1 << (2 * (x)))
+#define R300_VAP_PVS_FC_OPC_LOOP(x)         (2 << (2 * (x)))
+#define R300_VAP_PVS_FC_OPC_JSR(x)          (3 << (2 * (x)))
 
 /* The entire range from 0x2300 to 0x2AC inclusive seems to be used for
  * immediate vertices
@@ -505,6 +518,14 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 /* write 0 to indicate end of packet? */
 #define R300_VAP_VTX_END_OF_PKT             0x24AC
 
+#define R500_VAP_PVS_FLOW_CNTL_ADDRS_LW_0   0x2500
+#define R500_PVS_FC_ACT_ADRS(x)             ((x) << 0)
+#define R500_PVS_FC_LOOP_CNT_JMP_INST(x)    ((x) << 16)
+
+#define R500_VAP_PVS_FLOW_CNTL_ADDRS_UW_0   0x2504
+#define R500_PVS_FC_LAST_INST(x)            ((x) << 0)
+#define R500_PVS_FC_RTN_INST(x)             ((x) << 16)
+
 /* gap */
 
 /* These are values from r300_reg/r300_reg.h - they are known to be correct
diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c
index bb8f91491f5..cf89ab7ec3d 100644
--- a/src/mesa/drivers/dri/r300/r300_render.c
+++ b/src/mesa/drivers/dri/r300/r300_render.c
@@ -327,6 +327,8 @@ void r300RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim)
 	BATCH_LOCALS(&rmesa->radeon);
 	int type, num_verts;
 
+	radeon_prepare_render(&rmesa->radeon);
+
 	type = r300PrimitiveType(rmesa, prim);
 	num_verts = r300NumVerts(rmesa, end - start, prim);
 
diff --git a/src/mesa/drivers/dri/r300/r300_texstate.c b/src/mesa/drivers/dri/r300/r300_texstate.c
index 4ba6740e3d9..94588698265 100644
--- a/src/mesa/drivers/dri/r300/r300_texstate.c
+++ b/src/mesa/drivers/dri/r300/r300_texstate.c
@@ -152,8 +152,8 @@ int32_t r300TranslateTexFormat(gl_format mesaFormat)
 		case MESA_FORMAT_Z32:
 			return R300_EASY_TX_FORMAT(X, X, X, X, X32);
 		/* EXT_texture_sRGB */
-		case MESA_FORMAT_SRGBA8:
-			return R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8) | R300_TX_FORMAT_GAMMA;
+		case MESA_FORMAT_SARGB8:
+			return R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8) | R300_TX_FORMAT_GAMMA;
 		case MESA_FORMAT_SLA8:
 			return R300_EASY_TX_FORMAT(X, X, X, Y, Y8X8) | R300_TX_FORMAT_GAMMA;
 		case MESA_FORMAT_SL8:
diff --git a/src/mesa/drivers/dri/r600/r600_blit.c b/src/mesa/drivers/dri/r600/r600_blit.c
index 172f85eb264..27acff9c166 100644
--- a/src/mesa/drivers/dri/r600/r600_blit.c
+++ b/src/mesa/drivers/dri/r600/r600_blit.c
@@ -72,7 +72,7 @@ unsigned r600_check_blit(gl_format mesa_format)
     case MESA_FORMAT_Z24_S8:
     case MESA_FORMAT_Z16:
     case MESA_FORMAT_Z32:
-    case MESA_FORMAT_SRGBA8:
+    case MESA_FORMAT_SARGB8:
     case MESA_FORMAT_SLA8:
     case MESA_FORMAT_SL8:
 	    break;
@@ -320,9 +320,9 @@ set_render_target(context_t *context, struct radeon_bo *bo, gl_format mesa_forma
 	    CLEARbit(cb_color0_info, SOURCE_FORMAT_bit);
 	    SETfield(cb_color0_info, NUMBER_UNORM, NUMBER_TYPE_shift, NUMBER_TYPE_mask);
             break;
-    case MESA_FORMAT_SRGBA8:
+    case MESA_FORMAT_SARGB8:
             format = COLOR_8_8_8_8;
-            comp_swap = SWAP_STD_REV;
+            comp_swap = SWAP_ALT;
 	    SETbit(cb_color0_info, SOURCE_FORMAT_bit);
 	    SETfield(cb_color0_info, NUMBER_SRGB, NUMBER_TYPE_shift, NUMBER_TYPE_mask);
             break;
@@ -390,13 +390,20 @@ set_render_target(context_t *context, struct radeon_bo *bo, gl_format mesa_forma
 			 0, RADEON_GEM_DOMAIN_VRAM | RADEON_GEM_DOMAIN_GTT, 0);
     END_BATCH();
 
-    BEGIN_BATCH_NO_AUTOSTATE(12);
+    BEGIN_BATCH_NO_AUTOSTATE(9);
     R600_OUT_BATCH_REGVAL(CB_COLOR0_SIZE + (4 * id), cb_color0_size);
     R600_OUT_BATCH_REGVAL(CB_COLOR0_VIEW + (4 * id), cb_color0_view);
-    R600_OUT_BATCH_REGVAL(CB_COLOR0_INFO + (4 * id), cb_color0_info);
     R600_OUT_BATCH_REGVAL(CB_COLOR0_MASK + (4 * id), 0);
     END_BATCH();
 
+    BEGIN_BATCH_NO_AUTOSTATE(3 + 2);
+    R600_OUT_BATCH_REGVAL(CB_COLOR0_INFO + (4 * id), cb_color0_info);
+    R600_OUT_BATCH_RELOC(0,
+			 bo,
+			 0,
+			 0, RADEON_GEM_DOMAIN_VRAM | RADEON_GEM_DOMAIN_GTT, 0);
+    END_BATCH();
+
     COMMIT_BATCH();
 
 }
@@ -1043,17 +1050,17 @@ set_tex_resource(context_t * context,
 	    SETfield(sq_tex_resource4, SQ_SEL_X,
 		     SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
 	    break;
-    case MESA_FORMAT_SRGBA8:
+    case MESA_FORMAT_SARGB8:
 	    SETfield(sq_tex_resource1, FMT_8_8_8_8,
 		     SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
 
-	    SETfield(sq_tex_resource4, SQ_SEL_W,
-		     SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
 	    SETfield(sq_tex_resource4, SQ_SEL_Z,
-		     SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+		     SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
 	    SETfield(sq_tex_resource4, SQ_SEL_Y,
-		     SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		     SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
 	    SETfield(sq_tex_resource4, SQ_SEL_X,
+		     SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+	    SETfield(sq_tex_resource4, SQ_SEL_W,
 		     SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
 	    SETbit(sq_tex_resource4, SQ_TEX_RESOURCE_WORD4_0__FORCE_DEGAMMA_bit);
 	    break;
@@ -1477,7 +1484,6 @@ set_default_state(context_t *context)
                          (CLRCMP_SEL_SRC << CLRCMP_FCN_SEL_shift));
     R600_OUT_BATCH_REGVAL(SQ_VTX_BASE_VTX_LOC, 0);
     R600_OUT_BATCH_REGVAL(SQ_VTX_START_INST_LOC, 0);
-    R600_OUT_BATCH_REGVAL(DB_DEPTH_INFO, 0);
     R600_OUT_BATCH_REGVAL(DB_DEPTH_CONTROL, 0);
     R600_OUT_BATCH_REGVAL(CB_SHADER_MASK, (OUTPUT0_ENABLE_mask));
     R600_OUT_BATCH_REGVAL(CB_TARGET_MASK, (TARGET0_ENABLE_mask));
@@ -1526,6 +1532,7 @@ set_default_state(context_t *context)
     R600_OUT_BATCH(0);
 
     R600_OUT_BATCH_REGVAL(VGT_STRMOUT_BUFFER_EN, 0);
+    R600_OUT_BATCH_REGVAL(SX_ALPHA_TEST_CONTROL, 0);
 
     END_BATCH();
     COMMIT_BATCH();
@@ -1607,7 +1614,7 @@ unsigned r600_blit(GLcontext *ctx,
     /* Flush is needed to make sure that source buffer has correct data */
     radeonFlush(ctx);
 
-    rcommonEnsureCmdBufSpace(&context->radeon, 304, __FUNCTION__);
+    rcommonEnsureCmdBufSpace(&context->radeon, 308, __FUNCTION__);
 
     /* load shaders */
     load_shaders(context->radeon.glCtx);
@@ -1632,7 +1639,7 @@ unsigned r600_blit(GLcontext *ctx,
     set_tex_sampler(context);
 
     /* dst */
-    /* 27 */
+    /* 31 */
     set_render_target(context, dst_bo, dst_mesaformat,
 		      dst_pitch, dst_width, dst_height, dst_offset);
     /* scissors */
diff --git a/src/mesa/drivers/dri/r600/r600_context.c b/src/mesa/drivers/dri/r600/r600_context.c
index 84d9d423124..389b0412baa 100644
--- a/src/mesa/drivers/dri/r600/r600_context.c
+++ b/src/mesa/drivers/dri/r600/r600_context.c
@@ -72,6 +72,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define R600_ENABLE_GLSL_TEST 1
 
 #define need_GL_VERSION_2_0
+#define need_GL_VERSION_2_1
+#define need_GL_ARB_draw_elements_base_vertex
 #define need_GL_ARB_occlusion_query
 #define need_GL_ARB_point_parameters
 #define need_GL_ARB_vertex_program
@@ -140,6 +142,7 @@ static const struct dri_extension card_extensions[] = {
   {"GL_NV_vertex_program",		GL_NV_vertex_program_functions},
   {"GL_SGIS_generate_mipmap",		NULL},
   {"GL_ARB_pixel_buffer_object",        NULL},
+  {"GL_ARB_draw_elements_base_vertex",	GL_ARB_draw_elements_base_vertex_functions },
   {NULL,				NULL}
   /* *INDENT-ON* */
 };
@@ -157,6 +160,7 @@ static const struct dri_extension mm_extensions[] = {
 static const struct dri_extension gl_20_extension[] = {
 #ifdef R600_ENABLE_GLSL_TEST
     {"GL_ARB_shading_language_100",			GL_VERSION_2_0_functions },
+    {"GL_ARB_shading_language_120",			GL_VERSION_2_1_functions },
 #else
   {"GL_VERSION_2_0",			GL_VERSION_2_0_functions },
 #endif /* R600_ENABLE_GLSL_TEST */
diff --git a/src/mesa/drivers/dri/r600/r600_tex.c b/src/mesa/drivers/dri/r600/r600_tex.c
index 41419f84601..512a52ede3e 100644
--- a/src/mesa/drivers/dri/r600/r600_tex.c
+++ b/src/mesa/drivers/dri/r600/r600_tex.c
@@ -431,7 +431,7 @@ unsigned r600IsFormatRenderable(gl_format mesa_format)
 	case MESA_FORMAT_Z24_S8:
 	case MESA_FORMAT_Z16:
 	case MESA_FORMAT_Z32:
-	case MESA_FORMAT_SRGBA8:
+	case MESA_FORMAT_SARGB8:
 	case MESA_FORMAT_SLA8:
 	case MESA_FORMAT_SL8:
 		return 1;
diff --git a/src/mesa/drivers/dri/r600/r600_texstate.c b/src/mesa/drivers/dri/r600/r600_texstate.c
index 1600033b9bd..ba3690b70ed 100644
--- a/src/mesa/drivers/dri/r600/r600_texstate.c
+++ b/src/mesa/drivers/dri/r600/r600_texstate.c
@@ -605,17 +605,17 @@ static GLboolean r600GetTexFormat(struct gl_texture_object *tObj, gl_format mesa
 		}
 		break;
 	/* EXT_texture_sRGB */
-	case MESA_FORMAT_SRGBA8:
+	case MESA_FORMAT_SARGB8:
 		SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8,
 			 SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
 
-		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_W,
-			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
 		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
-			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
 		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
-			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
 		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+		SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_W,
 			 SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
 		SETbit(t->SQ_TEX_RESOURCE4, SQ_TEX_RESOURCE_WORD4_0__FORCE_DEGAMMA_bit);
 		break;
diff --git a/src/mesa/drivers/dri/r600/r700_assembler.c b/src/mesa/drivers/dri/r600/r700_assembler.c
index 99a33df4fcb..9c954cbf70c 100644
--- a/src/mesa/drivers/dri/r600/r700_assembler.c
+++ b/src/mesa/drivers/dri/r600/r700_assembler.c
@@ -275,7 +275,10 @@ GLuint GetSurfaceFormat(GLenum eType, GLuint nChannels, GLuint * pClient_size)
                 case 2:
                     format = FMT_8_8; break;
                 case 3:
-                    format = FMT_8_8_8; break;
+                    /* for some (small/unaligned) strides using 4 comps works
+                     * better, probably same as GL_SHORT below
+                     * test piglit/draw-vertices */
+                    format = FMT_8_8_8_8; break;
                 case 4:
                     format = FMT_8_8_8_8; break;
                 default:
@@ -2872,25 +2875,92 @@ GLboolean assemble_CMP(r700_AssemblerBase *pAsm)
 
 GLboolean assemble_TRIG(r700_AssemblerBase *pAsm, BITS opcode)
 {
+    /* 
+     * r600 - trunc to -PI..PI range
+     * r700 - normalize by dividing by 2PI
+     * see fdo bug 27901
+     */
+  
     int tmp;
     checkop1(pAsm);
 
     tmp = gethelpr(pAsm);
 
-    pAsm->D.dst.opcode = SQ_OP2_INST_MUL;
+    pAsm->D.dst.opcode = SQ_OP3_INST_MULADD;
+    pAsm->D.dst.op3    = 1;
+
     setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
     pAsm->D.dst.rtype  = DST_REG_TEMPORARY;
     pAsm->D.dst.reg    = tmp;
-    pAsm->D.dst.writex = 1;
 
     assemble_src(pAsm, 0, -1);
 
     pAsm->S[1].src.rtype = SRC_REC_LITERAL;
     setswizzle_PVSSRC(&(pAsm->S[1].src), SQ_SEL_X);
+    
+    pAsm->S[2].src.rtype = SRC_REC_LITERAL;
+    setswizzle_PVSSRC(&(pAsm->S[2].src), SQ_SEL_Y);
+
     pAsm->D2.dst2.literal_slots = 1;
     pAsm->C[0].f = 1/(3.1415926535 * 2);
-    pAsm->C[1].f = 0.0F;
-    next_ins(pAsm);
+    pAsm->C[1].f = 0.5f;
+    
+    if ( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_FRACT;
+
+    setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+    pAsm->D.dst.rtype  = DST_REG_TEMPORARY;
+    pAsm->D.dst.reg    = tmp;
+    pAsm->D.dst.writex = 1;
+
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+    pAsm->S[0].src.reg   = tmp;
+    setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X);
+
+    if(( GL_FALSE == next_ins(pAsm) ))
+    {
+        return GL_FALSE;
+    }
+    pAsm->D.dst.opcode = SQ_OP3_INST_MULADD;
+    pAsm->D.dst.op3    = 1;
+
+    setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+    pAsm->D.dst.rtype  = DST_REG_TEMPORARY;
+    pAsm->D.dst.reg    = tmp;
+
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+    pAsm->S[0].src.reg   = tmp;
+    setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X);
+
+    pAsm->S[1].src.rtype = SRC_REC_LITERAL;
+    setswizzle_PVSSRC(&(pAsm->S[1].src), SQ_SEL_X);
+
+    pAsm->S[2].src.rtype = SRC_REC_LITERAL;
+    setswizzle_PVSSRC(&(pAsm->S[2].src), SQ_SEL_Y);
+
+    pAsm->D2.dst2.literal_slots = 1;
+
+    if (pAsm->bR6xx)
+    {
+       pAsm->C[0].f = 3.1415926535897f * 2.0f;
+       pAsm->C[1].f = -3.1415926535897f;
+    }
+    else 
+    {
+       pAsm->C[0].f = 1.0f;
+       pAsm->C[1].f = -0.5f;
+    }
+
+    if(( GL_FALSE == next_ins(pAsm) ))
+    {
+        return GL_FALSE;
+    }
 
     pAsm->D.dst.opcode = opcode;
     pAsm->D.dst.math = 1;
@@ -4030,22 +4100,79 @@ GLboolean assemble_SCS(r700_AssemblerBase *pAsm)
     checkop1(pAsm);
 
     tmp = gethelpr(pAsm);
-    /* tmp.x = src /2*PI */
-    pAsm->D.dst.opcode = SQ_OP2_INST_MUL;
+
+    pAsm->D.dst.opcode = SQ_OP3_INST_MULADD;
+    pAsm->D.dst.op3    = 1;
+
     setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
     pAsm->D.dst.rtype  = DST_REG_TEMPORARY;
     pAsm->D.dst.reg    = tmp;
-    pAsm->D.dst.writex = 1;
 
     assemble_src(pAsm, 0, -1);
 
     pAsm->S[1].src.rtype = SRC_REC_LITERAL;
     setswizzle_PVSSRC(&(pAsm->S[1].src), SQ_SEL_X);
+
+    pAsm->S[2].src.rtype = SRC_REC_LITERAL;
+    setswizzle_PVSSRC(&(pAsm->S[2].src), SQ_SEL_Y);
+
     pAsm->D2.dst2.literal_slots = 1;
     pAsm->C[0].f = 1/(3.1415926535 * 2);
-    pAsm->C[1].f = 0.0F;
+    pAsm->C[1].f = 0.5F;
 
-    next_ins(pAsm);
+    if ( GL_FALSE == next_ins(pAsm) )
+    {
+        return GL_FALSE;
+    }
+
+    pAsm->D.dst.opcode = SQ_OP2_INST_FRACT;
+
+    setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+    pAsm->D.dst.rtype  = DST_REG_TEMPORARY;
+    pAsm->D.dst.reg    = tmp;
+    pAsm->D.dst.writex = 1;
+
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+    pAsm->S[0].src.reg   = tmp;
+    setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X);
+
+    if(( GL_FALSE == next_ins(pAsm) ))
+    {
+        return GL_FALSE;
+    }
+    pAsm->D.dst.opcode = SQ_OP3_INST_MULADD;
+    pAsm->D.dst.op3    = 1;
+
+    setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE);
+    pAsm->D.dst.rtype  = DST_REG_TEMPORARY;
+    pAsm->D.dst.reg    = tmp;
+
+    setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE);
+    pAsm->S[0].src.rtype = SRC_REG_TEMPORARY;
+    pAsm->S[0].src.reg   = tmp;
+    setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X);
+
+    pAsm->S[1].src.rtype = SRC_REC_LITERAL;
+    setswizzle_PVSSRC(&(pAsm->S[1].src), SQ_SEL_X);
+
+    pAsm->S[2].src.rtype = SRC_REC_LITERAL;
+    setswizzle_PVSSRC(&(pAsm->S[2].src), SQ_SEL_Y);
+
+    pAsm->D2.dst2.literal_slots = 1;
+
+    if(pAsm->bR6xx) {
+       pAsm->C[0].f = 3.1415926535897f * 2.0f;
+       pAsm->C[1].f = -3.1415926535897f;
+    } else {
+       pAsm->C[0].f = 1.0f;
+       pAsm->C[1].f = -0.5f;
+    }
+
+    if(( GL_FALSE == next_ins(pAsm) ))
+    {
+        return GL_FALSE;
+    }
 
     // COS dst.x,    a.x
     pAsm->D.dst.opcode = SQ_OP2_INST_COS;
@@ -6473,7 +6600,7 @@ GLboolean Process_Fragment_Exports(r700_AssemblerBase *pR700AsmCode,
      * results are undefined anyway */
     if(export_count == 0)
     {
-        Process_Export(pR700AsmCode, SQ_EXPORT_PIXEL, 0, 1, 0, GL_FALSE);
+        Process_Export(pR700AsmCode, SQ_EXPORT_PIXEL, 0, 1, pR700AsmCode->starting_export_register_number, GL_FALSE);
     }
     
     if(pR700AsmCode->cf_last_export_ptr != NULL) 
diff --git a/src/mesa/drivers/dri/r600/r700_chip.c b/src/mesa/drivers/dri/r600/r700_chip.c
index cefda3ac4ba..bf8063391a2 100644
--- a/src/mesa/drivers/dri/r600/r700_chip.c
+++ b/src/mesa/drivers/dri/r600/r700_chip.c
@@ -265,17 +265,6 @@ static void r700SendVTXState(GLcontext *ctx, struct radeon_state_atom *atom)
     if (context->radeon.tcl.aos_count == 0)
 	    return;
 
-    BEGIN_BATCH_NO_AUTOSTATE(6);
-    R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, 1));
-    R600_OUT_BATCH(mmSQ_VTX_BASE_VTX_LOC - ASIC_CTL_CONST_BASE_INDEX);
-    R600_OUT_BATCH(0);
-
-    R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, 1));
-    R600_OUT_BATCH(mmSQ_VTX_START_INST_LOC - ASIC_CTL_CONST_BASE_INDEX);
-    R600_OUT_BATCH(0);
-    END_BATCH();
-    COMMIT_BATCH();
-
     for(i=0; i<VERT_ATTRIB_MAX; i++) {
 	    if(vp->mesa_program->Base.InputsRead & (1 << i))
 	    {
@@ -523,9 +512,9 @@ static void r700SetRenderTarget(context_t *context, int id)
 		     CB_COLOR0_INFO__ARRAY_MODE_shift, CB_COLOR0_INFO__ARRAY_MODE_mask);
 	    CLEARbit(r700->render_target[id].CB_COLOR0_INFO.u32All, SOURCE_FORMAT_bit);
             break;
-    case MESA_FORMAT_SRGBA8:
+    case MESA_FORMAT_SARGB8:
             format = COLOR_8_8_8_8;
-            comp_swap = SWAP_STD_REV;
+            comp_swap = SWAP_ALT;
 	    number_type = NUMBER_SRGB;
 	    SETbit(r700->render_target[id].CB_COLOR0_INFO.u32All, SOURCE_FORMAT_bit);
             break;
@@ -617,18 +606,25 @@ static void r700SendDepthTargetState(GLcontext *ctx, struct radeon_state_atom *a
 
 	r700SetDepthTarget(context);
 
-        BEGIN_BATCH_NO_AUTOSTATE(8 + 2);
+        BEGIN_BATCH_NO_AUTOSTATE(7 + 2);
 	R600_OUT_BATCH_REGSEQ(DB_DEPTH_SIZE, 2);
 	R600_OUT_BATCH(r700->DB_DEPTH_SIZE.u32All);
 	R600_OUT_BATCH(r700->DB_DEPTH_VIEW.u32All);
-	R600_OUT_BATCH_REGSEQ(DB_DEPTH_BASE, 2);
+	R600_OUT_BATCH_REGSEQ(DB_DEPTH_BASE, 1);
 	R600_OUT_BATCH(r700->DB_DEPTH_BASE.u32All);
-	R600_OUT_BATCH(r700->DB_DEPTH_INFO.u32All);
 	R600_OUT_BATCH_RELOC(r700->DB_DEPTH_BASE.u32All,
 			     rrb->bo,
 			     r700->DB_DEPTH_BASE.u32All,
 			     0, RADEON_GEM_DOMAIN_VRAM, 0);
         END_BATCH();
+        BEGIN_BATCH_NO_AUTOSTATE(3 + 2);
+	R600_OUT_BATCH_REGSEQ(DB_DEPTH_INFO, 1);
+	R600_OUT_BATCH(r700->DB_DEPTH_INFO.u32All);
+	R600_OUT_BATCH_RELOC(r700->DB_DEPTH_INFO.u32All,
+			     rrb->bo,
+			     r700->DB_DEPTH_INFO.u32All,
+			     0, RADEON_GEM_DOMAIN_VRAM, 0);
+        END_BATCH();
 
 	if ((context->radeon.radeonScreen->chip_family > CHIP_FAMILY_R600) &&
 	    (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)) {
@@ -687,27 +683,35 @@ static void r700SendRenderTargetState(GLcontext *ctx, struct radeon_state_atom *
 	BEGIN_BATCH_NO_AUTOSTATE(3 + 2);
 	R600_OUT_BATCH_REGSEQ(CB_COLOR0_TILE + (4 * id), 1);
 	R600_OUT_BATCH(r700->render_target[id].CB_COLOR0_TILE.u32All);
-	R600_OUT_BATCH_RELOC(r700->render_target[id].CB_COLOR0_BASE.u32All,
+	R600_OUT_BATCH_RELOC(r700->render_target[id].CB_COLOR0_TILE.u32All,
 			     rrb->bo,
-			     r700->render_target[id].CB_COLOR0_BASE.u32All,
+			     r700->render_target[id].CB_COLOR0_TILE.u32All,
 			     0, RADEON_GEM_DOMAIN_VRAM, 0);
 	END_BATCH();
 	BEGIN_BATCH_NO_AUTOSTATE(3 + 2);
 	R600_OUT_BATCH_REGSEQ(CB_COLOR0_FRAG + (4 * id), 1);
 	R600_OUT_BATCH(r700->render_target[id].CB_COLOR0_FRAG.u32All);
-	R600_OUT_BATCH_RELOC(r700->render_target[id].CB_COLOR0_BASE.u32All,
+	R600_OUT_BATCH_RELOC(r700->render_target[id].CB_COLOR0_FRAG.u32All,
 			     rrb->bo,
-			     r700->render_target[id].CB_COLOR0_BASE.u32All,
+			     r700->render_target[id].CB_COLOR0_FRAG.u32All,
 			     0, RADEON_GEM_DOMAIN_VRAM, 0);
         END_BATCH();
 
-        BEGIN_BATCH_NO_AUTOSTATE(12);
+        BEGIN_BATCH_NO_AUTOSTATE(9);
 	R600_OUT_BATCH_REGVAL(CB_COLOR0_SIZE + (4 * id), r700->render_target[id].CB_COLOR0_SIZE.u32All);
 	R600_OUT_BATCH_REGVAL(CB_COLOR0_VIEW + (4 * id), r700->render_target[id].CB_COLOR0_VIEW.u32All);
-	R600_OUT_BATCH_REGVAL(CB_COLOR0_INFO + (4 * id), r700->render_target[id].CB_COLOR0_INFO.u32All);
 	R600_OUT_BATCH_REGVAL(CB_COLOR0_MASK + (4 * id), r700->render_target[id].CB_COLOR0_MASK.u32All);
         END_BATCH();
 
+	BEGIN_BATCH_NO_AUTOSTATE(3 + 2);
+	R600_OUT_BATCH_REGVAL(CB_COLOR0_INFO + (4 * id), r700->render_target[id].CB_COLOR0_INFO.u32All);
+	R600_OUT_BATCH_RELOC(r700->render_target[id].CB_COLOR0_INFO.u32All,
+			     rrb->bo,
+			     r700->render_target[id].CB_COLOR0_INFO.u32All,
+			     0, RADEON_GEM_DOMAIN_VRAM, 0);
+
+        END_BATCH();
+
 	COMMIT_BATCH();
 
 }
@@ -1465,9 +1469,6 @@ static int check_vtx(GLcontext *ctx, struct radeon_state_atom *atom)
 	context_t *context = R700_CONTEXT(ctx);
 	int count = context->radeon.tcl.aos_count * 18;
 
-	if (count)
-		count += 6;
-
 	radeon_print(RADEON_STATE, RADEON_TRACE, "%s %d\n", __func__, count);
 	return count;
 }
@@ -1567,7 +1568,7 @@ void r600InitAtoms(context_t *context)
 	ALLOC_STATE(sq, always, 34, r700SendSQConfig);
 	ALLOC_STATE(db, always, 17, r700SendDBState);
 	ALLOC_STATE(stencil, always, 4, r700SendStencilState);
-	ALLOC_STATE(db_target, always, 12, r700SendDepthTargetState);
+	ALLOC_STATE(db_target, always, 16, r700SendDepthTargetState);
 	ALLOC_STATE(sc, always, 15, r700SendSCState);
 	ALLOC_STATE(scissor, always, 22, r700SendScissorState);
 	ALLOC_STATE(aa, always, 12, r700SendAAState);
@@ -1578,7 +1579,7 @@ void r600InitAtoms(context_t *context)
 	ALLOC_STATE(poly, always, 10, r700SendPolyState);
 	ALLOC_STATE(cb, cb, 18, r700SendCBState);
 	ALLOC_STATE(clrcmp, always, 6, r700SendCBCLRCMPState);
-	ALLOC_STATE(cb_target, always, 29, r700SendRenderTargetState);
+	ALLOC_STATE(cb_target, always, 31, r700SendRenderTargetState);
 	ALLOC_STATE(blnd, blnd, (6 + (R700_MAX_RENDER_TARGETS * 3)), r700SendCBBlendState);
 	ALLOC_STATE(blnd_clr, always, 6, r700SendCBBlendColorState);
 	ALLOC_STATE(sx, always, 9, r700SendSXState);
@@ -1590,7 +1591,7 @@ void r600InitAtoms(context_t *context)
 	ALLOC_STATE(ps, always, 24, r700SendPSState);
 	ALLOC_STATE(vs_consts, vs_consts, (2 + (R700_MAX_DX9_CONSTS * 4)), r700SendVSConsts);
 	ALLOC_STATE(ps_consts, ps_consts, (2 + (R700_MAX_DX9_CONSTS * 4)), r700SendPSConsts);
-	ALLOC_STATE(vtx, vtx, (6 + (VERT_ATTRIB_MAX * 18)), r700SendVTXState);
+	ALLOC_STATE(vtx, vtx, (VERT_ATTRIB_MAX * 18), r700SendVTXState);
 	ALLOC_STATE(tx, tx, (R700_TEXTURE_NUMBERUNITS * 20), r700SendTexState);
 	ALLOC_STATE(tx_smplr, tx, (R700_TEXTURE_NUMBERUNITS * 5), r700SendTexSamplerState);
 	ALLOC_STATE(tx_brdr_clr, tx, (R700_TEXTURE_NUMBERUNITS * 6), r700SendTexBorderColorState);
diff --git a/src/mesa/drivers/dri/r600/r700_clear.c b/src/mesa/drivers/dri/r600/r700_clear.c
index 09c48565b68..d1008f28b9b 100644
--- a/src/mesa/drivers/dri/r600/r700_clear.c
+++ b/src/mesa/drivers/dri/r600/r700_clear.c
@@ -48,6 +48,7 @@ static GLboolean r700ClearFast(context_t *context, GLbitfield mask)
 void r700Clear(GLcontext * ctx, GLbitfield mask)
 {
     context_t *context = R700_CONTEXT(ctx);
+    radeonContextPtr radeon = &context->radeon;
     __DRIdrawable *dPriv = radeon_get_drawable(&context->radeon);
     const GLuint colorMask = *((GLuint *) & ctx->Color.ColorMask[0]);
     GLbitfield swrast_mask = 0, tri_mask = 0;
@@ -60,6 +61,8 @@ void r700Clear(GLcontext * ctx, GLbitfield mask)
         context->radeon.front_buffer_dirty = GL_TRUE;
     }
 
+    radeon_prepare_render(radeon);
+
     if( GL_TRUE == r700ClearFast(context, mask) )
     {
         return;
diff --git a/src/mesa/drivers/dri/r600/r700_render.c b/src/mesa/drivers/dri/r600/r700_render.c
index 1929b7cc129..c5771f9fd0b 100644
--- a/src/mesa/drivers/dri/r600/r700_render.c
+++ b/src/mesa/drivers/dri/r600/r700_render.c
@@ -244,7 +244,8 @@ static int r700NumVerts(int num_verts, int prim)
 	return num_verts - verts_off;
 }
 
-static void r700RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim)
+static void r700RunRenderPrimitive(GLcontext * ctx, int start, int end,
+				   int prim, GLint basevertex)
 {
     context_t *context = R700_CONTEXT(ctx);
     BATCH_LOCALS(&context->radeon);
@@ -282,6 +283,7 @@ static void r700RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim
     total_emit =   3  /* VGT_PRIMITIVE_TYPE */
 	         + 2  /* VGT_INDEX_TYPE */
 	         + 2  /* NUM_INSTANCES */
+		 + 4  /* VTX_BASE_VTX_LOC + VTX_START_INST_LOC */
 	         + 5 + 2; /* DRAW_INDEX */
 
     BEGIN_BATCH_NO_AUTOSTATE(total_emit);
@@ -294,6 +296,11 @@ static void r700RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim
     // num instances
     R600_OUT_BATCH(CP_PACKET3(R600_IT_NUM_INSTANCES, 0));
     R600_OUT_BATCH(1);
+    /* offset */
+    R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, 2));
+    R600_OUT_BATCH(mmSQ_VTX_BASE_VTX_LOC - ASIC_CTL_CONST_BASE_INDEX);
+    R600_OUT_BATCH(basevertex); //VTX_BASE_VTX_LOC
+    R600_OUT_BATCH(0);          //VTX_START_INST_LOC
     // draw packet
     R600_OUT_BATCH(CP_PACKET3(R600_IT_DRAW_INDEX, 3));
     R600_OUT_BATCH(context->ind_buf.bo_offset);
@@ -364,6 +371,7 @@ static void r700RunRenderPrimitiveImmediate(GLcontext * ctx, int start, int end,
     total_emit +=   3 /* VGT_PRIMITIVE_TYPE */
 	          + 2 /* VGT_INDEX_TYPE */
 	          + 2 /* NUM_INSTANCES */
+		  + 4 /* VTX_BASE_VTX_LOC + VTX_START_INST_LOC */
 	          + 3; /* DRAW */
 
     BEGIN_BATCH_NO_AUTOSTATE(total_emit);
@@ -376,6 +384,11 @@ static void r700RunRenderPrimitiveImmediate(GLcontext * ctx, int start, int end,
     // num instances
     R600_OUT_BATCH(CP_PACKET3(R600_IT_NUM_INSTANCES, 0));
     R600_OUT_BATCH(1);
+    /* offset */
+    R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, 2));
+    R600_OUT_BATCH(mmSQ_VTX_BASE_VTX_LOC - ASIC_CTL_CONST_BASE_INDEX);
+    R600_OUT_BATCH(0); //VTX_BASE_VTX_LOC
+    R600_OUT_BATCH(0); //VTX_START_INST_LOC
     // draw packet
     if(start == 0)
     {
@@ -433,16 +446,16 @@ static GLuint r700PredictRenderSize(GLcontext* ctx,
 
     dwords = PRE_EMIT_STATE_BUFSZ;
     if (ib)
-	    dwords += nr_prims * 14;
+	    dwords += nr_prims * 18;
     else {
 	    for (i = 0; i < nr_prims; ++i)
 	    {
 		    if (prim[i].start == 0)
-			    dwords += 10;
+			    dwords += 14;
 		    else if (prim[i].count > 0xffff)
-			    dwords += prim[i].count + 10;
+			    dwords += prim[i].count + 14;
 		    else
-			    dwords += ((prim[i].count + 1) / 2) + 10;
+			    dwords += ((prim[i].count + 1) / 2) + 14;
 	    }
     }
 
@@ -625,11 +638,11 @@ static void r700SetupStreams(GLcontext *ctx, const struct gl_client_array *input
 
         stride = (input[i]->StrideB == 0) ? getTypeSize(input[i]->Type) * input[i]->Size : input[i]->StrideB;
 
-        if (input[i]->Type == GL_DOUBLE || input[i]->Type == GL_UNSIGNED_INT || input[i]->Type == GL_INT ||
+        if (input[i]->Type == GL_DOUBLE || input[i]->Type == GL_UNSIGNED_INT || input[i]->Type == GL_INT
 #if MESA_BIG_ENDIAN
-            getTypeSize(input[i]->Type) != 4 || 
+            || getTypeSize(input[i]->Type) != 4
 #endif
-            stride < 4) 
+            ) 
         {
             r700ConvertAttrib(ctx, count, input[i], &context->stream_desc[index]);
         } 
@@ -637,19 +650,10 @@ static void r700SetupStreams(GLcontext *ctx, const struct gl_client_array *input
         {
             if (input[i]->BufferObj->Name) 
             {
-                if (stride % 4 != 0) 
-                {
-                    assert(((intptr_t) input[i]->Ptr) % input[i]->StrideB == 0);
-                    r700AlignDataToDword(ctx, input[i], count, &context->stream_desc[index]);
-                    context->stream_desc[index].is_named_bo = GL_FALSE;
-                } 
-                else 
-                {
-                    context->stream_desc[index].stride = input[i]->StrideB;
-                    context->stream_desc[index].bo_offset = (intptr_t) input[i]->Ptr;
-                    context->stream_desc[index].bo = get_radeon_buffer_object(input[i]->BufferObj)->bo;
-                    context->stream_desc[index].is_named_bo = GL_TRUE;
-                }
+                context->stream_desc[index].stride = input[i]->StrideB;
+                context->stream_desc[index].bo_offset = (intptr_t) input[i]->Ptr;
+                context->stream_desc[index].bo = get_radeon_buffer_object(input[i]->BufferObj)->bo;
+                context->stream_desc[index].is_named_bo = GL_TRUE;
             } 
             else 
             {
@@ -932,7 +936,8 @@ static GLboolean r700TryDrawPrims(GLcontext *ctx,
 		    r700RunRenderPrimitive(ctx,
 					   prim[i].start,
 					   prim[i].start + prim[i].count,
-					   prim[i].mode);
+					   prim[i].mode,
+					   prim[i].basevertex);
 	    else
 		    r700RunRenderPrimitiveImmediate(ctx,
 						    prim[i].start,
@@ -977,18 +982,24 @@ static void r700DrawPrims(GLcontext *ctx,
 {
 	GLboolean retval = GL_FALSE;
 
