Merge branch 'softpipe_0_1_branch' of git+ssh://[email protected]/git/mesa/mesa into softpipe_0_1_branch

60 files changed, 1676 insertions, 1121 deletions

diff --git a/src/mesa/drivers/dri/common/dri_util.c b/src/mesa/drivers/dri/common/dri_util.c
index dd52f7e9151..c30e66f1722 100644
--- a/src/mesa/drivers/dri/common/dri_util.c
+++ b/src/mesa/drivers/dri/common/dri_util.c
@@ -428,10 +428,12 @@ __driUtilUpdateDrawableInfo(__DRIdrawablePrivate *pdp)
 
     if (pdp->pClipRects) {
 	_mesa_free(pdp->pClipRects); 
+	pdp->pClipRects = NULL;
     }
 
     if (pdp->pBackClipRects) {
 	_mesa_free(pdp->pBackClipRects); 
+	pdp->pBackClipRects = NULL;
     }
 
     DRM_SPINUNLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID);
diff --git a/src/mesa/drivers/dri/i915/i915_vtbl.c b/src/mesa/drivers/dri/i915/i915_vtbl.c
index b0e5f87fc75..cc8a605e505 100644
--- a/src/mesa/drivers/dri/i915/i915_vtbl.c
+++ b/src/mesa/drivers/dri/i915/i915_vtbl.c
@@ -64,6 +64,7 @@ static void i915_reduced_primitive_state( intelContextPtr intel,
     st1 &= ~ST1_ENABLE;
 
     switch (rprim) {
+    case GL_QUADS: /* from RASTERIZE(GL_QUADS) in t_dd_tritemp.h */
     case GL_TRIANGLES:
        if (intel->ctx.Polygon.StippleFlag &&
 	   intel->hw_stipple)
diff --git a/src/mesa/drivers/dri/i915/intel_pixel.c b/src/mesa/drivers/dri/i915/intel_pixel.c
index d175870a0c5..c3030d42b04 100644
--- a/src/mesa/drivers/dri/i915/intel_pixel.c
+++ b/src/mesa/drivers/dri/i915/intel_pixel.c
@@ -450,10 +450,16 @@ intelDrawPixels( GLcontext *ctx,
        * wise happily run the fragment program on each pixel in the image).
        */
       struct gl_fragment_program *fpSave = ctx->FragmentProgram._Current;
+   /* can't just set current frag prog to 0 here as on buffer resize
+      we'll get new state checks which will segfault. Remains a hack. */
       ctx->FragmentProgram._Current = NULL;
+      ctx->FragmentProgram._UseTexEnvProgram = GL_FALSE;
+      ctx->FragmentProgram._Active = GL_FALSE;
       _swrast_DrawPixels( ctx, x, y, width, height, format, type,
                           unpack, pixels );
       ctx->FragmentProgram._Current = fpSave;
+      ctx->FragmentProgram._UseTexEnvProgram = GL_TRUE;
+      ctx->FragmentProgram._Active = GL_TRUE;
    }
    else {
       _swrast_DrawPixels( ctx, x, y, width, height, format, type,
diff --git a/src/mesa/drivers/dri/i915tex/i915_vtbl.c b/src/mesa/drivers/dri/i915tex/i915_vtbl.c
index f80e8d6327f..0e7ca9586e8 100644
--- a/src/mesa/drivers/dri/i915tex/i915_vtbl.c
+++ b/src/mesa/drivers/dri/i915tex/i915_vtbl.c
@@ -61,6 +61,7 @@ i915_reduced_primitive_state(struct intel_context *intel, GLenum rprim)
    st1 &= ~ST1_ENABLE;
 
    switch (rprim) {
+   case GL_QUADS: /* from RASTERIZE(GL_QUADS) in t_dd_tritemp.h */
    case GL_TRIANGLES:
       if (intel->ctx.Polygon.StippleFlag && intel->hw_stipple)
          st1 |= ST1_ENABLE;
@@ -463,6 +464,7 @@ i915_state_draw_region(struct intel_context *intel,
     * Set stride/cpp values
     */
    if (color_region) {
+//      fprintf(stderr, "color pitch %d\n", color_region->pitch);
       state->Buffer[I915_DESTREG_CBUFADDR0] = _3DSTATE_BUF_INFO_CMD;
       state->Buffer[I915_DESTREG_CBUFADDR1] =
          (BUF_3D_ID_COLOR_BACK |
@@ -471,6 +473,7 @@ i915_state_draw_region(struct intel_context *intel,
    }
 
    if (depth_region) {
+//      fprintf(stderr, "depth pitch %d\n", depth_region->pitch);
       state->Buffer[I915_DESTREG_DBUFADDR0] = _3DSTATE_BUF_INFO_CMD;
       state->Buffer[I915_DESTREG_DBUFADDR1] =
          (BUF_3D_ID_DEPTH |
diff --git a/src/mesa/drivers/dri/i915tex/intel_blit.c b/src/mesa/drivers/dri/i915tex/intel_blit.c
index dbe4ba2ac5e..c755eac6b2c 100644
--- a/src/mesa/drivers/dri/i915tex/intel_blit.c
+++ b/src/mesa/drivers/dri/i915tex/intel_blit.c
@@ -79,15 +79,25 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv,
     */
    LOCK_HARDWARE(intel);
 
+      if (intel->revalidateDrawable) {
+	 __DRIscreenPrivate *sPriv = intel->driScreen;
+	 if (dPriv) {
+	    DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv);
+	 }
+      }
+
    if (dPriv && dPriv->numClipRects) {
       struct intel_framebuffer *intel_fb = dPriv->driverPrivate;
       const struct intel_region *frontRegion
-	 = intel_get_rb_region(&intel_fb->Base, BUFFER_FRONT_LEFT);
+	 = intelScreen->front_region;
       const struct intel_region *backRegion
-	 = intel_get_rb_region(&intel_fb->Base, BUFFER_BACK_LEFT);
+	 = intel->ctx.DrawBuffer->_ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT ?
+	   intel_get_rb_region(&intel_fb->Base, BUFFER_FRONT_LEFT) :
+	   intel_get_rb_region(&intel_fb->Base, BUFFER_BACK_LEFT);
       const int nbox = dPriv->numClipRects;
       const drm_clip_rect_t *pbox = dPriv->pClipRects;
       const int pitch = frontRegion->pitch;
+      const int srcpitch = backRegion->pitch;
       const int cpp = frontRegion->cpp;
       int BR13, CMD;
       int i;
@@ -96,9 +106,12 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv,
       ASSERT(intel_fb->Base.Name == 0);    /* Not a user-created FBO */
       ASSERT(frontRegion);
       ASSERT(backRegion);
-      ASSERT(frontRegion->pitch == backRegion->pitch);
+//      ASSERT(frontRegion->pitch == backRegion->pitch);
       ASSERT(frontRegion->cpp == backRegion->cpp);
 
+      DBG("front pitch %d back pitch %d\n",
+	 frontRegion->pitch, backRegion->pitch);
+
       if (cpp == 2) {
 	 BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24);
 	 CMD = XY_SRC_COPY_BLT_CMD;
@@ -111,6 +124,7 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv,
 
       for (i = 0; i < nbox; i++, pbox++) {
 	 drm_clip_rect_t box;
+	 drm_clip_rect_t sbox;
 
 	 if (pbox->x1 > pbox->x2 ||
 	     pbox->y1 > pbox->y2 ||
@@ -133,6 +147,14 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv,
 	       continue;
 	 }
 
+	 DBG("box x1 x2 y1 y2 %d %d %d %d\n",
+	      box.x1, box.x2, box.y1, box.y2);
+
+	 /* XXX should make sure only the minimum area based on
+	    old draw buffer and new front clip rects is copied */
+	 sbox.x1 = box.x1 - dPriv->x;
+	 sbox.y1 = box.y1 - dPriv->y;
+
 	 BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS);
 	 OUT_BATCH(CMD);
 	 OUT_BATCH(BR13);
@@ -141,8 +163,8 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv,
 
 	 OUT_RELOC(frontRegion->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE,
 		   DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0);
-	 OUT_BATCH((pbox->y1 << 16) | pbox->x1);
-	 OUT_BATCH(BR13 & 0xffff);
+	 OUT_BATCH((sbox.y1 << 16) | sbox.x1);
+	 OUT_BATCH((srcpitch * cpp) & 0xffff);
 	 OUT_RELOC(backRegion->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ,
 		   DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0);
 
@@ -156,6 +178,12 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv,
    }
 
    UNLOCK_HARDWARE(intel);
+
+   if (intel->revalidateDrawable) {
+      intel->revalidateDrawable = GL_FALSE;
+      intelWindowMoved(intel);
+   }
+
 }
 
 
@@ -409,9 +437,8 @@ intelClearWithBlit(GLcontext * ctx, GLbitfield mask)
             b = *box;
          }
 
-         if (0)
-            _mesa_printf("clear %d,%d..%d,%d, mask %x\n",
-                         b.x1, b.y1, b.x2, b.y2, mask);
+         DBG("clear %d,%d..%d,%d, mask %x\n",
+                      b.x1, b.y1, b.x2, b.y2, mask);
 
          /* Loop over all renderbuffers */
          for (buf = 0; buf < BUFFER_COUNT && clearMask; buf++) {
diff --git a/src/mesa/drivers/dri/i915tex/intel_buffers.c b/src/mesa/drivers/dri/i915tex/intel_buffers.c
index 62ff54b0075..17b4feadf85 100644
--- a/src/mesa/drivers/dri/i915tex/intel_buffers.c
+++ b/src/mesa/drivers/dri/i915tex/intel_buffers.c
@@ -146,57 +146,30 @@ intelSetRenderbufferClipRects(struct intel_context *intel)
    intel->drawY = 0;
 }
 
-
 /**
- * As above, but for rendering to front buffer of a window.
- * \sa intelSetRenderbufferClipRects
+ * As above, but for rendering private front/back buffer of a window.
+ * \sa intelSetPrivbufClipRects
  */
-static void
-intelSetFrontClipRects(struct intel_context *intel)
-{
-   __DRIdrawablePrivate *dPriv = intel->driDrawable;
-
-   if (!dPriv)
-      return;
-
-   intel->numClipRects = dPriv->numClipRects;
-   intel->pClipRects = dPriv->pClipRects;
-   intel->drawX = dPriv->x;
-   intel->drawY = dPriv->y;
-}
 
-
-/**
- * As above, but for rendering to back buffer of a window.
- */
 static void
-intelSetBackClipRects(struct intel_context *intel)
+intelSetPrivbufClipRects(struct intel_context *intel)
 {
    __DRIdrawablePrivate *dPriv = intel->driDrawable;
-   struct intel_framebuffer *intel_fb;
-
    if (!dPriv)
       return;
 
-   intel_fb = dPriv->driverPrivate;
-
-   if (intel_fb->pf_active || dPriv->numBackClipRects == 0) {
-      /* use the front clip rects */
-      intel->numClipRects = dPriv->numClipRects;
-      intel->pClipRects = dPriv->pClipRects;
-      intel->drawX = dPriv->x;
-      intel->drawY = dPriv->y;
-   }
-   else {
-      /* use the back clip rects */
-      intel->numClipRects = dPriv->numBackClipRects;
-      intel->pClipRects = dPriv->pBackClipRects;
-      intel->drawX = dPriv->backX;
-      intel->drawY = dPriv->backY;
-   }
+   intel->fakeClipRect.x1 = 0;
+   intel->fakeClipRect.y1 = 0;
+   intel->fakeClipRect.x2 = dPriv->w;
+   intel->fakeClipRect.y2 = dPriv->h;
+   intel->numClipRects = 1;
+   intel->pClipRects = &intel->fakeClipRect;
+   intel->drawX = 0;
+   intel->drawY = 0;
 }
 
 
+
 /**
  * This will be called whenever the currently bound window is moved/resized.
  * XXX: actually, it seems to NOT be called when the window is only moved (BP).
@@ -210,27 +183,10 @@ intelWindowMoved(struct intel_context *intel)
 
    if (!intel->ctx.DrawBuffer) {
       /* when would this happen? -BP */
-      intelSetFrontClipRects(intel);
-   }
-   else if (intel->ctx.DrawBuffer->Name != 0) {
-      /* drawing to user-created FBO - do nothing */
-      /* Cliprects would be set from intelDrawBuffer() */
-   }
-   else {
-      /* drawing to a window */
-      switch (intel_fb->Base._ColorDrawBufferMask[0]) {
-      case BUFFER_BIT_FRONT_LEFT:
-         intelSetFrontClipRects(intel);
-         break;
-      case BUFFER_BIT_BACK_LEFT:
-         intelSetBackClipRects(intel);
-         break;
-      default:
-         /* glDrawBuffer(GL_NONE or GL_FRONT_AND_BACK): software fallback */
-         intelSetFrontClipRects(intel);
-      }
+      intel->numClipRects = 0;
    }
 
+
    if (intel->intelScreen->driScrnPriv->ddxMinor >= 7) {
       drmI830Sarea *sarea = intel->sarea;
       drm_clip_rect_t drw_rect = { .x1 = dPriv->x, .x2 = dPriv->x + dPriv->w,
@@ -316,7 +272,8 @@ intelWindowMoved(struct intel_context *intel)
 	 flags = intel_fb->vblank_flags & ~VBLANK_FLAG_SECONDARY;
       }
 
-      if (flags != intel_fb->vblank_flags) {
+      if (flags != intel_fb->vblank_flags && intel_fb->vblank_flags &&
+	  !(intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ)) {
 	 drmVBlank vbl;
 	 int i;
 
@@ -327,7 +284,9 @@ intelWindowMoved(struct intel_context *intel)
 	 }
 
 	 for (i = 0; i < intel_fb->pf_num_pages; i++) {
-	    if (!intel_fb->color_rb[i])
+	    if (!intel_fb->color_rb[i] ||
+		(intel_fb->vbl_waited - intel_fb->color_rb[i]->vbl_pending) <=
+		(1<<23))
 	       continue;
 
 	    vbl.request.sequence = intel_fb->color_rb[i]->vbl_pending;
@@ -479,6 +438,7 @@ void
 intelRotateWindow(struct intel_context *intel,
                   __DRIdrawablePrivate * dPriv, GLuint srcBuf)
 {
+
    intelScreenPrivate *screen = intel->intelScreen;
    drm_clip_rect_t fullRect;
    struct intel_framebuffer *intel_fb;
@@ -828,7 +788,8 @@ intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target)
    drm_i915_vblank_swap_t swap;
    GLboolean ret;
 
-   if ((intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ) ||
+   if (!intel_fb->vblank_flags ||
+       (intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ) ||
        intelScreen->current_rotation != 0 ||
        intelScreen->drmMinor < (intel_fb->pf_active ? 9 : 6))
       return GL_FALSE;
@@ -909,7 +870,7 @@ intelSwapBuffers(__DRIdrawablePrivate * dPriv)
 	 GLboolean missed_target;
 	 struct intel_framebuffer *intel_fb = dPriv->driverPrivate;
 	 int64_t ust;
-         
+
 	 _mesa_notifySwapBuffers(ctx);  /* flush pending rendering comands */
 
          if (screen->current_rotation != 0 ||
@@ -953,10 +914,17 @@ intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h)
 
       if (ctx->Visual.doubleBufferMode) {
          drm_clip_rect_t rect;
+#if 1
          rect.x1 = x + dPriv->x;
          rect.y1 = (dPriv->h - y - h) + dPriv->y;
          rect.x2 = rect.x1 + w;
          rect.y2 = rect.y1 + h;
+#else
+         rect.x1 = x;
+         rect.y1 = dPriv->h - y;
+         rect.x2 = rect.x1 + w;
+         rect.y2 = rect.y1 + h;
+#endif
          _mesa_notifySwapBuffers(ctx);  /* flush pending rendering comands */
          intelCopyBuffer(dPriv, &rect);
       }
@@ -991,7 +959,7 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb)
       return;
    }
 
-   /* Do this here, note core Mesa, since this function is called from
+   /* Do this here, not core Mesa, since this function is called from
     * many places within the driver.
     */
    if (ctx->NewState & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) {
@@ -1015,12 +983,7 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb)
    /*
     * How many color buffers are we drawing into?
     */
-   if (fb->_NumColorDrawBuffers[0] != 1
-#if 0
-       /* XXX FBO temporary - always use software rendering */
-       || 1
-#endif
-      ) {
+   if (fb->_NumColorDrawBuffers[0] != 1) {
       /* writing to 0 or 2 or 4 color buffers */
       /*_mesa_debug(ctx, "Software rendering\n");*/
       FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_TRUE);
@@ -1040,13 +1003,12 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb)
     * And set up cliprects.
     */
    if (fb->Name == 0) {
+      intelSetPrivbufClipRects(intel);
       /* drawing to window system buffer */
       if (front) {
-         intelSetFrontClipRects(intel);
          colorRegion = intel_get_rb_region(fb, BUFFER_FRONT_LEFT);
       }
       else {
-         intelSetBackClipRects(intel);
          colorRegion = intel_get_rb_region(fb, BUFFER_BACK_LEFT);
       }
    }
diff --git a/src/mesa/drivers/dri/i915tex/intel_context.c b/src/mesa/drivers/dri/i915tex/intel_context.c
index 6288bb1cd09..61842da29d9 100644
--- a/src/mesa/drivers/dri/i915tex/intel_context.c
+++ b/src/mesa/drivers/dri/i915tex/intel_context.c
@@ -289,18 +289,22 @@ intelFlush(GLcontext * ctx)
  * Check if we need to rotate/warp the front color buffer to the
  * rotated screen.  We generally need to do this when we get a glFlush
  * or glFinish after drawing to the front color buffer.
+ * If no rotation, just copy the private fake front buffer to the real one.
  */
 static void
-intelCheckFrontRotate(GLcontext * ctx)
+intelCheckFrontUpdate(GLcontext * ctx)
 {
    struct intel_context *intel = intel_context(ctx);
    if (intel->ctx.DrawBuffer->_ColorDrawBufferMask[0] ==
        BUFFER_BIT_FRONT_LEFT) {
       intelScreenPrivate *screen = intel->intelScreen;
+      __DRIdrawablePrivate *dPriv = intel->driDrawable;
       if (screen->current_rotation != 0) {
-         __DRIdrawablePrivate *dPriv = intel->driDrawable;
          intelRotateWindow(intel, dPriv, BUFFER_BIT_FRONT_LEFT);
       }
+      else {
+         intelCopyBuffer(dPriv, NULL);
+      }
    }
 }
 
@@ -312,7 +316,7 @@ static void
 intelglFlush(GLcontext * ctx)
 {
    intelFlush(ctx);
-   intelCheckFrontRotate(ctx);
+   intelCheckFrontUpdate(ctx);
 }
 
 void
@@ -326,7 +330,7 @@ intelFinish(GLcontext * ctx)
       driFenceUnReference(intel->batch->last_fence);
       intel->batch->last_fence = NULL;
    }
-   intelCheckFrontRotate(ctx);
+   intelCheckFrontUpdate(ctx);
 }
 
 
@@ -581,6 +585,13 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv,
                  __DRIdrawablePrivate * driReadPriv)
 {
 
+#if 0
+   if (driDrawPriv) {
+      fprintf(stderr, "x %d, y %d, width %d, height %d\n",
+	      driDrawPriv->x, driDrawPriv->y, driDrawPriv->w, driDrawPriv->h);
+   }
+#endif
+
    if (driContextPriv) {
       struct intel_context *intel =
          (struct intel_context *) driContextPriv->driverPrivate;
@@ -588,6 +599,9 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv,
 	 (struct intel_framebuffer *) driDrawPriv->driverPrivate;
       GLframebuffer *readFb = (GLframebuffer *) driReadPriv->driverPrivate;
 
+      /* this is a hack so we have a valid context when the region allocation
+         is done. Need a per-screen context? */
+      intel->intelScreen->dummyctxptr = intel;
 
       /* XXX FBO temporary fix-ups! */
       /* if the renderbuffers don't have regions, init them from the context */
@@ -625,18 +639,30 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv,
       }
 
       _mesa_make_current(&intel->ctx, &intel_fb->Base, readFb);
+      intel->intelScreen->dummyctxptr = &intel->ctx;
 
       /* The drawbuffer won't always be updated by _mesa_make_current: 
        */
       if (intel->ctx.DrawBuffer == &intel_fb->Base) {
 
 	 if (intel->driDrawable != driDrawPriv) {
-	    intel_fb->vblank_flags = (intel->intelScreen->irq_active != 0)
-	       ? driGetDefaultVBlankFlags(&intel->optionCache)
-	       : VBLANK_FLAG_NO_IRQ;
-	    (*dri_interface->getUST) (&intel_fb->swap_ust);
-	    driDrawableInitVBlank(driDrawPriv, intel_fb->vblank_flags,
-				  &intel_fb->vbl_seq);
+	    if (driDrawPriv->pdraw->swap_interval == (unsigned)-1) {
+	       int i;
+
+	       intel_fb->vblank_flags = (intel->intelScreen->irq_active != 0)
+		  ? driGetDefaultVBlankFlags(&intel->optionCache)
+		 : VBLANK_FLAG_NO_IRQ;
+
+	       (*dri_interface->getUST) (&intel_fb->swap_ust);
+	       driDrawableInitVBlank(driDrawPriv, intel_fb->vblank_flags,
+				     &intel_fb->vbl_seq);
+	       intel_fb->vbl_waited = intel_fb->vbl_seq;
+
+	       for (i = 0; i < (intel->intelScreen->third.handle ? 3 : 2); i++) {
+		  if (intel_fb->color_rb[i])
+		     intel_fb->color_rb[i]->vbl_pending = intel_fb->vbl_seq;
+	       }
+	    }
 	    intel->driDrawable = driDrawPriv;
 	    intelWindowMoved(intel);
 	 }
@@ -670,7 +696,8 @@ intelContendedLock(struct intel_context *intel, GLuint flags)
     * checking must be done *after* this call:
     */
    if (dPriv)
-      DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv);
+      intel->revalidateDrawable = GL_TRUE;
+//      DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv);
 
    if (sarea->width != intelScreen->width ||
        sarea->height != intelScreen->height ||
@@ -679,40 +706,31 @@ intelContendedLock(struct intel_context *intel, GLuint flags)
       intelUpdateScreenRotation(sPriv, sarea);
    }
 
+#if 0
    if (sarea->width != intel->width ||
        sarea->height != intel->height ||
        sarea->rotation != intel->current_rotation) {
-      
-      void *batchMap = intel->batch->map;
-      
+      int numClipRects = intel->numClipRects;
+
       /*
        * FIXME: Really only need to do this when drawing to a
        * common back- or front buffer.
        */
 
       /*
-       * This will drop the outstanding batchbuffer on the floor
+       * This will essentially drop the outstanding batchbuffer on the floor.
        */
+      intel->numClipRects = 0;
 
-      if (batchMap != NULL) {
-	 driBOUnmap(intel->batch->buffer);
-	 intel->batch->map = NULL;
-      }
-
-      intel_batchbuffer_reset(intel->batch);
+      if (intel->Fallback)
+	 _swrast_flush(&intel->ctx);
 
-      if (batchMap == NULL) {
-	 driBOUnmap(intel->batch->buffer);
-	 intel->batch->map = NULL;
-      }
+      INTEL_FIREVERTICES(intel);
 
-      /* lose all primitives */
-      intel->prim.primitive = ~0;
-      intel->prim.start_ptr = 0;
-      intel->prim.flush = 0;
+      if (intel->batch->map != intel->batch->ptr)
+	 intel_batchbuffer_flush(intel->batch);
 
-      /* re-emit all state */
-      intel->vtbl.lost_hardware(intel);
+      intel->numClipRects = numClipRects;
 
       /* force window update */
       intel->lastStamp = 0;
@@ -721,13 +739,7 @@ intelContendedLock(struct intel_context *intel, GLuint flags)
       intel->height = sarea->height;
       intel->current_rotation = sarea->rotation;
    }
-
-   /* Drawable changed?
-    */
-   if (dPriv && intel->lastStamp != dPriv->lastStamp) {
-      intelWindowMoved(intel);
-      intel->lastStamp = dPriv->lastStamp;
-   }
+#endif
 }
 
 
@@ -753,7 +765,9 @@ void LOCK_HARDWARE( struct intel_context *intel )
 				    BUFFER_BACK_LEFT);
     }
 
-    if (intel_rb && (intel_fb->vbl_waited - intel_rb->vbl_pending) > (1<<23)) {
+    if (intel_rb && intel_fb->vblank_flags &&
+	!(intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ) &&
+	(intel_fb->vbl_waited - intel_rb->vbl_pending) > (1<<23)) {
 	drmVBlank vbl;
 
 	vbl.request.type = DRM_VBLANK_ABSOLUTE;
diff --git a/src/mesa/drivers/dri/i915tex/intel_context.h b/src/mesa/drivers/dri/i915tex/intel_context.h
index 9d060eb866f..7b97c564a3e 100644
--- a/src/mesa/drivers/dri/i915tex/intel_context.h
+++ b/src/mesa/drivers/dri/i915tex/intel_context.h
@@ -254,6 +254,7 @@ struct intel_context
    GLuint numClipRects;         /**< cliprects for drawing */
    drm_clip_rect_t *pClipRects;
    drm_clip_rect_t fboRect;     /**< cliprect for FBO rendering */
+   drm_clip_rect_t fakeClipRect;     /**< cliprect for priv back/fake front buffers rendering */
 
    int perf_boxes;
 
@@ -272,6 +273,7 @@ struct intel_context
    drmI830Sarea *sarea;
 
    GLuint lastStamp;
+   GLuint revalidateDrawable;
 
