diff options
Diffstat (limited to 'src/mesa/drivers/dri')
346 files changed, 15754 insertions, 27541 deletions
diff --git a/src/mesa/drivers/dri/Makefile b/src/mesa/drivers/dri/Makefile index 32db0978613..264648c3fbc 100644 --- a/src/mesa/drivers/dri/Makefile +++ b/src/mesa/drivers/dri/Makefile @@ -25,7 +25,7 @@ pcedit = sed \ -e 's,@INSTALL_LIB_DIR@,$(INSTALL_LIB_DIR),' \ -e 's,@INSTALL_INC_DIR@,$(INSTALL_INC_DIR),' \ -e 's,@VERSION@,$(MESA_MAJOR).$(MESA_MINOR).$(MESA_TINY),' \ - -e 's,@DRI_DRIVER_DIR@,$(DRI_DRIVER_SEARCH_DIR),' \ + -e 's,@DRI_DRIVER_DIR@,$(DRI_DRIVER_INSTALL_DIR),' \ -e 's,@DRI_PC_REQ_PRIV@,$(DRI_PC_REQ_PRIV),' dri.pc: dri.pc.in diff --git a/src/mesa/drivers/dri/Makefile.template b/src/mesa/drivers/dri/Makefile.template index 18dbeba24a8..39d25ce3f47 100644 --- a/src/mesa/drivers/dri/Makefile.template +++ b/src/mesa/drivers/dri/Makefile.template @@ -60,18 +60,40 @@ SHARED_INCLUDES = \ ##### TARGETS ##### -default: symlinks depend $(LIBNAME) $(TOP)/$(LIB_DIR)/$(LIBNAME) +default: subdirs lib -$(LIBNAME): $(OBJECTS) $(MESA_MODULES) $(PIPE_DRIVERS) $(WINOBJ) Makefile $(TOP)/src/mesa/drivers/dri/Makefile.template +.PHONY: lib +lib: symlinks subdirs depend + @$(MAKE) $(LIBNAME) $(TOP)/$(LIB_DIR)/$(LIBNAME) + +$(LIBNAME): $(OBJECTS) $(MESA_MODULES) $(EXTRA_MODULES) $(WINOBJ) Makefile \ + $(TOP)/src/mesa/drivers/dri/Makefile.template $(MKLIB) -o $@ -noprefix -linker '$(CC)' -ldflags '$(LDFLAGS)' \ - $(OBJECTS) $(PIPE_DRIVERS) $(MESA_MODULES) $(WINOBJ) $(DRI_LIB_DEPS) + $(OBJECTS) $(MESA_MODULES) $(EXTRA_MODULES) $(WINOBJ) \ + $(DRI_LIB_DEPS) $(TOP)/$(LIB_DIR)/$(LIBNAME): $(LIBNAME) $(INSTALL) $(LIBNAME) $(TOP)/$(LIB_DIR) +# If the Makefile defined SUBDIRS, run make in each +.PHONY: subdirs +subdirs: + @if test -n "$(SUBDIRS)" ; then \ + for dir in $(SUBDIRS) ; do \ + if [ -d $$dir ] ; then \ + (cd $$dir && $(MAKE)) || exit 1; \ + fi \ + done \ + fi + + +.PHONY: symlinks +symlinks: + + depend: $(C_SOURCES) $(ASM_SOURCES) $(SYMLINKS) @ echo "running $(MKDEP)" @ rm -f depend diff --git a/src/mesa/drivers/dri/common/dri_util.c b/src/mesa/drivers/dri/common/dri_util.c index e48e10d7c06..da81ec9de51 100644 --- a/src/mesa/drivers/dri/common/dri_util.c +++ b/src/mesa/drivers/dri/common/dri_util.c @@ -167,11 +167,12 @@ static int driBindContext(__DRIcontext *pcp, __DRIdrawable *pdp, __DRIdrawable *prp) { - __DRIscreenPrivate *psp = pcp->driScreenPriv; + __DRIscreenPrivate *psp; /* Bind the drawable to the context */ if (pcp) { + psp = pcp->driScreenPriv; pcp->driDrawablePriv = pdp; pcp->driReadablePriv = prp; if (pdp) { @@ -498,11 +499,11 @@ static void dri_put_drawable(__DRIdrawable *pdp) { __DRIscreenPrivate *psp; - pdp->refcount--; - if (pdp->refcount) - return; - if (pdp) { + pdp->refcount--; + if (pdp->refcount) + return; + psp = pdp->driScreenPriv; (*psp->DriverAPI.DestroyBuffer)(pdp); if (pdp->pClipRects) { diff --git a/src/mesa/drivers/dri/common/drirenderbuffer.c b/src/mesa/drivers/dri/common/drirenderbuffer.c index 15af99136cd..4e7e92c82bc 100644 --- a/src/mesa/drivers/dri/common/drirenderbuffer.c +++ b/src/mesa/drivers/dri/common/drirenderbuffer.c @@ -1,5 +1,6 @@ #include "main/mtypes.h" +#include "main/formats.h" #include "main/framebuffer.h" #include "main/renderbuffer.h" #include "main/imports.h" @@ -53,7 +54,7 @@ driDeleteRenderbuffer(struct gl_renderbuffer *rb) * \param pitch pixels per row */ driRenderbuffer * -driNewRenderbuffer(GLenum format, GLvoid *addr, +driNewRenderbuffer(gl_format format, GLvoid *addr, GLint cpp, GLint offset, GLint pitch, __DRIdrawablePrivate *dPriv) { @@ -79,56 +80,47 @@ driNewRenderbuffer(GLenum format, GLvoid *addr, /* Make sure we're using a null-valued GetPointer routine */ assert(drb->Base.GetPointer(NULL, &drb->Base, 0, 0) == NULL); - drb->Base.InternalFormat = format; - - if (format == GL_RGBA || format == GL_RGB5 || format == GL_RGBA8) { - /* Color */ - drb->Base._BaseFormat = GL_RGBA; - drb->Base.DataType = GL_UNSIGNED_BYTE; - if (format == GL_RGB5) { - drb->Base.RedBits = 5; - drb->Base.GreenBits = 6; - drb->Base.BlueBits = 5; + switch (format) { + case MESA_FORMAT_ARGB8888: + if (cpp == 2) { + /* override format */ + format = MESA_FORMAT_RGB565; } - else { - drb->Base.RedBits = - drb->Base.GreenBits = - drb->Base.BlueBits = - drb->Base.AlphaBits = 8; - } - } - else if (format == GL_DEPTH_COMPONENT16) { - /* Depth */ - drb->Base._BaseFormat = GL_DEPTH_COMPONENT; - /* we always Get/Put 32-bit Z values */ - drb->Base.DataType = GL_UNSIGNED_INT; - drb->Base.DepthBits = 16; - } - else if (format == GL_DEPTH_COMPONENT24) { + drb->Base.DataType = GL_UNSIGNED_BYTE; + break; + case MESA_FORMAT_Z16: /* Depth */ - drb->Base._BaseFormat = GL_DEPTH_COMPONENT; /* we always Get/Put 32-bit Z values */ drb->Base.DataType = GL_UNSIGNED_INT; - drb->Base.DepthBits = 24; - } - else if (format == GL_DEPTH_COMPONENT32) { + assert(cpp == 2); + break; + case MESA_FORMAT_Z32: /* Depth */ - drb->Base._BaseFormat = GL_DEPTH_COMPONENT; /* we always Get/Put 32-bit Z values */ drb->Base.DataType = GL_UNSIGNED_INT; - drb->Base.DepthBits = 32; - } - else { + assert(cpp == 4); + break; + case MESA_FORMAT_Z24_S8: + drb->Base.DataType = GL_UNSIGNED_INT_24_8_EXT; + assert(cpp == 4); + break; + case MESA_FORMAT_S8_Z24: + drb->Base.DataType = GL_UNSIGNED_INT_24_8_EXT; + assert(cpp == 4); + break; + case MESA_FORMAT_S8: /* Stencil */ - ASSERT(format == GL_STENCIL_INDEX8_EXT); - drb->Base._BaseFormat = GL_STENCIL_INDEX; drb->Base.DataType = GL_UNSIGNED_BYTE; - drb->Base.StencilBits = 8; + break; + default: + _mesa_problem(NULL, "Bad format 0x%x in driNewRenderbuffer", format); + return NULL; } - /* XXX if we were allocating a user-created renderbuffer, we'd have - * to fill in the Red/Green/Blue/.../Bits values too. - */ + drb->Base.Format = format; + + drb->Base.InternalFormat = + drb->Base._BaseFormat = _mesa_get_format_base_format(format); drb->Base.AllocStorage = driRenderbufferStorage; drb->Base.Delete = driDeleteRenderbuffer; diff --git a/src/mesa/drivers/dri/common/drirenderbuffer.h b/src/mesa/drivers/dri/common/drirenderbuffer.h index cf55286b30f..3a5cbcdaac1 100644 --- a/src/mesa/drivers/dri/common/drirenderbuffer.h +++ b/src/mesa/drivers/dri/common/drirenderbuffer.h @@ -11,6 +11,7 @@ #define DRIRENDERBUFFER_H #include "main/mtypes.h" +#include "main/formats.h" #include "dri_util.h" @@ -56,14 +57,14 @@ typedef struct { * A handy flag to know if this is the back color buffer. * * \note - * This is currently only used by s3v and tdfx. + * This is currently only used by tdfx. */ GLboolean backBuffer; } driRenderbuffer; extern driRenderbuffer * -driNewRenderbuffer(GLenum format, GLvoid *addr, +driNewRenderbuffer(gl_format format, GLvoid *addr, GLint cpp, GLint offset, GLint pitch, __DRIdrawablePrivate *dPriv); diff --git a/src/mesa/drivers/dri/common/extension_helper.h b/src/mesa/drivers/dri/common/extension_helper.h deleted file mode 100644 index 2c89a9f58a9..00000000000 --- a/src/mesa/drivers/dri/common/extension_helper.h +++ /dev/null @@ -1,6639 +0,0 @@ -/* DO NOT EDIT - This file generated automatically by extension_helper.py (from Mesa) script */ - -/* - * (C) Copyright IBM Corporation 2005 - * 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 - * IBM, - * AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, - * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF - * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE - * SOFTWARE. - */ - -#include "utils.h" -#include "glapi/dispatch.h" - -#ifndef NULL -# define NULL 0 -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char UniformMatrix3fvARB_names[] = - "iiip\0" /* Parameter signature */ - "glUniformMatrix3fv\0" - "glUniformMatrix3fvARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) || defined(need_GL_ARB_multisample) -static const char SampleCoverageARB_names[] = - "fi\0" /* Parameter signature */ - "glSampleCoverage\0" - "glSampleCoverageARB\0" - ""; -#endif - -#if defined(need_GL_EXT_convolution) -static const char ConvolutionFilter1D_names[] = - "iiiiip\0" /* Parameter signature */ - "glConvolutionFilter1D\0" - "glConvolutionFilter1DEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_occlusion_query) -static const char BeginQueryARB_names[] = - "ii\0" /* Parameter signature */ - "glBeginQuery\0" - "glBeginQueryARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_NV_point_sprite) -static const char PointParameteriNV_names[] = - "ii\0" /* Parameter signature */ - "glPointParameteri\0" - "glPointParameteriNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) -static const char GetProgramiv_names[] = - "iip\0" /* Parameter signature */ - "glGetProgramiv\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord3sARB_names[] = - "iiii\0" /* Parameter signature */ - "glMultiTexCoord3s\0" - "glMultiTexCoord3sARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color) -static const char SecondaryColor3iEXT_names[] = - "iii\0" /* Parameter signature */ - "glSecondaryColor3i\0" - "glSecondaryColor3iEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos) -static const char WindowPos3fMESA_names[] = - "fff\0" /* Parameter signature */ - "glWindowPos3f\0" - "glWindowPos3fARB\0" - "glWindowPos3fMESA\0" - ""; -#endif - -#if defined(need_GL_SGIS_pixel_texture) -static const char PixelTexGenParameterfvSGIS_names[] = - "ip\0" /* Parameter signature */ - "glPixelTexGenParameterfvSGIS\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char ActiveTextureARB_names[] = - "i\0" /* Parameter signature */ - "glActiveTexture\0" - "glActiveTextureARB\0" - ""; -#endif - -#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_blit) -static const char BlitFramebufferEXT_names[] = - "iiiiiiiiii\0" /* Parameter signature */ - "glBlitFramebuffer\0" - "glBlitFramebufferEXT\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttrib4ubvNV_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib4ubvNV\0" - ""; -#endif - -#if defined(need_GL_NV_fragment_program) -static const char GetProgramNamedParameterdvNV_names[] = - "iipp\0" /* Parameter signature */ - "glGetProgramNamedParameterdvNV\0" - ""; -#endif - -#if defined(need_GL_EXT_histogram) -static const char Histogram_names[] = - "iiii\0" /* Parameter signature */ - "glHistogram\0" - "glHistogramEXT\0" - ""; -#endif - -#if defined(need_GL_SGIS_texture4D) -static const char TexImage4DSGIS_names[] = - "iiiiiiiiiip\0" /* Parameter signature */ - "glTexImage4DSGIS\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos) -static const char WindowPos2dvMESA_names[] = - "p\0" /* Parameter signature */ - "glWindowPos2dv\0" - "glWindowPos2dvARB\0" - "glWindowPos2dvMESA\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char ReplacementCodeuiColor3fVertex3fvSUN_names[] = - "ppp\0" /* Parameter signature */ - "glReplacementCodeuiColor3fVertex3fvSUN\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_EXT_blend_equation_separate) || defined(need_GL_ATI_blend_equation_separate) -static const char BlendEquationSeparateEXT_names[] = - "ii\0" /* Parameter signature */ - "glBlendEquationSeparate\0" - "glBlendEquationSeparateEXT\0" - "glBlendEquationSeparateATI\0" - ""; -#endif - -#if defined(need_GL_SGIX_list_priority) -static const char ListParameterfSGIX_names[] = - "iif\0" /* Parameter signature */ - "glListParameterfSGIX\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color) -static const char SecondaryColor3bEXT_names[] = - "iii\0" /* Parameter signature */ - "glSecondaryColor3b\0" - "glSecondaryColor3bEXT\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char TexCoord4fColor4fNormal3fVertex4fvSUN_names[] = - "pppp\0" /* Parameter signature */ - "glTexCoord4fColor4fNormal3fVertex4fvSUN\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttrib4svNV_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib4svNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_vertex_buffer_object) -static const char GetBufferSubDataARB_names[] = - "iiip\0" /* Parameter signature */ - "glGetBufferSubData\0" - "glGetBufferSubDataARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_vertex_buffer_object) -static const char BufferSubDataARB_names[] = - "iiip\0" /* Parameter signature */ - "glBufferSubData\0" - "glBufferSubDataARB\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char TexCoord2fColor4ubVertex3fvSUN_names[] = - "ppp\0" /* Parameter signature */ - "glTexCoord2fColor4ubVertex3fvSUN\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) -static const char AttachShader_names[] = - "ii\0" /* Parameter signature */ - "glAttachShader\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib2fARB_names[] = - "iff\0" /* Parameter signature */ - "glVertexAttrib2f\0" - "glVertexAttrib2fARB\0" - ""; -#endif - -#if defined(need_GL_MESA_shader_debug) -static const char GetDebugLogLengthMESA_names[] = - "iii\0" /* Parameter signature */ - "glGetDebugLogLengthMESA\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib3fARB_names[] = - "ifff\0" /* Parameter signature */ - "glVertexAttrib3f\0" - "glVertexAttrib3fARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_occlusion_query) -static const char GetQueryivARB_names[] = - "iip\0" /* Parameter signature */ - "glGetQueryiv\0" - "glGetQueryivARB\0" - ""; -#endif - -#if defined(need_GL_EXT_texture3D) -static const char TexImage3D_names[] = - "iiiiiiiiip\0" /* Parameter signature */ - "glTexImage3D\0" - "glTexImage3DEXT\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char ReplacementCodeuiVertex3fvSUN_names[] = - "pp\0" /* Parameter signature */ - "glReplacementCodeuiVertex3fvSUN\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_occlusion_query) -static const char GetQueryObjectivARB_names[] = - "iip\0" /* Parameter signature */ - "glGetQueryObjectiv\0" - "glGetQueryObjectivARB\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char ReplacementCodeuiTexCoord2fVertex3fvSUN_names[] = - "ppp\0" /* Parameter signature */ - "glReplacementCodeuiTexCoord2fVertex3fvSUN\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) || defined(need_GL_ARB_texture_compression) -static const char CompressedTexSubImage2DARB_names[] = - "iiiiiiiip\0" /* Parameter signature */ - "glCompressedTexSubImage2D\0" - "glCompressedTexSubImage2DARB\0" - ""; -#endif - -#if defined(need_GL_NV_register_combiners) -static const char CombinerOutputNV_names[] = - "iiiiiiiiii\0" /* Parameter signature */ - "glCombinerOutputNV\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttribs3fvNV_names[] = - "iip\0" /* Parameter signature */ - "glVertexAttribs3fvNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char Uniform2fARB_names[] = - "iff\0" /* Parameter signature */ - "glUniform2f\0" - "glUniform2fARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib1svARB_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib1sv\0" - "glVertexAttrib1svARB\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttribs1dvNV_names[] = - "iip\0" /* Parameter signature */ - "glVertexAttribs1dvNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char Uniform2ivARB_names[] = - "iip\0" /* Parameter signature */ - "glUniform2iv\0" - "glUniform2ivARB\0" - ""; -#endif - -#if defined(need_GL_HP_image_transform) -static const char GetImageTransformParameterfvHP_names[] = - "iip\0" /* Parameter signature */ - "glGetImageTransformParameterfvHP\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_blend) -static const char WeightubvARB_names[] = - "ip\0" /* Parameter signature */ - "glWeightubvARB\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttrib1fvNV_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib1fvNV\0" - ""; -#endif - -#if defined(need_GL_EXT_convolution) -static const char CopyConvolutionFilter1D_names[] = - "iiiii\0" /* Parameter signature */ - "glCopyConvolutionFilter1D\0" - "glCopyConvolutionFilter1DEXT\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char ReplacementCodeuiNormal3fVertex3fSUN_names[] = - "iffffff\0" /* Parameter signature */ - "glReplacementCodeuiNormal3fVertex3fSUN\0" - ""; -#endif - -#if defined(need_GL_ARB_sync) -static const char DeleteSync_names[] = - "i\0" /* Parameter signature */ - "glDeleteSync\0" - ""; -#endif - -#if defined(need_GL_SGIX_fragment_lighting) -static const char FragmentMaterialfvSGIX_names[] = - "iip\0" /* Parameter signature */ - "glFragmentMaterialfvSGIX\0" - ""; -#endif - -#if defined(need_GL_EXT_blend_color) -static const char BlendColor_names[] = - "ffff\0" /* Parameter signature */ - "glBlendColor\0" - "glBlendColorEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char UniformMatrix4fvARB_names[] = - "iiip\0" /* Parameter signature */ - "glUniformMatrix4fv\0" - "glUniformMatrix4fvARB\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_array_object) || defined(need_GL_APPLE_vertex_array_object) -static const char DeleteVertexArraysAPPLE_names[] = - "ip\0" /* Parameter signature */ - "glDeleteVertexArrays\0" - "glDeleteVertexArraysAPPLE\0" - ""; -#endif - -#if defined(need_GL_SGIX_instruments) -static const char ReadInstrumentsSGIX_names[] = - "i\0" /* Parameter signature */ - "glReadInstrumentsSGIX\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_1) -static const char UniformMatrix2x4fv_names[] = - "iiip\0" /* Parameter signature */ - "glUniformMatrix2x4fv\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char Color4ubVertex3fvSUN_names[] = - "pp\0" /* Parameter signature */ - "glColor4ubVertex3fvSUN\0" - ""; -#endif - -#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_texture_array) -static const char FramebufferTextureLayerEXT_names[] = - "iiiii\0" /* Parameter signature */ - "glFramebufferTextureLayer\0" - "glFramebufferTextureLayerEXT\0" - ""; -#endif - -#if defined(need_GL_SGIX_list_priority) -static const char GetListParameterfvSGIX_names[] = - "iip\0" /* Parameter signature */ - "glGetListParameterfvSGIX\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib4NusvARB_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib4Nusv\0" - "glVertexAttrib4NusvARB\0" - ""; -#endif - -#if defined(need_GL_MESA_window_pos) -static const char WindowPos4svMESA_names[] = - "p\0" /* Parameter signature */ - "glWindowPos4svMESA\0" - ""; -#endif - -#if defined(need_GL_ARB_shader_objects) -static const char CreateProgramObjectARB_names[] = - "\0" /* Parameter signature */ - "glCreateProgramObjectARB\0" - ""; -#endif - -#if defined(need_GL_SGIX_fragment_lighting) -static const char FragmentLightModelivSGIX_names[] = - "ip\0" /* Parameter signature */ - "glFragmentLightModelivSGIX\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_1) -static const char UniformMatrix4x3fv_names[] = - "iiip\0" /* Parameter signature */ - "glUniformMatrix4x3fv\0" - ""; -#endif - -#if defined(need_GL_EXT_texture_object) -static const char PrioritizeTextures_names[] = - "ipp\0" /* Parameter signature */ - "glPrioritizeTextures\0" - "glPrioritizeTexturesEXT\0" - ""; -#endif - -#if defined(need_GL_SGIX_async) -static const char AsyncMarkerSGIX_names[] = - "i\0" /* Parameter signature */ - "glAsyncMarkerSGIX\0" - ""; -#endif - -#if defined(need_GL_SUN_global_alpha) -static const char GlobalAlphaFactorubSUN_names[] = - "i\0" /* Parameter signature */ - "glGlobalAlphaFactorubSUN\0" - ""; -#endif - -#if defined(need_GL_MESA_shader_debug) -static const char ClearDebugLogMESA_names[] = - "iii\0" /* Parameter signature */ - "glClearDebugLogMESA\0" - ""; -#endif - -#if defined(need_GL_EXT_histogram) -static const char ResetHistogram_names[] = - "i\0" /* Parameter signature */ - "glResetHistogram\0" - "glResetHistogramEXT\0" - ""; -#endif - -#if defined(need_GL_NV_fragment_program) -static const char GetProgramNamedParameterfvNV_names[] = - "iipp\0" /* Parameter signature */ - "glGetProgramNamedParameterfvNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_point_parameters) || defined(need_GL_EXT_point_parameters) || defined(need_GL_SGIS_point_parameters) -static const char PointParameterfEXT_names[] = - "if\0" /* Parameter signature */ - "glPointParameterf\0" - "glPointParameterfARB\0" - "glPointParameterfEXT\0" - "glPointParameterfSGIS\0" - ""; -#endif - -#if defined(need_GL_SGIX_polynomial_ffd) -static const char LoadIdentityDeformationMapSGIX_names[] = - "i\0" /* Parameter signature */ - "glLoadIdentityDeformationMapSGIX\0" - ""; -#endif - -#if defined(need_GL_NV_fence) -static const char GenFencesNV_names[] = - "ip\0" /* Parameter signature */ - "glGenFencesNV\0" - ""; -#endif - -#if defined(need_GL_HP_image_transform) -static const char ImageTransformParameterfHP_names[] = - "iif\0" /* Parameter signature */ - "glImageTransformParameterfHP\0" - ""; -#endif - -#if defined(need_GL_ARB_matrix_palette) -static const char MatrixIndexusvARB_names[] = - "ip\0" /* Parameter signature */ - "glMatrixIndexusvARB\0" - ""; -#endif - -#if defined(need_GL_ARB_draw_elements_base_vertex) -static const char DrawElementsBaseVertex_names[] = - "iiipi\0" /* Parameter signature */ - "glDrawElementsBaseVertex\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char DisableVertexAttribArrayARB_names[] = - "i\0" /* Parameter signature */ - "glDisableVertexAttribArray\0" - "glDisableVertexAttribArrayARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) -static const char StencilMaskSeparate_names[] = - "ii\0" /* Parameter signature */ - "glStencilMaskSeparate\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_program) -static const char ProgramLocalParameter4dARB_names[] = - "iidddd\0" /* Parameter signature */ - "glProgramLocalParameter4dARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) || defined(need_GL_ARB_texture_compression) -static const char CompressedTexImage3DARB_names[] = - "iiiiiiiip\0" /* Parameter signature */ - "glCompressedTexImage3D\0" - "glCompressedTexImage3DARB\0" - ""; -#endif - -#if defined(need_GL_EXT_convolution) -static const char GetConvolutionParameteriv_names[] = - "iip\0" /* Parameter signature */ - "glGetConvolutionParameteriv\0" - "glGetConvolutionParameterivEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib1fARB_names[] = - "if\0" /* Parameter signature */ - "glVertexAttrib1f\0" - "glVertexAttrib1fARB\0" - ""; -#endif - -#if defined(need_GL_NV_fence) -static const char TestFenceNV_names[] = - "i\0" /* Parameter signature */ - "glTestFenceNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord1fvARB_names[] = - "ip\0" /* Parameter signature */ - "glMultiTexCoord1fv\0" - "glMultiTexCoord1fvARB\0" - ""; -#endif - -#if defined(need_GL_ATI_fragment_shader) -static const char ColorFragmentOp2ATI_names[] = - "iiiiiiiiii\0" /* Parameter signature */ - "glColorFragmentOp2ATI\0" - ""; -#endif - -#if defined(need_GL_IBM_vertex_array_lists) -static const char SecondaryColorPointerListIBM_names[] = - "iiipi\0" /* Parameter signature */ - "glSecondaryColorPointerListIBM\0" - ""; -#endif - -#if defined(need_GL_SGIS_pixel_texture) -static const char GetPixelTexGenParameterivSGIS_names[] = - "ip\0" /* Parameter signature */ - "glGetPixelTexGenParameterivSGIS\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttrib4fNV_names[] = - "iffff\0" /* Parameter signature */ - "glVertexAttrib4fNV\0" - ""; -#endif - -#if defined(need_GL_SUN_triangle_list) -static const char ReplacementCodeubSUN_names[] = - "i\0" /* Parameter signature */ - "glReplacementCodeubSUN\0" - ""; -#endif - -#if defined(need_GL_SGIX_async) -static const char FinishAsyncSGIX_names[] = - "p\0" /* Parameter signature */ - "glFinishAsyncSGIX\0" - ""; -#endif - -#if defined(need_GL_MESA_shader_debug) -static const char GetDebugLogMESA_names[] = - "iiiipp\0" /* Parameter signature */ - "glGetDebugLogMESA\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_fog_coord) -static const char FogCoorddEXT_names[] = - "d\0" /* Parameter signature */ - "glFogCoordd\0" - "glFogCoorddEXT\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char Color4ubVertex3fSUN_names[] = - "iiiifff\0" /* Parameter signature */ - "glColor4ubVertex3fSUN\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_fog_coord) -static const char FogCoordfEXT_names[] = - "f\0" /* Parameter signature */ - "glFogCoordf\0" - "glFogCoordfEXT\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char TexCoord2fVertex3fSUN_names[] = - "fffff\0" /* Parameter signature */ - "glTexCoord2fVertex3fSUN\0" - ""; -#endif - -#if defined(need_GL_SUN_global_alpha) -static const char GlobalAlphaFactoriSUN_names[] = - "i\0" /* Parameter signature */ - "glGlobalAlphaFactoriSUN\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttrib2dNV_names[] = - "idd\0" /* Parameter signature */ - "glVertexAttrib2dNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) -static const char GetProgramInfoLog_names[] = - "iipp\0" /* Parameter signature */ - "glGetProgramInfoLog\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib4NbvARB_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib4Nbv\0" - "glVertexAttrib4NbvARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_shader) -static const char GetActiveAttribARB_names[] = - "iiipppp\0" /* Parameter signature */ - "glGetActiveAttrib\0" - "glGetActiveAttribARB\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttrib4ubNV_names[] = - "iiiii\0" /* Parameter signature */ - "glVertexAttrib4ubNV\0" - ""; -#endif - -#if defined(need_GL_APPLE_texture_range) -static const char TextureRangeAPPLE_names[] = - "iip\0" /* Parameter signature */ - "glTextureRangeAPPLE\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char TexCoord2fColor4fNormal3fVertex3fSUN_names[] = - "ffffffffffff\0" /* Parameter signature */ - "glTexCoord2fColor4fNormal3fVertex3fSUN\0" - ""; -#endif - -#if defined(need_GL_NV_register_combiners) -static const char CombinerParameterfvNV_names[] = - "ip\0" /* Parameter signature */ - "glCombinerParameterfvNV\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttribs3dvNV_names[] = - "iip\0" /* Parameter signature */ - "glVertexAttribs3dvNV\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttribs4fvNV_names[] = - "iip\0" /* Parameter signature */ - "glVertexAttribs4fvNV\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_array_range) -static const char VertexArrayRangeNV_names[] = - "ip\0" /* Parameter signature */ - "glVertexArrayRangeNV\0" - ""; -#endif - -#if defined(need_GL_SGIX_fragment_lighting) -static const char FragmentLightiSGIX_names[] = - "iii\0" /* Parameter signature */ - "glFragmentLightiSGIX\0" - ""; -#endif - -#if defined(need_GL_EXT_polygon_offset) -static const char PolygonOffsetEXT_names[] = - "ff\0" /* Parameter signature */ - "glPolygonOffsetEXT\0" - ""; -#endif - -#if defined(need_GL_SGIX_async) -static const char PollAsyncSGIX_names[] = - "p\0" /* Parameter signature */ - "glPollAsyncSGIX\0" - ""; -#endif - -#if defined(need_GL_ATI_fragment_shader) -static const char DeleteFragmentShaderATI_names[] = - "i\0" /* Parameter signature */ - "glDeleteFragmentShaderATI\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char TexCoord2fNormal3fVertex3fvSUN_names[] = - "ppp\0" /* Parameter signature */ - "glTexCoord2fNormal3fVertex3fvSUN\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) || defined(need_GL_ARB_transpose_matrix) -static const char MultTransposeMatrixdARB_names[] = - "p\0" /* Parameter signature */ - "glMultTransposeMatrixd\0" - "glMultTransposeMatrixdARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos) -static const char WindowPos2svMESA_names[] = - "p\0" /* Parameter signature */ - "glWindowPos2sv\0" - "glWindowPos2svARB\0" - "glWindowPos2svMESA\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) || defined(need_GL_ARB_texture_compression) -static const char CompressedTexImage1DARB_names[] = - "iiiiiip\0" /* Parameter signature */ - "glCompressedTexImage1D\0" - "glCompressedTexImage1DARB\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttrib2sNV_names[] = - "iii\0" /* Parameter signature */ - "glVertexAttrib2sNV\0" - ""; -#endif - -#if defined(need_GL_IBM_vertex_array_lists) -static const char NormalPointerListIBM_names[] = - "iipi\0" /* Parameter signature */ - "glNormalPointerListIBM\0" - ""; -#endif - -#if defined(need_GL_EXT_vertex_array) -static const char IndexPointerEXT_names[] = - "iiip\0" /* Parameter signature */ - "glIndexPointerEXT\0" - ""; -#endif - -#if defined(need_GL_EXT_vertex_array) -static const char NormalPointerEXT_names[] = - "iiip\0" /* Parameter signature */ - "glNormalPointerEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord3dARB_names[] = - "iddd\0" /* Parameter signature */ - "glMultiTexCoord3d\0" - "glMultiTexCoord3dARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord2iARB_names[] = - "iii\0" /* Parameter signature */ - "glMultiTexCoord2i\0" - "glMultiTexCoord2iARB\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char ReplacementCodeuiTexCoord2fNormal3fVertex3fSUN_names[] = - "iffffffff\0" /* Parameter signature */ - "glReplacementCodeuiTexCoord2fNormal3fVertex3fSUN\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord2svARB_names[] = - "ip\0" /* Parameter signature */ - "glMultiTexCoord2sv\0" - "glMultiTexCoord2svARB\0" - ""; -#endif - -#if defined(need_GL_SUN_triangle_list) -static const char ReplacementCodeubvSUN_names[] = - "p\0" /* Parameter signature */ - "glReplacementCodeubvSUN\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char Uniform3iARB_names[] = - "iiii\0" /* Parameter signature */ - "glUniform3i\0" - "glUniform3iARB\0" - ""; -#endif - -#if defined(need_GL_SGIX_fragment_lighting) -static const char GetFragmentMaterialfvSGIX_names[] = - "iip\0" /* Parameter signature */ - "glGetFragmentMaterialfvSGIX\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) -static const char GetShaderInfoLog_names[] = - "iipp\0" /* Parameter signature */ - "glGetShaderInfoLog\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_blend) -static const char WeightivARB_names[] = - "ip\0" /* Parameter signature */ - "glWeightivARB\0" - ""; -#endif - -#if defined(need_GL_SGIX_instruments) -static const char PollInstrumentsSGIX_names[] = - "p\0" /* Parameter signature */ - "glPollInstrumentsSGIX\0" - ""; -#endif - -#if defined(need_GL_SUN_global_alpha) -static const char GlobalAlphaFactordSUN_names[] = - "d\0" /* Parameter signature */ - "glGlobalAlphaFactordSUN\0" - ""; -#endif - -#if defined(need_GL_NV_register_combiners) -static const char GetFinalCombinerInputParameterfvNV_names[] = - "iip\0" /* Parameter signature */ - "glGetFinalCombinerInputParameterfvNV\0" - ""; -#endif - -#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object) -static const char GenerateMipmapEXT_names[] = - "i\0" /* Parameter signature */ - "glGenerateMipmap\0" - "glGenerateMipmapEXT\0" - ""; -#endif - -#if defined(need_GL_ATI_fragment_shader) -static const char SetFragmentShaderConstantATI_names[] = - "ip\0" /* Parameter signature */ - "glSetFragmentShaderConstantATI\0" - ""; -#endif - -#if defined(need_GL_NV_evaluators) -static const char GetMapAttribParameterivNV_names[] = - "iiip\0" /* Parameter signature */ - "glGetMapAttribParameterivNV\0" - ""; -#endif - -#if defined(need_GL_ARB_shader_objects) -static const char CreateShaderObjectARB_names[] = - "i\0" /* Parameter signature */ - "glCreateShaderObjectARB\0" - ""; -#endif - -#if defined(need_GL_SGIS_sharpen_texture) -static const char GetSharpenTexFuncSGIS_names[] = - "ip\0" /* Parameter signature */ - "glGetSharpenTexFuncSGIS\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_vertex_buffer_object) -static const char BufferDataARB_names[] = - "iipi\0" /* Parameter signature */ - "glBufferData\0" - "glBufferDataARB\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_array_range) -static const char FlushVertexArrayRangeNV_names[] = - "\0" /* Parameter signature */ - "glFlushVertexArrayRangeNV\0" - ""; -#endif - -#if defined(need_GL_ATI_fragment_shader) -static const char SampleMapATI_names[] = - "iii\0" /* Parameter signature */ - "glSampleMapATI\0" - ""; -#endif - -#if defined(need_GL_EXT_vertex_array) -static const char VertexPointerEXT_names[] = - "iiiip\0" /* Parameter signature */ - "glVertexPointerEXT\0" - ""; -#endif - -#if defined(need_GL_SGIS_texture_filter4) -static const char GetTexFilterFuncSGIS_names[] = - "iip\0" /* Parameter signature */ - "glGetTexFilterFuncSGIS\0" - ""; -#endif - -#if defined(need_GL_NV_register_combiners) -static const char GetCombinerOutputParameterfvNV_names[] = - "iiip\0" /* Parameter signature */ - "glGetCombinerOutputParameterfvNV\0" - ""; -#endif - -#if defined(need_GL_EXT_subtexture) -static const char TexSubImage1D_names[] = - "iiiiiip\0" /* Parameter signature */ - "glTexSubImage1D\0" - "glTexSubImage1DEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib1sARB_names[] = - "ii\0" /* Parameter signature */ - "glVertexAttrib1s\0" - "glVertexAttrib1sARB\0" - ""; -#endif - -#if defined(need_GL_ARB_sync) -static const char FenceSync_names[] = - "ii\0" /* Parameter signature */ - "glFenceSync\0" - ""; -#endif - -#if defined(need_GL_NV_register_combiners) -static const char FinalCombinerInputNV_names[] = - "iiii\0" /* Parameter signature */ - "glFinalCombinerInputNV\0" - ""; -#endif - -#if defined(need_GL_SGIX_flush_raster) -static const char FlushRasterSGIX_names[] = - "\0" /* Parameter signature */ - "glFlushRasterSGIX\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char ReplacementCodeuiTexCoord2fVertex3fSUN_names[] = - "ifffff\0" /* Parameter signature */ - "glReplacementCodeuiTexCoord2fVertex3fSUN\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char Uniform1fARB_names[] = - "if\0" /* Parameter signature */ - "glUniform1f\0" - "glUniform1fARB\0" - ""; -#endif - -#if defined(need_GL_EXT_texture_object) -static const char AreTexturesResident_names[] = - "ipp\0" /* Parameter signature */ - "glAreTexturesResident\0" - "glAreTexturesResidentEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ATI_separate_stencil) -static const char StencilOpSeparate_names[] = - "iiii\0" /* Parameter signature */ - "glStencilOpSeparate\0" - "glStencilOpSeparateATI\0" - ""; -#endif - -#if defined(need_GL_SGI_color_table) -static const char ColorTableParameteriv_names[] = - "iip\0" /* Parameter signature */ - "glColorTableParameteriv\0" - "glColorTableParameterivSGI\0" - ""; -#endif - -#if defined(need_GL_IBM_vertex_array_lists) -static const char FogCoordPointerListIBM_names[] = - "iipi\0" /* Parameter signature */ - "glFogCoordPointerListIBM\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos) -static const char WindowPos3dMESA_names[] = - "ddd\0" /* Parameter signature */ - "glWindowPos3d\0" - "glWindowPos3dARB\0" - "glWindowPos3dMESA\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_point_parameters) || defined(need_GL_EXT_point_parameters) || defined(need_GL_SGIS_point_parameters) -static const char PointParameterfvEXT_names[] = - "ip\0" /* Parameter signature */ - "glPointParameterfv\0" - "glPointParameterfvARB\0" - "glPointParameterfvEXT\0" - "glPointParameterfvSGIS\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos) -static const char WindowPos2fvMESA_names[] = - "p\0" /* Parameter signature */ - "glWindowPos2fv\0" - "glWindowPos2fvARB\0" - "glWindowPos2fvMESA\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color) -static const char SecondaryColor3bvEXT_names[] = - "p\0" /* Parameter signature */ - "glSecondaryColor3bv\0" - "glSecondaryColor3bvEXT\0" - ""; -#endif - -#if defined(need_GL_IBM_vertex_array_lists) -static const char VertexPointerListIBM_names[] = - "iiipi\0" /* Parameter signature */ - "glVertexPointerListIBM\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_program) -static const char GetProgramLocalParameterfvARB_names[] = - "iip\0" /* Parameter signature */ - "glGetProgramLocalParameterfvARB\0" - ""; -#endif - -#if defined(need_GL_SGIX_fragment_lighting) -static const char FragmentMaterialfSGIX_names[] = - "iif\0" /* Parameter signature */ - "glFragmentMaterialfSGIX\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char TexCoord2fNormal3fVertex3fSUN_names[] = - "ffffffff\0" /* Parameter signature */ - "glTexCoord2fNormal3fVertex3fSUN\0" - ""; -#endif - -#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object) -static const char RenderbufferStorageEXT_names[] = - "iiii\0" /* Parameter signature */ - "glRenderbufferStorage\0" - "glRenderbufferStorageEXT\0" - ""; -#endif - -#if defined(need_GL_NV_fence) -static const char IsFenceNV_names[] = - "i\0" /* Parameter signature */ - "glIsFenceNV\0" - ""; -#endif - -#if defined(need_GL_ARB_shader_objects) -static const char AttachObjectARB_names[] = - "ii\0" /* Parameter signature */ - "glAttachObjectARB\0" - ""; -#endif - -#if defined(need_GL_SGIX_fragment_lighting) -static const char GetFragmentLightivSGIX_names[] = - "iip\0" /* Parameter signature */ - "glGetFragmentLightivSGIX\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char UniformMatrix2fvARB_names[] = - "iiip\0" /* Parameter signature */ - "glUniformMatrix2fv\0" - "glUniformMatrix2fvARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord2fARB_names[] = - "iff\0" /* Parameter signature */ - "glMultiTexCoord2f\0" - "glMultiTexCoord2fARB\0" - ""; -#endif - -#if defined(need_GL_SGI_color_table) || defined(need_GL_EXT_paletted_texture) -static const char ColorTable_names[] = - "iiiiip\0" /* Parameter signature */ - "glColorTable\0" - "glColorTableSGI\0" - "glColorTableEXT\0" - ""; -#endif - -#if defined(need_GL_NV_evaluators) -static const char MapControlPointsNV_names[] = - "iiiiiiiip\0" /* Parameter signature */ - "glMapControlPointsNV\0" - ""; -#endif - -#if defined(need_GL_EXT_convolution) -static const char ConvolutionFilter2D_names[] = - "iiiiiip\0" /* Parameter signature */ - "glConvolutionFilter2D\0" - "glConvolutionFilter2DEXT\0" - ""; -#endif - -#if defined(need_GL_NV_evaluators) -static const char MapParameterfvNV_names[] = - "iip\0" /* Parameter signature */ - "glMapParameterfvNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib3dvARB_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib3dv\0" - "glVertexAttrib3dvARB\0" - ""; -#endif - -#if defined(need_GL_PGI_misc_hints) -static const char HintPGI_names[] = - "ii\0" /* Parameter signature */ - "glHintPGI\0" - ""; -#endif - -#if defined(need_GL_EXT_convolution) -static const char ConvolutionParameteriv_names[] = - "iip\0" /* Parameter signature */ - "glConvolutionParameteriv\0" - "glConvolutionParameterivEXT\0" - ""; -#endif - -#if defined(need_GL_EXT_cull_vertex) -static const char CullParameterdvEXT_names[] = - "ip\0" /* Parameter signature */ - "glCullParameterdvEXT\0" - ""; -#endif - -#if defined(need_GL_NV_fragment_program) -static const char ProgramNamedParameter4fNV_names[] = - "iipffff\0" /* Parameter signature */ - "glProgramNamedParameter4fNV\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char Color3fVertex3fSUN_names[] = - "ffffff\0" /* Parameter signature */ - "glColor3fVertex3fSUN\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_program) || defined(need_GL_NV_vertex_program) -static const char ProgramEnvParameter4fvARB_names[] = - "iip\0" /* Parameter signature */ - "glProgramEnvParameter4fvARB\0" - "glProgramParameter4fvNV\0" - ""; -#endif - -#if defined(need_GL_SGIX_fragment_lighting) -static const char FragmentLightModeliSGIX_names[] = - "ii\0" /* Parameter signature */ - "glFragmentLightModeliSGIX\0" - ""; -#endif - -#if defined(need_GL_EXT_convolution) -static const char ConvolutionParameterfv_names[] = - "iip\0" /* Parameter signature */ - "glConvolutionParameterfv\0" - "glConvolutionParameterfvEXT\0" - ""; -#endif - -#if defined(need_GL_3DFX_tbuffer) -static const char TbufferMask3DFX_names[] = - "i\0" /* Parameter signature */ - "glTbufferMask3DFX\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char LoadProgramNV_names[] = - "iiip\0" /* Parameter signature */ - "glLoadProgramNV\0" - ""; -#endif - -#if defined(need_GL_ARB_sync) -static const char WaitSync_names[] = - "iii\0" /* Parameter signature */ - "glWaitSync\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttrib4fvNV_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib4fvNV\0" - ""; -#endif - -#if defined(need_GL_ARB_shader_objects) -static const char GetAttachedObjectsARB_names[] = - "iipp\0" /* Parameter signature */ - "glGetAttachedObjectsARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char Uniform3fvARB_names[] = - "iip\0" /* Parameter signature */ - "glUniform3fv\0" - "glUniform3fvARB\0" - ""; -#endif - -#if defined(need_GL_EXT_draw_range_elements) -static const char DrawRangeElements_names[] = - "iiiiip\0" /* Parameter signature */ - "glDrawRangeElements\0" - "glDrawRangeElementsEXT\0" - ""; -#endif - -#if defined(need_GL_SGIX_sprite) -static const char SpriteParameterfvSGIX_names[] = - "ip\0" /* Parameter signature */ - "glSpriteParameterfvSGIX\0" - ""; -#endif - -#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object) -static const char CheckFramebufferStatusEXT_names[] = - "i\0" /* Parameter signature */ - "glCheckFramebufferStatus\0" - "glCheckFramebufferStatusEXT\0" - ""; -#endif - -#if defined(need_GL_SUN_global_alpha) -static const char GlobalAlphaFactoruiSUN_names[] = - "i\0" /* Parameter signature */ - "glGlobalAlphaFactoruiSUN\0" - ""; -#endif - -#if defined(need_GL_ARB_shader_objects) -static const char GetHandleARB_names[] = - "i\0" /* Parameter signature */ - "glGetHandleARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char GetVertexAttribivARB_names[] = - "iip\0" /* Parameter signature */ - "glGetVertexAttribiv\0" - "glGetVertexAttribivARB\0" - ""; -#endif - -#if defined(need_GL_NV_register_combiners) -static const char GetCombinerInputParameterfvNV_names[] = - "iiiip\0" /* Parameter signature */ - "glGetCombinerInputParameterfvNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) -static const char CreateProgram_names[] = - "\0" /* Parameter signature */ - "glCreateProgram\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) || defined(need_GL_ARB_transpose_matrix) -static const char LoadTransposeMatrixdARB_names[] = - "p\0" /* Parameter signature */ - "glLoadTransposeMatrixd\0" - "glLoadTransposeMatrixdARB\0" - ""; -#endif - -#if defined(need_GL_EXT_histogram) -static const char GetMinmax_names[] = - "iiiip\0" /* Parameter signature */ - "glGetMinmax\0" - "glGetMinmaxEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) -static const char StencilFuncSeparate_names[] = - "iiii\0" /* Parameter signature */ - "glStencilFuncSeparate\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color) -static const char SecondaryColor3sEXT_names[] = - "iii\0" /* Parameter signature */ - "glSecondaryColor3s\0" - "glSecondaryColor3sEXT\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char Color3fVertex3fvSUN_names[] = - "pp\0" /* Parameter signature */ - "glColor3fVertex3fvSUN\0" - ""; -#endif - -#if defined(need_GL_SUN_global_alpha) -static const char GlobalAlphaFactorbSUN_names[] = - "i\0" /* Parameter signature */ - "glGlobalAlphaFactorbSUN\0" - ""; -#endif - -#if defined(need_GL_HP_image_transform) -static const char ImageTransformParameterfvHP_names[] = - "iip\0" /* Parameter signature */ - "glImageTransformParameterfvHP\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib4ivARB_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib4iv\0" - "glVertexAttrib4ivARB\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttrib3fNV_names[] = - "ifff\0" /* Parameter signature */ - "glVertexAttrib3fNV\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttribs2dvNV_names[] = - "iip\0" /* Parameter signature */ - "glVertexAttribs2dvNV\0" - ""; -#endif - -#if defined(need_GL_EXT_timer_query) -static const char GetQueryObjectui64vEXT_names[] = - "iip\0" /* Parameter signature */ - "glGetQueryObjectui64vEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord3fvARB_names[] = - "ip\0" /* Parameter signature */ - "glMultiTexCoord3fv\0" - "glMultiTexCoord3fvARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color) -static const char SecondaryColor3dEXT_names[] = - "ddd\0" /* Parameter signature */ - "glSecondaryColor3d\0" - "glSecondaryColor3dEXT\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char GetProgramParameterfvNV_names[] = - "iiip\0" /* Parameter signature */ - "glGetProgramParameterfvNV\0" - ""; -#endif - -#if defined(need_GL_EXT_coordinate_frame) -static const char TangentPointerEXT_names[] = - "iip\0" /* Parameter signature */ - "glTangentPointerEXT\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char Color4fNormal3fVertex3fvSUN_names[] = - "ppp\0" /* Parameter signature */ - "glColor4fNormal3fVertex3fvSUN\0" - ""; -#endif - -#if defined(need_GL_SGIX_instruments) -static const char GetInstrumentsSGIX_names[] = - "\0" /* Parameter signature */ - "glGetInstrumentsSGIX\0" - ""; -#endif - -#if defined(need_GL_NV_evaluators) -static const char EvalMapsNV_names[] = - "ii\0" /* Parameter signature */ - "glEvalMapsNV\0" - ""; -#endif - -#if defined(need_GL_EXT_subtexture) -static const char TexSubImage2D_names[] = - "iiiiiiiip\0" /* Parameter signature */ - "glTexSubImage2D\0" - "glTexSubImage2DEXT\0" - ""; -#endif - -#if defined(need_GL_SGIX_fragment_lighting) -static const char FragmentLightivSGIX_names[] = - "iip\0" /* Parameter signature */ - "glFragmentLightivSGIX\0" - ""; -#endif - -#if defined(need_GL_APPLE_texture_range) -static const char GetTexParameterPointervAPPLE_names[] = - "iip\0" /* Parameter signature */ - "glGetTexParameterPointervAPPLE\0" - ""; -#endif - -#if defined(need_GL_EXT_pixel_transform) -static const char PixelTransformParameterfvEXT_names[] = - "iip\0" /* Parameter signature */ - "glPixelTransformParameterfvEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib4bvARB_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib4bv\0" - "glVertexAttrib4bvARB\0" - ""; -#endif - -#if defined(need_GL_ATI_fragment_shader) -static const char AlphaFragmentOp2ATI_names[] = - "iiiiiiiii\0" /* Parameter signature */ - "glAlphaFragmentOp2ATI\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord4sARB_names[] = - "iiiii\0" /* Parameter signature */ - "glMultiTexCoord4s\0" - "glMultiTexCoord4sARB\0" - ""; -#endif - -#if defined(need_GL_SGIX_fragment_lighting) -static const char GetFragmentMaterialivSGIX_names[] = - "iip\0" /* Parameter signature */ - "glGetFragmentMaterialivSGIX\0" - ""; -#endif - -#if defined(need_GL_MESA_window_pos) -static const char WindowPos4dMESA_names[] = - "dddd\0" /* Parameter signature */ - "glWindowPos4dMESA\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_blend) -static const char WeightPointerARB_names[] = - "iiip\0" /* Parameter signature */ - "glWeightPointerARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos) -static const char WindowPos2dMESA_names[] = - "dd\0" /* Parameter signature */ - "glWindowPos2d\0" - "glWindowPos2dARB\0" - "glWindowPos2dMESA\0" - ""; -#endif - -#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object) -static const char FramebufferTexture3DEXT_names[] = - "iiiiii\0" /* Parameter signature */ - "glFramebufferTexture3D\0" - "glFramebufferTexture3DEXT\0" - ""; -#endif - -#if defined(need_GL_EXT_blend_minmax) -static const char BlendEquation_names[] = - "i\0" /* Parameter signature */ - "glBlendEquation\0" - "glBlendEquationEXT\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttrib3dNV_names[] = - "iddd\0" /* Parameter signature */ - "glVertexAttrib3dNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib3dARB_names[] = - "iddd\0" /* Parameter signature */ - "glVertexAttrib3d\0" - "glVertexAttrib3dARB\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char ReplacementCodeuiTexCoord2fColor4fNormal3fVertex3fvSUN_names[] = - "ppppp\0" /* Parameter signature */ - "glReplacementCodeuiTexCoord2fColor4fNormal3fVertex3fvSUN\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib4fARB_names[] = - "iffff\0" /* Parameter signature */ - "glVertexAttrib4f\0" - "glVertexAttrib4fARB\0" - ""; -#endif - -#if defined(need_GL_EXT_index_func) -static const char IndexFuncEXT_names[] = - "if\0" /* Parameter signature */ - "glIndexFuncEXT\0" - ""; -#endif - -#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object) -static const char FramebufferTexture2DEXT_names[] = - "iiiii\0" /* Parameter signature */ - "glFramebufferTexture2D\0" - "glFramebufferTexture2DEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord2dvARB_names[] = - "ip\0" /* Parameter signature */ - "glMultiTexCoord2dv\0" - "glMultiTexCoord2dvARB\0" - ""; -#endif - -#if defined(need_GL_EXT_cull_vertex) -static const char CullParameterfvEXT_names[] = - "ip\0" /* Parameter signature */ - "glCullParameterfvEXT\0" - ""; -#endif - -#if defined(need_GL_NV_fragment_program) -static const char ProgramNamedParameter4fvNV_names[] = - "iipp\0" /* Parameter signature */ - "glProgramNamedParameter4fvNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color) -static const char SecondaryColorPointerEXT_names[] = - "iiip\0" /* Parameter signature */ - "glSecondaryColorPointer\0" - "glSecondaryColorPointerEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib4fvARB_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib4fv\0" - "glVertexAttrib4fvARB\0" - ""; -#endif - -#if defined(need_GL_IBM_vertex_array_lists) -static const char ColorPointerListIBM_names[] = - "iiipi\0" /* Parameter signature */ - "glColorPointerListIBM\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char GetActiveUniformARB_names[] = - "iiipppp\0" /* Parameter signature */ - "glGetActiveUniform\0" - "glGetActiveUniformARB\0" - ""; -#endif - -#if defined(need_GL_HP_image_transform) -static const char ImageTransformParameteriHP_names[] = - "iii\0" /* Parameter signature */ - "glImageTransformParameteriHP\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord1svARB_names[] = - "ip\0" /* Parameter signature */ - "glMultiTexCoord1sv\0" - "glMultiTexCoord1svARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_occlusion_query) -static const char EndQueryARB_names[] = - "i\0" /* Parameter signature */ - "glEndQuery\0" - "glEndQueryARB\0" - ""; -#endif - -#if defined(need_GL_NV_fence) -static const char DeleteFencesNV_names[] = - "ip\0" /* Parameter signature */ - "glDeleteFencesNV\0" - ""; -#endif - -#if defined(need_GL_SGIX_polynomial_ffd) -static const char DeformationMap3dSGIX_names[] = - "iddiiddiiddiip\0" /* Parameter signature */ - "glDeformationMap3dSGIX\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) -static const char IsShader_names[] = - "i\0" /* Parameter signature */ - "glIsShader\0" - ""; -#endif - -#if defined(need_GL_HP_image_transform) -static const char GetImageTransformParameterivHP_names[] = - "iip\0" /* Parameter signature */ - "glGetImageTransformParameterivHP\0" - ""; -#endif - -#if defined(need_GL_MESA_window_pos) -static const char WindowPos4ivMESA_names[] = - "p\0" /* Parameter signature */ - "glWindowPos4ivMESA\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord3svARB_names[] = - "ip\0" /* Parameter signature */ - "glMultiTexCoord3sv\0" - "glMultiTexCoord3svARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord4iARB_names[] = - "iiiii\0" /* Parameter signature */ - "glMultiTexCoord4i\0" - "glMultiTexCoord4iARB\0" - ""; -#endif - -#if defined(need_GL_EXT_coordinate_frame) -static const char Binormal3ivEXT_names[] = - "p\0" /* Parameter signature */ - "glBinormal3ivEXT\0" - ""; -#endif - -#if defined(need_GL_MESA_resize_buffers) -static const char ResizeBuffersMESA_names[] = - "\0" /* Parameter signature */ - "glResizeBuffersMESA\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char GetUniformivARB_names[] = - "iip\0" /* Parameter signature */ - "glGetUniformiv\0" - "glGetUniformivARB\0" - ""; -#endif - -#if defined(need_GL_SGIS_pixel_texture) -static const char PixelTexGenParameteriSGIS_names[] = - "ii\0" /* Parameter signature */ - "glPixelTexGenParameteriSGIS\0" - ""; -#endif - -#if defined(need_GL_INTEL_parallel_arrays) -static const char VertexPointervINTEL_names[] = - "iip\0" /* Parameter signature */ - "glVertexPointervINTEL\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char ReplacementCodeuiColor4fNormal3fVertex3fvSUN_names[] = - "pppp\0" /* Parameter signature */ - "glReplacementCodeuiColor4fNormal3fVertex3fvSUN\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color) -static const char SecondaryColor3uiEXT_names[] = - "iii\0" /* Parameter signature */ - "glSecondaryColor3ui\0" - "glSecondaryColor3uiEXT\0" - ""; -#endif - -#if defined(need_GL_SGIX_instruments) -static const char StartInstrumentsSGIX_names[] = - "\0" /* Parameter signature */ - "glStartInstrumentsSGIX\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color) -static const char SecondaryColor3usvEXT_names[] = - "p\0" /* Parameter signature */ - "glSecondaryColor3usv\0" - "glSecondaryColor3usvEXT\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttrib2fvNV_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib2fvNV\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_program) -static const char ProgramLocalParameter4dvARB_names[] = - "iip\0" /* Parameter signature */ - "glProgramLocalParameter4dvARB\0" - ""; -#endif - -#if defined(need_GL_ARB_matrix_palette) -static const char MatrixIndexuivARB_names[] = - "ip\0" /* Parameter signature */ - "glMatrixIndexuivARB\0" - ""; -#endif - -#if defined(need_GL_ARB_framebuffer_object) -static const char RenderbufferStorageMultisample_names[] = - "iiiii\0" /* Parameter signature */ - "glRenderbufferStorageMultisample\0" - ""; -#endif - -#if defined(need_GL_EXT_coordinate_frame) -static const char Tangent3sEXT_names[] = - "iii\0" /* Parameter signature */ - "glTangent3sEXT\0" - ""; -#endif - -#if defined(need_GL_SUN_global_alpha) -static const char GlobalAlphaFactorfSUN_names[] = - "f\0" /* Parameter signature */ - "glGlobalAlphaFactorfSUN\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord3iARB_names[] = - "iiii\0" /* Parameter signature */ - "glMultiTexCoord3i\0" - "glMultiTexCoord3iARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) -static const char IsProgram_names[] = - "i\0" /* Parameter signature */ - "glIsProgram\0" - ""; -#endif - -#if defined(need_GL_IBM_vertex_array_lists) -static const char TexCoordPointerListIBM_names[] = - "iiipi\0" /* Parameter signature */ - "glTexCoordPointerListIBM\0" - ""; -#endif - -#if defined(need_GL_SUN_global_alpha) -static const char GlobalAlphaFactorusSUN_names[] = - "i\0" /* Parameter signature */ - "glGlobalAlphaFactorusSUN\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttrib2dvNV_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib2dvNV\0" - ""; -#endif - -#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object) -static const char FramebufferRenderbufferEXT_names[] = - "iiii\0" /* Parameter signature */ - "glFramebufferRenderbuffer\0" - "glFramebufferRenderbufferEXT\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttrib1dvNV_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib1dvNV\0" - ""; -#endif - -#if defined(need_GL_EXT_texture_object) -static const char GenTextures_names[] = - "ip\0" /* Parameter signature */ - "glGenTextures\0" - "glGenTexturesEXT\0" - ""; -#endif - -#if defined(need_GL_NV_fence) -static const char SetFenceNV_names[] = - "ii\0" /* Parameter signature */ - "glSetFenceNV\0" - ""; -#endif - -#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object) -static const char FramebufferTexture1DEXT_names[] = - "iiiii\0" /* Parameter signature */ - "glFramebufferTexture1D\0" - "glFramebufferTexture1DEXT\0" - ""; -#endif - -#if defined(need_GL_NV_register_combiners) -static const char GetCombinerOutputParameterivNV_names[] = - "iiip\0" /* Parameter signature */ - "glGetCombinerOutputParameterivNV\0" - ""; -#endif - -#if defined(need_GL_SGIS_pixel_texture) -static const char PixelTexGenParameterivSGIS_names[] = - "ip\0" /* Parameter signature */ - "glPixelTexGenParameterivSGIS\0" - ""; -#endif - -#if defined(need_GL_EXT_texture_perturb_normal) -static const char TextureNormalEXT_names[] = - "i\0" /* Parameter signature */ - "glTextureNormalEXT\0" - ""; -#endif - -#if defined(need_GL_IBM_vertex_array_lists) -static const char IndexPointerListIBM_names[] = - "iipi\0" /* Parameter signature */ - "glIndexPointerListIBM\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_blend) -static const char WeightfvARB_names[] = - "ip\0" /* Parameter signature */ - "glWeightfvARB\0" - ""; -#endif - -#if defined(need_GL_MESA_window_pos) -static const char WindowPos4fMESA_names[] = - "ffff\0" /* Parameter signature */ - "glWindowPos4fMESA\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos) -static const char WindowPos3dvMESA_names[] = - "p\0" /* Parameter signature */ - "glWindowPos3dv\0" - "glWindowPos3dvARB\0" - "glWindowPos3dvMESA\0" - ""; -#endif - -#if defined(need_GL_EXT_timer_query) -static const char GetQueryObjecti64vEXT_names[] = - "iip\0" /* Parameter signature */ - "glGetQueryObjecti64vEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord1dARB_names[] = - "id\0" /* Parameter signature */ - "glMultiTexCoord1d\0" - "glMultiTexCoord1dARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_NV_point_sprite) -static const char PointParameterivNV_names[] = - "ip\0" /* Parameter signature */ - "glPointParameteriv\0" - "glPointParameterivNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char Uniform2fvARB_names[] = - "iip\0" /* Parameter signature */ - "glUniform2fv\0" - "glUniform2fvARB\0" - ""; -#endif - -#if defined(need_GL_APPLE_flush_buffer_range) -static const char BufferParameteriAPPLE_names[] = - "iii\0" /* Parameter signature */ - "glBufferParameteriAPPLE\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord3dvARB_names[] = - "ip\0" /* Parameter signature */ - "glMultiTexCoord3dv\0" - "glMultiTexCoord3dvARB\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char ReplacementCodeuiTexCoord2fNormal3fVertex3fvSUN_names[] = - "pppp\0" /* Parameter signature */ - "glReplacementCodeuiTexCoord2fNormal3fVertex3fvSUN\0" - ""; -#endif - -#if defined(need_GL_ARB_shader_objects) -static const char DeleteObjectARB_names[] = - "i\0" /* Parameter signature */ - "glDeleteObjectARB\0" - ""; -#endif - -#if defined(need_GL_ARB_matrix_palette) -static const char MatrixIndexPointerARB_names[] = - "iiip\0" /* Parameter signature */ - "glMatrixIndexPointerARB\0" - ""; -#endif - -#if defined(need_GL_NV_fragment_program) -static const char ProgramNamedParameter4dvNV_names[] = - "iipp\0" /* Parameter signature */ - "glProgramNamedParameter4dvNV\0" - ""; -#endif - -#if defined(need_GL_EXT_coordinate_frame) -static const char Tangent3fvEXT_names[] = - "p\0" /* Parameter signature */ - "glTangent3fvEXT\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_array_object) -static const char GenVertexArrays_names[] = - "ip\0" /* Parameter signature */ - "glGenVertexArrays\0" - ""; -#endif - -#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object) -static const char BindFramebufferEXT_names[] = - "ii\0" /* Parameter signature */ - "glBindFramebuffer\0" - "glBindFramebufferEXT\0" - ""; -#endif - -#if defined(need_GL_SGIX_reference_plane) -static const char ReferencePlaneSGIX_names[] = - "p\0" /* Parameter signature */ - "glReferencePlaneSGIX\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char ValidateProgramARB_names[] = - "i\0" /* Parameter signature */ - "glValidateProgram\0" - "glValidateProgramARB\0" - ""; -#endif - -#if defined(need_GL_EXT_compiled_vertex_array) -static const char UnlockArraysEXT_names[] = - "\0" /* Parameter signature */ - "glUnlockArraysEXT\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char TexCoord2fColor3fVertex3fSUN_names[] = - "ffffffff\0" /* Parameter signature */ - "glTexCoord2fColor3fVertex3fSUN\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos) -static const char WindowPos3fvMESA_names[] = - "p\0" /* Parameter signature */ - "glWindowPos3fv\0" - "glWindowPos3fvARB\0" - "glWindowPos3fvMESA\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttrib1svNV_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib1svNV\0" - ""; -#endif - -#if defined(need_GL_EXT_copy_texture) -static const char CopyTexSubImage3D_names[] = - "iiiiiiiii\0" /* Parameter signature */ - "glCopyTexSubImage3D\0" - "glCopyTexSubImage3DEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib2dARB_names[] = - "idd\0" /* Parameter signature */ - "glVertexAttrib2d\0" - "glVertexAttrib2dARB\0" - ""; -#endif - -#if defined(need_GL_ARB_sync) -static const char GetInteger64v_names[] = - "ip\0" /* Parameter signature */ - "glGetInteger64v\0" - ""; -#endif - -#if defined(need_GL_SGIS_texture_color_mask) -static const char TextureColorMaskSGIS_names[] = - "iiii\0" /* Parameter signature */ - "glTextureColorMaskSGIS\0" - ""; -#endif - -#if defined(need_GL_SGI_color_table) || defined(need_GL_EXT_paletted_texture) -static const char GetColorTable_names[] = - "iiip\0" /* Parameter signature */ - "glGetColorTable\0" - "glGetColorTableSGI\0" - "glGetColorTableEXT\0" - ""; -#endif - -#if defined(need_GL_SGI_color_table) -static const char CopyColorTable_names[] = - "iiiii\0" /* Parameter signature */ - "glCopyColorTable\0" - "glCopyColorTableSGI\0" - ""; -#endif - -#if defined(need_GL_EXT_histogram) -static const char GetHistogramParameterfv_names[] = - "iip\0" /* Parameter signature */ - "glGetHistogramParameterfv\0" - "glGetHistogramParameterfvEXT\0" - ""; -#endif - -#if defined(need_GL_INTEL_parallel_arrays) -static const char ColorPointervINTEL_names[] = - "iip\0" /* Parameter signature */ - "glColorPointervINTEL\0" - ""; -#endif - -#if defined(need_GL_ATI_fragment_shader) -static const char AlphaFragmentOp1ATI_names[] = - "iiiiii\0" /* Parameter signature */ - "glAlphaFragmentOp1ATI\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord3ivARB_names[] = - "ip\0" /* Parameter signature */ - "glMultiTexCoord3iv\0" - "glMultiTexCoord3ivARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord2sARB_names[] = - "iii\0" /* Parameter signature */ - "glMultiTexCoord2s\0" - "glMultiTexCoord2sARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib1dvARB_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib1dv\0" - "glVertexAttrib1dvARB\0" - ""; -#endif - -#if defined(need_GL_EXT_texture_object) -static const char DeleteTextures_names[] = - "ip\0" /* Parameter signature */ - "glDeleteTextures\0" - "glDeleteTexturesEXT\0" - ""; -#endif - -#if defined(need_GL_EXT_vertex_array) -static const char TexCoordPointerEXT_names[] = - "iiiip\0" /* Parameter signature */ - "glTexCoordPointerEXT\0" - ""; -#endif - -#if defined(need_GL_SGIS_texture4D) -static const char TexSubImage4DSGIS_names[] = - "iiiiiiiiiiiip\0" /* Parameter signature */ - "glTexSubImage4DSGIS\0" - ""; -#endif - -#if defined(need_GL_NV_register_combiners2) -static const char CombinerStageParameterfvNV_names[] = - "iip\0" /* Parameter signature */ - "glCombinerStageParameterfvNV\0" - ""; -#endif - -#if defined(need_GL_SGIX_instruments) -static const char StopInstrumentsSGIX_names[] = - "i\0" /* Parameter signature */ - "glStopInstrumentsSGIX\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char TexCoord4fColor4fNormal3fVertex4fSUN_names[] = - "fffffffffffffff\0" /* Parameter signature */ - "glTexCoord4fColor4fNormal3fVertex4fSUN\0" - ""; -#endif - -#if defined(need_GL_SGIX_polynomial_ffd) -static const char DeformSGIX_names[] = - "i\0" /* Parameter signature */ - "glDeformSGIX\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char GetVertexAttribfvARB_names[] = - "iip\0" /* Parameter signature */ - "glGetVertexAttribfv\0" - "glGetVertexAttribfvARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color) -static const char SecondaryColor3ivEXT_names[] = - "p\0" /* Parameter signature */ - "glSecondaryColor3iv\0" - "glSecondaryColor3ivEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_1) -static const char UniformMatrix4x2fv_names[] = - "iiip\0" /* Parameter signature */ - "glUniformMatrix4x2fv\0" - ""; -#endif - -#if defined(need_GL_SGIS_detail_texture) -static const char GetDetailTexFuncSGIS_names[] = - "ip\0" /* Parameter signature */ - "glGetDetailTexFuncSGIS\0" - ""; -#endif - -#if defined(need_GL_NV_register_combiners2) -static const char GetCombinerStageParameterfvNV_names[] = - "iip\0" /* Parameter signature */ - "glGetCombinerStageParameterfvNV\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_array_object) -static const char BindVertexArray_names[] = - "i\0" /* Parameter signature */ - "glBindVertexArray\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char Color4ubVertex2fvSUN_names[] = - "pp\0" /* Parameter signature */ - "glColor4ubVertex2fvSUN\0" - ""; -#endif - -#if defined(need_GL_SGIS_texture_filter4) -static const char TexFilterFuncSGIS_names[] = - "iiip\0" /* Parameter signature */ - "glTexFilterFuncSGIS\0" - ""; -#endif - -#if defined(need_GL_SGIS_multisample) || defined(need_GL_EXT_multisample) -static const char SampleMaskSGIS_names[] = - "fi\0" /* Parameter signature */ - "glSampleMaskSGIS\0" - "glSampleMaskEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_shader) -static const char GetAttribLocationARB_names[] = - "ip\0" /* Parameter signature */ - "glGetAttribLocation\0" - "glGetAttribLocationARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib4ubvARB_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib4ubv\0" - "glVertexAttrib4ubvARB\0" - ""; -#endif - -#if defined(need_GL_SGIS_detail_texture) -static const char DetailTexFuncSGIS_names[] = - "iip\0" /* Parameter signature */ - "glDetailTexFuncSGIS\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char Normal3fVertex3fSUN_names[] = - "ffffff\0" /* Parameter signature */ - "glNormal3fVertex3fSUN\0" - ""; -#endif - -#if defined(need_GL_EXT_copy_texture) -static const char CopyTexImage2D_names[] = - "iiiiiiii\0" /* Parameter signature */ - "glCopyTexImage2D\0" - "glCopyTexImage2DEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_vertex_buffer_object) -static const char GetBufferPointervARB_names[] = - "iip\0" /* Parameter signature */ - "glGetBufferPointerv\0" - "glGetBufferPointervARB\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_program) || defined(need_GL_NV_vertex_program) -static const char ProgramEnvParameter4fARB_names[] = - "iiffff\0" /* Parameter signature */ - "glProgramEnvParameter4fARB\0" - "glProgramParameter4fNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char Uniform3ivARB_names[] = - "iip\0" /* Parameter signature */ - "glUniform3iv\0" - "glUniform3ivARB\0" - ""; -#endif - -#if defined(need_GL_NV_fence) -static const char GetFenceivNV_names[] = - "iip\0" /* Parameter signature */ - "glGetFenceivNV\0" - ""; -#endif - -#if defined(need_GL_MESA_window_pos) -static const char WindowPos4dvMESA_names[] = - "p\0" /* Parameter signature */ - "glWindowPos4dvMESA\0" - ""; -#endif - -#if defined(need_GL_EXT_color_subtable) -static const char ColorSubTable_names[] = - "iiiiip\0" /* Parameter signature */ - "glColorSubTable\0" - "glColorSubTableEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord4ivARB_names[] = - "ip\0" /* Parameter signature */ - "glMultiTexCoord4iv\0" - "glMultiTexCoord4ivARB\0" - ""; -#endif - -#if defined(need_GL_EXT_gpu_program_parameters) -static const char ProgramLocalParameters4fvEXT_names[] = - "iiip\0" /* Parameter signature */ - "glProgramLocalParameters4fvEXT\0" - ""; -#endif - -#if defined(need_GL_NV_evaluators) -static const char GetMapAttribParameterfvNV_names[] = - "iiip\0" /* Parameter signature */ - "glGetMapAttribParameterfvNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib4sARB_names[] = - "iiiii\0" /* Parameter signature */ - "glVertexAttrib4s\0" - "glVertexAttrib4sARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_occlusion_query) -static const char GetQueryObjectuivARB_names[] = - "iip\0" /* Parameter signature */ - "glGetQueryObjectuiv\0" - "glGetQueryObjectuivARB\0" - ""; -#endif - -#if defined(need_GL_NV_evaluators) -static const char MapParameterivNV_names[] = - "iip\0" /* Parameter signature */ - "glMapParameterivNV\0" - ""; -#endif - -#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object) -static const char GenRenderbuffersEXT_names[] = - "ip\0" /* Parameter signature */ - "glGenRenderbuffers\0" - "glGenRenderbuffersEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib2dvARB_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib2dv\0" - "glVertexAttrib2dvARB\0" - ""; -#endif - -#if defined(need_GL_EXT_vertex_array) -static const char EdgeFlagPointerEXT_names[] = - "iip\0" /* Parameter signature */ - "glEdgeFlagPointerEXT\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttribs2svNV_names[] = - "iip\0" /* Parameter signature */ - "glVertexAttribs2svNV\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_blend) -static const char WeightbvARB_names[] = - "ip\0" /* Parameter signature */ - "glWeightbvARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib2fvARB_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib2fv\0" - "glVertexAttrib2fvARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_vertex_buffer_object) -static const char GetBufferParameterivARB_names[] = - "iip\0" /* Parameter signature */ - "glGetBufferParameteriv\0" - "glGetBufferParameterivARB\0" - ""; -#endif - -#if defined(need_GL_SGIX_list_priority) -static const char ListParameteriSGIX_names[] = - "iii\0" /* Parameter signature */ - "glListParameteriSGIX\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char ReplacementCodeuiColor4fNormal3fVertex3fSUN_names[] = - "iffffffffff\0" /* Parameter signature */ - "glReplacementCodeuiColor4fNormal3fVertex3fSUN\0" - ""; -#endif - -#if defined(need_GL_SGIX_instruments) -static const char InstrumentsBufferSGIX_names[] = - "ip\0" /* Parameter signature */ - "glInstrumentsBufferSGIX\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib4NivARB_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib4Niv\0" - "glVertexAttrib4NivARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) -static const char GetAttachedShaders_names[] = - "iipp\0" /* Parameter signature */ - "glGetAttachedShaders\0" - ""; -#endif - -#if defined(need_GL_APPLE_vertex_array_object) -static const char GenVertexArraysAPPLE_names[] = - "ip\0" /* Parameter signature */ - "glGenVertexArraysAPPLE\0" - ""; -#endif - -#if defined(need_GL_EXT_gpu_program_parameters) -static const char ProgramEnvParameters4fvEXT_names[] = - "iiip\0" /* Parameter signature */ - "glProgramEnvParameters4fvEXT\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char TexCoord2fColor4fNormal3fVertex3fvSUN_names[] = - "pppp\0" /* Parameter signature */ - "glTexCoord2fColor4fNormal3fVertex3fvSUN\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos) -static const char WindowPos2iMESA_names[] = - "ii\0" /* Parameter signature */ - "glWindowPos2i\0" - "glWindowPos2iARB\0" - "glWindowPos2iMESA\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color) -static const char SecondaryColor3fvEXT_names[] = - "p\0" /* Parameter signature */ - "glSecondaryColor3fv\0" - "glSecondaryColor3fvEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) || defined(need_GL_ARB_texture_compression) -static const char CompressedTexSubImage1DARB_names[] = - "iiiiiip\0" /* Parameter signature */ - "glCompressedTexSubImage1D\0" - "glCompressedTexSubImage1DARB\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char GetVertexAttribivNV_names[] = - "iip\0" /* Parameter signature */ - "glGetVertexAttribivNV\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_program) -static const char GetProgramStringARB_names[] = - "iip\0" /* Parameter signature */ - "glGetProgramStringARB\0" - ""; -#endif - -#if defined(need_GL_ATI_envmap_bumpmap) -static const char TexBumpParameterfvATI_names[] = - "ip\0" /* Parameter signature */ - "glTexBumpParameterfvATI\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char CompileShaderARB_names[] = - "i\0" /* Parameter signature */ - "glCompileShader\0" - "glCompileShaderARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) -static const char DeleteShader_names[] = - "i\0" /* Parameter signature */ - "glDeleteShader\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char Uniform3fARB_names[] = - "ifff\0" /* Parameter signature */ - "glUniform3f\0" - "glUniform3fARB\0" - ""; -#endif - -#if defined(need_GL_SGIX_list_priority) -static const char ListParameterfvSGIX_names[] = - "iip\0" /* Parameter signature */ - "glListParameterfvSGIX\0" - ""; -#endif - -#if defined(need_GL_EXT_coordinate_frame) -static const char Tangent3dvEXT_names[] = - "p\0" /* Parameter signature */ - "glTangent3dvEXT\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char GetVertexAttribfvNV_names[] = - "iip\0" /* Parameter signature */ - "glGetVertexAttribfvNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos) -static const char WindowPos3sMESA_names[] = - "iii\0" /* Parameter signature */ - "glWindowPos3s\0" - "glWindowPos3sARB\0" - "glWindowPos3sMESA\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttrib2svNV_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib2svNV\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttribs1fvNV_names[] = - "iip\0" /* Parameter signature */ - "glVertexAttribs1fvNV\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char TexCoord2fVertex3fvSUN_names[] = - "pp\0" /* Parameter signature */ - "glTexCoord2fVertex3fvSUN\0" - ""; -#endif - -#if defined(need_GL_MESA_window_pos) -static const char WindowPos4sMESA_names[] = - "iiii\0" /* Parameter signature */ - "glWindowPos4sMESA\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib4NuivARB_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib4Nuiv\0" - "glVertexAttrib4NuivARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char ClientActiveTextureARB_names[] = - "i\0" /* Parameter signature */ - "glClientActiveTexture\0" - "glClientActiveTextureARB\0" - ""; -#endif - -#if defined(need_GL_SGIX_pixel_texture) -static const char PixelTexGenSGIX_names[] = - "i\0" /* Parameter signature */ - "glPixelTexGenSGIX\0" - ""; -#endif - -#if defined(need_GL_SUN_triangle_list) -static const char ReplacementCodeusvSUN_names[] = - "p\0" /* Parameter signature */ - "glReplacementCodeusvSUN\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char Uniform4fARB_names[] = - "iffff\0" /* Parameter signature */ - "glUniform4f\0" - "glUniform4fARB\0" - ""; -#endif - -#if defined(need_GL_ARB_map_buffer_range) -static const char FlushMappedBufferRange_names[] = - "iii\0" /* Parameter signature */ - "glFlushMappedBufferRange\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_program) || defined(need_GL_NV_vertex_program) -static const char IsProgramNV_names[] = - "i\0" /* Parameter signature */ - "glIsProgramARB\0" - "glIsProgramNV\0" - ""; -#endif - -#if defined(need_GL_APPLE_flush_buffer_range) -static const char FlushMappedBufferRangeAPPLE_names[] = - "iii\0" /* Parameter signature */ - "glFlushMappedBufferRangeAPPLE\0" - ""; -#endif - -#if defined(need_GL_SUN_triangle_list) -static const char ReplacementCodePointerSUN_names[] = - "iip\0" /* Parameter signature */ - "glReplacementCodePointerSUN\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_program) || defined(need_GL_NV_vertex_program) -static const char ProgramEnvParameter4dARB_names[] = - "iidddd\0" /* Parameter signature */ - "glProgramEnvParameter4dARB\0" - "glProgramParameter4dNV\0" - ""; -#endif - -#if defined(need_GL_SGI_color_table) -static const char ColorTableParameterfv_names[] = - "iip\0" /* Parameter signature */ - "glColorTableParameterfv\0" - "glColorTableParameterfvSGI\0" - ""; -#endif - -#if defined(need_GL_SGIX_fragment_lighting) -static const char FragmentLightModelfSGIX_names[] = - "if\0" /* Parameter signature */ - "glFragmentLightModelfSGIX\0" - ""; -#endif - -#if defined(need_GL_EXT_coordinate_frame) -static const char Binormal3bvEXT_names[] = - "p\0" /* Parameter signature */ - "glBinormal3bvEXT\0" - ""; -#endif - -#if defined(need_GL_EXT_texture_object) -static const char IsTexture_names[] = - "i\0" /* Parameter signature */ - "glIsTexture\0" - "glIsTextureEXT\0" - ""; -#endif - -#if defined(need_GL_EXT_vertex_weighting) -static const char VertexWeightfvEXT_names[] = - "p\0" /* Parameter signature */ - "glVertexWeightfvEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib1dARB_names[] = - "id\0" /* Parameter signature */ - "glVertexAttrib1d\0" - "glVertexAttrib1dARB\0" - ""; -#endif - -#if defined(need_GL_HP_image_transform) -static const char ImageTransformParameterivHP_names[] = - "iip\0" /* Parameter signature */ - "glImageTransformParameterivHP\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_occlusion_query) -static const char DeleteQueriesARB_names[] = - "ip\0" /* Parameter signature */ - "glDeleteQueries\0" - "glDeleteQueriesARB\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char Color4ubVertex2fSUN_names[] = - "iiiiff\0" /* Parameter signature */ - "glColor4ubVertex2fSUN\0" - ""; -#endif - -#if defined(need_GL_SGIX_fragment_lighting) -static const char FragmentColorMaterialSGIX_names[] = - "ii\0" /* Parameter signature */ - "glFragmentColorMaterialSGIX\0" - ""; -#endif - -#if defined(need_GL_ARB_matrix_palette) -static const char CurrentPaletteMatrixARB_names[] = - "i\0" /* Parameter signature */ - "glCurrentPaletteMatrixARB\0" - ""; -#endif - -#if defined(need_GL_SGIS_multisample) || defined(need_GL_EXT_multisample) -static const char SamplePatternSGIS_names[] = - "i\0" /* Parameter signature */ - "glSamplePatternSGIS\0" - "glSamplePatternEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_occlusion_query) -static const char IsQueryARB_names[] = - "i\0" /* Parameter signature */ - "glIsQuery\0" - "glIsQueryARB\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char ReplacementCodeuiColor4ubVertex3fSUN_names[] = - "iiiiifff\0" /* Parameter signature */ - "glReplacementCodeuiColor4ubVertex3fSUN\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib4usvARB_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib4usv\0" - "glVertexAttrib4usvARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char LinkProgramARB_names[] = - "i\0" /* Parameter signature */ - "glLinkProgram\0" - "glLinkProgramARB\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttrib2fNV_names[] = - "iff\0" /* Parameter signature */ - "glVertexAttrib2fNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char ShaderSourceARB_names[] = - "iipp\0" /* Parameter signature */ - "glShaderSource\0" - "glShaderSourceARB\0" - ""; -#endif - -#if defined(need_GL_SGIX_fragment_lighting) -static const char FragmentMaterialiSGIX_names[] = - "iii\0" /* Parameter signature */ - "glFragmentMaterialiSGIX\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib3svARB_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib3sv\0" - "glVertexAttrib3svARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) || defined(need_GL_ARB_texture_compression) -static const char CompressedTexSubImage3DARB_names[] = - "iiiiiiiiiip\0" /* Parameter signature */ - "glCompressedTexSubImage3D\0" - "glCompressedTexSubImage3DARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos) -static const char WindowPos2ivMESA_names[] = - "p\0" /* Parameter signature */ - "glWindowPos2iv\0" - "glWindowPos2ivARB\0" - "glWindowPos2ivMESA\0" - ""; -#endif - -#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object) -static const char IsFramebufferEXT_names[] = - "i\0" /* Parameter signature */ - "glIsFramebuffer\0" - "glIsFramebufferEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char Uniform4ivARB_names[] = - "iip\0" /* Parameter signature */ - "glUniform4iv\0" - "glUniform4ivARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char GetVertexAttribdvARB_names[] = - "iip\0" /* Parameter signature */ - "glGetVertexAttribdv\0" - "glGetVertexAttribdvARB\0" - ""; -#endif - -#if defined(need_GL_ATI_envmap_bumpmap) -static const char TexBumpParameterivATI_names[] = - "ip\0" /* Parameter signature */ - "glTexBumpParameterivATI\0" - ""; -#endif - -#if defined(need_GL_EXT_convolution) -static const char GetSeparableFilter_names[] = - "iiippp\0" /* Parameter signature */ - "glGetSeparableFilter\0" - "glGetSeparableFilterEXT\0" - ""; -#endif - -#if defined(need_GL_EXT_coordinate_frame) -static const char Binormal3dEXT_names[] = - "ddd\0" /* Parameter signature */ - "glBinormal3dEXT\0" - ""; -#endif - -#if defined(need_GL_SGIX_sprite) -static const char SpriteParameteriSGIX_names[] = - "ii\0" /* Parameter signature */ - "glSpriteParameteriSGIX\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char RequestResidentProgramsNV_names[] = - "ip\0" /* Parameter signature */ - "glRequestResidentProgramsNV\0" - ""; -#endif - -#if defined(need_GL_SGIX_tag_sample_buffer) -static const char TagSampleBufferSGIX_names[] = - "\0" /* Parameter signature */ - "glTagSampleBufferSGIX\0" - ""; -#endif - -#if defined(need_GL_SUN_triangle_list) -static const char ReplacementCodeusSUN_names[] = - "i\0" /* Parameter signature */ - "glReplacementCodeusSUN\0" - ""; -#endif - -#if defined(need_GL_SGIX_list_priority) -static const char ListParameterivSGIX_names[] = - "iip\0" /* Parameter signature */ - "glListParameterivSGIX\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_multi_draw_arrays) -static const char MultiDrawElementsEXT_names[] = - "ipipi\0" /* Parameter signature */ - "glMultiDrawElements\0" - "glMultiDrawElementsEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char Uniform1ivARB_names[] = - "iip\0" /* Parameter signature */ - "glUniform1iv\0" - "glUniform1ivARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos) -static const char WindowPos2sMESA_names[] = - "ii\0" /* Parameter signature */ - "glWindowPos2s\0" - "glWindowPos2sARB\0" - "glWindowPos2sMESA\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_blend) -static const char WeightusvARB_names[] = - "ip\0" /* Parameter signature */ - "glWeightusvARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_fog_coord) -static const char FogCoordPointerEXT_names[] = - "iip\0" /* Parameter signature */ - "glFogCoordPointer\0" - "glFogCoordPointerEXT\0" - ""; -#endif - -#if defined(need_GL_EXT_index_material) -static const char IndexMaterialEXT_names[] = - "ii\0" /* Parameter signature */ - "glIndexMaterialEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color) -static const char SecondaryColor3ubvEXT_names[] = - "p\0" /* Parameter signature */ - "glSecondaryColor3ubv\0" - "glSecondaryColor3ubvEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib4dvARB_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib4dv\0" - "glVertexAttrib4dvARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_shader) -static const char BindAttribLocationARB_names[] = - "iip\0" /* Parameter signature */ - "glBindAttribLocation\0" - "glBindAttribLocationARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord2dARB_names[] = - "idd\0" /* Parameter signature */ - "glMultiTexCoord2d\0" - "glMultiTexCoord2dARB\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char ExecuteProgramNV_names[] = - "iip\0" /* Parameter signature */ - "glExecuteProgramNV\0" - ""; -#endif - -#if defined(need_GL_SGIX_fragment_lighting) -static const char LightEnviSGIX_names[] = - "ii\0" /* Parameter signature */ - "glLightEnviSGIX\0" - ""; -#endif - -#if defined(need_GL_SUN_triangle_list) -static const char ReplacementCodeuiSUN_names[] = - "i\0" /* Parameter signature */ - "glReplacementCodeuiSUN\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttribPointerNV_names[] = - "iiiip\0" /* Parameter signature */ - "glVertexAttribPointerNV\0" - ""; -#endif - -#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object) -static const char GetFramebufferAttachmentParameterivEXT_names[] = - "iiip\0" /* Parameter signature */ - "glGetFramebufferAttachmentParameteriv\0" - "glGetFramebufferAttachmentParameterivEXT\0" - ""; -#endif - -#if defined(need_GL_EXT_pixel_transform) -static const char PixelTransformParameterfEXT_names[] = - "iif\0" /* Parameter signature */ - "glPixelTransformParameterfEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord4dvARB_names[] = - "ip\0" /* Parameter signature */ - "glMultiTexCoord4dv\0" - "glMultiTexCoord4dvARB\0" - ""; -#endif - -#if defined(need_GL_EXT_pixel_transform) -static const char PixelTransformParameteriEXT_names[] = - "iii\0" /* Parameter signature */ - "glPixelTransformParameteriEXT\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char TexCoord2fColor4ubVertex3fSUN_names[] = - "ffiiiifff\0" /* Parameter signature */ - "glTexCoord2fColor4ubVertex3fSUN\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char Uniform1iARB_names[] = - "ii\0" /* Parameter signature */ - "glUniform1i\0" - "glUniform1iARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttribPointerARB_names[] = - "iiiiip\0" /* Parameter signature */ - "glVertexAttribPointer\0" - "glVertexAttribPointerARB\0" - ""; -#endif - -#if defined(need_GL_SGIS_sharpen_texture) -static const char SharpenTexFuncSGIS_names[] = - "iip\0" /* Parameter signature */ - "glSharpenTexFuncSGIS\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord4fvARB_names[] = - "ip\0" /* Parameter signature */ - "glMultiTexCoord4fv\0" - "glMultiTexCoord4fvARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_1) -static const char UniformMatrix2x3fv_names[] = - "iiip\0" /* Parameter signature */ - "glUniformMatrix2x3fv\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char TrackMatrixNV_names[] = - "iiii\0" /* Parameter signature */ - "glTrackMatrixNV\0" - ""; -#endif - -#if defined(need_GL_NV_register_combiners) -static const char CombinerParameteriNV_names[] = - "ii\0" /* Parameter signature */ - "glCombinerParameteriNV\0" - ""; -#endif - -#if defined(need_GL_SGIX_async) -static const char DeleteAsyncMarkersSGIX_names[] = - "ii\0" /* Parameter signature */ - "glDeleteAsyncMarkersSGIX\0" - ""; -#endif - -#if defined(need_GL_SGIX_async) -static const char IsAsyncMarkerSGIX_names[] = - "i\0" /* Parameter signature */ - "glIsAsyncMarkerSGIX\0" - ""; -#endif - -#if defined(need_GL_SGIX_framezoom) -static const char FrameZoomSGIX_names[] = - "i\0" /* Parameter signature */ - "glFrameZoomSGIX\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char Normal3fVertex3fvSUN_names[] = - "pp\0" /* Parameter signature */ - "glNormal3fVertex3fvSUN\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib4NsvARB_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib4Nsv\0" - "glVertexAttrib4NsvARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib3fvARB_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib3fv\0" - "glVertexAttrib3fvARB\0" - ""; -#endif - -#if defined(need_GL_ARB_sync) -static const char GetSynciv_names[] = - "iiipp\0" /* Parameter signature */ - "glGetSynciv\0" - ""; -#endif - -#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object) -static const char DeleteFramebuffersEXT_names[] = - "ip\0" /* Parameter signature */ - "glDeleteFramebuffers\0" - "glDeleteFramebuffersEXT\0" - ""; -#endif - -#if defined(need_GL_SUN_global_alpha) -static const char GlobalAlphaFactorsSUN_names[] = - "i\0" /* Parameter signature */ - "glGlobalAlphaFactorsSUN\0" - ""; -#endif - -#if defined(need_GL_EXT_texture3D) -static const char TexSubImage3D_names[] = - "iiiiiiiiiip\0" /* Parameter signature */ - "glTexSubImage3D\0" - "glTexSubImage3DEXT\0" - ""; -#endif - -#if defined(need_GL_EXT_coordinate_frame) -static const char Tangent3fEXT_names[] = - "fff\0" /* Parameter signature */ - "glTangent3fEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color) -static const char SecondaryColor3uivEXT_names[] = - "p\0" /* Parameter signature */ - "glSecondaryColor3uiv\0" - "glSecondaryColor3uivEXT\0" - ""; -#endif - -#if defined(need_GL_ARB_matrix_palette) -static const char MatrixIndexubvARB_names[] = - "ip\0" /* Parameter signature */ - "glMatrixIndexubvARB\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char Color4fNormal3fVertex3fSUN_names[] = - "ffffffffff\0" /* Parameter signature */ - "glColor4fNormal3fVertex3fSUN\0" - ""; -#endif - -#if defined(need_GL_SGIS_pixel_texture) -static const char PixelTexGenParameterfSGIS_names[] = - "if\0" /* Parameter signature */ - "glPixelTexGenParameterfSGIS\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) -static const char CreateShader_names[] = - "i\0" /* Parameter signature */ - "glCreateShader\0" - ""; -#endif - -#if defined(need_GL_SGI_color_table) || defined(need_GL_EXT_paletted_texture) -static const char GetColorTableParameterfv_names[] = - "iip\0" /* Parameter signature */ - "glGetColorTableParameterfv\0" - "glGetColorTableParameterfvSGI\0" - "glGetColorTableParameterfvEXT\0" - ""; -#endif - -#if defined(need_GL_SGIX_fragment_lighting) -static const char FragmentLightModelfvSGIX_names[] = - "ip\0" /* Parameter signature */ - "glFragmentLightModelfvSGIX\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord3fARB_names[] = - "ifff\0" /* Parameter signature */ - "glMultiTexCoord3f\0" - "glMultiTexCoord3fARB\0" - ""; -#endif - -#if defined(need_GL_SGIS_pixel_texture) -static const char GetPixelTexGenParameterfvSGIS_names[] = - "ip\0" /* Parameter signature */ - "glGetPixelTexGenParameterfvSGIS\0" - ""; -#endif - -#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object) -static const char GenFramebuffersEXT_names[] = - "ip\0" /* Parameter signature */ - "glGenFramebuffers\0" - "glGenFramebuffersEXT\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char GetProgramParameterdvNV_names[] = - "iiip\0" /* Parameter signature */ - "glGetProgramParameterdvNV\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_array_object) || defined(need_GL_APPLE_vertex_array_object) -static const char IsVertexArrayAPPLE_names[] = - "i\0" /* Parameter signature */ - "glIsVertexArray\0" - "glIsVertexArrayAPPLE\0" - ""; -#endif - -#if defined(need_GL_SGIX_fragment_lighting) -static const char FragmentLightfvSGIX_names[] = - "iip\0" /* Parameter signature */ - "glFragmentLightfvSGIX\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) -static const char DetachShader_names[] = - "ii\0" /* Parameter signature */ - "glDetachShader\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib4NubARB_names[] = - "iiiii\0" /* Parameter signature */ - "glVertexAttrib4Nub\0" - "glVertexAttrib4NubARB\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_program) -static const char GetProgramEnvParameterfvARB_names[] = - "iip\0" /* Parameter signature */ - "glGetProgramEnvParameterfvARB\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char GetTrackMatrixivNV_names[] = - "iiip\0" /* Parameter signature */ - "glGetTrackMatrixivNV\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttrib3svNV_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib3svNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char Uniform4fvARB_names[] = - "iip\0" /* Parameter signature */ - "glUniform4fv\0" - "glUniform4fvARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) || defined(need_GL_ARB_transpose_matrix) -static const char MultTransposeMatrixfARB_names[] = - "p\0" /* Parameter signature */ - "glMultTransposeMatrixf\0" - "glMultTransposeMatrixfARB\0" - ""; -#endif - -#if defined(need_GL_ATI_fragment_shader) -static const char ColorFragmentOp1ATI_names[] = - "iiiiiii\0" /* Parameter signature */ - "glColorFragmentOp1ATI\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char GetUniformfvARB_names[] = - "iip\0" /* Parameter signature */ - "glGetUniformfv\0" - "glGetUniformfvARB\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char ReplacementCodeuiTexCoord2fColor4fNormal3fVertex3fSUN_names[] = - "iffffffffffff\0" /* Parameter signature */ - "glReplacementCodeuiTexCoord2fColor4fNormal3fVertex3fSUN\0" - ""; -#endif - -#if defined(need_GL_ARB_shader_objects) -static const char DetachObjectARB_names[] = - "ii\0" /* Parameter signature */ - "glDetachObjectARB\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_blend) -static const char VertexBlendARB_names[] = - "i\0" /* Parameter signature */ - "glVertexBlendARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos) -static const char WindowPos3iMESA_names[] = - "iii\0" /* Parameter signature */ - "glWindowPos3i\0" - "glWindowPos3iARB\0" - "glWindowPos3iMESA\0" - ""; -#endif - -#if defined(need_GL_EXT_convolution) -static const char SeparableFilter2D_names[] = - "iiiiiipp\0" /* Parameter signature */ - "glSeparableFilter2D\0" - "glSeparableFilter2DEXT\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char ReplacementCodeuiColor4ubVertex3fvSUN_names[] = - "ppp\0" /* Parameter signature */ - "glReplacementCodeuiColor4ubVertex3fvSUN\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) || defined(need_GL_ARB_texture_compression) -static const char CompressedTexImage2DARB_names[] = - "iiiiiiip\0" /* Parameter signature */ - "glCompressedTexImage2D\0" - "glCompressedTexImage2DARB\0" - ""; -#endif - -#if defined(need_GL_EXT_vertex_array) -static const char ArrayElement_names[] = - "i\0" /* Parameter signature */ - "glArrayElement\0" - "glArrayElementEXT\0" - ""; -#endif - -#if defined(need_GL_EXT_depth_bounds_test) -static const char DepthBoundsEXT_names[] = - "dd\0" /* Parameter signature */ - "glDepthBoundsEXT\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char ProgramParameters4fvNV_names[] = - "iiip\0" /* Parameter signature */ - "glProgramParameters4fvNV\0" - ""; -#endif - -#if defined(need_GL_SGIX_polynomial_ffd) -static const char DeformationMap3fSGIX_names[] = - "iffiiffiiffiip\0" /* Parameter signature */ - "glDeformationMap3fSGIX\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char GetProgramivNV_names[] = - "iip\0" /* Parameter signature */ - "glGetProgramivNV\0" - ""; -#endif - -#if defined(need_GL_EXT_histogram) -static const char GetMinmaxParameteriv_names[] = - "iip\0" /* Parameter signature */ - "glGetMinmaxParameteriv\0" - "glGetMinmaxParameterivEXT\0" - ""; -#endif - -#if defined(need_GL_EXT_copy_texture) -static const char CopyTexImage1D_names[] = - "iiiiiii\0" /* Parameter signature */ - "glCopyTexImage1D\0" - "glCopyTexImage1DEXT\0" - ""; -#endif - -#if defined(need_GL_ATI_fragment_shader) -static const char AlphaFragmentOp3ATI_names[] = - "iiiiiiiiiiii\0" /* Parameter signature */ - "glAlphaFragmentOp3ATI\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char GetVertexAttribdvNV_names[] = - "iip\0" /* Parameter signature */ - "glGetVertexAttribdvNV\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttrib3fvNV_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib3fvNV\0" - ""; -#endif - -#if defined(need_GL_NV_register_combiners) -static const char GetFinalCombinerInputParameterivNV_names[] = - "iip\0" /* Parameter signature */ - "glGetFinalCombinerInputParameterivNV\0" - ""; -#endif - -#if defined(need_GL_NV_evaluators) -static const char GetMapParameterivNV_names[] = - "iip\0" /* Parameter signature */ - "glGetMapParameterivNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char Uniform4iARB_names[] = - "iiiii\0" /* Parameter signature */ - "glUniform4i\0" - "glUniform4iARB\0" - ""; -#endif - -#if defined(need_GL_EXT_convolution) -static const char ConvolutionParameteri_names[] = - "iii\0" /* Parameter signature */ - "glConvolutionParameteri\0" - "glConvolutionParameteriEXT\0" - ""; -#endif - -#if defined(need_GL_EXT_coordinate_frame) -static const char Binormal3sEXT_names[] = - "iii\0" /* Parameter signature */ - "glBinormal3sEXT\0" - ""; -#endif - -#if defined(need_GL_EXT_convolution) -static const char ConvolutionParameterf_names[] = - "iif\0" /* Parameter signature */ - "glConvolutionParameterf\0" - "glConvolutionParameterfEXT\0" - ""; -#endif - -#if defined(need_GL_SGI_color_table) || defined(need_GL_EXT_paletted_texture) -static const char GetColorTableParameteriv_names[] = - "iip\0" /* Parameter signature */ - "glGetColorTableParameteriv\0" - "glGetColorTableParameterivSGI\0" - "glGetColorTableParameterivEXT\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_program) || defined(need_GL_NV_vertex_program) -static const char ProgramEnvParameter4dvARB_names[] = - "iip\0" /* Parameter signature */ - "glProgramEnvParameter4dvARB\0" - "glProgramParameter4dvNV\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttribs2fvNV_names[] = - "iip\0" /* Parameter signature */ - "glVertexAttribs2fvNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char UseProgramObjectARB_names[] = - "i\0" /* Parameter signature */ - "glUseProgram\0" - "glUseProgramObjectARB\0" - ""; -#endif - -#if defined(need_GL_NV_evaluators) -static const char GetMapParameterfvNV_names[] = - "iip\0" /* Parameter signature */ - "glGetMapParameterfvNV\0" - ""; -#endif - -#if defined(need_GL_ATI_fragment_shader) -static const char PassTexCoordATI_names[] = - "iii\0" /* Parameter signature */ - "glPassTexCoordATI\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) -static const char DeleteProgram_names[] = - "i\0" /* Parameter signature */ - "glDeleteProgram\0" - ""; -#endif - -#if defined(need_GL_EXT_coordinate_frame) -static const char Tangent3ivEXT_names[] = - "p\0" /* Parameter signature */ - "glTangent3ivEXT\0" - ""; -#endif - -#if defined(need_GL_EXT_coordinate_frame) -static const char Tangent3dEXT_names[] = - "ddd\0" /* Parameter signature */ - "glTangent3dEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color) -static const char SecondaryColor3dvEXT_names[] = - "p\0" /* Parameter signature */ - "glSecondaryColor3dv\0" - "glSecondaryColor3dvEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_multi_draw_arrays) -static const char MultiDrawArraysEXT_names[] = - "ippi\0" /* Parameter signature */ - "glMultiDrawArrays\0" - "glMultiDrawArraysEXT\0" - ""; -#endif - -#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object) -static const char BindRenderbufferEXT_names[] = - "ii\0" /* Parameter signature */ - "glBindRenderbuffer\0" - "glBindRenderbufferEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord4dARB_names[] = - "idddd\0" /* Parameter signature */ - "glMultiTexCoord4d\0" - "glMultiTexCoord4dARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color) -static const char SecondaryColor3usEXT_names[] = - "iii\0" /* Parameter signature */ - "glSecondaryColor3us\0" - "glSecondaryColor3usEXT\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_program) -static const char ProgramLocalParameter4fvARB_names[] = - "iip\0" /* Parameter signature */ - "glProgramLocalParameter4fvARB\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_program) || defined(need_GL_NV_vertex_program) -static const char DeleteProgramsNV_names[] = - "ip\0" /* Parameter signature */ - "glDeleteProgramsARB\0" - "glDeleteProgramsNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord1sARB_names[] = - "ii\0" /* Parameter signature */ - "glMultiTexCoord1s\0" - "glMultiTexCoord1sARB\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char ReplacementCodeuiColor3fVertex3fSUN_names[] = - "iffffff\0" /* Parameter signature */ - "glReplacementCodeuiColor3fVertex3fSUN\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) || defined(need_GL_NV_vertex_program) -static const char GetVertexAttribPointervNV_names[] = - "iip\0" /* Parameter signature */ - "glGetVertexAttribPointerv\0" - "glGetVertexAttribPointervARB\0" - "glGetVertexAttribPointervNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord1dvARB_names[] = - "ip\0" /* Parameter signature */ - "glMultiTexCoord1dv\0" - "glMultiTexCoord1dvARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char Uniform2iARB_names[] = - "iii\0" /* Parameter signature */ - "glUniform2i\0" - "glUniform2iARB\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char GetProgramStringNV_names[] = - "iip\0" /* Parameter signature */ - "glGetProgramStringNV\0" - ""; -#endif - -#if defined(need_GL_EXT_vertex_array) -static const char ColorPointerEXT_names[] = - "iiiip\0" /* Parameter signature */ - "glColorPointerEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_vertex_buffer_object) -static const char MapBufferARB_names[] = - "ii\0" /* Parameter signature */ - "glMapBuffer\0" - "glMapBufferARB\0" - ""; -#endif - -#if defined(need_GL_ARB_draw_elements_base_vertex) -static const char MultiDrawElementsBaseVertex_names[] = - "ipipip\0" /* Parameter signature */ - "glMultiDrawElementsBaseVertex\0" - ""; -#endif - -#if defined(need_GL_EXT_coordinate_frame) -static const char Binormal3svEXT_names[] = - "p\0" /* Parameter signature */ - "glBinormal3svEXT\0" - ""; -#endif - -#if defined(need_GL_EXT_light_texture) -static const char ApplyTextureEXT_names[] = - "i\0" /* Parameter signature */ - "glApplyTextureEXT\0" - ""; -#endif - -#if defined(need_GL_EXT_light_texture) -static const char TextureMaterialEXT_names[] = - "ii\0" /* Parameter signature */ - "glTextureMaterialEXT\0" - ""; -#endif - -#if defined(need_GL_EXT_light_texture) -static const char TextureLightEXT_names[] = - "i\0" /* Parameter signature */ - "glTextureLightEXT\0" - ""; -#endif - -#if defined(need_GL_EXT_histogram) -static const char ResetMinmax_names[] = - "i\0" /* Parameter signature */ - "glResetMinmax\0" - "glResetMinmaxEXT\0" - ""; -#endif - -#if defined(need_GL_SGIX_sprite) -static const char SpriteParameterfSGIX_names[] = - "if\0" /* Parameter signature */ - "glSpriteParameterfSGIX\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttrib4sNV_names[] = - "iiiii\0" /* Parameter signature */ - "glVertexAttrib4sNV\0" - ""; -#endif - -#if defined(need_GL_EXT_convolution) -static const char GetConvolutionParameterfv_names[] = - "iip\0" /* Parameter signature */ - "glGetConvolutionParameterfv\0" - "glGetConvolutionParameterfvEXT\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttribs4dvNV_names[] = - "iip\0" /* Parameter signature */ - "glVertexAttribs4dvNV\0" - ""; -#endif - -#if defined(need_GL_IBM_multimode_draw_arrays) -static const char MultiModeDrawArraysIBM_names[] = - "pppii\0" /* Parameter signature */ - "glMultiModeDrawArraysIBM\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib4dARB_names[] = - "idddd\0" /* Parameter signature */ - "glVertexAttrib4d\0" - "glVertexAttrib4dARB\0" - ""; -#endif - -#if defined(need_GL_ATI_envmap_bumpmap) -static const char GetTexBumpParameterfvATI_names[] = - "ip\0" /* Parameter signature */ - "glGetTexBumpParameterfvATI\0" - ""; -#endif - -#if defined(need_GL_NV_fragment_program) -static const char ProgramNamedParameter4dNV_names[] = - "iipdddd\0" /* Parameter signature */ - "glProgramNamedParameter4dNV\0" - ""; -#endif - -#if defined(need_GL_EXT_vertex_weighting) -static const char VertexWeightfEXT_names[] = - "f\0" /* Parameter signature */ - "glVertexWeightfEXT\0" - ""; -#endif - -#if defined(need_GL_EXT_coordinate_frame) -static const char Binormal3fEXT_names[] = - "fff\0" /* Parameter signature */ - "glBinormal3fEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_fog_coord) -static const char FogCoordfvEXT_names[] = - "p\0" /* Parameter signature */ - "glFogCoordfv\0" - "glFogCoordfvEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord1ivARB_names[] = - "ip\0" /* Parameter signature */ - "glMultiTexCoord1iv\0" - "glMultiTexCoord1ivARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color) -static const char SecondaryColor3ubEXT_names[] = - "iii\0" /* Parameter signature */ - "glSecondaryColor3ub\0" - "glSecondaryColor3ubEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord2ivARB_names[] = - "ip\0" /* Parameter signature */ - "glMultiTexCoord2iv\0" - "glMultiTexCoord2ivARB\0" - ""; -#endif - -#if defined(need_GL_SGIS_fog_function) -static const char