+	context_t *context = R700_CONTEXT(ctx);
+	radeonContextPtr radeon = &context->radeon;
+	radeon_prepare_render(radeon);
+
 	/* This check should get folded into just the places that
 	 * min/max index are really needed.
 	 */
-	if (!index_bounds_valid) {
-		vbo_get_minmax_index(ctx, prim, ib, &min_index, &max_index);
-	}
 
-	if (min_index) {
+	if (!vbo_all_varyings_in_vbos(arrays)) {
+	    if (!index_bounds_valid)
+		vbo_get_minmax_index(ctx, prim, ib, &min_index, &max_index);
+	    /* do we want to rebase, minimizes the 
+	     * amount of data to upload? */
+	    if (min_index) {
 		vbo_rebase_prims( ctx, arrays, prim, nr_prims, ib, min_index, max_index, r700DrawPrims );
 		return;
+	    }
 	}
-
 	/* Make an attempt at drawing */
 	retval = r700TryDrawPrims(ctx, arrays, prim, nr_prims, ib, min_index, max_index);
 
diff --git a/src/mesa/drivers/dri/r600/r700_vertprog.c b/src/mesa/drivers/dri/r600/r700_vertprog.c
index 137f3007ced..6a2a09eaf1a 100644
--- a/src/mesa/drivers/dri/r600/r700_vertprog.c
+++ b/src/mesa/drivers/dri/r600/r700_vertprog.c
@@ -461,11 +461,11 @@ static void r700TranslateAttrib(GLcontext *ctx, GLuint unLoc, int count, const s
 	stride = (input->StrideB == 0) ? getTypeSize(input->Type) * input->Size 
                                    : input->StrideB;
 
-    if (input->Type == GL_DOUBLE || input->Type == GL_UNSIGNED_INT || input->Type == GL_INT ||
+    if (input->Type == GL_DOUBLE || input->Type == GL_UNSIGNED_INT || input->Type == GL_INT
 #if MESA_BIG_ENDIAN
-        getTypeSize(input->Type) != 4 ||
+        || getTypeSize(input->Type) != 4
 #endif
-        stride < 4) 
+       ) 
     {
         pStreamDesc->type = GL_FLOAT;
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_chipset.h b/src/mesa/drivers/dri/radeon/radeon_chipset.h
index b7ee9a134bf..7d54fabebbc 100644
--- a/src/mesa/drivers/dri/radeon/radeon_chipset.h
+++ b/src/mesa/drivers/dri/radeon/radeon_chipset.h
@@ -414,9 +414,9 @@ enum {
    CHIP_FAMILY_R350,
    CHIP_FAMILY_RV350,
    CHIP_FAMILY_RV380,
+   CHIP_FAMILY_RS400,
    CHIP_FAMILY_R420,
    CHIP_FAMILY_RV410,
-   CHIP_FAMILY_RS400,
    CHIP_FAMILY_RS600,
    CHIP_FAMILY_RS690,
    CHIP_FAMILY_RS740,
diff --git a/src/mesa/drivers/dri/radeon/radeon_common.c b/src/mesa/drivers/dri/radeon/radeon_common.c
index 13f1f0611b8..c1a660af3d0 100644
--- a/src/mesa/drivers/dri/radeon/radeon_common.c
+++ b/src/mesa/drivers/dri/radeon/radeon_common.c
@@ -708,7 +708,6 @@ void radeon_draw_buffer(GLcontext *ctx, struct gl_framebuffer *fb)
 		if (fb->_ColorDrawBufferIndexes[0] == BUFFER_FRONT_LEFT) {
 			rrbColor = radeon_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
 			radeon->front_cliprects = GL_TRUE;
-			radeon->front_buffer_dirty = GL_TRUE;
 		} else {
 			rrbColor = radeon_renderbuffer(fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer);
 			radeon->front_cliprects = GL_FALSE;
@@ -1132,17 +1131,13 @@ flush_front:
 		if (screen->dri2.loader && (screen->dri2.loader->base.version >= 2)
 			&& (screen->dri2.loader->flushFrontBuffer != NULL)) {
 			__DRIdrawable * drawable = radeon_get_drawable(radeon);
-			(*screen->dri2.loader->flushFrontBuffer)(drawable, drawable->loaderPrivate);
 
-			/* Only clear the dirty bit if front-buffer rendering is no longer
-			 * enabled.  This is done so that the dirty bit can only be set in
-			 * glDrawBuffer.  Otherwise the dirty bit would have to be set at
-			 * each of N places that do rendering.  This has worse performances,
-			 * but it is much easier to get correct.
+			/* We set the dirty bit in radeon_prepare_render() if we're
+			 * front buffer rendering once we get there.
 			 */
-			if (!radeon->is_front_buffer_rendering) {
-				radeon->front_buffer_dirty = GL_FALSE;
-			}
+			radeon->front_buffer_dirty = GL_FALSE;
+
+			(*screen->dri2.loader->flushFrontBuffer)(drawable, drawable->loaderPrivate);
 		}
 	}
 }
diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.c b/src/mesa/drivers/dri/radeon/radeon_common_context.c
index 5a7d52c4d2f..92663bf66d7 100644
--- a/src/mesa/drivers/dri/radeon/radeon_common_context.c
+++ b/src/mesa/drivers/dri/radeon/radeon_common_context.c
@@ -493,6 +493,50 @@ radeon_bits_per_pixel(const struct radeon_renderbuffer *rb)
    return _mesa_get_format_bytes(rb->base.Format) * 8; 
 }
 
+/*
+ * Check if drawable has been invalidated by dri2InvalidateDrawable().
+ * Update renderbuffers if so. This prevents a client from accessing
+ * a backbuffer that has a swap pending but not yet completed.
+ *
+ * See intel_prepare_render for equivalent code in intel driver.
+ *
+ */
+void radeon_prepare_render(radeonContextPtr radeon)
+{
+    __DRIcontext *driContext = radeon->dri.context;
+    __DRIdrawable *drawable;
+    __DRIscreen *screen;
+
+    screen = driContext->driScreenPriv;
+    if (!screen->dri2.loader)
+        return;
+
+    drawable = driContext->driDrawablePriv;
+    if (drawable->dri2.stamp != driContext->dri2.draw_stamp) {
+	if (drawable->lastStamp != drawable->dri2.stamp)
+	    radeon_update_renderbuffers(driContext, drawable, GL_FALSE);
+
+	/* Intel driver does the equivalent of this, no clue if it is needed:
+	 * radeon_draw_buffer(radeon->glCtx, &(drawable->driverPrivate)->base);
+	 */
+	driContext->dri2.draw_stamp = drawable->dri2.stamp;
+    }
+
+    drawable = driContext->driReadablePriv;
+    if (drawable->dri2.stamp != driContext->dri2.read_stamp) {
+	if (drawable->lastStamp != drawable->dri2.stamp)
+	    radeon_update_renderbuffers(driContext, drawable, GL_FALSE);
+	driContext->dri2.read_stamp = drawable->dri2.stamp;
+    }
+
+    /* If we're currently rendering to the front buffer, the rendering
+     * that will happen next will probably dirty the front buffer.  So
+     * mark it as dirty here.
+     */
+    if (radeon->is_front_buffer_rendering)
+	radeon->front_buffer_dirty = GL_TRUE;
+}
+
 void
 radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable,
 			    GLboolean front_only)
@@ -514,6 +558,11 @@ radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable,
 	screen = context->driScreenPriv;
 	radeon = (radeonContextPtr) context->driverPrivate;
 
+	/* Set this up front, so that in case our buffers get invalidated
+	 * while we're getting new buffers, we don't clobber the stamp and
+	 * thus ignore the invalidate. */
+	drawable->lastStamp = drawable->dri2.stamp;
+
 	if (screen->dri2.loader
 	   && (screen->dri2.loader->base.version > 2)
 	   && (screen->dri2.loader->getBuffersWithFormat != NULL)) {
@@ -650,6 +699,13 @@ radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable,
 		rb->base.Height = drawable->h;
 		rb->has_surface = 0;
 
+		/* r6xx+ tiling */
+		rb->tile_config = radeon->radeonScreen->tile_config;
+		rb->group_bytes = radeon->radeonScreen->group_bytes;
+		rb->num_channels = radeon->radeonScreen->num_channels;
+		rb->num_banks = radeon->radeonScreen->num_banks;
+		rb->r7xx_bank_op = radeon->radeonScreen->r7xx_bank_op;
+
 		if (buffers[i].attachment == __DRI_BUFFER_STENCIL && depth_bo) {
 			if (RADEON_DEBUG & RADEON_DRI)
 				fprintf(stderr, "(reusing depth buffer as stencil)\n");
@@ -678,7 +734,7 @@ radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable,
 				bo->flags |= RADEON_BO_FLAGS_MACRO_TILE;
 			if (tiling_flags & RADEON_TILING_MICRO)
 				bo->flags |= RADEON_BO_FLAGS_MICRO_TILE;
-			
+
 		}
 
 		if (buffers[i].attachment == __DRI_BUFFER_DEPTH) {
diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.h b/src/mesa/drivers/dri/radeon/radeon_common_context.h
index 5156c5d0d0a..f06e5fdf244 100644
--- a/src/mesa/drivers/dri/radeon/radeon_common_context.h
+++ b/src/mesa/drivers/dri/radeon/radeon_common_context.h
@@ -93,6 +93,13 @@ struct radeon_renderbuffer
 	GLuint pf_pending;  /**< sequence number of pending flip */
 	GLuint vbl_pending;   /**< vblank sequence number of pending flip */
 	__DRIdrawable *dPriv;
+
+	/* r6xx+ tiling */
+	GLuint tile_config;
+	GLint group_bytes;
+	GLint num_channels;
+	GLint num_banks;
+	GLint r7xx_bank_op;
 };
 
 struct radeon_framebuffer
@@ -614,5 +621,6 @@ GLboolean radeonMakeCurrent(__DRIcontext * driContextPriv,
 			    __DRIdrawable * driDrawPriv,
 			    __DRIdrawable * driReadPriv);
 extern void radeonDestroyContext(__DRIcontext * driContextPriv);
+void radeon_prepare_render(radeonContextPtr radeon);
 
 #endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c
index c877e6c1765..c6e5f110ea3 100644
--- a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c
+++ b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c
@@ -133,7 +133,7 @@ static void compute_tex_image_offset(radeonContextPtr rmesa, radeon_mipmap_tree
 	height = _mesa_next_pow_two_32(lvl->height);
 
 	lvl->rowstride = get_texture_image_row_stride(rmesa, mt->mesaFormat, lvl->width, mt->tilebits);
-	lvl->size = get_texture_image_size(mt->mesaFormat, lvl->rowstride, lvl->height, lvl->depth, mt->tilebits);
+	lvl->size = get_texture_image_size(mt->mesaFormat, lvl->rowstride, height, lvl->depth, mt->tilebits);
 
 	assert(lvl->size > 0);
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_pixel_read.c b/src/mesa/drivers/dri/radeon/radeon_pixel_read.c
index dadb8002c7d..fb741173ca8 100644
--- a/src/mesa/drivers/dri/radeon/radeon_pixel_read.c
+++ b/src/mesa/drivers/dri/radeon/radeon_pixel_read.c
@@ -179,6 +179,9 @@ radeonReadPixels(GLcontext * ctx,
                  GLenum format, GLenum type,
                  const struct gl_pixelstore_attrib *pack, GLvoid * pixels)
 {
+    radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+    radeon_prepare_render(radeon);
+
     if (do_blit_readpixels(ctx, x, y, width, height, format, type, pack, pixels))
         return;
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.c b/src/mesa/drivers/dri/radeon/radeon_screen.c
index 82107cc6aeb..fa97a19302c 100644
--- a/src/mesa/drivers/dri/radeon/radeon_screen.c
+++ b/src/mesa/drivers/dri/radeon/radeon_screen.c
@@ -213,6 +213,10 @@ static const GLuint __driNConfigOptions = 17;
 
 static int getSwapInfo( __DRIdrawable *dPriv, __DRIswapInfo * sInfo );
 
+#ifndef RADEON_INFO_TILE_CONFIG
+#define RADEON_INFO_TILE_CONFIG 0x6
+#endif
+
 static int
 radeonGetParam(__DRIscreen *sPriv, int param, void *value)
 {
@@ -232,6 +236,9 @@ radeonGetParam(__DRIscreen *sPriv, int param, void *value)
       case RADEON_PARAM_NUM_Z_PIPES:
           info.request = RADEON_INFO_NUM_Z_PIPES;
           break;
+      case RADEON_INFO_TILE_CONFIG:
+	  info.request = RADEON_INFO_TILE_CONFIG;
+          break;
       default:
           return -EINVAL;
       }
@@ -376,6 +383,21 @@ static const __DRItexBufferExtension r600TexBufferExtension = {
 };
 #endif
 
+static void
+radeonDRI2Flush(__DRIdrawable *drawable)
+{
+    radeonContextPtr rmesa;
+
+    rmesa = (radeonContextPtr) drawable->driContextPriv->driverPrivate;
+    radeonFlush(rmesa->glCtx);
+}
+
+static const struct __DRI2flushExtensionRec radeonFlushExtension = {
+    { __DRI2_FLUSH, __DRI2_FLUSH_VERSION },
+    radeonDRI2Flush,
+    dri2InvalidateDrawable,
+};
+
 static int radeon_set_screen_flags(radeonScreenPtr screen, int device_id)
 {
    screen->device_id = device_id;
@@ -1305,6 +1327,56 @@ radeonCreateScreen2(__DRIscreen *sPriv)
    else
 	   screen->chip_flags |= RADEON_CLASS_R600;
 
+   /* r6xx+ tiling */
+   if (IS_R600_CLASS(screen) && (sPriv->drm_version.minor >= 6)) {
+	   ret = radeonGetParam(sPriv, RADEON_INFO_TILE_CONFIG, &temp);
+	   if (ret)
+		   fprintf(stderr, "failed to get tiling info\n");
+	   else {
+		   screen->tile_config = temp;
+		   screen->r7xx_bank_op = 0;
+		   switch((screen->tile_config & 0xe) >> 1) {
+		   case 0:
+			   screen->num_channels = 1;
+			   break;
+		   case 1:
+			   screen->num_channels = 2;
+			   break;
+		   case 2:
+			   screen->num_channels = 4;
+			   break;
+		   case 3:
+			   screen->num_channels = 8;
+			   break;
+		   default:
+			   fprintf(stderr, "bad channels\n");
+			   break;
+		   }
+		   switch((screen->tile_config & 0x30) >> 4) {
+		   case 0:
+			   screen->num_banks = 4;
+			   break;
+		   case 1:
+			   screen->num_banks = 8;
+			   break;
+		   default:
+			   fprintf(stderr, "bad banks\n");
+			   break;
+		   }
+		   switch((screen->tile_config & 0xc0) >> 6) {
+		   case 0:
+			   screen->group_bytes = 256;
+			   break;
+		   case 1:
+			   screen->group_bytes = 512;
+			   break;
+		   default:
+			   fprintf(stderr, "bad group_bytes\n");
+			   break;
+		   }
+	   }
+   }
+
    if (IS_R300_CLASS(screen)) {
        ret = radeonGetParam(sPriv, RADEON_PARAM_NUM_GB_PIPES, &temp);
        if (ret) {
@@ -1379,6 +1451,8 @@ radeonCreateScreen2(__DRIscreen *sPriv)
    screen->extensions[i++] = &r600TexBufferExtension.base;
 #endif
 
+   screen->extensions[i++] = &radeonFlushExtension.base;
+
    screen->extensions[i++] = NULL;
    sPriv->extensions = screen->extensions;
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.h b/src/mesa/drivers/dri/radeon/radeon_screen.h
index 0d7e335fa3a..2b33201a538 100644
--- a/src/mesa/drivers/dri/radeon/radeon_screen.h
+++ b/src/mesa/drivers/dri/radeon/radeon_screen.h
@@ -112,6 +112,13 @@ typedef struct radeon_screen {
    int kernel_mm;
    drm_radeon_sarea_t *sarea;	/* Private SAREA data */
    struct radeon_bo_manager *bom;
+
+   /* r6xx+ tiling */
+   GLuint tile_config;
+   GLint group_bytes;
+   GLint num_channels;
+   GLint num_banks;
+   GLint r7xx_bank_op;
 } radeonScreenRec, *radeonScreenPtr;
 
 #define IS_R100_CLASS(screen) \
diff --git a/src/mesa/drivers/dri/radeon/radeon_span.c b/src/mesa/drivers/dri/radeon/radeon_span.c
index 1adb6096033..9dfe2dd2433 100644
--- a/src/mesa/drivers/dri/radeon/radeon_span.c
+++ b/src/mesa/drivers/dri/radeon/radeon_span.c
@@ -111,7 +111,6 @@ static GLubyte *r200_depth_4byte(const struct radeon_renderbuffer * rrb,
  * two main types:
  * - 1D (akin to macro-linear/micro-tiled on older asics)
  * - 2D (akin to macro-tiled/micro-tiled on older asics)
- * only 1D tiling is implemented below
  */
 #if defined(RADEON_R600)
 static inline GLint r600_1d_tile_helper(const struct radeon_renderbuffer * rrb,
@@ -208,12 +207,190 @@ static inline GLint r600_1d_tile_helper(const struct radeon_renderbuffer * rrb,
     return offset;
 }
 
+static inline GLint r600_log2(GLint n)
+{
+	GLint log2 = 0;
+
+	while (n >>= 1)
+		++log2;
+	return log2;
+}
+
+static inline GLint r600_2d_tile_helper(const struct radeon_renderbuffer * rrb,
+					GLint x, GLint y, GLint is_depth, GLint is_stencil)
+{
+	GLint group_bytes = rrb->group_bytes;
+	GLint num_channels = rrb->num_channels;
+	GLint num_banks = rrb->num_banks;
+	GLint r7xx_bank_op = rrb->r7xx_bank_op;
+	/* */
+	GLint group_bits = r600_log2(group_bytes);
+	GLint channel_bits = r600_log2(num_channels);
+	GLint bank_bits = r600_log2(num_banks);
+	GLint element_bytes = rrb->cpp;
+	GLint num_samples = 1;
+	GLint tile_width = 8;
+	GLint tile_height = 8;
+	GLint tile_thickness = 1;
+	GLint macro_tile_width = num_banks;
+	GLint macro_tile_height = num_channels;
+	GLint pitch_elements = (rrb->pitch / element_bytes) / tile_width;
+	GLint height = rrb->base.Height / tile_height;
+	GLint z = 0;
+	GLint sample_number = 0;
+	/* */
+	GLint tile_bytes;
+	GLint macro_tile_bytes;
+	GLint macro_tiles_per_row;
+	GLint macro_tiles_per_slice;
+	GLint slice_offset;
+	GLint macro_tile_row_index;
+	GLint macro_tile_column_index;
+	GLint macro_tile_offset;
+	GLint pixel_number = 0;
+	GLint element_offset;
+	GLint bank = 0;
+	GLint channel = 0;
+	GLint total_offset;
+	GLint group_mask = (1 << group_bits) - 1;
+	GLint offset_low;
+	GLint offset_high;
+	GLint offset = 0;
+
+	switch (num_channels) {
+	case 2:
+	default:
+		// channel[0] = x[3] ^ y[3]
+		channel |= (((x >> 3) ^ (y >> 3)) & 1) << 0;
+		break;
+	case 4:
+		// channel[0] = x[4] ^ y[3]
+		channel |= (((x >> 4) ^ (y >> 3)) & 1) << 0;
+		// channel[1] = x[3] ^ y[4]
+		channel |= (((x >> 3) ^ (y >> 4)) & 1) << 1;
+		break;
+	case 8:
+		// channel[0] = x[5] ^ y[3]
+		channel |= (((x >> 5) ^ (y >> 3)) & 1) << 0;
+		// channel[0] = x[4] ^ x[5] ^ y[4]
+		channel |= (((x >> 4) ^ (x >> 5) ^ (y >> 4)) & 1) << 1;
+		// channel[0] = x[3] ^ y[5]
+		channel |= (((x >> 3) ^ (y >> 5)) & 1) << 2;
+		break;
+	}
+
+	switch (num_banks) {
+	case 4:
+		// bank[0] = x[3] ^ y[4 + log2(num_channels)]
+		bank |= (((x >> 3) ^ (y >> (4 + channel_bits))) & 1) << 0;
+		if (r7xx_bank_op)
+			// bank[1] = x[3] ^ y[4 + log2(num_channels)] ^ x[5]
+			bank |= (((x >> 4) ^ (y >> (3 + channel_bits)) ^ (x >> 5)) & 1) << 1;
+		else
+			// bank[1] = x[4] ^ y[3 + log2(num_channels)]
+			bank |= (((x >> 4) ^ (y >> (3 + channel_bits))) & 1) << 1;
+		break;
+	case 8:
+		// bank[0] = x[3] ^ y[5 + log2(num_channels)]
+		bank |= (((x >> 3) ^ (y >> (5 + channel_bits))) & 1) << 0;
+		// bank[1] = x[4] ^ y[4 + log2(num_channels)] ^ y[5 + log2(num_channels)]
+		bank |= (((x >> 4) ^ (y >> (4 + channel_bits)) ^ (y >> (5 + channel_bits))) & 1) << 1;
+		if (r7xx_bank_op)
+			// bank[2] = x[5] ^ y[3 + log2(num_channels)] ^ x[6]
+			bank |= (((x >> 5) ^ (y >> (3 + channel_bits)) ^ (x >> 6)) & 1) << 2;
+		else
+			// bank[2] = x[5] ^ y[3 + log2(num_channels)]
+			bank |= (((x >> 5) ^ (y >> (3 + channel_bits))) & 1) << 2;
+		break;
+	}
+
+	tile_bytes = tile_width * tile_height * tile_thickness * element_bytes * num_samples;
+	macro_tile_bytes = macro_tile_width * macro_tile_height * tile_bytes;
+	macro_tiles_per_row = pitch_elements / macro_tile_width;
+	macro_tiles_per_slice = macro_tiles_per_row * (height / macro_tile_height);
+	slice_offset = (z / tile_thickness) * macro_tiles_per_slice * macro_tile_bytes;
+	macro_tile_row_index = (y / tile_height) / macro_tile_height;
+	macro_tile_column_index = (x / tile_width) / macro_tile_width;
+	macro_tile_offset = ((macro_tile_row_index * macro_tiles_per_row) + macro_tile_column_index) * macro_tile_bytes;
+
+	if (is_depth) {
+		GLint pixel_offset = 0;
+
+		pixel_number |= ((x >> 0) & 1) << 0; // pn[0] = x[0]
+		pixel_number |= ((y >> 0) & 1) << 1; // pn[1] = y[0]
+		pixel_number |= ((x >> 1) & 1) << 2; // pn[2] = x[1]
+		pixel_number |= ((y >> 1) & 1) << 3; // pn[3] = y[1]
+		pixel_number |= ((x >> 2) & 1) << 4; // pn[4] = x[2]
+		pixel_number |= ((y >> 2) & 1) << 5; // pn[5] = y[2]
+		switch (element_bytes) {
+		case 2:
+			pixel_offset = pixel_number * element_bytes * num_samples;
+			break;
+		case 4:
+			/* stencil and depth data are stored separately within a tile.
+			 * stencil is stored in a contiguous tile before the depth tile.
+			 * stencil element is 1 byte, depth element is 3 bytes.
+			 * stencil tile is 64 bytes.
+			 */
+			if (is_stencil)
+				pixel_offset = pixel_number * 1 * num_samples;
+			else
+				pixel_offset = (pixel_number * 3 * num_samples) + 64;
+			break;
+		}
+		element_offset = pixel_offset + (sample_number * element_bytes);
+	} else {
+		GLint sample_offset;
+
+		switch (element_bytes) {
+		case 1:
+			pixel_number |= ((x >> 0) & 1) << 0; // pn[0] = x[0]
+			pixel_number |= ((x >> 1) & 1) << 1; // pn[1] = x[1]
+			pixel_number |= ((x >> 2) & 1) << 2; // pn[2] = x[2]
+			pixel_number |= ((y >> 1) & 1) << 3; // pn[3] = y[1]
+			pixel_number |= ((y >> 0) & 1) << 4; // pn[4] = y[0]
+			pixel_number |= ((y >> 2) & 1) << 5; // pn[5] = y[2]
+			break;
+		case 2:
+			pixel_number |= ((x >> 0) & 1) << 0; // pn[0] = x[0]
+			pixel_number |= ((x >> 1) & 1) << 1; // pn[1] = x[1]
+			pixel_number |= ((x >> 2) & 1) << 2; // pn[2] = x[2]
+			pixel_number |= ((y >> 0) & 1) << 3; // pn[3] = y[0]
+			pixel_number |= ((y >> 1) & 1) << 4; // pn[4] = y[1]
+			pixel_number |= ((y >> 2) & 1) << 5; // pn[5] = y[2]
+			break;
+		case 4:
+			pixel_number |= ((x >> 0) & 1) << 0; // pn[0] = x[0]
+			pixel_number |= ((x >> 1) & 1) << 1; // pn[1] = x[1]
+			pixel_number |= ((y >> 0) & 1) << 2; // pn[2] = y[0]
+			pixel_number |= ((x >> 2) & 1) << 3; // pn[3] = x[2]
+			pixel_number |= ((y >> 1) & 1) << 4; // pn[4] = y[1]
+			pixel_number |= ((y >> 2) & 1) << 5; // pn[5] = y[2]
+			break;
+		}
+		sample_offset = sample_number * (tile_bytes / num_samples);
+		element_offset = sample_offset + (pixel_number * element_bytes);
+	}
+	total_offset = (slice_offset + macro_tile_offset) >> (channel_bits + bank_bits);
+	total_offset += element_offset;
+
+	offset_low = total_offset & group_mask;
+	offset_high = (total_offset & ~group_mask) << (channel_bits + bank_bits);
+	offset = (bank << (group_bits + channel_bits)) + (channel << group_bits) + offset_low + offset_high;
+
+	return offset;
+}
+
 /* depth buffers */
 static GLubyte *r600_ptr_depth(const struct radeon_renderbuffer * rrb,
 			       GLint x, GLint y)
 {
     GLubyte *ptr = rrb->bo->ptr;
-    GLint offset = r600_1d_tile_helper(rrb, x, y, 1, 0);
+    GLint offset;
+    if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE)
+	    offset = r600_2d_tile_helper(rrb, x, y, 1, 0);
+    else
+	    offset = r600_1d_tile_helper(rrb, x, y, 1, 0);
     return &ptr[offset];
 }
 
@@ -221,7 +398,11 @@ static GLubyte *r600_ptr_stencil(const struct radeon_renderbuffer * rrb,
 				 GLint x, GLint y)
 {
     GLubyte *ptr = rrb->bo->ptr;
-    GLint offset = r600_1d_tile_helper(rrb, x, y, 1, 1);
+    GLint offset;
+    if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE)
+	    offset = r600_2d_tile_helper(rrb, x, y, 1, 1);
+    else
+	    offset = r600_1d_tile_helper(rrb, x, y, 1, 1);
     return &ptr[offset];
 }
 
@@ -235,7 +416,10 @@ static GLubyte *r600_ptr_color(const struct radeon_renderbuffer * rrb,
     if (rrb->has_surface || !(rrb->bo->flags & mask)) {
         offset = x * rrb->cpp + y * rrb->pitch;
     } else {
-	    offset = r600_1d_tile_helper(rrb, x, y, 0, 0);
+	    if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE)
+		    offset = r600_2d_tile_helper(rrb, x, y, 0, 0);
+	    else
+		    offset = r600_1d_tile_helper(rrb, x, y, 0, 0);
     }
     return &ptr[offset];
 }
diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c
index f2fcb46688a..29defe73a70 100644
--- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c
+++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c
@@ -40,7 +40,10 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/macros.h"
 #include "main/simple_list.h"
 
+#include "math/m_xform.h"
+
 #include "swrast_setup/swrast_setup.h"
+
 #include "tnl/tnl.h"
 #include "tnl/t_context.h"
 #include "tnl/t_pipeline.h"
@@ -408,6 +411,8 @@ static GLboolean radeon_run_render( GLcontext *ctx,
        !radeon_dma_validate_render( ctx, VB ))
       return GL_TRUE;		
 
+   radeon_prepare_render(&rmesa->radeon);
+
    tnl->Driver.Render.Start( ctx );
 
    for (i = 0 ; i < VB->PrimitiveCount ; i++)
diff --git a/src/mesa/drivers/dri/radeon/radeon_tcl.c b/src/mesa/drivers/dri/radeon/radeon_tcl.c
index ea796e1a45f..5e1718f9dfc 100644
--- a/src/mesa/drivers/dri/radeon/radeon_tcl.c
+++ b/src/mesa/drivers/dri/radeon/radeon_tcl.c
@@ -252,6 +252,8 @@ void radeonTclPrimitive( GLcontext *ctx,
    GLuint se_cntl;
    GLuint newprim = hw_prim | RADEON_CP_VC_CNTL_TCL_ENABLE;
 
+   radeon_prepare_render(&rmesa->radeon);
+
    if (newprim != rmesa->tcl.hw_primitive ||
        !discrete_prim[hw_prim&0xf]) {
       RADEON_NEWPRIM( rmesa );
diff --git a/src/mesa/drivers/dri/radeon/radeon_tex_copy.c b/src/mesa/drivers/dri/radeon/radeon_tex_copy.c
index 29fd31ac23f..4cb0bb60c85 100644
--- a/src/mesa/drivers/dri/radeon/radeon_tex_copy.c
+++ b/src/mesa/drivers/dri/radeon/radeon_tex_copy.c
@@ -153,6 +153,9 @@ radeonCopyTexImage2D(GLcontext *ctx, GLenum target, GLint level,
         _mesa_select_tex_image(ctx, texObj, target, level);
     int srcx, srcy, dstx, dsty;
 
+    radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+    radeon_prepare_render(radeon);
+
     if (border)
         goto fail;
 
@@ -202,6 +205,9 @@ radeonCopyTexSubImage2D(GLcontext *ctx, GLenum target, GLint level,
     struct gl_texture_object *texObj = _mesa_select_tex_object(ctx, texUnit, target);
     struct gl_texture_image *texImage = _mesa_select_tex_image(ctx, texObj, target, level);
 
+    radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+    radeon_prepare_render(radeon);
+
     if (!do_copy_texsubimage(ctx, target, level,
                              radeon_tex_obj(texObj), (radeon_texture_image *)texImage,
                              xoffset, yoffset, x, y, width, height)) {
diff --git a/src/mesa/drivers/dri/radeon/radeon_texture.c b/src/mesa/drivers/dri/radeon/radeon_texture.c
index d2b190e42e0..8c6a50d2f0d 100644
--- a/src/mesa/drivers/dri/radeon/radeon_texture.c
+++ b/src/mesa/drivers/dri/radeon/radeon_texture.c
@@ -551,7 +551,7 @@ gl_format radeonChooseTextureFormat(GLcontext * ctx,
 	case GL_SRGB8_ALPHA8:
 	case GL_COMPRESSED_SRGB:
 	case GL_COMPRESSED_SRGB_ALPHA:
-		return MESA_FORMAT_SRGBA8;
+		return MESA_FORMAT_SARGB8;
 
 	case GL_SLUMINANCE:
 	case GL_SLUMINANCE8:
diff --git a/src/mesa/drivers/dri/savage/savagerender.c b/src/mesa/drivers/dri/savage/savagerender.c
index c369bb124c2..2d9e80e29c4 100644
--- a/src/mesa/drivers/dri/savage/savagerender.c
+++ b/src/mesa/drivers/dri/savage/savagerender.c
@@ -33,6 +33,8 @@
 #include "main/imports.h"
 #include "main/mtypes.h"
 
+#include "math/m_xform.h"
+
 #include "tnl/t_context.h"
 
 #include "savagecontext.h"
diff --git a/src/mesa/drivers/dri/unichrome/via_render.c b/src/mesa/drivers/dri/unichrome/via_render.c
index 896c43db1b0..4351f119555 100644
--- a/src/mesa/drivers/dri/unichrome/via_render.c
+++ b/src/mesa/drivers/dri/unichrome/via_render.c
@@ -33,6 +33,8 @@
 #include "main/macros.h"
 #include "main/mtypes.h"
 
+#include "math/m_xform.h"
+
 #include "tnl/t_context.h"
 
 #include "via_context.h"
diff --git a/src/mesa/main/arbprogram.h b/src/mesa/main/arbprogram.h
index 787ffd62f4b..e2e535e911e 100644
--- a/src/mesa/main/arbprogram.h
+++ b/src/mesa/main/arbprogram.h
@@ -27,7 +27,6 @@
 #define ARBPROGRAM_H
 
 
-#include "compiler.h"
 #include "glheader.h"
 
 
diff --git a/src/mesa/main/arrayobj.h b/src/mesa/main/arrayobj.h
index 8999edc724f..fdf7e2bca46 100644
--- a/src/mesa/main/arrayobj.h
+++ b/src/mesa/main/arrayobj.h
@@ -28,7 +28,7 @@
 #ifndef ARRAYOBJ_H
 #define ARRAYOBJ_H
 
-#include "context.h"
+#include "mtypes.h"
 
 /**
  * \file arrayobj.h
diff --git a/src/mesa/main/attrib.c b/src/mesa/main/attrib.c
index 002448fedb8..753949be503 100644
--- a/src/mesa/main/attrib.c
+++ b/src/mesa/main/attrib.c
@@ -41,6 +41,7 @@
 #include "hint.h"
 #include "light.h"
 #include "lines.h"
+#include "macros.h"
 #include "matrix.h"
 #include "multisample.h"
 #include "points.h"
diff --git a/src/mesa/main/bufferobj.h b/src/mesa/main/bufferobj.h
index 912529cfdf9..f234d06c6cc 100644
--- a/src/mesa/main/bufferobj.h
+++ b/src/mesa/main/bufferobj.h
@@ -29,7 +29,7 @@
 #define BUFFEROBJ_H
 
 
-#include "context.h"
+#include "mtypes.h"
 
 
 /*
diff --git a/src/mesa/main/clear.c b/src/mesa/main/clear.c
index e76ab5527b0..49d86b3b1f1 100644
--- a/src/mesa/main/clear.c
+++ b/src/mesa/main/clear.c
@@ -35,6 +35,7 @@
 #include "context.h"
 #include "colormac.h"
 #include "enums.h"
+#include "macros.h"
 #include "state.h"
 
 
diff --git a/src/mesa/main/clear.h b/src/mesa/main/clear.h
index 4c78eeda488..6657370c4b6 100644
--- a/src/mesa/main/clear.h
+++ b/src/mesa/main/clear.h
@@ -27,7 +27,7 @@
 #define CLEAR_H
 
 
-#include "main/mtypes.h"
+#include "glheader.h"
 
 
 extern void GLAPIENTRY
diff --git a/src/mesa/main/clip.h b/src/mesa/main/clip.h
index d53afb45bdf..ac472d66e08 100644
--- a/src/mesa/main/clip.h
+++ b/src/mesa/main/clip.h
@@ -31,7 +31,7 @@
 #ifndef CLIP_H
 #define CLIP_H
 
-#include "mtypes.h"
+#include "glheader.h"
 
 extern void GLAPIENTRY 
 _mesa_ClipPlane( GLenum plane, const GLdouble *equation );
diff --git a/src/mesa/main/colormac.h b/src/mesa/main/colormac.h
index 905f4e22837..245fb658bb3 100644
--- a/src/mesa/main/colormac.h
+++ b/src/mesa/main/colormac.h
@@ -33,9 +33,9 @@
 #define COLORMAC_H
 
 
-#include "imports.h"
 #include "config.h"
 #include "macros.h"
+#include "mtypes.h"
 