    /**
     * Configuration cache
@@ -353,7 +355,7 @@ __memcpy(void *to, const void *from, size_t n)
 /* ================================================================
  * Debugging:
  */
-#define DO_DEBUG		0
+#define DO_DEBUG		1
 #if DO_DEBUG
 extern int INTEL_DEBUG;
 #else
diff --git a/src/mesa/drivers/dri/i915tex/intel_fbo.c b/src/mesa/drivers/dri/i915tex/intel_fbo.c
index 349912ffecd..2dd3617f93a 100644
--- a/src/mesa/drivers/dri/i915tex/intel_fbo.c
+++ b/src/mesa/drivers/dri/i915tex/intel_fbo.c
@@ -176,7 +176,7 @@ intel_get_pointer(GLcontext * ctx, struct gl_renderbuffer *rb,
 
 /**
  * Called via glRenderbufferStorageEXT() to set the format and allocate
- * storage for a user-created renderbuffer.
+ * storage for a user-created (or priv buffer) renderbuffer.
  */
 static GLboolean
 intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
@@ -188,8 +188,6 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
    GLboolean softwareBuffer = GL_FALSE;
    int cpp;
 
-   ASSERT(rb->Name != 0);
-
    switch (internalFormat) {
    case GL_R3_G3_B2:
    case GL_RGB4:
@@ -257,7 +255,7 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
       break;
    default:
       _mesa_problem(ctx,
-                    "Unexpected format in intel_alloc_renderbuffer_storage");
+                    "Unexpected format (%x) in intel_alloc_renderbuffer_storage", internalFormat);
       return GL_FALSE;
    }
 
@@ -453,6 +451,29 @@ intel_create_renderbuffer(GLenum intFormat, GLsizei width, GLsizei height,
    return irb;
 }
 
+struct gl_renderbuffer *
+intel_new_renderbuffer_fb(GLcontext * ctx, GLuint intFormat)
+{
+   struct intel_renderbuffer *irb;
+
+   irb = CALLOC_STRUCT(intel_renderbuffer);
+   if (!irb) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "creating renderbuffer");
+      return NULL;
+   }
+
+   _mesa_init_renderbuffer(&irb->Base, 0);
+   irb->Base.ClassID = INTEL_RB_CLASS;
+   irb->Base.InternalFormat = intFormat;
+
+   /* intel-specific methods */
+   irb->Base.Delete = intel_delete_renderbuffer;
+   irb->Base.AllocStorage = intel_alloc_renderbuffer_storage;
+   irb->Base.GetPointer = intel_get_pointer;
+   /* span routines set in alloc_storage function */
+
+   return &irb->Base;
+}
 
 /**
  * Create a new renderbuffer object.
diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c b/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c
index 77c67c821eb..e4e57cb3a7e 100644
--- a/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c
+++ b/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c
@@ -370,10 +370,16 @@ intelDrawPixels(GLcontext * ctx,
        * wise happily run the fragment program on each pixel in the image).
        */
       struct gl_fragment_program *fpSave = ctx->FragmentProgram._Current;
+   /* can't just set current frag prog to 0 here as on buffer resize
+      we'll get new state checks which will segfault. Remains a hack. */
       ctx->FragmentProgram._Current = NULL;
+      ctx->FragmentProgram._UseTexEnvProgram = GL_FALSE;
+      ctx->FragmentProgram._Active = GL_FALSE;
       _swrast_DrawPixels( ctx, x, y, width, height, format, type,
                           unpack, pixels );
       ctx->FragmentProgram._Current = fpSave;
+      ctx->FragmentProgram._UseTexEnvProgram = GL_TRUE;
+      ctx->FragmentProgram._Active = GL_TRUE;
    }
    else {
       _swrast_DrawPixels( ctx, x, y, width, height, format, type,
diff --git a/src/mesa/drivers/dri/i915tex/intel_screen.c b/src/mesa/drivers/dri/i915tex/intel_screen.c
index 2acdead63d5..08057a66006 100644
--- a/src/mesa/drivers/dri/i915tex/intel_screen.c
+++ b/src/mesa/drivers/dri/i915tex/intel_screen.c
@@ -88,6 +88,7 @@ intelMapScreenRegions(__DRIscreenPrivate * sPriv)
       _mesa_warning(NULL, "no front buffer handle in intelMapScreenRegions!");
    }
 
+#if 0
    if (0)
       _mesa_printf("Back 0x%08x ", intelScreen->back.handle);
    if (drmMap(sPriv->fd,
@@ -119,6 +120,7 @@ intelMapScreenRegions(__DRIscreenPrivate * sPriv)
       intelUnmapScreenRegions(intelScreen);
       return GL_FALSE;
    }
+#endif
 
 #if 0
    _mesa_printf("TEX 0x%08x ", intelScreen->tex.handle);
@@ -156,7 +158,7 @@ intel_recreate_static(intelScreenPrivate *intelScreen,
   }
   return region;
 }
-    
+
 
 /* Create intel_region structs to describe the static front,back,depth
  * buffers created by the xserver. 
@@ -172,6 +174,7 @@ intel_recreate_static(intelScreenPrivate *intelScreen,
 static void
 intel_recreate_static_regions(intelScreenPrivate *intelScreen)
 {
+/* this is the real front buffer which is only used for blitting to */
    intelScreen->front_region =
       intel_recreate_static(intelScreen,
 			    intelScreen->front_region,
@@ -192,7 +195,7 @@ intel_recreate_static_regions(intelScreenPrivate *intelScreen)
 			    intelScreen->rotated.pitch /
 			    intelScreen->cpp, intelScreen->height);
 
-
+#if 0
    intelScreen->back_region =
       intel_recreate_static(intelScreen,
 			    intelScreen->back_region,
@@ -226,6 +229,7 @@ intel_recreate_static_regions(intelScreenPrivate *intelScreen)
 			    intelScreen->cpp,
 			    intelScreen->depth.pitch / intelScreen->cpp,
 			    intelScreen->height);
+#endif
 }
 
 /**
@@ -382,7 +386,7 @@ intelUpdateScreenFromSAREA(intelScreenPrivate * intelScreen,
    intelScreen->rotatedWidth = sarea->virtualX;
    intelScreen->rotatedHeight = sarea->virtualY;
 
-   if (0)
+   if (1)
       intelPrintSAREA(sarea);
 }
 
@@ -600,6 +604,7 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv,
 
       _mesa_initialize_framebuffer(&intel_fb->Base, mesaVis);
 
+#if 0
       /* setup the hardware-based renderbuffers */
       {
          intel_fb->color_rb[0]
@@ -640,7 +645,6 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv,
 	    _mesa_reference_renderbuffer(&tmp_rb, &intel_fb->color_rb[2]->Base);
 	 }
       }
-
       if (mesaVis->depthBits == 24 && mesaVis->stencilBits == 8) {
          /* combined depth/stencil buffer */
          struct intel_renderbuffer *depthStencilRb
@@ -670,6 +674,50 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv,
          _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, &depthRb->Base);
       }
 
+#else
+      {
+	 /* fake frontbuffer */
+	 /* XXX allocation should only happen in the unusual case
+            it's actually needed */
+         intel_fb->color_rb[0]
+            = intel_new_renderbuffer_fb(NULL, rgbFormat);
+         _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT,
+				&intel_fb->color_rb[0]->Base);
+      }
+
+      if (mesaVis->doubleBufferMode) {
+         intel_fb->color_rb[1]
+            = intel_new_renderbuffer_fb(NULL, rgbFormat);
+         _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT,
+				&intel_fb->color_rb[1]->Base);
+
+	 if (screen->third.handle) {
+	    struct gl_renderbuffer *tmp_rb = NULL;
+
+	    intel_fb->color_rb[2]
+	       = intel_new_renderbuffer_fb(NULL, rgbFormat);
+	    _mesa_reference_renderbuffer(&tmp_rb, &intel_fb->color_rb[2]->Base);
+	 }
+      }
+      if (mesaVis->depthBits == 24 && mesaVis->stencilBits == 8) {
+         /* combined depth/stencil buffer */
+         struct intel_renderbuffer *depthStencilRb
+            = intel_new_renderbuffer_fb(NULL, GL_DEPTH24_STENCIL8_EXT);
+         /* note: bind RB to two attachment points */
+         _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH,
+				&depthStencilRb->Base);
+         _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_STENCIL,
+				&depthStencilRb->Base);
+      }
+      else if (mesaVis->depthBits == 16) {
+         /* just 16-bit depth buffer, no hw stencil */
+         struct intel_renderbuffer *depthRb
+            = intel_new_renderbuffer_fb(NULL, GL_DEPTH_COMPONENT16);
+         _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, &depthRb->Base);
+      }
+
+#endif
+
       /* now add any/all software-based renderbuffers we may need */
       _mesa_add_soft_renderbuffers(&intel_fb->Base,
                                    GL_FALSE, /* never sw color */
@@ -939,11 +987,18 @@ struct intel_context *intelScreenContext(intelScreenPrivate *intelScreen)
    * context at screen creation. For now just use the current context.
    */
 
-  GET_CURRENT_CONTEXT(ctx);
+/*  GET_CURRENT_CONTEXT(ctx);
   if (ctx == NULL) {
      _mesa_problem(NULL, "No current context in intelScreenContext\n");
      return NULL;
   }
   return intel_context(ctx);
+*/
+  if (intelScreen->dummyctxptr == NULL) {
+     _mesa_problem(NULL, "No current context in intelScreenContext\n");
+     return NULL;
+  }
+  return intelScreen->dummyctxptr;
+
 }
 
diff --git a/src/mesa/drivers/dri/i915tex/intel_screen.h b/src/mesa/drivers/dri/i915tex/intel_screen.h
index bac43aadddf..783ee35524b 100644
--- a/src/mesa/drivers/dri/i915tex/intel_screen.h
+++ b/src/mesa/drivers/dri/i915tex/intel_screen.h
@@ -96,6 +96,7 @@ typedef struct
    struct _DriBufferPool *staticPool;
    unsigned int maxBatchSize;
    GLboolean havePools;
+   struct intel_context *dummyctxptr;
 } intelScreenPrivate;
 
 
diff --git a/src/mesa/drivers/dri/i915tex/intel_state.c b/src/mesa/drivers/dri/i915tex/intel_state.c
index 271511037e9..b2773990e4b 100644
--- a/src/mesa/drivers/dri/i915tex/intel_state.c
+++ b/src/mesa/drivers/dri/i915tex/intel_state.c
@@ -205,25 +205,17 @@ intelCalcViewport(GLcontext * ctx)
    GLfloat *m = intel->ViewportMatrix.m;
    GLfloat yScale, yBias;
 
-   if (ctx->DrawBuffer->Name) {
-      /* User created FBO */
-      struct intel_renderbuffer *irb
+   struct intel_renderbuffer *irb
          = intel_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0][0]);
-      if (irb && !irb->RenderToTexture) {
-         /* y=0=top */
-         yScale = -1.0;
-         yBias = irb->Base.Height;
-      }
-      else {
-         /* y=0=bottom */
-         yScale = 1.0;
-         yBias = 0.0;
-      }
+   if (irb && !irb->RenderToTexture) {
+      /* y=0=top */
+      yScale = -1.0;
+      yBias = irb->Base.Height;
    }
    else {
-      /* window buffer, y=0=top */
-      yScale = -1.0;
-      yBias = (intel->driDrawable) ? intel->driDrawable->h : 0.0F;
+      /* y=0=bottom */
+      yScale = 1.0;
+      yBias = 0.0;
    }
 
    m[MAT_SX] = v[MAT_SX];
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_buffers.c b/src/mesa/drivers/dri/nouveau/nouveau_buffers.c
index 857cd30584c..f98d6665635 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_buffers.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_buffers.c
@@ -69,7 +69,7 @@ void
 nouveau_mem_free(GLcontext *ctx, nouveau_mem *mem)
 {
    nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
-   drm_nouveau_mem_free_t memf;
+   struct drm_nouveau_mem_free memf;
 
    if (NOUVEAU_DEBUG & DEBUG_MEM)  {
       fprintf(stderr, "%s: type=0x%x, offset=0x%x, size=0x%x\n",
@@ -78,8 +78,8 @@ nouveau_mem_free(GLcontext *ctx, nouveau_mem *mem)
 
    if (mem->map)
       drmUnmap(mem->map, mem->size);
-   memf.flags         = mem->type;
-   memf.region_offset = mem->offset;
+   memf.flags  = mem->type;
+   memf.offset = mem->offset;
    drmCommandWrite(nmesa->driFd, DRM_NOUVEAU_MEM_FREE, &memf, sizeof(memf));
    FREE(mem);
 }
@@ -88,7 +88,7 @@ nouveau_mem *
 nouveau_mem_alloc(GLcontext *ctx, int type, GLuint size, GLuint align)
 {
    nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
-   drm_nouveau_mem_alloc_t mema;
+   struct drm_nouveau_mem_alloc mema;
    nouveau_mem *mem;
    int ret;
 
@@ -111,7 +111,7 @@ nouveau_mem_alloc(GLcontext *ctx, int type, GLuint size, GLuint align)
       FREE(mem);
       return NULL;
    }
-   mem->offset = mema.region_offset;
+   mem->offset = mema.offset;
    mem->type   = mema.flags;
 
    if (NOUVEAU_DEBUG & DEBUG_MEM)  {
@@ -120,7 +120,7 @@ nouveau_mem_alloc(GLcontext *ctx, int type, GLuint size, GLuint align)
    }
 
    if (type & NOUVEAU_MEM_MAPPED)
-      ret = drmMap(nmesa->driFd, mem->offset, mem->size, &mem->map);
+      ret = drmMap(nmesa->driFd, mema.map_handle, mem->size, &mem->map);
    if (ret) {
       mem->map = NULL;
       nouveau_mem_free(ctx, mem);
@@ -135,12 +135,7 @@ nouveau_mem_gpu_offset_get(GLcontext *ctx, nouveau_mem *mem)
 {
    nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
 
-   if (mem->type & NOUVEAU_MEM_FB)
-      return (uint32_t)mem->offset - nmesa->vram_phys;
-   else if (mem->type & NOUVEAU_MEM_AGP)
-      return (uint32_t)mem->offset - nmesa->gart_phys;
-   else
-      return 0xDEADF00D;
+   return mem->offset;
 }
 
 static GLboolean
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_context.c b/src/mesa/drivers/dri/nouveau/nouveau_context.c
index d96b00242cd..44392c0267c 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_context.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_context.c
@@ -180,7 +180,7 @@ GLboolean nouveauCreateContext( const __GLcontextModes *glVisual,
 	driParseConfigFiles (&nmesa->optionCache, &screen->optionCache,
 			screen->driScreen->myNum, "nouveau");
 
-	nmesa->sarea = (drm_nouveau_sarea_t *)((char *)sPriv->pSAREA +
+	nmesa->sarea = (struct drm_nouveau_sarea *)((char *)sPriv->pSAREA +
 			screen->sarea_priv_offset);
 
 	/* Enable any supported extensions */
@@ -224,6 +224,8 @@ GLboolean nouveauCreateContext( const __GLcontextModes *glVisual,
 			nv04TriInitFunctions( ctx );
 			break;
 		case NV_10:
+		case NV_11:
+		case NV_17:
 		case NV_20:
 		case NV_30:
 		case NV_40:
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_context.h b/src/mesa/drivers/dri/nouveau/nouveau_context.h
index 10d2ed6e172..9a0be2cb2a5 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_context.h
+++ b/src/mesa/drivers/dri/nouveau/nouveau_context.h
@@ -109,12 +109,12 @@ typedef struct nouveau_context {
 	uint64_t gart_size;
 
 	/* Channel synchronisation */
-	drm_nouveau_notifier_alloc_t *syncNotifier;
+	struct drm_nouveau_notifier_alloc *syncNotifier;
 
 	/* ARB_occlusion_query / EXT_timer_query */
 	GLuint		  query_object_max;
 	GLboolean *	  query_alloc;
-	drm_nouveau_notifier_alloc_t *queryNotifier;
+	struct drm_nouveau_notifier_alloc *queryNotifier;
 
 	/* Additional hw-specific functions */
 	nouveau_hw_func hw_func;
@@ -168,7 +168,7 @@ typedef struct nouveau_context {
 	nouveauShader *passthrough_fp;
 
 	nouveauScreenRec *screen;
-	drm_nouveau_sarea_t *sarea;
+	struct drm_nouveau_sarea *sarea;
 
 	__DRIcontextPrivate  *driContext;    /* DRI context */
 	__DRIscreenPrivate   *driScreen;     /* DRI screen */
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_driver.c b/src/mesa/drivers/dri/nouveau/nouveau_driver.c
index 00956aa8f8b..ddc9535624b 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_driver.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_driver.c
@@ -41,7 +41,7 @@ GLboolean nouveauDRMGetParam(nouveauContextPtr nmesa,
 			     unsigned int      param,
 			     uint64_t*         value)
 {
-	drm_nouveau_getparam_t getp;
+	struct drm_nouveau_getparam getp;
 
 	getp.param = param;
 	if (!value || drmCommandWriteRead(nmesa->driFd, DRM_NOUVEAU_GETPARAM,
@@ -56,7 +56,7 @@ GLboolean nouveauDRMSetParam(nouveauContextPtr nmesa,
 			     unsigned int      param,
 			     uint64_t          value)
 {
-	drm_nouveau_setparam_t setp;
+	struct drm_nouveau_setparam setp;
 
 	setp.param = param;
 	setp.value = value;
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_fifo.c b/src/mesa/drivers/dri/nouveau/nouveau_fifo.c
index e9320918f9f..7b5e96b4c26 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_fifo.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_fifo.c
@@ -98,7 +98,7 @@ void nouveauWaitForIdle(nouveauContextPtr nmesa)
 // here we call the fifo initialization ioctl and fill in stuff accordingly
 GLboolean nouveauFifoInit(nouveauContextPtr nmesa)
 {
-	drm_nouveau_fifo_alloc_t fifo_init;
+	struct drm_nouveau_fifo_alloc fifo_init;
 	int i, ret;
 
 #ifdef NOUVEAU_RING_DEBUG
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_lock.c b/src/mesa/drivers/dri/nouveau/nouveau_lock.c
index c119d14dd75..aa86c9e7838 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_lock.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_lock.c
@@ -44,7 +44,7 @@ void nouveauGetLock( nouveauContextPtr nmesa, GLuint flags )
 {
    __DRIdrawablePrivate *dPriv = nmesa->driDrawable;
    __DRIscreenPrivate *sPriv = nmesa->driScreen;
-   drm_nouveau_sarea_t *sarea = nmesa->sarea;
+   struct drm_nouveau_sarea *sarea = nmesa->sarea;
 
    drmGetLock( nmesa->driFd, nmesa->hHWContext, flags );
 
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_object.c b/src/mesa/drivers/dri/nouveau/nouveau_object.c
index 69f8dbf7946..a143488e8d5 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_object.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_object.c
@@ -7,7 +7,7 @@
 GLboolean nouveauCreateContextObject(nouveauContextPtr nmesa,
 				     uint32_t handle, int class)
 {
-	drm_nouveau_grobj_alloc_t cto;
+	struct drm_nouveau_grobj_alloc cto;
 	int ret;
 
 	cto.channel = nmesa->fifo.channel;
@@ -34,10 +34,13 @@ void nouveauObjectInit(nouveauContextPtr nmesa)
 	nouveauCreateContextObject(nmesa, Nv3D, nmesa->screen->card->class_3d);
 	if (nmesa->screen->card->type>=NV_10) {
 		nouveauCreateContextObject(nmesa, NvCtxSurf2D, NV10_CONTEXT_SURFACES_2D);
-		nouveauCreateContextObject(nmesa, NvImageBlit, NV10_IMAGE_BLIT);
 	} else {
 		nouveauCreateContextObject(nmesa, NvCtxSurf2D, NV04_CONTEXT_SURFACES_2D);
 		nouveauCreateContextObject(nmesa, NvCtxSurf3D, NV04_CONTEXT_SURFACES_3D);
+	}
+	if (nmesa->screen->card->type>=NV_11) {
+		nouveauCreateContextObject(nmesa, NvImageBlit, NV10_IMAGE_BLIT);
+	} else {
 		nouveauCreateContextObject(nmesa, NvImageBlit, NV_IMAGE_BLIT);
 	}
 	nouveauCreateContextObject(nmesa, NvMemFormat, NV_MEMORY_TO_MEMORY_FORMAT);
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_screen.c b/src/mesa/drivers/dri/nouveau/nouveau_screen.c
index bc7f39b042a..69b0691bb7b 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_screen.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_screen.c
@@ -328,7 +328,7 @@ void * __driCreateNewScreen_20050727( __DRInativeDisplay *dpy, int scrn, __DRIsc
 	static const __DRIversion ddx_expected = { 1, 2, 0 };
 	static const __DRIversion dri_expected = { 4, 0, 0 };
 	static const __DRIversion drm_expected = { 0, 0, NOUVEAU_DRM_HEADER_PATCHLEVEL };
-#if NOUVEAU_DRM_HEADER_PATCHLEVEL != 7
+#if NOUVEAU_DRM_HEADER_PATCHLEVEL != 9
 #error nouveau_drm.h version doesn't match expected version
 #endif
 	dri_interface = interface;
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_state.c b/src/mesa/drivers/dri/nouveau/nouveau_state.c
index 7cb805902a7..f618dcfc99b 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_state.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_state.c
@@ -162,6 +162,8 @@ void nouveauDDInitState(nouveauContextPtr nmesa)
             nv04InitStateFuncs(nmesa->glCtx, &nmesa->glCtx->Driver);
             break;
         case NV_10:
+        case NV_11:
+        case NV_17:
             nv10InitStateFuncs(nmesa->glCtx, &nmesa->glCtx->Driver);
             break;
         case NV_20:
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_sync.c b/src/mesa/drivers/dri/nouveau/nouveau_sync.c
index 1d1eeede18b..8abc847e1e2 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_sync.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_sync.c
@@ -39,11 +39,11 @@
 	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx); \
 	volatile uint32_t *__v = (void*)nmesa->notifier_block + notifier->offset
 
-drm_nouveau_notifier_alloc_t *
+struct drm_nouveau_notifier_alloc *
 nouveau_notifier_new(GLcontext *ctx, GLuint handle, GLuint count)
 {
 	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
-	drm_nouveau_notifier_alloc_t *notifier;
+	struct drm_nouveau_notifier_alloc *notifier;
 	int ret;
 
 #ifdef NOUVEAU_RING_DEBUG
@@ -69,14 +69,16 @@ nouveau_notifier_new(GLcontext *ctx, GLuint handle, GLuint count)
 }
 
 void
-nouveau_notifier_destroy(GLcontext *ctx, drm_nouveau_notifier_alloc_t *notifier)
+nouveau_notifier_destroy(GLcontext *ctx,
+			 struct drm_nouveau_notifier_alloc *notifier)
 {
 	/*XXX: free notifier object.. */
 	FREE(notifier);
 }
 
 void
-nouveau_notifier_reset(GLcontext *ctx, drm_nouveau_notifier_alloc_t *notifier,
+nouveau_notifier_reset(GLcontext *ctx,
+		       struct drm_nouveau_notifier_alloc *notifier,
 		       GLuint id)
 {
 	NOTIFIER(n);
@@ -93,7 +95,8 @@ nouveau_notifier_reset(GLcontext *ctx, drm_nouveau_notifier_alloc_t *notifier,
 }
 
 GLuint
-nouveau_notifier_status(GLcontext *ctx, drm_nouveau_notifier_alloc_t *notifier,
+nouveau_notifier_status(GLcontext *ctx,
+			struct drm_nouveau_notifier_alloc *notifier,
 			GLuint id)
 {
 	NOTIFIER(n);
@@ -103,7 +106,8 @@ nouveau_notifier_status(GLcontext *ctx, drm_nouveau_notifier_alloc_t *notifier,
 
 GLuint
 nouveau_notifier_return_val(GLcontext *ctx,
-			    drm_nouveau_notifier_alloc_t *notifier, GLuint id)
+			    struct drm_nouveau_notifier_alloc *notifier,
+			    GLuint id)
 {
 	NOTIFIER(n);
 
@@ -112,8 +116,8 @@ nouveau_notifier_return_val(GLcontext *ctx,
 
 GLboolean
 nouveau_notifier_wait_status(GLcontext *ctx,
-			     drm_nouveau_notifier_alloc_t *notifier, GLuint id,
-			     GLuint status, GLuint timeout)
+			     struct drm_nouveau_notifier_alloc *notifier,
+			     GLuint id, GLuint status, GLuint timeout)
 {
 	NOTIFIER(n);
 	unsigned int time = 0;
@@ -146,7 +150,8 @@ nouveau_notifier_wait_status(GLcontext *ctx,
 
 void
 nouveau_notifier_wait_nop(GLcontext *ctx,
-			  drm_nouveau_notifier_alloc_t *notifier, GLuint subc)
+			  struct drm_nouveau_notifier_alloc *notifier,
+			  GLuint subc)
 {
 	NOTIFIER(n);
 	GLboolean ret;
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_sync.h b/src/mesa/drivers/dri/nouveau/nouveau_sync.h
index b56cc5fb544..b76af172762 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_sync.h
+++ b/src/mesa/drivers/dri/nouveau/nouveau_sync.h
@@ -47,22 +47,24 @@
 #define NV_NOTIFY                                                     0x00000104
 #define NV_NOTIFY_STYLE_WRITE_ONLY                                             0
 