FogFuncSGIS_names[] = - "ip\0" /* Parameter signature */ - "glFogFuncSGIS\0" - ""; -#endif - -#if defined(need_GL_EXT_copy_texture) -static const char CopyTexSubImage2D_names[] = - "iiiiiiii\0" /* Parameter signature */ - "glCopyTexSubImage2D\0" - "glCopyTexSubImage2DEXT\0" - ""; -#endif - -#if defined(need_GL_ARB_shader_objects) -static const char GetObjectParameterivARB_names[] = - "iip\0" /* Parameter signature */ - "glGetObjectParameterivARB\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char TexCoord4fVertex4fSUN_names[] = - "ffffffff\0" /* Parameter signature */ - "glTexCoord4fVertex4fSUN\0" - ""; -#endif - -#if defined(need_GL_APPLE_vertex_array_object) -static const char BindVertexArrayAPPLE_names[] = - "i\0" /* Parameter signature */ - "glBindVertexArrayAPPLE\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_program) -static const char GetProgramLocalParameterdvARB_names[] = - "iip\0" /* Parameter signature */ - "glGetProgramLocalParameterdvARB\0" - ""; -#endif - -#if defined(need_GL_EXT_histogram) -static const char GetHistogramParameteriv_names[] = - "iip\0" /* Parameter signature */ - "glGetHistogramParameteriv\0" - "glGetHistogramParameterivEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord1iARB_names[] = - "ii\0" /* Parameter signature */ - "glMultiTexCoord1i\0" - "glMultiTexCoord1iARB\0" - ""; -#endif - -#if defined(need_GL_EXT_convolution) -static const char GetConvolutionFilter_names[] = - "iiip\0" /* Parameter signature */ - "glGetConvolutionFilter\0" - "glGetConvolutionFilterEXT\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_program) -static const char GetProgramivARB_names[] = - "iip\0" /* Parameter signature */ - "glGetProgramivARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_blend_func_separate) || defined(need_GL_INGR_blend_func_separate) -static const char BlendFuncSeparateEXT_names[] = - "iiii\0" /* Parameter signature */ - "glBlendFuncSeparate\0" - "glBlendFuncSeparateEXT\0" - "glBlendFuncSeparateINGR\0" - ""; -#endif - -#if defined(need_GL_ARB_map_buffer_range) -static const char MapBufferRange_names[] = - "iiii\0" /* Parameter signature */ - "glMapBufferRange\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char ProgramParameters4dvNV_names[] = - "iiip\0" /* Parameter signature */ - "glProgramParameters4dvNV\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char TexCoord2fColor3fVertex3fvSUN_names[] = - "ppp\0" /* Parameter signature */ - "glTexCoord2fColor3fVertex3fvSUN\0" - ""; -#endif - -#if defined(need_GL_EXT_coordinate_frame) -static const char Binormal3dvEXT_names[] = - "p\0" /* Parameter signature */ - "glBinormal3dvEXT\0" - ""; -#endif - -#if defined(need_GL_NV_fence) -static const char FinishFenceNV_names[] = - "i\0" /* Parameter signature */ - "glFinishFenceNV\0" - ""; -#endif - -#if defined(need_GL_SGIS_fog_function) -static const char GetFogFuncSGIS_names[] = - "p\0" /* Parameter signature */ - "glGetFogFuncSGIS\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char GetUniformLocationARB_names[] = - "ip\0" /* Parameter signature */ - "glGetUniformLocation\0" - "glGetUniformLocationARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color) -static const char SecondaryColor3fEXT_names[] = - "fff\0" /* Parameter signature */ - "glSecondaryColor3f\0" - "glSecondaryColor3fEXT\0" - ""; -#endif - -#if defined(need_GL_NV_register_combiners) -static const char CombinerInputNV_names[] = - "iiiiii\0" /* Parameter signature */ - "glCombinerInputNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib3sARB_names[] = - "iiii\0" /* Parameter signature */ - "glVertexAttrib3s\0" - "glVertexAttrib3sARB\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char ReplacementCodeuiNormal3fVertex3fvSUN_names[] = - "ppp\0" /* Parameter signature */ - "glReplacementCodeuiNormal3fVertex3fvSUN\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_program) -static const char ProgramStringARB_names[] = - "iiip\0" /* Parameter signature */ - "glProgramStringARB\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char TexCoord4fVertex4fvSUN_names[] = - "pp\0" /* Parameter signature */ - "glTexCoord4fVertex4fvSUN\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttrib3sNV_names[] = - "iiii\0" /* Parameter signature */ - "glVertexAttrib3sNV\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttrib1fNV_names[] = - "if\0" /* Parameter signature */ - "glVertexAttrib1fNV\0" - ""; -#endif - -#if defined(need_GL_SGIX_fragment_lighting) -static const char FragmentLightfSGIX_names[] = - "iif\0" /* Parameter signature */ - "glFragmentLightfSGIX\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) || defined(need_GL_ARB_texture_compression) -static const char GetCompressedTexImageARB_names[] = - "iip\0" /* Parameter signature */ - "glGetCompressedTexImage\0" - "glGetCompressedTexImageARB\0" - ""; -#endif - -#if defined(need_GL_EXT_vertex_weighting) -static const char VertexWeightPointerEXT_names[] = - "iiip\0" /* Parameter signature */ - "glVertexWeightPointerEXT\0" - ""; -#endif - -#if defined(need_GL_EXT_histogram) -static const char GetHistogram_names[] = - "iiiip\0" /* Parameter signature */ - "glGetHistogram\0" - "glGetHistogramEXT\0" - ""; -#endif - -#if defined(need_GL_EXT_stencil_two_side) -static const char ActiveStencilFaceEXT_names[] = - "i\0" /* Parameter signature */ - "glActiveStencilFaceEXT\0" - ""; -#endif - -#if defined(need_GL_ATI_separate_stencil) -static const char StencilFuncSeparateATI_names[] = - "iiii\0" /* Parameter signature */ - "glStencilFuncSeparateATI\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char GetShaderSourceARB_names[] = - "iipp\0" /* Parameter signature */ - "glGetShaderSource\0" - "glGetShaderSourceARB\0" - ""; -#endif - -#if defined(need_GL_SGIX_igloo_interface) -static const char IglooInterfaceSGIX_names[] = - "ip\0" /* Parameter signature */ - "glIglooInterfaceSGIX\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttrib4dNV_names[] = - "idddd\0" /* Parameter signature */ - "glVertexAttrib4dNV\0" - ""; -#endif - -#if defined(need_GL_IBM_multimode_draw_arrays) -static const char MultiModeDrawElementsIBM_names[] = - "ppipii\0" /* Parameter signature */ - "glMultiModeDrawElementsIBM\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord4svARB_names[] = - "ip\0" /* Parameter signature */ - "glMultiTexCoord4sv\0" - "glMultiTexCoord4svARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_occlusion_query) -static const char GenQueriesARB_names[] = - "ip\0" /* Parameter signature */ - "glGenQueries\0" - "glGenQueriesARB\0" - ""; -#endif - -#if defined(need_GL_SUN_vertex) -static const char ReplacementCodeuiVertex3fSUN_names[] = - "ifff\0" /* Parameter signature */ - "glReplacementCodeuiVertex3fSUN\0" - ""; -#endif - -#if defined(need_GL_EXT_coordinate_frame) -static const char Tangent3iEXT_names[] = - "iii\0" /* Parameter signature */ - "glTangent3iEXT\0" - ""; -#endif - -#if defined(need_GL_SUN_mesh_array) -static const char DrawMeshArraysSUN_names[] = - "iiii\0" /* Parameter signature */ - "glDrawMeshArraysSUN\0" - ""; -#endif - -#if defined(need_GL_ARB_sync) -static const char IsSync_names[] = - "i\0" /* Parameter signature */ - "glIsSync\0" - ""; -#endif - -#if defined(need_GL_NV_evaluators) -static const char GetMapControlPointsNV_names[] = - "iiiiiip\0" /* Parameter signature */ - "glGetMapControlPointsNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_draw_buffers) || defined(need_GL_ATI_draw_buffers) -static const char DrawBuffersARB_names[] = - "ip\0" /* Parameter signature */ - "glDrawBuffers\0" - "glDrawBuffersARB\0" - "glDrawBuffersATI\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_program) -static const char ProgramLocalParameter4fARB_names[] = - "iiffff\0" /* Parameter signature */ - "glProgramLocalParameter4fARB\0" - ""; -#endif - -#if defined(need_GL_SGIX_sprite) -static const char SpriteParameterivSGIX_names[] = - "ip\0" /* Parameter signature */ - "glSpriteParameterivSGIX\0" - ""; -#endif - -#if defined(need_GL_EXT_provoking_vertex) -static const char ProvokingVertexEXT_names[] = - "i\0" /* Parameter signature */ - "glProvokingVertexEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord1fARB_names[] = - "if\0" /* Parameter signature */ - "glMultiTexCoord1f\0" - "glMultiTexCoord1fARB\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttribs4ubvNV_names[] = - "iip\0" /* Parameter signature */ - "glVertexAttribs4ubvNV\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_blend) -static const char WeightsvARB_names[] = - "ip\0" /* Parameter signature */ - "glWeightsvARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) -static const char Uniform1fvARB_names[] = - "iip\0" /* Parameter signature */ - "glUniform1fv\0" - "glUniform1fvARB\0" - ""; -#endif - -#if defined(need_GL_EXT_copy_texture) -static const char CopyTexSubImage1D_names[] = - "iiiiii\0" /* Parameter signature */ - "glCopyTexSubImage1D\0" - "glCopyTexSubImage1DEXT\0" - ""; -#endif - -#if defined(need_GL_EXT_texture_object) -static const char BindTexture_names[] = - "ii\0" /* Parameter signature */ - "glBindTexture\0" - "glBindTextureEXT\0" - ""; -#endif - -#if defined(need_GL_ATI_fragment_shader) -static const char BeginFragmentShaderATI_names[] = - "\0" /* Parameter signature */ - "glBeginFragmentShaderATI\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord4fARB_names[] = - "iffff\0" /* Parameter signature */ - "glMultiTexCoord4f\0" - "glMultiTexCoord4fARB\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttribs3svNV_names[] = - "iip\0" /* Parameter signature */ - "glVertexAttribs3svNV\0" - ""; -#endif - -#if defined(need_GL_SUN_triangle_list) -static const char ReplacementCodeuivSUN_names[] = - "p\0" /* Parameter signature */ - "glReplacementCodeuivSUN\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char EnableVertexAttribArrayARB_names[] = - "i\0" /* Parameter signature */ - "glEnableVertexAttribArray\0" - "glEnableVertexAttribArrayARB\0" - ""; -#endif - -#if defined(need_GL_INTEL_parallel_arrays) -static const char NormalPointervINTEL_names[] = - "ip\0" /* Parameter signature */ - "glNormalPointervINTEL\0" - ""; -#endif - -#if defined(need_GL_EXT_convolution) -static const char CopyConvolutionFilter2D_names[] = - "iiiiii\0" /* Parameter signature */ - "glCopyConvolutionFilter2D\0" - "glCopyConvolutionFilter2DEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos) -static const char WindowPos3ivMESA_names[] = - "p\0" /* Parameter signature */ - "glWindowPos3iv\0" - "glWindowPos3ivARB\0" - "glWindowPos3ivMESA\0" - ""; -#endif - -#if defined(need_GL_ARB_copy_buffer) -static const char CopyBufferSubData_names[] = - "iiiii\0" /* Parameter signature */ - "glCopyBufferSubData\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_vertex_buffer_object) -static const char IsBufferARB_names[] = - "i\0" /* Parameter signature */ - "glIsBuffer\0" - "glIsBufferARB\0" - ""; -#endif - -#if defined(need_GL_MESA_window_pos) -static const char WindowPos4iMESA_names[] = - "iiii\0" /* Parameter signature */ - "glWindowPos4iMESA\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib4uivARB_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib4uiv\0" - "glVertexAttrib4uivARB\0" - ""; -#endif - -#if defined(need_GL_EXT_coordinate_frame) -static const char Tangent3bvEXT_names[] = - "p\0" /* Parameter signature */ - "glTangent3bvEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_1) -static const char UniformMatrix3x4fv_names[] = - "iiip\0" /* Parameter signature */ - "glUniformMatrix3x4fv\0" - ""; -#endif - -#if defined(need_GL_ARB_draw_elements_base_vertex) -static const char DrawRangeElementsBaseVertex_names[] = - "iiiiipi\0" /* Parameter signature */ - "glDrawRangeElementsBaseVertex\0" - ""; -#endif - -#if defined(need_GL_INTEL_parallel_arrays) -static const char TexCoordPointervINTEL_names[] = - "iip\0" /* Parameter signature */ - "glTexCoordPointervINTEL\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_vertex_buffer_object) -static const char DeleteBuffersARB_names[] = - "ip\0" /* Parameter signature */ - "glDeleteBuffers\0" - "glDeleteBuffersARB\0" - ""; -#endif - -#if defined(need_GL_MESA_window_pos) -static const char WindowPos4fvMESA_names[] = - "p\0" /* Parameter signature */ - "glWindowPos4fvMESA\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttrib1sNV_names[] = - "ii\0" /* Parameter signature */ - "glVertexAttrib1sNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_secondary_color) -static const char SecondaryColor3svEXT_names[] = - "p\0" /* Parameter signature */ - "glSecondaryColor3sv\0" - "glSecondaryColor3svEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) || defined(need_GL_ARB_transpose_matrix) -static const char LoadTransposeMatrixfARB_names[] = - "p\0" /* Parameter signature */ - "glLoadTransposeMatrixf\0" - "glLoadTransposeMatrixfARB\0" - ""; -#endif - -#if defined(need_GL_EXT_vertex_array) -static const char GetPointerv_names[] = - "ip\0" /* Parameter signature */ - "glGetPointerv\0" - "glGetPointervEXT\0" - ""; -#endif - -#if defined(need_GL_EXT_coordinate_frame) -static const char Tangent3bEXT_names[] = - "iii\0" /* Parameter signature */ - "glTangent3bEXT\0" - ""; -#endif - -#if defined(need_GL_NV_register_combiners) -static const char CombinerParameterfNV_names[] = - "if\0" /* Parameter signature */ - "glCombinerParameterfNV\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_program) || defined(need_GL_NV_vertex_program) -static const char BindProgramNV_names[] = - "ii\0" /* Parameter signature */ - "glBindProgramARB\0" - "glBindProgramNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib4svARB_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib4sv\0" - "glVertexAttrib4svARB\0" - ""; -#endif - -#if defined(need_GL_MESA_shader_debug) -static const char CreateDebugObjectMESA_names[] = - "\0" /* Parameter signature */ - "glCreateDebugObjectMESA\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) -static const char GetShaderiv_names[] = - "iip\0" /* Parameter signature */ - "glGetShaderiv\0" - ""; -#endif - -#if defined(need_GL_ARB_sync) -static const char ClientWaitSync_names[] = - "iii\0" /* Parameter signature */ - "glClientWaitSync\0" - ""; -#endif - -#if defined(need_GL_ATI_fragment_shader) -static const char BindFragmentShaderATI_names[] = - "i\0" /* Parameter signature */ - "glBindFragmentShaderATI\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_vertex_buffer_object) -static const char UnmapBufferARB_names[] = - "i\0" /* Parameter signature */ - "glUnmapBuffer\0" - "glUnmapBufferARB\0" - ""; -#endif - -#if defined(need_GL_EXT_histogram) -static const char Minmax_names[] = - "iii\0" /* Parameter signature */ - "glMinmax\0" - "glMinmaxEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_EXT_fog_coord) -static const char FogCoorddvEXT_names[] = - "p\0" /* Parameter signature */ - "glFogCoorddv\0" - "glFogCoorddvEXT\0" - ""; -#endif - -#if defined(need_GL_SUNX_constant_data) -static const char FinishTextureSUNX_names[] = - "\0" /* Parameter signature */ - "glFinishTextureSUNX\0" - ""; -#endif - -#if defined(need_GL_SGIX_fragment_lighting) -static const char GetFragmentLightfvSGIX_names[] = - "iip\0" /* Parameter signature */ - "glGetFragmentLightfvSGIX\0" - ""; -#endif - -#if defined(need_GL_EXT_coordinate_frame) -static const char Binormal3fvEXT_names[] = - "p\0" /* Parameter signature */ - "glBinormal3fvEXT\0" - ""; -#endif - -#if defined(need_GL_ATI_fragment_shader) -static const char ColorFragmentOp3ATI_names[] = - "iiiiiiiiiiiii\0" /* Parameter signature */ - "glColorFragmentOp3ATI\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib2svARB_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib2sv\0" - "glVertexAttrib2svARB\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char AreProgramsResidentNV_names[] = - "ipp\0" /* Parameter signature */ - "glAreProgramsResidentNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos) -static const char WindowPos3svMESA_names[] = - "p\0" /* Parameter signature */ - "glWindowPos3sv\0" - "glWindowPos3svARB\0" - "glWindowPos3svMESA\0" - ""; -#endif - -#if defined(need_GL_EXT_color_subtable) -static const char CopyColorSubTable_names[] = - "iiiii\0" /* Parameter signature */ - "glCopyColorSubTable\0" - "glCopyColorSubTableEXT\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_blend) -static const char WeightdvARB_names[] = - "ip\0" /* Parameter signature */ - "glWeightdvARB\0" - ""; -#endif - -#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object) -static const char DeleteRenderbuffersEXT_names[] = - "ip\0" /* Parameter signature */ - "glDeleteRenderbuffers\0" - "glDeleteRenderbuffersEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib4NubvARB_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib4Nubv\0" - "glVertexAttrib4NubvARB\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttrib3dvNV_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib3dvNV\0" - ""; -#endif - -#if defined(need_GL_ARB_shader_objects) -static const char GetObjectParameterfvARB_names[] = - "iip\0" /* Parameter signature */ - "glGetObjectParameterfvARB\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_program) -static const char GetProgramEnvParameterdvARB_names[] = - "iip\0" /* Parameter signature */ - "glGetProgramEnvParameterdvARB\0" - ""; -#endif - -#if defined(need_GL_EXT_compiled_vertex_array) -static const char LockArraysEXT_names[] = - "ii\0" /* Parameter signature */ - "glLockArraysEXT\0" - ""; -#endif - -#if defined(need_GL_EXT_pixel_transform) -static const char PixelTransformParameterivEXT_names[] = - "iip\0" /* Parameter signature */ - "glPixelTransformParameterivEXT\0" - ""; -#endif - -#if defined(need_GL_EXT_coordinate_frame) -static const char BinormalPointerEXT_names[] = - "iip\0" /* Parameter signature */ - "glBinormalPointerEXT\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttrib1dNV_names[] = - "id\0" /* Parameter signature */ - "glVertexAttrib1dNV\0" - ""; -#endif - -#if defined(need_GL_NV_register_combiners) -static const char GetCombinerInputParameterivNV_names[] = - "iiiip\0" /* Parameter signature */ - "glGetCombinerInputParameterivNV\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_3) -static const char MultiTexCoord2fvARB_names[] = - "ip\0" /* Parameter signature */ - "glMultiTexCoord2fv\0" - "glMultiTexCoord2fvARB\0" - ""; -#endif - -#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object) -static const char GetRenderbufferParameterivEXT_names[] = - "iip\0" /* Parameter signature */ - "glGetRenderbufferParameteriv\0" - "glGetRenderbufferParameterivEXT\0" - ""; -#endif - -#if defined(need_GL_NV_register_combiners) -static const char CombinerParameterivNV_names[] = - "ip\0" /* Parameter signature */ - "glCombinerParameterivNV\0" - ""; -#endif - -#if defined(need_GL_ATI_fragment_shader) -static const char GenFragmentShadersATI_names[] = - "i\0" /* Parameter signature */ - "glGenFragmentShadersATI\0" - ""; -#endif - -#if defined(need_GL_EXT_vertex_array) -static const char DrawArrays_names[] = - "iii\0" /* Parameter signature */ - "glDrawArrays\0" - "glDrawArraysEXT\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_blend) -static const char WeightuivARB_names[] = - "ip\0" /* Parameter signature */ - "glWeightuivARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib2sARB_names[] = - "iii\0" /* Parameter signature */ - "glVertexAttrib2s\0" - "glVertexAttrib2sARB\0" - ""; -#endif - -#if defined(need_GL_SGIX_async) -static const char GenAsyncMarkersSGIX_names[] = - "i\0" /* Parameter signature */ - "glGenAsyncMarkersSGIX\0" - ""; -#endif - -#if defined(need_GL_EXT_coordinate_frame) -static const char Tangent3svEXT_names[] = - "p\0" /* Parameter signature */ - "glTangent3svEXT\0" - ""; -#endif - -#if defined(need_GL_SGIX_list_priority) -static const char GetListParameterivSGIX_names[] = - "iip\0" /* Parameter signature */ - "glGetListParameterivSGIX\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_vertex_buffer_object) -static const char BindBufferARB_names[] = - "ii\0" /* Parameter signature */ - "glBindBuffer\0" - "glBindBufferARB\0" - ""; -#endif - -#if defined(need_GL_ARB_shader_objects) -static const char GetInfoLogARB_names[] = - "iipp\0" /* Parameter signature */ - "glGetInfoLogARB\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttribs4svNV_names[] = - "iip\0" /* Parameter signature */ - "glVertexAttribs4svNV\0" - ""; -#endif - -#if defined(need_GL_IBM_vertex_array_lists) -static const char EdgeFlagPointerListIBM_names[] = - "ipi\0" /* Parameter signature */ - "glEdgeFlagPointerListIBM\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_1) -static const char UniformMatrix3x2fv_names[] = - "iiip\0" /* Parameter signature */ - "glUniformMatrix3x2fv\0" - ""; -#endif - -#if defined(need_GL_EXT_histogram) -static const char GetMinmaxParameterfv_names[] = - "iip\0" /* Parameter signature */ - "glGetMinmaxParameterfv\0" - "glGetMinmaxParameterfvEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) -static const char VertexAttrib1fvARB_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib1fv\0" - "glVertexAttrib1fvARB\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_5) || defined(need_GL_ARB_vertex_buffer_object) -static const char GenBuffersARB_names[] = - "ip\0" /* Parameter signature */ - "glGenBuffers\0" - "glGenBuffersARB\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttribs1svNV_names[] = - "iip\0" /* Parameter signature */ - "glVertexAttribs1svNV\0" - ""; -#endif - -#if defined(need_GL_ATI_envmap_bumpmap) -static const char GetTexBumpParameterivATI_names[] = - "ip\0" /* Parameter signature */ - "glGetTexBumpParameterivATI\0" - ""; -#endif - -#if defined(need_GL_EXT_coordinate_frame) -static const char Binormal3bEXT_names[] = - "iii\0" /* Parameter signature */ - "glBinormal3bEXT\0" - ""; -#endif - -#if defined(need_GL_SGIX_fragment_lighting) -static const char FragmentMaterialivSGIX_names[] = - "iip\0" /* Parameter signature */ - "glFragmentMaterialivSGIX\0" - ""; -#endif - -#if defined(need_GL_ARB_framebuffer_object) || defined(need_GL_EXT_framebuffer_object) -static const char IsRenderbufferEXT_names[] = - "i\0" /* Parameter signature */ - "glIsRenderbuffer\0" - "glIsRenderbufferEXT\0" - ""; -#endif - -#if defined(need_GL_ARB_vertex_program) || defined(need_GL_NV_vertex_program) -static const char GenProgramsNV_names[] = - "ip\0" /* Parameter signature */ - "glGenProgramsARB\0" - "glGenProgramsNV\0" - ""; -#endif - -#if defined(need_GL_NV_vertex_program) -static const char VertexAttrib4dvNV_names[] = - "ip\0" /* Parameter signature */ - "glVertexAttrib4dvNV\0" - ""; -#endif - -#if defined(need_GL_ATI_fragment_shader) -static const char EndFragmentShaderATI_names[] = - "\0" /* Parameter signature */ - "glEndFragmentShaderATI\0" - ""; -#endif - -#if defined(need_GL_EXT_coordinate_frame) -static const char Binormal3iEXT_names[] = - "iii\0" /* Parameter signature */ - "glBinormal3iEXT\0" - ""; -#endif - -#if defined(need_GL_VERSION_1_4) || defined(need_GL_ARB_window_pos) || defined(need_GL_MESA_window_pos) -static const char WindowPos2fMESA_names[] = - "ff\0" /* Parameter signature */ - "glWindowPos2f\0" - "glWindowPos2fARB\0" - "glWindowPos2fMESA\0" - ""; -#endif - -#if defined(need_GL_3DFX_tbuffer) -static const struct dri_extension_function GL_3DFX_tbuffer_functions[] = { - { TbufferMask3DFX_names, TbufferMask3DFX_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_APPLE_flush_buffer_range) -static const struct dri_extension_function GL_APPLE_flush_buffer_range_functions[] = { - { BufferParameteriAPPLE_names, BufferParameteriAPPLE_remap_index, -1 }, - { FlushMappedBufferRangeAPPLE_names, FlushMappedBufferRangeAPPLE_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_APPLE_texture_range) -static const struct dri_extension_function GL_APPLE_texture_range_functions[] = { - { TextureRangeAPPLE_names, TextureRangeAPPLE_remap_index, -1 }, - { GetTexParameterPointervAPPLE_names, GetTexParameterPointervAPPLE_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_APPLE_vertex_array_object) -static const struct dri_extension_function GL_APPLE_vertex_array_object_functions[] = { - { DeleteVertexArraysAPPLE_names, DeleteVertexArraysAPPLE_remap_index, -1 }, - { GenVertexArraysAPPLE_names, GenVertexArraysAPPLE_remap_index, -1 }, - { IsVertexArrayAPPLE_names, IsVertexArrayAPPLE_remap_index, -1 }, - { BindVertexArrayAPPLE_names, BindVertexArrayAPPLE_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_ARB_copy_buffer) -static const struct dri_extension_function GL_ARB_copy_buffer_functions[] = { - { CopyBufferSubData_names, CopyBufferSubData_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_ARB_draw_buffers) -static const struct dri_extension_function GL_ARB_draw_buffers_functions[] = { - { DrawBuffersARB_names, DrawBuffersARB_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_ARB_draw_elements_base_vertex) -static const struct dri_extension_function GL_ARB_draw_elements_base_vertex_functions[] = { - { DrawElementsBaseVertex_names, DrawElementsBaseVertex_remap_index, -1 }, - { MultiDrawElementsBaseVertex_names, MultiDrawElementsBaseVertex_remap_index, -1 }, - { DrawRangeElementsBaseVertex_names, DrawRangeElementsBaseVertex_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_ARB_framebuffer_object) -static const struct dri_extension_function GL_ARB_framebuffer_object_functions[] = { - { BlitFramebufferEXT_names, BlitFramebufferEXT_remap_index, -1 }, - { FramebufferTextureLayerEXT_names, FramebufferTextureLayerEXT_remap_index, -1 }, - { GenerateMipmapEXT_names, GenerateMipmapEXT_remap_index, -1 }, - { RenderbufferStorageEXT_names, RenderbufferStorageEXT_remap_index, -1 }, - { CheckFramebufferStatusEXT_names, CheckFramebufferStatusEXT_remap_index, -1 }, - { FramebufferTexture3DEXT_names, FramebufferTexture3DEXT_remap_index, -1 }, - { FramebufferTexture2DEXT_names, FramebufferTexture2DEXT_remap_index, -1 }, - { RenderbufferStorageMultisample_names, RenderbufferStorageMultisample_remap_index, -1 }, - { FramebufferRenderbufferEXT_names, FramebufferRenderbufferEXT_remap_index, -1 }, - { FramebufferTexture1DEXT_names, FramebufferTexture1DEXT_remap_index, -1 }, - { BindFramebufferEXT_names, BindFramebufferEXT_remap_index, -1 }, - { GenRenderbuffersEXT_names, GenRenderbuffersEXT_remap_index, -1 }, - { IsFramebufferEXT_names, IsFramebufferEXT_remap_index, -1 }, - { GetFramebufferAttachmentParameterivEXT_names, GetFramebufferAttachmentParameterivEXT_remap_index, -1 }, - { DeleteFramebuffersEXT_names, DeleteFramebuffersEXT_remap_index, -1 }, - { GenFramebuffersEXT_names, GenFramebuffersEXT_remap_index, -1 }, - { BindRenderbufferEXT_names, BindRenderbufferEXT_remap_index, -1 }, - { DeleteRenderbuffersEXT_names, DeleteRenderbuffersEXT_remap_index, -1 }, - { GetRenderbufferParameterivEXT_names, GetRenderbufferParameterivEXT_remap_index, -1 }, - { IsRenderbufferEXT_names, IsRenderbufferEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_ARB_map_buffer_range) -static const struct dri_extension_function GL_ARB_map_buffer_range_functions[] = { - { FlushMappedBufferRange_names, FlushMappedBufferRange_remap_index, -1 }, - { MapBufferRange_names, MapBufferRange_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_ARB_matrix_palette) -static const struct dri_extension_function GL_ARB_matrix_palette_functions[] = { - { MatrixIndexusvARB_names, MatrixIndexusvARB_remap_index, -1 }, - { MatrixIndexuivARB_names, MatrixIndexuivARB_remap_index, -1 }, - { MatrixIndexPointerARB_names, MatrixIndexPointerARB_remap_index, -1 }, - { CurrentPaletteMatrixARB_names, CurrentPaletteMatrixARB_remap_index, -1 }, - { MatrixIndexubvARB_names, MatrixIndexubvARB_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_ARB_multisample) -static const struct dri_extension_function GL_ARB_multisample_functions[] = { - { SampleCoverageARB_names, SampleCoverageARB_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_ARB_occlusion_query) -static const struct dri_extension_function GL_ARB_occlusion_query_functions[] = { - { BeginQueryARB_names, BeginQueryARB_remap_index, -1 }, - { GetQueryivARB_names, GetQueryivARB_remap_index, -1 }, - { GetQueryObjectivARB_names, GetQueryObjectivARB_remap_index, -1 }, - { EndQueryARB_names, EndQueryARB_remap_index, -1 }, - { GetQueryObjectuivARB_names, GetQueryObjectuivARB_remap_index, -1 }, - { DeleteQueriesARB_names, DeleteQueriesARB_remap_index, -1 }, - { IsQueryARB_names, IsQueryARB_remap_index, -1 }, - { GenQueriesARB_names, GenQueriesARB_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_ARB_point_parameters) -static const struct dri_extension_function GL_ARB_point_parameters_functions[] = { - { PointParameterfEXT_names, PointParameterfEXT_remap_index, -1 }, - { PointParameterfvEXT_names, PointParameterfvEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_ARB_shader_objects) -static const struct dri_extension_function GL_ARB_shader_objects_functions[] = { - { UniformMatrix3fvARB_names, UniformMatrix3fvARB_remap_index, -1 }, - { Uniform2fARB_names, Uniform2fARB_remap_index, -1 }, - { Uniform2ivARB_names, Uniform2ivARB_remap_index, -1 }, - { UniformMatrix4fvARB_names, UniformMatrix4fvARB_remap_index, -1 }, - { CreateProgramObjectARB_names, CreateProgramObjectARB_remap_index, -1 }, - { Uniform3iARB_names, Uniform3iARB_remap_index, -1 }, - { CreateShaderObjectARB_names, CreateShaderObjectARB_remap_index, -1 }, - { Uniform1fARB_names, Uniform1fARB_remap_index, -1 }, - { AttachObjectARB_names, AttachObjectARB_remap_index, -1 }, - { UniformMatrix2fvARB_names, UniformMatrix2fvARB_remap_index, -1 }, - { GetAttachedObjectsARB_names, GetAttachedObjectsARB_remap_index, -1 }, - { Uniform3fvARB_names, Uniform3fvARB_remap_index, -1 }, - { GetHandleARB_names, GetHandleARB_remap_index, -1 }, - { GetActiveUniformARB_names, GetActiveUniformARB_remap_index, -1 }, - { GetUniformivARB_names, GetUniformivARB_remap_index, -1 }, - { Uniform2fvARB_names, Uniform2fvARB_remap_index, -1 }, - { DeleteObjectARB_names, DeleteObjectARB_remap_index, -1 }, - { ValidateProgramARB_names, ValidateProgramARB_remap_index, -1 }, - { Uniform3ivARB_names, Uniform3ivARB_remap_index, -1 }, - { CompileShaderARB_names, CompileShaderARB_remap_index, -1 }, - { Uniform3fARB_names, Uniform3fARB_remap_index, -1 }, - { Uniform4fARB_names, Uniform4fARB_remap_index, -1 }, - { LinkProgramARB_names, LinkProgramARB_remap_index, -1 }, - { ShaderSourceARB_names, ShaderSourceARB_remap_index, -1 }, - { Uniform4ivARB_names, Uniform4ivARB_remap_index, -1 }, - { Uniform1ivARB_names, Uniform1ivARB_remap_index, -1 }, - { Uniform1iARB_names, Uniform1iARB_remap_index, -1 }, - { Uniform4fvARB_names, Uniform4fvARB_remap_index, -1 }, - { GetUniformfvARB_names, GetUniformfvARB_remap_index, -1 }, - { DetachObjectARB_names, DetachObjectARB_remap_index, -1 }, - { Uniform4iARB_names, Uniform4iARB_remap_index, -1 }, - { UseProgramObjectARB_names, UseProgramObjectARB_remap_index, -1 }, - { Uniform2iARB_names, Uniform2iARB_remap_index, -1 }, - { GetObjectParameterivARB_names, GetObjectParameterivARB_remap_index, -1 }, - { GetUniformLocationARB_names, GetUniformLocationARB_remap_index, -1 }, - { GetShaderSourceARB_names, GetShaderSourceARB_remap_index, -1 }, - { Uniform1fvARB_names, Uniform1fvARB_remap_index, -1 }, - { GetObjectParameterfvARB_names, GetObjectParameterfvARB_remap_index, -1 }, - { GetInfoLogARB_names, GetInfoLogARB_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_ARB_sync) -static const struct dri_extension_function GL_ARB_sync_functions[] = { - { DeleteSync_names, DeleteSync_remap_index, -1 }, - { FenceSync_names, FenceSync_remap_index, -1 }, - { WaitSync_names, WaitSync_remap_index, -1 }, - { GetInteger64v_names, GetInteger64v_remap_index, -1 }, - { GetSynciv_names, GetSynciv_remap_index, -1 }, - { IsSync_names, IsSync_remap_index, -1 }, - { ClientWaitSync_names, ClientWaitSync_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_ARB_texture_compression) -static const struct dri_extension_function GL_ARB_texture_compression_functions[] = { - { CompressedTexSubImage2DARB_names, CompressedTexSubImage2DARB_remap_index, -1 }, - { CompressedTexImage3DARB_names, CompressedTexImage3DARB_remap_index, -1 }, - { CompressedTexImage1DARB_names, CompressedTexImage1DARB_remap_index, -1 }, - { CompressedTexSubImage1DARB_names, CompressedTexSubImage1DARB_remap_index, -1 }, - { CompressedTexSubImage3DARB_names, CompressedTexSubImage3DARB_remap_index, -1 }, - { CompressedTexImage2DARB_names, CompressedTexImage2DARB_remap_index, -1 }, - { GetCompressedTexImageARB_names, GetCompressedTexImageARB_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_ARB_transpose_matrix) -static const struct dri_extension_function GL_ARB_transpose_matrix_functions[] = { - { MultTransposeMatrixdARB_names, MultTransposeMatrixdARB_remap_index, -1 }, - { LoadTransposeMatrixdARB_names, LoadTransposeMatrixdARB_remap_index, -1 }, - { MultTransposeMatrixfARB_names, MultTransposeMatrixfARB_remap_index, -1 }, - { LoadTransposeMatrixfARB_names, LoadTransposeMatrixfARB_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_ARB_vertex_array_object) -static const struct dri_extension_function GL_ARB_vertex_array_object_functions[] = { - { DeleteVertexArraysAPPLE_names, DeleteVertexArraysAPPLE_remap_index, -1 }, - { GenVertexArrays_names, GenVertexArrays_remap_index, -1 }, - { BindVertexArray_names, BindVertexArray_remap_index, -1 }, - { IsVertexArrayAPPLE_names, IsVertexArrayAPPLE_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_ARB_vertex_blend) -static const struct dri_extension_function GL_ARB_vertex_blend_functions[] = { - { WeightubvARB_names, WeightubvARB_remap_index, -1 }, - { WeightivARB_names, WeightivARB_remap_index, -1 }, - { WeightPointerARB_names, WeightPointerARB_remap_index, -1 }, - { WeightfvARB_names, WeightfvARB_remap_index, -1 }, - { WeightbvARB_names, WeightbvARB_remap_index, -1 }, - { WeightusvARB_names, WeightusvARB_remap_index, -1 }, - { VertexBlendARB_names, VertexBlendARB_remap_index, -1 }, - { WeightsvARB_names, WeightsvARB_remap_index, -1 }, - { WeightdvARB_names, WeightdvARB_remap_index, -1 }, - { WeightuivARB_names, WeightuivARB_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_ARB_vertex_buffer_object) -static const struct dri_extension_function GL_ARB_vertex_buffer_object_functions[] = { - { GetBufferSubDataARB_names, GetBufferSubDataARB_remap_index, -1 }, - { BufferSubDataARB_names, BufferSubDataARB_remap_index, -1 }, - { BufferDataARB_names, BufferDataARB_remap_index, -1 }, - { GetBufferPointervARB_names, GetBufferPointervARB_remap_index, -1 }, - { GetBufferParameterivARB_names, GetBufferParameterivARB_remap_index, -1 }, - { MapBufferARB_names, MapBufferARB_remap_index, -1 }, - { IsBufferARB_names, IsBufferARB_remap_index, -1 }, - { DeleteBuffersARB_names, DeleteBuffersARB_remap_index, -1 }, - { UnmapBufferARB_names, UnmapBufferARB_remap_index, -1 }, - { BindBufferARB_names, BindBufferARB_remap_index, -1 }, - { GenBuffersARB_names, GenBuffersARB_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_ARB_vertex_program) -static const struct dri_extension_function GL_ARB_vertex_program_functions[] = { - { VertexAttrib2fARB_names, VertexAttrib2fARB_remap_index, -1 }, - { VertexAttrib3fARB_names, VertexAttrib3fARB_remap_index, -1 }, - { VertexAttrib1svARB_names, VertexAttrib1svARB_remap_index, -1 }, - { VertexAttrib4NusvARB_names, VertexAttrib4NusvARB_remap_index, -1 }, - { DisableVertexAttribArrayARB_names, DisableVertexAttribArrayARB_remap_index, -1 }, - { ProgramLocalParameter4dARB_names, ProgramLocalParameter4dARB_remap_index, -1 }, - { VertexAttrib1fARB_names, VertexAttrib1fARB_remap_index, -1 }, - { VertexAttrib4NbvARB_names, VertexAttrib4NbvARB_remap_index, -1 }, - { VertexAttrib1sARB_names, VertexAttrib1sARB_remap_index, -1 }, - { GetProgramLocalParameterfvARB_names, GetProgramLocalParameterfvARB_remap_index, -1 }, - { VertexAttrib3dvARB_names, VertexAttrib3dvARB_remap_index, -1 }, - { ProgramEnvParameter4fvARB_names, ProgramEnvParameter4fvARB_remap_index, -1 }, - { GetVertexAttribivARB_names, GetVertexAttribivARB_remap_index, -1 }, - { VertexAttrib4ivARB_names, VertexAttrib4ivARB_remap_index, -1 }, - { VertexAttrib4bvARB_names, VertexAttrib4bvARB_remap_index, -1 }, - { VertexAttrib3dARB_names, VertexAttrib3dARB_remap_index, -1 }, - { VertexAttrib4fARB_names, VertexAttrib4fARB_remap_index, -1 }, - { VertexAttrib4fvARB_names, VertexAttrib4fvARB_remap_index, -1 }, - { ProgramLocalParameter4dvARB_names, ProgramLocalParameter4dvARB_remap_index, -1 }, - { VertexAttrib2dARB_names, VertexAttrib2dARB_remap_index, -1 }, - { VertexAttrib1dvARB_names, VertexAttrib1dvARB_remap_index, -1 }, - { GetVertexAttribfvARB_names, GetVertexAttribfvARB_remap_index, -1 }, - { VertexAttrib4ubvARB_names, VertexAttrib4ubvARB_remap_index, -1 }, - { ProgramEnvParameter4fARB_names, ProgramEnvParameter4fARB_remap_index, -1 }, - { VertexAttrib4sARB_names, VertexAttrib4sARB_remap_index, -1 }, - { VertexAttrib2dvARB_names, VertexAttrib2dvARB_remap_index, -1 }, - { VertexAttrib2fvARB_names, VertexAttrib2fvARB_remap_index, -1 }, - { VertexAttrib4NivARB_names, VertexAttrib4NivARB_remap_index, -1 }, - { GetProgramStringARB_names, GetProgramStringARB_remap_index, -1 }, - { VertexAttrib4NuivARB_names, VertexAttrib4NuivARB_remap_index, -1 }, - { IsProgramNV_names, IsProgramNV_remap_index, -1 }, - { ProgramEnvParameter4dARB_names, ProgramEnvParameter4dARB_remap_index, -1 }, - { VertexAttrib1dARB_names, VertexAttrib1dARB_remap_index, -1 }, - { VertexAttrib4usvARB_names, VertexAttrib4usvARB_remap_index, -1 }, - { VertexAttrib3svARB_names, VertexAttrib3svARB_remap_index, -1 }, - { GetVertexAttribdvARB_names, GetVertexAttribdvARB_remap_index, -1 }, - { VertexAttrib4dvARB_names, VertexAttrib4dvARB_remap_index, -1 }, - { VertexAttribPointerARB_names, VertexAttribPointerARB_remap_index, -1 }, - { VertexAttrib4NsvARB_names, VertexAttrib4NsvARB_remap_index, -1 }, - { VertexAttrib3fvARB_names, VertexAttrib3fvARB_remap_index, -1 }, - { VertexAttrib4NubARB_names, VertexAttrib4NubARB_remap_index, -1 }, - { GetProgramEnvParameterfvARB_names, GetProgramEnvParameterfvARB_remap_index, -1 }, - { ProgramEnvParameter4dvARB_names, ProgramEnvParameter4dvARB_remap_index, -1 }, - { ProgramLocalParameter4fvARB_names, ProgramLocalParameter4fvARB_remap_index, -1 }, - { DeleteProgramsNV_names, DeleteProgramsNV_remap_index, -1 }, - { GetVertexAttribPointervNV_names, GetVertexAttribPointervNV_remap_index, -1 }, - { VertexAttrib4dARB_names, VertexAttrib4dARB_remap_index, -1 }, - { GetProgramLocalParameterdvARB_names, GetProgramLocalParameterdvARB_remap_index, -1 }, - { GetProgramivARB_names, GetProgramivARB_remap_index, -1 }, - { VertexAttrib3sARB_names, VertexAttrib3sARB_remap_index, -1 }, - { ProgramStringARB_names, ProgramStringARB_remap_index, -1 }, - { ProgramLocalParameter4fARB_names, ProgramLocalParameter4fARB_remap_index, -1 }, - { EnableVertexAttribArrayARB_names, EnableVertexAttribArrayARB_remap_index, -1 }, - { VertexAttrib4uivARB_names, VertexAttrib4uivARB_remap_index, -1 }, - { BindProgramNV_names, BindProgramNV_remap_index, -1 }, - { VertexAttrib4svARB_names, VertexAttrib4svARB_remap_index, -1 }, - { VertexAttrib2svARB_names, VertexAttrib2svARB_remap_index, -1 }, - { VertexAttrib4NubvARB_names, VertexAttrib4NubvARB_remap_index, -1 }, - { GetProgramEnvParameterdvARB_names, GetProgramEnvParameterdvARB_remap_index, -1 }, - { VertexAttrib2sARB_names, VertexAttrib2sARB_remap_index, -1 }, - { VertexAttrib1fvARB_names, VertexAttrib1fvARB_remap_index, -1 }, - { GenProgramsNV_names, GenProgramsNV_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_ARB_vertex_shader) -static const struct dri_extension_function GL_ARB_vertex_shader_functions[] = { - { GetActiveAttribARB_names, GetActiveAttribARB_remap_index, -1 }, - { GetAttribLocationARB_names, GetAttribLocationARB_remap_index, -1 }, - { BindAttribLocationARB_names, BindAttribLocationARB_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_ARB_window_pos) -static const struct dri_extension_function GL_ARB_window_pos_functions[] = { - { WindowPos3fMESA_names, WindowPos3fMESA_remap_index, -1 }, - { WindowPos2dvMESA_names, WindowPos2dvMESA_remap_index, -1 }, - { WindowPos2svMESA_names, WindowPos2svMESA_remap_index, -1 }, - { WindowPos3dMESA_names, WindowPos3dMESA_remap_index, -1 }, - { WindowPos2fvMESA_names, WindowPos2fvMESA_remap_index, -1 }, - { WindowPos2dMESA_names, WindowPos2dMESA_remap_index, -1 }, - { WindowPos3dvMESA_names, WindowPos3dvMESA_remap_index, -1 }, - { WindowPos3fvMESA_names, WindowPos3fvMESA_remap_index, -1 }, - { WindowPos2iMESA_names, WindowPos2iMESA_remap_index, -1 }, - { WindowPos3sMESA_names, WindowPos3sMESA_remap_index, -1 }, - { WindowPos2ivMESA_names, WindowPos2ivMESA_remap_index, -1 }, - { WindowPos2sMESA_names, WindowPos2sMESA_remap_index, -1 }, - { WindowPos3iMESA_names, WindowPos3iMESA_remap_index, -1 }, - { WindowPos3ivMESA_names, WindowPos3ivMESA_remap_index, -1 }, - { WindowPos3svMESA_names, WindowPos3svMESA_remap_index, -1 }, - { WindowPos2fMESA_names, WindowPos2fMESA_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_ATI_blend_equation_separate) -static const struct dri_extension_function GL_ATI_blend_equation_separate_functions[] = { - { BlendEquationSeparateEXT_names, BlendEquationSeparateEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_ATI_draw_buffers) -static const struct dri_extension_function GL_ATI_draw_buffers_functions[] = { - { DrawBuffersARB_names, DrawBuffersARB_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_ATI_envmap_bumpmap) -static const struct dri_extension_function GL_ATI_envmap_bumpmap_functions[] = { - { TexBumpParameterfvATI_names, TexBumpParameterfvATI_remap_index, -1 }, - { TexBumpParameterivATI_names, TexBumpParameterivATI_remap_index, -1 }, - { GetTexBumpParameterfvATI_names, GetTexBumpParameterfvATI_remap_index, -1 }, - { GetTexBumpParameterivATI_names, GetTexBumpParameterivATI_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_ATI_fragment_shader) -static const struct dri_extension_function GL_ATI_fragment_shader_functions[] = { - { ColorFragmentOp2ATI_names, ColorFragmentOp2ATI_remap_index, -1 }, - { DeleteFragmentShaderATI_names, DeleteFragmentShaderATI_remap_index, -1 }, - { SetFragmentShaderConstantATI_names, SetFragmentShaderConstantATI_remap_index, -1 }, - { SampleMapATI_names, SampleMapATI_remap_index, -1 }, - { AlphaFragmentOp2ATI_names, AlphaFragmentOp2ATI_remap_index, -1 }, - { AlphaFragmentOp1ATI_names, AlphaFragmentOp1ATI_remap_index, -1 }, - { ColorFragmentOp1ATI_names, ColorFragmentOp1ATI_remap_index, -1 }, - { AlphaFragmentOp3ATI_names, AlphaFragmentOp3ATI_remap_index, -1 }, - { PassTexCoordATI_names, PassTexCoordATI_remap_index, -1 }, - { BeginFragmentShaderATI_names, BeginFragmentShaderATI_remap_index, -1 }, - { BindFragmentShaderATI_names, BindFragmentShaderATI_remap_index, -1 }, - { ColorFragmentOp3ATI_names, ColorFragmentOp3ATI_remap_index, -1 }, - { GenFragmentShadersATI_names, GenFragmentShadersATI_remap_index, -1 }, - { EndFragmentShaderATI_names, EndFragmentShaderATI_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_ATI_separate_stencil) -static const struct dri_extension_function GL_ATI_separate_stencil_functions[] = { - { StencilOpSeparate_names, StencilOpSeparate_remap_index, -1 }, - { StencilFuncSeparateATI_names, StencilFuncSeparateATI_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_blend_color) -static const struct dri_extension_function GL_EXT_blend_color_functions[] = { - { BlendColor_names, -1, 336 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_blend_equation_separate) -static const struct dri_extension_function GL_EXT_blend_equation_separate_functions[] = { - { BlendEquationSeparateEXT_names, BlendEquationSeparateEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_blend_func_separate) -static const struct dri_extension_function GL_EXT_blend_func_separate_functions[] = { - { BlendFuncSeparateEXT_names, BlendFuncSeparateEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_blend_minmax) -static const struct dri_extension_function GL_EXT_blend_minmax_functions[] = { - { BlendEquation_names, -1, 337 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_color_subtable) -static const struct dri_extension_function GL_EXT_color_subtable_functions[] = { - { ColorSubTable_names, -1, 346 }, - { CopyColorSubTable_names, -1, 347 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_compiled_vertex_array) -static const struct dri_extension_function GL_EXT_compiled_vertex_array_functions[] = { - { UnlockArraysEXT_names, UnlockArraysEXT_remap_index, -1 }, - { LockArraysEXT_names, LockArraysEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_convolution) -static const struct dri_extension_function GL_EXT_convolution_functions[] = { - { ConvolutionFilter1D_names, -1, 348 }, - { CopyConvolutionFilter1D_names, -1, 354 }, - { GetConvolutionParameteriv_names, -1, 358 }, - { ConvolutionFilter2D_names, -1, 349 }, - { ConvolutionParameteriv_names, -1, 353 }, - { ConvolutionParameterfv_names, -1, 351 }, - { GetSeparableFilter_names, -1, 359 }, - { SeparableFilter2D_names, -1, 360 }, - { ConvolutionParameteri_names, -1, 352 }, - { ConvolutionParameterf_names, -1, 350 }, - { GetConvolutionParameterfv_names, -1, 357 }, - { GetConvolutionFilter_names, -1, 356 }, - { CopyConvolutionFilter2D_names, -1, 355 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_coordinate_frame) -static const struct dri_extension_function GL_EXT_coordinate_frame_functions[] = { - { TangentPointerEXT_names, TangentPointerEXT_remap_index, -1 }, - { Binormal3ivEXT_names, Binormal3ivEXT_remap_index, -1 }, - { Tangent3sEXT_names, Tangent3sEXT_remap_index, -1 }, - { Tangent3fvEXT_names, Tangent3fvEXT_remap_index, -1 }, - { Tangent3dvEXT_names, Tangent3dvEXT_remap_index, -1 }, - { Binormal3bvEXT_names, Binormal3bvEXT_remap_index, -1 }, - { Binormal3dEXT_names, Binormal3dEXT_remap_index, -1 }, - { Tangent3fEXT_names, Tangent3fEXT_remap_index, -1 }, - { Binormal3sEXT_names, Binormal3sEXT_remap_index, -1 }, - { Tangent3ivEXT_names, Tangent3ivEXT_remap_index, -1 }, - { Tangent3dEXT_names, Tangent3dEXT_remap_index, -1 }, - { Binormal3svEXT_names, Binormal3svEXT_remap_index, -1 }, - { Binormal3fEXT_names, Binormal3fEXT_remap_index, -1 }, - { Binormal3dvEXT_names, Binormal3dvEXT_remap_index, -1 }, - { Tangent3iEXT_names, Tangent3iEXT_remap_index, -1 }, - { Tangent3bvEXT_names, Tangent3bvEXT_remap_index, -1 }, - { Tangent3bEXT_names, Tangent3bEXT_remap_index, -1 }, - { Binormal3fvEXT_names, Binormal3fvEXT_remap_index, -1 }, - { BinormalPointerEXT_names, BinormalPointerEXT_remap_index, -1 }, - { Tangent3svEXT_names, Tangent3svEXT_remap_index, -1 }, - { Binormal3bEXT_names, Binormal3bEXT_remap_index, -1 }, - { Binormal3iEXT_names, Binormal3iEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_copy_texture) -static const struct dri_extension_function GL_EXT_copy_texture_functions[] = { - { CopyTexSubImage3D_names, -1, 373 }, - { CopyTexImage2D_names, -1, 324 }, - { CopyTexImage1D_names, -1, 323 }, - { CopyTexSubImage2D_names, -1, 326 }, - { CopyTexSubImage1D_names, -1, 325 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_cull_vertex) -static const struct dri_extension_function GL_EXT_cull_vertex_functions[] = { - { CullParameterdvEXT_names, CullParameterdvEXT_remap_index, -1 }, - { CullParameterfvEXT_names, CullParameterfvEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_depth_bounds_test) -static const struct dri_extension_function GL_EXT_depth_bounds_test_functions[] = { - { DepthBoundsEXT_names, DepthBoundsEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_draw_range_elements) -static const struct dri_extension_function GL_EXT_draw_range_elements_functions[] = { - { DrawRangeElements_names, -1, 338 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_fog_coord) -static const struct dri_extension_function GL_EXT_fog_coord_functions[] = { - { FogCoorddEXT_names, FogCoorddEXT_remap_index, -1 }, - { FogCoordfEXT_names, FogCoordfEXT_remap_index, -1 }, - { FogCoordPointerEXT_names, FogCoordPointerEXT_remap_index, -1 }, - { FogCoordfvEXT_names, FogCoordfvEXT_remap_index, -1 }, - { FogCoorddvEXT_names, FogCoorddvEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_framebuffer_blit) -static const struct dri_extension_function GL_EXT_framebuffer_blit_functions[] = { - { BlitFramebufferEXT_names, BlitFramebufferEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_framebuffer_object) -static const struct dri_extension_function GL_EXT_framebuffer_object_functions[] = { - { GenerateMipmapEXT_names, GenerateMipmapEXT_remap_index, -1 }, - { RenderbufferStorageEXT_names, RenderbufferStorageEXT_remap_index, -1 }, - { CheckFramebufferStatusEXT_names, CheckFramebufferStatusEXT_remap_index, -1 }, - { FramebufferTexture3DEXT_names, FramebufferTexture3DEXT_remap_index, -1 }, - { FramebufferTexture2DEXT_names, FramebufferTexture2DEXT_remap_index, -1 }, - { FramebufferRenderbufferEXT_names, FramebufferRenderbufferEXT_remap_index, -1 }, - { FramebufferTexture1DEXT_names, FramebufferTexture1DEXT_remap_index, -1 }, - { BindFramebufferEXT_names, BindFramebufferEXT_remap_index, -1 }, - { GenRenderbuffersEXT_names, GenRenderbuffersEXT_remap_index, -1 }, - { IsFramebufferEXT_names, IsFramebufferEXT_remap_index, -1 }, - { GetFramebufferAttachmentParameterivEXT_names, GetFramebufferAttachmentParameterivEXT_remap_index, -1 }, - { DeleteFramebuffersEXT_names, DeleteFramebuffersEXT_remap_index, -1 }, - { GenFramebuffersEXT_names, GenFramebuffersEXT_remap_index, -1 }, - { BindRenderbufferEXT_names, BindRenderbufferEXT_remap_index, -1 }, - { DeleteRenderbuffersEXT_names, DeleteRenderbuffersEXT_remap_index, -1 }, - { GetRenderbufferParameterivEXT_names, GetRenderbufferParameterivEXT_remap_index, -1 }, - { IsRenderbufferEXT_names, IsRenderbufferEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_gpu_program_parameters) -static const struct dri_extension_function GL_EXT_gpu_program_parameters_functions[] = { - { ProgramLocalParameters4fvEXT_names, ProgramLocalParameters4fvEXT_remap_index, -1 }, - { ProgramEnvParameters4fvEXT_names, ProgramEnvParameters4fvEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_histogram) -static const struct dri_extension_function GL_EXT_histogram_functions[] = { - { Histogram_names, -1, 367 }, - { ResetHistogram_names, -1, 369 }, - { GetMinmax_names, -1, 364 }, - { GetHistogramParameterfv_names, -1, 362 }, - { GetMinmaxParameteriv_names, -1, 366 }, - { ResetMinmax_names, -1, 370 }, - { GetHistogramParameteriv_names, -1, 363 }, - { GetHistogram_names, -1, 361 }, - { Minmax_names, -1, 368 }, - { GetMinmaxParameterfv_names, -1, 365 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_index_func) -static const struct dri_extension_function GL_EXT_index_func_functions[] = { - { IndexFuncEXT_names, IndexFuncEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_index_material) -static const struct dri_extension_function GL_EXT_index_material_functions[] = { - { IndexMaterialEXT_names, IndexMaterialEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_light_texture) -static const struct dri_extension_function GL_EXT_light_texture_functions[] = { - { ApplyTextureEXT_names, ApplyTextureEXT_remap_index, -1 }, - { TextureMaterialEXT_names, TextureMaterialEXT_remap_index, -1 }, - { TextureLightEXT_names, TextureLightEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_multi_draw_arrays) -static const struct dri_extension_function GL_EXT_multi_draw_arrays_functions[] = { - { MultiDrawElementsEXT_names, MultiDrawElementsEXT_remap_index, -1 }, - { MultiDrawArraysEXT_names, MultiDrawArraysEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_multisample) -static const struct dri_extension_function GL_EXT_multisample_functions[] = { - { SampleMaskSGIS_names, SampleMaskSGIS_remap_index, -1 }, - { SamplePatternSGIS_names, SamplePatternSGIS_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_paletted_texture) -static const struct dri_extension_function GL_EXT_paletted_texture_functions[] = { - { ColorTable_names, -1, 339 }, - { GetColorTable_names, -1, 343 }, - { GetColorTableParameterfv_names, -1, 344 }, - { GetColorTableParameteriv_names, -1, 345 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_pixel_transform) -static const struct dri_extension_function GL_EXT_pixel_transform_functions[] = { - { PixelTransformParameterfvEXT_names, PixelTransformParameterfvEXT_remap_index, -1 }, - { PixelTransformParameterfEXT_names, PixelTransformParameterfEXT_remap_index, -1 }, - { PixelTransformParameteriEXT_names, PixelTransformParameteriEXT_remap_index, -1 }, - { PixelTransformParameterivEXT_names, PixelTransformParameterivEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_point_parameters) -static const struct dri_extension_function GL_EXT_point_parameters_functions[] = { - { PointParameterfEXT_names, PointParameterfEXT_remap_index, -1 }, - { PointParameterfvEXT_names, PointParameterfvEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_polygon_offset) -static const struct dri_extension_function GL_EXT_polygon_offset_functions[] = { - { PolygonOffsetEXT_names, PolygonOffsetEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_provoking_vertex) -static const struct dri_extension_function GL_EXT_provoking_vertex_functions[] = { - { ProvokingVertexEXT_names, ProvokingVertexEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_secondary_color) -static const struct dri_extension_function GL_EXT_secondary_color_functions[] = { - { SecondaryColor3iEXT_names, SecondaryColor3iEXT_remap_index, -1 }, - { SecondaryColor3bEXT_names, SecondaryColor3bEXT_remap_index, -1 }, - { SecondaryColor3bvEXT_names, SecondaryColor3bvEXT_remap_index, -1 }, - { SecondaryColor3sEXT_names, SecondaryColor3sEXT_remap_index, -1 }, - { SecondaryColor3dEXT_names, SecondaryColor3dEXT_remap_index, -1 }, - { SecondaryColorPointerEXT_names, SecondaryColorPointerEXT_remap_index, -1 }, - { SecondaryColor3uiEXT_names, SecondaryColor3uiEXT_remap_index, -1 }, - { SecondaryColor3usvEXT_names, SecondaryColor3usvEXT_remap_index, -1 }, - { SecondaryColor3ivEXT_names, SecondaryColor3ivEXT_remap_index, -1 }, - { SecondaryColor3fvEXT_names, SecondaryColor3fvEXT_remap_index, -1 }, - { SecondaryColor3ubvEXT_names, SecondaryColor3ubvEXT_remap_index, -1 }, - { SecondaryColor3uivEXT_names, SecondaryColor3uivEXT_remap_index, -1 }, - { SecondaryColor3dvEXT_names, SecondaryColor3dvEXT_remap_index, -1 }, - { SecondaryColor3usEXT_names, SecondaryColor3usEXT_remap_index, -1 }, - { SecondaryColor3ubEXT_names, SecondaryColor3ubEXT_remap_index, -1 }, - { SecondaryColor3fEXT_names, SecondaryColor3fEXT_remap_index, -1 }, - { SecondaryColor3svEXT_names, SecondaryColor3svEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_stencil_two_side) -static const struct dri_extension_function GL_EXT_stencil_two_side_functions[] = { - { ActiveStencilFaceEXT_names, ActiveStencilFaceEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_subtexture) -static const struct dri_extension_function GL_EXT_subtexture_functions[] = { - { TexSubImage1D_names, -1, 332 }, - { TexSubImage2D_names, -1, 333 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_texture3D) -static const struct dri_extension_function GL_EXT_texture3D_functions[] = { - { TexImage3D_names, -1, 371 }, - { TexSubImage3D_names, -1, 372 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_texture_array) -static const struct dri_extension_function GL_EXT_texture_array_functions[] = { - { FramebufferTextureLayerEXT_names, FramebufferTextureLayerEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_texture_object) -static const struct dri_extension_function GL_EXT_texture_object_functions[] = { - { PrioritizeTextures_names, -1, 331 }, - { AreTexturesResident_names, -1, 322 }, - { GenTextures_names, -1, 328 }, - { DeleteTextures_names, -1, 327 }, - { IsTexture_names, -1, 330 }, - { BindTexture_names, -1, 307 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_texture_perturb_normal) -static const struct dri_extension_function GL_EXT_texture_perturb_normal_functions[] = { - { TextureNormalEXT_names, TextureNormalEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_timer_query) -static const struct dri_extension_function GL_EXT_timer_query_functions[] = { - { GetQueryObjectui64vEXT_names, GetQueryObjectui64vEXT_remap_index, -1 }, - { GetQueryObjecti64vEXT_names, GetQueryObjecti64vEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_vertex_array) -static const struct dri_extension_function GL_EXT_vertex_array_functions[] = { - { IndexPointerEXT_names, IndexPointerEXT_remap_index, -1 }, - { NormalPointerEXT_names, NormalPointerEXT_remap_index, -1 }, - { VertexPointerEXT_names, VertexPointerEXT_remap_index, -1 }, - { TexCoordPointerEXT_names, TexCoordPointerEXT_remap_index, -1 }, - { EdgeFlagPointerEXT_names, EdgeFlagPointerEXT_remap_index, -1 }, - { ArrayElement_names, -1, 306 }, - { ColorPointerEXT_names, ColorPointerEXT_remap_index, -1 }, - { GetPointerv_names, -1, 329 }, - { DrawArrays_names, -1, 310 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_EXT_vertex_weighting) -static const struct dri_extension_function GL_EXT_vertex_weighting_functions[] = { - { VertexWeightfvEXT_names, VertexWeightfvEXT_remap_index, -1 }, - { VertexWeightfEXT_names, VertexWeightfEXT_remap_index, -1 }, - { VertexWeightPointerEXT_names, VertexWeightPointerEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_HP_image_transform) -static const struct dri_extension_function GL_HP_image_transform_functions[] = { - { GetImageTransformParameterfvHP_names, GetImageTransformParameterfvHP_remap_index, -1 }, - { ImageTransformParameterfHP_names, ImageTransformParameterfHP_remap_index, -1 }, - { ImageTransformParameterfvHP_names, ImageTransformParameterfvHP_remap_index, -1 }, - { ImageTransformParameteriHP_names, ImageTransformParameteriHP_remap_index, -1 }, - { GetImageTransformParameterivHP_names, GetImageTransformParameterivHP_remap_index, -1 }, - { ImageTransformParameterivHP_names, ImageTransformParameterivHP_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_IBM_multimode_draw_arrays) -static const struct dri_extension_function GL_IBM_multimode_draw_arrays_functions[] = { - { MultiModeDrawArraysIBM_names, MultiModeDrawArraysIBM_remap_index, -1 }, - { MultiModeDrawElementsIBM_names, MultiModeDrawElementsIBM_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_IBM_vertex_array_lists) -static const struct dri_extension_function GL_IBM_vertex_array_lists_functions[] = { - { SecondaryColorPointerListIBM_names, SecondaryColorPointerListIBM_remap_index, -1 }, - { NormalPointerListIBM_names, NormalPointerListIBM_remap_index, -1 }, - { FogCoordPointerListIBM_names, FogCoordPointerListIBM_remap_index, -1 }, - { VertexPointerListIBM_names, VertexPointerListIBM_remap_index, -1 }, - { ColorPointerListIBM_names, ColorPointerListIBM_remap_index, -1 }, - { TexCoordPointerListIBM_names, TexCoordPointerListIBM_remap_index, -1 }, - { IndexPointerListIBM_names, IndexPointerListIBM_remap_index, -1 }, - { EdgeFlagPointerListIBM_names, EdgeFlagPointerListIBM_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_INGR_blend_func_separate) -static const struct dri_extension_function GL_INGR_blend_func_separate_functions[] = { - { BlendFuncSeparateEXT_names, BlendFuncSeparateEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_INTEL_parallel_arrays) -static const struct dri_extension_function GL_INTEL_parallel_arrays_functions[] = { - { VertexPointervINTEL_names, VertexPointervINTEL_remap_index, -1 }, - { ColorPointervINTEL_names, ColorPointervINTEL_remap_index, -1 }, - { NormalPointervINTEL_names, NormalPointervINTEL_remap_index, -1 }, - { TexCoordPointervINTEL_names, TexCoordPointervINTEL_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_MESA_resize_buffers) -static const struct dri_extension_function GL_MESA_resize_buffers_functions[] = { - { ResizeBuffersMESA_names, ResizeBuffersMESA_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_MESA_shader_debug) -static const struct dri_extension_function GL_MESA_shader_debug_functions[] = { - { GetDebugLogLengthMESA_names, GetDebugLogLengthMESA_remap_index, -1 }, - { ClearDebugLogMESA_names, ClearDebugLogMESA_remap_index, -1 }, - { GetDebugLogMESA_names, GetDebugLogMESA_remap_index, -1 }, - { CreateDebugObjectMESA_names, CreateDebugObjectMESA_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_MESA_window_pos) -static const struct dri_extension_function GL_MESA_window_pos_functions[] = { - { WindowPos3fMESA_names, WindowPos3fMESA_remap_index, -1 }, - { WindowPos2dvMESA_names, WindowPos2dvMESA_remap_index, -1 }, - { WindowPos4svMESA_names, WindowPos4svMESA_remap_index, -1 }, - { WindowPos2svMESA_names, WindowPos2svMESA_remap_index, -1 }, - { WindowPos3dMESA_names, WindowPos3dMESA_remap_index, -1 }, - { WindowPos2fvMESA_names, WindowPos2fvMESA_remap_index, -1 }, - { WindowPos4dMESA_names, WindowPos4dMESA_remap_index, -1 }, - { WindowPos2dMESA_names, WindowPos2dMESA_remap_index, -1 }, - { WindowPos4ivMESA_names, WindowPos4ivMESA_remap_index, -1 }, - { WindowPos4fMESA_names, WindowPos4fMESA_remap_index, -1 }, - { WindowPos3dvMESA_names, WindowPos3dvMESA_remap_index, -1 }, - { WindowPos3fvMESA_names, WindowPos3fvMESA_remap_index, -1 }, - { WindowPos4dvMESA_names, WindowPos4dvMESA_remap_index, -1 }, - { WindowPos2iMESA_names, WindowPos2iMESA_remap_index, -1 }, - { WindowPos3sMESA_names, WindowPos3sMESA_remap_index, -1 }, - { WindowPos4sMESA_names, WindowPos4sMESA_remap_index, -1 }, - { WindowPos2ivMESA_names, WindowPos2ivMESA_remap_index, -1 }, - { WindowPos2sMESA_names, WindowPos2sMESA_remap_index, -1 }, - { WindowPos3iMESA_names, WindowPos3iMESA_remap_index, -1 }, - { WindowPos3ivMESA_names, WindowPos3ivMESA_remap_index, -1 }, - { WindowPos4iMESA_names, WindowPos4iMESA_remap_index, -1 }, - { WindowPos4fvMESA_names, WindowPos4fvMESA_remap_index, -1 }, - { WindowPos3svMESA_names, WindowPos3svMESA_remap_index, -1 }, - { WindowPos2fMESA_names, WindowPos2fMESA_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_NV_evaluators) -static const struct dri_extension_function GL_NV_evaluators_functions[] = { - { GetMapAttribParameterivNV_names, GetMapAttribParameterivNV_remap_index, -1 }, - { MapControlPointsNV_names, MapControlPointsNV_remap_index, -1 }, - { MapParameterfvNV_names, MapParameterfvNV_remap_index, -1 }, - { EvalMapsNV_names, EvalMapsNV_remap_index, -1 }, - { GetMapAttribParameterfvNV_names, GetMapAttribParameterfvNV_remap_index, -1 }, - { MapParameterivNV_names, MapParameterivNV_remap_index, -1 }, - { GetMapParameterivNV_names, GetMapParameterivNV_remap_index, -1 }, - { GetMapParameterfvNV_names, GetMapParameterfvNV_remap_index, -1 }, - { GetMapControlPointsNV_names, GetMapControlPointsNV_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_NV_fence) -static const struct dri_extension_function GL_NV_fence_functions[] = { - { GenFencesNV_names, GenFencesNV_remap_index, -1 }, - { TestFenceNV_names, TestFenceNV_remap_index, -1 }, - { IsFenceNV_names, IsFenceNV_remap_index, -1 }, - { DeleteFencesNV_names, DeleteFencesNV_remap_index, -1 }, - { SetFenceNV_names, SetFenceNV_remap_index, -1 }, - { GetFenceivNV_names, GetFenceivNV_remap_index, -1 }, - { FinishFenceNV_names, FinishFenceNV_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_NV_fragment_program) -static const struct dri_extension_function GL_NV_fragment_program_functions[] = { - { GetProgramNamedParameterdvNV_names, GetProgramNamedParameterdvNV_remap_index, -1 }, - { GetProgramNamedParameterfvNV_names, GetProgramNamedParameterfvNV_remap_index, -1 }, - { ProgramNamedParameter4fNV_names, ProgramNamedParameter4fNV_remap_index, -1 }, - { ProgramNamedParameter4fvNV_names, ProgramNamedParameter4fvNV_remap_index, -1 }, - { ProgramNamedParameter4dvNV_names, ProgramNamedParameter4dvNV_remap_index, -1 }, - { ProgramNamedParameter4dNV_names, ProgramNamedParameter4dNV_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_NV_point_sprite) -static const struct dri_extension_function GL_NV_point_sprite_functions[] = { - { PointParameteriNV_names, PointParameteriNV_remap_index, -1 }, - { PointParameterivNV_names, PointParameterivNV_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_NV_register_combiners) -static const struct dri_extension_function GL_NV_register_combiners_functions[] = { - { CombinerOutputNV_names, CombinerOutputNV_remap_index, -1 }, - { CombinerParameterfvNV_names, CombinerParameterfvNV_remap_index, -1 }, - { GetFinalCombinerInputParameterfvNV_names, GetFinalCombinerInputParameterfvNV_remap_index, -1 }, - { GetCombinerOutputParameterfvNV_names, GetCombinerOutputParameterfvNV_remap_index, -1 }, - { FinalCombinerInputNV_names, FinalCombinerInputNV_remap_index, -1 }, - { GetCombinerInputParameterfvNV_names, GetCombinerInputParameterfvNV_remap_index, -1 }, - { GetCombinerOutputParameterivNV_names, GetCombinerOutputParameterivNV_remap_index, -1 }, - { CombinerParameteriNV_names, CombinerParameteriNV_remap_index, -1 }, - { GetFinalCombinerInputParameterivNV_names, GetFinalCombinerInputParameterivNV_remap_index, -1 }, - { CombinerInputNV_names, CombinerInputNV_remap_index, -1 }, - { CombinerParameterfNV_names, CombinerParameterfNV_remap_index, -1 }, - { GetCombinerInputParameterivNV_names, GetCombinerInputParameterivNV_remap_index, -1 }, - { CombinerParameterivNV_names, CombinerParameterivNV_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_NV_register_combiners2) -static const struct dri_extension_function GL_NV_register_combiners2_functions[] = { - { CombinerStageParameterfvNV_names, CombinerStageParameterfvNV_remap_index, -1 }, - { GetCombinerStageParameterfvNV_names, GetCombinerStageParameterfvNV_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_NV_vertex_array_range) -static const struct dri_extension_function GL_NV_vertex_array_range_functions[] = { - { VertexArrayRangeNV_names, VertexArrayRangeNV_remap_index, -1 }, - { FlushVertexArrayRangeNV_names, FlushVertexArrayRangeNV_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_NV_vertex_program) -static const struct dri_extension_function GL_NV_vertex_program_functions[] = { - { VertexAttrib4ubvNV_names, VertexAttrib4ubvNV_remap_index, -1 }, - { VertexAttrib4svNV_names, VertexAttrib4svNV_remap_index, -1 }, - { VertexAttribs3fvNV_names, VertexAttribs3fvNV_remap_index, -1 }, - { VertexAttribs1dvNV_names, VertexAttribs1dvNV_remap_index, -1 }, - { VertexAttrib1fvNV_names, VertexAttrib1fvNV_remap_index, -1 }, - { VertexAttrib4fNV_names, VertexAttrib4fNV_remap_index, -1 }, - { VertexAttrib2dNV_names, VertexAttrib2dNV_remap_index, -1 }, - { VertexAttrib4ubNV_names, VertexAttrib4ubNV_remap_index, -1 }, - { VertexAttribs3dvNV_names, VertexAttribs3dvNV_remap_index, -1 }, - { VertexAttribs4fvNV_names, VertexAttribs4fvNV_remap_index, -1 }, - { VertexAttrib2sNV_names, VertexAttrib2sNV_remap_index, -1 }, - { ProgramEnvParameter4fvARB_names, ProgramEnvParameter4fvARB_remap_index, -1 }, - { LoadProgramNV_names, LoadProgramNV_remap_index, -1 }, - { VertexAttrib4fvNV_names, VertexAttrib4fvNV_remap_index, -1 }, - { VertexAttrib3fNV_names, VertexAttrib3fNV_remap_index, -1 }, - { VertexAttribs2dvNV_names, VertexAttribs2dvNV_remap_index, -1 }, - { GetProgramParameterfvNV_names, GetProgramParameterfvNV_remap_index, -1 }, - { VertexAttrib3dNV_names, VertexAttrib3dNV_remap_index, -1 }, - { VertexAttrib2fvNV_names, VertexAttrib2fvNV_remap_index, -1 }, - { VertexAttrib2dvNV_names, VertexAttrib2dvNV_remap_index, -1 }, - { VertexAttrib1dvNV_names, VertexAttrib1dvNV_remap_index, -1 }, - { VertexAttrib1svNV_names, VertexAttrib1svNV_remap_index, -1 }, - { ProgramEnvParameter4fARB_names, ProgramEnvParameter4fARB_remap_index, -1 }, - { VertexAttribs2svNV_names, VertexAttribs2svNV_remap_index, -1 }, - { GetVertexAttribivNV_names, GetVertexAttribivNV_remap_index, -1 }, - { GetVertexAttribfvNV_names, GetVertexAttribfvNV_remap_index, -1 }, - { VertexAttrib2svNV_names, VertexAttrib2svNV_remap_index, -1 }, - { VertexAttribs1fvNV_names, VertexAttribs1fvNV_remap_index, -1 }, - { IsProgramNV_names, IsProgramNV_remap_index, -1 }, - { ProgramEnvParameter4dARB_names, ProgramEnvParameter4dARB_remap_index, -1 }, - { VertexAttrib2fNV_names, VertexAttrib2fNV_remap_index, -1 }, - { RequestResidentProgramsNV_names, RequestResidentProgramsNV_remap_index, -1 }, - { ExecuteProgramNV_names, ExecuteProgramNV_remap_index, -1 }, - { VertexAttribPointerNV_names, VertexAttribPointerNV_remap_index, -1 }, - { TrackMatrixNV_names, TrackMatrixNV_remap_index, -1 }, - { GetProgramParameterdvNV_names, GetProgramParameterdvNV_remap_index, -1 }, - { GetTrackMatrixivNV_names, GetTrackMatrixivNV_remap_index, -1 }, - { VertexAttrib3svNV_names, VertexAttrib3svNV_remap_index, -1 }, - { ProgramParameters4fvNV_names, ProgramParameters4fvNV_remap_index, -1 }, - { GetProgramivNV_names, GetProgramivNV_remap_index, -1 }, - { GetVertexAttribdvNV_names, GetVertexAttribdvNV_remap_index, -1 }, - { VertexAttrib3fvNV_names, VertexAttrib3fvNV_remap_index, -1 }, - { ProgramEnvParameter4dvARB_names, ProgramEnvParameter4dvARB_remap_index, -1 }, - { VertexAttribs2fvNV_names, VertexAttribs2fvNV_remap_index, -1 }, - { DeleteProgramsNV_names, DeleteProgramsNV_remap_index, -1 }, - { GetVertexAttribPointervNV_names, GetVertexAttribPointervNV_remap_index, -1 }, - { GetProgramStringNV_names, GetProgramStringNV_remap_index, -1 }, - { VertexAttrib4sNV_names, VertexAttrib4sNV_remap_index, -1 }, - { VertexAttribs4dvNV_names, VertexAttribs4dvNV_remap_index, -1 }, - { ProgramParameters4dvNV_names, ProgramParameters4dvNV_remap_index, -1 }, - { VertexAttrib3sNV_names, VertexAttrib3sNV_remap_index, -1 }, - { VertexAttrib1fNV_names, VertexAttrib1fNV_remap_index, -1 }, - { VertexAttrib4dNV_names, VertexAttrib4dNV_remap_index, -1 }, - { VertexAttribs4ubvNV_names, VertexAttribs4ubvNV_remap_index, -1 }, - { VertexAttribs3svNV_names, VertexAttribs3svNV_remap_index, -1 }, - { VertexAttrib1sNV_names, VertexAttrib1sNV_remap_index, -1 }, - { BindProgramNV_names, BindProgramNV_remap_index, -1 }, - { AreProgramsResidentNV_names, AreProgramsResidentNV_remap_index, -1 }, - { VertexAttrib3dvNV_names, VertexAttrib3dvNV_remap_index, -1 }, - { VertexAttrib1dNV_names, VertexAttrib1dNV_remap_index, -1 }, - { VertexAttribs4svNV_names, VertexAttribs4svNV_remap_index, -1 }, - { VertexAttribs1svNV_names, VertexAttribs1svNV_remap_index, -1 }, - { GenProgramsNV_names, GenProgramsNV_remap_index, -1 }, - { VertexAttrib4dvNV_names, VertexAttrib4dvNV_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_PGI_misc_hints) -static const struct dri_extension_function GL_PGI_misc_hints_functions[] = { - { HintPGI_names, HintPGI_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SGIS_detail_texture) -static const struct dri_extension_function GL_SGIS_detail_texture_functions[] = { - { GetDetailTexFuncSGIS_names, GetDetailTexFuncSGIS_remap_index, -1 }, - { DetailTexFuncSGIS_names, DetailTexFuncSGIS_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SGIS_fog_function) -static const struct dri_extension_function GL_SGIS_fog_function_functions[] = { - { FogFuncSGIS_names, FogFuncSGIS_remap_index, -1 }, - { GetFogFuncSGIS_names, GetFogFuncSGIS_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SGIS_multisample) -static const struct dri_extension_function GL_SGIS_multisample_functions[] = { - { SampleMaskSGIS_names, SampleMaskSGIS_remap_index, -1 }, - { SamplePatternSGIS_names, SamplePatternSGIS_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SGIS_pixel_texture) -static const struct dri_extension_function GL_SGIS_pixel_texture_functions[] = { - { PixelTexGenParameterfvSGIS_names, PixelTexGenParameterfvSGIS_remap_index, -1 }, - { GetPixelTexGenParameterivSGIS_names, GetPixelTexGenParameterivSGIS_remap_index, -1 }, - { PixelTexGenParameteriSGIS_names, PixelTexGenParameteriSGIS_remap_index, -1 }, - { PixelTexGenParameterivSGIS_names, PixelTexGenParameterivSGIS_remap_index, -1 }, - { PixelTexGenParameterfSGIS_names, PixelTexGenParameterfSGIS_remap_index, -1 }, - { GetPixelTexGenParameterfvSGIS_names, GetPixelTexGenParameterfvSGIS_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SGIS_point_parameters) -static const struct dri_extension_function GL_SGIS_point_parameters_functions[] = { - { PointParameterfEXT_names, PointParameterfEXT_remap_index, -1 }, - { PointParameterfvEXT_names, PointParameterfvEXT_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SGIS_sharpen_texture) -static const struct dri_extension_function GL_SGIS_sharpen_texture_functions[] = { - { GetSharpenTexFuncSGIS_names, GetSharpenTexFuncSGIS_remap_index, -1 }, - { SharpenTexFuncSGIS_names, SharpenTexFuncSGIS_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SGIS_texture4D) -static const struct dri_extension_function GL_SGIS_texture4D_functions[] = { - { TexImage4DSGIS_names, TexImage4DSGIS_remap_index, -1 }, - { TexSubImage4DSGIS_names, TexSubImage4DSGIS_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SGIS_texture_color_mask) -static const struct dri_extension_function GL_SGIS_texture_color_mask_functions[] = { - { TextureColorMaskSGIS_names, TextureColorMaskSGIS_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SGIS_texture_filter4) -static const struct dri_extension_function GL_SGIS_texture_filter4_functions[] = { - { GetTexFilterFuncSGIS_names, GetTexFilterFuncSGIS_remap_index, -1 }, - { TexFilterFuncSGIS_names, TexFilterFuncSGIS_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SGIX_async) -static const struct dri_extension_function GL_SGIX_async_functions[] = { - { AsyncMarkerSGIX_names, AsyncMarkerSGIX_remap_index, -1 }, - { FinishAsyncSGIX_names, FinishAsyncSGIX_remap_index, -1 }, - { PollAsyncSGIX_names, PollAsyncSGIX_remap_index, -1 }, - { DeleteAsyncMarkersSGIX_names, DeleteAsyncMarkersSGIX_remap_index, -1 }, - { IsAsyncMarkerSGIX_names, IsAsyncMarkerSGIX_remap_index, -1 }, - { GenAsyncMarkersSGIX_names, GenAsyncMarkersSGIX_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SGIX_flush_raster) -static const struct dri_extension_function GL_SGIX_flush_raster_functions[] = { - { FlushRasterSGIX_names, FlushRasterSGIX_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SGIX_fragment_lighting) -static const struct dri_extension_function GL_SGIX_fragment_lighting_functions[] = { - { FragmentMaterialfvSGIX_names, FragmentMaterialfvSGIX_remap_index, -1 }, - { FragmentLightModelivSGIX_names, FragmentLightModelivSGIX_remap_index, -1 }, - { FragmentLightiSGIX_names, FragmentLightiSGIX_remap_index, -1 }, - { GetFragmentMaterialfvSGIX_names, GetFragmentMaterialfvSGIX_remap_index, -1 }, - { FragmentMaterialfSGIX_names, FragmentMaterialfSGIX_remap_index, -1 }, - { GetFragmentLightivSGIX_names, GetFragmentLightivSGIX_remap_index, -1 }, - { FragmentLightModeliSGIX_names, FragmentLightModeliSGIX_remap_index, -1 }, - { FragmentLightivSGIX_names, FragmentLightivSGIX_remap_index, -1 }, - { GetFragmentMaterialivSGIX_names, GetFragmentMaterialivSGIX_remap_index, -1 }, - { FragmentLightModelfSGIX_names, FragmentLightModelfSGIX_remap_index, -1 }, - { FragmentColorMaterialSGIX_names, FragmentColorMaterialSGIX_remap_index, -1 }, - { FragmentMaterialiSGIX_names, FragmentMaterialiSGIX_remap_index, -1 }, - { LightEnviSGIX_names, LightEnviSGIX_remap_index, -1 }, - { FragmentLightModelfvSGIX_names, FragmentLightModelfvSGIX_remap_index, -1 }, - { FragmentLightfvSGIX_names, FragmentLightfvSGIX_remap_index, -1 }, - { FragmentLightfSGIX_names, FragmentLightfSGIX_remap_index, -1 }, - { GetFragmentLightfvSGIX_names, GetFragmentLightfvSGIX_remap_index, -1 }, - { FragmentMaterialivSGIX_names, FragmentMaterialivSGIX_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SGIX_framezoom) -static const struct dri_extension_function GL_SGIX_framezoom_functions[] = { - { FrameZoomSGIX_names, FrameZoomSGIX_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SGIX_igloo_interface) -static const struct dri_extension_function GL_SGIX_igloo_interface_functions[] = { - { IglooInterfaceSGIX_names, IglooInterfaceSGIX_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SGIX_instruments) -static const struct dri_extension_function GL_SGIX_instruments_functions[] = { - { ReadInstrumentsSGIX_names, ReadInstrumentsSGIX_remap_index, -1 }, - { PollInstrumentsSGIX_names, PollInstrumentsSGIX_remap_index, -1 }, - { GetInstrumentsSGIX_names, GetInstrumentsSGIX_remap_index, -1 }, - { StartInstrumentsSGIX_names, StartInstrumentsSGIX_remap_index, -1 }, - { StopInstrumentsSGIX_names, StopInstrumentsSGIX_remap_index, -1 }, - { InstrumentsBufferSGIX_names, InstrumentsBufferSGIX_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SGIX_list_priority) -static const struct dri_extension_function GL_SGIX_list_priority_functions[] = { - { ListParameterfSGIX_names, ListParameterfSGIX_remap_index, -1 }, - { GetListParameterfvSGIX_names, GetListParameterfvSGIX_remap_index, -1 }, - { ListParameteriSGIX_names, ListParameteriSGIX_remap_index, -1 }, - { ListParameterfvSGIX_names, ListParameterfvSGIX_remap_index, -1 }, - { ListParameterivSGIX_names, ListParameterivSGIX_remap_index, -1 }, - { GetListParameterivSGIX_names, GetListParameterivSGIX_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SGIX_pixel_texture) -static const struct dri_extension_function GL_SGIX_pixel_texture_functions[] = { - { PixelTexGenSGIX_names, PixelTexGenSGIX_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SGIX_polynomial_ffd) -static const struct dri_extension_function