 
 /** \def BYTE_TO_CHAN
diff --git a/src/mesa/main/config.h b/src/mesa/main/config.h
index 32f7d969d8d..0f2d1a8f8da 100644
--- a/src/mesa/main/config.h
+++ b/src/mesa/main/config.h
@@ -177,7 +177,7 @@
 /**
  * Per-program constants (power of two)
  *
- * \c MAX_PROGRAM_LOCAL_PARAMS and \c MAX_UNIFORMS are just the assmebly shader
+ * \c MAX_PROGRAM_LOCAL_PARAMS and \c MAX_UNIFORMS are just the assembly shader
  * and GLSL shader names for the same thing.  They should \b always have the
  * same value.  Each refers to the number of vec4 values supplied as
  * per-program parameters.
diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c
index a369532e99c..b01fed1781e 100644
--- a/src/mesa/main/context.c
+++ b/src/mesa/main/context.c
@@ -462,7 +462,7 @@ _mesa_init_current(GLcontext *ctx)
 
 
 /**
- * Init vertex/fragment program limits.
+ * Init vertex/fragment/geometry program limits.
  * Important: drivers should override these with actual limits.
  */
 static void
@@ -477,16 +477,18 @@ init_program_limits(GLenum type, struct gl_program_constants *prog)
    prog->MaxLocalParams = MAX_PROGRAM_LOCAL_PARAMS;
    prog->MaxUniformComponents = 4 * MAX_UNIFORMS;
 
-   if (type == GL_VERTEX_PROGRAM_ARB) {
+   switch (type) {
+   case GL_VERTEX_PROGRAM_ARB:
       prog->MaxParameters = MAX_VERTEX_PROGRAM_PARAMS;
       prog->MaxAttribs = MAX_NV_VERTEX_PROGRAM_INPUTS;
       prog->MaxAddressRegs = MAX_VERTEX_PROGRAM_ADDRESS_REGS;
-   }
-   else if (type == GL_FRAGMENT_PROGRAM_ARB) {
+      break;
+   case GL_FRAGMENT_PROGRAM_ARB:
       prog->MaxParameters = MAX_NV_FRAGMENT_PROGRAM_PARAMS;
       prog->MaxAttribs = MAX_NV_FRAGMENT_PROGRAM_INPUTS;
       prog->MaxAddressRegs = MAX_FRAGMENT_PROGRAM_ADDRESS_REGS;
-   } else {
+      break;
+   case MESA_GEOMETRY_PROGRAM:
       prog->MaxParameters = MAX_NV_VERTEX_PROGRAM_PARAMS;
       prog->MaxAttribs = MAX_NV_VERTEX_PROGRAM_INPUTS;
       prog->MaxAddressRegs = MAX_VERTEX_PROGRAM_ADDRESS_REGS;
@@ -497,6 +499,9 @@ init_program_limits(GLenum type, struct gl_program_constants *prog)
       prog->MaxGeometryUniformComponents = MAX_GEOMETRY_UNIFORM_COMPONENTS;
       prog->MaxGeometryOutputVertices = MAX_GEOMETRY_OUTPUT_VERTICES;
       prog->MaxGeometryTotalOutputComponents = MAX_GEOMETRY_TOTAL_OUTPUT_COMPONENTS;
+      break;
+   default:
+      assert(0 && "Bad program type in init_program_limits()");
    }
 
    /* Set the native limits to zero.  This implies that there is no native
diff --git a/src/mesa/main/convolve.c b/src/mesa/main/convolve.c
index 15e8dffc230..f63bddc44d5 100644
--- a/src/mesa/main/convolve.c
+++ b/src/mesa/main/convolve.c
@@ -37,6 +37,7 @@
 #include "convolve.h"
 #include "context.h"
 #include "image.h"
+#include "macros.h"
 #include "mtypes.h"
 #include "state.h"
 #include "main/dispatch.h"
diff --git a/src/mesa/main/debug.h b/src/mesa/main/debug.h
index 0449cb1798a..b517cc8259f 100644
--- a/src/mesa/main/debug.h
+++ b/src/mesa/main/debug.h
@@ -36,6 +36,9 @@
 #ifndef _DEBUG_H
 #define _DEBUG_H
 
+#include "glheader.h"
+#include "mtypes.h"
+
 #if _HAVE_FULL_GL
 
 extern void _mesa_print_tri_caps( const char *name, GLuint flags );
diff --git a/src/mesa/main/depthstencil.h b/src/mesa/main/depthstencil.h
index 3dde081f5a5..afbac77f0e2 100644
--- a/src/mesa/main/depthstencil.h
+++ b/src/mesa/main/depthstencil.h
@@ -26,6 +26,7 @@
 #ifndef DEPTHSTENCIL_H
 #define DEPTHSTENCIL_H
 
+#include "mtypes.h"
 
 extern struct gl_renderbuffer *
 _mesa_new_z24_renderbuffer_wrapper(GLcontext *ctx,
diff --git a/src/mesa/main/fbobject.c b/src/mesa/main/fbobject.c
index 8c86b392c7b..9a84e5a79cf 100644
--- a/src/mesa/main/fbobject.c
+++ b/src/mesa/main/fbobject.c
@@ -97,6 +97,11 @@ _mesa_init_fbobjects(GLcontext *ctx)
    DummyRenderbuffer.Delete = delete_dummy_renderbuffer;
 }
 
+struct gl_framebuffer *
+_mesa_get_incomplete_framebuffer(void)
+{
+   return &DummyFramebuffer;
+}
 
 /**
  * Helper routine for getting a gl_renderbuffer.
diff --git a/src/mesa/main/fbobject.h b/src/mesa/main/fbobject.h
index ff946033a4d..9850ee9aa23 100644
--- a/src/mesa/main/fbobject.h
+++ b/src/mesa/main/fbobject.h
@@ -26,10 +26,14 @@
 #ifndef FBOBJECT_H
 #define FBOBJECT_H
 
+#include "mtypes.h"
 
 extern void
 _mesa_init_fbobjects(GLcontext *ctx);
 
+extern struct gl_framebuffer *
+_mesa_get_incomplete_framebuffer(void);
+
 extern struct gl_renderbuffer *
 _mesa_lookup_renderbuffer(GLcontext *ctx, GLuint id);
 
diff --git a/src/mesa/main/fog.c b/src/mesa/main/fog.c
index 269ff3f8b99..9f26c012d66 100644
--- a/src/mesa/main/fog.c
+++ b/src/mesa/main/fog.c
@@ -27,6 +27,7 @@
 #include "colormac.h"
 #include "context.h"
 #include "fog.h"
+#include "macros.h"
 #include "mtypes.h"
 
 
diff --git a/src/mesa/main/formats.c b/src/mesa/main/formats.c
index 49463fcc3c2..90449cc04f0 100644
--- a/src/mesa/main/formats.c
+++ b/src/mesa/main/formats.c
@@ -26,7 +26,6 @@
 
 #include "imports.h"
 #include "formats.h"
-#include "config.h"
 
 
 /**
diff --git a/src/mesa/main/formats.h b/src/mesa/main/formats.h
index aa14185628f..ad176caaa0f 100644
--- a/src/mesa/main/formats.h
+++ b/src/mesa/main/formats.h
@@ -33,7 +33,7 @@
 #define FORMATS_H
 
 
-#include "main/mtypes.h"
+#include <GL/gl.h>
 
 
 
diff --git a/src/mesa/main/framebuffer.c b/src/mesa/main/framebuffer.c
index 56558cfcc1e..e0aac26f62b 100644
--- a/src/mesa/main/framebuffer.c
+++ b/src/mesa/main/framebuffer.c
@@ -75,7 +75,6 @@ compute_depth_max(struct gl_framebuffer *fb)
    fb->_MRD = (GLfloat)1.0 / fb->_DepthMaxF;
 }
 
-
 /**
  * Create and initialize a gl_framebuffer object.
  * This is intended for creating _window_system_ framebuffers, not generic
diff --git a/src/mesa/main/framebuffer.h b/src/mesa/main/framebuffer.h
index 1b6e3b1f0cb..2e9844282f8 100644
--- a/src/mesa/main/framebuffer.h
+++ b/src/mesa/main/framebuffer.h
@@ -26,6 +26,7 @@
 #ifndef FRAMEBUFFER_H
 #define FRAMEBUFFER_H
 
+#include "mtypes.h"
 
 extern struct gl_framebuffer *
 _mesa_create_framebuffer(const GLvisual *visual);
diff --git a/src/mesa/main/get.h b/src/mesa/main/get.h
index 320492b4ce2..99a004b71dd 100644
--- a/src/mesa/main/get.h
+++ b/src/mesa/main/get.h
@@ -32,7 +32,7 @@
 #define GET_H
 
 
-#include "mtypes.h"
+#include "glheader.h"
 
 
 extern void GLAPIENTRY
diff --git a/src/mesa/main/histogram.c b/src/mesa/main/histogram.c
index 3a65bb19260..4e482bcd54b 100644
--- a/src/mesa/main/histogram.c
+++ b/src/mesa/main/histogram.c
@@ -29,6 +29,7 @@
 #include "context.h"
 #include "image.h"
 #include "histogram.h"
+#include "macros.h"
 #include "main/dispatch.h"
 
 
diff --git a/src/mesa/main/image.c b/src/mesa/main/image.c
index 63c28342f26..86aa6d0d702 100644
--- a/src/mesa/main/image.c
+++ b/src/mesa/main/image.c
@@ -32,7 +32,6 @@
 
 #include "glheader.h"
 #include "colormac.h"
-#include "context.h"
 #include "enums.h"
 #include "image.h"
 #include "imports.h"
diff --git a/src/mesa/main/imports.c b/src/mesa/main/imports.c
index 25080db40c4..46e5c932d0f 100644
--- a/src/mesa/main/imports.c
+++ b/src/mesa/main/imports.c
@@ -756,7 +756,7 @@ _mesa_strdup( const char *s )
 float
 _mesa_strtof( const char *s, char **end )
 {
-#ifdef _GNU_SOURCE
+#if defined(_GNU_SOURCE) && !defined(__CYGWIN__) && !defined(__FreeBSD__)
    static locale_t loc = NULL;
    if (!loc) {
       loc = newlocale(LC_CTYPE_MASK, "C", NULL);
diff --git a/src/mesa/main/mm.c b/src/mesa/main/mm.c
index 3ef38e94be9..25a0293703c 100644
--- a/src/mesa/main/mm.c
+++ b/src/mesa/main/mm.c
@@ -22,6 +22,11 @@
  *
  */
 
+#include <assert.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+#include "compiler.h"
 #include "mm.h"
 
 
diff --git a/src/mesa/main/mm.h b/src/mesa/main/mm.h
index df340808ac9..228721ca2a5 100644
--- a/src/mesa/main/mm.h
+++ b/src/mesa/main/mm.h
@@ -32,9 +32,6 @@
 #define MM_H
 
 
-#include "imports.h"
-
-
 struct mem_block {
    struct mem_block *next, *prev;
    struct mem_block *next_free, *prev_free;
diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h
index cbb9eb84f33..8d92892ad7d 100644
--- a/src/mesa/main/mtypes.h
+++ b/src/mesa/main/mtypes.h
@@ -36,7 +36,6 @@
 
 #include "main/glheader.h"
 #include "main/config.h"
-#include "main/compiler.h"
 #include "main/mfeatures.h"
 #include "glapi/glapi.h"
 #include "math/m_matrix.h"	/* GLmatrix */
@@ -1808,6 +1807,11 @@ struct gl_program
    /** Which texture target is being sampled (TEXTURE_1D/2D/3D/etc_INDEX) */
    gl_texture_index SamplerTargets[MAX_SAMPLERS];
 
+   /** Bitmask of which register files are read/written with indirect
+    * addressing.  Mask of (1 << PROGRAM_x) bits.
+    */
+   GLbitfield IndirectRegisterFiles;
+
    /** Logical counts */
    /*@{*/
    GLuint NumInstructions;
@@ -2477,29 +2481,29 @@ struct gl_framebuffer
 
 
 /**
- * Limits for vertex and fragment programs.
+ * Limits for vertex and fragment programs/shaders.
  */
 struct gl_program_constants
 {
    /* logical limits */
    GLuint MaxInstructions;
-   GLuint MaxAluInstructions; /* fragment programs only, for now */
-   GLuint MaxTexInstructions; /* fragment programs only, for now */
-   GLuint MaxTexIndirections; /* fragment programs only, for now */
+   GLuint MaxAluInstructions;
+   GLuint MaxTexInstructions;
+   GLuint MaxTexIndirections;
    GLuint MaxAttribs;
    GLuint MaxTemps;
-   GLuint MaxAddressRegs; /* vertex program only, for now */
+   GLuint MaxAddressRegs;
    GLuint MaxParameters;
    GLuint MaxLocalParams;
    GLuint MaxEnvParams;
    /* native/hardware limits */
    GLuint MaxNativeInstructions;
-   GLuint MaxNativeAluInstructions; /* fragment programs only, for now */
-   GLuint MaxNativeTexInstructions; /* fragment programs only, for now */
-   GLuint MaxNativeTexIndirections; /* fragment programs only, for now */
+   GLuint MaxNativeAluInstructions;
+   GLuint MaxNativeTexInstructions;
+   GLuint MaxNativeTexIndirections;
    GLuint MaxNativeAttribs;
    GLuint MaxNativeTemps;
-   GLuint MaxNativeAddressRegs; /* vertex program only, for now */
+   GLuint MaxNativeAddressRegs;
    GLuint MaxNativeParameters;
    /* For shaders */
    GLuint MaxUniformComponents;
diff --git a/src/mesa/main/multisample.h b/src/mesa/main/multisample.h
index 4305900cc49..998488ef420 100644
--- a/src/mesa/main/multisample.h
+++ b/src/mesa/main/multisample.h
@@ -26,6 +26,7 @@
 #ifndef MULTISAMPLE_H
 #define MULTISAMPLE_H
 
+#include "mtypes.h"
 
 extern void GLAPIENTRY
 _mesa_SampleCoverageARB(GLclampf value, GLboolean invert);
diff --git a/src/mesa/main/nvprogram.h b/src/mesa/main/nvprogram.h
index 8ee59661bd0..260a25ba9e9 100644
--- a/src/mesa/main/nvprogram.h
+++ b/src/mesa/main/nvprogram.h
@@ -29,6 +29,8 @@
 #ifndef NVPROGRAM_H
 #define NVPROGRAM_H
 
+#include "glheader.h"
+#include "mtypes.h"
 
 extern void GLAPIENTRY
 _mesa_ExecuteProgramNV(GLenum target, GLuint id, const GLfloat *params);
diff --git a/src/mesa/main/pixelstore.h b/src/mesa/main/pixelstore.h
index ee963f9ba3c..47bff4276d1 100644
--- a/src/mesa/main/pixelstore.h
+++ b/src/mesa/main/pixelstore.h
@@ -33,6 +33,7 @@
 
 
 #include "glheader.h"
+#include "mtypes.h"
 
 
 extern void GLAPIENTRY
diff --git a/src/mesa/main/querymatrix.c b/src/mesa/main/querymatrix.c
index 6f62415ba8c..32aaa79f7fb 100644
--- a/src/mesa/main/querymatrix.c
+++ b/src/mesa/main/querymatrix.c
@@ -36,9 +36,9 @@
 #define INT_TO_FIXED(x) ((GLfixed) ((x) << 16))
 #define FLOAT_TO_FIXED(x) ((GLfixed) ((x) * 65536.0))
 
-#if defined(WIN32) || defined(_WIN32_WCE)
+#if defined(_MSC_VER)
 /* Oddly, the fpclassify() function doesn't exist in such a form
- * on Windows.  This is an implementation using slightly different
+ * on MSVC.  This is an implementation using slightly different
  * lower-level Windows functions.
  */
 #include <float.h>
@@ -72,7 +72,7 @@ fpclassify(double x)
 
 #elif defined(__APPLE__) || defined(__CYGWIN__) || defined(__FreeBSD__) || \
      defined(__OpenBSD__) || defined(__NetBSD__) || defined(__DragonFly__) || \
-     (defined(__sun) && defined(__C99FEATURES__))
+     (defined(__sun) && defined(__C99FEATURES__)) || defined(__MINGW32__)
 
 /* fpclassify is available. */
 
diff --git a/src/mesa/main/remap.h b/src/mesa/main/remap.h
index 7afdee36f5b..a2a55f615d5 100644
--- a/src/mesa/main/remap.h
+++ b/src/mesa/main/remap.h
@@ -28,6 +28,7 @@
 #define REMAP_H
 
 
+#include "main/compiler.h"
 #include "main/mfeatures.h"
 
 struct gl_function_pool_remap {
diff --git a/src/mesa/main/renderbuffer.h b/src/mesa/main/renderbuffer.h
index 7c205e141c1..bc92b269821 100644
--- a/src/mesa/main/renderbuffer.h
+++ b/src/mesa/main/renderbuffer.h
@@ -26,6 +26,11 @@
 #ifndef RENDERBUFFER_H
 #define RENDERBUFFER_H
 
+#include "glheader.h"
+#include "mtypes.h"
+
+struct gl_framebuffer;
+struct gl_renderbuffer;
 
 extern void
 _mesa_init_renderbuffer(struct gl_renderbuffer *rb, GLuint name);
diff --git a/src/mesa/main/restart.h b/src/mesa/main/restart.h
index 931cd701281..25f58f24c35 100644
--- a/src/mesa/main/restart.h
+++ b/src/mesa/main/restart.h
@@ -28,6 +28,7 @@
 #ifndef RESTART_H
 #define RESTART_H
 
+#include "glheader.h"
 
 extern void GLAPIENTRY
 _mesa_PrimitiveRestart(void);
diff --git a/src/mesa/main/shared.c b/src/mesa/main/shared.c
index f9d10f3bbea..cbe004518a0 100644
--- a/src/mesa/main/shared.c
+++ b/src/mesa/main/shared.c
@@ -32,7 +32,6 @@
 #include "imports.h"
 #include "mtypes.h"
 #include "hash.h"
-#include "arrayobj.h"
 #if FEATURE_ATI_fragment_shader
 #include "atifragshader.h"
 #endif
diff --git a/src/mesa/main/shared.h b/src/mesa/main/shared.h
index ef164a14590..5166a0ce51f 100644
--- a/src/mesa/main/shared.h
+++ b/src/mesa/main/shared.h
@@ -25,6 +25,7 @@
 #ifndef SHARED_H
 #define SHARED_H
 
+#include "mtypes.h"
 
 struct gl_shared_state *
 _mesa_alloc_shared_state(GLcontext *ctx);
diff --git a/src/mesa/main/syncobj.h b/src/mesa/main/syncobj.h
index f23fa281e20..c53511995b1 100644
--- a/src/mesa/main/syncobj.h
+++ b/src/mesa/main/syncobj.h
@@ -31,7 +31,10 @@
 #ifndef SYNCOBJ_H
 #define SYNCOBJ_H
 
-#include "context.h"
+#include "glheader.h"
+#include "mtypes.h"
+
+struct dd_function_table;
 
 extern void
 _mesa_init_sync_object_functions(struct dd_function_table *driver);
diff --git a/src/mesa/main/texcompress.c b/src/mesa/main/texcompress.c
index f4b1119eb17..e911524cbc5 100644
--- a/src/mesa/main/texcompress.c
+++ b/src/mesa/main/texcompress.c
@@ -33,7 +33,6 @@
 #include "glheader.h"
 #include "imports.h"
 #include "colormac.h"
-#include "context.h"
 #include "formats.h"
 #include "texcompress.h"
 
diff --git a/src/mesa/main/texcompress_fxt1.c b/src/mesa/main/texcompress_fxt1.c
index 04acf05e528..c8b45bd3a55 100644
--- a/src/mesa/main/texcompress_fxt1.c
+++ b/src/mesa/main/texcompress_fxt1.c
@@ -32,9 +32,9 @@
 #include "glheader.h"
 #include "imports.h"
 #include "colormac.h"
-#include "context.h"
 #include "convolve.h"
 #include "image.h"
+#include "macros.h"
 #include "mipmap.h"
 #include "texcompress.h"
 #include "texcompress_fxt1.h"
diff --git a/src/mesa/main/texcompress_fxt1.h b/src/mesa/main/texcompress_fxt1.h
index d63ca71e212..38048b26ccb 100644
--- a/src/mesa/main/texcompress_fxt1.h
+++ b/src/mesa/main/texcompress_fxt1.h
@@ -25,9 +25,11 @@
 #ifndef TEXCOMPRESS_FXT1_H
 #define TEXCOMPRESS_FXT1_H
 
-#include "main/mtypes.h"
+#include "glheader.h"
 #include "texstore.h"
 
+struct gl_texture_image;
+
 #if FEATURE_texture_fxt1
 
 extern GLboolean
diff --git a/src/mesa/main/texcompress_s3tc.c b/src/mesa/main/texcompress_s3tc.c
index 85c394b051f..c70792cab61 100644
--- a/src/mesa/main/texcompress_s3tc.c
+++ b/src/mesa/main/texcompress_s3tc.c
@@ -36,10 +36,10 @@
 #include "glheader.h"
 #include "imports.h"
 #include "colormac.h"
-#include "context.h"
 #include "convolve.h"
 #include "dlopen.h"
 #include "image.h"
+#include "macros.h"
 #include "texcompress.h"
 #include "texcompress_s3tc.h"
 #include "texstore.h"
diff --git a/src/mesa/main/texfetch.c b/src/mesa/main/texfetch.c
index fe002082cca..c03bc71cd7a 100644
--- a/src/mesa/main/texfetch.c
+++ b/src/mesa/main/texfetch.c
@@ -34,7 +34,7 @@
 
 
 #include "colormac.h"
-#include "context.h"
+#include "macros.h"
 #include "texcompress.h"
 #include "texcompress_fxt1.h"
 #include "texcompress_s3tc.h"
diff --git a/src/mesa/main/texgen.h b/src/mesa/main/texgen.h
index 397d89e630f..2224a937611 100644
--- a/src/mesa/main/texgen.h
+++ b/src/mesa/main/texgen.h
@@ -27,7 +27,10 @@
 #define TEXGEN_H
 
 
-#include "main/mtypes.h"
+#include "compiler.h"
+#include "glheader.h"
+
+struct _glapi_table;
 
 
 #if FEATURE_texgen
diff --git a/src/mesa/main/texgetimage.h b/src/mesa/main/texgetimage.h
index 088d27c7e17..866ab704945 100644
--- a/src/mesa/main/texgetimage.h
+++ b/src/mesa/main/texgetimage.h
@@ -27,6 +27,7 @@
 #ifndef TEXGETIMAGE_H
 #define TEXGETIMAGE_H
 
+#include "mtypes.h"
 
 extern void
 _mesa_get_teximage(GLcontext *ctx, GLenum target, GLint level,
diff --git a/src/mesa/main/texrender.c b/src/mesa/main/texrender.c
index d29af5a5b2f..c68105b3951 100644
--- a/src/mesa/main/texrender.c
+++ b/src/mesa/main/texrender.c
@@ -1,6 +1,7 @@
 
 #include "context.h"
 #include "colormac.h"
+#include "macros.h"
 #include "texfetch.h"
 #include "texrender.h"
 #include "renderbuffer.h"
diff --git a/src/mesa/main/texrender.h b/src/mesa/main/texrender.h
index 7c3fb0871bd..1e87d594a28 100644
--- a/src/mesa/main/texrender.h
+++ b/src/mesa/main/texrender.h
@@ -1,6 +1,7 @@
 #ifndef TEXRENDER_H
 #define TEXRENDER_H
 
+#include "mtypes.h"
 
 extern void
 _mesa_render_texture(GLcontext *ctx,
diff --git a/src/mesa/main/texstate.h b/src/mesa/main/texstate.h
index 17ac68000c5..912cb677985 100644
--- a/src/mesa/main/texstate.h
+++ b/src/mesa/main/texstate.h
@@ -32,6 +32,7 @@
 #define TEXSTATE_H
 
 
+#include "compiler.h"
 #include "mtypes.h"
 
 
diff --git a/src/mesa/main/texstore.c b/src/mesa/main/texstore.c
index 0f21395af39..2989fdb72ed 100644
--- a/src/mesa/main/texstore.c
+++ b/src/mesa/main/texstore.c
@@ -55,7 +55,6 @@
 #include "glheader.h"
 #include "bufferobj.h"
 #include "colormac.h"
-#include "context.h"
 #include "convolve.h"
 #include "image.h"
 #include "macros.h"
diff --git a/src/mesa/main/uniforms.h b/src/mesa/main/uniforms.h
index 29f77cb35a0..ef98fe16bb1 100644
--- a/src/mesa/main/uniforms.h
+++ b/src/mesa/main/uniforms.h
@@ -25,6 +25,10 @@
 #ifndef UNIFORMS_H
 #define UNIFORMS_H
 
+#include "glheader.h"
+
+struct gl_program;
+struct _glapi_table;
 
 extern void GLAPIENTRY
 _mesa_Uniform1fARB(GLint, GLfloat);
diff --git a/src/mesa/main/viewport.h b/src/mesa/main/viewport.h
index f08fef27978..ec054a7c597 100644
--- a/src/mesa/main/viewport.h
+++ b/src/mesa/main/viewport.h
@@ -27,6 +27,8 @@
 #ifndef VIEWPORT_H
 #define VIEWPORT_H
 
+#include "glheader.h"
+#include "mtypes.h"
 
 extern void GLAPIENTRY
 _mesa_Viewport(GLint x, GLint y, GLsizei width, GLsizei height);
diff --git a/src/mesa/main/vtxfmt.h b/src/mesa/main/vtxfmt.h
index fb6c23abe98..aad38b87c35 100644
--- a/src/mesa/main/vtxfmt.h
+++ b/src/mesa/main/vtxfmt.h
@@ -33,6 +33,9 @@
 #ifndef _VTXFMT_H_
 #define _VTXFMT_H_
 
+#include "compiler.h"
+#include "mtypes.h"
+
 #if FEATURE_beginend
 
 extern void _mesa_init_exec_vtxfmt( GLcontext *ctx );
diff --git a/src/mesa/math/m_matrix.h b/src/mesa/math/m_matrix.h
index 3bc5de6cd4d..a69afb8589a 100644
--- a/src/mesa/math/m_matrix.h
+++ b/src/mesa/math/m_matrix.h
@@ -32,6 +32,8 @@
 #define _M_MATRIX_H
 
 
+#include "main/glheader.h"
+
 
 /**
  * \name Symbolic names to some of the entries in the matrix
diff --git a/src/mesa/math/m_translate.c b/src/mesa/math/m_translate.c
index b12b07957cb..51daf7bfd37 100644
--- a/src/mesa/math/m_translate.c
+++ b/src/mesa/math/m_translate.c
@@ -29,8 +29,8 @@
 
 
 #include "main/glheader.h"
+#include "main/macros.h"
 #include "main/mtypes.h"		/* GLchan hack */
-#include "main/colormac.h"
 
 #include "m_translate.h"
 
diff --git a/src/mesa/math/m_translate.h b/src/mesa/math/m_translate.h
index c677682d506..58041031163 100644
--- a/src/mesa/math/m_translate.h
+++ b/src/mesa/math/m_translate.h
@@ -26,7 +26,8 @@
 #ifndef _M_TRANSLATE_H_
 #define _M_TRANSLATE_H_
 
-#include "main/config.h"
+#include "main/compiler.h"
+#include "main/glheader.h"
 #include "main/mtypes.h"		/* hack for GLchan */
 
 
diff --git a/src/mesa/math/m_xform.h b/src/mesa/math/m_xform.h
index 33421ad1c0a..14ac956a7bc 100644
--- a/src/mesa/math/m_xform.h
+++ b/src/mesa/math/m_xform.h
@@ -27,10 +27,10 @@
 #define _M_XFORM_H
 
 
+#include "main/compiler.h"
 #include "main/glheader.h"
-#include "main/config.h"
-#include "math/m_vector.h"
 #include "math/m_matrix.h"
+#include "math/m_vector.h"
 
 #ifdef USE_X86_ASM
 #define _XFORMAPI _ASMAPI
diff --git a/src/mesa/program/arbprogparse.c b/src/mesa/program/arbprogparse.c
index 6373529e4e8..f834aaf5686 100644
--- a/src/mesa/program/arbprogparse.c
+++ b/src/mesa/program/arbprogparse.c
@@ -109,6 +109,7 @@ _mesa_parse_arb_fragment_program(GLcontext* ctx, GLenum target,
    program->Base.NumNativeTexIndirections = prog.NumTexIndirections;
    program->Base.InputsRead      = prog.InputsRead;
    program->Base.OutputsWritten  = prog.OutputsWritten;
+   program->Base.IndirectRegisterFiles = prog.IndirectRegisterFiles;
    for (i = 0; i < MAX_TEXTURE_IMAGE_UNITS; i++) {
       program->Base.TexturesUsed[i] = prog.TexturesUsed[i];
       if (prog.TexturesUsed[i])
@@ -199,6 +200,7 @@ _mesa_parse_arb_vertex_program(GLcontext *ctx, GLenum target,
    program->Base.NumNativeAddressRegs = prog.NumNativeAddressRegs;
    program->Base.InputsRead     = prog.InputsRead;
    program->Base.OutputsWritten = prog.OutputsWritten;
+   program->Base.IndirectRegisterFiles = prog.IndirectRegisterFiles;
    program->IsPositionInvariant = (state.option.PositionInvariant)
       ? GL_TRUE : GL_FALSE;
 
diff --git a/src/mesa/program/hash_table.h b/src/mesa/program/hash_table.h
index 7b302f5dbee..e750906f961 100644
--- a/src/mesa/program/hash_table.h
+++ b/src/mesa/program/hash_table.h
@@ -31,8 +31,6 @@
 #ifndef HASH_TABLE_H
 #define HASH_TABLE_H
 
-#include <string.h>
-
 struct hash_table;
 
 typedef unsigned (*hash_func_t)(const void *key);
diff --git a/src/mesa/program/nvfragparse.h b/src/mesa/program/nvfragparse.h
index 544ab80c56c..e28a6c49349 100644
--- a/src/mesa/program/nvfragparse.h
+++ b/src/mesa/program/nvfragparse.h
@@ -30,6 +30,7 @@
 #ifndef NVFRAGPARSE_H
 #define NVFRAGPARSE_H
 
+#include "main/mtypes.h"
 
 extern void
 _mesa_parse_nv_fragment_program(GLcontext *ctx, GLenum target,
diff --git a/src/mesa/program/nvvertparse.c b/src/mesa/program/nvvertparse.c
index e2afcfd4ce6..1ac83d0e59d 100644
--- a/src/mesa/program/nvvertparse.c
+++ b/src/mesa/program/nvvertparse.c
@@ -64,6 +64,7 @@ struct parse_state {
    GLbitfield inputsRead;
    GLbitfield outputsWritten;
    GLboolean anyProgRegsWritten;
+   GLboolean indirectRegisterFiles;
    GLuint numInst;                 /* number of instructions parsed */
 };
 
@@ -410,6 +411,7 @@ Parse_ParamReg(struct parse_state *parseState, struct prog_src_register *srcReg)
 
       srcReg->RelAddr = GL_TRUE;
       srcReg->File = PROGRAM_ENV_PARAM;
+      parseState->indirectRegisterFiles |= (1 << srcReg->File);
       /* Look for +/-N offset */
       if (!Peek_Token(parseState, token))
          RETURN_ERROR;
@@ -1308,6 +1310,7 @@ _mesa_parse_nv_vertex_program(GLcontext *ctx, GLenum dstTarget,
    parseState.inputsRead = 0;
    parseState.outputsWritten = 0;
    parseState.anyProgRegsWritten = GL_FALSE;
+   parseState.indirectRegisterFiles = 0x0;
 
    /* Reset error state */
    _mesa_set_program_error(ctx, -1, NULL);
@@ -1408,6 +1411,8 @@ _mesa_parse_nv_vertex_program(GLcontext *ctx, GLenum dstTarget,
       program->Base.Parameters = _mesa_new_parameter_list ();
       program->Base.NumParameters = 0;
 
+      program->Base.IndirectRegisterFiles = parseState.indirectRegisterFiles;
+
       state_tokens[0] = STATE_VERTEX_PROGRAM;
       state_tokens[1] = STATE_ENV;
       /* Add refs to all of the potential params, in order.  If we want to not
diff --git a/src/mesa/program/nvvertparse.h b/src/mesa/program/nvvertparse.h
index 9919e22388d..91ef79e6c3c 100644
--- a/src/mesa/program/nvvertparse.h
+++ b/src/mesa/program/nvvertparse.h
@@ -29,6 +29,7 @@
 #ifndef NVVERTPARSE_H
 #define NVVERTPARSE_H
 
+#include "main/mtypes.h"
 
 extern void
 _mesa_parse_nv_vertex_program(GLcontext *ctx, GLenum target,
diff --git a/src/mesa/program/prog_cache.h b/src/mesa/program/prog_cache.h
index 4e1ccac03ff..bfe8f99d445 100644
--- a/src/mesa/program/prog_cache.h
+++ b/src/mesa/program/prog_cache.h
@@ -30,6 +30,9 @@
 #define PROG_CACHE_H
 
 
+#include "main/mtypes.h"
+
+
 /** Opaque type */
 struct gl_program_cache;
 
diff --git a/src/mesa/program/prog_execute.c b/src/mesa/program/prog_execute.c
index f85c6513f31..1670c91b6ad 100644
--- a/src/mesa/program/prog_execute.c
+++ b/src/mesa/program/prog_execute.c
@@ -37,7 +37,7 @@
 
 #include "main/glheader.h"
 #include "main/colormac.h"
-#include "main/context.h"
+#include "main/macros.h"
 #include "prog_execute.h"
 #include "prog_instruction.h"
 #include "prog_parameter.h"
@@ -81,6 +81,22 @@ static const GLfloat ZeroVec[4] = { 0.0F, 0.0F, 0.0F, 0.0F };
 
 
 /**
+ * Return TRUE for +0 and other positive values, FALSE otherwise.
+ * Used for RCC opcode.
+ */
+static INLINE GLboolean
+positive(float x)
+{
+   fi_type fi;
+   fi.f = x;
+   if (fi.i & 0x80000000)
+      return GL_FALSE;
+   return GL_TRUE;
+}
+
+
+
+/**
  * Return a pointer to the 4-element float vector specified by the given
  * source register.
  */
@@ -1340,6 +1356,44 @@ _mesa_execute_program(GLcontext * ctx,
             store_vector4(inst, machine, result);
          }
          break;
+      case OPCODE_RCC:  /* clamped riciprocal */
+         {
+            const float largest = 1.884467e+19, smallest = 5.42101e-20;
+            GLfloat a[4], r, result[4];
+            fetch_vector1(&inst->SrcReg[0], machine, a);
+            if (DEBUG_PROG) {
+               if (a[0] == 0)
+                  printf("RCC(0)\n");
+               else if (IS_INF_OR_NAN(a[0]))
+                  printf("RCC(inf)\n");
+            }
+            if (a[0] == 1.0F) {
+               r = 1.0F;
+            }
+            else {
+               r = 1.0F / a[0];
+            }
+            if (positive(r)) {
+               if (r > largest) {
+                  r = largest;
+               }
+               else if (r < smallest) {
+                  r = smallest;
+               }
+            }
+            else {
+               if (r < -largest) {
+                  r = -largest;
+               }
+               else if (r > -smallest) {
+                  r = -smallest;
+               }
+            }
+            result[0] = result[1] = result[2] = result[3] = r;
+            store_vector4(inst, machine, result);
+         }
+         break;
+
       case OPCODE_RCP:
          {
             GLfloat a[4], result[4];
diff --git a/src/mesa/program/prog_execute.h b/src/mesa/program/prog_execute.h
index adefc5439de..f59b65176ff 100644
--- a/src/mesa/program/prog_execute.h
+++ b/src/mesa/program/prog_execute.h
@@ -26,6 +26,7 @@
 #define PROG_EXECUTE_H
 
 #include "main/config.h"
+#include "main/mtypes.h"
 
 
 typedef void (*FetchTexelLodFunc)(GLcontext *ctx, const GLfloat texcoord[4],
diff --git a/src/mesa/program/prog_instruction.h b/src/mesa/program/prog_instruction.h
index dacbc33704b..098b366ab56 100644
--- a/src/mesa/program/prog_instruction.h
+++ b/src/mesa/program/prog_instruction.h
@@ -38,7 +38,7 @@
 #define PROG_INSTRUCTION_H
 
 
-#include "main/mfeatures.h"
+#include "main/glheader.h"
 