-extern drm_nouveau_notifier_alloc_t *
+extern struct drm_nouveau_notifier_alloc *
 nouveau_notifier_new(GLcontext *, GLuint handle, GLuint count);
 extern void 
-nouveau_notifier_destroy(GLcontext *, drm_nouveau_notifier_alloc_t *);
+nouveau_notifier_destroy(GLcontext *, struct drm_nouveau_notifier_alloc *);
 extern void
-nouveau_notifier_reset(GLcontext *, drm_nouveau_notifier_alloc_t *, GLuint id);
+nouveau_notifier_reset(GLcontext *, struct drm_nouveau_notifier_alloc *,
+		       GLuint id);
 extern GLuint
-nouveau_notifier_status(GLcontext *, drm_nouveau_notifier_alloc_t *, GLuint id);
+nouveau_notifier_status(GLcontext *, struct drm_nouveau_notifier_alloc *,
+			GLuint id);
 extern GLuint
-nouveau_notifier_return_val(GLcontext *, drm_nouveau_notifier_alloc_t *,
+nouveau_notifier_return_val(GLcontext *, struct drm_nouveau_notifier_alloc *,
 			    GLuint id);
 extern GLboolean
-nouveau_notifier_wait_status(GLcontext *, drm_nouveau_notifier_alloc_t *,
+nouveau_notifier_wait_status(GLcontext *, struct drm_nouveau_notifier_alloc *,
 			     GLuint id, GLuint status, GLuint timeout);
 extern void
-nouveau_notifier_wait_nop(GLcontext *ctx, drm_nouveau_notifier_alloc_t *,
+nouveau_notifier_wait_nop(GLcontext *ctx, struct drm_nouveau_notifier_alloc *,
 			  GLuint subc);
 
 extern GLboolean nouveauSyncInitFuncs(GLcontext *ctx);
diff --git a/src/mesa/drivers/dri/nouveau/nv10_state.c b/src/mesa/drivers/dri/nouveau/nv10_state.c
index 5f304ccab92..47c4b14ba6b 100644
--- a/src/mesa/drivers/dri/nouveau/nv10_state.c
+++ b/src/mesa/drivers/dri/nouveau/nv10_state.c
@@ -697,8 +697,7 @@ static GLboolean nv10InitCard(nouveauContextPtr nmesa)
 	BEGIN_RING_SIZE(NvSub3D, 0x03f4, 1);
 	OUT_RING(0);
 
-	/* not for nv10, only for >= nv11 */
-	if ((nmesa->screen->card->id>>4) >= 0x11) {
+	if (nmesa->screen->card->type >= NV_11) {
 	        BEGIN_RING_SIZE(NvSub3D, 0x120, 3);
         	OUT_RING(0);
 	        OUT_RING(1);
@@ -739,11 +738,11 @@ static GLboolean nv10BindBuffers(nouveauContextPtr nmesa, int num_color,
 	OUT_RING_CACHE(depth ? depth->offset : color[0]->offset);
 
 	/* Always set to bottom left of buffer */
-	BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_VIEWPORT_ORIGIN_X, 4);
+	/*BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_VIEWPORT_ORIGIN_X, 4);
 	OUT_RING_CACHEf (0.0);
 	OUT_RING_CACHEf ((GLfloat) h);
 	OUT_RING_CACHEf (0.0);
-	OUT_RING_CACHEf (0.0);
+	OUT_RING_CACHEf (0.0);*/
 
 	return GL_TRUE;
 }
diff --git a/src/mesa/drivers/dri/nouveau/nv10_swtcl.c b/src/mesa/drivers/dri/nouveau/nv10_swtcl.c
index 4576c1ede4d..611469b6e41 100644
--- a/src/mesa/drivers/dri/nouveau/nv10_swtcl.c
+++ b/src/mesa/drivers/dri/nouveau/nv10_swtcl.c
@@ -58,7 +58,7 @@ static void nv10ResetLineStipple( GLcontext *ctx );
 
 static inline void nv10StartPrimitive(struct nouveau_context* nmesa,uint32_t primitive,uint32_t size)
 {
-	if (nmesa->screen->card->type==NV_10)
+	if ((nmesa->screen->card->type>=NV_10) && (nmesa->screen->card->type<=NV_17))
 		BEGIN_RING_SIZE(NvSub3D,NV10_TCL_PRIMITIVE_3D_BEGIN_END,1);
 	else if (nmesa->screen->card->type==NV_20)
 		BEGIN_RING_SIZE(NvSub3D,NV20_TCL_PRIMITIVE_3D_BEGIN_END,1);
@@ -66,7 +66,7 @@ static inline void nv10StartPrimitive(struct nouveau_context* nmesa,uint32_t pri
 		BEGIN_RING_SIZE(NvSub3D,NV30_TCL_PRIMITIVE_3D_BEGIN_END,1);
 	OUT_RING(primitive);
 
-	if (nmesa->screen->card->type==NV_10)
+	if ((nmesa->screen->card->type>=NV_10) && (nmesa->screen->card->type<=NV_17))
 		BEGIN_RING_SIZE(NvSub3D,NV10_TCL_PRIMITIVE_3D_VERTEX_ARRAY_DATA|NONINC_METHOD,size);
 	else if (nmesa->screen->card->type==NV_20)
 		BEGIN_RING_SIZE(NvSub3D,NV20_TCL_PRIMITIVE_3D_VERTEX_DATA|NONINC_METHOD,size);
@@ -76,7 +76,7 @@ static inline void nv10StartPrimitive(struct nouveau_context* nmesa,uint32_t pri
 
 inline void nv10FinishPrimitive(struct nouveau_context *nmesa)
 {
-	if (nmesa->screen->card->type==NV_10)
+	if ((nmesa->screen->card->type>=NV_10) && (nmesa->screen->card->type<=NV_17))
 		BEGIN_RING_SIZE(NvSub3D,NV10_TCL_PRIMITIVE_3D_BEGIN_END,1);
 	else if (nmesa->screen->card->type==NV_20)
 		BEGIN_RING_SIZE(NvSub3D,NV20_TCL_PRIMITIVE_3D_BEGIN_END,1);
@@ -454,7 +454,7 @@ static inline void nv10OutputVertexFormat(struct nouveau_context* nmesa)
 	/* 
 	 * Tell the hardware about the vertex format
 	 */
-	if (nmesa->screen->card->type==NV_10) {
+	if ((nmesa->screen->card->type>=NV_10) && (nmesa->screen->card->type<=NV_17)) {
 		int size;
 
 #define NV_VERTEX_ATTRIBUTE_TYPE_FLOAT 2
diff --git a/src/mesa/drivers/dri/nouveau/nv20_state.c b/src/mesa/drivers/dri/nouveau/nv20_state.c
index 3d8d83a865a..ccf2f6148b4 100644
--- a/src/mesa/drivers/dri/nouveau/nv20_state.c
+++ b/src/mesa/drivers/dri/nouveau/nv20_state.c
@@ -568,10 +568,10 @@ static void nv20Scissor(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h)
 	   y += nmesa->drawY;
 	}
 
-        BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_SCISSOR_X2_X1, 1);
+        /*BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_SCISSOR_X2_X1, 1);
         OUT_RING_CACHE((w << 16) | x );
         BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_SCISSOR_Y2_Y1, 1);
-        OUT_RING_CACHE((h << 16) | y );
+        OUT_RING_CACHE((h << 16) | y );*/
 
 }
 
@@ -764,11 +764,11 @@ static GLboolean nv20BindBuffers(nouveauContextPtr nmesa, int num_color,
 	}
 
 	/* Always set to bottom left of buffer */
-	BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_VIEWPORT_ORIGIN_X, 4);
+	/*BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_VIEWPORT_ORIGIN_X, 4);
 	OUT_RING_CACHEf (0.0);
 	OUT_RING_CACHEf ((GLfloat) h);
 	OUT_RING_CACHEf (0.0);
-	OUT_RING_CACHEf (0.0);
+	OUT_RING_CACHEf (0.0);*/
 
 	return GL_TRUE;
 }
diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h
index 6615bc79fb3..0349bac9a26 100644
--- a/src/mesa/drivers/dri/r300/r300_context.h
+++ b/src/mesa/drivers/dri/r300/r300_context.h
@@ -568,19 +568,10 @@ struct r300_vertex_shader_fragment {
 	union {
 		GLuint d[VSF_MAX_FRAGMENT_LENGTH];
 		float f[VSF_MAX_FRAGMENT_LENGTH];
-		VERTEX_SHADER_INSTRUCTION i[VSF_MAX_FRAGMENT_LENGTH / 4];
+		struct r300_vertprog_instruction i[VSF_MAX_FRAGMENT_LENGTH / 4];
 	} body;
 };
 
-#define VSF_DEST_PROGRAM	0x0
-#define VSF_DEST_MATRIX0	0x200
-#define VSF_DEST_MATRIX1	0x204
-#define VSF_DEST_MATRIX2	0x208
-#define VSF_DEST_VECTOR0	0x20c
-#define VSF_DEST_VECTOR1	0x20d
-#define VSF_DEST_UNKNOWN1	0x400
-#define VSF_DEST_UNKNOWN2	0x406
-
 struct r300_vertex_shader_state {
 	struct r300_vertex_shader_fragment program;
 };
diff --git a/src/mesa/drivers/dri/r300/r300_emit.c b/src/mesa/drivers/dri/r300/r300_emit.c
index 229439dfa87..424bf44e595 100644
--- a/src/mesa/drivers/dri/r300/r300_emit.c
+++ b/src/mesa/drivers/dri/r300/r300_emit.c
@@ -299,13 +299,14 @@ GLuint r300VAPOutputCntl0(GLcontext * ctx, GLuint OutputsWritten)
 	if (OutputsWritten & (1 << VERT_RESULT_COL1))
 		ret |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_1_PRESENT;
 
-#if 0
-	if (OutputsWritten & (1 << VERT_RESULT_BFC0))
-		ret |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_2_PRESENT;
-
-	if (OutputsWritten & (1 << VERT_RESULT_BFC1))
-		ret |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_3_PRESENT;
+	if (OutputsWritten & (1 << VERT_RESULT_BFC0)
+	    || OutputsWritten & (1 << VERT_RESULT_BFC1))
+		ret |=
+		    R300_VAP_OUTPUT_VTX_FMT_0__COLOR_1_PRESENT |
+		    R300_VAP_OUTPUT_VTX_FMT_0__COLOR_2_PRESENT |
+		    R300_VAP_OUTPUT_VTX_FMT_0__COLOR_3_PRESENT;
 
+#if 0
 	if (OutputsWritten & (1 << VERT_RESULT_FOGC)) ;
 #endif
 
diff --git a/src/mesa/drivers/dri/r300/r300_reg.h b/src/mesa/drivers/dri/r300/r300_reg.h
index e59919be493..1baa74c5269 100644
--- a/src/mesa/drivers/dri/r300/r300_reg.h
+++ b/src/mesa/drivers/dri/r300/r300_reg.h
@@ -282,9 +282,32 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
  */
 #define R300_VAP_PVS_UPLOAD_ADDRESS         0x2200
 #       define R300_PVS_UPLOAD_PROGRAM           0x00000000
+/* gap */
 #       define R300_PVS_UPLOAD_PARAMETERS        0x00000200
+/* gap */
+#       define R300_PVS_UPLOAD_CLIP_PLANE0       0x00000400
+#       define R300_PVS_UPLOAD_CLIP_PLANE1       0x00000401
+#       define R300_PVS_UPLOAD_CLIP_PLANE2       0x00000402
+#       define R300_PVS_UPLOAD_CLIP_PLANE3       0x00000403
+#       define R300_PVS_UPLOAD_CLIP_PLANE4       0x00000404
+#       define R300_PVS_UPLOAD_CLIP_PLANE5       0x00000405
 #       define R300_PVS_UPLOAD_POINTSIZE         0x00000406
 
+/*
+ * These are obsolete defines form r300_context.h, but they might give some
+ * clues when investigating the addresses further...
+ */
+#if 0
+#define VSF_DEST_PROGRAM        0x0
+#define VSF_DEST_MATRIX0        0x200
+#define VSF_DEST_MATRIX1        0x204
+#define VSF_DEST_MATRIX2        0x208
+#define VSF_DEST_VECTOR0        0x20c
+#define VSF_DEST_VECTOR1        0x20d
+#define VSF_DEST_UNKNOWN1       0x400
+#define VSF_DEST_UNKNOWN2       0x406
+#endif
+
 /* gap */
 
 #define R300_VAP_PVS_UPLOAD_DATA            0x2208
diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c
index b5cf21d644a..6789efd4281 100644
--- a/src/mesa/drivers/dri/r300/r300_state.c
+++ b/src/mesa/drivers/dri/r300/r300_state.c
@@ -82,6 +82,8 @@ static void r300BlendColor(GLcontext * ctx, const GLfloat cf[4])
 
 	rmesa->hw.blend_color.cmd[1] = PACK_COLOR_8888(color[3], color[0],
 						       color[1], color[2]);
+	rmesa->hw.blend_color.cmd[2] = 0;
+	rmesa->hw.blend_color.cmd[3] = 0;
 }
 
 /**
@@ -317,20 +319,34 @@ static void r300UpdateCulling(GLcontext * ctx)
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
 	uint32_t val = 0;
 
-	R300_STATECHANGE(r300, cul);
 	if (ctx->Polygon.CullFlag) {
-		if (ctx->Polygon.CullFaceMode == GL_FRONT_AND_BACK)
-			val = R300_CULL_FRONT | R300_CULL_BACK;
-		else if (ctx->Polygon.CullFaceMode == GL_FRONT)
+		switch (ctx->Polygon.CullFaceMode) {
+		case GL_FRONT:
 			val = R300_CULL_FRONT;
-		else
+			break;
+		case GL_BACK:
 			val = R300_CULL_BACK;
+			break;
+		case GL_FRONT_AND_BACK:
+			val = R300_CULL_FRONT | R300_CULL_BACK;
+			break;
+		default:
+			break;
+		}
+	}
 
-		if (ctx->Polygon.FrontFace == GL_CW)
-			val |= R300_FRONT_FACE_CW;
-		else
-			val |= R300_FRONT_FACE_CCW;
+	switch (ctx->Polygon.FrontFace) {
+	case GL_CW:
+		val |= R300_FRONT_FACE_CW;
+		break;
+	case GL_CCW:
+		val |= R300_FRONT_FACE_CCW;
+		break;
+	default:
+		break;
 	}
+
+	R300_STATECHANGE(r300, cul);
 	r300->hw.cul.cmd[R300_CUL_CULL] = val;
 }
 
@@ -344,6 +360,20 @@ static void r300SetEarlyZState(GLcontext * ctx)
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
 
 	R300_STATECHANGE(r300, zstencil_format);
+	switch (ctx->Visual.depthBits) {
+	case 16:
+		r300->hw.zstencil_format.cmd[1] = R300_DEPTH_FORMAT_16BIT_INT_Z;
+		break;
+	case 24:
+		r300->hw.zstencil_format.cmd[1] = R300_DEPTH_FORMAT_24BIT_INT_Z;
+		break;
+	default:
+		fprintf(stderr, "Error: Unsupported depth %d... exiting\n", ctx->Visual.depthBits);
+		_mesa_exit(-1);
+	}
+
+	// r300->hw.zstencil_format.cmd[1] |= R300_DEPTH_FORMAT_UNK32;
+
 	if (ctx->Color.AlphaEnabled && ctx->Color.AlphaFunc != GL_ALWAYS)
 		/* disable early Z */
 		r300->hw.zstencil_format.cmd[2] = R300_EARLY_Z_DISABLE;
@@ -355,6 +385,9 @@ static void r300SetEarlyZState(GLcontext * ctx)
 			/* disable early Z */
 			r300->hw.zstencil_format.cmd[2] = R300_EARLY_Z_DISABLE;
 	}
+
+	r300->hw.zstencil_format.cmd[3] = 0x00000003;
+	r300->hw.zstencil_format.cmd[4] = 0x00000000;
 }
 
 static void r300SetAlphaState(GLcontext * ctx)
@@ -403,6 +436,7 @@ static void r300SetAlphaState(GLcontext * ctx)
 
 	R300_STATECHANGE(r300, at);
 	r300->hw.at.cmd[R300_AT_ALPHA_TEST] = pp_misc;
+	r300->hw.at.cmd[R300_AT_UNKNOWN] = 0;
 
 	r300SetEarlyZState(ctx);
 }
@@ -513,6 +547,9 @@ static void r300UpdatePolygonMode(GLcontext * ctx)
 		R300_STATECHANGE(r300, polygon_mode);
 		r300->hw.polygon_mode.cmd[1] = hw_mode;
 	}
+
+	r300->hw.polygon_mode.cmd[2] = 0x00000001;
+	r300->hw.polygon_mode.cmd[3] = 0x00000000;
 }
 
 /**
@@ -762,6 +799,7 @@ static void r300ShadeModel(GLcontext * ctx, GLenum mode)
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 
 	R300_STATECHANGE(rmesa, shade);
+	rmesa->hw.shade.cmd[1] = 0x00000002;
 	switch (mode) {
 	case GL_FLAT:
 		rmesa->hw.shade.cmd[2] = R300_RE_SHADE_MODEL_FLAT;
@@ -772,6 +810,8 @@ static void r300ShadeModel(GLcontext * ctx, GLenum mode)
 	default:
 		return;
 	}
+	rmesa->hw.shade.cmd[3] = 0x00000000;
+	rmesa->hw.shade.cmd[4] = 0x00000000;
 }
 
 static void r300StencilFuncSeparate(GLcontext * ctx, GLenum face,
@@ -1526,7 +1566,7 @@ static void r300SetupDefaultVertexProgram(r300ContextPtr rmesa)
 
 	for (i = VERT_ATTRIB_POS; i < VERT_ATTRIB_MAX; i++) {
 		if (rmesa->state.sw_tcl_inputs[i] != -1) {
-			prog->program.body.i[program_end].op = EASY_VSF_OP(MUL, o_reg++, ALL, RESULT);
+			prog->program.body.i[program_end].opcode = EASY_VSF_OP(MUL, o_reg++, ALL, RESULT);
 			prog->program.body.i[program_end].src[0] = VSF_REG(rmesa->state.sw_tcl_inputs[i]);
 			prog->program.body.i[program_end].src[1] = VSF_ATTR_UNITY(rmesa->state.sw_tcl_inputs[i]);
 			prog->program.body.i[program_end].src[2] = VSF_UNITY(rmesa->state.sw_tcl_inputs[i]);
@@ -1536,7 +1576,7 @@ static void r300SetupDefaultVertexProgram(r300ContextPtr rmesa)
 
 	prog->program.length = program_end * 4;
 
-	r300SetupVertexProgramFragment(rmesa, VSF_DEST_PROGRAM, &(prog->program));
+	r300SetupVertexProgramFragment(rmesa, R300_PVS_UPLOAD_PROGRAM, &(prog->program));
 	inst_count = (prog->program.length / 4) - 1;
 
 	R300_STATECHANGE(rmesa, pvs);
@@ -1570,7 +1610,7 @@ static void r300SetupRealVertexProgram(r300ContextPtr rmesa)
 	bump_vpu_count(rmesa->hw.vpp.cmd, param_count);
 	param_count /= 4;
 
-	r300SetupVertexProgramFragment(rmesa, VSF_DEST_PROGRAM, &(prog->program));
+	r300SetupVertexProgramFragment(rmesa, R300_PVS_UPLOAD_PROGRAM, &(prog->program));
 	inst_count = (prog->program.length / 4) - 1;
 
 	R300_STATECHANGE(rmesa, pvs);
@@ -1848,15 +1888,10 @@ static void r300ResetHwState(r300ContextPtr r300)
 	r300->hw.unk4260.cmd[2] = r300PackFloat32(0.0);
 	r300->hw.unk4260.cmd[3] = r300PackFloat32(1.0);
 
-	r300->hw.shade.cmd[1] = 0x00000002;
 	r300ShadeModel(ctx, ctx->Light.ShadeModel);
-	r300->hw.shade.cmd[3] = 0x00000000;
-	r300->hw.shade.cmd[4] = 0x00000000;
 
 	r300PolygonMode(ctx, GL_FRONT, ctx->Polygon.FrontMode);
 	r300PolygonMode(ctx, GL_BACK, ctx->Polygon.BackMode);
-	r300->hw.polygon_mode.cmd[2] = 0x00000001;
-	r300->hw.polygon_mode.cmd[3] = 0x00000000;
 	r300->hw.zbias_cntl.cmd[1] = 0x00000000;
 
 	r300PolygonOffset(ctx, ctx->Polygon.OffsetFactor,
@@ -1887,14 +1922,11 @@ static void r300ResetHwState(r300ContextPtr r300)
 	r300Fogfv(ctx, GL_FOG_COLOR, ctx->Fog.Color);
 	r300Fogfv(ctx, GL_FOG_COORDINATE_SOURCE_EXT, NULL);
 
-	r300->hw.at.cmd[R300_AT_UNKNOWN] = 0;
 	r300->hw.unk4BD8.cmd[1] = 0;
 
 	r300->hw.unk4E00.cmd[1] = 0;
 
 	r300BlendColor(ctx, ctx->Color.BlendColor);
-	r300->hw.blend_color.cmd[2] = 0;
-	r300->hw.blend_color.cmd[3] = 0;
 
 	/* Again, r300ClearBuffer uses this */
 	r300->hw.cb.cmd[R300_CB_OFFSET] =
@@ -1925,25 +1957,6 @@ static void r300ResetHwState(r300ContextPtr r300)
 	r300->hw.unk4EA0.cmd[1] = 0x00000000;
 	r300->hw.unk4EA0.cmd[2] = 0xffffffff;
 
-	switch (ctx->Visual.depthBits) {
-	case 16:
-		r300->hw.zstencil_format.cmd[1] = R300_DEPTH_FORMAT_16BIT_INT_Z;
-		break;
-	case 24:
-		r300->hw.zstencil_format.cmd[1] = R300_DEPTH_FORMAT_24BIT_INT_Z;
-		break;
-	default:
-		fprintf(stderr, "Error: Unsupported depth %d... exiting\n",
-			ctx->Visual.depthBits);
-		_mesa_exit(-1);
-
-	}
-	/* z compress? */
-	//r300->hw.zstencil_format.cmd[1] |= R300_DEPTH_FORMAT_UNK32;
-
-	r300->hw.zstencil_format.cmd[3] = 0x00000003;
-	r300->hw.zstencil_format.cmd[4] = 0x00000000;
-
 	r300->hw.zb.cmd[R300_ZB_OFFSET] =
 	    r300->radeon.radeonScreen->depthOffset +
 	    r300->radeon.radeonScreen->fbLocation;
diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.c b/src/mesa/drivers/dri/r300/r300_vertprog.c
index 16dddf6557d..0fb6110494b 100644
--- a/src/mesa/drivers/dri/r300/r300_vertprog.c
+++ b/src/mesa/drivers/dri/r300/r300_vertprog.c
@@ -29,6 +29,8 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
  * \file
  *
  * \author Aapo Tahkola <[email protected]>
+ *
+ * \author Oliver McFadden <[email protected]>
  */
 
 #include "glheader.h"
@@ -55,54 +57,58 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #error Cannot change these!
 #endif
 
-#define SCALAR_FLAG (1<<31)
-#define FLAG_MASK (1<<31)
-#define OP_MASK	(0xf)		/* we are unlikely to have more than 15 */
-#define OPN(operator, ip) {#operator, OPCODE_##operator, ip}
-
-static struct {
-	char *name;
-	int opcode;
-	unsigned long ip;	/* number of input operands and flags */
-} op_names[] = {
-	/* *INDENT-OFF* */
-	OPN(ABS, 1),
-	OPN(ADD, 2),
-	OPN(ARL, 1 | SCALAR_FLAG),
-	OPN(DP3, 2),
-	OPN(DP4, 2),
-	OPN(DPH, 2),
-	OPN(DST, 2),
-	OPN(EX2, 1 | SCALAR_FLAG),
-	OPN(EXP, 1 | SCALAR_FLAG),
-	OPN(FLR, 1),
-	OPN(FRC, 1),
-	OPN(LG2, 1 | SCALAR_FLAG),
-	OPN(LIT, 1),
-	OPN(LOG, 1 | SCALAR_FLAG),
-	OPN(MAD, 3),
-	OPN(MAX, 2),
-	OPN(MIN, 2),
-	OPN(MOV, 1),
-	OPN(MUL, 2),
-	OPN(POW, 2 | SCALAR_FLAG),
-	OPN(RCP, 1 | SCALAR_FLAG),
-	OPN(RSQ, 1 | SCALAR_FLAG),
-	OPN(SGE, 2),
-	OPN(SLT, 2),
-	OPN(SUB, 2),
-	OPN(SWZ, 1),
-	OPN(XPD, 2),
-	OPN(RCC, 0),	//extra
-	OPN(PRINT, 0),
-	OPN(END, 0)
-	/* *INDENT-ON* */
-};
-
-#undef OPN
+/* TODO: Get rid of t_src_class call */
+#define CMP_SRCS(a, b) ((a.RelAddr != b.RelAddr) || (a.Index != b.Index && \
+		       ((t_src_class(a.File) == VSF_IN_CLASS_PARAM && \
+			 t_src_class(b.File) == VSF_IN_CLASS_PARAM) || \
+			(t_src_class(a.File) == VSF_IN_CLASS_ATTR && \
+			 t_src_class(b.File) == VSF_IN_CLASS_ATTR)))) \
+
+#define ZERO_SRC_0 (MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), \
+				    SWIZZLE_ZERO, SWIZZLE_ZERO, \
+				    SWIZZLE_ZERO, SWIZZLE_ZERO, \
+				    t_src_class(src[0].File), VSF_FLAG_NONE) | (src[0].RelAddr << 4))
+
+#define ZERO_SRC_1 (MAKE_VSF_SOURCE(t_src_index(vp, &src[1]), \
+				    SWIZZLE_ZERO, SWIZZLE_ZERO, \
+				    SWIZZLE_ZERO, SWIZZLE_ZERO, \
+				    t_src_class(src[1].File), VSF_FLAG_NONE) | (src[1].RelAddr << 4))
+
+#define ZERO_SRC_2 (MAKE_VSF_SOURCE(t_src_index(vp, &src[2]), \
+				    SWIZZLE_ZERO, SWIZZLE_ZERO, \
+				    SWIZZLE_ZERO, SWIZZLE_ZERO, \
+				    t_src_class(src[2].File), VSF_FLAG_NONE) | (src[2].RelAddr << 4))
+
+#define ONE_SRC_0 (MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), \
+				    SWIZZLE_ONE, SWIZZLE_ONE, \
+				    SWIZZLE_ONE, SWIZZLE_ONE, \
+				    t_src_class(src[0].File), VSF_FLAG_NONE) | (src[0].RelAddr << 4))
+
+#define ONE_SRC_1 (MAKE_VSF_SOURCE(t_src_index(vp, &src[1]), \
+				    SWIZZLE_ONE, SWIZZLE_ONE, \
+				    SWIZZLE_ONE, SWIZZLE_ONE, \
+				    t_src_class(src[1].File), VSF_FLAG_NONE) | (src[1].RelAddr << 4))
+
+#define ONE_SRC_2 (MAKE_VSF_SOURCE(t_src_index(vp, &src[2]), \
+				    SWIZZLE_ONE, SWIZZLE_ONE, \
+				    SWIZZLE_ONE, SWIZZLE_ONE, \
+				    t_src_class(src[2].File), VSF_FLAG_NONE) | (src[2].RelAddr << 4))
+
+/* DP4 version seems to trigger some hw peculiarity */
+//#define PREFER_DP4
+
+#define FREE_TEMPS() \
+	do { \
+		if(u_temp_i < vp->num_temporaries) { \
+			WARN_ONCE("Ran out of temps, num temps %d, us %d\n", vp->num_temporaries, u_temp_i); \
+			vp->native = GL_FALSE; \
+		} \
+		u_temp_i=VSF_MAX_FRAGMENT_TEMPS-1; \
+	} while (0)
 
 int r300VertexProgUpdateParams(GLcontext * ctx,
-			       struct r300_vertex_program_cont *vp, float *dst)
+			       struct r300_vertex_program_cont *vp,
+			       float *dst)
 {
 	int pi;
 	struct gl_vertex_program *mesa_vp = &vp->mesa_program;
@@ -234,8 +240,8 @@ static void vp_dump_inputs(struct r300_vertex_program *vp, char *caller)
 	int i;
 
 	if (vp == NULL) {
-		fprintf(stderr, "vp null in call to %s from %s\n", __FUNCTION__,
-			caller);
+		fprintf(stderr, "vp null in call to %s from %s\n",
+			__FUNCTION__, caller);
 		return;
 	}
 
@@ -276,6 +282,8 @@ static unsigned long t_src_index(struct r300_vertex_program *vp,
 	}
 }
 
+/* these two functions should probably be merged... */
+
 static unsigned long t_src(struct r300_vertex_program *vp,
 			   struct prog_src_register *src)
 {
@@ -294,7 +302,9 @@ static unsigned long t_src(struct r300_vertex_program *vp,
 static unsigned long t_src_scalar(struct r300_vertex_program *vp,
 				  struct prog_src_register *src)
 {
-
+	/* src->NegateBase uses the NEGATE_ flags from program_instruction.h,
+	 * which equal our VSF_FLAGS_ values, so it's safe to just pass it here.
+	 */
 	return MAKE_VSF_SOURCE(t_src_index(vp, src),
 			       t_swizzle(GET_SWZ(src->Swizzle, 0)),
 			       t_swizzle(GET_SWZ(src->Swizzle, 0)),
@@ -306,128 +316,727 @@ static unsigned long t_src_scalar(struct r300_vertex_program *vp,
 	    (src->RelAddr << 4);
 }
 
-static unsigned long t_opcode(enum prog_opcode opcode)
+static GLboolean valid_dst(struct r300_vertex_program *vp,
+			   struct prog_dst_register *dst)
 {
+	if (dst->File == PROGRAM_OUTPUT && vp->outputs[dst->Index] == -1) {
+		return GL_FALSE;
+	} else if (dst->File == PROGRAM_ADDRESS) {
+		assert(dst->Index == 0);
+	}
+
+	return GL_TRUE;
+}
 
-	switch (opcode) {
-	/* *INDENT-OFF* */
-	case OPCODE_ARL: return R300_VPI_OUT_OP_ARL;
-	case OPCODE_DST: return R300_VPI_OUT_OP_DST;
-	case OPCODE_EX2: return R300_VPI_OUT_OP_EX2;
-	case OPCODE_EXP: return R300_VPI_OUT_OP_EXP;
-	case OPCODE_FRC: return R300_VPI_OUT_OP_FRC;
-	case OPCODE_LG2: return R300_VPI_OUT_OP_LG2;
-	case OPCODE_LOG: return R300_VPI_OUT_OP_LOG;
-	case OPCODE_MAX: return R300_VPI_OUT_OP_MAX;
-	case OPCODE_MIN: return R300_VPI_OUT_OP_MIN;
-	case OPCODE_MUL: return R300_VPI_OUT_OP_MUL;
-	case OPCODE_RCP: return R300_VPI_OUT_OP_RCP;
-	case OPCODE_RSQ: return R300_VPI_OUT_OP_RSQ;
-	case OPCODE_SGE: return R300_VPI_OUT_OP_SGE;
-	case OPCODE_SLT: return R300_VPI_OUT_OP_SLT;
-	case OPCODE_DP4: return R300_VPI_OUT_OP_DOT;
-	/* *INDENT-ON* */
+/*
+ * Instruction    Inputs  Output   Description
+ * -----------    ------  ------   --------------------------------
+ * ABS            v       v        absolute value
+ * ADD            v,v     v        add
+ * ARL            s       a        address register load
+ * DP3            v,v     ssss     3-component dot product
+ * DP4            v,v     ssss     4-component dot product
+ * DPH            v,v     ssss     homogeneous dot product
+ * DST            v,v     v        distance vector
+ * EX2            s       ssss     exponential base 2
+ * EXP            s       v        exponential base 2 (approximate)
+ * FLR            v       v        floor
+ * FRC            v       v        fraction
+ * LG2            s       ssss     logarithm base 2
+ * LIT            v       v        compute light coefficients
+ * LOG            s       v        logarithm base 2 (approximate)
+ * MAD            v,v,v   v        multiply and add
+ * MAX            v,v     v        maximum
+ * MIN            v,v     v        minimum
+ * MOV            v       v        move
+ * MUL            v,v     v        multiply
+ * POW            s,s     ssss     exponentiate
+ * RCP            s       ssss     reciprocal
+ * RSQ            s       ssss     reciprocal square root
+ * SGE            v,v     v        set on greater than or equal
+ * SLT            v,v     v        set on less than
+ * SUB            v,v     v        subtract
+ * SWZ            v       v        extended swizzle
+ * XPD            v,v     v        cross product
+ *
+ * Table X.5:  Summary of vertex program instructions.  "v" indicates a
+ * floating-point vector input or output, "s" indicates a floating-point
+ * scalar input, "ssss" indicates a scalar output replicated across a
+ * 4-component result vector, and "a" indicates a single address register
+ * component.
+ */
 
-	default:
-		fprintf(stderr, "%s: Should not be called with opcode %d!",
-			__FUNCTION__, opcode);
-	}
-	_mesa_exit(-1);
-	return 0;
+static void t_opcode_abs(struct r300_vertex_program *vp,
+			 struct prog_instruction *vpi,
+			 struct r300_vertprog_instruction *o_inst,
+			 struct prog_src_register src[3])
+{
+	//MAX RESULT 1.X Y Z W PARAM 0{} {X Y Z W} PARAM 0{X Y Z W } {X Y Z W} neg Xneg Yneg Zneg W
+
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_MAX, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+
+	o_inst->src[0] = t_src(vp, &src[0]);
+	o_inst->src[1] =
+	    MAKE_VSF_SOURCE(t_src_index(vp, &src[0]),
+			    t_swizzle(GET_SWZ(src[0].Swizzle, 0)),
+			    t_swizzle(GET_SWZ(src[0].Swizzle, 1)),
+			    t_swizzle(GET_SWZ(src[0].Swizzle, 2)),
+			    t_swizzle(GET_SWZ(src[0].Swizzle, 3)),
+			    t_src_class(src[0].File),
+			    (!src[0].
+			     NegateBase) ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
+	    (src[0].RelAddr << 4);
+	o_inst->src[2] = 0;
 }
 
-static unsigned long op_operands(enum prog_opcode opcode)
+static void t_opcode_add(struct r300_vertex_program *vp,
+			 struct prog_instruction *vpi,
+			 struct r300_vertprog_instruction *o_inst,
+			 struct prog_src_register src[3])
 {
-	int i;
+	unsigned long hw_op;
 
-	/* Can we trust mesas opcodes to be in order ? */
-	for (i = 0; i < sizeof(op_names) / sizeof(*op_names); i++)
-		if (op_names[i].opcode == opcode)
-			return op_names[i].ip;
+#if 1
+	hw_op = (src[0].File == PROGRAM_TEMPORARY
+		 && src[1].File ==
+		 PROGRAM_TEMPORARY) ? R300_VPI_OUT_OP_MAD_2 :
+	    R300_VPI_OUT_OP_MAD;
+
+	o_inst->opcode =
+	    MAKE_VSF_OP(hw_op, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+	o_inst->src[0] = ONE_SRC_0;
+	o_inst->src[1] = t_src(vp, &src[0]);
+	o_inst->src[2] = t_src(vp, &src[1]);
+#else
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_ADD, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+	o_inst->src[0] = t_src(vp, &src[0]);
+	o_inst->src[1] = t_src(vp, &src[1]);
+	o_inst->src[2] = ZERO_SRC_1;
 
-	fprintf(stderr, "op %d not found in op_names\n", opcode);
-	_mesa_exit(-1);
-	return 0;
+#endif
 }
 
-static GLboolean valid_dst(struct r300_vertex_program *vp,
-			   struct prog_dst_register *dst)
+static void t_opcode_arl(struct r300_vertex_program *vp,
+			 struct prog_instruction *vpi,
+			 struct r300_vertprog_instruction *o_inst,
+			 struct prog_src_register src[3])
 {
-	if (dst->File == PROGRAM_OUTPUT && vp->outputs[dst->Index] == -1) {
-		return GL_FALSE;
-	} else if (dst->File == PROGRAM_ADDRESS) {
-		assert(dst->Index == 0);
-	}
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_ARL, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+
+	o_inst->src[0] = t_src(vp, &src[0]);
+	o_inst->src[1] = ZERO_SRC_0;
+	o_inst->src[2] = ZERO_SRC_0;
+}
 
-	return GL_TRUE;
+static void t_opcode_dp3(struct r300_vertex_program *vp,
+			 struct prog_instruction *vpi,
+			 struct r300_vertprog_instruction *o_inst,
+			 struct prog_src_register src[3])
+{
+	//DOT RESULT 1.X Y Z W PARAM 0{} {X Y Z ZERO} PARAM 0{} {X Y Z ZERO}
+
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_DOT, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+
+	o_inst->src[0] =
+	    MAKE_VSF_SOURCE(t_src_index(vp, &src[0]),
+			    t_swizzle(GET_SWZ(src[0].Swizzle, 0)),
+			    t_swizzle(GET_SWZ(src[0].Swizzle, 1)),
+			    t_swizzle(GET_SWZ(src[0].Swizzle, 2)),
+			    SWIZZLE_ZERO, t_src_class(src[0].File),
+			    src[0].
+			    NegateBase ? VSF_FLAG_XYZ : VSF_FLAG_NONE) |
+	    (src[0].RelAddr << 4);
+
+	o_inst->src[1] =
+	    MAKE_VSF_SOURCE(t_src_index(vp, &src[1]),
+			    t_swizzle(GET_SWZ(src[1].Swizzle, 0)),
+			    t_swizzle(GET_SWZ(src[1].Swizzle, 1)),
+			    t_swizzle(GET_SWZ(src[1].Swizzle, 2)),
+			    SWIZZLE_ZERO, t_src_class(src[1].File),
+			    src[1].
+			    NegateBase ? VSF_FLAG_XYZ : VSF_FLAG_NONE) |
+	    (src[1].RelAddr << 4);
+
+	o_inst->src[2] = ZERO_SRC_1;
 }
 
-/* TODO: Get rid of t_src_class call */
-#define CMP_SRCS(a, b) ((a.RelAddr != b.RelAddr) || (a.Index != b.Index && \
-		       ((t_src_class(a.File) == VSF_IN_CLASS_PARAM && \
-			 t_src_class(b.File) == VSF_IN_CLASS_PARAM) || \
-			(t_src_class(a.File) == VSF_IN_CLASS_ATTR && \
-			 t_src_class(b.File) == VSF_IN_CLASS_ATTR)))) \
+static void t_opcode_dp4(struct r300_vertex_program *vp,
+			 struct prog_instruction *vpi,
+			 struct r300_vertprog_instruction *o_inst,
+			 struct prog_src_register src[3])
+{
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_DOT, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+
+	o_inst->src[0] = t_src(vp, &src[0]);
+	o_inst->src[1] = t_src(vp, &src[1]);
+	o_inst->src[2] = ZERO_SRC_1;
+}
 
-#define ZERO_SRC_0 (MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), \
-				    SWIZZLE_ZERO, SWIZZLE_ZERO, \
-				    SWIZZLE_ZERO, SWIZZLE_ZERO, \
-				    t_src_class(src[0].File), VSF_FLAG_NONE) | (src[0].RelAddr << 4))
+static void t_opcode_dph(struct r300_vertex_program *vp,
+			 struct prog_instruction *vpi,
+			 struct r300_vertprog_instruction *o_inst,
+			 struct prog_src_register src[3])
+{
+	//DOT RESULT 1.X Y Z W PARAM 0{} {X Y Z ONE} PARAM 0{} {X Y Z W}
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_DOT, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+
+	o_inst->src[0] =
+	    MAKE_VSF_SOURCE(t_src_index(vp, &src[0]),
+			    t_swizzle(GET_SWZ(src[0].Swizzle, 0)),
+			    t_swizzle(GET_SWZ(src[0].Swizzle, 1)),
+			    t_swizzle(GET_SWZ(src[0].Swizzle, 2)),
+			    VSF_IN_COMPONENT_ONE, t_src_class(src[0].File),
+			    src[0].
+			    NegateBase ? VSF_FLAG_XYZ : VSF_FLAG_NONE) |
+	    (src[0].RelAddr << 4);
+	o_inst->src[1] = t_src(vp, &src[1]);
+	o_inst->src[2] = ZERO_SRC_1;
+}
 
-#define ZERO_SRC_1 (MAKE_VSF_SOURCE(t_src_index(vp, &src[1]), \
-				    SWIZZLE_ZERO, SWIZZLE_ZERO, \
-				    SWIZZLE_ZERO, SWIZZLE_ZERO, \
-				    t_src_class(src[1].File), VSF_FLAG_NONE) | (src[1].RelAddr << 4))
+static void t_opcode_dst(struct r300_vertex_program *vp,
+			 struct prog_instruction *vpi,
+			 struct r300_vertprog_instruction *o_inst,
+			 struct prog_src_register src[3])
+{
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_DST, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+
+	o_inst->src[0] = t_src(vp, &src[0]);
+	o_inst->src[1] = t_src(vp, &src[1]);
+	o_inst->src[2] = ZERO_SRC_1;
+}
 
-#define ZERO_SRC_2 (MAKE_VSF_SOURCE(t_src_index(vp, &src[2]), \
-				    SWIZZLE_ZERO, SWIZZLE_ZERO, \
-				    SWIZZLE_ZERO, SWIZZLE_ZERO, \
-				    t_src_class(src[2].File), VSF_FLAG_NONE) | (src[2].RelAddr << 4))
+static void t_opcode_ex2(struct r300_vertex_program *vp,
+			 struct prog_instruction *vpi,
+			 struct r300_vertprog_instruction *o_inst,
+			 struct prog_src_register src[3])
+{
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_EX2, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+
+	o_inst->src[0] = t_src_scalar(vp, &src[0]);
+	o_inst->src[1] = ZERO_SRC_0;
+	o_inst->src[2] = ZERO_SRC_0;
+}
 
-#define ONE_SRC_0 (MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), \
-				    SWIZZLE_ONE, SWIZZLE_ONE, \
-				    SWIZZLE_ONE, SWIZZLE_ONE, \
-				    t_src_class(src[0].File), VSF_FLAG_NONE) | (src[0].RelAddr << 4))
+static void t_opcode_exp(struct r300_vertex_program *vp,
+			 struct prog_instruction *vpi,
+			 struct r300_vertprog_instruction *o_inst,
+			 struct prog_src_register src[3])
+{
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_EXP, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+
+	o_inst->src[0] = t_src_scalar(vp, &src[0]);
+	o_inst->src[1] = ZERO_SRC_0;
+	o_inst->src[2] = ZERO_SRC_0;
+}
 
-#define ONE_SRC_1 (MAKE_VSF_SOURCE(t_src_index(vp, &src[1]), \
-				    SWIZZLE_ONE, SWIZZLE_ONE, \
-				    SWIZZLE_ONE, SWIZZLE_ONE, \
-				    t_src_class(src[1].File), VSF_FLAG_NONE) | (src[1].RelAddr << 4))
+static struct r300_vertprog_instruction *t_opcode_flr(struct r300_vertex_program *vp,
+						      struct prog_instruction *vpi,
+						      struct r300_vertprog_instruction *o_inst,
+						      struct prog_src_register src[3],
+						      int *u_temp_i)
+{
+	/* FRC TMP 0.X Y Z W PARAM 0{} {X Y Z W}
+	   ADD RESULT 1.X Y Z W PARAM 0{} {X Y Z W} TMP 0{X Y Z W } {X Y Z W} neg Xneg Yneg Zneg W */
+
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_FRC, *u_temp_i,
+			t_dst_mask(vpi->DstReg.WriteMask),
+			VSF_OUT_CLASS_TMP);
+
+	o_inst->src[0] = t_src(vp, &src[0]);
+	o_inst->src[1] = ZERO_SRC_0;
+	o_inst->src[2] = ZERO_SRC_0;
+	o_inst++;
+
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_ADD, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+
+	o_inst->src[0] = t_src(vp, &src[0]);
+	o_inst->src[1] =
+	    MAKE_VSF_SOURCE(*u_temp_i, VSF_IN_COMPONENT_X,
+			    VSF_IN_COMPONENT_Y, VSF_IN_COMPONENT_Z,
+			    VSF_IN_COMPONENT_W, VSF_IN_CLASS_TMP,
+			    /* Not 100% sure about this */
+			    (!src[0].
+			     NegateBase) ? VSF_FLAG_ALL : VSF_FLAG_NONE
+			    /*VSF_FLAG_ALL */ );
+
+	o_inst->src[2] = ZERO_SRC_0;
+	(*u_temp_i)--;
+
+	return o_inst;
+}
 
-#define ONE_SRC_2 (MAKE_VSF_SOURCE(t_src_index(vp, &src[2]), \
-				    SWIZZLE_ONE, SWIZZLE_ONE, \
-				    SWIZZLE_ONE, SWIZZLE_ONE, \
-				    t_src_class(src[2].File), VSF_FLAG_NONE) | (src[2].RelAddr << 4))
+static void t_opcode_frc(struct r300_vertex_program *vp,
+			 struct prog_instruction *vpi,
+			 struct r300_vertprog_instruction *o_inst,
+			 struct prog_src_register src[3])
+{
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_FRC, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+
+	o_inst->src[0] = t_src(vp, &src[0]);
+	o_inst->src[1] = ZERO_SRC_0;
+	o_inst->src[2] = ZERO_SRC_0;
+}
 
-/* DP4 version seems to trigger some hw peculiarity */
-//#define PREFER_DP4
+static void t_opcode_lg2(struct r300_vertex_program *vp,
+			 struct prog_instruction *vpi,
+			 struct r300_vertprog_instruction *o_inst,
+			 struct prog_src_register src[3])
+{
+	// LG2 RESULT 1.X Y Z W PARAM 0{} {X X X X}
+
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_LG2, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+
+	o_inst->src[0] =
+	    MAKE_VSF_SOURCE(t_src_index(vp, &src[0]),
+			    t_swizzle(GET_SWZ(src[0].Swizzle, 0)),
+			    t_swizzle(GET_SWZ(src[0].Swizzle, 0)),
+			    t_swizzle(GET_SWZ(src[0].Swizzle, 0)),
+			    t_swizzle(GET_SWZ(src[0].Swizzle, 0)),
+			    t_src_class(src[0].File),
+			    src[0].
+			    NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
+	    (src[0].RelAddr << 4);
+	o_inst->src[1] = ZERO_SRC_0;
+	o_inst->src[2] = ZERO_SRC_0;
+}
 
-#define FREE_TEMPS() \
-	do { \
-		if(u_temp_i < vp->num_temporaries) { \
-			WARN_ONCE("Ran out of temps, num temps %d, us %d\n", vp->num_temporaries, u_temp_i); \
-			vp->native = GL_FALSE; \
-		} \
-		u_temp_i=VSF_MAX_FRAGMENT_TEMPS-1; \
-	} while (0)
+static void t_opcode_lit(struct r300_vertex_program *vp,
+			 struct prog_instruction *vpi,
+			 struct r300_vertprog_instruction *o_inst,
+			 struct prog_src_register src[3])
+{
+	//LIT TMP 1.Y Z TMP 1{} {X W Z Y} TMP 1{} {Y W Z X} TMP 1{} {Y X Z W}
+
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_LIT, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+	/* NOTE: Users swizzling might not work. */
+	o_inst->src[0] = MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 0)),	// x
+					 t_swizzle(GET_SWZ(src[0].Swizzle, 3)),	// w
+					 VSF_IN_COMPONENT_ZERO,	// z
+					 t_swizzle(GET_SWZ(src[0].Swizzle, 1)),	// y
+					 t_src_class(src[0].File),
+					 src[0].
+					 NegateBase ? VSF_FLAG_ALL :
+					 VSF_FLAG_NONE) | (src[0].
+							   RelAddr << 4);
+	o_inst->src[1] = MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 1)),	// y
+					 t_swizzle(GET_SWZ(src[0].Swizzle, 3)),	// w
+					 VSF_IN_COMPONENT_ZERO,	// z
+					 t_swizzle(GET_SWZ(src[0].Swizzle, 0)),	// x
+					 t_src_class(src[0].File),
+					 src[0].
+					 NegateBase ? VSF_FLAG_ALL :
+					 VSF_FLAG_NONE) | (src[0].
+							   RelAddr << 4);
+	o_inst->src[2] = MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 1)),	// y
+					 t_swizzle(GET_SWZ(src[0].Swizzle, 0)),	// x
+					 VSF_IN_COMPONENT_ZERO,	// z
+					 t_swizzle(GET_SWZ(src[0].Swizzle, 3)),	// w
+					 t_src_class(src[0].File),
+					 src[0].
+					 NegateBase ? VSF_FLAG_ALL :
+					 VSF_FLAG_NONE) | (src[0].
+							   RelAddr << 4);
+}
 
-static void r300TranslateVertexShader(struct r300_vertex_program *vp,
-				      struct prog_instruction *vpi)
+static void t_opcode_log(struct r300_vertex_program *vp,
+			 struct prog_instruction *vpi,
+			 struct r300_vertprog_instruction *o_inst,
+			 struct prog_src_register src[3])
+{
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_LOG, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+
+	o_inst->src[0] = t_src_scalar(vp, &src[0]);
+	o_inst->src[1] = ZERO_SRC_0;
+	o_inst->src[2] = ZERO_SRC_0;
+}
+
+static void t_opcode_mad(struct r300_vertex_program *vp,
+			 struct prog_instruction *vpi,
+			 struct r300_vertprog_instruction *o_inst,
+			 struct prog_src_register src[3])
 {
-	int i, cur_reg = 0;
-	VERTEX_SHADER_INSTRUCTION *o_inst;
-	unsigned long operands;
-	int are_srcs_scalar;
 	unsigned long hw_op;
-	/* Initial value should be last tmp reg that hw supports.
-	   Strangely enough r300 doesnt mind even though these would be out of range.
-	   Smart enough to realize that it doesnt need it? */
-	int u_temp_i = VSF_MAX_FRAGMENT_TEMPS - 1;
-	struct prog_src_register src[3];
 