 
 /**
@@ -149,20 +149,20 @@ typedef enum prog_opcode {
    OPCODE_ADD,       /*   X        X       X       X         X   */
    OPCODE_AND,       /*                                          */
    OPCODE_ARA,       /*                    2                     */
-   OPCODE_ARL,       /*   X                X                     */
+   OPCODE_ARL,       /*   X                X                 X   */
    OPCODE_ARL_NV,    /*                    2                     */
    OPCODE_ARR,       /*                    2                     */
    OPCODE_BGNLOOP,   /*                                     opt  */
    OPCODE_BGNSUB,    /*                                     opt  */
    OPCODE_BRA,       /*                    2                 X   */
    OPCODE_BRK,       /*                    2                opt  */
-   OPCODE_CAL,       /*                    2       2             */
-   OPCODE_CMP,       /*            X                             */
+   OPCODE_CAL,       /*                    2       2         X   */
+   OPCODE_CMP,       /*            X                         X   */
    OPCODE_CONT,      /*                                     opt  */
    OPCODE_COS,       /*            X       2       X         X   */
    OPCODE_DDX,       /*                            X         X   */
    OPCODE_DDY,       /*                            X         X   */
-   OPCODE_DP2,       /*                            2             */
+   OPCODE_DP2,       /*                            2         X   */
    OPCODE_DP2A,      /*                            2             */
    OPCODE_DP3,       /*   X        X       X       X         X   */
    OPCODE_DP4,       /*   X        X       X       X         X   */
@@ -185,7 +185,7 @@ typedef enum prog_opcode {
    OPCODE_LG2,       /*   X        X       2       X         X   */
    OPCODE_LIT,       /*   X        X       X       X             */
    OPCODE_LOG,       /*   X                X                 X   */
-   OPCODE_LRP,       /*            X               X             */
+   OPCODE_LRP,       /*            X               X         X   */
    OPCODE_MAD,       /*   X        X       X       X         X   */
    OPCODE_MAX,       /*   X        X       X       X         X   */
    OPCODE_MIN,       /*   X        X       X       X         X   */
@@ -196,8 +196,8 @@ typedef enum prog_opcode {
    OPCODE_NOISE3,    /*                                      X   */
    OPCODE_NOISE4,    /*                                      X   */
    OPCODE_NOT,       /*                                          */
-   OPCODE_NRM3,      /*                                          */
-   OPCODE_NRM4,      /*                                          */
+   OPCODE_NRM3,      /*                                      X   */
+   OPCODE_NRM4,      /*                                      X   */
    OPCODE_OR,        /*                                          */
    OPCODE_PK2H,      /*                            X             */
    OPCODE_PK2US,     /*                            X             */
@@ -209,7 +209,7 @@ typedef enum prog_opcode {
    OPCODE_PUSHA,     /*                    3                     */
    OPCODE_RCC,       /*                    1.1                   */
    OPCODE_RCP,       /*   X        X       X       X         X   */
-   OPCODE_RET,       /*                    2       2             */
+   OPCODE_RET,       /*                    2       2         X   */
    OPCODE_RFL,       /*            X               X             */
    OPCODE_RSQ,       /*   X        X       X       X         X   */
    OPCODE_SCS,       /*            X                             */
diff --git a/src/mesa/program/prog_noise.h b/src/mesa/program/prog_noise.h
index c4779479f9b..dd7986efcdb 100644
--- a/src/mesa/program/prog_noise.h
+++ b/src/mesa/program/prog_noise.h
@@ -25,6 +25,8 @@
 #ifndef PROG_NOISE
 #define PROG_NOISE
 
+#include "main/glheader.h"
+
 extern GLfloat _mesa_noise1(GLfloat);
 extern GLfloat _mesa_noise2(GLfloat, GLfloat);
 extern GLfloat _mesa_noise3(GLfloat, GLfloat, GLfloat);
diff --git a/src/mesa/program/prog_optimize.c b/src/mesa/program/prog_optimize.c
index 2941a17da3f..c78187c983d 100644
--- a/src/mesa/program/prog_optimize.c
+++ b/src/mesa/program/prog_optimize.c
@@ -38,40 +38,117 @@
 
 static GLboolean dbg = GL_FALSE;
 
-/* Returns the mask of channels read from the given srcreg in this instruction.
+#define NO_MASK 0xf
+
+/**
+ * Returns the mask of channels (bitmask of WRITEMASK_X,Y,Z,W) which
+ * are read from the given src in this instruction, We also provide
+ * one optional masks which may mask other components in the dst
+ * register
  */
 static GLuint
-get_src_arg_mask(const struct prog_instruction *inst, int arg)
+get_src_arg_mask(const struct prog_instruction *inst,
+                 GLuint arg, GLuint dst_mask)
 {
-   int writemask = inst->DstReg.WriteMask;
+   GLuint read_mask, channel_mask;
+   GLuint comp;
 
-   if (inst->CondUpdate)
-      writemask = WRITEMASK_XYZW;
+   ASSERT(arg < _mesa_num_inst_src_regs(inst->Opcode));
 
-   switch (inst->Opcode) {
-   case OPCODE_MOV:
-   case OPCODE_ABS:
-   case OPCODE_ADD:
-   case OPCODE_MUL:
-   case OPCODE_SUB:
-      return writemask;
-   case OPCODE_RCP:
-   case OPCODE_SIN:
-   case OPCODE_COS:
-   case OPCODE_RSQ:
-   case OPCODE_POW:
-   case OPCODE_EX2:
-      return WRITEMASK_X;
-   case OPCODE_DP2:
-      return WRITEMASK_XY;
-   case OPCODE_DP3:
-   case OPCODE_XPD:
-      return WRITEMASK_XYZ;
-   default:
-      return WRITEMASK_XYZW;
+   /* Form the dst register, find the written channels */
+   if (inst->CondUpdate) {
+      channel_mask = WRITEMASK_XYZW;
+   }
+   else {
+      switch (inst->Opcode) {
+      case OPCODE_MOV:
+      case OPCODE_MIN:
+      case OPCODE_MAX:
+      case OPCODE_ABS:
+      case OPCODE_ADD:
+      case OPCODE_MAD:
+      case OPCODE_MUL:
+      case OPCODE_SUB:
+         channel_mask = inst->DstReg.WriteMask & dst_mask;
+         break;
+      case OPCODE_RCP:
+      case OPCODE_SIN:
+      case OPCODE_COS:
+      case OPCODE_RSQ:
+      case OPCODE_POW:
+      case OPCODE_EX2:
+      case OPCODE_LOG:
+         channel_mask = WRITEMASK_X;
+         break;
+      case OPCODE_DP2:
+         channel_mask = WRITEMASK_XY;
+         break;
+      case OPCODE_DP3:
+      case OPCODE_XPD:
+         channel_mask = WRITEMASK_XYZ;
+         break;
+      default:
+         channel_mask = WRITEMASK_XYZW;
+         break;
+      }
    }
+
+   /* Now, given the src swizzle and the written channels, find which
+    * components are actually read
+    */
+   read_mask = 0x0;
+   for (comp = 0; comp < 4; ++comp) {
+      const GLuint coord = GET_SWZ(inst->SrcReg[arg].Swizzle, comp);
+      ASSERT(coord < 4);
+      if (channel_mask & (1 << comp) && coord <= SWIZZLE_W)
+         read_mask |= 1 << coord;
+   }
+
+   return read_mask;
+}
+
+
+/**
+ * For a MOV instruction, compute a write mask when src register also has
+ * a mask
+ */
+static GLuint
+get_dst_mask_for_mov(const struct prog_instruction *mov, GLuint src_mask)
+{
+   const GLuint mask = mov->DstReg.WriteMask;
+   GLuint comp;
+   GLuint updated_mask = 0x0;
+
+   ASSERT(mov->Opcode == OPCODE_MOV);
+
+   for (comp = 0; comp < 4; ++comp) {
+      GLuint src_comp;
+      if ((mask & (1 << comp)) == 0)
+         continue;
+      src_comp = GET_SWZ(mov->SrcReg[0].Swizzle, comp);
+      if ((src_mask & (1 << src_comp)) == 0)
+         continue;
+      updated_mask |= 1 << comp;
+   }
+
+   return updated_mask;
+}
+
+
+/**
+ * Ensure that the swizzle is regular.  That is, all of the swizzle
+ * terms are SWIZZLE_X,Y,Z,W and not SWIZZLE_ZERO or SWIZZLE_ONE.
+ */
+static GLboolean
+is_swizzle_regular(GLuint swz)
+{
+   return GET_SWZ(swz,0) <= SWIZZLE_W &&
+          GET_SWZ(swz,1) <= SWIZZLE_W &&
+          GET_SWZ(swz,2) <= SWIZZLE_W &&
+          GET_SWZ(swz,3) <= SWIZZLE_W;
 }
 
+
 /**
  * In 'prog' remove instruction[i] if removeFlags[i] == TRUE.
  * \return number of instructions removed
@@ -148,82 +225,13 @@ replace_regs(struct gl_program *prog, gl_register_file file, const GLint map[])
 
 
 /**
- * Consolidate temporary registers to use low numbers.  For example, if the
- * shader only uses temps 4, 5, 8, replace them with 0, 1, 2.
- */
-static void
-_mesa_consolidate_registers(struct gl_program *prog)
-{
-   GLboolean tempUsed[MAX_PROGRAM_TEMPS];
-   GLint tempMap[MAX_PROGRAM_TEMPS];
-   GLuint tempMax = 0, i;
-
-   if (dbg) {
-      printf("Optimize: Begin register consolidation\n");
-   }
-
-   memset(tempUsed, 0, sizeof(tempUsed));
-
-   for (i = 0; i < MAX_PROGRAM_TEMPS; i++) {
-      tempMap[i] = -1;
-   }
-
-   /* set tempUsed[i] if temporary [i] is referenced */
-   for (i = 0; i < prog->NumInstructions; i++) {
-      const struct prog_instruction *inst = prog->Instructions + i;
-      const GLuint numSrc = _mesa_num_inst_src_regs(inst->Opcode);
-      GLuint j;
-      for (j = 0; j < numSrc; j++) {
-         if (inst->SrcReg[j].File == PROGRAM_TEMPORARY) {
-            const GLuint index = inst->SrcReg[j].Index;
-            ASSERT(index < MAX_PROGRAM_TEMPS);
-            tempUsed[index] = GL_TRUE;
-            tempMax = MAX2(tempMax, index);
-            break;
-         }
-      }
-      if (inst->DstReg.File == PROGRAM_TEMPORARY) {
-         const GLuint index = inst->DstReg.Index;
-         ASSERT(index < MAX_PROGRAM_TEMPS);
-         tempUsed[index] = GL_TRUE;
-         tempMax = MAX2(tempMax, index);
-      }
-   }
-
-   /* allocate a new index for each temp that's used */
-   {
-      GLuint freeTemp = 0;
-      for (i = 0; i <= tempMax; i++) {
-         if (tempUsed[i]) {
-            tempMap[i] = freeTemp++;
-            /*printf("replace %u with %u\n", i, tempMap[i]);*/
-         }
-      }
-      if (freeTemp == tempMax + 1) {
-         /* no consolidation possible */
-         return;
-      }         
-      if (dbg) {
-         printf("Replace regs 0..%u with 0..%u\n", tempMax, freeTemp-1);
-      }
-   }
-
-   replace_regs(prog, PROGRAM_TEMPORARY, tempMap);
-
-   if (dbg) {
-      printf("Optimize: End register consolidation\n");
-   }
-}
-
-
-/**
  * Remove dead instructions from the given program.
  * This is very primitive for now.  Basically look for temp registers
  * that are written to but never read.  Remove any instructions that
  * write to such registers.  Be careful with condition code setters.
  */
-static void
-_mesa_remove_dead_code(struct gl_program *prog)
+static GLboolean
+_mesa_remove_dead_code_global(struct gl_program *prog)
 {
    GLboolean tempRead[MAX_PROGRAM_TEMPS][4];
    GLboolean *removeInst; /* per-instruction removal flag */
@@ -251,7 +259,7 @@ _mesa_remove_dead_code(struct gl_program *prog)
             const GLuint index = inst->SrcReg[j].Index;
             GLuint read_mask;
             ASSERT(index < MAX_PROGRAM_TEMPS);
-	    read_mask = get_src_arg_mask(inst, j);
+	    read_mask = get_src_arg_mask(inst, j, NO_MASK);
 
             if (inst->SrcReg[j].RelAddr) {
                if (dbg)
@@ -260,25 +268,12 @@ _mesa_remove_dead_code(struct gl_program *prog)
             }
 
 	    for (comp = 0; comp < 4; comp++) {
-	       GLuint swz = (inst->SrcReg[j].Swizzle >> (3 * comp)) & 0x7;
-
-	       if ((read_mask & (1 << comp)) == 0)
+	       const GLuint swz = GET_SWZ(inst->SrcReg[j].Swizzle, comp);
+	       ASSERT(swz < 4);
+               if ((read_mask & (1 << swz)) == 0)
 		  continue;
-
-	       switch (swz) {
-	       case SWIZZLE_X:
-		  tempRead[index][0] = GL_TRUE;
-		  break;
-	       case SWIZZLE_Y:
-		  tempRead[index][1] = GL_TRUE;
-		  break;
-	       case SWIZZLE_Z:
-		  tempRead[index][2] = GL_TRUE;
-		  break;
-	       case SWIZZLE_W:
-		  tempRead[index][3] = GL_TRUE;
-		  break;
-	       }
+               if (swz <= SWIZZLE_W)
+                  tempRead[index][swz] = GL_TRUE;
 	    }
          }
       }
@@ -348,10 +343,11 @@ _mesa_remove_dead_code(struct gl_program *prog)
 
 done:
    free(removeInst);
+   return rem != 0;
 }
 
 
-enum temp_use
+enum inst_use
 {
    READ,
    WRITE,
@@ -359,13 +355,19 @@ enum temp_use
    END
 };
 
+
 /**
- * Scan forward in program from 'start' for the next occurance of TEMP[index].
+ * Scan forward in program from 'start' for the next occurances of TEMP[index].
+ * We look if an instruction reads the component given by the masks and if they
+ * are overwritten.
  * Return READ, WRITE, FLOW or END to indicate the next usage or an indicator
  * that we can't look further.
  */
-static enum temp_use
-find_next_temp_use(const struct gl_program *prog, GLuint start, GLuint index)
+static enum inst_use
+find_next_use(const struct gl_program *prog,
+              GLuint start,
+              GLuint index,
+              GLuint mask)
 {
    GLuint i;
 
@@ -373,30 +375,50 @@ find_next_temp_use(const struct gl_program *prog, GLuint start, GLuint index)
       const struct prog_instruction *inst = prog->Instructions + i;
       switch (inst->Opcode) {
       case OPCODE_BGNLOOP:
-      case OPCODE_ENDLOOP:
       case OPCODE_BGNSUB:
+      case OPCODE_BRA:
+      case OPCODE_CAL:
+      case OPCODE_CONT:
+      case OPCODE_IF:
+      case OPCODE_ELSE:
+      case OPCODE_ENDIF:
+      case OPCODE_ENDLOOP:
       case OPCODE_ENDSUB:
+      case OPCODE_RET:
          return FLOW;
+      case OPCODE_END:
+         return END;
       default:
          {
             const GLuint numSrc = _mesa_num_inst_src_regs(inst->Opcode);
             GLuint j;
             for (j = 0; j < numSrc; j++) {
-               if (inst->SrcReg[j].File == PROGRAM_TEMPORARY &&
-                   inst->SrcReg[j].Index == index)
+               if (inst->SrcReg[j].RelAddr ||
+                   (inst->SrcReg[j].File == PROGRAM_TEMPORARY &&
+                   inst->SrcReg[j].Index == index &&
+                   (get_src_arg_mask(inst,j,NO_MASK) & mask)))
                   return READ;
             }
-            if (inst->DstReg.File == PROGRAM_TEMPORARY &&
-                inst->DstReg.Index == index)
-               return WRITE;
+            if (_mesa_num_inst_dst_regs(inst->Opcode) == 1 &&
+                inst->DstReg.File == PROGRAM_TEMPORARY &&
+                inst->DstReg.Index == index) {
+               mask &= ~inst->DstReg.WriteMask;
+               if (mask == 0)
+                  return WRITE;
+            }
          }
       }
    }
-
    return END;
 }
 
-static GLboolean _mesa_is_flow_control_opcode(enum prog_opcode opcode)
+
+/**
+ * Is the given instruction opcode a flow-control opcode?
+ * XXX maybe move this into prog_instruction.[ch]
+ */
+static GLboolean
+_mesa_is_flow_control_opcode(enum prog_opcode opcode)
 {
    switch (opcode) {
    case OPCODE_BGNLOOP:
@@ -417,6 +439,37 @@ static GLboolean _mesa_is_flow_control_opcode(enum prog_opcode opcode)
    }
 }
 
+
+/**
+ * Test if the given instruction is a simple MOV (no conditional updating,
+ * not relative addressing, no negation/abs, etc).
+ */
+static GLboolean
+can_downward_mov_be_modifed(const struct prog_instruction *mov)
+{
+   return
+      mov->Opcode == OPCODE_MOV &&
+      mov->CondUpdate == GL_FALSE &&
+      mov->SrcReg[0].RelAddr == 0 &&
+      mov->SrcReg[0].Negate == 0 &&
+      mov->SrcReg[0].Abs == 0 &&
+      mov->SrcReg[0].HasIndex2 == 0 &&
+      mov->SrcReg[0].RelAddr2 == 0 &&
+      mov->DstReg.RelAddr == 0 &&
+      mov->DstReg.CondMask == COND_TR &&
+      mov->SaturateMode == SATURATE_OFF;
+}
+
+
+static GLboolean
+can_upward_mov_be_modifed(const struct prog_instruction *mov)
+{
+   return
+      can_downward_mov_be_modifed(mov) &&
+      mov->DstReg.File == PROGRAM_TEMPORARY;
+}
+
+
 /**
  * Try to remove use of extraneous MOV instructions, to free them up for dead
  * code removal.
@@ -444,14 +497,15 @@ _mesa_remove_extra_move_use(struct gl_program *prog)
 
    for (i = 0; i + 1 < prog->NumInstructions; i++) {
       const struct prog_instruction *mov = prog->Instructions + i;
+      GLuint dst_mask, src_mask;
+      if (can_upward_mov_be_modifed(mov) == GL_FALSE)
+         continue;
 
-      if (mov->Opcode != OPCODE_MOV ||
-	  mov->DstReg.File != PROGRAM_TEMPORARY ||
-	  mov->DstReg.RelAddr ||
-	  mov->DstReg.CondMask != COND_TR ||
-	  mov->SaturateMode != SATURATE_OFF ||
-	  mov->SrcReg[0].RelAddr)
-	 continue;
+      /* Scanning the code, we maintain the components which are still active in
+       * these two masks
+       */
+      dst_mask = mov->DstReg.WriteMask;
+      src_mask = get_src_arg_mask(mov, 0, NO_MASK);
 
       /* Walk through remaining instructions until the or src reg gets
        * rewritten or we get into some flow-control, eliminating the use of
@@ -459,61 +513,60 @@ _mesa_remove_extra_move_use(struct gl_program *prog)
        */
       for (j = i + 1; j < prog->NumInstructions; j++) {
 	 struct prog_instruction *inst2 = prog->Instructions + j;
-	 GLuint arg;
+         GLuint arg;
 
 	 if (_mesa_is_flow_control_opcode(inst2->Opcode))
 	     break;
 
 	 /* First rewrite this instruction's args if appropriate. */
 	 for (arg = 0; arg < _mesa_num_inst_src_regs(inst2->Opcode); arg++) {
-	    int comp;
-	    int read_mask = get_src_arg_mask(inst2, arg);
+	    GLuint comp, read_mask;
 
 	    if (inst2->SrcReg[arg].File != mov->DstReg.File ||
 		inst2->SrcReg[arg].Index != mov->DstReg.Index ||
 		inst2->SrcReg[arg].RelAddr ||
 		inst2->SrcReg[arg].Abs)
 	       continue;
+            read_mask = get_src_arg_mask(inst2, arg, NO_MASK);
 
-	    /* Check that all the sources for this arg of inst2 come from inst1
-	     * or constants.
-	     */
-	    for (comp = 0; comp < 4; comp++) {
-	       int src_swz = GET_SWZ(inst2->SrcReg[arg].Swizzle, comp);
-
-	       /* If the MOV didn't write that channel, can't use it. */
-	       if ((read_mask & (1 << comp)) &&
-		   src_swz <= SWIZZLE_W &&
-		   (mov->DstReg.WriteMask & (1 << src_swz)) == 0)
-		  break;
-	    }
-	    if (comp != 4)
-	       continue;
-
-	    /* Adjust the swizzles of inst2 to point at MOV's source */
-	    for (comp = 0; comp < 4; comp++) {
-	       int inst2_swz = GET_SWZ(inst2->SrcReg[arg].Swizzle, comp);
-
-	       if (inst2_swz <= SWIZZLE_W) {
-		  GLuint s = GET_SWZ(mov->SrcReg[0].Swizzle, inst2_swz);
-		  inst2->SrcReg[arg].Swizzle &= ~(7 << (3 * comp));
-		  inst2->SrcReg[arg].Swizzle |= s << (3 * comp);
-		  inst2->SrcReg[arg].Negate ^= (((mov->SrcReg[0].Negate >>
-						  inst2_swz) & 0x1) << comp);
-	       }
-	    }
-	    inst2->SrcReg[arg].File = mov->SrcReg[0].File;
-	    inst2->SrcReg[arg].Index = mov->SrcReg[0].Index;
+	    /* Adjust the swizzles of inst2 to point at MOV's source if ALL the
+             * components read still come from the mov instructions
+             */
+            if (is_swizzle_regular(inst2->SrcReg[arg].Swizzle) &&
+               (read_mask & dst_mask) == read_mask) {
+               for (comp = 0; comp < 4; comp++) {
+                  const GLuint inst2_swz =
+                     GET_SWZ(inst2->SrcReg[arg].Swizzle, comp);
+                  const GLuint s = GET_SWZ(mov->SrcReg[0].Swizzle, inst2_swz);
+                  inst2->SrcReg[arg].Swizzle &= ~(7 << (3 * comp));
+                  inst2->SrcReg[arg].Swizzle |= s << (3 * comp);
+                  inst2->SrcReg[arg].Negate ^= (((mov->SrcReg[0].Negate >>
+                                                  inst2_swz) & 0x1) << comp);
+               }
+               inst2->SrcReg[arg].File = mov->SrcReg[0].File;
+               inst2->SrcReg[arg].Index = mov->SrcReg[0].Index;
+            }
 	 }
 
-	 /* If this instruction overwrote part of the move, our time is up. */
-	 if ((inst2->DstReg.File == mov->DstReg.File &&
-	      (inst2->DstReg.RelAddr ||
-	       inst2->DstReg.Index == mov->DstReg.Index)) ||
-	     (inst2->DstReg.File == mov->SrcReg[0].File &&
-	      (inst2->DstReg.RelAddr ||
-	       inst2->DstReg.Index == mov->SrcReg[0].Index)))
-	    break;
+	 /* The source of MOV is written. This potentially deactivates some
+          * components from the src and dst of the MOV instruction
+          */
+	 if (inst2->DstReg.File == mov->DstReg.File &&
+	     (inst2->DstReg.RelAddr ||
+	      inst2->DstReg.Index == mov->DstReg.Index)) {
+            dst_mask &= ~inst2->DstReg.WriteMask;
+            src_mask = get_src_arg_mask(mov, 0, dst_mask);
+         }
+
+         /* Idem when the destination of mov is written */
+	 if (inst2->DstReg.File == mov->SrcReg[0].File &&
+	     (inst2->DstReg.RelAddr ||
+	      inst2->DstReg.Index == mov->SrcReg[0].Index)) {
+            src_mask &= ~inst2->DstReg.WriteMask;
+            dst_mask &= get_dst_mask_for_mov(mov, src_mask);
+         }
+         if (dst_mask == 0)
+            break;
       }
    }
 
@@ -523,14 +576,151 @@ _mesa_remove_extra_move_use(struct gl_program *prog)
    }
 }
 
+
+/**
+ * Complements dead_code_global. Try to remove code in block of code by
+ * carefully monitoring the swizzles. Both functions should be merged into one
+ * with a proper control flow graph
+ */
+static GLboolean
+_mesa_remove_dead_code_local(struct gl_program *prog)
+{
+   GLboolean *removeInst;
+   GLuint i, arg, rem = 0;
+
+   removeInst = (GLboolean *)
+      calloc(1, prog->NumInstructions * sizeof(GLboolean));
+
+   for (i = 0; i < prog->NumInstructions; i++) {
+      const struct prog_instruction *inst = prog->Instructions + i;
+      const GLuint index = inst->DstReg.Index;
+      const GLuint mask = inst->DstReg.WriteMask;
+      enum inst_use use;
+
+      /* We must deactivate the pass as soon as some indirection is used */
+      if (inst->DstReg.RelAddr)
+         goto done;
+      for (arg = 0; arg < _mesa_num_inst_src_regs(inst->Opcode); arg++)
+         if (inst->SrcReg[arg].RelAddr)
+            goto done;
+
+      if (_mesa_is_flow_control_opcode(inst->Opcode) ||
+          _mesa_num_inst_dst_regs(inst->Opcode) == 0 ||
+          inst->DstReg.File != PROGRAM_TEMPORARY ||
+          inst->DstReg.RelAddr)
+         continue;
+
+      use = find_next_use(prog, i+1, index, mask);
+      if (use == WRITE || use == END)
+         removeInst[i] = GL_TRUE;
+   }
+
+   rem = remove_instructions(prog, removeInst);
+
+done:
+   free(removeInst);
+   return rem != 0;
+}
+
+
+/**
+ * Try to inject the destination of mov as the destination of inst and recompute
+ * the swizzles operators for the sources of inst if required. Return GL_TRUE
+ * of the substitution was possible, GL_FALSE otherwise
+ */
+static GLboolean
+_mesa_merge_mov_into_inst(struct prog_instruction *inst,
+                          const struct prog_instruction *mov)
+{
+   /* Indirection table which associates destination and source components for
+    * the mov instruction
+    */
+   const GLuint mask = get_src_arg_mask(mov, 0, NO_MASK);
+
+   /* Some components are not written by inst. We cannot remove the mov */
+   if (mask != (inst->DstReg.WriteMask & mask))
+      return GL_FALSE;
+
+   /* Depending on the instruction, we may need to recompute the swizzles.
+    * Also, some other instructions (like TEX) are not linear. We will only
+    * consider completely active sources and destinations
+    */
+   switch (inst->Opcode) {
+
+   /* Carstesian instructions: we compute the swizzle */
+   case OPCODE_MOV:
+   case OPCODE_MIN:
+   case OPCODE_MAX:
+   case OPCODE_ABS:
+   case OPCODE_ADD:
+   case OPCODE_MAD:
+   case OPCODE_MUL:
+   case OPCODE_SUB:
+   {
+      GLuint dst_to_src_comp[4] = {0,0,0,0};
+      GLuint dst_comp, arg;
+      for (dst_comp = 0; dst_comp < 4; ++dst_comp) {
+         if (mov->DstReg.WriteMask & (1 << dst_comp)) {
+            const GLuint src_comp = GET_SWZ(mov->SrcReg[0].Swizzle, dst_comp);
+            ASSERT(src_comp < 4);
+            dst_to_src_comp[dst_comp] = src_comp;
+         }
+      }
+
+      /* Patch each source of the instruction */
+      for (arg = 0; arg < _mesa_num_inst_src_regs(inst->Opcode); arg++) {
+         const GLuint arg_swz = inst->SrcReg[arg].Swizzle;
+         inst->SrcReg[arg].Swizzle = 0;
+
+         /* Reset each active component of the swizzle */
+         for (dst_comp = 0; dst_comp < 4; ++dst_comp) {
+            GLuint src_comp, arg_comp;
+            if ((mov->DstReg.WriteMask & (1 << dst_comp)) == 0)
+               continue;
+            src_comp = dst_to_src_comp[dst_comp];
+            ASSERT(src_comp < 4);
+            arg_comp = GET_SWZ(arg_swz, src_comp);
+            ASSERT(arg_comp < 4);
+            inst->SrcReg[arg].Swizzle |= arg_comp << (3*dst_comp);
+         }
+      }
+      inst->DstReg = mov->DstReg;
+      return GL_TRUE;
+   }
+
+   /* Dot products and scalar instructions: we only change the destination */
+   case OPCODE_RCP:
+   case OPCODE_SIN:
+   case OPCODE_COS:
+   case OPCODE_RSQ:
+   case OPCODE_POW:
+   case OPCODE_EX2:
+   case OPCODE_LOG:
+   case OPCODE_DP2:
+   case OPCODE_DP3:
+   case OPCODE_DP4:
+      inst->DstReg = mov->DstReg;
+      return GL_TRUE;
+
+   /* All other instructions require fully active components with no swizzle */
+   default:
+      if (mov->SrcReg[0].Swizzle != SWIZZLE_XYZW ||
+          inst->DstReg.WriteMask != WRITEMASK_XYZW)
+         return GL_FALSE;
+      inst->DstReg = mov->DstReg;
+      return GL_TRUE;
+   }
+}
+
+
 /**
  * Try to remove extraneous MOV instructions from the given program.
  */
-static void
+static GLboolean
 _mesa_remove_extra_moves(struct gl_program *prog)
 {
    GLboolean *removeInst; /* per-instruction removal flag */
-   GLuint i, rem, loopNesting = 0, subroutineNesting = 0;
+   GLuint i, rem = 0, nesting = 0;
 
    if (dbg) {
       printf("Optimize: Begin remove extra moves\n");
@@ -549,29 +739,24 @@ _mesa_remove_extra_moves(struct gl_program *prog)
     */
 
    for (i = 0; i < prog->NumInstructions; i++) {
-      const struct prog_instruction *inst = prog->Instructions + i;
+      const struct prog_instruction *mov = prog->Instructions + i;
 
-      switch (inst->Opcode) {
+      switch (mov->Opcode) {
       case OPCODE_BGNLOOP:
-         loopNesting++;
-         break;
-      case OPCODE_ENDLOOP:
-         loopNesting--;
-         break;
       case OPCODE_BGNSUB:
-         subroutineNesting++;
+      case OPCODE_IF:
+         nesting++;
          break;
+      case OPCODE_ENDLOOP:
       case OPCODE_ENDSUB:
-         subroutineNesting--;
+      case OPCODE_ENDIF:
+         nesting--;
          break;
       case OPCODE_MOV:
-         if (i > 0 &&
-             loopNesting == 0 &&
-             subroutineNesting == 0 &&
-             inst->SrcReg[0].File == PROGRAM_TEMPORARY &&
-             inst->SrcReg[0].Swizzle == SWIZZLE_XYZW) {
+         if (i > 0 && can_downward_mov_be_modifed(mov) && nesting == 0) {
+
             /* see if this MOV can be removed */
-            const GLuint tempIndex = inst->SrcReg[0].Index;
+            const GLuint id = mov->SrcReg[0].Index;
             struct prog_instruction *prevInst;
             GLuint prevI;
 
@@ -582,11 +767,13 @@ _mesa_remove_extra_moves(struct gl_program *prog)
             prevInst = prog->Instructions + prevI;
 
             if (prevInst->DstReg.File == PROGRAM_TEMPORARY &&
-                prevInst->DstReg.Index == tempIndex &&
-                prevInst->DstReg.WriteMask == WRITEMASK_XYZW) {
+                prevInst->DstReg.Index == id &&
+                prevInst->DstReg.RelAddr == 0 &&
+                prevInst->DstReg.CondSrc == 0 && 
+                prevInst->DstReg.CondMask == COND_TR) {
 
-               enum temp_use next_use =
-                  find_next_temp_use(prog, i + 1, tempIndex);
+               const GLuint dst_mask = prevInst->DstReg.WriteMask;
+               enum inst_use next_use = find_next_use(prog, i+1, id, dst_mask);
 
                if (next_use == WRITE || next_use == END) {
                   /* OK, we can safely remove this MOV instruction.
@@ -596,18 +783,13 @@ _mesa_remove_extra_moves(struct gl_program *prog)
                    * Into:
                    *   prevI: FOO z, x, y;
                    */
-
-                  /* patch up prev inst */
-                  prevInst->DstReg.File = inst->DstReg.File;
-                  prevInst->DstReg.Index = inst->DstReg.Index;
-
-                  /* flag this instruction for removal */
-                  removeInst[i] = GL_TRUE;
-
-                  if (dbg) {
-                     printf("Remove MOV at %u\n", i);
-                     printf("new prev inst %u: ", prevI);
-                     _mesa_print_instruction(prevInst);
+                  if (_mesa_merge_mov_into_inst(prevInst, mov)) {
+                     removeInst[i] = GL_TRUE;
+                     if (dbg) {
+                        printf("Remove MOV at %u\n", i);
+                        printf("new prev inst %u: ", prevI);
+                        _mesa_print_instruction(prevInst);
+                     }
                   }
                }
             }
@@ -627,6 +809,8 @@ _mesa_remove_extra_moves(struct gl_program *prog)
       printf("Optimize: End remove extra moves.  %u instructions removed\n", rem);
       /*_mesa_print_program(prog);*/
    }
+
+   return rem != 0;
 }
 
 
@@ -713,6 +897,7 @@ compare_start(const void *a, const void *b)
       return 0;
 }
 
+
 /** sort the interval list according to interval starts */
 static void
 sort_interval_list_by_start(struct interval_list *list)
@@ -1013,6 +1198,17 @@ _mesa_reallocate_registers(struct gl_program *prog)
 }
 
 
+#if 0
+static void
+print_it(GLcontext *ctx, struct gl_program *program, const char *txt) {
+   fprintf(stderr, "%s (%u inst):\n", txt, program->NumInstructions);
+   _mesa_print_program(program);
+   _mesa_print_program_parameters(ctx, program);
+   fprintf(stderr, "\n\n");
+}
+#endif
+
+
 /**
  * Apply optimizations to the given program to eliminate unnecessary
  * instructions, temp regs, etc.
@@ -1020,16 +1216,19 @@ _mesa_reallocate_registers(struct gl_program *prog)
 void
 _mesa_optimize_program(GLcontext *ctx, struct gl_program *program)
 {
-   _mesa_remove_extra_move_use(program);
-
-   if (1)
-      _mesa_remove_dead_code(program);
-
-   if (0) /* not tested much yet */
-      _mesa_remove_extra_moves(program);
-
-   if (0)
-      _mesa_consolidate_registers(program);
-   else
+   GLboolean any_change;
+
+   /* Stop when no modifications were output */
+   do {
+      any_change = GL_FALSE;
+      _mesa_remove_extra_move_use(program);
+      if (_mesa_remove_dead_code_global(program))
+         any_change = GL_TRUE;
+      if (_mesa_remove_extra_moves(program))
+         any_change = GL_TRUE;
+      if (_mesa_remove_dead_code_local(program))
+         any_change = GL_TRUE;
       _mesa_reallocate_registers(program);
+   } while (any_change);
 }
+
diff --git a/src/mesa/program/prog_optimize.h b/src/mesa/program/prog_optimize.h
index 43894a27237..06cd9cb2c20 100644
--- a/src/mesa/program/prog_optimize.h
+++ b/src/mesa/program/prog_optimize.h
@@ -27,6 +27,7 @@
 
 
 #include "main/config.h"
+#include "main/mtypes.h"
 
 
 struct gl_program;
diff --git a/src/mesa/program/prog_parameter_layout.c b/src/mesa/program/prog_parameter_layout.c
index a8885738321..d7dc97edbfb 100644
--- a/src/mesa/program/prog_parameter_layout.c
+++ b/src/mesa/program/prog_parameter_layout.c
@@ -28,6 +28,7 @@
  * \author Ian Romanick <[email protected]>
  */
 