-	vp->pos_end = 0;	/* Not supported yet */
-	vp->program.length = 0;
-	/*vp->num_temporaries=mesa_vp->Base.NumTemporaries; */
+	hw_op = (src[0].File == PROGRAM_TEMPORARY
+		 && src[1].File == PROGRAM_TEMPORARY
+		 && src[2].File ==
+		 PROGRAM_TEMPORARY) ? R300_VPI_OUT_OP_MAD_2 :
+	    R300_VPI_OUT_OP_MAD;
+
+	o_inst->opcode =
+	    MAKE_VSF_OP(hw_op, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+	o_inst->src[0] = t_src(vp, &src[0]);
+	o_inst->src[1] = t_src(vp, &src[1]);
+	o_inst->src[2] = t_src(vp, &src[2]);
+}
+
+static void t_opcode_max(struct r300_vertex_program *vp,
+			 struct prog_instruction *vpi,
+			 struct r300_vertprog_instruction *o_inst,
+			 struct prog_src_register src[3])
+{
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_MAX, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+
+	o_inst->src[0] = t_src(vp, &src[0]);
+	o_inst->src[1] = t_src(vp, &src[1]);
+	o_inst->src[2] = ZERO_SRC_1;
+}
+
+static void t_opcode_min(struct r300_vertex_program *vp,
+			 struct prog_instruction *vpi,
+			 struct r300_vertprog_instruction *o_inst,
+			 struct prog_src_register src[3])
+{
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_MIN, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+
+	o_inst->src[0] = t_src(vp, &src[0]);
+	o_inst->src[1] = t_src(vp, &src[1]);
+	o_inst->src[2] = ZERO_SRC_1;
+}
+
+static void t_opcode_mov(struct r300_vertex_program *vp,
+			 struct prog_instruction *vpi,
+			 struct r300_vertprog_instruction *o_inst,
+			 struct prog_src_register src[3])
+{
+	//ADD RESULT 1.X Y Z W PARAM 0{} {X Y Z W} PARAM 0{} {ZERO ZERO ZERO ZERO}
+
+#if 1
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_ADD, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+	o_inst->src[0] = t_src(vp, &src[0]);
+	o_inst->src[1] = ZERO_SRC_0;
+	o_inst->src[2] = ZERO_SRC_0;
+#else
+	hw_op =
+	    (src[0].File ==
+	     PROGRAM_TEMPORARY) ? R300_VPI_OUT_OP_MAD_2 :
+	    R300_VPI_OUT_OP_MAD;
+
+	o_inst->opcode =
+	    MAKE_VSF_OP(hw_op, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+	o_inst->src[0] = t_src(vp, &src[0]);
+	o_inst->src[1] = ONE_SRC_0;
+	o_inst->src[2] = ZERO_SRC_0;
+#endif
+}
+
+static void t_opcode_mul(struct r300_vertex_program *vp,
+			 struct prog_instruction *vpi,
+			 struct r300_vertprog_instruction *o_inst,
+			 struct prog_src_register src[3])
+{
+	unsigned long hw_op;
+
+	// HW mul can take third arg but appears to have some other limitations.
+
+	hw_op = (src[0].File == PROGRAM_TEMPORARY
+		 && src[1].File ==
+		 PROGRAM_TEMPORARY) ? R300_VPI_OUT_OP_MAD_2 :
+	    R300_VPI_OUT_OP_MAD;
+
+	o_inst->opcode =
+	    MAKE_VSF_OP(hw_op, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+	o_inst->src[0] = t_src(vp, &src[0]);
+	o_inst->src[1] = t_src(vp, &src[1]);
+
+	o_inst->src[2] = ZERO_SRC_1;
+}
+
+static void t_opcode_pow(struct r300_vertex_program *vp,
+			 struct prog_instruction *vpi,
+			 struct r300_vertprog_instruction *o_inst,
+			 struct prog_src_register src[3])
+{
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_POW, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+	o_inst->src[0] = t_src_scalar(vp, &src[0]);
+	o_inst->src[1] = ZERO_SRC_0;
+	o_inst->src[2] = t_src_scalar(vp, &src[1]);
+}
+
+static void t_opcode_rcp(struct r300_vertex_program *vp,
+			 struct prog_instruction *vpi,
+			 struct r300_vertprog_instruction *o_inst,
+			 struct prog_src_register src[3])
+{
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_RCP, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+
+	o_inst->src[0] = t_src_scalar(vp, &src[0]);
+	o_inst->src[1] = ZERO_SRC_0;
+	o_inst->src[2] = ZERO_SRC_0;
+}
+
+static void t_opcode_rsq(struct r300_vertex_program *vp,
+			 struct prog_instruction *vpi,
+			 struct r300_vertprog_instruction *o_inst,
+			 struct prog_src_register src[3])
+{
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_RSQ, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+
+	o_inst->src[0] = t_src_scalar(vp, &src[0]);
+	o_inst->src[1] = ZERO_SRC_0;
+	o_inst->src[2] = ZERO_SRC_0;
+}
+
+static void t_opcode_sge(struct r300_vertex_program *vp,
+			 struct prog_instruction *vpi,
+			 struct r300_vertprog_instruction *o_inst,
+			 struct prog_src_register src[3])
+{
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_SGE, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+
+	o_inst->src[0] = t_src(vp, &src[0]);
+	o_inst->src[1] = t_src(vp, &src[1]);
+	o_inst->src[2] = ZERO_SRC_1;
+}
+
+static void t_opcode_slt(struct r300_vertex_program *vp,
+			 struct prog_instruction *vpi,
+			 struct r300_vertprog_instruction *o_inst,
+			 struct prog_src_register src[3])
+{
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_SLT, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+
+	o_inst->src[0] = t_src(vp, &src[0]);
+	o_inst->src[1] = t_src(vp, &src[1]);
+	o_inst->src[2] = ZERO_SRC_1;
+}
+
+static void t_opcode_sub(struct r300_vertex_program *vp,
+			 struct prog_instruction *vpi,
+			 struct r300_vertprog_instruction *o_inst,
+			 struct prog_src_register src[3])
+{
+	unsigned long hw_op;
+
+	//ADD RESULT 1.X Y Z W TMP 0{} {X Y Z W} PARAM 1{X Y Z W } {X Y Z W} neg Xneg Yneg Zneg W
+
+#if 1
+	hw_op = (src[0].File == PROGRAM_TEMPORARY
+		 && src[1].File ==
+		 PROGRAM_TEMPORARY) ? R300_VPI_OUT_OP_MAD_2 :
+	    R300_VPI_OUT_OP_MAD;
+
+	o_inst->opcode =
+	    MAKE_VSF_OP(hw_op, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+	o_inst->src[0] = t_src(vp, &src[0]);
+	o_inst->src[1] = ONE_SRC_0;
+	o_inst->src[2] =
+	    MAKE_VSF_SOURCE(t_src_index(vp, &src[1]),
+			    t_swizzle(GET_SWZ(src[1].Swizzle, 0)),
+			    t_swizzle(GET_SWZ(src[1].Swizzle, 1)),
+			    t_swizzle(GET_SWZ(src[1].Swizzle, 2)),
+			    t_swizzle(GET_SWZ(src[1].Swizzle, 3)),
+			    t_src_class(src[1].File),
+			    (!src[1].
+			     NegateBase) ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
+	    (src[1].RelAddr << 4);
+#else
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_ADD, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+
+	o_inst->src[0] = t_src(vp, &src[0]);
+	o_inst->src[1] =
+	    MAKE_VSF_SOURCE(t_src_index(vp, &src[1]),
+			    t_swizzle(GET_SWZ(src[1].Swizzle, 0)),
+			    t_swizzle(GET_SWZ(src[1].Swizzle, 1)),
+			    t_swizzle(GET_SWZ(src[1].Swizzle, 2)),
+			    t_swizzle(GET_SWZ(src[1].Swizzle, 3)),
+			    t_src_class(src[1].File),
+			    (!src[1].
+			     NegateBase) ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
+	    (src[1].RelAddr << 4);
+	o_inst->src[2] = 0;
+#endif
+}
+
+static void t_opcode_swz(struct r300_vertex_program *vp,
+			 struct prog_instruction *vpi,
+			 struct r300_vertprog_instruction *o_inst,
+			 struct prog_src_register src[3])
+{
+	//ADD RESULT 1.X Y Z W PARAM 0{} {X Y Z W} PARAM 0{} {ZERO ZERO ZERO ZERO}
+
+#if 1
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_ADD, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+	o_inst->src[0] = t_src(vp, &src[0]);
+	o_inst->src[1] = ZERO_SRC_0;
+	o_inst->src[2] = ZERO_SRC_0;
+#else
+	hw_op =
+	    (src[0].File ==
+	     PROGRAM_TEMPORARY) ? R300_VPI_OUT_OP_MAD_2 :
+	    R300_VPI_OUT_OP_MAD;
+
+	o_inst->opcode =
+	    MAKE_VSF_OP(hw_op, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+	o_inst->src[0] = t_src(vp, &src[0]);
+	o_inst->src[1] = ONE_SRC_0;
+	o_inst->src[2] = ZERO_SRC_0;
+#endif
+}
+
+static struct r300_vertprog_instruction *t_opcode_xpd(struct r300_vertex_program *vp,
+						      struct prog_instruction *vpi,
+						      struct r300_vertprog_instruction *o_inst,
+						      struct prog_src_register src[3],
+						      int *u_temp_i)
+{
+	/* mul r0, r1.yzxw, r2.zxyw
+	   mad r0, -r2.yzxw, r1.zxyw, r0
+	   NOTE: might need MAD_2
+	 */
+
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_MAD, *u_temp_i,
+			t_dst_mask(vpi->DstReg.WriteMask),
+			VSF_OUT_CLASS_TMP);
+
+	o_inst->src[0] = MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 1)),	// y
+					 t_swizzle(GET_SWZ(src[0].Swizzle, 2)),	// z
+					 t_swizzle(GET_SWZ(src[0].Swizzle, 0)),	// x
+					 t_swizzle(GET_SWZ(src[0].Swizzle, 3)),	// w
+					 t_src_class(src[0].File),
+					 src[0].
+					 NegateBase ? VSF_FLAG_ALL :
+					 VSF_FLAG_NONE) | (src[0].
+							   RelAddr << 4);
+
+	o_inst->src[1] = MAKE_VSF_SOURCE(t_src_index(vp, &src[1]), t_swizzle(GET_SWZ(src[1].Swizzle, 2)),	// z
+					 t_swizzle(GET_SWZ(src[1].Swizzle, 0)),	// x
+					 t_swizzle(GET_SWZ(src[1].Swizzle, 1)),	// y
+					 t_swizzle(GET_SWZ(src[1].Swizzle, 3)),	// w
+					 t_src_class(src[1].File),
+					 src[1].
+					 NegateBase ? VSF_FLAG_ALL :
+					 VSF_FLAG_NONE) | (src[1].
+							   RelAddr << 4);
+
+	o_inst->src[2] = ZERO_SRC_1;
+	o_inst++;
+	(*u_temp_i)--;
+
+	o_inst->opcode =
+	    MAKE_VSF_OP(R300_VPI_OUT_OP_MAD, t_dst_index(vp, &vpi->DstReg),
+			t_dst_mask(vpi->DstReg.WriteMask),
+			t_dst_class(vpi->DstReg.File));
+
+	o_inst->src[0] = MAKE_VSF_SOURCE(t_src_index(vp, &src[1]), t_swizzle(GET_SWZ(src[1].Swizzle, 1)),	// y
+					 t_swizzle(GET_SWZ(src[1].Swizzle, 2)),	// z
+					 t_swizzle(GET_SWZ(src[1].Swizzle, 0)),	// x
+					 t_swizzle(GET_SWZ(src[1].Swizzle, 3)),	// w
+					 t_src_class(src[1].File),
+					 (!src[1].
+					  NegateBase) ? VSF_FLAG_ALL :
+					 VSF_FLAG_NONE) | (src[1].
+							   RelAddr << 4);
+
+	o_inst->src[1] = MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 2)),	// z
+					 t_swizzle(GET_SWZ(src[0].Swizzle, 0)),	// x
+					 t_swizzle(GET_SWZ(src[0].Swizzle, 1)),	// y
+					 t_swizzle(GET_SWZ(src[0].Swizzle, 3)),	// w
+					 t_src_class(src[0].File),
+					 src[0].
+					 NegateBase ? VSF_FLAG_ALL :
+					 VSF_FLAG_NONE) | (src[0].
+							   RelAddr << 4);
+
+	o_inst->src[2] =
+	    MAKE_VSF_SOURCE(*u_temp_i + 1, VSF_IN_COMPONENT_X,
+			    VSF_IN_COMPONENT_Y, VSF_IN_COMPONENT_Z,
+			    VSF_IN_COMPONENT_W, VSF_IN_CLASS_TMP,
+			    VSF_FLAG_NONE);
+
+	return o_inst;
+}
+
+static void t_inputs_outputs(struct r300_vertex_program *vp)
+{
+	int i;
+	int cur_reg = 0;
 
 	for (i = 0; i < VERT_ATTRIB_MAX; i++)
 		vp->inputs[i] = -1;
@@ -437,39 +1046,71 @@ static void r300TranslateVertexShader(struct r300_vertex_program *vp,
 
 	assert(vp->key.OutputsWritten & (1 << VERT_RESULT_HPOS));
 
-	/* Assign outputs */
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_HPOS))
+	if (vp->key.OutputsWritten & (1 << VERT_RESULT_HPOS)) {
 		vp->outputs[VERT_RESULT_HPOS] = cur_reg++;
+	}
 
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_PSIZ))
+	if (vp->key.OutputsWritten & (1 << VERT_RESULT_PSIZ)) {
 		vp->outputs[VERT_RESULT_PSIZ] = cur_reg++;
+	}
 
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_COL0))
+	if (vp->key.OutputsWritten & (1 << VERT_RESULT_COL0)) {
 		vp->outputs[VERT_RESULT_COL0] = cur_reg++;
+	}
 
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_COL1))
-		vp->outputs[VERT_RESULT_COL1] = cur_reg++;
-
-#if 0				/* Not supported yet */
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_BFC0))
-		vp->outputs[VERT_RESULT_BFC0] = cur_reg++;
+	if (vp->key.OutputsWritten & (1 << VERT_RESULT_COL1)) {
+		vp->outputs[VERT_RESULT_COL1] =
+		    vp->outputs[VERT_RESULT_COL0] + 1;
+		cur_reg = vp->outputs[VERT_RESULT_COL1] + 1;
+	}
 
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_BFC1))
-		vp->outputs[VERT_RESULT_BFC1] = cur_reg++;
+	if (vp->key.OutputsWritten & (1 << VERT_RESULT_BFC0)) {
+		vp->outputs[VERT_RESULT_BFC0] =
+		    vp->outputs[VERT_RESULT_COL0] + 2;
+		cur_reg = vp->outputs[VERT_RESULT_BFC0] + 1;
+	}
 
-	if (vp->key.OutputsWritten & (1 << VERT_RESULT_FOGC))
+	if (vp->key.OutputsWritten & (1 << VERT_RESULT_BFC1)) {
+		vp->outputs[VERT_RESULT_BFC1] =
+		    vp->outputs[VERT_RESULT_COL0] + 3;
+		cur_reg = vp->outputs[VERT_RESULT_BFC1] + 1;
+	}
+#if 0
+	if (vp->key.OutputsWritten & (1 << VERT_RESULT_FOGC)) {
 		vp->outputs[VERT_RESULT_FOGC] = cur_reg++;
+	}
 #endif
 
-	for (i = VERT_RESULT_TEX0; i <= VERT_RESULT_TEX7; i++)
-		if (vp->key.OutputsWritten & (1 << i))
+	for (i = VERT_RESULT_TEX0; i <= VERT_RESULT_TEX7; i++) {
+		if (vp->key.OutputsWritten & (1 << i)) {
 			vp->outputs[i] = cur_reg++;
+		}
+	}
+}
+
+static void r300TranslateVertexShader(struct r300_vertex_program *vp,
+				      struct prog_instruction *vpi)
+{
+	int i;
+	struct r300_vertprog_instruction *o_inst;
+	unsigned long num_operands;
+	/* Initial value should be last tmp reg that hw supports.
+	   Strangely enough r300 doesnt mind even though these would be out of range.
+	   Smart enough to realize that it doesnt need it? */
+	int u_temp_i = VSF_MAX_FRAGMENT_TEMPS - 1;
+	struct prog_src_register src[3];
 
+	vp->pos_end = 0;	/* Not supported yet */
+	vp->program.length = 0;
+	/*vp->num_temporaries=mesa_vp->Base.NumTemporaries; */
 	vp->translated = GL_TRUE;
 	vp->native = GL_TRUE;
 
-	o_inst = vp->program.body.i;
-	for (; vpi->Opcode != OPCODE_END; vpi++, o_inst++) {
+	t_inputs_outputs(vp);
+
+	for (o_inst = vp->program.body.i; vpi->Opcode != OPCODE_END;
+	     vpi++, o_inst++) {
+
 		FREE_TEMPS();
 
 		if (!valid_dst(vp, &vpi->DstReg)) {
@@ -478,29 +1119,30 @@ static void r300TranslateVertexShader(struct r300_vertex_program *vp,
 			vpi->DstReg.Index = u_temp_i;
 		}
 
-		operands = op_operands(vpi->Opcode);
-		are_srcs_scalar = operands & SCALAR_FLAG;
-		operands &= OP_MASK;
+		num_operands = _mesa_num_inst_src_regs(vpi->Opcode);
 
-		for (i = 0; i < operands; i++)
+		/* copy the sources (src) from mesa into a local variable... is this needed? */
+		for (i = 0; i < num_operands; i++) {
 			src[i] = vpi->SrcReg[i];
+		}
 
-		if (operands == 3) {	/* TODO: scalars */
+		if (num_operands == 3) {	/* TODO: scalars */
 			if (CMP_SRCS(src[1], src[2])
 			    || CMP_SRCS(src[0], src[2])) {
-				o_inst->op =
-				    MAKE_VSF_OP(R300_VPI_OUT_OP_ADD, u_temp_i,
-						VSF_FLAG_ALL,
+				o_inst->opcode =
+				    MAKE_VSF_OP(R300_VPI_OUT_OP_ADD,
+						u_temp_i, VSF_FLAG_ALL,
 						VSF_OUT_CLASS_TMP);
 
 				o_inst->src[0] =
-				    MAKE_VSF_SOURCE(t_src_index(vp, &src[2]),
+				    MAKE_VSF_SOURCE(t_src_index
+						    (vp, &src[2]),
 						    SWIZZLE_X, SWIZZLE_Y,
 						    SWIZZLE_Z, SWIZZLE_W,
-						    t_src_class(src[2].File),
-						    VSF_FLAG_NONE) | (src[2].
-								      RelAddr <<
-								      4);
+						    t_src_class(src[2].
+								File),
+						    VSF_FLAG_NONE) |
+				    (src[2].RelAddr << 4);
 
 				o_inst->src[1] = ZERO_SRC_2;
 				o_inst->src[2] = ZERO_SRC_2;
@@ -511,24 +1153,24 @@ static void r300TranslateVertexShader(struct r300_vertex_program *vp,
 				src[2].RelAddr = 0;
 				u_temp_i--;
 			}
-
 		}
 
-		if (operands >= 2) {
+		if (num_operands >= 2) {
 			if (CMP_SRCS(src[1], src[0])) {
-				o_inst->op =
-				    MAKE_VSF_OP(R300_VPI_OUT_OP_ADD, u_temp_i,
-						VSF_FLAG_ALL,
+				o_inst->opcode =
+				    MAKE_VSF_OP(R300_VPI_OUT_OP_ADD,
+						u_temp_i, VSF_FLAG_ALL,
 						VSF_OUT_CLASS_TMP);
 
 				o_inst->src[0] =
-				    MAKE_VSF_SOURCE(t_src_index(vp, &src[0]),
+				    MAKE_VSF_SOURCE(t_src_index
+						    (vp, &src[0]),
 						    SWIZZLE_X, SWIZZLE_Y,
 						    SWIZZLE_Z, SWIZZLE_W,
-						    t_src_class(src[0].File),
-						    VSF_FLAG_NONE) | (src[0].
-								      RelAddr <<
-								      4);
+						    t_src_class(src[0].
+								File),
+						    VSF_FLAG_NONE) |
+				    (src[0].RelAddr << 4);
 
 				o_inst->src[1] = ZERO_SRC_0;
 				o_inst->src[2] = ZERO_SRC_0;
@@ -541,517 +1183,101 @@ static void r300TranslateVertexShader(struct r300_vertex_program *vp,
 			}
 		}
 
-		/* These ops need special handling. */
 		switch (vpi->Opcode) {
-		case OPCODE_POW:
-			o_inst->op =
-			    MAKE_VSF_OP(R300_VPI_OUT_OP_POW,
-					t_dst_index(vp, &vpi->DstReg),
-					t_dst_mask(vpi->DstReg.WriteMask),
-					t_dst_class(vpi->DstReg.File));
-			o_inst->src[0] = t_src_scalar(vp, &src[0]);
-			o_inst->src[1] = ZERO_SRC_0;
-			o_inst->src[2] = t_src_scalar(vp, &src[1]);
-			goto next;
-
-		case OPCODE_MOV:	//ADD RESULT 1.X Y Z W PARAM 0{} {X Y Z W} PARAM 0{} {ZERO ZERO ZERO ZERO}
-		case OPCODE_SWZ:
-#if 1
-			o_inst->op =
-			    MAKE_VSF_OP(R300_VPI_OUT_OP_ADD,
-					t_dst_index(vp, &vpi->DstReg),
-					t_dst_mask(vpi->DstReg.WriteMask),
-					t_dst_class(vpi->DstReg.File));
-			o_inst->src[0] = t_src(vp, &src[0]);
-			o_inst->src[1] = ZERO_SRC_0;
-			o_inst->src[2] = ZERO_SRC_0;
-#else
-			hw_op =
-			    (src[0].File ==
-			     PROGRAM_TEMPORARY) ? R300_VPI_OUT_OP_MAD_2 :
-			    R300_VPI_OUT_OP_MAD;
-
-			o_inst->op =
-			    MAKE_VSF_OP(hw_op, t_dst_index(vp, &vpi->DstReg),
-					t_dst_mask(vpi->DstReg.WriteMask),
-					t_dst_class(vpi->DstReg.File));
-			o_inst->src[0] = t_src(vp, &src[0]);
-			o_inst->src[1] = ONE_SRC_0;
-			o_inst->src[2] = ZERO_SRC_0;
-#endif
-
-			goto next;
-
+		case OPCODE_ABS:
+			t_opcode_abs(vp, vpi, o_inst, src);
+			break;
 		case OPCODE_ADD:
-#if 1
-			hw_op = (src[0].File == PROGRAM_TEMPORARY &&
-				 src[1].File ==
-				 PROGRAM_TEMPORARY) ? R300_VPI_OUT_OP_MAD_2 :
-			    R300_VPI_OUT_OP_MAD;
-
-			o_inst->op =
-			    MAKE_VSF_OP(hw_op, t_dst_index(vp, &vpi->DstReg),
-					t_dst_mask(vpi->DstReg.WriteMask),
-					t_dst_class(vpi->DstReg.File));
-			o_inst->src[0] = ONE_SRC_0;
-			o_inst->src[1] = t_src(vp, &src[0]);
-			o_inst->src[2] = t_src(vp, &src[1]);
-#else
-			o_inst->op =
-			    MAKE_VSF_OP(R300_VPI_OUT_OP_ADD,
-					t_dst_index(vp, &vpi->DstReg),
-					t_dst_mask(vpi->DstReg.WriteMask),
-					t_dst_class(vpi->DstReg.File));
-			o_inst->src[0] = t_src(vp, &src[0]);
-			o_inst->src[1] = t_src(vp, &src[1]);
-			o_inst->src[2] = ZERO_SRC_1;
-
-#endif
-			goto next;
-
-		case OPCODE_MAD:
-			hw_op = (src[0].File == PROGRAM_TEMPORARY &&
-				 src[1].File == PROGRAM_TEMPORARY &&
-				 src[2].File ==
-				 PROGRAM_TEMPORARY) ? R300_VPI_OUT_OP_MAD_2 :
-			    R300_VPI_OUT_OP_MAD;
-
-			o_inst->op =
-			    MAKE_VSF_OP(hw_op, t_dst_index(vp, &vpi->DstReg),
-					t_dst_mask(vpi->DstReg.WriteMask),
-					t_dst_class(vpi->DstReg.File));
-			o_inst->src[0] = t_src(vp, &src[0]);
-			o_inst->src[1] = t_src(vp, &src[1]);
-			o_inst->src[2] = t_src(vp, &src[2]);
-			goto next;
-
-		case OPCODE_MUL:	/* HW mul can take third arg but appears to have some other limitations. */
-			hw_op = (src[0].File == PROGRAM_TEMPORARY &&
-				 src[1].File ==
-				 PROGRAM_TEMPORARY) ? R300_VPI_OUT_OP_MAD_2 :
-			    R300_VPI_OUT_OP_MAD;
-
-			o_inst->op =
-			    MAKE_VSF_OP(hw_op, t_dst_index(vp, &vpi->DstReg),
-					t_dst_mask(vpi->DstReg.WriteMask),
-					t_dst_class(vpi->DstReg.File));
-			o_inst->src[0] = t_src(vp, &src[0]);
-			o_inst->src[1] = t_src(vp, &src[1]);
-
-			o_inst->src[2] = ZERO_SRC_1;
-			goto next;
-
-		case OPCODE_DP3:	//DOT RESULT 1.X Y Z W PARAM 0{} {X Y Z ZERO} PARAM 0{} {X Y Z ZERO}
-			o_inst->op =
-			    MAKE_VSF_OP(R300_VPI_OUT_OP_DOT,
-					t_dst_index(vp, &vpi->DstReg),
-					t_dst_mask(vpi->DstReg.WriteMask),
-					t_dst_class(vpi->DstReg.File));
-
-			o_inst->src[0] =
-			    MAKE_VSF_SOURCE(t_src_index(vp, &src[0]),
-					    t_swizzle(GET_SWZ
-						      (src[0].Swizzle, 0)),
-					    t_swizzle(GET_SWZ
-						      (src[0].Swizzle, 1)),
-					    t_swizzle(GET_SWZ
-						      (src[0].Swizzle, 2)),
-					    SWIZZLE_ZERO,
-					    t_src_class(src[0].File),
-					    src[0].
-					    NegateBase ? VSF_FLAG_XYZ :
-					    VSF_FLAG_NONE) | (src[0].
-							      RelAddr << 4);
-
-			o_inst->src[1] =
-			    MAKE_VSF_SOURCE(t_src_index(vp, &src[1]),
-					    t_swizzle(GET_SWZ
-						      (src[1].Swizzle, 0)),
-					    t_swizzle(GET_SWZ
-						      (src[1].Swizzle, 1)),
-					    t_swizzle(GET_SWZ
-						      (src[1].Swizzle, 2)),
-					    SWIZZLE_ZERO,
-					    t_src_class(src[1].File),
-					    src[1].
-					    NegateBase ? VSF_FLAG_XYZ :
-					    VSF_FLAG_NONE) | (src[1].
-							      RelAddr << 4);
-
-			o_inst->src[2] = ZERO_SRC_1;
-			goto next;
-
-		case OPCODE_SUB:	//ADD RESULT 1.X Y Z W TMP 0{} {X Y Z W} PARAM 1{X Y Z W } {X Y Z W} neg Xneg Yneg Zneg W
-#if 1
-			hw_op = (src[0].File == PROGRAM_TEMPORARY &&
-				 src[1].File ==
-				 PROGRAM_TEMPORARY) ? R300_VPI_OUT_OP_MAD_2 :
-			    R300_VPI_OUT_OP_MAD;
-
-			o_inst->op =
-			    MAKE_VSF_OP(hw_op, t_dst_index(vp, &vpi->DstReg),
-					t_dst_mask(vpi->DstReg.WriteMask),
-					t_dst_class(vpi->DstReg.File));
-			o_inst->src[0] = t_src(vp, &src[0]);
-			o_inst->src[1] = ONE_SRC_0;
-			o_inst->src[2] =
-			    MAKE_VSF_SOURCE(t_src_index(vp, &src[1]),
-					    t_swizzle(GET_SWZ
-						      (src[1].Swizzle, 0)),
-					    t_swizzle(GET_SWZ
-						      (src[1].Swizzle, 1)),
-					    t_swizzle(GET_SWZ
-						      (src[1].Swizzle, 2)),
-					    t_swizzle(GET_SWZ
-						      (src[1].Swizzle, 3)),
-					    t_src_class(src[1].File),
-					    (!src[1].
-					     NegateBase) ? VSF_FLAG_ALL :
-					    VSF_FLAG_NONE) | (src[1].
-							      RelAddr << 4);
-#else
-			o_inst->op =
-			    MAKE_VSF_OP(R300_VPI_OUT_OP_ADD,
-					t_dst_index(vp, &vpi->DstReg),
-					t_dst_mask(vpi->DstReg.WriteMask),
-					t_dst_class(vpi->DstReg.File));
-
-			o_inst->src[0] = t_src(vp, &src[0]);
-			o_inst->src[1] =
-			    MAKE_VSF_SOURCE(t_src_index(vp, &src[1]),
-					    t_swizzle(GET_SWZ
-						      (src[1].Swizzle, 0)),
-					    t_swizzle(GET_SWZ
-						      (src[1].Swizzle, 1)),
-					    t_swizzle(GET_SWZ
-						      (src[1].Swizzle, 2)),
-					    t_swizzle(GET_SWZ
-						      (src[1].Swizzle, 3)),
-					    t_src_class(src[1].File),
-					    (!src[1].
-					     NegateBase) ? VSF_FLAG_ALL :
-					    VSF_FLAG_NONE) | (src[1].
-							      RelAddr << 4);
-			o_inst->src[2] = 0;
-#endif
-			goto next;
-
-		case OPCODE_ABS:	//MAX RESULT 1.X Y Z W PARAM 0{} {X Y Z W} PARAM 0{X Y Z W } {X Y Z W} neg Xneg Yneg Zneg W
-			o_inst->op =
-			    MAKE_VSF_OP(R300_VPI_OUT_OP_MAX,
-					t_dst_index(vp, &vpi->DstReg),
-					t_dst_mask(vpi->DstReg.WriteMask),
-					t_dst_class(vpi->DstReg.File));
-
-			o_inst->src[0] = t_src(vp, &src[0]);
-			o_inst->src[1] =
-			    MAKE_VSF_SOURCE(t_src_index(vp, &src[0]),
-					    t_swizzle(GET_SWZ
-						      (src[0].Swizzle, 0)),
-					    t_swizzle(GET_SWZ
-						      (src[0].Swizzle, 1)),
-					    t_swizzle(GET_SWZ
-						      (src[0].Swizzle, 2)),
-					    t_swizzle(GET_SWZ
-						      (src[0].Swizzle, 3)),
-					    t_src_class(src[0].File),
-					    (!src[0].
-					     NegateBase) ? VSF_FLAG_ALL :
-					    VSF_FLAG_NONE) | (src[0].
-							      RelAddr << 4);
-			o_inst->src[2] = 0;
-			goto next;
-
+			t_opcode_add(vp, vpi, o_inst, src);
+			break;
+		case OPCODE_ARL:
+			t_opcode_arl(vp, vpi, o_inst, src);
+			break;
+		case OPCODE_DP3:
+			t_opcode_dp3(vp, vpi, o_inst, src);
+			break;
+		case OPCODE_DP4:
+			t_opcode_dp4(vp, vpi, o_inst, src);
+			break;
+		case OPCODE_DPH:
+			t_opcode_dph(vp, vpi, o_inst, src);
+			break;
+		case OPCODE_DST:
+			t_opcode_dst(vp, vpi, o_inst, src);
+			break;
+		case OPCODE_EX2:
+			t_opcode_ex2(vp, vpi, o_inst, src);
+			break;
+		case OPCODE_EXP:
+			t_opcode_exp(vp, vpi, o_inst, src);
+			break;
 		case OPCODE_FLR:
-			/* FRC TMP 0.X Y Z W PARAM 0{} {X Y Z W}
-			   ADD RESULT 1.X Y Z W PARAM 0{} {X Y Z W} TMP 0{X Y Z W } {X Y Z W} neg Xneg Yneg Zneg W */
-
-			o_inst->op = MAKE_VSF_OP(R300_VPI_OUT_OP_FRC, u_temp_i,
-						 t_dst_mask(vpi->DstReg.
-							    WriteMask),
-						 VSF_OUT_CLASS_TMP);
-
-			o_inst->src[0] = t_src(vp, &src[0]);
-			o_inst->src[1] = ZERO_SRC_0;
-			o_inst->src[2] = ZERO_SRC_0;
-			o_inst++;
-
-			o_inst->op =
-			    MAKE_VSF_OP(R300_VPI_OUT_OP_ADD,
-					t_dst_index(vp, &vpi->DstReg),
-					t_dst_mask(vpi->DstReg.WriteMask),
-					t_dst_class(vpi->DstReg.File));
-
-			o_inst->src[0] = t_src(vp, &src[0]);
-			o_inst->src[1] = MAKE_VSF_SOURCE(u_temp_i,
-							 VSF_IN_COMPONENT_X,
-							 VSF_IN_COMPONENT_Y,
-							 VSF_IN_COMPONENT_Z,
-							 VSF_IN_COMPONENT_W,
-							 VSF_IN_CLASS_TMP,
-							 /* Not 100% sure about this */
-							 (!src[0].
-							  NegateBase) ?
-							 VSF_FLAG_ALL :
-							 VSF_FLAG_NONE
-							 /*VSF_FLAG_ALL */ );
-
-			o_inst->src[2] = ZERO_SRC_0;
-			u_temp_i--;
-			goto next;
-
-		case OPCODE_LG2:	// LG2 RESULT 1.X Y Z W PARAM 0{} {X X X X}
-			o_inst->op =
-			    MAKE_VSF_OP(R300_VPI_OUT_OP_LG2,
-					t_dst_index(vp, &vpi->DstReg),
-					t_dst_mask(vpi->DstReg.WriteMask),
-					t_dst_class(vpi->DstReg.File));
-
-			o_inst->src[0] =
-			    MAKE_VSF_SOURCE(t_src_index(vp, &src[0]),
-					    t_swizzle(GET_SWZ
-						      (src[0].Swizzle, 0)),
-					    t_swizzle(GET_SWZ
-						      (src[0].Swizzle, 0)),
-					    t_swizzle(GET_SWZ
-						      (src[0].Swizzle, 0)),
-					    t_swizzle(GET_SWZ
-						      (src[0].Swizzle, 0)),
-					    t_src_class(src[0].File),
-					    src[0].
-					    NegateBase ? VSF_FLAG_ALL :
-					    VSF_FLAG_NONE) | (src[0].
-							      RelAddr << 4);
-			o_inst->src[1] = ZERO_SRC_0;
-			o_inst->src[2] = ZERO_SRC_0;
-			goto next;
-
-		case OPCODE_LIT:	//LIT TMP 1.Y Z TMP 1{} {X W Z Y} TMP 1{} {Y W Z X} TMP 1{} {Y X Z W}
-			o_inst->op =
-			    MAKE_VSF_OP(R300_VPI_OUT_OP_LIT,
-					t_dst_index(vp, &vpi->DstReg),
-					t_dst_mask(vpi->DstReg.WriteMask),
-					t_dst_class(vpi->DstReg.File));
-			/* NOTE: Users swizzling might not work. */
-			o_inst->src[0] = MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 0)),	// x
-							 t_swizzle(GET_SWZ(src[0].Swizzle, 3)),	// w
-							 VSF_IN_COMPONENT_ZERO,	// z
-							 t_swizzle(GET_SWZ(src[0].Swizzle, 1)),	// y
-							 t_src_class(src[0].
-								     File),
-							 src[0].
-							 NegateBase ?
-							 VSF_FLAG_ALL :
-							 VSF_FLAG_NONE) |
-			    (src[0].RelAddr << 4);
-			o_inst->src[1] = MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 1)),	// y
-							 t_swizzle(GET_SWZ(src[0].Swizzle, 3)),	// w
-							 VSF_IN_COMPONENT_ZERO,	// z
-							 t_swizzle(GET_SWZ(src[0].Swizzle, 0)),	// x
-							 t_src_class(src[0].
-								     File),
-							 src[0].
-							 NegateBase ?
-							 VSF_FLAG_ALL :
-							 VSF_FLAG_NONE) |
-			    (src[0].RelAddr << 4);
-			o_inst->src[2] = MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 1)),	// y
-							 t_swizzle(GET_SWZ(src[0].Swizzle, 0)),	// x
-							 VSF_IN_COMPONENT_ZERO,	// z
-							 t_swizzle(GET_SWZ(src[0].Swizzle, 3)),	// w
-							 t_src_class(src[0].
-								     File),
-							 src[0].
-							 NegateBase ?
-							 VSF_FLAG_ALL :
-							 VSF_FLAG_NONE) |
-			    (src[0].RelAddr << 4);
-			goto next;
-
-		case OPCODE_DPH:	//DOT RESULT 1.X Y Z W PARAM 0{} {X Y Z ONE} PARAM 0{} {X Y Z W}
-			o_inst->op =
-			    MAKE_VSF_OP(R300_VPI_OUT_OP_DOT,
-					t_dst_index(vp, &vpi->DstReg),
-					t_dst_mask(vpi->DstReg.WriteMask),
-					t_dst_class(vpi->DstReg.File));
-
-			o_inst->src[0] =
-			    MAKE_VSF_SOURCE(t_src_index(vp, &src[0]),
-					    t_swizzle(GET_SWZ
-						      (src[0].Swizzle, 0)),
-					    t_swizzle(GET_SWZ
-						      (src[0].Swizzle, 1)),
-					    t_swizzle(GET_SWZ
-						      (src[0].Swizzle, 2)),
-					    VSF_IN_COMPONENT_ONE,
-					    t_src_class(src[0].File),
-					    src[0].
-					    NegateBase ? VSF_FLAG_XYZ :
-					    VSF_FLAG_NONE) | (src[0].
-							      RelAddr << 4);
-			o_inst->src[1] = t_src(vp, &src[1]);
-			o_inst->src[2] = ZERO_SRC_1;
-			goto next;
-
-		case OPCODE_XPD:
-			/* mul r0, r1.yzxw, r2.zxyw
-			   mad r0, -r2.yzxw, r1.zxyw, r0
-			   NOTE: might need MAD_2
-			 */
-
-			o_inst->op = MAKE_VSF_OP(R300_VPI_OUT_OP_MAD, u_temp_i,
-						 t_dst_mask(vpi->DstReg.
-							    WriteMask),
-						 VSF_OUT_CLASS_TMP);
-
-			o_inst->src[0] = MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 1)),	// y
-							 t_swizzle(GET_SWZ(src[0].Swizzle, 2)),	// z
-							 t_swizzle(GET_SWZ(src[0].Swizzle, 0)),	// x
-							 t_swizzle(GET_SWZ(src[0].Swizzle, 3)),	// w
-							 t_src_class(src[0].
-								     File),
-							 src[0].
-							 NegateBase ?
-							 VSF_FLAG_ALL :
-							 VSF_FLAG_NONE) |
-			    (src[0].RelAddr << 4);
-
-			o_inst->src[1] = MAKE_VSF_SOURCE(t_src_index(vp, &src[1]), t_swizzle(GET_SWZ(src[1].Swizzle, 2)),	// z
-							 t_swizzle(GET_SWZ(src[1].Swizzle, 0)),	// x
-							 t_swizzle(GET_SWZ(src[1].Swizzle, 1)),	// y
-							 t_swizzle(GET_SWZ(src[1].Swizzle, 3)),	// w
-							 t_src_class(src[1].
-								     File),
-							 src[1].
-							 NegateBase ?
-							 VSF_FLAG_ALL :
-							 VSF_FLAG_NONE) |
-			    (src[1].RelAddr << 4);
-
-			o_inst->src[2] = ZERO_SRC_1;
-			o_inst++;
-			u_temp_i--;
-
-			o_inst->op =
-			    MAKE_VSF_OP(R300_VPI_OUT_OP_MAD,
-					t_dst_index(vp, &vpi->DstReg),
-					t_dst_mask(vpi->DstReg.WriteMask),
-					t_dst_class(vpi->DstReg.File));
-
-			o_inst->src[0] = MAKE_VSF_SOURCE(t_src_index(vp, &src[1]), t_swizzle(GET_SWZ(src[1].Swizzle, 1)),	// y
-							 t_swizzle(GET_SWZ(src[1].Swizzle, 2)),	// z
-							 t_swizzle(GET_SWZ(src[1].Swizzle, 0)),	// x
-							 t_swizzle(GET_SWZ(src[1].Swizzle, 3)),	// w
-							 t_src_class(src[1].
-								     File),
-							 (!src[1].
-							  NegateBase) ?
-							 VSF_FLAG_ALL :
-							 VSF_FLAG_NONE) |
-			    (src[1].RelAddr << 4);
-
-			o_inst->src[1] = MAKE_VSF_SOURCE(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 2)),	// z
-							 t_swizzle(GET_SWZ(src[0].Swizzle, 0)),	// x
-							 t_swizzle(GET_SWZ(src[0].Swizzle, 1)),	// y
-							 t_swizzle(GET_SWZ(src[0].Swizzle, 3)),	// w
-							 t_src_class(src[0].
-								     File),
-							 src[0].
-							 NegateBase ?
-							 VSF_FLAG_ALL :
-							 VSF_FLAG_NONE) |
-			    (src[0].RelAddr << 4);
-
-			o_inst->src[2] = MAKE_VSF_SOURCE(u_temp_i + 1,
-							 VSF_IN_COMPONENT_X,
-							 VSF_IN_COMPONENT_Y,
-							 VSF_IN_COMPONENT_Z,
-							 VSF_IN_COMPONENT_W,
-							 VSF_IN_CLASS_TMP,
-							 VSF_FLAG_NONE);
-
-			goto next;
-
-		case OPCODE_RCC:
-			fprintf(stderr, "Dont know how to handle op %d yet\n",
-				vpi->Opcode);
-			_mesa_exit(-1);
+			/* FIXME */
+			o_inst = t_opcode_flr(vp, vpi, o_inst, src, &u_temp_i);
+			break;
+		case OPCODE_FRC:
+			t_opcode_frc(vp, vpi, o_inst, src);
+			break;
+		case OPCODE_LG2:
+			t_opcode_lg2(vp, vpi, o_inst, src);
+			break;
+		case OPCODE_LIT:
+			t_opcode_lit(vp, vpi, o_inst, src);
+			break;
+		case OPCODE_LOG:
+			t_opcode_log(vp, vpi, o_inst, src);
+			break;
+		case OPCODE_MAD:
+			t_opcode_mad(vp, vpi, o_inst, src);
+			break;
+		case OPCODE_MAX:
+			t_opcode_max(vp, vpi, o_inst, src);
+			break;
+		case OPCODE_MIN:
+			t_opcode_min(vp, vpi, o_inst, src);
+			break;
+		case OPCODE_MOV:
+			t_opcode_mov(vp, vpi, o_inst, src);
 			break;
-		case OPCODE_END:
+		case OPCODE_MUL:
+			t_opcode_mul(vp, vpi, o_inst, src);
+			break;
+		case OPCODE_POW:
+			t_opcode_pow(vp, vpi, o_inst, src);
+			break;
+		case OPCODE_RCP:
+			t_opcode_rcp(vp, vpi, o_inst, src);
+			break;
+		case OPCODE_RSQ:
+			t_opcode_rsq(vp, vpi, o_inst, src);
+			break;
+		case OPCODE_SGE:
+			t_opcode_sge(vp, vpi, o_inst, src);
+			break;
+		case OPCODE_SLT:
+			t_opcode_slt(vp, vpi, o_inst, src);
+			break;
+		case OPCODE_SUB:
+			t_opcode_sub(vp, vpi, o_inst, src);
+			break;
+		case OPCODE_SWZ:
+			t_opcode_swz(vp, vpi, o_inst, src);
+			break;
+		case OPCODE_XPD:
+			/* FIXME */
+			o_inst = t_opcode_xpd(vp, vpi, o_inst, src, &u_temp_i);
 			break;
 		default:
+			assert(0);
 			break;
 		}
-
-		o_inst->op =
-		    MAKE_VSF_OP(t_opcode(vpi->Opcode),
-				t_dst_index(vp, &vpi->DstReg),
-				t_dst_mask(vpi->DstReg.WriteMask),
-				t_dst_class(vpi->DstReg.File));
-
-		if (are_srcs_scalar) {
-			switch (operands) {
-			case 1:
-				o_inst->src[0] = t_src_scalar(vp, &src[0]);
-				o_inst->src[1] = ZERO_SRC_0;
-				o_inst->src[2] = ZERO_SRC_0;
-				break;
-
-			case 2:
-				o_inst->src[0] = t_src_scalar(vp, &src[0]);
-				o_inst->src[1] = t_src_scalar(vp, &src[1]);
-				o_inst->src[2] = ZERO_SRC_1;
-				break;
-
-			case 3:
-				o_inst->src[0] = t_src_scalar(vp, &src[0]);
-				o_inst->src[1] = t_src_scalar(vp, &src[1]);
-				o_inst->src[2] = t_src_scalar(vp, &src[2]);
-				break;
-
-			default:
-				fprintf(stderr,
-					"scalars and op RCC not handled yet");
-				_mesa_exit(-1);
-				break;
-			}
-		} else {
-			switch (operands) {
-			case 1:
-				o_inst->src[0] = t_src(vp, &src[0]);
-				o_inst->src[1] = ZERO_SRC_0;
-				o_inst->src[2] = ZERO_SRC_0;
-				break;
-
-			case 2:
-				o_inst->src[0] = t_src(vp, &src[0]);
-				o_inst->src[1] = t_src(vp, &src[1]);
-				o_inst->src[2] = ZERO_SRC_1;
-				break;
-
-			case 3:
-				o_inst->src[0] = t_src(vp, &src[0]);
-				o_inst->src[1] = t_src(vp, &src[1]);
-				o_inst->src[2] = t_src(vp, &src[2]);
-				break;
-
-			default:
-				fprintf(stderr,
-					"scalars and op RCC not handled yet");
-				_mesa_exit(-1);
-				break;
-			}
-		}
-	      next:;
 	}
 
-	/* Will most likely segfault before we get here... fix later. */
-	if (o_inst - vp->program.body.i >= VSF_MAX_FRAGMENT_LENGTH / 4) {
+	vp->program.length = (o_inst - vp->program.body.i) * 4;
+	if (vp->program.length >= VSF_MAX_FRAGMENT_LENGTH) {
 		vp->program.length = 0;
 		vp->native = GL_FALSE;
-		return;
 	}
-	vp->program.length = (o_inst - vp->program.body.i) * 4;
 #if 0
 	fprintf(stderr, "hw program:\n");
 	for (i = 0; i < vp->program.length; i++)
@@ -1065,7 +1291,8 @@ static void position_invariant(struct gl_program *prog)
 	struct gl_program_parameter_list *paramList;
 	int i;
 
-	gl_state_index tokens[STATE_LENGTH] = { STATE_MVP_MATRIX, 0, 0, 0, 0 };
+	gl_state_index tokens[STATE_LENGTH] =
+	    { STATE_MVP_MATRIX, 0, 0, 0, 0 };
 
 	/* tokens[4] = matrix modifier */
 #ifdef PREFER_DP4
@@ -1159,8 +1386,8 @@ static void insert_wpos(struct r300_vertex_program *vp,
 				prog->NumInstructions - 1);
 	/* END */
 	_mesa_copy_instructions(&vpi[prog->NumInstructions + 1],
-				&prog->Instructions[prog->NumInstructions - 1],
-				1);
+				&prog->Instructions[prog->NumInstructions -
+						    1], 1);
 	vpi_insert = &vpi[prog->NumInstructions - 1];
 
 	vpi_insert[i].Opcode = OPCODE_MOV;
@@ -1206,8 +1433,8 @@ static void pos_as_texcoord(struct r300_vertex_program *vp,
 	prog->NumTemporaries++;
 
 	for (vpi = prog->Instructions; vpi->Opcode != OPCODE_END; vpi++) {
-		if (vpi->DstReg.File == PROGRAM_OUTPUT &&
-		    vpi->DstReg.Index == VERT_RESULT_HPOS) {
+		if (vpi->DstReg.File == PROGRAM_OUTPUT
+		    && vpi->DstReg.Index == VERT_RESULT_HPOS) {
 			vpi->DstReg.File = PROGRAM_TEMPORARY;
 			vpi->DstReg.Index = tempregi;
 		}
@@ -1223,20 +1450,18 @@ static struct r300_vertex_program *build_program(struct r300_vertex_program_key
 
 	vp = _mesa_calloc(sizeof(*vp));
 	_mesa_memcpy(&vp->key, wanted_key, sizeof(vp->key));
-
 	vp->wpos_idx = wpos_idx;
 
 	if (mesa_vp->IsPositionInvariant) {
 		position_invariant(&mesa_vp->Base);
 	}
 
-	if (wpos_idx > -1)
+	if (wpos_idx > -1) {
 		pos_as_texcoord(vp, &mesa_vp->Base);
+	}
 
 	assert(mesa_vp->Base.NumInstructions);
-
 	vp->num_temporaries = mesa_vp->Base.NumTemporaries;
-
 	r300TranslateVertexShader(vp, mesa_vp->Base.Instructions);
 
 	return vp;
@@ -1252,11 +1477,10 @@ void r300SelectVertexShader(r300ContextPtr r300)
 	struct r300_vertex_program *vp;
 	GLint wpos_idx;
 
-	vpc = (struct r300_vertex_program_cont *)ctx->VertexProgram._Current;
+	vpc =
+	    (struct r300_vertex_program_cont *)ctx->VertexProgram._Current;
 	InputsRead = ctx->FragmentProgram._Current->Base.InputsRead;
 
-	wanted_key.OutputsWritten |= 1 << VERT_RESULT_HPOS;
-
 	wpos_idx = -1;
 	if (InputsRead & FRAG_BIT_WPOS) {
 		for (i = 0; i < ctx->Const.MaxTextureUnits; i++)
@@ -1271,28 +1495,35 @@ void r300SelectVertexShader(r300ContextPtr r300)
 		InputsRead |= (FRAG_BIT_TEX0 << i);
 		wpos_idx = i;
 	}
+	wanted_key.InputsRead = vpc->mesa_program.Base.InputsRead;
+	wanted_key.OutputsWritten = vpc->mesa_program.Base.OutputsWritten;
+
+	wanted_key.OutputsWritten |= 1 << VERT_RESULT_HPOS;
 
-	if (InputsRead & FRAG_BIT_COL0)
+	if (InputsRead & FRAG_BIT_COL0) {
 		wanted_key.OutputsWritten |= 1 << VERT_RESULT_COL0;
+	}
 
-	if ((InputsRead & FRAG_BIT_COL1)	/*||
-						   (InputsRead & FRAG_BIT_FOGC) */ )
+	if ((InputsRead & FRAG_BIT_COL1)) {
 		wanted_key.OutputsWritten |= 1 << VERT_RESULT_COL1;
+	}
 
-	for (i = 0; i < ctx->Const.MaxTextureUnits; i++)
-		if (InputsRead & (FRAG_BIT_TEX0 << i))
+	for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
+		if (InputsRead & (FRAG_BIT_TEX0 << i)) {
 			wanted_key.OutputsWritten |=
 			    1 << (VERT_RESULT_TEX0 + i);
+		}
+	}
 
-	wanted_key.InputsRead = vpc->mesa_program.Base.InputsRead;
 	if (vpc->mesa_program.IsPositionInvariant) {
 		/* we wan't position don't we ? */
 		wanted_key.InputsRead |= (1 << VERT_ATTRIB_POS);
+		wanted_key.OutputsWritten |= (1 << VERT_RESULT_HPOS);
 	}
 
 	for (vp = vpc->progs; vp; vp = vp->next)
-		if (_mesa_memcmp(&vp->key, &wanted_key, sizeof(wanted_key)) ==
-		    0) {
+		if (_mesa_memcmp(&vp->key, &wanted_key, sizeof(wanted_key))
+		    == 0) {
 			r300->selected_vp = vp;
 			return;
 		}
diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.h b/src/mesa/drivers/dri/r300/r300_vertprog.h
index 252d5a901f4..2d399e243a8 100644
--- a/src/mesa/drivers/dri/r300/r300_vertprog.h
+++ b/src/mesa/drivers/dri/r300/r300_vertprog.h
@@ -3,10 +3,10 @@
 
 #include "r300_reg.h"
 
-typedef struct {
-	GLuint op;
+struct r300_vertprog_instruction {
+	GLuint opcode;
 	GLuint src[3];
-} VERTEX_SHADER_INSTRUCTION;
+};
 
 #define VSF_FLAG_X	1
 #define VSF_FLAG_Y	2
diff --git a/src/mesa/main/api_validate.c b/src/mesa/main/api_validate.c
index 3d20ba7d144..841c6a53026 100644
--- a/src/mesa/main/api_validate.c
+++ b/src/mesa/main/api_validate.c
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.1
+ * Version:  7.0.1
  *
- * Copyright (C) 1999-2003  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2007  Brian Paul   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"),
@@ -100,6 +100,11 @@ _mesa_validate_DrawElements(GLcontext *ctx,
                          (const GLubyte *) indices);
       }
    }
+   else {
+      /* not using a VBO */
+      if (!indices)
+         return GL_FALSE;
+   }
 
    if (ctx->Const.CheckArrayBounds) {
       /* find max array index */
@@ -170,6 +175,16 @@ _mesa_validate_DrawRangeElements(GLcontext *ctx, GLenum mode,
        && !(ctx->VertexProgram._Enabled && ctx->Array.ArrayObj->VertexAttrib[0].Enabled))
       return GL_FALSE;
 