+#include "main/compiler.h"
 #include "main/mtypes.h"
 #include "prog_parameter.h"
 #include "prog_parameter_layout.h"
diff --git a/src/mesa/program/prog_print.c b/src/mesa/program/prog_print.c
index 6ab199aa02b..6056c459e4c 100644
--- a/src/mesa/program/prog_print.c
+++ b/src/mesa/program/prog_print.c
@@ -924,6 +924,8 @@ _mesa_fprint_program_parameters(FILE *f,
    fprintf(f, "NumParameters=%d\n", prog->NumParameters);
    fprintf(f, "NumAttributes=%d\n", prog->NumAttributes);
    fprintf(f, "NumAddressRegs=%d\n", prog->NumAddressRegs);
+   fprintf(f, "IndirectRegisterFiles: 0x%x (0b%s)\n",
+           prog->IndirectRegisterFiles, binary(prog->IndirectRegisterFiles));
    fprintf(f, "SamplersUsed: 0x%x (0b%s)\n",
                  prog->SamplersUsed, binary(prog->SamplersUsed));
    fprintf(f, "Samplers=[ ");
diff --git a/src/mesa/program/prog_print.h b/src/mesa/program/prog_print.h
index 9ab74560169..4667373f379 100644
--- a/src/mesa/program/prog_print.h
+++ b/src/mesa/program/prog_print.h
@@ -26,6 +26,16 @@
 #ifndef PROG_PRINT_H
 #define PROG_PRINT_H
 
+#include <stdio.h>
+
+#include "main/glheader.h"
+#include "main/mtypes.h"
+
+struct gl_program;
+struct gl_program_parameter_list;
+struct gl_shader;
+struct prog_instruction;
+
 
 /**
  * The output style to use when printing programs.
diff --git a/src/mesa/program/prog_uniform.h b/src/mesa/program/prog_uniform.h
index a671d30bfe8..7988d534a7d 100644
--- a/src/mesa/program/prog_uniform.h
+++ b/src/mesa/program/prog_uniform.h
@@ -31,8 +31,7 @@
 #ifndef PROG_UNIFORM_H
 #define PROG_UNIFORM_H
 
-#include "main/mtypes.h"
-#include "prog_statevars.h"
+#include "main/glheader.h"
 
 
 /**
diff --git a/src/mesa/program/program.c b/src/mesa/program/program.c
index cf46095ce84..3b6d6827446 100644
--- a/src/mesa/program/program.c
+++ b/src/mesa/program/program.c
@@ -55,13 +55,21 @@ _mesa_init_program(GLcontext *ctx)
 
    /*
     * If this assertion fails, we need to increase the field
-    * size for register indexes.
+    * size for register indexes (see INST_INDEX_BITS).
     */
    ASSERT(ctx->Const.VertexProgram.MaxUniformComponents / 4
           <= (1 << INST_INDEX_BITS));
    ASSERT(ctx->Const.FragmentProgram.MaxUniformComponents / 4
           <= (1 << INST_INDEX_BITS));
 
+   ASSERT(ctx->Const.VertexProgram.MaxTemps <= (1 << INST_INDEX_BITS));
+   ASSERT(ctx->Const.VertexProgram.MaxLocalParams <= (1 << INST_INDEX_BITS));
+   ASSERT(ctx->Const.FragmentProgram.MaxTemps <= (1 << INST_INDEX_BITS));
+   ASSERT(ctx->Const.FragmentProgram.MaxLocalParams <= (1 << INST_INDEX_BITS));
+
+   ASSERT(ctx->Const.VertexProgram.MaxUniformComponents <= 4 * MAX_UNIFORMS);
+   ASSERT(ctx->Const.FragmentProgram.MaxUniformComponents <= 4 * MAX_UNIFORMS);
+
    /* If this fails, increase prog_instruction::TexSrcUnit size */
    ASSERT(MAX_TEXTURE_UNITS < (1 << 5));
 
@@ -512,6 +520,7 @@ _mesa_clone_program(GLcontext *ctx, const struct gl_program *prog)
    if (prog->Attributes)
       clone->Attributes = _mesa_clone_parameter_list(prog->Attributes);
    memcpy(clone->LocalParams, prog->LocalParams, sizeof(clone->LocalParams));
+   clone->IndirectRegisterFiles = prog->IndirectRegisterFiles;
    clone->NumInstructions = prog->NumInstructions;
    clone->NumTemporaries = prog->NumTemporaries;
    clone->NumParameters = prog->NumParameters;
diff --git a/src/mesa/program/program_parse.tab.c b/src/mesa/program/program_parse.tab.c
index 6421d1f58aa..31a609600b7 100644
--- a/src/mesa/program/program_parse.tab.c
+++ b/src/mesa/program/program_parse.tab.c
@@ -798,29 +798,29 @@ static const yytype_uint16 yyrline[] =
      415,   459,   464,   474,   518,   524,   525,   526,   527,   528,
      529,   530,   531,   532,   533,   534,   535,   538,   550,   558,
      575,   582,   601,   612,   632,   657,   664,   697,   704,   719,
-     774,   817,   826,   847,   857,   861,   890,   909,   909,   911,
-     918,   930,   931,   932,   935,   949,   963,   983,   994,  1006,
-    1008,  1009,  1010,  1011,  1014,  1014,  1014,  1014,  1015,  1018,
-    1022,  1027,  1034,  1041,  1048,  1071,  1094,  1095,  1096,  1097,
-    1098,  1099,  1102,  1121,  1125,  1131,  1135,  1139,  1143,  1152,
-    1161,  1165,  1170,  1176,  1187,  1187,  1188,  1190,  1194,  1198,
-    1202,  1208,  1208,  1210,  1228,  1254,  1257,  1268,  1274,  1280,
-    1281,  1288,  1294,  1300,  1308,  1314,  1320,  1328,  1334,  1340,
-    1348,  1349,  1352,  1353,  1354,  1355,  1356,  1357,  1358,  1359,
-    1360,  1361,  1362,  1365,  1374,  1378,  1382,  1388,  1397,  1401,
-    1405,  1414,  1418,  1424,  1430,  1437,  1442,  1450,  1460,  1462,
-    1470,  1476,  1480,  1484,  1490,  1501,  1510,  1514,  1519,  1523,
-    1527,  1531,  1537,  1544,  1548,  1554,  1562,  1573,  1580,  1584,
-    1590,  1600,  1611,  1615,  1633,  1642,  1645,  1651,  1655,  1659,
-    1665,  1676,  1681,  1686,  1691,  1696,  1701,  1709,  1712,  1717,
-    1730,  1738,  1749,  1757,  1757,  1759,  1759,  1761,  1771,  1776,
-    1783,  1793,  1802,  1807,  1814,  1824,  1834,  1846,  1846,  1847,
-    1847,  1849,  1859,  1867,  1877,  1885,  1893,  1902,  1913,  1917,
-    1923,  1924,  1925,  1928,  1928,  1931,  1966,  1970,  1970,  1973,
-    1980,  1989,  2003,  2012,  2021,  2025,  2034,  2043,  2054,  2061,
-    2066,  2075,  2087,  2090,  2099,  2110,  2111,  2112,  2115,  2116,
-    2117,  2120,  2121,  2124,  2125,  2128,  2129,  2132,  2143,  2154,
-    2165,  2191,  2192
+     774,   817,   826,   848,   858,   862,   891,   910,   910,   912,
+     919,   931,   932,   933,   936,   950,   964,   984,   995,  1007,
+    1009,  1010,  1011,  1012,  1015,  1015,  1015,  1015,  1016,  1019,
+    1023,  1028,  1035,  1042,  1049,  1072,  1095,  1096,  1097,  1098,
+    1099,  1100,  1103,  1122,  1126,  1132,  1136,  1140,  1144,  1153,
+    1162,  1166,  1171,  1177,  1188,  1188,  1189,  1191,  1195,  1199,
+    1203,  1209,  1209,  1211,  1229,  1255,  1258,  1269,  1275,  1281,
+    1282,  1289,  1295,  1301,  1309,  1315,  1321,  1329,  1335,  1341,
+    1349,  1350,  1353,  1354,  1355,  1356,  1357,  1358,  1359,  1360,
+    1361,  1362,  1363,  1366,  1375,  1379,  1383,  1389,  1398,  1402,
+    1406,  1415,  1419,  1425,  1431,  1438,  1443,  1451,  1461,  1463,
+    1471,  1477,  1481,  1485,  1491,  1502,  1511,  1515,  1520,  1524,
+    1528,  1532,  1538,  1545,  1549,  1555,  1563,  1574,  1581,  1585,
+    1591,  1601,  1612,  1616,  1634,  1643,  1646,  1652,  1656,  1660,
+    1666,  1677,  1682,  1687,  1692,  1697,  1702,  1710,  1713,  1718,
+    1731,  1739,  1750,  1758,  1758,  1760,  1760,  1762,  1772,  1777,
+    1784,  1794,  1803,  1808,  1815,  1825,  1835,  1847,  1847,  1848,
+    1848,  1850,  1860,  1868,  1878,  1886,  1894,  1903,  1914,  1918,
+    1924,  1925,  1926,  1929,  1929,  1932,  1967,  1971,  1971,  1974,
+    1981,  1990,  2004,  2013,  2022,  2026,  2035,  2044,  2055,  2062,
+    2067,  2076,  2088,  2091,  2100,  2111,  2112,  2113,  2116,  2117,
+    2118,  2121,  2122,  2125,  2126,  2129,  2130,  2133,  2144,  2155,
+    2166,  2192,  2193
 };
 #endif
 
@@ -2844,6 +2844,7 @@ yyreduce:
 	   (yyval.src_reg).Base.File = (yyvsp[(1) - (4)].sym)->param_binding_type;
 
 	   if ((yyvsp[(3) - (4)].src_reg).Base.RelAddr) {
+              state->prog->IndirectRegisterFiles |= (1 << (yyval.src_reg).Base.File);
 	      (yyvsp[(1) - (4)].sym)->param_accessed_indirectly = 1;
 
 	      (yyval.src_reg).Base.RelAddr = 1;
@@ -2858,7 +2859,7 @@ yyreduce:
   case 63:
 
 /* Line 1455 of yacc.c  */
-#line 848 "program_parse.y"
+#line 849 "program_parse.y"
     {
            gl_register_file file = ((yyvsp[(1) - (1)].temp_sym).name != NULL) 
 	      ? (yyvsp[(1) - (1)].temp_sym).param_binding_type
@@ -2871,7 +2872,7 @@ yyreduce:
   case 64:
 
 /* Line 1455 of yacc.c  */
-#line 858 "program_parse.y"
+#line 859 "program_parse.y"
     {
 	   set_dst_reg(& (yyval.dst_reg), PROGRAM_OUTPUT, (yyvsp[(1) - (1)].result));
 	;}
@@ -2880,7 +2881,7 @@ yyreduce:
   case 65:
 
 /* Line 1455 of yacc.c  */
-#line 862 "program_parse.y"
+#line 863 "program_parse.y"
     {
 	   struct asm_symbol *const s = (struct asm_symbol *)
 	      _mesa_symbol_table_find_symbol(state->st, 0, (yyvsp[(1) - (1)].string));
@@ -2912,7 +2913,7 @@ yyreduce:
   case 66:
 
 /* Line 1455 of yacc.c  */
-#line 891 "program_parse.y"
+#line 892 "program_parse.y"
     {
 	   struct asm_symbol *const s = (struct asm_symbol *)
 	      _mesa_symbol_table_find_symbol(state->st, 0, (yyvsp[(1) - (1)].string));
@@ -2934,7 +2935,7 @@ yyreduce:
   case 69:
 
 /* Line 1455 of yacc.c  */
-#line 912 "program_parse.y"
+#line 913 "program_parse.y"
     {
 	   init_src_reg(& (yyval.src_reg));
 	   (yyval.src_reg).Base.Index = (yyvsp[(1) - (1)].integer);
@@ -2944,7 +2945,7 @@ yyreduce:
   case 70:
 
 /* Line 1455 of yacc.c  */
-#line 919 "program_parse.y"
+#line 920 "program_parse.y"
     {
 	   /* FINISHME: Add support for multiple address registers.
 	    */
@@ -2959,30 +2960,30 @@ yyreduce:
   case 71:
 
 /* Line 1455 of yacc.c  */
-#line 930 "program_parse.y"
+#line 931 "program_parse.y"
     { (yyval.integer) = 0; ;}
     break;
 
   case 72:
 
 /* Line 1455 of yacc.c  */
-#line 931 "program_parse.y"
+#line 932 "program_parse.y"
     { (yyval.integer) = (yyvsp[(2) - (2)].integer); ;}
     break;
 
   case 73:
 
 /* Line 1455 of yacc.c  */
-#line 932 "program_parse.y"
+#line 933 "program_parse.y"
     { (yyval.integer) = -(yyvsp[(2) - (2)].integer); ;}
     break;
 
   case 74:
 
 /* Line 1455 of yacc.c  */
-#line 936 "program_parse.y"
+#line 937 "program_parse.y"
     {
-	   if (((yyvsp[(1) - (1)].integer) < 0) || ((yyvsp[(1) - (1)].integer) > 63)) {
+	   if (((yyvsp[(1) - (1)].integer) < 0) || ((yyvsp[(1) - (1)].integer) > 4095)) {
               char s[100];
               _mesa_snprintf(s, sizeof(s),
                              "relative address offset too large (%d)", (yyvsp[(1) - (1)].integer));
@@ -2997,9 +2998,9 @@ yyreduce:
   case 75:
 
 /* Line 1455 of yacc.c  */
-#line 950 "program_parse.y"
+#line 951 "program_parse.y"
     {
-	   if (((yyvsp[(1) - (1)].integer) < 0) || ((yyvsp[(1) - (1)].integer) > 64)) {
+	   if (((yyvsp[(1) - (1)].integer) < 0) || ((yyvsp[(1) - (1)].integer) > 4096)) {
               char s[100];
               _mesa_snprintf(s, sizeof(s),
                              "relative address offset too large (%d)", (yyvsp[(1) - (1)].integer));
@@ -3014,7 +3015,7 @@ yyreduce:
   case 76:
 
 /* Line 1455 of yacc.c  */
-#line 964 "program_parse.y"
+#line 965 "program_parse.y"
     {
 	   struct asm_symbol *const s = (struct asm_symbol *)
 	      _mesa_symbol_table_find_symbol(state->st, 0, (yyvsp[(1) - (1)].string));
@@ -3037,7 +3038,7 @@ yyreduce:
   case 77:
 
 /* Line 1455 of yacc.c  */
-#line 984 "program_parse.y"
+#line 985 "program_parse.y"
     {
 	   if ((yyvsp[(1) - (1)].swiz_mask).mask != WRITEMASK_X) {
 	      yyerror(& (yylsp[(1) - (1)]), state, "invalid address component selector");
@@ -3051,7 +3052,7 @@ yyreduce:
   case 78:
 
 /* Line 1455 of yacc.c  */
-#line 995 "program_parse.y"
+#line 996 "program_parse.y"
     {
 	   if ((yyvsp[(1) - (1)].swiz_mask).mask != WRITEMASK_X) {
 	      yyerror(& (yylsp[(1) - (1)]), state,
@@ -3066,21 +3067,21 @@ yyreduce:
   case 83:
 
 /* Line 1455 of yacc.c  */
-#line 1011 "program_parse.y"
+#line 1012 "program_parse.y"
     { (yyval.swiz_mask).swizzle = SWIZZLE_NOOP; (yyval.swiz_mask).mask = WRITEMASK_XYZW; ;}
     break;
 
   case 88:
 
 /* Line 1455 of yacc.c  */
-#line 1015 "program_parse.y"
+#line 1016 "program_parse.y"
     { (yyval.swiz_mask).swizzle = SWIZZLE_NOOP; (yyval.swiz_mask).mask = WRITEMASK_XYZW; ;}
     break;
 
   case 89:
 
 /* Line 1455 of yacc.c  */
-#line 1019 "program_parse.y"
+#line 1020 "program_parse.y"
     {
 	   (yyval.dst_reg) = (yyvsp[(2) - (3)].dst_reg);
 	;}
@@ -3089,7 +3090,7 @@ yyreduce:
   case 90:
 
 /* Line 1455 of yacc.c  */
-#line 1023 "program_parse.y"
+#line 1024 "program_parse.y"
     {
 	   (yyval.dst_reg) = (yyvsp[(2) - (3)].dst_reg);
 	;}
@@ -3098,7 +3099,7 @@ yyreduce:
   case 91:
 
 /* Line 1455 of yacc.c  */
-#line 1027 "program_parse.y"
+#line 1028 "program_parse.y"
     {
 	   (yyval.dst_reg).CondMask = COND_TR;
 	   (yyval.dst_reg).CondSwizzle = SWIZZLE_NOOP;
@@ -3109,7 +3110,7 @@ yyreduce:
   case 92:
 
 /* Line 1455 of yacc.c  */
-#line 1035 "program_parse.y"
+#line 1036 "program_parse.y"
     {
 	   (yyval.dst_reg) = (yyvsp[(1) - (2)].dst_reg);
 	   (yyval.dst_reg).CondSwizzle = (yyvsp[(2) - (2)].swiz_mask).swizzle;
@@ -3119,7 +3120,7 @@ yyreduce:
   case 93:
 
 /* Line 1455 of yacc.c  */
-#line 1042 "program_parse.y"
+#line 1043 "program_parse.y"
     {
 	   (yyval.dst_reg) = (yyvsp[(1) - (2)].dst_reg);
 	   (yyval.dst_reg).CondSwizzle = (yyvsp[(2) - (2)].swiz_mask).swizzle;
@@ -3129,7 +3130,7 @@ yyreduce:
   case 94:
 
 /* Line 1455 of yacc.c  */
-#line 1049 "program_parse.y"
+#line 1050 "program_parse.y"
     {
 	   const int cond = _mesa_parse_cc((yyvsp[(1) - (1)].string));
 	   if ((cond == 0) || ((yyvsp[(1) - (1)].string)[2] != '\0')) {
@@ -3155,7 +3156,7 @@ yyreduce:
   case 95:
 
 /* Line 1455 of yacc.c  */
-#line 1072 "program_parse.y"
+#line 1073 "program_parse.y"
     {
 	   const int cond = _mesa_parse_cc((yyvsp[(1) - (1)].string));
 	   if ((cond == 0) || ((yyvsp[(1) - (1)].string)[2] != '\0')) {
@@ -3181,7 +3182,7 @@ yyreduce:
   case 102:
 
 /* Line 1455 of yacc.c  */
-#line 1103 "program_parse.y"
+#line 1104 "program_parse.y"
     {
 	   struct asm_symbol *const s =
 	      declare_variable(state, (yyvsp[(2) - (4)].string), at_attrib, & (yylsp[(2) - (4)]));
@@ -3203,7 +3204,7 @@ yyreduce:
   case 103:
 
 /* Line 1455 of yacc.c  */
-#line 1122 "program_parse.y"
+#line 1123 "program_parse.y"
     {
 	   (yyval.attrib) = (yyvsp[(2) - (2)].attrib);
 	;}
@@ -3212,7 +3213,7 @@ yyreduce:
   case 104:
 
 /* Line 1455 of yacc.c  */
-#line 1126 "program_parse.y"
+#line 1127 "program_parse.y"
     {
 	   (yyval.attrib) = (yyvsp[(2) - (2)].attrib);
 	;}
@@ -3221,7 +3222,7 @@ yyreduce:
   case 105:
 
 /* Line 1455 of yacc.c  */
-#line 1132 "program_parse.y"
+#line 1133 "program_parse.y"
     {
 	   (yyval.attrib) = VERT_ATTRIB_POS;
 	;}
@@ -3230,7 +3231,7 @@ yyreduce:
   case 106:
 
 /* Line 1455 of yacc.c  */
-#line 1136 "program_parse.y"
+#line 1137 "program_parse.y"
     {
 	   (yyval.attrib) = VERT_ATTRIB_WEIGHT;
 	;}
@@ -3239,7 +3240,7 @@ yyreduce:
   case 107:
 
 /* Line 1455 of yacc.c  */
-#line 1140 "program_parse.y"
+#line 1141 "program_parse.y"
     {
 	   (yyval.attrib) = VERT_ATTRIB_NORMAL;
 	;}
@@ -3248,7 +3249,7 @@ yyreduce:
   case 108:
 
 /* Line 1455 of yacc.c  */
-#line 1144 "program_parse.y"
+#line 1145 "program_parse.y"
     {
 	   if (!state->ctx->Extensions.EXT_secondary_color) {
 	      yyerror(& (yylsp[(2) - (2)]), state, "GL_EXT_secondary_color not supported");
@@ -3262,7 +3263,7 @@ yyreduce:
   case 109:
 
 /* Line 1455 of yacc.c  */
-#line 1153 "program_parse.y"
+#line 1154 "program_parse.y"
     {
 	   if (!state->ctx->Extensions.EXT_fog_coord) {
 	      yyerror(& (yylsp[(1) - (1)]), state, "GL_EXT_fog_coord not supported");
@@ -3276,7 +3277,7 @@ yyreduce:
   case 110:
 
 /* Line 1455 of yacc.c  */
-#line 1162 "program_parse.y"
+#line 1163 "program_parse.y"
     {
 	   (yyval.attrib) = VERT_ATTRIB_TEX0 + (yyvsp[(2) - (2)].integer);
 	;}
@@ -3285,7 +3286,7 @@ yyreduce:
   case 111:
 
 /* Line 1455 of yacc.c  */
-#line 1166 "program_parse.y"
+#line 1167 "program_parse.y"
     {
 	   yyerror(& (yylsp[(1) - (4)]), state, "GL_ARB_matrix_palette not supported");
 	   YYERROR;
@@ -3295,7 +3296,7 @@ yyreduce:
   case 112:
 
 /* Line 1455 of yacc.c  */
-#line 1171 "program_parse.y"
+#line 1172 "program_parse.y"
     {
 	   (yyval.attrib) = VERT_ATTRIB_GENERIC0 + (yyvsp[(3) - (4)].integer);
 	;}
@@ -3304,7 +3305,7 @@ yyreduce:
   case 113:
 
 /* Line 1455 of yacc.c  */
-#line 1177 "program_parse.y"
+#line 1178 "program_parse.y"
     {
 	   if ((unsigned) (yyvsp[(1) - (1)].integer) >= state->limits->MaxAttribs) {
 	      yyerror(& (yylsp[(1) - (1)]), state, "invalid vertex attribute reference");
@@ -3318,7 +3319,7 @@ yyreduce:
   case 117:
 
 /* Line 1455 of yacc.c  */
-#line 1191 "program_parse.y"
+#line 1192 "program_parse.y"
     {
 	   (yyval.attrib) = FRAG_ATTRIB_WPOS;
 	;}
@@ -3327,7 +3328,7 @@ yyreduce:
   case 118:
 
 /* Line 1455 of yacc.c  */
-#line 1195 "program_parse.y"
+#line 1196 "program_parse.y"
     {
 	   (yyval.attrib) = FRAG_ATTRIB_COL0 + (yyvsp[(2) - (2)].integer);
 	;}
@@ -3336,7 +3337,7 @@ yyreduce:
   case 119:
 
 /* Line 1455 of yacc.c  */
-#line 1199 "program_parse.y"
+#line 1200 "program_parse.y"
     {
 	   (yyval.attrib) = FRAG_ATTRIB_FOGC;
 	;}
@@ -3345,7 +3346,7 @@ yyreduce:
   case 120:
 
 /* Line 1455 of yacc.c  */
-#line 1203 "program_parse.y"
+#line 1204 "program_parse.y"
     {
 	   (yyval.attrib) = FRAG_ATTRIB_TEX0 + (yyvsp[(2) - (2)].integer);
 	;}
@@ -3354,7 +3355,7 @@ yyreduce:
   case 123:
 
 /* Line 1455 of yacc.c  */
-#line 1211 "program_parse.y"
+#line 1212 "program_parse.y"
     {
 	   struct asm_symbol *const s =
 	      declare_variable(state, (yyvsp[(2) - (3)].string), at_param, & (yylsp[(2) - (3)]));
@@ -3375,7 +3376,7 @@ yyreduce:
   case 124:
 
 /* Line 1455 of yacc.c  */
-#line 1229 "program_parse.y"
+#line 1230 "program_parse.y"
     {
 	   if (((yyvsp[(4) - (6)].integer) != 0) && ((unsigned) (yyvsp[(4) - (6)].integer) != (yyvsp[(6) - (6)].temp_sym).param_binding_length)) {
 	      free((yyvsp[(2) - (6)].string));
@@ -3403,7 +3404,7 @@ yyreduce:
   case 125:
 
 /* Line 1455 of yacc.c  */
-#line 1254 "program_parse.y"
+#line 1255 "program_parse.y"
     {
 	   (yyval.integer) = 0;
 	;}
@@ -3412,7 +3413,7 @@ yyreduce:
   case 126:
 
 /* Line 1455 of yacc.c  */
-#line 1258 "program_parse.y"
+#line 1259 "program_parse.y"
     {
 	   if (((yyvsp[(1) - (1)].integer) < 1) || ((unsigned) (yyvsp[(1) - (1)].integer) > state->limits->MaxParameters)) {
 	      yyerror(& (yylsp[(1) - (1)]), state, "invalid parameter array size");
@@ -3426,7 +3427,7 @@ yyreduce:
   case 127:
 
 /* Line 1455 of yacc.c  */
-#line 1269 "program_parse.y"
+#line 1270 "program_parse.y"
     {
 	   (yyval.temp_sym) = (yyvsp[(2) - (2)].temp_sym);
 	;}
@@ -3435,7 +3436,7 @@ yyreduce:
   case 128:
 
 /* Line 1455 of yacc.c  */
-#line 1275 "program_parse.y"
+#line 1276 "program_parse.y"
     {
 	   (yyval.temp_sym) = (yyvsp[(3) - (4)].temp_sym);
 	;}
@@ -3444,7 +3445,7 @@ yyreduce:
   case 130:
 
 /* Line 1455 of yacc.c  */
-#line 1282 "program_parse.y"
+#line 1283 "program_parse.y"
     {
 	   (yyvsp[(1) - (3)].temp_sym).param_binding_length += (yyvsp[(3) - (3)].temp_sym).param_binding_length;
 	   (yyval.temp_sym) = (yyvsp[(1) - (3)].temp_sym);
@@ -3454,7 +3455,7 @@ yyreduce:
   case 131:
 
 /* Line 1455 of yacc.c  */
-#line 1289 "program_parse.y"
+#line 1290 "program_parse.y"
     {
 	   memset(& (yyval.temp_sym), 0, sizeof((yyval.temp_sym)));
 	   (yyval.temp_sym).param_binding_begin = ~0;
@@ -3465,7 +3466,7 @@ yyreduce:
   case 132:
 
 /* Line 1455 of yacc.c  */
-#line 1295 "program_parse.y"
+#line 1296 "program_parse.y"
     {
 	   memset(& (yyval.temp_sym), 0, sizeof((yyval.temp_sym)));
 	   (yyval.temp_sym).param_binding_begin = ~0;
@@ -3476,7 +3477,7 @@ yyreduce:
   case 133:
 
 /* Line 1455 of yacc.c  */
-#line 1301 "program_parse.y"
+#line 1302 "program_parse.y"
     {
 	   memset(& (yyval.temp_sym), 0, sizeof((yyval.temp_sym)));
 	   (yyval.temp_sym).param_binding_begin = ~0;
@@ -3487,7 +3488,7 @@ yyreduce:
   case 134:
 
 /* Line 1455 of yacc.c  */
-#line 1309 "program_parse.y"
+#line 1310 "program_parse.y"
     {
 	   memset(& (yyval.temp_sym), 0, sizeof((yyval.temp_sym)));
 	   (yyval.temp_sym).param_binding_begin = ~0;
@@ -3498,7 +3499,7 @@ yyreduce:
   case 135:
 
 /* Line 1455 of yacc.c  */
-#line 1315 "program_parse.y"
+#line 1316 "program_parse.y"
     {
 	   memset(& (yyval.temp_sym), 0, sizeof((yyval.temp_sym)));
 	   (yyval.temp_sym).param_binding_begin = ~0;
@@ -3509,7 +3510,7 @@ yyreduce:
   case 136:
 
 /* Line 1455 of yacc.c  */
-#line 1321 "program_parse.y"
+#line 1322 "program_parse.y"
     {
 	   memset(& (yyval.temp_sym), 0, sizeof((yyval.temp_sym)));
 	   (yyval.temp_sym).param_binding_begin = ~0;
@@ -3520,7 +3521,7 @@ yyreduce:
   case 137:
 
 /* Line 1455 of yacc.c  */
-#line 1329 "program_parse.y"
+#line 1330 "program_parse.y"
     {
 	   memset(& (yyval.temp_sym), 0, sizeof((yyval.temp_sym)));
 	   (yyval.temp_sym).param_binding_begin = ~0;
@@ -3531,7 +3532,7 @@ yyreduce:
   case 138:
 
 /* Line 1455 of yacc.c  */
-#line 1335 "program_parse.y"
+#line 1336 "program_parse.y"
     {
 	   memset(& (yyval.temp_sym), 0, sizeof((yyval.temp_sym)));
 	   (yyval.temp_sym).param_binding_begin = ~0;
@@ -3542,7 +3543,7 @@ yyreduce:
   case 139:
 
 /* Line 1455 of yacc.c  */
-#line 1341 "program_parse.y"
+#line 1342 "program_parse.y"
     {
 	   memset(& (yyval.temp_sym), 0, sizeof((yyval.temp_sym)));
 	   (yyval.temp_sym).param_binding_begin = ~0;
@@ -3553,98 +3554,98 @@ yyreduce:
   case 140:
 
 /* Line 1455 of yacc.c  */
-#line 1348 "program_parse.y"
+#line 1349 "program_parse.y"
     { memcpy((yyval.state), (yyvsp[(1) - (1)].state), sizeof((yyval.state))); ;}
     break;
 
   case 141:
 
 /* Line 1455 of yacc.c  */
-#line 1349 "program_parse.y"
+#line 1350 "program_parse.y"
     { memcpy((yyval.state), (yyvsp[(2) - (2)].state), sizeof((yyval.state))); ;}
     break;
 
   case 142:
 
 /* Line 1455 of yacc.c  */
-#line 1352 "program_parse.y"
+#line 1353 "program_parse.y"
     { memcpy((yyval.state), (yyvsp[(2) - (2)].state), sizeof((yyval.state))); ;}
     break;
 
   case 143:
 
 /* Line 1455 of yacc.c  */
-#line 1353 "program_parse.y"
+#line 1354 "program_parse.y"
     { memcpy((yyval.state), (yyvsp[(2) - (2)].state), sizeof((yyval.state))); ;}
     break;
 
   case 144:
 
 /* Line 1455 of yacc.c  */
-#line 1354 "program_parse.y"
+#line 1355 "program_parse.y"
     { memcpy((yyval.state), (yyvsp[(2) - (2)].state), sizeof((yyval.state))); ;}
     break;
 
   case 145:
 
 /* Line 1455 of yacc.c  */
-#line 1355 "program_parse.y"
+#line 1356 "program_parse.y"
     { memcpy((yyval.state), (yyvsp[(2) - (2)].state), sizeof((yyval.state))); ;}
     break;
 
   case 146:
 
 /* Line 1455 of yacc.c  */
-#line 1356 "program_parse.y"
+#line 1357 "program_parse.y"
     { memcpy((yyval.state), (yyvsp[(2) - (2)].state), sizeof((yyval.state))); ;}
     break;
 
   case 147:
 
 /* Line 1455 of yacc.c  */
-#line 1357 "program_parse.y"
+#line 1358 "program_parse.y"
     { memcpy((yyval.state), (yyvsp[(2) - (2)].state), sizeof((yyval.state))); ;}
     break;
 
   case 148:
 
 /* Line 1455 of yacc.c  */
-#line 1358 "program_parse.y"
+#line 1359 "program_parse.y"
     { memcpy((yyval.state), (yyvsp[(2) - (2)].state), sizeof((yyval.state))); ;}
     break;
 
   case 149:
 
 /* Line 1455 of yacc.c  */
-#line 1359 "program_parse.y"
+#line 1360 "program_parse.y"
     { memcpy((yyval.state), (yyvsp[(2) - (2)].state), sizeof((yyval.state))); ;}
     break;
 
   case 150:
 
 /* Line 1455 of yacc.c  */
-#line 1360 "program_parse.y"
+#line 1361 "program_parse.y"
     { memcpy((yyval.state), (yyvsp[(2) - (2)].state), sizeof((yyval.state))); ;}
     break;
 
   case 151:
 
 /* Line 1455 of yacc.c  */
-#line 1361 "program_parse.y"
+#line 1362 "program_parse.y"
     { memcpy((yyval.state), (yyvsp[(2) - (2)].state), sizeof((yyval.state))); ;}
     break;
 
   case 152:
 
 /* Line 1455 of yacc.c  */
-#line 1362 "program_parse.y"
+#line 1363 "program_parse.y"
     { memcpy((yyval.state), (yyvsp[(2) - (2)].state), sizeof((yyval.state))); ;}
     break;
 
   case 153:
 
 /* Line 1455 of yacc.c  */
-#line 1366 "program_parse.y"
+#line 1367 "program_parse.y"
     {
 	   memset((yyval.state), 0, sizeof((yyval.state)));
 	   (yyval.state)[0] = STATE_MATERIAL;
@@ -3656,7 +3657,7 @@ yyreduce:
   case 154:
 
 /* Line 1455 of yacc.c  */
-#line 1375 "program_parse.y"
+#line 1376 "program_parse.y"
     {
 	   (yyval.integer) = (yyvsp[(1) - (1)].integer);
 	;}
@@ -3665,7 +3666,7 @@ yyreduce:
   case 155:
 
 /* Line 1455 of yacc.c  */
-#line 1379 "program_parse.y"
+#line 1380 "program_parse.y"
     {
 	   (yyval.integer) = STATE_EMISSION;
 	;}
@@ -3674,7 +3675,7 @@ yyreduce:
   case 156:
 
 /* Line 1455 of yacc.c  */
-#line 1383 "program_parse.y"
+#line 1384 "program_parse.y"
     {
 	   (yyval.integer) = STATE_SHININESS;
 	;}
@@ -3683,7 +3684,7 @@ yyreduce:
   case 157:
 
 /* Line 1455 of yacc.c  */
-#line 1389 "program_parse.y"
+#line 1390 "program_parse.y"
     {
 	   memset((yyval.state), 0, sizeof((yyval.state)));
 	   (yyval.state)[0] = STATE_LIGHT;
@@ -3695,7 +3696,7 @@ yyreduce:
   case 158:
 
 /* Line 1455 of yacc.c  */
-#line 1398 "program_parse.y"
+#line 1399 "program_parse.y"
     {
 	   (yyval.integer) = (yyvsp[(1) - (1)].integer);
 	;}
@@ -3704,7 +3705,7 @@ yyreduce:
   case 159:
 
 /* Line 1455 of yacc.c  */
-#line 1402 "program_parse.y"
+#line 1403 "program_parse.y"
     {
 	   (yyval.integer) = STATE_POSITION;
 	;}
@@ -3713,7 +3714,7 @@ yyreduce:
   case 160:
 
 /* Line 1455 of yacc.c  */
-#line 1406 "program_parse.y"
+#line 1407 "program_parse.y"
     {
 	   if (!state->ctx->Extensions.EXT_point_parameters) {
 	      yyerror(& (yylsp[(1) - (1)]), state, "GL_ARB_point_parameters not supported");
@@ -3727,7 +3728,7 @@ yyreduce:
   case 161:
 
 /* Line 1455 of yacc.c  */
-#line 1415 "program_parse.y"
+#line 1416 "program_parse.y"
     {
 	   (yyval.integer) = (yyvsp[(2) - (2)].integer);
 	;}
@@ -3736,7 +3737,7 @@ yyreduce:
   case 162:
 
 /* Line 1455 of yacc.c  */
-#line 1419 "program_parse.y"
+#line 1420 "program_parse.y"
     {
 	   (yyval.integer) = STATE_HALF_VECTOR;
 	;}
@@ -3745,7 +3746,7 @@ yyreduce:
   case 163:
 
 /* Line 1455 of yacc.c  */
-#line 1425 "program_parse.y"
+#line 1426 "program_parse.y"
     {
 	   (yyval.integer) = STATE_SPOT_DIRECTION;
 	;}
@@ -3754,7 +3755,7 @@ yyreduce:
   case 164:
 
 /* Line 1455 of yacc.c  */
-#line 1431 "program_parse.y"
+#line 1432 "program_parse.y"
     {
 	   (yyval.state)[0] = (yyvsp[(2) - (2)].state)[0];
 	   (yyval.state)[1] = (yyvsp[(2) - (2)].state)[1];
@@ -3764,7 +3765,7 @@ yyreduce:
   case 165:
 
 /* Line 1455 of yacc.c  */
-#line 1438 "program_parse.y"
+#line 1439 "program_parse.y"
     {
 	   memset((yyval.state), 0, sizeof((yyval.state)));
 	   (yyval.state)[0] = STATE_LIGHTMODEL_AMBIENT;
@@ -3774,7 +3775,7 @@ yyreduce:
   case 166:
 
 /* Line 1455 of yacc.c  */
-#line 1443 "program_parse.y"
+#line 1444 "program_parse.y"
     {
 	   memset((yyval.state), 0, sizeof((yyval.state)));
 	   (yyval.state)[0] = STATE_LIGHTMODEL_SCENECOLOR;
@@ -3785,7 +3786,7 @@ yyreduce:
   case 167:
 
 /* Line 1455 of yacc.c  */
-#line 1451 "program_parse.y"
+#line 1452 "program_parse.y"
     {
 	   memset((yyval.state), 0, sizeof((yyval.state)));
 	   (yyval.state)[0] = STATE_LIGHTPROD;
@@ -3798,7 +3799,7 @@ yyreduce:
   case 169:
 
 /* Line 1455 of yacc.c  */
-#line 1463 "program_parse.y"
+#line 1464 "program_parse.y"
     {
 	   memset((yyval.state), 0, sizeof((yyval.state)));
 	   (yyval.state)[0] = (yyvsp[(3) - (3)].integer);
@@ -3809,7 +3810,7 @@ yyreduce:
   case 170:
 
 /* Line 1455 of yacc.c  */
-#line 1471 "program_parse.y"
+#line 1472 "program_parse.y"
     {
 	   (yyval.integer) = STATE_TEXENV_COLOR;
 	;}
@@ -3818,7 +3819,7 @@ yyreduce:
   case 171:
 
 /* Line 1455 of yacc.c  */
-#line 1477 "program_parse.y"
+#line 1478 "program_parse.y"
     {
 	   (yyval.integer) = STATE_AMBIENT;
 	;}
@@ -3827,7 +3828,7 @@ yyreduce:
   case 172:
 
 /* Line 1455 of yacc.c  */
-#line 1481 "program_parse.y"
+#line 1482 "program_parse.y"
     {
 	   (yyval.integer) = STATE_DIFFUSE;
 	;}
@@ -3836,7 +3837,7 @@ yyreduce:
   case 173:
 
 /* Line 1455 of yacc.c  */
-#line 1485 "program_parse.y"
+#line 1486 "program_parse.y"
     {
 	   (yyval.integer) = STATE_SPECULAR;
 	;}
@@ -3845,7 +3846,7 @@ yyreduce:
   case 174:
 
 /* Line 1455 of yacc.c  */
-#line 1491 "program_parse.y"
+#line 1492 "program_parse.y"
     {
 	   if ((unsigned) (yyvsp[(1) - (1)].integer) >= state->MaxLights) {
 	      yyerror(& (yylsp[(1) - (1)]), state, "invalid light selector");
@@ -3859,7 +3860,7 @@ yyreduce:
   case 175:
 
 /* Line 1455 of yacc.c  */
-#line 1502 "program_parse.y"
+#line 1503 "program_parse.y"
     {
 	   memset((yyval.state), 0, sizeof((yyval.state)));
 	   (yyval.state)[0] = STATE_TEXGEN;
@@ -3871,7 +3872,7 @@ yyreduce:
   case 176:
 
 /* Line 1455 of yacc.c  */
-#line 1511 "program_parse.y"
+#line 1512 "program_parse.y"
     {
 	   (yyval.integer) = STATE_TEXGEN_EYE_S;
 	;}
@@ -3880,7 +3881,7 @@ yyreduce:
   case 177:
 
 /* Line 1455 of yacc.c  */
-#line 1515 "program_parse.y"
+#line 1516 "program_parse.y"
     {
 	   (yyval.integer) = STATE_TEXGEN_OBJECT_S;
 	;}
@@ -3889,7 +3890,7 @@ yyreduce:
   case 178:
 
 /* Line 1455 of yacc.c  */
-#line 1520 "program_parse.y"
+#line 1521 "program_parse.y"
     {
 	   (yyval.integer) = STATE_TEXGEN_EYE_S - STATE_TEXGEN_EYE_S;
 	;}
@@ -3898,7 +3899,7 @@ yyreduce:
   case 179:
 
 /* Line 1455 of yacc.c  */
-#line 1524 "program_parse.y"
+#line 1525 "program_parse.y"
     {
 	   (yyval.integer) = STATE_TEXGEN_EYE_T - STATE_TEXGEN_EYE_S;
 	;}
@@ -3907,7 +3908,7 @@ yyreduce:
   case 180:
 
 /* Line 1455 of yacc.c  */
-#line 1528 "program_parse.y"
+#line 1529 "program_parse.y"
     {
 	   (yyval.integer) = STATE_TEXGEN_EYE_R - STATE_TEXGEN_EYE_S;
 	;}
@@ -3916,7 +3917,7 @@ yyreduce:
   case 181:
 
 /* Line 1455 of yacc.c  */
-#line 1532 "program_parse.y"
+#line 1533 "program_parse.y"
     {
 	   (yyval.integer) = STATE_TEXGEN_EYE_Q - STATE_TEXGEN_EYE_S;
 	;}
@@ -3925,7 +3926,7 @@ yyreduce:
   case 182:
 
 /* Line 1455 of yacc.c  */
-#line 1538 "program_parse.y"
+#line 1539 "program_parse.y"
     {
 	   memset((yyval.state), 0, sizeof((yyval.state)));
 	   (yyval.state)[0] = (yyvsp[(2) - (2)].integer);
@@ -3935,7 +3936,7 @@ yyreduce:
   case 183:
 
 /* Line 1455 of yacc.c  */
-#line 1545 "program_parse.y"
+#line 1546 "program_parse.y"
     {
 	   (yyval.integer) = STATE_FOG_COLOR;
 	;}
@@ -3944,7 +3945,7 @@ yyreduce:
   case 184:
 
 /* Line 1455 of yacc.c  */
-#line 1549 "program_parse.y"
+#line 1550 "program_parse.y"
     {
 	   (yyval.integer) = STATE_FOG_PARAMS;
 	;}
@@ -3953,7 +3954,7 @@ yyreduce:
   case 185:
 
 /* Line 1455 of yacc.c  */
-#line 1555 "program_parse.y"
+#line 1556 "program_parse.y"
     {
 	   memset((yyval.state), 0, sizeof((yyval.state)));
 	   (yyval.state)[0] = STATE_CLIPPLANE;
@@ -3964,7 +3965,7 @@ yyreduce:
   case 186:
 
 /* Line 1455 of yacc.c  */
-#line 1563 "program_parse.y"
+#line 1564 "program_parse.y"
     {
 	   if ((unsigned) (yyvsp[(1) - (1)].integer) >= state->MaxClipPlanes) {
 	      yyerror(& (yylsp[(1) - (1)]), state, "invalid clip plane selector");
@@ -3978,7 +3979,7 @@ yyreduce:
   case 187:
 
 /* Line 1455 of yacc.c  */
-#line 1574 "program_parse.y"
+#line 1575 "program_parse.y"
     {
 	   memset((yyval.state), 0, sizeof((yyval.state)));
 	   (yyval.state)[0] = (yyvsp[(2) - (2)].integer);
@@ -3988,7 +3989,7 @@ yyreduce:
   case 188:
 
 /* Line 1455 of yacc.c  */
-#line 1581 "program_parse.y"
+#line 1582 "program_parse.y"
     {
 	   (yyval.integer) = STATE_POINT_SIZE;
 	;}
@@ -3997,7 +3998,7 @@ yyreduce:
   case 189:
 
 /* Line 1455 of yacc.c  */
-#line 1585 "program_parse.y"
+#line 1586 "program_parse.y"
     {
 	   (yyval.integer) = STATE_POINT_ATTENUATION;
 	;}
@@ -4006,7 +4007,7 @@ yyreduce:
   case 190:
 
 /* Line 1455 of yacc.c  */
-#line 1591 "program_parse.y"
+#line 1592 "program_parse.y"
     {
 	   (yyval.state)[0] = (yyvsp[(1) - (5)].state)[0];
 	   (yyval.state)[1] = (yyvsp[(1) - (5)].state)[1];
@@ -4019,7 +4020,7 @@ yyreduce:
   case 191:
 
 /* Line 1455 of yacc.c  */
-#line 1601 "program_parse.y"
+#line 1602 "program_parse.y"
     {
 	   (yyval.state)[0] = (yyvsp[(1) - (2)].state)[0];
 	   (yyval.state)[1] = (yyvsp[(1) - (2)].state)[1];
@@ -4032,7 +4033,7 @@ yyreduce:
   case 192:
 
 /* Line 1455 of yacc.c  */
-#line 1611 "program_parse.y"
+#line 1612 "program_parse.y"
     {
 	   (yyval.state)[2] = 0;
 	   (yyval.state)[3] = 3;
@@ -4042,7 +4043,7 @@ yyreduce:
   case 193:
 
 /* Line 1455 of yacc.c  */
-#line 1616 "program_parse.y"
+#line 1617 "program_parse.y"
     {
 	   /* It seems logical that the matrix row range specifier would have
 	    * to specify a range or more than one row (i.e., $5 > $3).
@@ -4063,7 +4064,7 @@ yyreduce:
   case 194:
 
 /* Line 1455 of yacc.c  */
-#line 1634 "program_parse.y"
+#line 1635 "program_parse.y"
     {
 	   (yyval.state)[0] = (yyvsp[(2) - (3)].state)[0];
 	   (yyval.state)[1] = (yyvsp[(2) - (3)].state)[1];
@@ -4074,7 +4075,7 @@ yyreduce:
   case 195:
 
 /* Line 1455 of yacc.c  */
-#line 1642 "program_parse.y"
+#line 1643 "program_parse.y"
     {
 	   (yyval.integer) = 0;
 	;}
@@ -4083,7 +4084,7 @@ yyreduce:
   case 196:
 
 /* Line 1455 of yacc.c  */
-#line 1646 "program_parse.y"
+#line 1647 "program_parse.y"
     {
 	   (yyval.integer) = (yyvsp[(1) - (1)].integer);
 	;}
@@ -4092,7 +4093,7 @@ yyreduce:
   case 197:
 
 /* Line 1455 of yacc.c  */
-#line 1652 "program_parse.y"
+#line 1653 "program_parse.y"
     {
 	   (yyval.integer) = STATE_MATRIX_INVERSE;
 	;}
@@ -4101,7 +4102,7 @@ yyreduce:
   case 198:
 
 /* Line 1455 of yacc.c  */
-#line 1656 "program_parse.y"
+#line 1657 "program_parse.y"
     {
 	   (yyval.integer) = STATE_MATRIX_TRANSPOSE;
 	;}
@@ -4110,7 +4111,7 @@ yyreduce:
   case 199:
 
 /* Line 1455 of yacc.c  */
-#line 1660 "program_parse.y"
+#line 1661 "program_parse.y"
     {
 	   (yyval.integer) = STATE_MATRIX_INVTRANS;
 	;}
@@ -4119,7 +4120,7 @@ yyreduce:
   case 200:
 
 /* Line 1455 of yacc.c  */
-#line 1666 "program_parse.y"
+#line 1667 "program_parse.y"
     {
 	   if ((yyvsp[(1) - (1)].integer) > 3) {
 	      yyerror(& (yylsp[(1) - (1)]), state, "invalid matrix row reference");
@@ -4133,7 +4134,7 @@ yyreduce:
   case 201:
 
 /* Line 1455 of yacc.c  */
-#line 1677 "program_parse.y"
+#line 1678 "program_parse.y"
     {
 	   (yyval.state)[0] = STATE_MODELVIEW_MATRIX;
 	   (yyval.state)[1] = (yyvsp[(2) - (2)].integer);
@@ -4143,7 +4144,7 @@ yyreduce:
   case 202:
 
 /* Line 1455 of yacc.c  */
-#line 1682 "program_parse.y"
+#line 1683 "program_parse.y"
     {
 	   (yyval.state)[0] = STATE_PROJECTION_MATRIX;
 	   (yyval.state)[1] = 0;
@@ -4153,7 +4154,7 @@ yyreduce:
   case 203:
 
 /* Line 1455 of yacc.c  */
-#line 1687 "program_parse.y"
+#line 1688 "program_parse.y"
     {
 	   (yyval.state)[0] = STATE_MVP_MATRIX;
 	   (yyval.state)[1] = 0;
@@ -4163,7 +4164,7 @@ yyreduce:
   case 204:
 
 /* Line 1455 of yacc.c  */
-#line 1692 "program_parse.y"
+#line 1693 "program_parse.y"
     {
 	   (yyval.state)[0] = STATE_TEXTURE_MATRIX;
 	   (yyval.state)[1] = (yyvsp[(2) - (2)].integer);
@@ -4173,7 +4174,7 @@ yyreduce:
   case 205:
 
 /* Line 1455 of yacc.c  */
-#line 1697 "program_parse.y"
+#line 1698 "program_parse.y"
     {
 	   yyerror(& (yylsp[(1) - (4)]), state, "GL_ARB_matrix_palette not supported");
 	   YYERROR;
@@ -4183,7 +4184,7 @@ yyreduce:
   case 206:
 
 /* Line 1455 of yacc.c  */
-#line 1702 "program_parse.y"
+#line 1703 "program_parse.y"
     {
 	   (yyval.state)[0] = STATE_PROGRAM_MATRIX;
 	   (yyval.state)[1] = (yyvsp[(3) - (4)].integer);
@@ -4193,7 +4194,7 @@ yyreduce:
   case 207:
 
 /* Line 1455 of yacc.c  */
-#line 1709 "program_parse.y"
+#line 1710 "program_parse.y"
     {
 	   (yyval.integer) = 0;
 	;}
@@ -4202,7 +4203,7 @@ yyreduce:
   case 208:
 
 /* Line 1455 of yacc.c  */
-#line 1713 "program_parse.y"
+#line 1714 "program_parse.y"
     {
 	   (yyval.integer) = (yyvsp[(2) - (3)].integer);
 	;}
@@ -4211,7 +4212,7 @@ yyreduce:
   case 209:
 
 /* Line 1455 of yacc.c  */
-#line 1718 "program_parse.y"
+#line 1719 "program_parse.y"
     {
 	   /* Since GL_ARB_vertex_blend isn't supported, only modelview matrix
 	    * zero is valid.
@@ -4228,7 +4229,7 @@ yyreduce:
   case 210:
 
 /* Line 1455 of yacc.c  */
-#line 1731 "program_parse.y"
+#line 1732 "program_parse.y"
     {
 	   /* Since GL_ARB_matrix_palette isn't supported, just let any value
 	    * through here.  The error will be generated later.
@@ -4240,7 +4241,7 @@ yyreduce:
   case 211:
 
 /* Line 1455 of yacc.c  */
-#line 1739 "program_parse.y"
+#line 1740 "program_parse.y"
     {
 	   if ((unsigned) (yyvsp[(1) - (1)].integer) >= state->MaxProgramMatrices) {
 	      yyerror(& (yylsp[(1) - (1)]), state, "invalid program matrix selector");
@@ -4254,7 +4255,7 @@ yyreduce:
   case 212:
 
 /* Line 1455 of yacc.c  */
-#line 1750 "program_parse.y"
+#line 1751 "program_parse.y"
     {
 	   memset((yyval.state), 0, sizeof((yyval.state)));
 	   (yyval.state)[0] = STATE_DEPTH_RANGE;
@@ -4264,7 +4265,7 @@ yyreduce:
   case 217:
 
 /* Line 1455 of yacc.c  */
-#line 1762 "program_parse.y"
+#line 1763 "program_parse.y"
     {
 	   memset((yyval.state), 0, sizeof((yyval.state)));
 	   (yyval.state)[0] = state->state_param_enum;
@@ -4277,7 +4278,7 @@ yyreduce:
   case 218:
 
 /* Line 1455 of yacc.c  */
-#line 1772 "program_parse.y"
+#line 1773 "program_parse.y"
     {
 	   (yyval.state)[0] = (yyvsp[(1) - (1)].integer);
 	   (yyval.state)[1] = (yyvsp[(1) - (1)].integer);
@@ -4287,7 +4288,7 @@ yyreduce:
   case 219:
 
 /* Line 1455 of yacc.c  */
-#line 1777 "program_parse.y"
+#line 1778 "program_parse.y"
     {
 	   (yyval.state)[0] = (yyvsp[(1) - (3)].integer);
 	   (yyval.state)[1] = (yyvsp[(3) - (3)].integer);
@@ -4297,7 +4298,7 @@ yyreduce:
   case 220:
 
 /* Line 1455 of yacc.c  */
-#line 1784 "program_parse.y"
+#line 1785 "program_parse.y"
     {
 	   memset((yyval.state), 0, sizeof((yyval.state)));
 	   (yyval.state)[0] = state->state_param_enum;
@@ -4310,7 +4311,7 @@ yyreduce:
   case 221:
 
 /* Line 1455 of yacc.c  */
-#line 1794 "program_parse.y"
+#line 1795 "program_parse.y"
     {
 	   memset((yyval.state), 0, sizeof((yyval.state)));
 	   (yyval.state)[0] = state->state_param_enum;
@@ -4323,7 +4324,7 @@ yyreduce:
   case 222:
 
 /* Line 1455 of yacc.c  */
-#line 1803 "program_parse.y"
+#line 1804 "program_parse.y"
     {
 	   (yyval.state)[0] = (yyvsp[(1) - (1)].integer);
 	   (yyval.state)[1] = (yyvsp[(1) - (1)].integer);
@@ -4333,7 +4334,7 @@ yyreduce:
   case 223:
 
 /* Line 1455 of yacc.c  */
-#line 1808 "program_parse.y"
+#line 1809 "program_parse.y"
     {
 	   (yyval.state)[0] = (yyvsp[(1) - (3)].integer);
 	   (yyval.state)[1] = (yyvsp[(3) - (3)].integer);
@@ -4343,7 +4344,7 @@ yyreduce:
   case 224:
 
 /* Line 1455 of yacc.c  */
-#line 1815 "program_parse.y"
+#line 1816 "program_parse.y"
     {
 	   memset((yyval.state), 0, sizeof((yyval.state)));
 	   (yyval.state)[0] = state->state_param_enum;
@@ -4356,7 +4357,7 @@ yyreduce:
   case 225:
 
 /* Line 1455 of yacc.c  */
-#line 1825 "program_parse.y"
+#line 1826 "program_parse.y"
     {
 	   if ((unsigned) (yyvsp[(1) - (1)].integer) >= state->limits->MaxEnvParams) {
 	      yyerror(& (yylsp[(1) - (1)]), state, "invalid environment parameter reference");
@@ -4369,7 +4370,7 @@ yyreduce:
   case 226:
 
 /* Line 1455 of yacc.c  */
-#line 1835 "program_parse.y"
+#line 1836 "program_parse.y"
     {
 	   if ((unsigned) (yyvsp[(1) - (1)].integer) >= state->limits->MaxLocalParams) {
 	      yyerror(& (yylsp[(1) - (1)]), state, "invalid local parameter reference");
@@ -4382,7 +4383,7 @@ yyreduce:
   case 231:
 
 /* Line 1455 of yacc.c  */
-#line 1850 "program_parse.y"
+#line 1851 "program_parse.y"
     {
 	   (yyval.vector).count = 4;
 	   (yyval.vector).data[0] = (yyvsp[(1) - (1)].real);
@@ -4395,7 +4396,7 @@ yyreduce:
   case 232:
 
 /* Line 1455 of yacc.c  */
-#line 1860 "program_parse.y"
+#line 1861 "program_parse.y"
     {
 	   (yyval.vector).count = 1;
 	   (yyval.vector).data[0] = (yyvsp[(1) - (1)].real);
@@ -4408,7 +4409,7 @@ yyreduce:
   case 233:
 
 /* Line 1455 of yacc.c  */
-#line 1868 "program_parse.y"
+#line 1869 "program_parse.y"
     {
 	   (yyval.vector).count = 1;
 	   (yyval.vector).data[0] = (float) (yyvsp[(1) - (1)].integer);
@@ -4421,7 +4422,7 @@ yyreduce:
   case 234:
 
 /* Line 1455 of yacc.c  */
-#line 1878 "program_parse.y"
+#line 1879 "program_parse.y"
     {
 	   (yyval.vector).count = 4;
 	   (yyval.vector).data[0] = (yyvsp[(2) - (3)].real);
@@ -4434,7 +4435,7 @@ yyreduce:
   case 235:
 
 /* Line 1455 of yacc.c  */
-#line 1886 "program_parse.y"
+#line 1887 "program_parse.y"
     {
 	   (yyval.vector).count = 4;
 	   (yyval.vector).data[0] = (yyvsp[(2) - (5)].real);
@@ -4447,7 +4448,7 @@ yyreduce:
   case 236:
 
 /* Line 1455 of yacc.c  */
-#line 1895 "program_parse.y"
+#line 1896 "program_parse.y"
     {
 	   (yyval.vector).count = 4;
 	   (yyval.vector).data[0] = (yyvsp[(2) - (7)].real);
@@ -4460,7 +4461,7 @@ yyreduce:
   case 237:
 
 /* Line 1455 of yacc.c  */
-#line 1904 "program_parse.y"
+#line 1905 "program_parse.y"
     {
 	   (yyval.vector).count = 4;
 	   (yyval.vector).data[0] = (yyvsp[(2) - (9)].real);
@@ -4473,7 +4474,7 @@ yyreduce:
   case 238:
 
 /* Line 1455 of yacc.c  */
-#line 1914 "program_parse.y"
+#line 1915 "program_parse.y"
     {
 	   (yyval.real) = ((yyvsp[(1) - (2)].negate)) ? -(yyvsp[(2) - (2)].real) : (yyvsp[(2) - (2)].real);
 	;}
@@ -4482,7 +4483,7 @@ yyreduce:
   case 239:
 
 /* Line 1455 of yacc.c  */
-#line 1918 "program_parse.y"
+#line 1919 "program_parse.y"
     {
 	   (yyval.real) = (float)(((yyvsp[(1) - (2)].negate)) ? -(yyvsp[(2) - (2)].integer) : (yyvsp[(2) - (2)].integer));
 	;}
@@ -4491,35 +4492,35 @@ yyreduce:
   case 240:
 
 /* Line 1455 of yacc.c  */
-#line 1923 "program_parse.y"
+#line 1924 "program_parse.y"
     { (yyval.negate) = FALSE; ;}
     break;
 
   case 241:
 
 /* Line 1455 of yacc.c  */
-#line 1924 "program_parse.y"
+#line 1925 "program_parse.y"
     { (yyval.negate) = TRUE;  ;}
     break;
 
   case 242:
 
 /* Line 1455 of yacc.c  */
-#line 1925 "program_parse.y"
+#line 1926 "program_parse.y"
     { (yyval.negate) = FALSE; ;}
     break;
 
   case 243:
 
 /* Line 1455 of yacc.c  */
-#line 1928 "program_parse.y"
+#line 1929 "program_parse.y"
     { (yyval.integer) = (yyvsp[(2) - (2)].integer); ;}
     break;
 
   case 245:
 
 /* Line 1455 of yacc.c  */
-#line 1932 "program_parse.y"
+#line 1933 "program_parse.y"
     {
 	   /* NV_fragment_program_option defines the size qualifiers in a
 	    * fairly broken way.  "SHORT" or "LONG" can optionally be used
@@ -4558,7 +4559,7 @@ yyreduce:
   case 246:
 
 /* Line 1455 of yacc.c  */
-#line 1966 "program_parse.y"
+#line 1967 "program_parse.y"
     {
 	;}
     break;
@@ -4566,14 +4567,14 @@ yyreduce:
   case 247:
 
 /* Line 1455 of yacc.c  */
-#line 1970 "program_parse.y"
+#line 1971 "program_parse.y"
     { (yyval.integer) = (yyvsp[(1) - (1)].integer); ;}
     break;
 
   case 249:
 
 /* Line 1455 of yacc.c  */
-#line 1974 "program_parse.y"
+#line 1975 "program_parse.y"
     {
 	   if (!declare_variable(state, (yyvsp[(3) - (3)].string), (yyvsp[(0) - (3)].integer), & (yylsp[(3) - (3)]))) {
 	      free((yyvsp[(3) - (3)].string));
@@ -4585,7 +4586,7 @@ yyreduce:
   case 250:
 
 /* Line 1455 of yacc.c  */
-#line 1981 "program_parse.y"
+#line 1982 "program_parse.y"
     {
 	   if (!declare_variable(state, (yyvsp[(1) - (1)].string), (yyvsp[(0) - (1)].integer), & (yylsp[(1) - (1)]))) {
 	      free((yyvsp[(1) - (1)].string));
@@ -4597,7 +4598,7 @@ yyreduce:
   case 251:
 
 /* Line 1455 of yacc.c  */
-#line 1990 "program_parse.y"
+#line 1991 "program_parse.y"
     {
 	   struct asm_symbol *const s =
 	      declare_variable(state, (yyvsp[(3) - (5)].string), at_output, & (yylsp[(3) - (5)]));
@@ -4614,7 +4615,7 @@ yyreduce:
   case 252:
 
 /* Line 1455 of yacc.c  */
-#line 2004 "program_parse.y"
+#line 2005 "program_parse.y"
     {
 	   if (state->mode == ARB_vertex) {
 	      (yyval.result) = VERT_RESULT_HPOS;
@@ -4628,7 +4629,7 @@ yyreduce:
   case 253:
 
 /* Line 1455 of yacc.c  */
-#line 2013 "program_parse.y"
+#line 2014 "program_parse.y"
     {
 	   if (state->mode == ARB_vertex) {
 	      (yyval.result) = VERT_RESULT_FOGC;
@@ -4642,7 +4643,7 @@ yyreduce:
   case 254:
 
 /* Line 1455 of yacc.c  */
-#line 2022 "program_parse.y"
+#line 2023 "program_parse.y"
     {
 	   (yyval.result) = (yyvsp[(2) - (2)].result);
 	;}
@@ -4651,7 +4652,7 @@ yyreduce:
   case 255:
 
 /* Line 1455 of yacc.c  */
-#line 2026 "program_parse.y"
+#line 2027 "program_parse.y"
     {
 	   if (state->mode == ARB_vertex) {
 	      (yyval.result) = VERT_RESULT_PSIZ;
@@ -4665,7 +4666,7 @@ yyreduce:
   case 256:
 
 /* Line 1455 of yacc.c  */
-#line 2035 "program_parse.y"
+#line 2036 "program_parse.y"
     {
 	   if (state->mode == ARB_vertex) {
 	      (yyval.result) = VERT_RESULT_TEX0 + (yyvsp[(3) - (3)].integer);
@@ -4679,7 +4680,7 @@ yyreduce:
   case 257:
 
 /* Line 1455 of yacc.c  */
-#line 2044 "program_parse.y"
+#line 2045 "program_parse.y"
     {
 	   if (state->mode == ARB_fragment) {
 	      (yyval.result) = FRAG_RESULT_DEPTH;
@@ -4693,7 +4694,7 @@ yyreduce:
   case 258:
 
 /* Line 1455 of yacc.c  */
-#line 2055 "program_parse.y"
+#line 2056 "program_parse.y"
     {
 	   (yyval.result) = (yyvsp[(2) - (3)].integer) + (yyvsp[(3) - (3)].integer);
 	;}
@@ -4702,7 +4703,7 @@ yyreduce:
   case 259:
 
 /* Line 1455 of yacc.c  */
-#line 2061 "program_parse.y"
+#line 2062 "program_parse.y"
     {
 	   (yyval.integer) = (state->mode == ARB_vertex)
 	      ? VERT_RESULT_COL0
@@ -4713,7 +4714,7 @@ yyreduce:
   case 260:
 
 /* Line 1455 of yacc.c  */
-#line 2067 "program_parse.y"
+#line 2068 "program_parse.y"
     {
 	   if (state->mode == ARB_vertex) {
 	      (yyval.integer) = VERT_RESULT_COL0;
@@ -4727,7 +4728,7 @@ yyreduce:
   case 261:
 
 /* Line 1455 of yacc.c  */
-#line 2076 "program_parse.y"
+#line 2077 "program_parse.y"
     {
 	   if (state->mode == ARB_vertex) {
 	      (yyval.integer) = VERT_RESULT_BFC0;
@@ -4741,7 +4742,7 @@ yyreduce:
   case 262:
 
 /* Line 1455 of yacc.c  */
-#line 2087 "program_parse.y"
+#line 2088 "program_parse.y"
     {
 	   (yyval.integer) = 0; 
 	;}
@@ -4750,7 +4751,7 @@ yyreduce:
   case 263:
 
 /* Line 1455 of yacc.c  */
-#line 2091 "program_parse.y"
+#line 2092 "program_parse.y"
     {
 	   if (state->mode == ARB_vertex) {
 	      (yyval.integer) = 0;
@@ -4764,7 +4765,7 @@ yyreduce:
   case 264:
 
 /* Line 1455 of yacc.c  */
-#line 2100 "program_parse.y"
+#line 2101 "program_parse.y"
     {
 	   if (state->mode == ARB_vertex) {
 	      (yyval.integer) = 1;
@@ -4778,91 +4779,91 @@ yyreduce:
   case 265:
 
 /* Line 1455 of yacc.c  */
-#line 2110 "program_parse.y"
+#line 2111 "program_parse.y"
     { (yyval.integer) = 0; ;}
     break;
 
   case 266:
 
 /* Line 1455 of yacc.c  */
-#line 2111 "program_parse.y"
+#line 2112 "program_parse.y"
     { (yyval.integer) = 0; ;}
     break;
 
   case 267:
 
 /* Line 1455 of yacc.c  */
-#line 2112 "program_parse.y"
+#line 2113 "program_parse.y"
     { (yyval.integer) = 1; ;}
     break;
 
   case 268:
 
 /* Line 1455 of yacc.c  */
-#line 2115 "program_parse.y"
+#line 2116 "program_parse.y"
     { (yyval.integer) = 0; ;}
     break;
 
   case 269:
 
 /* Line 1455 of yacc.c  */
-#line 2116 "program_parse.y"
+#line 2117 "program_parse.y"
     { (yyval.integer) = 0; ;}
     break;
 
   case 270:
 
 /* Line 1455 of yacc.c  */
-#line 2117 "program_parse.y"
+#line 2118 "program_parse.y"
     { (yyval.integer) = 1; ;}
     break;
 
   case 271:
 
 /* Line 1455 of yacc.c  */
-#line 2120 "program_parse.y"
+#line 2121 "program_parse.y"
     { (yyval.integer) = 0; ;}
     break;
 
   case 272:
 
 /* Line 1455 of yacc.c  */
-#line 2121 "program_parse.y"
+#line 2122 "program_parse.y"
     { (yyval.integer) = (yyvsp[(2) - (3)].integer); ;}
     break;
 
   case 273:
 
 /* Line 1455 of yacc.c  */
-#line 2124 "program_parse.y"
+#line 2125 "program_parse.y"
     { (yyval.integer) = 0; ;}
     break;
 
   case 274:
 
 /* Line 1455 of yacc.c  */
-#line 2125 "program_parse.y"
+#line 2126 "program_parse.y"
     { (yyval.integer) = (yyvsp[(2) - (3)].integer); ;}
     break;
 
   case 275:
 
 /* Line 1455 of yacc.c  */
-#line 2128 "program_parse.y"
+#line 2129 "program_parse.y"
     { (yyval.integer) = 0; ;}
     break;
 
   case 276:
 
 /* Line 1455 of yacc.c  */
-#line 2129 "program_parse.y"
+#line 2130 "program_parse.y"
     { (yyval.integer) = (yyvsp[(2) - (3)].integer); ;}
     break;
 
   case 277:
 
 /* Line 1455 of yacc.c  */
-#line 2133 "program_parse.y"
+#line 2134 "program_parse.y"
     {
 	   if ((unsigned) (yyvsp[(1) - (1)].integer) >= state->MaxTextureCoordUnits) {
 	      yyerror(& (yylsp[(1) - (1)]), state, "invalid texture coordinate unit selector");
@@ -4876,7 +4877,7 @@ yyreduce:
   case 278:
 
 /* Line 1455 of yacc.c  */
-#line 2144 "program_parse.y"
+#line 2145 "program_parse.y"
     {
 	   if ((unsigned) (yyvsp[(1) - (1)].integer) >= state->MaxTextureImageUnits) {
 	      yyerror(& (yylsp[(1) - (1)]), state, "invalid texture image unit selector");
@@ -4890,7 +4891,7 @@ yyreduce:
   case 279:
 
 /* Line 1455 of yacc.c  */
-#line 2155 "program_parse.y"
+#line 2156 "program_parse.y"
     {
 	   if ((unsigned) (yyvsp[(1) - (1)].integer) >= state->MaxTextureUnits) {
 	      yyerror(& (yylsp[(1) - (1)]), state, "invalid texture unit selector");
@@ -4904,7 +4905,7 @@ yyreduce:
   case 280:
 
 /* Line 1455 of yacc.c  */
-#line 2166 "program_parse.y"
+#line 2167 "program_parse.y"
     {
 	   struct asm_symbol *exist = (struct asm_symbol *)
 	      _mesa_symbol_table_find_symbol(state->st, 0, (yyvsp[(2) - (4)].string));
@@ -4933,7 +4934,7 @@ yyreduce:
 
 
 /* Line 1455 of yacc.c  */
-#line 4937 "program_parse.tab.c"
+#line 4938 "program_parse.tab.c"
       default: break;
     }
   YY_SYMBOL_PRINT ("-> $$ =", yyr1[yyn], &yyval, &yyloc);
@@ -5152,7 +5153,7 @@ yyreturn:
 
 
 /* Line 1675 of yacc.c  */
-#line 2195 "program_parse.y"
+#line 2196 "program_parse.y"
 
 
 void
diff --git a/src/mesa/program/program_parse.y b/src/mesa/program/program_parse.y
index 861927c744c..fb6ef85a9fc 100644
--- a/src/mesa/program/program_parse.y
+++ b/src/mesa/program/program_parse.y
@@ -835,6 +835,7 @@ srcReg: USED_IDENTIFIER /* temporaryReg | progParamSingle */
 	   $$.Base.File = $1->param_binding_type;
 
 	   if ($3.Base.RelAddr) {
+              state->prog->IndirectRegisterFiles |= (1 << $$.Base.File);
 	      $1->param_accessed_indirectly = 1;
 
 	      $$.Base.RelAddr = 1;
@@ -934,7 +935,7 @@ addrRegRelOffset:              { $$ = 0; }
 
 addrRegPosOffset: INTEGER
 	{
-	   if (($1 < 0) || ($1 > 63)) {
+	   if (($1 < 0) || ($1 > 4095)) {
               char s[100];
               _mesa_snprintf(s, sizeof(s),
                              "relative address offset too large (%d)", $1);
@@ -948,7 +949,7 @@ addrRegPosOffset: INTEGER
 
 addrRegNegOffset: INTEGER
 	{
-	   if (($1 < 0) || ($1 > 64)) {
+	   if (($1 < 0) || ($1 > 4096)) {
               char s[100];
               _mesa_snprintf(s, sizeof(s),
                              "relative address offset too large (%d)", $1);
diff --git a/src/mesa/program/programopt.h b/src/mesa/program/programopt.h
index 21fac07849a..4af6357f976 100644
--- a/src/mesa/program/programopt.h
+++ b/src/mesa/program/programopt.h
@@ -26,6 +26,7 @@
 #ifndef PROGRAMOPT_H
 #define PROGRAMOPT_H 1
 
+#include "main/mtypes.h"
 
 extern void
 _mesa_insert_mvp_code(GLcontext *ctx, struct gl_vertex_program *vprog);
diff --git a/src/mesa/slang/library/slang_common_builtin.gc b/src/mesa/slang/library/slang_common_builtin.gc
index d75354deffe..1f5ddbc1ee2 100644
--- a/src/mesa/slang/library/slang_common_builtin.gc
+++ b/src/mesa/slang/library/slang_common_builtin.gc
@@ -411,7 +411,7 @@ float atan(const float y, const float x)
    if (abs(x) > 1.0e-4) {
       r = atan(y / x);
       if (x < 0.0) {
-         r = r + sign(y) * 3.141593;
+         r = r + 3.141593 - 6.283186 * float(y < 0.0);
       }
    }
    else {
diff --git a/src/mesa/slang/slang_builtin.h b/src/mesa/slang/slang_builtin.h
index ed9ae80b3c3..dc92f83f8ef 100644
--- a/src/mesa/slang/slang_builtin.h
+++ b/src/mesa/slang/slang_builtin.h
@@ -26,8 +26,8 @@
 #ifndef SLANG_BUILTIN_H
 #define SLANG_BUILTIN_H
 