+   /* Vertex buffer object tests */
+   if (ctx->Array.ElementArrayBufferObj->Name) {
+      /* XXX re-use code from above? */
+   }
+   else {
+      /* not using VBO */
+      if (!indices)
+         return GL_FALSE;
+   }
+
    if (ctx->Const.CheckArrayBounds) {
       /* Find max array index.
        * We don't trust the user's start and end values.
diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c
index 21adcf32106..2ad1badac70 100644
--- a/src/mesa/main/context.c
+++ b/src/mesa/main/context.c
@@ -1495,9 +1495,20 @@ _mesa_make_current( GLcontext *newCtx, GLframebuffer *drawBuffer,
           */
          if (!newCtx->DrawBuffer || newCtx->DrawBuffer->Name == 0) {
             _mesa_reference_framebuffer(&newCtx->DrawBuffer, drawBuffer);
+         /* fix up the fb fields - these will end up wrong otherwise
+            if the DRIdrawable changes, and someone may rely on them.
+          */
+            /* What a mess!?! */
+            int i;
+            GLenum buffers[MAX_DRAW_BUFFERS];
+            for(i = 0; i < newCtx->Const.MaxDrawBuffers; i++) {
+               buffers[i] = newCtx->Color.DrawBuffer[i];
+            }
+            _mesa_drawbuffers(newCtx, newCtx->Const.MaxDrawBuffers, buffers, NULL);
          }
          if (!newCtx->ReadBuffer || newCtx->ReadBuffer->Name == 0) {
             _mesa_reference_framebuffer(&newCtx->ReadBuffer, readBuffer);
+            _mesa_ReadBuffer(newCtx->Pixel.ReadBuffer);
          }
 
 	 newCtx->NewState |= _NEW_BUFFERS;
diff --git a/src/mesa/main/framebuffer.c b/src/mesa/main/framebuffer.c
index 1fd31a53211..dc10d9ffbc8 100644
--- a/src/mesa/main/framebuffer.c
+++ b/src/mesa/main/framebuffer.c
@@ -65,7 +65,9 @@ compute_depth_max(struct gl_framebuffer *fb)
       fb->_DepthMax = 0xffffffff;
    }
    fb->_DepthMaxF = (GLfloat) fb->_DepthMax;
-   fb->_MRD = 1.0;  /* Minimum resolvable depth value, for polygon offset */
+
+   /* Minimum resolvable depth value, for polygon offset */
+   fb->_MRD = 1.0 / fb->_DepthMaxF;
 }
 
 
diff --git a/src/mesa/main/image.c b/src/mesa/main/image.c
index e874719e645..ba46cdc1b17 100644
--- a/src/mesa/main/image.c
+++ b/src/mesa/main/image.c
@@ -3882,9 +3882,19 @@ _mesa_pack_stencil_span( const GLcontext *ctx, GLuint n,
         }                                                               \
     } while (0)
 
+
+/**
+ * Unpack a row of depth/z values from memory, returning GLushort, GLuint
+ * or GLfloat values.
+ * The glPixelTransfer (scale/bias) params will be applied.
+ *
+ * \param dstType  one of GL_UNSIGNED_SHORT, GL_UNSIGNED_INT, GL_FLOAT
+ * \param depthMax  max value for returned GLushort or GLuint values
+ *                  (ignored for GLfloat).
+ */
 void
 _mesa_unpack_depth_span( const GLcontext *ctx, GLuint n,
-                         GLenum dstType, GLvoid *dest, GLfloat depthScale,
+                         GLenum dstType, GLvoid *dest, GLuint depthMax,
                          GLenum srcType, const GLvoid *source,
                          const struct gl_pixelstore_attrib *srcPacking )
 {
@@ -3907,7 +3917,9 @@ _mesa_unpack_depth_span( const GLcontext *ctx, GLuint n,
          }
          return;
       }
-      if (srcType == GL_UNSIGNED_SHORT && dstType == GL_UNSIGNED_INT) {
+      if (srcType == GL_UNSIGNED_SHORT
+          && dstType == GL_UNSIGNED_INT
+          && depthMax == 0xffffffff) {
          const GLushort *src = (const GLushort *) source;
          GLuint *dst = (GLuint *) dest;
          GLuint i;
@@ -3955,7 +3967,7 @@ _mesa_unpack_depth_span( const GLcontext *ctx, GLuint n,
          break;
       case GL_UNSIGNED_INT_24_8_EXT: /* GL_EXT_packed_depth_stencil */
          if (dstType == GL_UNSIGNED_INT &&
-             depthScale == (GLfloat) 0xffffff &&
+             depthMax == 0xffffff &&
              ctx->Pixel.DepthScale == 1.0 &&
              ctx->Pixel.DepthBias == 0.0) {
             const GLuint *src = (const GLuint *) source;
@@ -4033,16 +4045,16 @@ _mesa_unpack_depth_span( const GLcontext *ctx, GLuint n,
    if (dstType == GL_UNSIGNED_INT) {
       GLuint *zValues = (GLuint *) dest;
       GLuint i;
-      if (depthScale <= (GLfloat) 0xffffff) {
+      if (depthMax <= 0xffffff) {
          /* no overflow worries */
          for (i = 0; i < n; i++) {
-            zValues[i] = (GLuint) (depthValues[i] * depthScale);
+            zValues[i] = (GLuint) (depthValues[i] * (GLfloat) depthMax);
          }
       }
       else {
          /* need to use double precision to prevent overflow problems */
          for (i = 0; i < n; i++) {
-            GLdouble z = depthValues[i] * depthScale;
+            GLdouble z = depthValues[i] * (GLfloat) depthMax;
             if (z >= (GLdouble) 0xffffffff)
                zValues[i] = 0xffffffff;
             else
@@ -4053,14 +4065,14 @@ _mesa_unpack_depth_span( const GLcontext *ctx, GLuint n,
    else if (dstType == GL_UNSIGNED_SHORT) {
       GLushort *zValues = (GLushort *) dest;
       GLuint i;
-      ASSERT(depthScale <= 65535.0);
+      ASSERT(depthMax <= 0xffff);
       for (i = 0; i < n; i++) {
-         zValues[i] = (GLushort) (depthValues[i] * depthScale);
+         zValues[i] = (GLushort) (depthValues[i] * (GLfloat) depthMax);
       }
    }
    else {
       ASSERT(dstType == GL_FLOAT);
-      ASSERT(depthScale == 1.0F);
+      /*ASSERT(depthMax == 1.0F);*/
    }
 }
 
diff --git a/src/mesa/main/image.h b/src/mesa/main/image.h
index 990398a7c45..2a16989fa7e 100644
--- a/src/mesa/main/image.h
+++ b/src/mesa/main/image.h
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5.2
+ * Version:  7.1
  *
- * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2007  Brian Paul   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"),
@@ -181,7 +181,7 @@ _mesa_pack_stencil_span( const GLcontext *ctx, GLuint n,
 
 extern void
 _mesa_unpack_depth_span( const GLcontext *ctx, GLuint n,
-                         GLenum dstType, GLvoid *dest, GLfloat depthScale,
+                         GLenum dstType, GLvoid *dest, GLuint depthMax,
                          GLenum srcType, const GLvoid *source,
                          const struct gl_pixelstore_attrib *srcPacking );
 
diff --git a/src/mesa/main/teximage.c b/src/mesa/main/teximage.c
index f902365b9bd..3420d8e2baf 100644
--- a/src/mesa/main/teximage.c
+++ b/src/mesa/main/teximage.c
@@ -1264,6 +1264,10 @@ _mesa_init_teximage_fields(GLcontext *ctx, GLenum target,
  * A hardware driver might override this function if, for example, the
  * max 3D texture size is 512x512x64 (i.e. not a cube).
  *
+ * Note that width, height, depth == 0 is not an error.  However, a
+ * texture with zero width/height/depth will be considered "incomplete"
+ * and texturing will effectively be disabled.
+ *
  * \param target  one of GL_PROXY_TEXTURE_1D, GL_PROXY_TEXTURE_2D,
  *                GL_PROXY_TEXTURE_3D, GL_PROXY_TEXTURE_RECTANGLE_NV,
  *                GL_PROXY_TEXTURE_CUBE_MAP_ARB.
@@ -1293,7 +1297,7 @@ _mesa_test_proxy_teximage(GLcontext *ctx, GLenum target, GLint level,
       maxSize = 1 << (ctx->Const.MaxTextureLevels - 1);
       if (width < 2 * border || width > 2 + maxSize ||
           (!ctx->Extensions.ARB_texture_non_power_of_two &&
-           _mesa_bitcount(width - 2 * border) != 1) ||
+           width >0 && _mesa_bitcount(width - 2 * border) != 1) ||
           level >= ctx->Const.MaxTextureLevels) {
          /* bad width or level */
          return GL_FALSE;
@@ -1303,10 +1307,10 @@ _mesa_test_proxy_teximage(GLcontext *ctx, GLenum target, GLint level,
       maxSize = 1 << (ctx->Const.MaxTextureLevels - 1);
       if (width < 2 * border || width > 2 + maxSize ||
           (!ctx->Extensions.ARB_texture_non_power_of_two &&
-           _mesa_bitcount(width - 2 * border) != 1) ||
+           width > 0 && _mesa_bitcount(width - 2 * border) != 1) ||
           height < 2 * border || height > 2 + maxSize ||
           (!ctx->Extensions.ARB_texture_non_power_of_two &&
-           _mesa_bitcount(height - 2 * border) != 1) ||
+           height > 0 && _mesa_bitcount(height - 2 * border) != 1) ||
           level >= ctx->Const.MaxTextureLevels) {
          /* bad width or height or level */
          return GL_FALSE;
@@ -1316,21 +1320,21 @@ _mesa_test_proxy_teximage(GLcontext *ctx, GLenum target, GLint level,
       maxSize = 1 << (ctx->Const.Max3DTextureLevels - 1);
       if (width < 2 * border || width > 2 + maxSize ||
           (!ctx->Extensions.ARB_texture_non_power_of_two &&
-           _mesa_bitcount(width - 2 * border) != 1) ||
+           width > 0 && _mesa_bitcount(width - 2 * border) != 1) ||
           height < 2 * border || height > 2 + maxSize ||
           (!ctx->Extensions.ARB_texture_non_power_of_two &&
-           _mesa_bitcount(height - 2 * border) != 1) ||
+           height > 0 && _mesa_bitcount(height - 2 * border) != 1) ||
           depth < 2 * border || depth > 2 + maxSize ||
           (!ctx->Extensions.ARB_texture_non_power_of_two &&
-           _mesa_bitcount(depth - 2 * border) != 1) ||
+           depth > 0 && _mesa_bitcount(depth - 2 * border) != 1) ||
           level >= ctx->Const.Max3DTextureLevels) {
          /* bad width or height or depth or level */
          return GL_FALSE;
       }
       return GL_TRUE;
    case GL_PROXY_TEXTURE_RECTANGLE_NV:
-      if (width < 1 || width > ctx->Const.MaxTextureRectSize ||
-          height < 1 || height > ctx->Const.MaxTextureRectSize ||
+      if (width < 0 || width > ctx->Const.MaxTextureRectSize ||
+          height < 0 || height > ctx->Const.MaxTextureRectSize ||
           level != 0) {
          /* bad width or height or level */
          return GL_FALSE;
@@ -1340,10 +1344,10 @@ _mesa_test_proxy_teximage(GLcontext *ctx, GLenum target, GLint level,
       maxSize = 1 << (ctx->Const.MaxCubeTextureLevels - 1);
       if (width < 2 * border || width > 2 + maxSize ||
           (!ctx->Extensions.ARB_texture_non_power_of_two &&
-           _mesa_bitcount(width - 2 * border) != 1) ||
+           width > 0 && _mesa_bitcount(width - 2 * border) != 1) ||
           height < 2 * border || height > 2 + maxSize ||
           (!ctx->Extensions.ARB_texture_non_power_of_two &&
-           _mesa_bitcount(height - 2 * border) != 1) ||
+           height > 0 && _mesa_bitcount(height - 2 * border) != 1) ||
           level >= ctx->Const.MaxCubeTextureLevels) {
          /* bad width or height */
          return GL_FALSE;
@@ -1353,7 +1357,7 @@ _mesa_test_proxy_teximage(GLcontext *ctx, GLenum target, GLint level,
       maxSize = 1 << (ctx->Const.MaxTextureLevels - 1);
       if (width < 2 * border || width > 2 + maxSize ||
           (!ctx->Extensions.ARB_texture_non_power_of_two &&
-           _mesa_bitcount(width - 2 * border) != 1) ||
+           width > 0 && _mesa_bitcount(width - 2 * border) != 1) ||
           level >= ctx->Const.MaxTextureLevels) {
          /* bad width or level */
          return GL_FALSE;
@@ -1367,10 +1371,10 @@ _mesa_test_proxy_teximage(GLcontext *ctx, GLenum target, GLint level,
       maxSize = 1 << (ctx->Const.MaxTextureLevels - 1);
       if (width < 2 * border || width > 2 + maxSize ||
           (!ctx->Extensions.ARB_texture_non_power_of_two &&
-           _mesa_bitcount(width - 2 * border) != 1) ||
+           width > 0 && _mesa_bitcount(width - 2 * border) != 1) ||
           height < 2 * border || height > 2 + maxSize ||
           (!ctx->Extensions.ARB_texture_non_power_of_two &&
-           _mesa_bitcount(height - 2 * border) != 1) ||
+           height > 0 && _mesa_bitcount(height - 2 * border) != 1) ||
           level >= ctx->Const.MaxTextureLevels) {
          /* bad width or height or level */
          return GL_FALSE;
@@ -1472,7 +1476,7 @@ texture_error_check( GLcontext *ctx, GLenum target,
       if (target == GL_PROXY_TEXTURE_1D || target == GL_TEXTURE_1D) {
          proxy_target = GL_PROXY_TEXTURE_1D;
          height = 1;
-         width = 1;
+         depth = 1;
       }
       else {
          _mesa_error( ctx, GL_INVALID_ENUM, "glTexImage1D(target)" );
diff --git a/src/mesa/main/texstore.c b/src/mesa/main/texstore.c
index 9b8a06df145..3b5151ed171 100644
--- a/src/mesa/main/texstore.c
+++ b/src/mesa/main/texstore.c
@@ -1077,7 +1077,7 @@ _mesa_texstore_rgba(TEXSTORE_PARAMS)
 GLboolean
 _mesa_texstore_z32(TEXSTORE_PARAMS)
 {
-   const GLfloat depthScale = (GLfloat) 0xffffffff;
+   const GLuint depthScale = 0xffffffff;
    (void) dims;
    ASSERT(dstFormat == &_mesa_texformat_z32);
    ASSERT(dstFormat->TexelBytes == sizeof(GLuint));
@@ -1124,7 +1124,7 @@ _mesa_texstore_z32(TEXSTORE_PARAMS)
 GLboolean
 _mesa_texstore_z16(TEXSTORE_PARAMS)
 {
-   const GLfloat depthScale = 65535.0f;
+   const GLuint depthScale = 0xffff;
    (void) dims;
    ASSERT(dstFormat == &_mesa_texformat_z16);
    ASSERT(dstFormat->TexelBytes == sizeof(GLushort));
@@ -2319,6 +2319,8 @@ _mesa_texstore_ycbcr(TEXSTORE_PARAMS)
 GLboolean
 _mesa_texstore_z24_s8(TEXSTORE_PARAMS)
 {
+   const GLuint depthScale = 0xffffff;
+
    ASSERT(dstFormat == &_mesa_texformat_z24_s8);
    ASSERT(srcFormat == GL_DEPTH_STENCIL_EXT);
    ASSERT(srcType == GL_UNSIGNED_INT_24_8_EXT);
@@ -2357,7 +2359,7 @@ _mesa_texstore_z24_s8(TEXSTORE_PARAMS)
             _mesa_unpack_depth_span(ctx, srcWidth,
                                     GL_UNSIGNED_INT, /* dst type */
                                     dstRow, /* dst addr */
-                                    (GLfloat) 0xffffff, /* depthScale */
+                                    depthScale,
                                     srcType, src, srcPacking);
             /* get the 8-bit stencil values */
             _mesa_unpack_stencil_span(ctx, srcWidth,
diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h
index fd5e7ad3af9..e3f62a80ad7 100644
--- a/src/mesa/pipe/p_state.h
+++ b/src/mesa/pipe/p_state.h
@@ -121,8 +121,11 @@ struct pipe_clip_state {
    GLuint nr;
 };
 
+
 struct pipe_fs_state {
-   struct gl_fragment_program *fp;
+   GLuint inputs_read;		/* FRAG_ATTRIB_* */
+   const struct tgsi_token *tokens;
+   
 };
 
 struct pipe_constant_buffer {
diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c
index 7b1c90cba66..536365f2193 100644
--- a/src/mesa/pipe/softpipe/sp_quad_fs.c
+++ b/src/mesa/pipe/softpipe/sp_quad_fs.c
@@ -37,6 +37,7 @@
 #include "sp_context.h"
 #include "sp_headers.h"
 #include "sp_quad.h"
+#include "core/tgsi_core.h"
 
 struct exec_machine {
    const struct setup_coefficient *coef; /**< will point to quad->coef */
@@ -180,8 +181,47 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad )
 #endif
    }
 
-#if 0
-   softpipe->run_fs( tri->fp, quad, &tri->outputs );
+#if 1
+   /*softpipe->run_fs( tri->fp, quad, &tri->outputs );*/
+
+   {
+      struct tgsi_exec_machine machine;
+      struct tgsi_exec_vector inputs[FRAG_ATTRIB_MAX + 1];
+      struct tgsi_exec_vector outputs[FRAG_ATTRIB_MAX + 1];
+      struct tgsi_exec_vector *ainputs;
+      struct tgsi_exec_vector *aoutputs;
+      GLuint i, total;
+
+      ainputs = (struct tgsi_exec_vector *) tgsi_align_128bit( inputs );
+      aoutputs = (struct tgsi_exec_vector *) tgsi_align_128bit( outputs );
+
+      for( i = total = 0; i < PIPE_ATTRIB_MAX; i++ ) {
+         GLuint attr;
+
+         attr = softpipe->fp_attr_to_slot[i];
+         if( attr ) {
+            assert( total < FRAG_ATTRIB_MAX );
+            assert( attr < FRAG_ATTRIB_MAX );
+            assert( sizeof( ainputs[0] ) == sizeof( exec.attr[0] ) );
+
+            memcpy(
+               &ainputs[total],
+               exec.attr[attr],
+               sizeof( ainputs[0] ) );
+            total++;
+         }
+      }
+
+      tgsi_exec_machine_init(
+         &machine,
+         softpipe->fs.tokens );
+
+      machine.Inputs = ainputs;
+      machine.Outputs = aoutputs;
+
+      tgsi_exec_machine_run(
+         &machine );
+   }
 #else
    {
       GLuint attr = softpipe->fp_attr_to_slot[FRAG_ATTRIB_COL0];
diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c
index 18dfb50e381..fcdedb54a93 100644
--- a/src/mesa/pipe/softpipe/sp_state_derived.c
+++ b/src/mesa/pipe/softpipe/sp_state_derived.c
@@ -28,7 +28,6 @@
 #include "main/glheader.h"
 #include "main/macros.h"
 #include "main/enums.h"
-#include "shader/program.h"
 
 #include "vf/vf.h"
 #include "pipe/draw/draw_context.h"
@@ -68,8 +67,7 @@ static const GLuint frag_to_vf[FRAG_ATTRIB_MAX] =
  */
 static void calculate_vertex_layout( struct softpipe_context *softpipe )
 {
-   struct gl_fragment_program *fp = softpipe->fs.fp;
-   const GLuint inputsRead = fp->Base.InputsRead;
+   const GLuint inputsRead = softpipe->fs.inputs_read;
    GLuint slot_to_vf_attr[VF_ATTRIB_MAX];
    GLbitfield attr_mask = 0x0;
    GLuint i;
diff --git a/src/mesa/pipe/tgsi/core/tgsi_dump.c b/src/mesa/pipe/tgsi/core/tgsi_dump.c
index fecb246ab1a..0345fd93f7f 100644
--- a/src/mesa/pipe/tgsi/core/tgsi_dump.c
+++ b/src/mesa/pipe/tgsi/core/tgsi_dump.c
@@ -400,12 +400,16 @@ tgsi_dump(
    GLuint deflt = !(flags & TGSI_DUMP_NO_DEFAULT);
 
    {
+#if 0
       static GLuint counter = 0;
       char buffer[64];
 
       sprintf( buffer, "sbir-dump-%.4u.txt", counter++ );
       dump.file = fopen( buffer, "wt" );
+#else
+      dump.file = stderr;
       dump.tabs = 0;
+#endif
    }
 
    tgsi_parse_init( &parse, tokens );
diff --git a/src/mesa/pipe/tgsi/core/tgsi_util.c b/src/mesa/pipe/tgsi/core/tgsi_util.c
index 2331affdfdf..38d6d6e6bcd 100644
--- a/src/mesa/pipe/tgsi/core/tgsi_util.c
+++ b/src/mesa/pipe/tgsi/core/tgsi_util.c
@@ -1,15 +1,22 @@
 #include "tgsi_platform.h"
 #include "tgsi_core.h"
 
+union pointer_hack
+{
+   void *pointer;
+   unsigned long long uint64;
+};
+
 void *
 tgsi_align_128bit(
    void *unaligned )
 {
-   GLuint *ptr, addr;
+   union pointer_hack ph;
 
-   ptr = (GLuint *) unaligned;
-   addr = (*(GLuint *) &ptr + 15) & ~15;
-   return *(void **) &addr;
+   ph.uint64 = 0;
+   ph.pointer = unaligned;
+   ph.uint64 = (ph.uint64 + 15) & ~15;
+   return ph.pointer;
 }
 
 GLuint
diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h
index 4c1141e5796..9256318997a 100644
--- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h
+++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h
@@ -5,6 +5,8 @@
 extern "C" {

 #endif // defined __cplusplus

 

+struct tgsi_token;

+

 GLboolean

 tgsi_mesa_compile_fp_program(

    const struct gl_fragment_program *program,

diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c
index 66509d56db1..1a931326afa 100644
--- a/src/mesa/shader/shader_api.c
+++ b/src/mesa/shader/shader_api.c
@@ -779,7 +779,7 @@ _mesa_get_programiv(GLcontext *ctx, GLuint program,
       *params = shProg->Validated;
       break;
    case GL_INFO_LOG_LENGTH:
-      *params = shProg->InfoLog ? strlen(shProg->InfoLog) : 0;
+      *params = shProg->InfoLog ? strlen(shProg->InfoLog) + 1 : 0;
       break;
    case GL_ATTACHED_SHADERS:
       *params = shProg->NumShaders;
@@ -831,10 +831,10 @@ _mesa_get_shaderiv(GLcontext *ctx, GLuint name, GLenum pname, GLint *params)
       *params = shader->CompileStatus;
       break;
    case GL_INFO_LOG_LENGTH:
-      *params = shader->InfoLog ? strlen(shader->InfoLog) : 0;
+      *params = shader->InfoLog ? strlen(shader->InfoLog) + 1 : 0;
       break;
    case GL_SHADER_SOURCE_LENGTH:
-      *params = shader->Source ? strlen((char *) shader->Source) : 0;
+      *params = shader->Source ? strlen((char *) shader->Source) + 1 : 0;
       break;
    default:
       _mesa_error(ctx, GL_INVALID_ENUM, "glGetShaderiv(pname)");
diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c
index 2b5196f095f..35e80e04523 100644
--- a/src/mesa/shader/slang/slang_codegen.c
+++ b/src/mesa/shader/slang/slang_codegen.c
@@ -1,6 +1,6 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5.3
+ * Version:  7.1
  *
  * Copyright (C) 2005-2007  Brian Paul   All Rights Reserved.
  *
@@ -2328,7 +2328,8 @@ _slang_gen_field(slang_assemble_ctx * A, slang_operation *oper)
          n = _slang_gen_swizzle(n, swizzle);
       return n;
    }
-   else if (ti.spec.type == SLANG_SPEC_FLOAT) {
+   else if (   ti.spec.type == SLANG_SPEC_FLOAT
+            || ti.spec.type == SLANG_SPEC_INT) {
       const GLuint rows = 1;
       slang_swizzle swz;
       slang_ir_node *n;
diff --git a/src/mesa/sources b/src/mesa/sources
index a76e41bdda3..a589ae4373b 100644
--- a/src/mesa/sources
+++ b/src/mesa/sources
@@ -206,6 +206,7 @@ STATETRACKER_SOURCES = \
 	state_tracker/st_atom_clip.c \
 	state_tracker/st_atom_depth.c \
 	state_tracker/st_atom_fs.c \
+	state_tracker/st_atom_vs.c \
 	state_tracker/st_atom_framebuffer.c \
 	state_tracker/st_atom_sampler.c \
 	state_tracker/st_atom_scissor.c \
diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c
index dfebfb4768c..85c99bc1822 100644
--- a/src/mesa/state_tracker/st_atom.c
+++ b/src/mesa/state_tracker/st_atom.c
@@ -46,6 +46,7 @@ static const struct st_tracked_state *atoms[] =
    &st_update_clear_color,
    &st_update_depth,
    &st_update_clip,
+   &st_update_vs,
    &st_update_fs,
    &st_update_setup,
    &st_update_polygon_stipple,
diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h
index a56483ac396..1b70e279339 100644
--- a/src/mesa/state_tracker/st_atom.h
+++ b/src/mesa/state_tracker/st_atom.h
@@ -49,6 +49,7 @@ const struct st_tracked_state st_update_clip;
 const struct st_tracked_state st_update_clear_color;
 const struct st_tracked_state st_update_depth;
 const struct st_tracked_state st_update_fs;
+const struct st_tracked_state st_update_vs;
 const struct st_tracked_state st_update_setup;
 const struct st_tracked_state st_update_polygon_stipple;
 const struct st_tracked_state st_update_viewport;
diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c
index 9c6bc1ce2af..6fa4f53c734 100644
--- a/src/mesa/state_tracker/st_atom_fs.c
+++ b/src/mesa/state_tracker/st_atom_fs.c
@@ -32,15 +32,38 @@
 #include "st_context.h"
 #include "pipe/p_context.h"
 #include "st_atom.h"
+#include "st_program.h"
+#include "pipe/tgsi/mesa/mesa_to_tgsi.h"
+#include "pipe/tgsi/core/tgsi_dump.h"
+
+static void compile_fs( struct st_context *st,
+			struct st_fragment_program *fs )
+{
+   /* XXX: fix static allocation of tokens:
+    */
+   tgsi_mesa_compile_fp_program( &fs->Base, fs->tokens, ST_FP_MAX_TOKENS );
+
+   tgsi_dump( fs->tokens, TGSI_DUMP_VERBOSE );
+}
 
 
 static void update_fs( struct st_context *st )
 {
    struct pipe_fs_state fs;
+   struct st_fragment_program *fp = st_fragment_program(st->ctx->FragmentProgram._Current);
+
+   memset( &fs, 0, sizeof(fs) );
 