-#include "program/prog_parameter.h"
-#include "slang_utility.h"
+#include "main/glheader.h"
+#include "main/mtypes.h"
 #include "slang_ir.h"
 
 
diff --git a/src/mesa/slang/slang_codegen.h b/src/mesa/slang/slang_codegen.h
index 461633fe346..ff0279bbfed 100644
--- a/src/mesa/slang/slang_codegen.h
+++ b/src/mesa/slang/slang_codegen.h
@@ -27,9 +27,13 @@
 #define SLANG_CODEGEN_H
 
 
-#include "main/imports.h"
+#include "main/glheader.h"
 #include "slang_compile.h"
+#include "slang_compile_variable.h"
+#include "slang_typeinfo.h"
+#include "slang_utility.h"
 
+struct slang_function_;
 
 #define MAX_LOOP_DEPTH 30
 
diff --git a/src/mesa/slang/slang_compile.c b/src/mesa/slang/slang_compile.c
index 12ab4666aed..de1bb56cd9a 100644
--- a/src/mesa/slang/slang_compile.c
+++ b/src/mesa/slang/slang_compile.c
@@ -36,6 +36,7 @@
 #include "program/prog_print.h"
 #include "program/prog_parameter.h"
 #include "../../glsl/pp/sl_pp_public.h"
+#include "../../glsl/pp/sl_pp_purify.h"
 #include "../../glsl/cl/sl_cl_parse.h"
 #include "slang_codegen.h"
 #include "slang_compile.h"
diff --git a/src/mesa/slang/slang_compile.h b/src/mesa/slang/slang_compile.h
index 71fcaa39931..6061f878e75 100644
--- a/src/mesa/slang/slang_compile.h
+++ b/src/mesa/slang/slang_compile.h
@@ -25,13 +25,14 @@
 #if !defined SLANG_COMPILE_H
 #define SLANG_COMPILE_H
 
-#include "main/imports.h"
+#include "main/glheader.h"
 #include "main/mtypes.h"
-#include "slang_typeinfo.h"
-#include "slang_compile_variable.h"
-#include "slang_compile_struct.h"
-#include "slang_compile_operation.h"
 #include "slang_compile_function.h"
+#include "slang_compile_struct.h"
+#include "slang_compile_variable.h"
+#include "slang_utility.h"
+
+struct slang_code_object_;
 
 #if defined __cplusplus
 extern "C" {
diff --git a/src/mesa/slang/slang_compile_function.h b/src/mesa/slang/slang_compile_function.h
index a5445ec2537..0eced3ca1a1 100644
--- a/src/mesa/slang/slang_compile_function.h
+++ b/src/mesa/slang/slang_compile_function.h
@@ -25,6 +25,14 @@
 #ifndef SLANG_COMPILE_FUNCTION_H
 #define SLANG_COMPILE_FUNCTION_H
 
+#include "main/glheader.h"
+#include "slang_compile_operation.h"
+#include "slang_compile_variable.h"
+#include "slang_log.h"
+#include "slang_utility.h"
+
+struct slang_name_space_;
+struct slang_operation_;
 
 /**
  * Types of functions.
diff --git a/src/mesa/slang/slang_compile_operation.h b/src/mesa/slang/slang_compile_operation.h
index 1f15c198963..b8c5f214cf0 100644
--- a/src/mesa/slang/slang_compile_operation.h
+++ b/src/mesa/slang/slang_compile_operation.h
@@ -26,6 +26,10 @@
 #define SLANG_COMPILE_OPERATION_H
 
 
+#include "main/glheader.h"
+#include "slang_compile_variable.h"
+#include "slang_utility.h"
+
 /**
  * Types of slang operations.
  * These are the types of the AST (abstract syntax tree) nodes.
diff --git a/src/mesa/slang/slang_compile_struct.h b/src/mesa/slang/slang_compile_struct.h
index 90c5512f4d3..7be6f204e11 100644
--- a/src/mesa/slang/slang_compile_struct.h
+++ b/src/mesa/slang/slang_compile_struct.h
@@ -29,6 +29,9 @@
 extern "C" {
 #endif
 
+#include "main/glheader.h"
+#include "slang_utility.h"
+
 struct slang_function_;
 
 typedef struct slang_struct_scope_
diff --git a/src/mesa/slang/slang_compile_variable.h b/src/mesa/slang/slang_compile_variable.h
index 5c9d248b354..48dc6efca4b 100644
--- a/src/mesa/slang/slang_compile_variable.h
+++ b/src/mesa/slang/slang_compile_variable.h
@@ -26,7 +26,9 @@
 #define SLANG_COMPILE_VARIABLE_H
 
 
-struct slang_ir_storage_;
+#include "main/glheader.h"
+#include "slang_typeinfo.h"
+#include "slang_utility.h"
 
 
 /**
diff --git a/src/mesa/slang/slang_emit.h b/src/mesa/slang/slang_emit.h
index ab4c202d673..f93d6b00d69 100644
--- a/src/mesa/slang/slang_emit.h
+++ b/src/mesa/slang/slang_emit.h
@@ -25,11 +25,9 @@
 #ifndef SLANG_EMIT_H
 #define SLANG_EMIT_H
 
-
-#include "main/imports.h"
-#include "slang_compile.h"
+#include "main/glheader.h"
 #include "slang_ir.h"
-#include "main/mtypes.h"
+#include "slang_vartable.h"
 
 
 extern GLuint
diff --git a/src/mesa/slang/slang_ir.h b/src/mesa/slang/slang_ir.h
index b7a373746b4..ce9a6c5a483 100644
--- a/src/mesa/slang/slang_ir.h
+++ b/src/mesa/slang/slang_ir.h
@@ -37,6 +37,7 @@
 #include "slang_compile.h"
 #include "slang_label.h"
 #include "main/mtypes.h"
+#include "program/prog_instruction.h"
 
 
 /**
diff --git a/src/mesa/slang/slang_label.c b/src/mesa/slang/slang_label.c
index 8e3a8ebc1aa..24881d5b6e6 100644
--- a/src/mesa/slang/slang_label.c
+++ b/src/mesa/slang/slang_label.c
@@ -7,6 +7,8 @@
  */
 
 
+#include "main/mtypes.h"
+#include "program/prog_instruction.h"
 #include "slang_label.h"
 #include "slang_mem.h"
 
diff --git a/src/mesa/slang/slang_label.h b/src/mesa/slang/slang_label.h
index 4d04df18d25..b0cff3a8e89 100644
--- a/src/mesa/slang/slang_label.h
+++ b/src/mesa/slang/slang_label.h
@@ -1,10 +1,9 @@
 #ifndef SLANG_LABEL_H
 #define SLANG_LABEL_H 1
 
-#include "main/imports.h"
-#include "main/mtypes.h"
-#include "program/prog_instruction.h"
+#include "main/glheader.h"
 
+struct gl_program;
 
 struct slang_label_
 {
diff --git a/src/mesa/slang/slang_link.c b/src/mesa/slang/slang_link.c
index 00c2c13cc67..c21f67256a5 100644
--- a/src/mesa/slang/slang_link.c
+++ b/src/mesa/slang/slang_link.c
@@ -756,6 +756,8 @@ _slang_update_inputs_outputs(struct gl_program *prog)
    prog->InputsRead = 0x0;
    prog->OutputsWritten = 0x0;
 
+   prog->IndirectRegisterFiles = 0x0;
+
    for (i = 0; i < prog->NumInstructions; i++) {
       const struct prog_instruction *inst = prog->Instructions + i;
       const GLuint numSrc = _mesa_num_inst_src_regs(inst->Opcode);
@@ -774,6 +776,9 @@ _slang_update_inputs_outputs(struct gl_program *prog)
          else if (inst->SrcReg[j].File == PROGRAM_ADDRESS) {
             maxAddrReg = MAX2(maxAddrReg, (GLuint) (inst->SrcReg[j].Index + 1));
          }
+
+         if (inst->SrcReg[j].RelAddr)
+            prog->IndirectRegisterFiles |= (1 << inst->SrcReg[j].File);
       }
 
       if (inst->DstReg.File == PROGRAM_OUTPUT) {
@@ -784,6 +789,8 @@ _slang_update_inputs_outputs(struct gl_program *prog)
       else if (inst->DstReg.File == PROGRAM_ADDRESS) {
          maxAddrReg = MAX2(maxAddrReg, inst->DstReg.Index + 1);
       }
+      if (inst->DstReg.RelAddr)
+         prog->IndirectRegisterFiles |= (1 << inst->DstReg.File);
    }
    prog->NumAddressRegs = maxAddrReg;
 }
@@ -1199,11 +1206,11 @@ _slang_link(GLcontext *ctx,
       vertNotify = ctx->Driver.ProgramStringNotify(ctx, GL_FRAGMENT_PROGRAM_ARB,
                                                  &shProg->FragmentProgram->Base);
       if (ctx->Shader.Flags & GLSL_DUMP) {
-         printf("Mesa pre-link fragment program:\n");
+         fprintf(stderr, "Mesa pre-link fragment program:\n");
          _mesa_print_program(&fragProg->Base);
          _mesa_print_program_parameters(ctx, &fragProg->Base);
 
-         printf("Mesa post-link fragment program:\n");
+         fprintf(stderr, "Mesa post-link fragment program:\n");
          _mesa_print_program(&shProg->FragmentProgram->Base);
          _mesa_print_program_parameters(ctx, &shProg->FragmentProgram->Base);
       }
@@ -1222,11 +1229,11 @@ _slang_link(GLcontext *ctx,
       geomNotify = ctx->Driver.ProgramStringNotify(ctx, MESA_GEOMETRY_PROGRAM,
                                                    &shProg->GeometryProgram->Base);
       if (ctx->Shader.Flags & GLSL_DUMP) {
-         printf("Mesa pre-link geometry program:\n");
+         fprintf(stderr, "Mesa pre-link geometry program:\n");
          _mesa_print_program(&geomProg->Base);
          _mesa_print_program_parameters(ctx, &geomProg->Base);
 
-         printf("Mesa post-link geometry program:\n");
+         fprintf(stderr, "Mesa post-link geometry program:\n");
          _mesa_print_program(&shProg->GeometryProgram->Base);
          _mesa_print_program_parameters(ctx, &shProg->GeometryProgram->Base);
       }
@@ -1240,11 +1247,11 @@ _slang_link(GLcontext *ctx,
       fragNotify = ctx->Driver.ProgramStringNotify(ctx, GL_VERTEX_PROGRAM_ARB,
                                                    &shProg->VertexProgram->Base);
       if (ctx->Shader.Flags & GLSL_DUMP) {
-         printf("Mesa pre-link vertex program:\n");
+         fprintf(stderr, "Mesa pre-link vertex program:\n");
          _mesa_print_program(&vertProg->Base);
          _mesa_print_program_parameters(ctx, &vertProg->Base);
 
-         printf("Mesa post-link vertex program:\n");
+         fprintf(stderr, "Mesa post-link vertex program:\n");
          _mesa_print_program(&shProg->VertexProgram->Base);
          _mesa_print_program_parameters(ctx, &shProg->VertexProgram->Base);
       }
@@ -1259,10 +1266,10 @@ _slang_link(GLcontext *ctx,
    }
 
    if (ctx->Shader.Flags & GLSL_DUMP) {
-      printf("Varying vars:\n");
+      fprintf(stderr, "Varying vars:\n");
       _mesa_print_parameter_list(shProg->Varying);
       if (shProg->InfoLog) {
-         printf("Info Log: %s\n", shProg->InfoLog);
+         fprintf(stderr, "Info Log: %s\n", shProg->InfoLog);
       }
    }
 
diff --git a/src/mesa/slang/slang_link.h b/src/mesa/slang/slang_link.h
index 2b44d20787a..3e9fa2d743d 100644
--- a/src/mesa/slang/slang_link.h
+++ b/src/mesa/slang/slang_link.h
@@ -25,7 +25,7 @@
 #ifndef SLANG_LINK_H
 #define SLANG_LINK_H 1
 
-#include "slang_compile.h"
+#include "main/mtypes.h"
 
 
 extern void
diff --git a/src/mesa/slang/slang_log.h b/src/mesa/slang/slang_log.h
index dcaba0285a7..544a26654e7 100644
--- a/src/mesa/slang/slang_log.h
+++ b/src/mesa/slang/slang_log.h
@@ -27,6 +27,8 @@
 #define SLANG_LOG_H
 
 
+#include "main/glheader.h"
+
 typedef struct slang_info_log_
 {
    char *text;
diff --git a/src/mesa/slang/slang_print.h b/src/mesa/slang/slang_print.h
index 46605c80610..99da3041437 100644
--- a/src/mesa/slang/slang_print.h
+++ b/src/mesa/slang/slang_print.h
@@ -3,6 +3,12 @@
 #ifndef SLANG_PRINT
 #define SLANG_PRINT
 
+#include "main/glheader.h"
+#include "slang_compile_function.h"
+#include "slang_compile_operation.h"
+#include "slang_compile_variable.h"
+#include "slang_typeinfo.h"
+
 extern void
 slang_print_function(const slang_function *f, GLboolean body);
 
diff --git a/src/mesa/slang/slang_simplify.h b/src/mesa/slang/slang_simplify.h
index 8689c23b1a0..37fb938d4fb 100644
--- a/src/mesa/slang/slang_simplify.h
+++ b/src/mesa/slang/slang_simplify.h
@@ -26,6 +26,13 @@
 #define SLANG_SIMPLIFY_H
 
 
+#include "main/glheader.h"
+#include "slang_compile.h"
+#include "slang_compile_function.h"
+#include "slang_compile_operation.h"
+#include "slang_log.h"
+#include "slang_utility.h"
+
 extern GLint
 _slang_lookup_constant(const char *name);
 
diff --git a/src/mesa/slang/slang_utility.h b/src/mesa/slang/slang_utility.h
index 2c0d0bcbb2a..cb9b6d2aaaa 100644
--- a/src/mesa/slang/slang_utility.h
+++ b/src/mesa/slang/slang_utility.h
@@ -26,6 +26,8 @@
 #define SLANG_UTILITY_H
 
 
+#include "main/glheader.h"
+
 /* Compile-time assertions.  If the expression is zero, try to declare an
  * array of size [-1] to cause compilation error.
  */
diff --git a/src/mesa/slang/slang_vartable.h b/src/mesa/slang/slang_vartable.h
index 94bcd63f45a..97945b89d03 100644
--- a/src/mesa/slang/slang_vartable.h
+++ b/src/mesa/slang/slang_vartable.h
@@ -2,6 +2,9 @@
 #ifndef SLANG_VARTABLE_H
 #define SLANG_VARTABLE_H
 
+#include "main/glheader.h"
+#include "slang_utility.h"
+
 struct slang_ir_storage_;
 
 typedef struct slang_var_table_ slang_var_table;
diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h
index 1f0fef63df5..c7a04951bff 100644
--- a/src/mesa/state_tracker/st_atom.h
+++ b/src/mesa/state_tracker/st_atom.h
@@ -34,6 +34,8 @@
 #ifndef ST_ATOM_H
 #define ST_ATOM_H
 
+#include "main/glheader.h"
+
 struct st_context;
 struct st_tracked_state;
 
diff --git a/src/mesa/state_tracker/st_atom_constbuf.h b/src/mesa/state_tracker/st_atom_constbuf.h
index f707534e2cf..97b076629ee 100644
--- a/src/mesa/state_tracker/st_atom_constbuf.h
+++ b/src/mesa/state_tracker/st_atom_constbuf.h
@@ -29,6 +29,9 @@
 #ifndef ST_ATOM_CONSTBUF_H
 #define ST_ATOM_CONSTBUF_H
 
+struct gl_program_parameter_list;
+struct st_context;
+
 
 void st_upload_constants( struct st_context *st,
                           struct gl_program_parameter_list *params,
diff --git a/src/mesa/state_tracker/st_atom_depth.c b/src/mesa/state_tracker/st_atom_depth.c
index 3c07afba9aa..1616e945fea 100644
--- a/src/mesa/state_tracker/st_atom_depth.c
+++ b/src/mesa/state_tracker/st_atom_depth.c
@@ -33,6 +33,8 @@
   */
  
 
+#include <assert.h>
+
 #include "st_context.h"
 #include "st_atom.h"
 #include "pipe/p_context.h"
diff --git a/src/mesa/state_tracker/st_atom_pixeltransfer.c b/src/mesa/state_tracker/st_atom_pixeltransfer.c
index b88c74fa03a..8a8d17599ec 100644
--- a/src/mesa/state_tracker/st_atom_pixeltransfer.c
+++ b/src/mesa/state_tracker/st_atom_pixeltransfer.c
@@ -37,6 +37,7 @@
 #include "main/image.h"
 #include "main/macros.h"
 #include "program/program.h"
+#include "program/prog_cache.h"
 #include "program/prog_instruction.h"
 #include "program/prog_parameter.h"
 #include "program/prog_print.h"
diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c
index cebaad5f000..05442ef91b5 100644
--- a/src/mesa/state_tracker/st_atom_shader.c
+++ b/src/mesa/state_tracker/st_atom_shader.c
@@ -40,7 +40,6 @@
 #include "program/program.h"
 
 #include "pipe/p_context.h"
-#include "pipe/p_shader_tokens.h"
 
 #include "util/u_simple_shaders.h"
 
diff --git a/src/mesa/state_tracker/st_atom_shader.h b/src/mesa/state_tracker/st_atom_shader.h
index 8403bc66c92..56d4c68f4f7 100644
--- a/src/mesa/state_tracker/st_atom_shader.h
+++ b/src/mesa/state_tracker/st_atom_shader.h
@@ -30,6 +30,9 @@
 #define ST_ATOM_SHADER_H
 
 
+struct st_context;
+struct translated_vertex_program;
+
 extern void
 st_free_translated_vertex_programs(struct st_context *st,
                                    struct translated_vertex_program *xvp);
diff --git a/src/mesa/state_tracker/st_atom_stipple.c b/src/mesa/state_tracker/st_atom_stipple.c
index 31e124b3293..ecdd9f06f6a 100644
--- a/src/mesa/state_tracker/st_atom_stipple.c
+++ b/src/mesa/state_tracker/st_atom_stipple.c
@@ -33,6 +33,8 @@
   */
  
 
+#include <assert.h>
+
 #include "st_context.h"
 #include "st_atom.h"
 #include "pipe/p_context.h"
diff --git a/src/mesa/state_tracker/st_cache.h b/src/mesa/state_tracker/st_cache.h
index b81de316ec9..6d5de7b13ad 100644
--- a/src/mesa/state_tracker/st_cache.h
+++ b/src/mesa/state_tracker/st_cache.h
@@ -33,10 +33,11 @@
 #ifndef ST_CACHE_H
 #define ST_CACHE_H
 
-#include "cso_cache/cso_cache.h"
-
 struct pipe_blend_state;
+struct pipe_depth_stencil_alpha_state;
+struct pipe_rasterizer_state;
 struct pipe_sampler_state;
+struct pipe_shader_state;
 struct st_context;
 
 
diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c
index ba600ccef6d..0b8ecd27cb9 100644
--- a/src/mesa/state_tracker/st_cb_bitmap.c
+++ b/src/mesa/state_tracker/st_cb_bitmap.c
@@ -46,6 +46,7 @@
 
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
+#include "pipe/p_shader_tokens.h"
 #include "util/u_inlines.h"
 #include "util/u_draw_quad.h"
 #include "util/u_simple_shaders.h"
diff --git a/src/mesa/state_tracker/st_cb_bitmap.h b/src/mesa/state_tracker/st_cb_bitmap.h
index 8af975b74fc..d04b2b67795 100644
--- a/src/mesa/state_tracker/st_cb_bitmap.h
+++ b/src/mesa/state_tracker/st_cb_bitmap.h
@@ -30,7 +30,10 @@
 #define ST_CB_BITMAP_H
 
 
-#include "main/mtypes.h"
+#include "main/compiler.h"
+
+struct dd_function_table;
+struct st_context;
 
 #if FEATURE_drawpix
 
diff --git a/src/mesa/state_tracker/st_cb_blit.h b/src/mesa/state_tracker/st_cb_blit.h
index 7ab9a54df90..c230652cefc 100644
--- a/src/mesa/state_tracker/st_cb_blit.h
+++ b/src/mesa/state_tracker/st_cb_blit.h
@@ -29,8 +29,10 @@
 #define ST_CB_BLIT_H
 
 
-#include "main/mtypes.h"
-#include "st_context.h"
+#include "main/compiler.h"
+
+struct dd_function_table;
+struct st_context;
 
 
 extern void
diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.h b/src/mesa/state_tracker/st_cb_bufferobjects.h
index a27daac2bf0..1c991d20837 100644
--- a/src/mesa/state_tracker/st_cb_bufferobjects.h
+++ b/src/mesa/state_tracker/st_cb_bufferobjects.h
@@ -28,9 +28,12 @@
 #ifndef ST_CB_BUFFEROBJECTS_H
 #define ST_CB_BUFFEROBJECTS_H
 
-struct st_context;
-struct gl_buffer_object;
+#include "main/compiler.h"
+#include "main/mtypes.h"
+
+struct dd_function_table;
 struct pipe_resource;
+struct st_context;
 
 /**
  * State_tracker vertex/pixel buffer object, derived from Mesa's
diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c
index ea2414c4a00..246ab2e9579 100644
--- a/src/mesa/state_tracker/st_cb_clear.c
+++ b/src/mesa/state_tracker/st_cb_clear.c
@@ -45,6 +45,7 @@
 #include "st_program.h"
 
 #include "pipe/p_context.h"
+#include "pipe/p_shader_tokens.h"
 #include "pipe/p_state.h"
 #include "pipe/p_defines.h"
 #include "util/u_format.h"
diff --git a/src/mesa/state_tracker/st_cb_clear.h b/src/mesa/state_tracker/st_cb_clear.h
index bc035ac25ca..b27c09d10e4 100644
--- a/src/mesa/state_tracker/st_cb_clear.h
+++ b/src/mesa/state_tracker/st_cb_clear.h
@@ -30,6 +30,9 @@
 #define ST_CB_CLEAR_H
 
 
+struct dd_function_table;
+struct st_context;
+
 extern void
 st_init_clear(struct st_context *st);
 
diff --git a/src/mesa/state_tracker/st_cb_condrender.h b/src/mesa/state_tracker/st_cb_condrender.h
index 891f1cbcd8c..79d0db8d08a 100644
--- a/src/mesa/state_tracker/st_cb_condrender.h
+++ b/src/mesa/state_tracker/st_cb_condrender.h
@@ -29,6 +29,8 @@
 #define ST_CB_CONDRENDER_H
 
 
+struct dd_function_table;
+
 extern void st_init_cond_render_functions(struct dd_function_table *functions);
 
 
diff --git a/src/mesa/state_tracker/st_cb_drawpixels.h b/src/mesa/state_tracker/st_cb_drawpixels.h
index 7d5e901ccc5..575f169e08e 100644
--- a/src/mesa/state_tracker/st_cb_drawpixels.h
+++ b/src/mesa/state_tracker/st_cb_drawpixels.h
@@ -30,7 +30,10 @@
 #define ST_CB_DRAWPIXELS_H
 
 
-#include "main/mtypes.h"
+#include "main/compiler.h"
+
+struct dd_function_table;
+struct st_context;
 
 #if FEATURE_drawpix
 
diff --git a/src/mesa/state_tracker/st_cb_drawtex.c b/src/mesa/state_tracker/st_cb_drawtex.c
index b191a7f8902..c99a8d792ed 100644
--- a/src/mesa/state_tracker/st_cb_drawtex.c
+++ b/src/mesa/state_tracker/st_cb_drawtex.c
@@ -14,7 +14,6 @@
 
 #include "main/imports.h"
 #include "main/image.h"
-#include "main/bufferobj.h"
 #include "main/macros.h"
 #include "program/program.h"
 #include "program/prog_print.h"
diff --git a/src/mesa/state_tracker/st_cb_drawtex.h b/src/mesa/state_tracker/st_cb_drawtex.h
index a3f54a349cc..d21262f8977 100644
--- a/src/mesa/state_tracker/st_cb_drawtex.h
+++ b/src/mesa/state_tracker/st_cb_drawtex.h
@@ -10,7 +10,10 @@
 #define ST_CB_DRAWTEX_H
 
 
-#include "main/mtypes.h"
+#include "main/compiler.h"
+
+struct dd_function_table;
+struct st_context;
 
 #if FEATURE_OES_draw_texture
 
diff --git a/src/mesa/state_tracker/st_cb_eglimage.c b/src/mesa/state_tracker/st_cb_eglimage.c
index 4aaf91d5a19..037e576fabe 100644
--- a/src/mesa/state_tracker/st_cb_eglimage.c
+++ b/src/mesa/state_tracker/st_cb_eglimage.c
@@ -33,6 +33,7 @@
 #include "util/u_format.h"
 #include "st_cb_eglimage.h"
 #include "st_cb_fbo.h"
+#include "st_context.h"
 #include "st_texture.h"
 #include "st_format.h"
 #include "st_manager.h"
diff --git a/src/mesa/state_tracker/st_cb_eglimage.h b/src/mesa/state_tracker/st_cb_eglimage.h
index d6953e99f69..b6e44d5aff5 100644
--- a/src/mesa/state_tracker/st_cb_eglimage.h
+++ b/src/mesa/state_tracker/st_cb_eglimage.h
@@ -29,8 +29,9 @@
 #ifndef ST_CB_EGLIMAGE_H
 #define ST_CB_EGLIMAGE_H
 
-#include "main/mtypes.h"
-#include "main/dd.h"
+#include "main/compiler.h"
+
+struct dd_function_table;
 
 #if FEATURE_OES_EGL_image
 
diff --git a/src/mesa/state_tracker/st_cb_fbo.h b/src/mesa/state_tracker/st_cb_fbo.h
index 43b6c1e75f4..62a9bbcb25f 100644
--- a/src/mesa/state_tracker/st_cb_fbo.h
+++ b/src/mesa/state_tracker/st_cb_fbo.h
@@ -29,6 +29,15 @@
 #ifndef ST_CB_FBO_H
 #define ST_CB_FBO_H
 
+#include "main/compiler.h"
+#include "main/glheader.h"
+#include "main/mtypes.h"
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_format.h"
+
+struct dd_function_table;
+struct pipe_context;
 
 /**
  * Derived renderbuffer class.  Just need to add a pointer to the
diff --git a/src/mesa/state_tracker/st_cb_feedback.h b/src/mesa/state_tracker/st_cb_feedback.h
index 706d84960f7..f2342f58238 100644
--- a/src/mesa/state_tracker/st_cb_feedback.h
+++ b/src/mesa/state_tracker/st_cb_feedback.h
@@ -30,7 +30,9 @@
 #define ST_CB_FEEDBACK_H
 
 
-#include "main/mtypes.h"
+#include "main/compiler.h"
+
+struct dd_function_table;
 
 #if FEATURE_feedback
 
diff --git a/src/mesa/state_tracker/st_cb_flush.h b/src/mesa/state_tracker/st_cb_flush.h
index 7fca0176a30..7672b4cf1da 100644
--- a/src/mesa/state_tracker/st_cb_flush.h
+++ b/src/mesa/state_tracker/st_cb_flush.h
@@ -30,6 +30,12 @@
 #define ST_CB_FLUSH_H
 
 
+#include "pipe/p_compiler.h"
+
+struct dd_function_table;
+struct pipe_fence_handle;
+struct st_context;
+
 extern void
 st_init_flush_functions(struct dd_function_table *functions);
 
diff --git a/src/mesa/state_tracker/st_cb_program.h b/src/mesa/state_tracker/st_cb_program.h
index 0de96f2fd22..0fd179ef3df 100644
--- a/src/mesa/state_tracker/st_cb_program.h
+++ b/src/mesa/state_tracker/st_cb_program.h
@@ -29,6 +29,10 @@
 #define ST_CB_PROGRAM_H
 
 
+#include "main/mtypes.h"
+
+struct dd_function_table;
+
 extern void
 st_init_program_functions(struct dd_function_table *functions);
 
diff --git a/src/mesa/state_tracker/st_cb_rasterpos.h b/src/mesa/state_tracker/st_cb_rasterpos.h
index d2ed7297f15..2dc109bb184 100644
--- a/src/mesa/state_tracker/st_cb_rasterpos.h
+++ b/src/mesa/state_tracker/st_cb_rasterpos.h
@@ -29,7 +29,9 @@
 #define ST_CB_RASTERPOS_H
 
 
-#include "main/mtypes.h"
+#include "main/compiler.h"
+
+struct dd_function_table;
 
 #if FEATURE_rastpos
 
diff --git a/src/mesa/state_tracker/st_cb_readpixels.h b/src/mesa/state_tracker/st_cb_readpixels.h
index c90ef029062..9e1f7b4925e 100644
--- a/src/mesa/state_tracker/st_cb_readpixels.h
+++ b/src/mesa/state_tracker/st_cb_readpixels.h
@@ -29,6 +29,10 @@
 #ifndef ST_CB_READPIXELS_H
 #define ST_CB_READPIXELS_H
 
+#include "main/mtypes.h"
+
+struct dd_function_table;
+
 extern struct st_renderbuffer *
 st_get_color_read_renderbuffer(GLcontext *ctx);
 
diff --git a/src/mesa/state_tracker/st_cb_strings.h b/src/mesa/state_tracker/st_cb_strings.h
index 3b765aaa592..92d5d2d9ba7 100644
--- a/src/mesa/state_tracker/st_cb_strings.h
+++ b/src/mesa/state_tracker/st_cb_strings.h
@@ -30,6 +30,8 @@
 #define ST_CB_STRINGS_H
 
 
+struct dd_function_table;
+
 extern void
 st_init_string_functions(struct dd_function_table *functions);
 
diff --git a/src/mesa/state_tracker/st_cb_texture.h b/src/mesa/state_tracker/st_cb_texture.h
index 1cd9fc3a50f..6942478e815 100644
--- a/src/mesa/state_tracker/st_cb_texture.h
+++ b/src/mesa/state_tracker/st_cb_texture.h
@@ -30,6 +30,13 @@
 #define ST_CB_TEXTURE_H
 
 
+#include "main/glheader.h"
+#include "main/mtypes.h"
+
+struct dd_function_table;
+struct pipe_context;
+struct st_context;
+
 extern GLboolean
 st_finalize_texture(GLcontext *ctx,
 		    struct pipe_context *pipe, 
diff --git a/src/mesa/state_tracker/st_cb_viewport.h b/src/mesa/state_tracker/st_cb_viewport.h
index db7dd6eab82..bcfd7cb68af 100644
--- a/src/mesa/state_tracker/st_cb_viewport.h
+++ b/src/mesa/state_tracker/st_cb_viewport.h
@@ -25,5 +25,12 @@
  *
  **************************************************************************/
 
+#ifndef ST_CB_VIEWPORT_H
+#define ST_CB_VIEWPORT_H
+
+struct dd_function_table;
+
 extern void
 st_init_viewport_functions(struct dd_function_table *functions);
+
+#endif /* ST_CB_VIEW_PORT_H */
diff --git a/src/mesa/state_tracker/st_cb_xformfb.h b/src/mesa/state_tracker/st_cb_xformfb.h
index 50efcb9293f..574cf481e18 100644
--- a/src/mesa/state_tracker/st_cb_xformfb.h
+++ b/src/mesa/state_tracker/st_cb_xformfb.h
@@ -29,6 +29,10 @@
 #define ST_CB_XFORMFB_H
 
 
+#include "main/compiler.h"
+
+struct dd_function_table;
+
 #if FEATURE_EXT_transform_feedback
 
 extern void
diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c
index 7eb5f32611d..2ce5f087536 100644
--- a/src/mesa/state_tracker/st_context.c
+++ b/src/mesa/state_tracker/st_context.c
@@ -28,6 +28,7 @@
 #include "main/imports.h"
 #include "main/context.h"
 #include "main/shaderobj.h"
+#include "program/prog_cache.h"
 #include "vbo/vbo.h"
 #include "glapi/glapi.h"
 #include "st_context.h"
@@ -62,6 +63,9 @@
 #include "cso_cache/cso_context.h"
 
 
+DEBUG_GET_ONCE_BOOL_OPTION(mesa_mvp_dp4, "MESA_MVP_DP4", FALSE)
+
+
 /**
  * Called via ctx->Driver.UpdateState()
  */
@@ -169,7 +173,7 @@ struct st_context *st_create_context(gl_api api, struct pipe_context *pipe,
    /* XXX: need a capability bit in gallium to query if the pipe
     * driver prefers DP4 or MUL/MAD for vertex transformation.
     */
-   if (debug_get_bool_option("MESA_MVP_DP4", FALSE))
+   if (debug_get_option_mesa_mvp_dp4())
       _mesa_set_mvp_with_dp4( ctx, GL_TRUE );
 
    return st_create_context_priv(ctx, pipe);
diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h
index a147a021176..60c25fb8f00 100644
--- a/src/mesa/state_tracker/st_context.h
+++ b/src/mesa/state_tracker/st_context.h
@@ -1,3 +1,4 @@
+//struct dd_function_table;
 /**************************************************************************
  * 
  * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
@@ -29,21 +30,17 @@
 #define ST_CONTEXT_H
 
 #include "main/mtypes.h"
-#include "program/prog_cache.h"
 #include "pipe/p_state.h"
 #include "state_tracker/st_api.h"
 
-
-struct st_context;
-struct st_texture_object;
-struct st_fragment_program;
+struct bitmap_cache;
+struct blit_state;
+struct dd_function_table;
 struct draw_context;
 struct draw_stage;
-struct cso_cache;
-struct cso_blend;
 struct gen_mipmap_state;
-struct blit_state;
-struct bitmap_cache;
+struct st_context;
+struct st_fragment_program;
 
 
 #define ST_NEW_MESA                    0x1 /* Mesa state has changed */
diff --git a/src/mesa/state_tracker/st_debug.c b/src/mesa/state_tracker/st_debug.c
index ebf6ec6e7e2..df32491d044 100644
--- a/src/mesa/state_tracker/st_debug.c
+++ b/src/mesa/state_tracker/st_debug.c
@@ -55,6 +55,8 @@ static const struct debug_named_value st_debug_flags[] = {
    { "query",    DEBUG_QUERY, NULL },
    DEBUG_NAMED_VALUE_END
 };
+
+DEBUG_GET_ONCE_FLAGS_OPTION(st_debug, "ST_DEBUG", st_debug_flags, 0)
 #endif
 
 
@@ -62,7 +64,7 @@ void
 st_debug_init(void)
 {
 #ifdef DEBUG
-   ST_DEBUG = debug_get_flags_option("ST_DEBUG", st_debug_flags, 0 );
+   ST_DEBUG = debug_get_option_st_debug();
 #endif
 }
 
diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c
index 5821da4889d..5b054892702 100644
--- a/src/mesa/state_tracker/st_draw.c
+++ b/src/mesa/state_tracker/st_draw.c
@@ -58,6 +58,7 @@
 #include "util/u_inlines.h"
 #include "util/u_format.h"
 #include "util/u_prim.h"
+#include "util/u_draw_quad.h"
 #include "draw/draw_context.h"
 #include "cso_cache/cso_context.h"
 
@@ -494,6 +495,49 @@ setup_non_interleaved_attribs(GLcontext *ctx,
 }
 
 
+static void
+setup_index_buffer(GLcontext *ctx,
+                   const struct _mesa_index_buffer *ib,
+                   struct pipe_index_buffer *ibuffer)
+{
+   struct st_context *st = st_context(ctx);
+   struct pipe_context *pipe = st->pipe;
+
+   memset(ibuffer, 0, sizeof(*ibuffer));
+   if (ib) {
+      struct gl_buffer_object *bufobj = ib->obj;
+
+      switch (ib->type) {
+      case GL_UNSIGNED_INT:
+         ibuffer->index_size = 4;
+         break;
+      case GL_UNSIGNED_SHORT:
+         ibuffer->index_size = 2;
+         break;
+      case GL_UNSIGNED_BYTE:
+         ibuffer->index_size = 1;
+         break;
+      default:
+         assert(0);
+	 return;
+      }
+
+      /* get/create the index buffer object */
+      if (bufobj && bufobj->Name) {
+         /* elements/indexes are in a real VBO */
+         struct st_buffer_object *stobj = st_buffer_object(bufobj);
+         pipe_resource_reference(&ibuffer->buffer, stobj->buffer);
+         ibuffer->offset = pointer_to_offset(ib->ptr);
+      }
+      else {
+         /* element/indicies are in user space memory */
+         ibuffer->buffer =
+            pipe_user_buffer_create(pipe->screen, (void *) ib->ptr,
+                                    ib->count * ibuffer->index_size,
+                                    PIPE_BIND_INDEX_BUFFER);
+      }
+   }
+}
 