-   fs.fp = st->ctx->FragmentProgram._Current;
+   if (fp->dirty)
+      compile_fs( st, fp );
+   
+   fs.inputs_read = fp->Base.Base.InputsRead;
+   fs.tokens = &fp->tokens[0];
    
-   if (memcmp(&fs, &st->state.fs, sizeof(fs)) != 0) {
+   if (memcmp(&fs, &st->state.fs, sizeof(fs)) != 0 ||
+       fp->dirty) 
+   {
+      fp->dirty = 0;
       st->state.fs = fs;
       st->pipe->set_fs_state(st->pipe, &fs);
    }
diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c
new file mode 100644
index 00000000000..6a26bfdd19f
--- /dev/null
+++ b/src/mesa/state_tracker/st_atom_vs.c
@@ -0,0 +1,49 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 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.
+ * 
+ **************************************************************************/
+ /*
+  * Authors:
+  *   Keith Whitwell <[email protected]>
+  */
+                   
+#include "st_context.h"
+#include "pipe/p_context.h"
+#include "st_atom.h"
+
+
+ 
+static void update_vs( struct st_context *st )
+{
+}
+
+
+const struct st_tracked_state st_update_vs = {
+   .dirty = {
+      .mesa  = 0,
+      .st   = ST_NEW_VERTEX_PROGRAM,
+   },
+   .update = update_vs
+};
diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c
index 327b627722f..18061ca69cf 100644
--- a/src/mesa/state_tracker/st_cb_program.c
+++ b/src/mesa/state_tracker/st_cb_program.c
@@ -52,6 +52,7 @@ static void st_bind_program( GLcontext *ctx,
 
    switch (target) {
    case GL_VERTEX_PROGRAM_ARB: 
+      st->dirty.st |= ST_NEW_VERTEX_PROGRAM;
       break;
    case GL_FRAGMENT_PROGRAM_ARB:
       st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM;
@@ -66,16 +67,23 @@ static struct gl_program *st_new_program( GLcontext *ctx,
    struct st_context *st = st_context(ctx);
 
    switch (target) {
-   case GL_VERTEX_PROGRAM_ARB:
-      return _mesa_init_vertex_program(ctx, 
-				       CALLOC_STRUCT(gl_vertex_program),
-                                       target, 
-				       id);
+   case GL_VERTEX_PROGRAM_ARB: {
+      struct st_vertex_program *prog = CALLOC_STRUCT(st_vertex_program);
+
+      prog->id = st->program_id++;
+      prog->dirty = 1;
+
+      return _mesa_init_vertex_program( ctx, 
+					&prog->Base,
+					target, 
+					id );
+   }
 
    case GL_FRAGMENT_PROGRAM_ARB: {
       struct st_fragment_program *prog = CALLOC_STRUCT(st_fragment_program);
 
       prog->id = st->program_id++;
+      prog->dirty = 1;
 
       return _mesa_init_fragment_program( ctx, 
 					  &prog->Base,
@@ -106,40 +114,25 @@ static void st_program_string_notify( GLcontext *ctx,
 				      GLenum target,
 				      struct gl_program *prog )
 {
-   if (target == GL_FRAGMENT_PROGRAM_ARB) {
-      struct st_context *st = st_context(ctx);
+   struct st_context *st = st_context(ctx);
 
-      if (prog == &st->ctx->FragmentProgram._Current->Base) 
-      {
-	 struct st_fragment_program *p = 
-	    (struct st_fragment_program *) prog;
+   if (target == GL_FRAGMENT_PROGRAM_ARB) {
+      struct st_fragment_program *p = (struct st_fragment_program *)prog;
 
+      if (prog == &ctx->FragmentProgram._Current->Base)
 	 st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM;
 
-	 p->id = st->program_id++;
-#if 0
-	 p->param_state = p->Base.Base.Parameters->StateFlags; 
-	 p->translated = 0;
-#endif
-
-	 /* Gack! do this in the compiler: 
-	  */
-	 if (p->Base.FogOption) {
-	    /* add extra instructions to do fog, then turn off FogOption field */
-	    _mesa_append_fog_code(ctx, &p->Base);
-	    p->Base.FogOption = GL_NONE;
-	 }
-
-         /* XXX: Not hooked-up yet. */
-	 {
-	    struct tgsi_token tokens[1024];
-
-	    tgsi_mesa_compile_fp_program( prog, tokens, 1024 );
-	    tgsi_dump( tokens, TGSI_DUMP_VERBOSE );
-	 }
-      }
+      p->id = st->program_id++;      
+      p->param_state = p->Base.Base.Parameters->StateFlags;
    }
    else if (target == GL_VERTEX_PROGRAM_ARB) {
+      struct st_vertex_program *p = (struct st_vertex_program *)prog;
+
+      if (prog == &ctx->VertexProgram._Current->Base)
+	 st->dirty.st |= ST_NEW_VERTEX_PROGRAM;
+
+      p->id = st->program_id++;      
+      p->param_state = p->Base.Base.Parameters->StateFlags;
 
       /* Also tell tnl about it:
        */
diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h
index c1d868604cf..ef3cdb3b091 100644
--- a/src/mesa/state_tracker/st_context.h
+++ b/src/mesa/state_tracker/st_context.h
@@ -40,6 +40,7 @@ struct st_fragment_program;
 
 #define ST_NEW_MESA                    0x1 /* Mesa state has changed */
 #define ST_NEW_FRAGMENT_PROGRAM        0x2
+#define ST_NEW_VERTEX_PROGRAM          0x4
 
 struct st_state_flags {
    GLuint mesa;
diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h
index a47059d7a69..b28887946d5 100644
--- a/src/mesa/state_tracker/st_program.h
+++ b/src/mesa/state_tracker/st_program.h
@@ -34,6 +34,12 @@
 #ifndef ST_PROGRAM_H
 #define ST_PROGRAM_H
 
+#include "mtypes.h"
+#include "pipe/tgsi/core/tgsi_token.h"
+
+#define ST_FP_MAX_TOKENS 1024
+
+
 struct st_fragment_program
 {
    struct gl_fragment_program Base;
@@ -43,6 +49,11 @@ struct st_fragment_program
 				 * ProgramStringNotify changes.
 				 */
 
+
+   struct tgsi_token tokens[ST_FP_MAX_TOKENS];
+   GLboolean dirty;
+   
+
 #if 0   
    GLfloat (*cbuffer)[4];
    GLuint nr_constants;
@@ -56,13 +67,38 @@ struct st_fragment_program
       const GLfloat *values;    /* Pointer to tracked values */
    } *param;
    GLuint nr_params;
+#endif
 
    GLuint param_state;
-#endif
 };
 
 
+struct st_vertex_program
+{
+   struct gl_vertex_program Base;
+   GLboolean error;             /* If program is malformed for any reason. */
+
+   GLuint    id;		/* String id, for tracking
+				 * ProgramStringNotify changes.
+				 */
+
+   GLboolean dirty;
+   GLuint param_state;
+};
+
 void st_init_cb_program( struct st_context *st );
 void st_destroy_cb_program( struct st_context *st );
 
+static inline struct st_fragment_program *
+st_fragment_program( struct gl_fragment_program *fp )
+{
+   return (struct st_fragment_program *)fp;
+}
+
+static inline struct st_vertex_program *
+st_vertex_program( struct gl_vertex_program *vp )
+{
+   return (struct st_vertex_program *)vp;
+}
+
 #endif
diff --git a/src/mesa/swrast/s_drawpix.c b/src/mesa/swrast/s_drawpix.c
index 1c9f64b275e..0cf425e1c62 100644
--- a/src/mesa/swrast/s_drawpix.c
+++ b/src/mesa/swrast/s_drawpix.c
@@ -484,7 +484,7 @@ draw_depth_pixels( GLcontext *ctx, GLint x, GLint y,
    }
    else {
       /* General case */
-      const GLfloat depthMax = ctx->DrawBuffer->_DepthMaxF;
+      const GLuint depthMax = ctx->DrawBuffer->_DepthMax;
       GLint skipPixels = 0;
 
       /* in case width > MAX_WIDTH do the copy in chunks */
@@ -695,7 +695,7 @@ draw_depth_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
    const GLint imgX = x, imgY = y;
    const GLboolean scaleOrBias
       = ctx->Pixel.DepthScale != 1.0 || ctx->Pixel.DepthBias != 0.0;
-   const GLfloat depthScale = ctx->DrawBuffer->_DepthMaxF;
+   const GLuint depthMax = ctx->DrawBuffer->_DepthMax;
    const GLuint stencilMask = ctx->Stencil.WriteMask[0];
    const GLuint stencilType = (STENCIL_BITS == 8) ? 
       GL_UNSIGNED_BYTE : GL_UNSIGNED_SHORT;
@@ -783,7 +783,7 @@ draw_depth_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
                /* general case */
                GLuint zValues[MAX_WIDTH];  /* 16 or 32-bit Z value storage */
                _mesa_unpack_depth_span(ctx, width,
-                                       depthRb->DataType, zValues, depthScale,
+                                       depthRb->DataType, zValues, depthMax,
                                        type, depthStencilSrc, &clippedUnpack);
                if (zoom) {
                   _swrast_write_zoomed_z_span(ctx, imgX, imgY, width, x,
diff --git a/src/mesa/swrast/s_stencil.c b/src/mesa/swrast/s_stencil.c
index 89991fad02d..d0cbdd6917d 100644
--- a/src/mesa/swrast/s_stencil.c
+++ b/src/mesa/swrast/s_stencil.c
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5
+ * Version:  7.1
  *
- * Copyright (C) 1999-2005  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2007  Brian Paul   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"),
@@ -392,6 +392,23 @@ do_stencil_test( GLcontext *ctx, GLuint face, GLuint n, GLstencil stencil[],
 }
 
 
+/**
+ * Compute the zpass/zfail masks by comparing the pre- and post-depth test
+ * masks.
+ */
+static INLINE void
+compute_pass_fail_masks(GLuint n, const GLubyte origMask[],
+                        const GLubyte newMask[],
+                        GLubyte passMask[], GLubyte failMask[])
+{
+   GLuint i;
+   for (i = 0; i < n; i++) {
+      ASSERT(newMask[i] == 0 || newMask[i] == 1);
+      passMask[i] = origMask[i] & newMask[i];
+      failMask[i] = origMask[i] & (newMask[i] ^ 1);
+   }
+}
+
 
 /**
  * Apply stencil and depth testing to the span of pixels.
@@ -460,39 +477,24 @@ stencil_and_ztest_span(GLcontext *ctx, SWspan *span, GLuint face)
       /*
        * Perform depth buffering, then apply zpass or zfail stencil function.
        */
-      GLubyte passmask[MAX_WIDTH], failmask[MAX_WIDTH], oldmask[MAX_WIDTH];
-      GLuint i;
+      GLubyte passMask[MAX_WIDTH], failMask[MAX_WIDTH], origMask[MAX_WIDTH];
 
       /* save the current mask bits */
-      _mesa_memcpy(oldmask, mask, n * sizeof(GLubyte));
+      _mesa_memcpy(origMask, mask, n * sizeof(GLubyte));
 
       /* apply the depth test */
       _swrast_depth_test_span(ctx, span);
 
-      /* Set the stencil pass/fail flags according to result of depth testing.
-       * if oldmask[i] == 0 then
-       *    Don't touch the stencil value
-       * else if oldmask[i] and newmask[i] then
-       *    Depth test passed
-       * else
-       *    assert(oldmask[i] && !newmask[i])
-       *    Depth test failed
-       * endif
-       */
-      for (i=0;i<n;i++) {
-         ASSERT(mask[i] == 0 || mask[i] == 1);
-         passmask[i] = oldmask[i] & mask[i];
-         failmask[i] = oldmask[i] & (mask[i] ^ 1);
-      }
+      compute_pass_fail_masks(n, origMask, mask, passMask, failMask);
 
       /* apply the pass and fail operations */
       if (ctx->Stencil.ZFailFunc[face] != GL_KEEP) {
          apply_stencil_op( ctx, ctx->Stencil.ZFailFunc[face], face,
-                           n, stencil, failmask );
+                           n, stencil, failMask );
       }
       if (ctx->Stencil.ZPassFunc[face] != GL_KEEP) {
          apply_stencil_op( ctx, ctx->Stencil.ZPassFunc[face], face,
-                           n, stencil, passmask );
+                           n, stencil, passMask );
       }
    }
 
@@ -902,6 +904,7 @@ stencil_test_pixels( GLcontext *ctx, GLuint face, GLuint n,
 static GLboolean
 stencil_and_ztest_pixels( GLcontext *ctx, SWspan *span, GLuint face )
 {
+   GLubyte passMask[MAX_WIDTH], failMask[MAX_WIDTH], origMask[MAX_WIDTH];
    struct gl_framebuffer *fb = ctx->DrawBuffer;
    struct gl_renderbuffer *rb = fb->_StencilBuffer;
    const GLuint n = span->end;
@@ -916,13 +919,10 @@ stencil_and_ztest_pixels( GLcontext *ctx, SWspan *span, GLuint face )
    if (!rb->GetPointer(ctx, rb, 0, 0)) {
       /* No direct access */
       GLstencil stencil[MAX_WIDTH];
-      GLubyte origMask[MAX_WIDTH];
 
       ASSERT(rb->DataType == GL_UNSIGNED_BYTE);
       _swrast_get_values(ctx, rb, n, x, y, stencil, sizeof(GLubyte));
 
-      _mesa_memcpy(origMask, mask, n * sizeof(GLubyte));
-
       (void) do_stencil_test(ctx, face, n, stencil, mask);
 
       if (ctx->Depth.Test == GL_FALSE) {
@@ -930,27 +930,19 @@ stencil_and_ztest_pixels( GLcontext *ctx, SWspan *span, GLuint face )
                           n, stencil, mask);
       }
       else {
+         _mesa_memcpy(origMask, mask, n * sizeof(GLubyte));
+
          _swrast_depth_test_span(ctx, span);
 
+         compute_pass_fail_masks(n, origMask, mask, passMask, failMask);
+
          if (ctx->Stencil.ZFailFunc[face] != GL_KEEP) {
-            GLubyte failmask[MAX_WIDTH];
-            GLuint i;
-            for (i = 0; i < n; i++) {
-               ASSERT(mask[i] == 0 || mask[i] == 1);
-               failmask[i] = origMask[i] & (mask[i] ^ 1);
-            }
             apply_stencil_op(ctx, ctx->Stencil.ZFailFunc[face], face,
-                             n, stencil, failmask);
+                             n, stencil, failMask);
          }
          if (ctx->Stencil.ZPassFunc[face] != GL_KEEP) {
-            GLubyte passmask[MAX_WIDTH];
-            GLuint i;
-            for (i = 0; i < n; i++) {
-               ASSERT(mask[i] == 0 || mask[i] == 1);
-               passmask[i] = origMask[i] & mask[i];
-            }
             apply_stencil_op(ctx, ctx->Stencil.ZPassFunc[face], face,
-                             n, stencil, passmask);
+                             n, stencil, passMask);
          }
       }
 
@@ -972,28 +964,21 @@ stencil_and_ztest_pixels( GLcontext *ctx, SWspan *span, GLuint face )
                                     ctx->Stencil.ZPassFunc[face], face, mask);
       }
       else {
-         GLubyte passmask[MAX_WIDTH], failmask[MAX_WIDTH], oldmask[MAX_WIDTH];
-         GLuint i;
-
-         _mesa_memcpy(oldmask, mask, n * sizeof(GLubyte));
+         _mesa_memcpy(origMask, mask, n * sizeof(GLubyte));
 
          _swrast_depth_test_span(ctx, span);
 
-         for (i=0;i<n;i++) {
-            ASSERT(mask[i] == 0 || mask[i] == 1);
-            passmask[i] = oldmask[i] & mask[i];
-            failmask[i] = oldmask[i] & (mask[i] ^ 1);
-         }
+         compute_pass_fail_masks(n, origMask, mask, passMask, failMask);
 
          if (ctx->Stencil.ZFailFunc[face] != GL_KEEP) {
             apply_stencil_op_to_pixels(ctx, n, x, y,
                                        ctx->Stencil.ZFailFunc[face],
-                                       face, failmask);
+                                       face, failMask);
          }
          if (ctx->Stencil.ZPassFunc[face] != GL_KEEP) {
             apply_stencil_op_to_pixels(ctx, n, x, y,
                                        ctx->Stencil.ZPassFunc[face],
-                                       face, passmask);
+                                       face, passMask);
          }
       }
 
diff --git a/src/mesa/swrast_setup/ss_context.c b/src/mesa/swrast_setup/ss_context.c
index f8a1cadfa5a..a9c7d941e5b 100644
--- a/src/mesa/swrast_setup/ss_context.c
+++ b/src/mesa/swrast_setup/ss_context.c
@@ -280,26 +280,29 @@ _swsetup_Translate( GLcontext *ctx, const void *vertex, SWvertex *dest )
 
    /** XXX try to limit these loops someday */
    for (i = 0 ; i < ctx->Const.MaxTextureCoordUnits ; i++)
-      _tnl_get_attr( ctx, vertex, _TNL_ATTRIB_TEX0+i,
+      _tnl_get_attr( ctx, vertex, _TNL_ATTRIB_TEX0 + i,
                      dest->attrib[FRAG_ATTRIB_TEX0 + i] );
 
    for (i = 0 ; i < ctx->Const.MaxVarying ; i++)
-      _tnl_get_attr( ctx, vertex, _TNL_ATTRIB_GENERIC0+i,
+      _tnl_get_attr( ctx, vertex, _TNL_ATTRIB_GENERIC0 + i,
                      dest->attrib[FRAG_ATTRIB_VAR0 + i] );
 
-   _tnl_get_attr( ctx, vertex, _TNL_ATTRIB_COLOR0, tmp );
-   /* XXX need float color FRAG_ATTRIB_COL0?? */
-   UNCLAMPED_FLOAT_TO_RGBA_CHAN( dest->color, tmp );
+   if (ctx->Visual.rgbMode) {
+      _tnl_get_attr( ctx, vertex, _TNL_ATTRIB_COLOR0,
+                     dest->attrib[FRAG_ATTRIB_COL0] );
+      UNCLAMPED_FLOAT_TO_RGBA_CHAN( dest->color, tmp );
 
-   _tnl_get_attr( ctx, vertex, _TNL_ATTRIB_COLOR1, tmp );
-   COPY_4V(dest->attrib[FRAG_ATTRIB_COL1], tmp);
+      _tnl_get_attr( ctx, vertex, _TNL_ATTRIB_COLOR1,
+                     dest->attrib[FRAG_ATTRIB_COL1]);
+   }
+   else {
+      _tnl_get_attr( ctx, vertex, _TNL_ATTRIB_COLOR_INDEX, tmp );
+      dest->attrib[FRAG_ATTRIB_CI][0] = tmp[0];
+   }
 
    _tnl_get_attr( ctx, vertex, _TNL_ATTRIB_FOG, tmp );
    dest->attrib[FRAG_ATTRIB_FOGC][0] = tmp[0];
 
-   _tnl_get_attr( ctx, vertex, _TNL_ATTRIB_COLOR_INDEX, tmp );
-   dest->attrib[FRAG_ATTRIB_CI][0] = tmp[0];
-
    /* XXX See _tnl_get_attr about pointsize ... */
    _tnl_get_attr( ctx, vertex, _TNL_ATTRIB_POINTSIZE, tmp );
    dest->pointSize = tmp[0];
diff --git a/src/mesa/swrast_setup/ss_tritmp.h b/src/mesa/swrast_setup/ss_tritmp.h
index c14468e9514..f6b738d60df 100644
--- a/src/mesa/swrast_setup/ss_tritmp.h
+++ b/src/mesa/swrast_setup/ss_tritmp.h
@@ -37,7 +37,7 @@ static void TAG(triangle)(GLcontext *ctx, GLuint e0, GLuint e1, GLuint e2 )
    SWvertex *verts = SWSETUP_CONTEXT(ctx)->verts;
    SWvertex *v[3];
    GLfloat z[3];
-   GLfloat offset;
+   GLfloat offset, oz0, oz1, oz2;
    GLenum mode = GL_FILL;
    GLuint facing = 0;
    GLchan saved_color[3][4];
@@ -142,12 +142,16 @@ static void TAG(triangle)(GLcontext *ctx, GLuint e0, GLuint e1, GLuint e2 )
 	 }
       }
 
-      if (IND & SS_OFFSET_BIT)
-      {
-	 offset = ctx->Polygon.OffsetUnits * ctx->DrawBuffer->_MRD;
+      if (IND & SS_OFFSET_BIT) {
+         const GLfloat max = ctx->DrawBuffer->_DepthMaxF;
+         /* save original Z values (restored later) */
 	 z[0] = v[0]->attrib[FRAG_ATTRIB_WPOS][2];
 	 z[1] = v[1]->attrib[FRAG_ATTRIB_WPOS][2];
 	 z[2] = v[2]->attrib[FRAG_ATTRIB_WPOS][2];
+         /* Note that Z values are already scaled to [0,65535] (for example)
+          * so no MRD value is used here.
+          */
+	 offset = ctx->Polygon.OffsetUnits;
 	 if (cc * cc > 1e-16) {
 	    const GLfloat ez = z[0] - z[2];
 	    const GLfloat fz = z[1] - z[2];
@@ -155,35 +159,33 @@ static void TAG(triangle)(GLcontext *ctx, GLuint e0, GLuint e1, GLuint e2 )
 	    const GLfloat dzdx = FABSF((ey * fz - ez * fy) * oneOverArea);
 	    const GLfloat dzdy = FABSF((ez * fx - ex * fz) * oneOverArea);
 	    offset += MAX2(dzdx, dzdy) * ctx->Polygon.OffsetFactor;
-            /* Unfortunately, we need to clamp to prevent negative Zs below.
-             * Technically, we should do the clamping per-fragment.
-             */
-            offset = MAX2(offset, -v[0]->attrib[FRAG_ATTRIB_WPOS][2]);
-            offset = MAX2(offset, -v[1]->attrib[FRAG_ATTRIB_WPOS][2]);
-            offset = MAX2(offset, -v[2]->attrib[FRAG_ATTRIB_WPOS][2]);
 	 }
+         /* new Z values */
+         oz0 = CLAMP(v[0]->attrib[FRAG_ATTRIB_WPOS][2] + offset, 0.0, max);
+         oz1 = CLAMP(v[1]->attrib[FRAG_ATTRIB_WPOS][2] + offset, 0.0, max);
+         oz2 = CLAMP(v[2]->attrib[FRAG_ATTRIB_WPOS][2] + offset, 0.0, max);
       }
    }
 
    if (mode == GL_POINT) {
       if ((IND & SS_OFFSET_BIT) && ctx->Polygon.OffsetPoint) {
-	 v[0]->attrib[FRAG_ATTRIB_WPOS][2] += offset;
-	 v[1]->attrib[FRAG_ATTRIB_WPOS][2] += offset;
-	 v[2]->attrib[FRAG_ATTRIB_WPOS][2] += offset;
+	 v[0]->attrib[FRAG_ATTRIB_WPOS][2] = oz0;
+	 v[1]->attrib[FRAG_ATTRIB_WPOS][2] = oz1;
+	 v[2]->attrib[FRAG_ATTRIB_WPOS][2] = oz2;
       }
       _swsetup_render_point_tri( ctx, e0, e1, e2, facing );
    } else if (mode == GL_LINE) {
       if ((IND & SS_OFFSET_BIT) && ctx->Polygon.OffsetLine) {
-	 v[0]->attrib[FRAG_ATTRIB_WPOS][2] += offset;
-	 v[1]->attrib[FRAG_ATTRIB_WPOS][2] += offset;
-	 v[2]->attrib[FRAG_ATTRIB_WPOS][2] += offset;
+	 v[0]->attrib[FRAG_ATTRIB_WPOS][2] = oz0;
+	 v[1]->attrib[FRAG_ATTRIB_WPOS][2] = oz1;
+	 v[2]->attrib[FRAG_ATTRIB_WPOS][2] = oz2;
       }
       _swsetup_render_line_tri( ctx, e0, e1, e2, facing );
    } else {
       if ((IND & SS_OFFSET_BIT) && ctx->Polygon.OffsetFill) {
-	 v[0]->attrib[FRAG_ATTRIB_WPOS][2] += offset;
-	 v[1]->attrib[FRAG_ATTRIB_WPOS][2] += offset;
-	 v[2]->attrib[FRAG_ATTRIB_WPOS][2] += offset;
+	 v[0]->attrib[FRAG_ATTRIB_WPOS][2] = oz0;
+	 v[1]->attrib[FRAG_ATTRIB_WPOS][2] = oz1;
+	 v[2]->attrib[FRAG_ATTRIB_WPOS][2] = oz2;
       }
       _swrast_Triangle( ctx, v[0], v[1], v[2] );
    }