 /**
  * Prior to drawing, check that any uniforms referenced by the
@@ -568,8 +612,11 @@ st_draw_vbo(GLcontext *ctx,
    GLuint attr;
    struct pipe_vertex_element velements[PIPE_MAX_ATTRIBS];
    unsigned num_vbuffers, num_velements;
+   struct pipe_index_buffer ibuffer;
    GLboolean userSpace = GL_FALSE;
    GLboolean vertDataEdgeFlags;
+   struct pipe_draw_info info;
+   unsigned i;
 
    /* Mesa core state should have been validated already */
    assert(ctx->NewState == 0x0);
@@ -647,113 +694,35 @@ st_draw_vbo(GLcontext *ctx,
    if (num_vbuffers == 0 || num_velements == 0)
       return;
 
-   /* do actual drawing */
-   if (ib) {
-      /* indexed primitive */
-      struct gl_buffer_object *bufobj = ib->obj;
-      struct pipe_resource *indexBuf = NULL;
-      unsigned indexSize, indexOffset, i;
+   setup_index_buffer(ctx, ib, &ibuffer);
+   pipe->set_index_buffer(pipe, &ibuffer);
 
-      switch (ib->type) {
-      case GL_UNSIGNED_INT:
-         indexSize = 4;
-         break;
-      case GL_UNSIGNED_SHORT:
-         indexSize = 2;
-         break;
-      case GL_UNSIGNED_BYTE:
-         indexSize = 1;
-         break;
-      default:
-         assert(0);
-	 return;
-      }
-
-      /* get/create the index buffer object */
-      if (bufobj && bufobj->Name) {
-         /* elements/indexes are in a real VBO */
-         struct st_buffer_object *stobj = st_buffer_object(bufobj);
-         pipe_resource_reference(&indexBuf, stobj->buffer);
-         indexOffset = pointer_to_offset(ib->ptr) / indexSize;
-      }
-      else {
-         /* element/indicies are in user space memory */
-         indexBuf = pipe_user_buffer_create(pipe->screen, (void *) ib->ptr,
-                                            ib->count * indexSize,
-					    PIPE_BIND_INDEX_BUFFER);
-         indexOffset = 0;
+   util_draw_init_info(&info);
+   if (ib) {
+      info.indexed = TRUE;
+      if (min_index != ~0 && max_index != ~0) {
+         info.min_index = min_index;
+         info.max_index = max_index;
       }
+   }
 
-      /* draw */
-      if (pipe->draw_range_elements && min_index != ~0 && max_index != ~0) {
-         /* XXX: exercise temporary path to pass min/max directly
-          * through to driver & draw module.  These interfaces still
-          * need a bit of work...
-          */
-         for (i = 0; i < nr_prims; i++) {
-            unsigned vcount = prims[i].count;
-            unsigned prim = translate_prim(ctx, prims[i].mode);
-
-            if (u_trim_pipe_prim(prims[i].mode, &vcount)) {
-               pipe->draw_range_elements(pipe, indexBuf, indexSize,
-                                         prims[i].basevertex,
-                                         min_index, max_index, prim,
-                                         prims[i].start + indexOffset, vcount);
-            }
-         }
-      }
-      else {
-         for (i = 0; i < nr_prims; i++) {
-            unsigned vcount = prims[i].count;
-            unsigned prim = translate_prim(ctx, prims[i].mode);
-            
-            if (u_trim_pipe_prim(prims[i].mode, &vcount)) {
-               if (prims[i].num_instances == 1) {
-                  pipe->draw_elements(pipe, indexBuf,
-                                      indexSize,
-                                      prims[i].basevertex,
-                                      prim,
-                                      prims[i].start + indexOffset,
-                                      vcount);
-               }
-               else {
-                  pipe->draw_elements_instanced(pipe, indexBuf,
-                                                indexSize,
-                                                prims[i].basevertex,
-                                                prim,
-                                                prims[i].start + indexOffset,
-                                                vcount,
-                                                0, /* startInstance */
-                                                prims[i].num_instances);
-               }
-            }
-         }
+   /* do actual drawing */
+   for (i = 0; i < nr_prims; i++) {
+      info.mode = translate_prim( ctx, prims[i].mode );
+      info.start = prims[i].start;
+      info.count = prims[i].count;
+      info.instance_count = prims[i].num_instances;
+      info.index_bias = prims[i].basevertex;
+      if (!ib) {
+         info.min_index = info.start;
+         info.max_index = info.start + info.count - 1;
       }
 
-      pipe_resource_reference(&indexBuf, NULL);
+      if (u_trim_pipe_prim(info.mode, &info.count))
+         pipe->draw_vbo(pipe, &info);
    }
-   else {
-      /* non-indexed */
-      GLuint i;
-
-      for (i = 0; i < nr_prims; i++) {
-         unsigned vcount = prims[i].count;
-         unsigned prim = translate_prim(ctx, prims[i].mode);
 
-         if (u_trim_pipe_prim(prims[i].mode, &vcount)) {
-            if (prims[i].num_instances == 1) {
-               pipe->draw_arrays(pipe, prim, prims[i].start, vcount);
-            }
-            else {
-               pipe->draw_arrays_instanced(pipe, prim,
-                                           prims[i].start,
-                                           vcount,
-                                           0, /* startInstance */
-                                           prims[i].num_instances);
-            }
-         }
-      }
-   }
+   pipe_resource_reference(&ibuffer.buffer, NULL);
 
    /* unreference buffers (frees wrapped user-space buffer objects) */
    for (attr = 0; attr < num_vbuffers; attr++) {
diff --git a/src/mesa/state_tracker/st_draw.h b/src/mesa/state_tracker/st_draw.h
index 3e0face656b..f36184487a6 100644
--- a/src/mesa/state_tracker/st_draw.h
+++ b/src/mesa/state_tracker/st_draw.h
@@ -34,8 +34,13 @@
 #ifndef ST_DRAW_H
 #define ST_DRAW_H
 
-struct _mesa_prim;
+#include "main/compiler.h"
+#include "main/glheader.h"
+#include "main/mtypes.h"
+
 struct _mesa_index_buffer;
+struct _mesa_prim;
+struct st_context;
 
 void st_init_draw( struct st_context *st );
 
diff --git a/src/mesa/state_tracker/st_extensions.h b/src/mesa/state_tracker/st_extensions.h
index 2994f16dd33..aa9b2b2b914 100644
--- a/src/mesa/state_tracker/st_extensions.h
+++ b/src/mesa/state_tracker/st_extensions.h
@@ -30,6 +30,8 @@
 #define ST_EXTENSIONS_H
 
 
+struct st_context;
+
 extern void st_init_limits(struct st_context *st);
 
 extern void st_init_extensions(struct st_context *st);
diff --git a/src/mesa/state_tracker/st_format.h b/src/mesa/state_tracker/st_format.h
index 29768f296d6..841c58cadc8 100644
--- a/src/mesa/state_tracker/st_format.h
+++ b/src/mesa/state_tracker/st_format.h
@@ -31,7 +31,12 @@
 #define ST_FORMAT_H
 
 #include "main/formats.h"
+#include "main/mtypes.h"
 
+#include "pipe/p_defines.h"
+#include "pipe/p_format.h"
+
+struct pipe_screen;
 
 extern GLenum
 st_format_datatype(enum pipe_format format);
diff --git a/src/mesa/state_tracker/st_gen_mipmap.h b/src/mesa/state_tracker/st_gen_mipmap.h
index 00fbae93026..016bf3f4bba 100644
--- a/src/mesa/state_tracker/st_gen_mipmap.h
+++ b/src/mesa/state_tracker/st_gen_mipmap.h
@@ -30,6 +30,10 @@
 #define ST_GEN_MIPMAP_H
 
 
+#include "main/mtypes.h"
+
+struct st_context;
+
 extern void
 st_init_generate_mipmap(struct st_context *st);
 
diff --git a/src/mesa/state_tracker/st_gl_api.h b/src/mesa/state_tracker/st_gl_api.h
index fe1aec207ea..57c6d9f24d2 100644
--- a/src/mesa/state_tracker/st_gl_api.h
+++ b/src/mesa/state_tracker/st_gl_api.h
@@ -2,8 +2,6 @@
 #ifndef ST_GL_API_H
 #define ST_GL_API_H
 
-#include "state_tracker/st_api.h"
-
 struct st_api *st_gl_api_create(void);
 struct st_api *st_gl_api_create_es1(void);
 struct st_api *st_gl_api_create_es2(void);
diff --git a/src/mesa/state_tracker/st_manager.h b/src/mesa/state_tracker/st_manager.h
index cd2887b1e0f..48a9d4d99a6 100644
--- a/src/mesa/state_tracker/st_manager.h
+++ b/src/mesa/state_tracker/st_manager.h
@@ -29,8 +29,11 @@
 #ifndef ST_MANAGER_H
 #define ST_MANAGER_H
 
-#include "state_tracker/st_api.h"
-#include "st_context.h"
+#include "main/mtypes.h"
+
+#include "pipe/p_compiler.h"
+
+struct st_context;
 
 struct pipe_surface *
 st_manager_get_egl_image_surface(struct st_context *st,
diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c
index bacd091853b..a19dcc92534 100644
--- a/src/mesa/state_tracker/st_mesa_to_tgsi.c
+++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c
@@ -44,6 +44,15 @@
 #include "util/u_math.h"
 #include "util/u_memory.h"
 
+
+#define PROGRAM_ANY_CONST ((1 << PROGRAM_LOCAL_PARAM) |  \
+                           (1 << PROGRAM_ENV_PARAM) |    \
+                           (1 << PROGRAM_STATE_VAR) |    \
+                           (1 << PROGRAM_NAMED_PARAM) |  \
+                           (1 << PROGRAM_CONSTANT) |     \
+                           (1 << PROGRAM_UNIFORM))
+
+
 struct label {
    unsigned branch_target;
    unsigned token;
@@ -205,7 +214,7 @@ src_register( struct st_translate *t,
       return ureg_src_undef();
 
    case PROGRAM_TEMPORARY:
-      ASSERT(index >= 0);
+      assert(index >= 0);
       if (ureg_dst_is_undef(t->temps[index]))
          t->temps[index] = ureg_DECL_temporary( t->ureg );
       assert(index < Elements(t->temps));
@@ -215,7 +224,7 @@ src_register( struct st_translate *t,
    case PROGRAM_ENV_PARAM:
    case PROGRAM_LOCAL_PARAM:
    case PROGRAM_UNIFORM:
-      ASSERT(index >= 0);
+      assert(index >= 0);
       return t->constants[index];
    case PROGRAM_STATE_VAR:
    case PROGRAM_CONSTANT:       /* ie, immediate */
@@ -738,9 +747,11 @@ emit_adjusted_wpos( struct st_translate *t,
    struct ureg_dst wpos_temp = ureg_DECL_temporary(ureg);
    struct ureg_src wpos_input = t->inputs[t->inputMapping[FRAG_ATTRIB_WPOS]];
 
-   ureg_ADD(ureg,
-            ureg_writemask(wpos_temp, TGSI_WRITEMASK_X | TGSI_WRITEMASK_Y),
-            wpos_input, ureg_imm1f(ureg, value));
+   /* Note that we bias X and Y and pass Z and W through unchanged.
+    * The shader might also use gl_FragCoord.w and .z.
+    */
+   ureg_ADD(ureg, wpos_temp, wpos_input,
+            ureg_imm4f(ureg, value, value, 0.0f, 0.0f));
 
    t->inputs[t->inputMapping[FRAG_ATTRIB_WPOS]] = ureg_src(wpos_temp);
 }
@@ -1057,6 +1068,16 @@ st_translate_mesa_program(
       t->address[0] = ureg_DECL_address( ureg );
    }
 
+   if (program->IndirectRegisterFiles & (1 << PROGRAM_TEMPORARY)) {
+      /* If temps are accessed with indirect addressing, declare temporaries
+       * in sequential order.  Else, we declare them on demand elsewhere.
+       */
+      for (i = 0; i < program->NumTemporaries; i++) {
+         /* XXX use TGSI_FILE_TEMPORARY_ARRAY when it's supported by ureg */
+         t->temps[i] = ureg_DECL_temporary( t->ureg );
+      }
+   }
+
    /* Emit constants and immediates.  Mesa uses a single index space
     * for these, so we put all the translated regs in t->constants.
     */
@@ -1067,7 +1088,7 @@ st_translate_mesa_program(
          ret = PIPE_ERROR_OUT_OF_MEMORY;
          goto out;
       }
-      
+
       for (i = 0; i < program->Parameters->NumParameters; i++) {
          switch (program->Parameters->Parameters[i].Type) {
          case PROGRAM_ENV_PARAM:
@@ -1078,13 +1099,14 @@ st_translate_mesa_program(
             t->constants[i] = ureg_DECL_constant( ureg, i );
             break;
 
-            /* Emit immediates only when there is no address register
-             * in use.  FIXME: Be smarter and recognize param arrays:
+            /* Emit immediates only when there's no indirect addressing of
+             * the const buffer.
+             * FIXME: Be smarter and recognize param arrays:
              * indirect addressing is only valid within the referenced
              * array.
              */
          case PROGRAM_CONSTANT:
-            if (program->NumAddressRegs > 0) 
+            if (program->IndirectRegisterFiles & PROGRAM_ANY_CONST)
                t->constants[i] = ureg_DECL_constant( ureg, i );
             else
                t->constants[i] = 
diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.h b/src/mesa/state_tracker/st_mesa_to_tgsi.h
index e3c5bd1d94d..ca076ce3622 100644
--- a/src/mesa/state_tracker/st_mesa_to_tgsi.h
+++ b/src/mesa/state_tracker/st_mesa_to_tgsi.h
@@ -30,8 +30,10 @@
 #define ST_MESA_TO_TGSI_H
 
 #include "main/mtypes.h"
-#include "tgsi/tgsi_ureg.h"
 
+#include "pipe/p_compiler.h"
+
+struct ureg_program;
 
 #if defined __cplusplus
 extern "C" {
diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c
index 6f3ecdbce11..91528c227b2 100644
--- a/src/mesa/state_tracker/st_program.c
+++ b/src/mesa/state_tracker/st_program.c
@@ -41,6 +41,7 @@
 #include "pipe/p_shader_tokens.h"
 #include "draw/draw_context.h"
 #include "tgsi/tgsi_dump.h"
+#include "tgsi/tgsi_ureg.h"
 
 #include "st_debug.h"
 #include "st_context.h"
diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h
index d779d5a6dde..3805b9a725e 100644
--- a/src/mesa/state_tracker/st_program.h
+++ b/src/mesa/state_tracker/st_program.h
@@ -36,11 +36,8 @@
 
 #include "main/mtypes.h"
 #include "program/program.h"
-#include "pipe/p_shader_tokens.h"
-
-
-struct cso_fragment_shader;
-struct cso_vertex_shader;
+#include "pipe/p_state.h"
+#include "st_context.h"
 
 
 /**
diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c
index dbdf1ea1ad0..add6e949dfb 100644
--- a/src/mesa/state_tracker/st_texture.c
+++ b/src/mesa/state_tracker/st_texture.c
@@ -25,14 +25,14 @@
  * 
  **************************************************************************/
 
+#include <stdio.h>
+
 #include "st_context.h"
 #include "st_format.h"
 #include "st_texture.h"
 #include "st_cb_fbo.h"
 #include "main/enums.h"
 
-#undef Elements  /* fix re-defined macro warning */
-
 #include "pipe/p_state.h"
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
diff --git a/src/mesa/swrast/s_aaline.h b/src/mesa/swrast/s_aaline.h
index f1d708ec801..922eb230e51 100644
--- a/src/mesa/swrast/s_aaline.h
+++ b/src/mesa/swrast/s_aaline.h
@@ -28,7 +28,7 @@
 #define S_AALINE_H
 
 
-#include "swrast.h"
+#include "main/mtypes.h"
 
 
 extern void
diff --git a/src/mesa/swrast/s_aatriangle.h b/src/mesa/swrast/s_aatriangle.h
index 4b57fa73a27..9aed41a1915 100644
--- a/src/mesa/swrast/s_aatriangle.h
+++ b/src/mesa/swrast/s_aatriangle.h
@@ -28,7 +28,7 @@
 #define S_AATRIANGLE_H
 
 
-#include "swrast.h"
+#include "main/mtypes.h"
 
 
 extern void
diff --git a/src/mesa/swrast/s_alpha.h b/src/mesa/swrast/s_alpha.h
index 7a5b72e650a..239484a9743 100644
--- a/src/mesa/swrast/s_alpha.h
+++ b/src/mesa/swrast/s_alpha.h
@@ -28,7 +28,8 @@
 #define S_ALPHA_H
 
 
-#include "s_context.h"
+#include "main/mtypes.h"
+#include "s_span.h"
 
 
 extern GLint
diff --git a/src/mesa/swrast/s_atifragshader.c b/src/mesa/swrast/s_atifragshader.c
index fa280e72e40..1338b6802d4 100644
--- a/src/mesa/swrast/s_atifragshader.c
+++ b/src/mesa/swrast/s_atifragshader.c
@@ -21,10 +21,10 @@
 
 #include "main/glheader.h"
 #include "main/colormac.h"
-#include "main/context.h"
 #include "main/macros.h"
 #include "main/atifragshader.h"
 #include "swrast/s_atifragshader.h"
+#include "swrast/s_context.h"
 
 
 /**
diff --git a/src/mesa/swrast/s_atifragshader.h b/src/mesa/swrast/s_atifragshader.h
index 871a0c04559..cce455a0465 100644
--- a/src/mesa/swrast/s_atifragshader.h
+++ b/src/mesa/swrast/s_atifragshader.h
@@ -27,7 +27,8 @@
 #define S_ATIFRAGSHADER_H
 
 
-#include "s_context.h"
+#include "main/mtypes.h"
+#include "s_span.h"
 
 
 extern void
diff --git a/src/mesa/swrast/s_blend.h b/src/mesa/swrast/s_blend.h
index 8d5a81635d5..9cedde3bf20 100644
--- a/src/mesa/swrast/s_blend.h
+++ b/src/mesa/swrast/s_blend.h
@@ -27,7 +27,8 @@
 #define S_BLEND_H
 
 
-#include "s_context.h"
+#include "main/mtypes.h"
+#include "s_span.h"
 
 
 extern void
diff --git a/src/mesa/swrast/s_context.c b/src/mesa/swrast/s_context.c
index 6d2d17c61d9..d8d8a80b7d7 100644
--- a/src/mesa/swrast/s_context.c
+++ b/src/mesa/swrast/s_context.c
@@ -28,7 +28,6 @@
 
 #include "main/imports.h"
 #include "main/bufferobj.h"
-#include "main/context.h"
 #include "main/colormac.h"
 #include "main/mtypes.h"
 #include "main/teximage.h"
diff --git a/src/mesa/swrast/s_context.h b/src/mesa/swrast/s_context.h
index c9755e6da18..6d81f74768f 100644
--- a/src/mesa/swrast/s_context.h
+++ b/src/mesa/swrast/s_context.h
@@ -43,6 +43,7 @@
 #ifndef S_CONTEXT_H
 #define S_CONTEXT_H
 
+#include "main/compiler.h"
 #include "main/mtypes.h"
 #include "program/prog_execute.h"
 #include "swrast.h"
diff --git a/src/mesa/swrast/s_depth.c b/src/mesa/swrast/s_depth.c
index ed637cac124..f952fd6baa7 100644
--- a/src/mesa/swrast/s_depth.c
+++ b/src/mesa/swrast/s_depth.c
@@ -30,7 +30,6 @@
 #include "main/imports.h"
 
 #include "s_depth.h"
-#include "s_context.h"
 #include "s_span.h"
 
 
diff --git a/src/mesa/swrast/s_depth.h b/src/mesa/swrast/s_depth.h
index 7eae3667428..878d242f5e5 100644
--- a/src/mesa/swrast/s_depth.h
+++ b/src/mesa/swrast/s_depth.h
@@ -27,7 +27,8 @@
 #define S_DEPTH_H
 
 
-#include "s_context.h"
+#include "main/mtypes.h"
+#include "s_span.h"
 
 
 extern GLuint
diff --git a/src/mesa/swrast/s_feedback.c b/src/mesa/swrast/s_feedback.c
index 373b1416e28..6ac8ac73b0b 100644
--- a/src/mesa/swrast/s_feedback.c
+++ b/src/mesa/swrast/s_feedback.c
@@ -24,7 +24,6 @@
 
 #include "main/glheader.h"
 #include "main/colormac.h"
-#include "main/context.h"
 #include "main/feedback.h"
 #include "main/macros.h"
 
diff --git a/src/mesa/swrast/s_fog.c b/src/mesa/swrast/s_fog.c
index 3fc84392133..689500a613a 100644
--- a/src/mesa/swrast/s_fog.c
+++ b/src/mesa/swrast/s_fog.c
@@ -25,7 +25,6 @@
 
 #include "main/glheader.h"
 #include "main/colormac.h"
-#include "main/context.h"
 #include "main/macros.h"
 
 #include "s_context.h"
diff --git a/src/mesa/swrast/s_fog.h b/src/mesa/swrast/s_fog.h
index 06107de3f9d..a496746d106 100644
--- a/src/mesa/swrast/s_fog.h
+++ b/src/mesa/swrast/s_fog.h
@@ -28,7 +28,8 @@
 #define S_FOG_H
 
 
-#include "swrast.h"
+#include "main/mtypes.h"
+#include "s_span.h"
 
 
 extern GLfloat
diff --git a/src/mesa/swrast/s_fragprog.c b/src/mesa/swrast/s_fragprog.c
index 413f136cd59..9facb44d9bf 100644
--- a/src/mesa/swrast/s_fragprog.c
+++ b/src/mesa/swrast/s_fragprog.c
@@ -24,9 +24,9 @@
 
 #include "main/glheader.h"
 #include "main/colormac.h"
-#include "main/context.h"
 #include "program/prog_instruction.h"
 
+#include "s_context.h"
 #include "s_fragprog.h"
 #include "s_span.h"
 
diff --git a/src/mesa/swrast/s_fragprog.h b/src/mesa/swrast/s_fragprog.h
index e1b7e679185..92b9d01e173 100644
--- a/src/mesa/swrast/s_fragprog.h
+++ b/src/mesa/swrast/s_fragprog.h
@@ -27,7 +27,8 @@
 #define S_FRAGPROG_H
 
 
-#include "s_context.h"
+#include "main/mtypes.h"
+#include "s_span.h"
 
 
 extern void
diff --git a/src/mesa/swrast/s_logic.h b/src/mesa/swrast/s_logic.h
index e8cfae33f23..d609513348d 100644
--- a/src/mesa/swrast/s_logic.h
+++ b/src/mesa/swrast/s_logic.h
@@ -27,7 +27,8 @@
 #define S_LOGIC_H
 
 
-#include "swrast.h"
+#include "main/mtypes.h"
+#include "s_span.h"
 
 extern void
 _swrast_logicop_rgba_span(GLcontext *ctx, struct gl_renderbuffer *rb,
diff --git a/src/mesa/swrast/s_masking.h b/src/mesa/swrast/s_masking.h
index 3ba4f8356cb..cb000da0fd8 100644
--- a/src/mesa/swrast/s_masking.h
+++ b/src/mesa/swrast/s_masking.h
@@ -27,7 +27,8 @@
 #define S_MASKING_H
 
 
-#include "swrast.h"
+#include "main/mtypes.h"
+#include "s_span.h"
 
 
 extern void
diff --git a/src/mesa/swrast/s_points.c b/src/mesa/swrast/s_points.c
index 1663ece8294..12431662c47 100644
--- a/src/mesa/swrast/s_points.c
+++ b/src/mesa/swrast/s_points.c
@@ -25,7 +25,6 @@
 
 #include "main/glheader.h"
 #include "main/colormac.h"
-#include "main/context.h"
 #include "main/macros.h"
 #include "s_context.h"
 #include "s_feedback.h"
diff --git a/src/mesa/swrast/s_readpix.c b/src/mesa/swrast/s_readpix.c
index 6ad9aceec77..553fd9a76d8 100644
--- a/src/mesa/swrast/s_readpix.c
+++ b/src/mesa/swrast/s_readpix.c
@@ -27,7 +27,6 @@
 #include "main/bufferobj.h"
 #include "main/colormac.h"
 #include "main/convolve.h"
-#include "main/context.h"
 #include "main/feedback.h"
 #include "main/formats.h"
 #include "main/image.h"
diff --git a/src/mesa/swrast/s_span.c b/src/mesa/swrast/s_span.c
index 687c8eb0bf8..8931cdec1bc 100644
--- a/src/mesa/swrast/s_span.c
+++ b/src/mesa/swrast/s_span.c
@@ -33,7 +33,6 @@
 
 #include "main/glheader.h"
 #include "main/colormac.h"
-#include "main/context.h"
 #include "main/macros.h"
 #include "main/imports.h"
 #include "main/image.h"
@@ -971,6 +970,10 @@ shade_texture_span(GLcontext *ctx, SWspan *span)
       if (span->primitive == GL_BITMAP && span->array->ChanType != GL_FLOAT) {
          convert_color_type(span, GL_FLOAT, 0);
       }
+      else {
+         span->array->rgba = (void *) span->array->attribs[FRAG_ATTRIB_COL0];
+      }
+
       if (span->primitive != GL_POINT ||
 	  (span->interpMask & SPAN_RGBA) ||
 	  ctx->Point.PointSprite) {
@@ -1222,9 +1225,22 @@ _swrast_write_rgba_span( GLcontext *ctx, SWspan *span)
             GLchan rgbaSave[MAX_WIDTH][4];
             const GLuint fragOutput = multiFragOutputs ? buf : 0;
 
+            /* set span->array->rgba to colors for render buffer's datatype */
             if (rb->DataType != span->array->ChanType || fragOutput > 0) {
                convert_color_type(span, rb->DataType, fragOutput);
             }
+            else {
+               if (rb->DataType == GL_UNSIGNED_BYTE) {
+                  span->array->rgba = span->array->rgba8;
+               }
+               else if (rb->DataType == GL_UNSIGNED_SHORT) {
+                  span->array->rgba = (void *) span->array->rgba16;
+               }
+               else {
+                  span->array->rgba = (void *)
+                     span->array->attribs[FRAG_ATTRIB_COL0];
+               }
+            }
 
             if (!multiFragOutputs && numBuffers > 1) {
                /* save colors for second, third renderbuffer writes */
diff --git a/src/mesa/swrast/s_stencil.h b/src/mesa/swrast/s_stencil.h
index cd6cbc57b0b..c076ebbe2a1 100644
--- a/src/mesa/swrast/s_stencil.h
+++ b/src/mesa/swrast/s_stencil.h
@@ -27,7 +27,8 @@
 #define S_STENCIL_H
 
 
-#include "swrast.h"
+#include "main/mtypes.h"
+#include "s_span.h"
 
 
 
diff --git a/src/mesa/swrast/s_texcombine.h b/src/mesa/swrast/s_texcombine.h
index 9ed96efb879..4f5dfbe1afe 100644
--- a/src/mesa/swrast/s_texcombine.h
+++ b/src/mesa/swrast/s_texcombine.h
@@ -27,7 +27,8 @@
 #define S_TEXCOMBINE_H
 
 
-#include "swrast.h"
+#include "main/mtypes.h"
+#include "s_span.h"
 
 extern void
 _swrast_texture_span( GLcontext *ctx, SWspan *span );
diff --git a/src/mesa/swrast/s_texfilter.h b/src/mesa/swrast/s_texfilter.h
index 2e265d685c5..eceab59658e 100644
--- a/src/mesa/swrast/s_texfilter.h
+++ b/src/mesa/swrast/s_texfilter.h
@@ -27,7 +27,8 @@
 #define S_TEXFILTER_H
 
 
-#include "swrast.h"
+#include "main/mtypes.h"
+#include "s_context.h"
 
 
 extern texture_sample_func
diff --git a/src/mesa/swrast/s_zoom.h b/src/mesa/swrast/s_zoom.h
index 43917be65fc..09f624efad5 100644
--- a/src/mesa/swrast/s_zoom.h
+++ b/src/mesa/swrast/s_zoom.h
@@ -25,7 +25,8 @@
 #ifndef S_ZOOM_H
 #define S_ZOOM_H
 
-#include "swrast.h"
+#include "main/mtypes.h"
+#include "s_span.h"
 
 
 extern void
diff --git a/src/mesa/swrast_setup/ss_context.h b/src/mesa/swrast_setup/ss_context.h
index 1ec293fade1..56551ab273c 100644
--- a/src/mesa/swrast_setup/ss_context.h
+++ b/src/mesa/swrast_setup/ss_context.h
@@ -28,9 +28,8 @@
 #ifndef SS_CONTEXT_H
 #define SS_CONTEXT_H
 
-#include "main/mtypes.h"
+#include "main/glheader.h"
 #include "swrast/swrast.h"
-#include "swrast_setup.h"
 #include "tnl/t_context.h"
 
 typedef struct {
diff --git a/src/mesa/swrast_setup/ss_triangle.h b/src/mesa/swrast_setup/ss_triangle.h
index 007fa2e9141..ac553cbd018 100644
--- a/src/mesa/swrast_setup/ss_triangle.h
+++ b/src/mesa/swrast_setup/ss_triangle.h
@@ -29,7 +29,7 @@
 #ifndef SS_TRIANGLE_H
 #define SS_TRIANGLE_H
 
-#include "ss_context.h"
+#include "main/mtypes.h"
 
 
 void _swsetup_trifuncs_init( GLcontext *ctx );
diff --git a/src/mesa/swrast_setup/ss_vb.h b/src/mesa/swrast_setup/ss_vb.h
index 2ad1f56f396..944a3b78d8c 100644
--- a/src/mesa/swrast_setup/ss_vb.h
+++ b/src/mesa/swrast_setup/ss_vb.h
@@ -30,7 +30,6 @@
 #define SS_VB_H
 
 #include "main/mtypes.h"
-#include "swrast_setup.h"
 
 void _swsetup_vb_init( GLcontext *ctx );
 void _swsetup_choose_rastersetup_func( GLcontext *ctx );
diff --git a/src/mesa/tnl/t_context.h b/src/mesa/tnl/t_context.h
index ebaae6335b9..258906f7956 100644
--- a/src/mesa/tnl/t_context.h
+++ b/src/mesa/tnl/t_context.h
@@ -53,9 +53,7 @@
 #include "main/bitset.h"
 #include "main/mtypes.h"
 
-#include "math/m_matrix.h"
 #include "math/m_vector.h"
-#include "math/m_xform.h"
 
 #include "vbo/vbo.h"
 
diff --git a/src/mesa/tnl/t_rasterpos.c b/src/mesa/tnl/t_rasterpos.c
index 3596d162b23..d82d5b50736 100644
--- a/src/mesa/tnl/t_rasterpos.c
+++ b/src/mesa/tnl/t_rasterpos.c
@@ -25,7 +25,6 @@
 
 #include "main/glheader.h"
 #include "main/colormac.h"
-#include "main/context.h"
 #include "main/feedback.h"
 #include "main/light.h"
 #include "main/macros.h"
diff --git a/src/mesa/tnl/t_vb_cull.c b/src/mesa/tnl/t_vb_cull.c
index 712901acf30..22df7166735 100644
--- a/src/mesa/tnl/t_vb_cull.c
+++ b/src/mesa/tnl/t_vb_cull.c
@@ -28,7 +28,6 @@
 
 #include "main/glheader.h"
 #include "main/colormac.h"
-#include "main/context.h"
 #include "main/macros.h"
 #include "main/imports.h"
 #include "main/mtypes.h"
diff --git a/src/mesa/tnl/t_vb_fog.c b/src/mesa/tnl/t_vb_fog.c
index 4a0e6ad4f99..9faae24ec6d 100644
--- a/src/mesa/tnl/t_vb_fog.c
+++ b/src/mesa/tnl/t_vb_fog.c
@@ -28,7 +28,6 @@
 
 #include "main/glheader.h"
 #include "main/colormac.h"
-#include "main/context.h"
 #include "main/macros.h"
 #include "main/imports.h"
 #include "main/mtypes.h"
diff --git a/src/mesa/tnl/t_vb_normals.c b/src/mesa/tnl/t_vb_normals.c
index 61ac4095733..c2aa655674c 100644
--- a/src/mesa/tnl/t_vb_normals.c
+++ b/src/mesa/tnl/t_vb_normals.c
@@ -28,7 +28,6 @@
 
 #include "main/glheader.h"
 #include "main/colormac.h"
-#include "main/context.h"
 #include "main/macros.h"
 #include "main/imports.h"
 #include "main/mtypes.h"
diff --git a/src/mesa/tnl/t_vb_program.c b/src/mesa/tnl/t_vb_program.c
index 614c67d05eb..f3a338ef1ed 100644
--- a/src/mesa/tnl/t_vb_program.c
+++ b/src/mesa/tnl/t_vb_program.c
@@ -33,9 +33,9 @@
 
 #include "main/glheader.h"
 #include "main/colormac.h"
-#include "main/context.h"
 #include "main/macros.h"
 #include "main/imports.h"
+#include "math/m_xform.h"
 #include "program/prog_instruction.h"
 #include "program/prog_statevars.h"
 #include "program/prog_execute.h"
diff --git a/src/mesa/tnl/t_vb_render.c b/src/mesa/tnl/t_vb_render.c
index c1bebc99423..7d991009a14 100644
--- a/src/mesa/tnl/t_vb_render.c
+++ b/src/mesa/tnl/t_vb_render.c
@@ -44,6 +44,7 @@
 #include "main/macros.h"
 #include "main/imports.h"
 #include "main/mtypes.h"
+#include "math/m_xform.h"
 
 #include "t_pipeline.h"
 
diff --git a/src/mesa/tnl/t_vb_texgen.c b/src/mesa/tnl/t_vb_texgen.c
index 9ef13bc96d8..950e0f54e9f 100644
--- a/src/mesa/tnl/t_vb_texgen.c
+++ b/src/mesa/tnl/t_vb_texgen.c
@@ -37,7 +37,6 @@
 
 #include "main/glheader.h"
 #include "main/colormac.h"
-#include "main/context.h"
 #include "main/macros.h"
 #include "main/imports.h"
 #include "main/mtypes.h"
diff --git a/src/mesa/tnl/t_vb_texmat.c b/src/mesa/tnl/t_vb_texmat.c
index 83688290e59..985d137e5cc 100644
--- a/src/mesa/tnl/t_vb_texmat.c
+++ b/src/mesa/tnl/t_vb_texmat.c
@@ -28,7 +28,6 @@
 
 #include "main/glheader.h"
 #include "main/colormac.h"
-#include "main/context.h"
 #include "main/macros.h"
 #include "main/imports.h"
 #include "main/mtypes.h"
diff --git a/src/mesa/tnl/t_vb_vertex.c b/src/mesa/tnl/t_vb_vertex.c
index a2753425633..453479227b7 100644
--- a/src/mesa/tnl/t_vb_vertex.c
+++ b/src/mesa/tnl/t_vb_vertex.c
@@ -28,7 +28,6 @@
 
 #include "main/glheader.h"
 #include "main/colormac.h"
-#include "main/context.h"
 #include "main/macros.h"
 #include "main/imports.h"
 #include "main/mtypes.h"
diff --git a/src/mesa/vbo/vbo_exec_draw.c b/src/mesa/vbo/vbo_exec_draw.c
index 045af46da8d..84ae1b87f93 100644
--- a/src/mesa/vbo/vbo_exec_draw.c
+++ b/src/mesa/vbo/vbo_exec_draw.c
@@ -27,7 +27,7 @@
 
 #include "main/glheader.h"
 #include "main/bufferobj.h"
-#include "main/context.h"
+#include "main/compiler.h"
 #include "main/enums.h"
 #include "main/state.h"
 
diff --git a/src/mesa/vf/vf.h b/src/mesa/vf/vf.h
index 83d7547619c..5fe392bbe51 100644
--- a/src/mesa/vf/vf.h
+++ b/src/mesa/vf/vf.h
@@ -28,7 +28,7 @@
 #ifndef VF_VERTEX_H
 #define VF_VERTEX_H
 
-#include "main/mtypes.h"
+#include "main/glheader.h"
 #include "math/m_vector.h"
 
 enum {
diff --git a/src/mesa/vf/vf_generic.c b/src/mesa/vf/vf_generic.c
index 0af8893c302..95a317e99db 100644
--- a/src/mesa/vf/vf_generic.c
+++ b/src/mesa/vf/vf_generic.c
@@ -29,6 +29,7 @@
 #include "main/glheader.h"
 #include "main/context.h"
 #include "main/colormac.h"
+#include "main/macros.h"
 #include "main/simple_list.h"
 
 #include "vf/vf.h"