GL_SGIX_polynomial_ffd_functions[] = { - { LoadIdentityDeformationMapSGIX_names, LoadIdentityDeformationMapSGIX_remap_index, -1 }, - { DeformationMap3dSGIX_names, DeformationMap3dSGIX_remap_index, -1 }, - { DeformSGIX_names, DeformSGIX_remap_index, -1 }, - { DeformationMap3fSGIX_names, DeformationMap3fSGIX_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SGIX_reference_plane) -static const struct dri_extension_function GL_SGIX_reference_plane_functions[] = { - { ReferencePlaneSGIX_names, ReferencePlaneSGIX_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SGIX_sprite) -static const struct dri_extension_function GL_SGIX_sprite_functions[] = { - { SpriteParameterfvSGIX_names, SpriteParameterfvSGIX_remap_index, -1 }, - { SpriteParameteriSGIX_names, SpriteParameteriSGIX_remap_index, -1 }, - { SpriteParameterfSGIX_names, SpriteParameterfSGIX_remap_index, -1 }, - { SpriteParameterivSGIX_names, SpriteParameterivSGIX_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SGIX_tag_sample_buffer) -static const struct dri_extension_function GL_SGIX_tag_sample_buffer_functions[] = { - { TagSampleBufferSGIX_names, TagSampleBufferSGIX_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SGI_color_table) -static const struct dri_extension_function GL_SGI_color_table_functions[] = { - { ColorTableParameteriv_names, -1, 341 }, - { ColorTable_names, -1, 339 }, - { GetColorTable_names, -1, 343 }, - { CopyColorTable_names, -1, 342 }, - { ColorTableParameterfv_names, -1, 340 }, - { GetColorTableParameterfv_names, -1, 344 }, - { GetColorTableParameteriv_names, -1, 345 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SUNX_constant_data) -static const struct dri_extension_function GL_SUNX_constant_data_functions[] = { - { FinishTextureSUNX_names, FinishTextureSUNX_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SUN_global_alpha) -static const struct dri_extension_function GL_SUN_global_alpha_functions[] = { - { GlobalAlphaFactorubSUN_names, GlobalAlphaFactorubSUN_remap_index, -1 }, - { GlobalAlphaFactoriSUN_names, GlobalAlphaFactoriSUN_remap_index, -1 }, - { GlobalAlphaFactordSUN_names, GlobalAlphaFactordSUN_remap_index, -1 }, - { GlobalAlphaFactoruiSUN_names, GlobalAlphaFactoruiSUN_remap_index, -1 }, - { GlobalAlphaFactorbSUN_names, GlobalAlphaFactorbSUN_remap_index, -1 }, - { GlobalAlphaFactorfSUN_names, GlobalAlphaFactorfSUN_remap_index, -1 }, - { GlobalAlphaFactorusSUN_names, GlobalAlphaFactorusSUN_remap_index, -1 }, - { GlobalAlphaFactorsSUN_names, GlobalAlphaFactorsSUN_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SUN_mesh_array) -static const struct dri_extension_function GL_SUN_mesh_array_functions[] = { - { DrawMeshArraysSUN_names, DrawMeshArraysSUN_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SUN_triangle_list) -static const struct dri_extension_function GL_SUN_triangle_list_functions[] = { - { ReplacementCodeubSUN_names, ReplacementCodeubSUN_remap_index, -1 }, - { ReplacementCodeubvSUN_names, ReplacementCodeubvSUN_remap_index, -1 }, - { ReplacementCodeusvSUN_names, ReplacementCodeusvSUN_remap_index, -1 }, - { ReplacementCodePointerSUN_names, ReplacementCodePointerSUN_remap_index, -1 }, - { ReplacementCodeusSUN_names, ReplacementCodeusSUN_remap_index, -1 }, - { ReplacementCodeuiSUN_names, ReplacementCodeuiSUN_remap_index, -1 }, - { ReplacementCodeuivSUN_names, ReplacementCodeuivSUN_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_SUN_vertex) -static const struct dri_extension_function GL_SUN_vertex_functions[] = { - { ReplacementCodeuiColor3fVertex3fvSUN_names, ReplacementCodeuiColor3fVertex3fvSUN_remap_index, -1 }, - { TexCoord4fColor4fNormal3fVertex4fvSUN_names, TexCoord4fColor4fNormal3fVertex4fvSUN_remap_index, -1 }, - { TexCoord2fColor4ubVertex3fvSUN_names, TexCoord2fColor4ubVertex3fvSUN_remap_index, -1 }, - { ReplacementCodeuiVertex3fvSUN_names, ReplacementCodeuiVertex3fvSUN_remap_index, -1 }, - { ReplacementCodeuiTexCoord2fVertex3fvSUN_names, ReplacementCodeuiTexCoord2fVertex3fvSUN_remap_index, -1 }, - { ReplacementCodeuiNormal3fVertex3fSUN_names, ReplacementCodeuiNormal3fVertex3fSUN_remap_index, -1 }, - { Color4ubVertex3fvSUN_names, Color4ubVertex3fvSUN_remap_index, -1 }, - { Color4ubVertex3fSUN_names, Color4ubVertex3fSUN_remap_index, -1 }, - { TexCoord2fVertex3fSUN_names, TexCoord2fVertex3fSUN_remap_index, -1 }, - { TexCoord2fColor4fNormal3fVertex3fSUN_names, TexCoord2fColor4fNormal3fVertex3fSUN_remap_index, -1 }, - { TexCoord2fNormal3fVertex3fvSUN_names, TexCoord2fNormal3fVertex3fvSUN_remap_index, -1 }, - { ReplacementCodeuiTexCoord2fNormal3fVertex3fSUN_names, ReplacementCodeuiTexCoord2fNormal3fVertex3fSUN_remap_index, -1 }, - { ReplacementCodeuiTexCoord2fVertex3fSUN_names, ReplacementCodeuiTexCoord2fVertex3fSUN_remap_index, -1 }, - { TexCoord2fNormal3fVertex3fSUN_names, TexCoord2fNormal3fVertex3fSUN_remap_index, -1 }, - { Color3fVertex3fSUN_names, Color3fVertex3fSUN_remap_index, -1 }, - { Color3fVertex3fvSUN_names, Color3fVertex3fvSUN_remap_index, -1 }, - { Color4fNormal3fVertex3fvSUN_names, Color4fNormal3fVertex3fvSUN_remap_index, -1 }, - { ReplacementCodeuiTexCoord2fColor4fNormal3fVertex3fvSUN_names, ReplacementCodeuiTexCoord2fColor4fNormal3fVertex3fvSUN_remap_index, -1 }, - { ReplacementCodeuiColor4fNormal3fVertex3fvSUN_names, ReplacementCodeuiColor4fNormal3fVertex3fvSUN_remap_index, -1 }, - { ReplacementCodeuiTexCoord2fNormal3fVertex3fvSUN_names, ReplacementCodeuiTexCoord2fNormal3fVertex3fvSUN_remap_index, -1 }, - { TexCoord2fColor3fVertex3fSUN_names, TexCoord2fColor3fVertex3fSUN_remap_index, -1 }, - { TexCoord4fColor4fNormal3fVertex4fSUN_names, TexCoord4fColor4fNormal3fVertex4fSUN_remap_index, -1 }, - { Color4ubVertex2fvSUN_names, Color4ubVertex2fvSUN_remap_index, -1 }, - { Normal3fVertex3fSUN_names, Normal3fVertex3fSUN_remap_index, -1 }, - { ReplacementCodeuiColor4fNormal3fVertex3fSUN_names, ReplacementCodeuiColor4fNormal3fVertex3fSUN_remap_index, -1 }, - { TexCoord2fColor4fNormal3fVertex3fvSUN_names, TexCoord2fColor4fNormal3fVertex3fvSUN_remap_index, -1 }, - { TexCoord2fVertex3fvSUN_names, TexCoord2fVertex3fvSUN_remap_index, -1 }, - { Color4ubVertex2fSUN_names, Color4ubVertex2fSUN_remap_index, -1 }, - { ReplacementCodeuiColor4ubVertex3fSUN_names, ReplacementCodeuiColor4ubVertex3fSUN_remap_index, -1 }, - { TexCoord2fColor4ubVertex3fSUN_names, TexCoord2fColor4ubVertex3fSUN_remap_index, -1 }, - { Normal3fVertex3fvSUN_names, Normal3fVertex3fvSUN_remap_index, -1 }, - { Color4fNormal3fVertex3fSUN_names, Color4fNormal3fVertex3fSUN_remap_index, -1 }, - { ReplacementCodeuiTexCoord2fColor4fNormal3fVertex3fSUN_names, ReplacementCodeuiTexCoord2fColor4fNormal3fVertex3fSUN_remap_index, -1 }, - { ReplacementCodeuiColor4ubVertex3fvSUN_names, ReplacementCodeuiColor4ubVertex3fvSUN_remap_index, -1 }, - { ReplacementCodeuiColor3fVertex3fSUN_names, ReplacementCodeuiColor3fVertex3fSUN_remap_index, -1 }, - { TexCoord4fVertex4fSUN_names, TexCoord4fVertex4fSUN_remap_index, -1 }, - { TexCoord2fColor3fVertex3fvSUN_names, TexCoord2fColor3fVertex3fvSUN_remap_index, -1 }, - { ReplacementCodeuiNormal3fVertex3fvSUN_names, ReplacementCodeuiNormal3fVertex3fvSUN_remap_index, -1 }, - { TexCoord4fVertex4fvSUN_names, TexCoord4fVertex4fvSUN_remap_index, -1 }, - { ReplacementCodeuiVertex3fSUN_names, ReplacementCodeuiVertex3fSUN_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_VERSION_1_3) -static const struct dri_extension_function GL_VERSION_1_3_functions[] = { - { SampleCoverageARB_names, SampleCoverageARB_remap_index, -1 }, - { MultiTexCoord3sARB_names, -1, 398 }, - { ActiveTextureARB_names, -1, 374 }, - { CompressedTexSubImage2DARB_names, CompressedTexSubImage2DARB_remap_index, -1 }, - { CompressedTexImage3DARB_names, CompressedTexImage3DARB_remap_index, -1 }, - { MultiTexCoord1fvARB_names, -1, 379 }, - { MultTransposeMatrixdARB_names, MultTransposeMatrixdARB_remap_index, -1 }, - { CompressedTexImage1DARB_names, CompressedTexImage1DARB_remap_index, -1 }, - { MultiTexCoord3dARB_names, -1, 392 }, - { MultiTexCoord2iARB_names, -1, 388 }, - { MultiTexCoord2svARB_names, -1, 391 }, - { MultiTexCoord2fARB_names, -1, 386 }, - { LoadTransposeMatrixdARB_names, LoadTransposeMatrixdARB_remap_index, -1 }, - { MultiTexCoord3fvARB_names, -1, 395 }, - { MultiTexCoord4sARB_names, -1, 406 }, - { MultiTexCoord2dvARB_names, -1, 385 }, - { MultiTexCoord1svARB_names, -1, 383 }, - { MultiTexCoord3svARB_names, -1, 399 }, - { MultiTexCoord4iARB_names, -1, 404 }, - { MultiTexCoord3iARB_names, -1, 396 }, - { MultiTexCoord1dARB_names, -1, 376 }, - { MultiTexCoord3dvARB_names, -1, 393 }, - { MultiTexCoord3ivARB_names, -1, 397 }, - { MultiTexCoord2sARB_names, -1, 390 }, - { MultiTexCoord4ivARB_names, -1, 405 }, - { CompressedTexSubImage1DARB_names, CompressedTexSubImage1DARB_remap_index, -1 }, - { ClientActiveTextureARB_names, -1, 375 }, - { CompressedTexSubImage3DARB_names, CompressedTexSubImage3DARB_remap_index, -1 }, - { MultiTexCoord2dARB_names, -1, 384 }, - { MultiTexCoord4dvARB_names, -1, 401 }, - { MultiTexCoord4fvARB_names, -1, 403 }, - { MultiTexCoord3fARB_names, -1, 394 }, - { MultTransposeMatrixfARB_names, MultTransposeMatrixfARB_remap_index, -1 }, - { CompressedTexImage2DARB_names, CompressedTexImage2DARB_remap_index, -1 }, - { MultiTexCoord4dARB_names, -1, 400 }, - { MultiTexCoord1sARB_names, -1, 382 }, - { MultiTexCoord1dvARB_names, -1, 377 }, - { MultiTexCoord1ivARB_names, -1, 381 }, - { MultiTexCoord2ivARB_names, -1, 389 }, - { MultiTexCoord1iARB_names, -1, 380 }, - { GetCompressedTexImageARB_names, GetCompressedTexImageARB_remap_index, -1 }, - { MultiTexCoord4svARB_names, -1, 407 }, - { MultiTexCoord1fARB_names, -1, 378 }, - { MultiTexCoord4fARB_names, -1, 402 }, - { LoadTransposeMatrixfARB_names, LoadTransposeMatrixfARB_remap_index, -1 }, - { MultiTexCoord2fvARB_names, -1, 387 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_VERSION_1_4) -static const struct dri_extension_function GL_VERSION_1_4_functions[] = { - { PointParameteriNV_names, PointParameteriNV_remap_index, -1 }, - { SecondaryColor3iEXT_names, SecondaryColor3iEXT_remap_index, -1 }, - { WindowPos3fMESA_names, WindowPos3fMESA_remap_index, -1 }, - { WindowPos2dvMESA_names, WindowPos2dvMESA_remap_index, -1 }, - { SecondaryColor3bEXT_names, SecondaryColor3bEXT_remap_index, -1 }, - { PointParameterfEXT_names, PointParameterfEXT_remap_index, -1 }, - { FogCoorddEXT_names, FogCoorddEXT_remap_index, -1 }, - { FogCoordfEXT_names, FogCoordfEXT_remap_index, -1 }, - { WindowPos2svMESA_names, WindowPos2svMESA_remap_index, -1 }, - { WindowPos3dMESA_names, WindowPos3dMESA_remap_index, -1 }, - { PointParameterfvEXT_names, PointParameterfvEXT_remap_index, -1 }, - { WindowPos2fvMESA_names, WindowPos2fvMESA_remap_index, -1 }, - { SecondaryColor3bvEXT_names, SecondaryColor3bvEXT_remap_index, -1 }, - { SecondaryColor3sEXT_names, SecondaryColor3sEXT_remap_index, -1 }, - { SecondaryColor3dEXT_names, SecondaryColor3dEXT_remap_index, -1 }, - { WindowPos2dMESA_names, WindowPos2dMESA_remap_index, -1 }, - { SecondaryColorPointerEXT_names, SecondaryColorPointerEXT_remap_index, -1 }, - { SecondaryColor3uiEXT_names, SecondaryColor3uiEXT_remap_index, -1 }, - { SecondaryColor3usvEXT_names, SecondaryColor3usvEXT_remap_index, -1 }, - { WindowPos3dvMESA_names, WindowPos3dvMESA_remap_index, -1 }, - { PointParameterivNV_names, PointParameterivNV_remap_index, -1 }, - { WindowPos3fvMESA_names, WindowPos3fvMESA_remap_index, -1 }, - { SecondaryColor3ivEXT_names, SecondaryColor3ivEXT_remap_index, -1 }, - { WindowPos2iMESA_names, WindowPos2iMESA_remap_index, -1 }, - { SecondaryColor3fvEXT_names, SecondaryColor3fvEXT_remap_index, -1 }, - { WindowPos3sMESA_names, WindowPos3sMESA_remap_index, -1 }, - { WindowPos2ivMESA_names, WindowPos2ivMESA_remap_index, -1 }, - { MultiDrawElementsEXT_names, MultiDrawElementsEXT_remap_index, -1 }, - { WindowPos2sMESA_names, WindowPos2sMESA_remap_index, -1 }, - { FogCoordPointerEXT_names, FogCoordPointerEXT_remap_index, -1 }, - { SecondaryColor3ubvEXT_names, SecondaryColor3ubvEXT_remap_index, -1 }, - { SecondaryColor3uivEXT_names, SecondaryColor3uivEXT_remap_index, -1 }, - { WindowPos3iMESA_names, WindowPos3iMESA_remap_index, -1 }, - { SecondaryColor3dvEXT_names, SecondaryColor3dvEXT_remap_index, -1 }, - { MultiDrawArraysEXT_names, MultiDrawArraysEXT_remap_index, -1 }, - { SecondaryColor3usEXT_names, SecondaryColor3usEXT_remap_index, -1 }, - { FogCoordfvEXT_names, FogCoordfvEXT_remap_index, -1 }, - { SecondaryColor3ubEXT_names, SecondaryColor3ubEXT_remap_index, -1 }, - { BlendFuncSeparateEXT_names, BlendFuncSeparateEXT_remap_index, -1 }, - { SecondaryColor3fEXT_names, SecondaryColor3fEXT_remap_index, -1 }, - { WindowPos3ivMESA_names, WindowPos3ivMESA_remap_index, -1 }, - { SecondaryColor3svEXT_names, SecondaryColor3svEXT_remap_index, -1 }, - { FogCoorddvEXT_names, FogCoorddvEXT_remap_index, -1 }, - { WindowPos3svMESA_names, WindowPos3svMESA_remap_index, -1 }, - { WindowPos2fMESA_names, WindowPos2fMESA_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_VERSION_1_5) -static const struct dri_extension_function GL_VERSION_1_5_functions[] = { - { BeginQueryARB_names, BeginQueryARB_remap_index, -1 }, - { GetBufferSubDataARB_names, GetBufferSubDataARB_remap_index, -1 }, - { BufferSubDataARB_names, BufferSubDataARB_remap_index, -1 }, - { GetQueryivARB_names, GetQueryivARB_remap_index, -1 }, - { GetQueryObjectivARB_names, GetQueryObjectivARB_remap_index, -1 }, - { BufferDataARB_names, BufferDataARB_remap_index, -1 }, - { EndQueryARB_names, EndQueryARB_remap_index, -1 }, - { GetBufferPointervARB_names, GetBufferPointervARB_remap_index, -1 }, - { GetQueryObjectuivARB_names, GetQueryObjectuivARB_remap_index, -1 }, - { GetBufferParameterivARB_names, GetBufferParameterivARB_remap_index, -1 }, - { DeleteQueriesARB_names, DeleteQueriesARB_remap_index, -1 }, - { IsQueryARB_names, IsQueryARB_remap_index, -1 }, - { MapBufferARB_names, MapBufferARB_remap_index, -1 }, - { GenQueriesARB_names, GenQueriesARB_remap_index, -1 }, - { IsBufferARB_names, IsBufferARB_remap_index, -1 }, - { DeleteBuffersARB_names, DeleteBuffersARB_remap_index, -1 }, - { UnmapBufferARB_names, UnmapBufferARB_remap_index, -1 }, - { BindBufferARB_names, BindBufferARB_remap_index, -1 }, - { GenBuffersARB_names, GenBuffersARB_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_VERSION_2_0) -static const struct dri_extension_function GL_VERSION_2_0_functions[] = { - { UniformMatrix3fvARB_names, UniformMatrix3fvARB_remap_index, -1 }, - { GetProgramiv_names, GetProgramiv_remap_index, -1 }, - { BlendEquationSeparateEXT_names, BlendEquationSeparateEXT_remap_index, -1 }, - { AttachShader_names, AttachShader_remap_index, -1 }, - { VertexAttrib2fARB_names, VertexAttrib2fARB_remap_index, -1 }, - { VertexAttrib3fARB_names, VertexAttrib3fARB_remap_index, -1 }, - { Uniform2fARB_names, Uniform2fARB_remap_index, -1 }, - { VertexAttrib1svARB_names, VertexAttrib1svARB_remap_index, -1 }, - { Uniform2ivARB_names, Uniform2ivARB_remap_index, -1 }, - { UniformMatrix4fvARB_names, UniformMatrix4fvARB_remap_index, -1 }, - { VertexAttrib4NusvARB_names, VertexAttrib4NusvARB_remap_index, -1 }, - { DisableVertexAttribArrayARB_names, DisableVertexAttribArrayARB_remap_index, -1 }, - { StencilMaskSeparate_names, StencilMaskSeparate_remap_index, -1 }, - { VertexAttrib1fARB_names, VertexAttrib1fARB_remap_index, -1 }, - { GetProgramInfoLog_names, GetProgramInfoLog_remap_index, -1 }, - { VertexAttrib4NbvARB_names, VertexAttrib4NbvARB_remap_index, -1 }, - { GetActiveAttribARB_names, GetActiveAttribARB_remap_index, -1 }, - { Uniform3iARB_names, Uniform3iARB_remap_index, -1 }, - { GetShaderInfoLog_names, GetShaderInfoLog_remap_index, -1 }, - { VertexAttrib1sARB_names, VertexAttrib1sARB_remap_index, -1 }, - { Uniform1fARB_names, Uniform1fARB_remap_index, -1 }, - { StencilOpSeparate_names, StencilOpSeparate_remap_index, -1 }, - { UniformMatrix2fvARB_names, UniformMatrix2fvARB_remap_index, -1 }, - { VertexAttrib3dvARB_names, VertexAttrib3dvARB_remap_index, -1 }, - { Uniform3fvARB_names, Uniform3fvARB_remap_index, -1 }, - { GetVertexAttribivARB_names, GetVertexAttribivARB_remap_index, -1 }, - { CreateProgram_names, CreateProgram_remap_index, -1 }, - { StencilFuncSeparate_names, StencilFuncSeparate_remap_index, -1 }, - { VertexAttrib4ivARB_names, VertexAttrib4ivARB_remap_index, -1 }, - { VertexAttrib4bvARB_names, VertexAttrib4bvARB_remap_index, -1 }, - { VertexAttrib3dARB_names, VertexAttrib3dARB_remap_index, -1 }, - { VertexAttrib4fARB_names, VertexAttrib4fARB_remap_index, -1 }, - { VertexAttrib4fvARB_names, VertexAttrib4fvARB_remap_index, -1 }, - { GetActiveUniformARB_names, GetActiveUniformARB_remap_index, -1 }, - { IsShader_names, IsShader_remap_index, -1 }, - { GetUniformivARB_names, GetUniformivARB_remap_index, -1 }, - { IsProgram_names, IsProgram_remap_index, -1 }, - { Uniform2fvARB_names, Uniform2fvARB_remap_index, -1 }, - { ValidateProgramARB_names, ValidateProgramARB_remap_index, -1 }, - { VertexAttrib2dARB_names, VertexAttrib2dARB_remap_index, -1 }, - { VertexAttrib1dvARB_names, VertexAttrib1dvARB_remap_index, -1 }, - { GetVertexAttribfvARB_names, GetVertexAttribfvARB_remap_index, -1 }, - { GetAttribLocationARB_names, GetAttribLocationARB_remap_index, -1 }, - { VertexAttrib4ubvARB_names, VertexAttrib4ubvARB_remap_index, -1 }, - { Uniform3ivARB_names, Uniform3ivARB_remap_index, -1 }, - { VertexAttrib4sARB_names, VertexAttrib4sARB_remap_index, -1 }, - { VertexAttrib2dvARB_names, VertexAttrib2dvARB_remap_index, -1 }, - { VertexAttrib2fvARB_names, VertexAttrib2fvARB_remap_index, -1 }, - { VertexAttrib4NivARB_names, VertexAttrib4NivARB_remap_index, -1 }, - { GetAttachedShaders_names, GetAttachedShaders_remap_index, -1 }, - { CompileShaderARB_names, CompileShaderARB_remap_index, -1 }, - { DeleteShader_names, DeleteShader_remap_index, -1 }, - { Uniform3fARB_names, Uniform3fARB_remap_index, -1 }, - { VertexAttrib4NuivARB_names, VertexAttrib4NuivARB_remap_index, -1 }, - { Uniform4fARB_names, Uniform4fARB_remap_index, -1 }, - { VertexAttrib1dARB_names, VertexAttrib1dARB_remap_index, -1 }, - { VertexAttrib4usvARB_names, VertexAttrib4usvARB_remap_index, -1 }, - { LinkProgramARB_names, LinkProgramARB_remap_index, -1 }, - { ShaderSourceARB_names, ShaderSourceARB_remap_index, -1 }, - { VertexAttrib3svARB_names, VertexAttrib3svARB_remap_index, -1 }, - { Uniform4ivARB_names, Uniform4ivARB_remap_index, -1 }, - { GetVertexAttribdvARB_names, GetVertexAttribdvARB_remap_index, -1 }, - { Uniform1ivARB_names, Uniform1ivARB_remap_index, -1 }, - { VertexAttrib4dvARB_names, VertexAttrib4dvARB_remap_index, -1 }, - { BindAttribLocationARB_names, BindAttribLocationARB_remap_index, -1 }, - { Uniform1iARB_names, Uniform1iARB_remap_index, -1 }, - { VertexAttribPointerARB_names, VertexAttribPointerARB_remap_index, -1 }, - { VertexAttrib4NsvARB_names, VertexAttrib4NsvARB_remap_index, -1 }, - { VertexAttrib3fvARB_names, VertexAttrib3fvARB_remap_index, -1 }, - { CreateShader_names, CreateShader_remap_index, -1 }, - { DetachShader_names, DetachShader_remap_index, -1 }, - { VertexAttrib4NubARB_names, VertexAttrib4NubARB_remap_index, -1 }, - { Uniform4fvARB_names, Uniform4fvARB_remap_index, -1 }, - { GetUniformfvARB_names, GetUniformfvARB_remap_index, -1 }, - { Uniform4iARB_names, Uniform4iARB_remap_index, -1 }, - { UseProgramObjectARB_names, UseProgramObjectARB_remap_index, -1 }, - { DeleteProgram_names, DeleteProgram_remap_index, -1 }, - { GetVertexAttribPointervNV_names, GetVertexAttribPointervNV_remap_index, -1 }, - { Uniform2iARB_names, Uniform2iARB_remap_index, -1 }, - { VertexAttrib4dARB_names, VertexAttrib4dARB_remap_index, -1 }, - { GetUniformLocationARB_names, GetUniformLocationARB_remap_index, -1 }, - { VertexAttrib3sARB_names, VertexAttrib3sARB_remap_index, -1 }, - { GetShaderSourceARB_names, GetShaderSourceARB_remap_index, -1 }, - { DrawBuffersARB_names, DrawBuffersARB_remap_index, -1 }, - { Uniform1fvARB_names, Uniform1fvARB_remap_index, -1 }, - { EnableVertexAttribArrayARB_names, EnableVertexAttribArrayARB_remap_index, -1 }, - { VertexAttrib4uivARB_names, VertexAttrib4uivARB_remap_index, -1 }, - { VertexAttrib4svARB_names, VertexAttrib4svARB_remap_index, -1 }, - { GetShaderiv_names, GetShaderiv_remap_index, -1 }, - { VertexAttrib2svARB_names, VertexAttrib2svARB_remap_index, -1 }, - { VertexAttrib4NubvARB_names, VertexAttrib4NubvARB_remap_index, -1 }, - { VertexAttrib2sARB_names, VertexAttrib2sARB_remap_index, -1 }, - { VertexAttrib1fvARB_names, VertexAttrib1fvARB_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - -#if defined(need_GL_VERSION_2_1) -static const struct dri_extension_function GL_VERSION_2_1_functions[] = { - { UniformMatrix2x4fv_names, UniformMatrix2x4fv_remap_index, -1 }, - { UniformMatrix4x3fv_names, UniformMatrix4x3fv_remap_index, -1 }, - { UniformMatrix4x2fv_names, UniformMatrix4x2fv_remap_index, -1 }, - { UniformMatrix2x3fv_names, UniformMatrix2x3fv_remap_index, -1 }, - { UniformMatrix3x4fv_names, UniformMatrix3x4fv_remap_index, -1 }, - { UniformMatrix3x2fv_names, UniformMatrix3x2fv_remap_index, -1 }, - { NULL, 0, 0 } -}; -#endif - diff --git a/src/mesa/drivers/dri/common/spantmp2.h b/src/mesa/drivers/dri/common/spantmp2.h index 89c815722f6..447f3d15b95 100644 --- a/src/mesa/drivers/dri/common/spantmp2.h +++ b/src/mesa/drivers/dri/common/spantmp2.h @@ -82,6 +82,72 @@ rgba[3] = 0xff; \ } while (0) +#elif (SPANTMP_PIXEL_FMT == GL_RGB) && (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_SHORT_5_6_5_REV) + +/** + ** GL_RGB, GL_UNSIGNED_SHORT_5_6_5_REV + **/ + +#ifndef GET_VALUE +#ifndef GET_PTR +#define GET_PTR(_x, _y) (buf + (_x) * 2 + (_y) * pitch) +#endif + +#define GET_VALUE(_x, _y) *(volatile GLushort *)(GET_PTR(_x, _y)) +#define PUT_VALUE(_x, _y, _v) *(volatile GLushort *)(GET_PTR(_x, _y)) = (_v) +#endif /* GET_VALUE */ + +#define INIT_MONO_PIXEL(p, color) \ + p = PACK_COLOR_565_REV( color[0], color[1], color[2] ) + +#define WRITE_RGBA( _x, _y, r, g, b, a ) \ + PUT_VALUE(_x, _y, PACK_COLOR_565_REV( r, g, b )) + +#define WRITE_PIXEL( _x, _y, p ) PUT_VALUE(_x, _y, p) + +#define READ_RGBA( rgba, _x, _y ) \ + do { \ + GLushort p = GET_VALUE(_x, _y); \ + p = p << 8 | p >> 8; \ + rgba[0] = ((p >> 8) & 0xf8) * 255 / 0xf8; \ + rgba[1] = ((p >> 3) & 0xfc) * 255 / 0xfc; \ + rgba[2] = ((p << 3) & 0xf8) * 255 / 0xf8; \ + rgba[3] = 0xff; \ + } while (0) + +#elif (SPANTMP_PIXEL_FMT == GL_BGRA) && (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_SHORT_4_4_4_4) + +/** + ** GL_BGRA, GL_UNSIGNED_SHORT_4_4_4_4 + **/ + +#ifndef GET_VALUE +#ifndef GET_PTR +#define GET_PTR(_x, _y) (buf + (_x) * 2 + (_y) * pitch) +#endif + +#define GET_VALUE(_x, _y) *(volatile GLushort *)(GET_PTR(_x, _y)) +#define PUT_VALUE(_x, _y, _v) *(volatile GLushort *)(GET_PTR(_x, _y)) = (_v) +#endif /* GET_VALUE */ + +#define INIT_MONO_PIXEL(p, color) \ + p = PACK_COLOR_4444_REV(color[3], color[0], color[1], color[2]) + +#define WRITE_RGBA( _x, _y, r, g, b, a ) \ + PUT_VALUE(_x, _y, PACK_COLOR_4444_REV(a, r, g, b)) \ + +#define WRITE_PIXEL( _x, _y, p ) PUT_VALUE(_x, _y, p) + +#define READ_RGBA( rgba, _x, _y ) \ + do { \ + GLushort p = GET_VALUE(_x, _y); \ + rgba[0] = ((p >> 0) & 0xf) * 0x11; \ + rgba[1] = ((p >> 12) & 0xf) * 0x11; \ + rgba[2] = ((p >> 4) & 0xf) * 0x11; \ + rgba[3] = ((p >> 8) & 0xf) * 0x11; \ + } while (0) + + #elif (SPANTMP_PIXEL_FMT == GL_BGRA) && (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_SHORT_4_4_4_4_REV) /** @@ -147,6 +213,39 @@ rgba[3] = ((p >> 15) & 0x1) * 0xff; \ } while (0) +#elif (SPANTMP_PIXEL_FMT == GL_BGRA) && (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_SHORT_1_5_5_5) + +/** + ** GL_BGRA, GL_UNSIGNED_SHORT_1_5_5_5 + **/ + +#ifndef GET_VALUE +#ifndef GET_PTR +#define GET_PTR(_x, _y) (buf + (_x) * 2 + (_y) * pitch) +#endif + +#define GET_VALUE(_x, _y) *(volatile GLushort *)(GET_PTR(_x, _y)) +#define PUT_VALUE(_x, _y, _v) *(volatile GLushort *)(GET_PTR(_x, _y)) = (_v) +#endif /* GET_VALUE */ + +#define INIT_MONO_PIXEL(p, color) \ + p = PACK_COLOR_1555_REV(color[3], color[0], color[1], color[2]) + +#define WRITE_RGBA( _x, _y, r, g, b, a ) \ + PUT_VALUE(_x, _y, PACK_COLOR_1555_REV(a, r, g, b)) \ + +#define WRITE_PIXEL( _x, _y, p ) PUT_VALUE(_x, _y, p) + +#define READ_RGBA( rgba, _x, _y ) \ + do { \ + GLushort p = GET_VALUE(_x, _y); \ + p = p << 8 | p >> 8; \ + rgba[0] = ((p >> 7) & 0xf8) * 255 / 0xf8; \ + rgba[1] = ((p >> 2) & 0xf8) * 255 / 0xf8; \ + rgba[2] = ((p << 3) & 0xf8) * 255 / 0xf8; \ + rgba[3] = ((p >> 15) & 0x1) * 0xff; \ + } while (0) + #elif (SPANTMP_PIXEL_FMT == GL_BGRA) && (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_INT_8_8_8_8_REV) /** @@ -202,6 +301,118 @@ } while (0) # endif +#elif (SPANTMP_PIXEL_FMT == GL_BGRA) && (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_INT_8_8_8_8) + +/** + ** GL_BGRA, GL_UNSIGNED_INT_8_8_8_8 + **/ + +#ifndef GET_VALUE +#ifndef GET_PTR +#define GET_PTR(_x, _y) ( buf + (_x) * 4 + (_y) * pitch) +#endif + +#define GET_VALUE(_x, _y) *(volatile GLuint *)(GET_PTR(_x, _y)) +#define PUT_VALUE(_x, _y, _v) *(volatile GLuint *)(GET_PTR(_x, _y)) = (_v) +#endif /* GET_VALUE */ + +# define INIT_MONO_PIXEL(p, color) \ + p = PACK_COLOR_8888(color[2], color[1], color[0], color[3]) + +# define WRITE_RGBA(_x, _y, r, g, b, a) \ + PUT_VALUE(_x, _y, ((r << 8) | \ + (g << 16) | \ + (b << 24) | \ + (a << 0))) + +#define WRITE_PIXEL(_x, _y, p) PUT_VALUE(_x, _y, p) + +# if defined( USE_X86_ASM ) +# define READ_RGBA(rgba, _x, _y) \ + do { \ + GLuint p = GET_VALUE(_x, _y); \ + __asm__ __volatile__( "rorl $8, %0" \ + : "=r" (p) : "0" (p) ); \ + ((GLuint *)rgba)[0] = p; \ + } while (0) +# elif defined( MESA_BIG_ENDIAN ) + /* On PowerPC with GCC 3.4.2 the shift madness below becomes a single + * rotlwi instruction. It also produces good code on SPARC. + */ +# define READ_RGBA( rgba, _x, _y ) \ + do { \ + GLuint p = CPU_TO_LE32(GET_VALUE(_x, _y)); \ + GLuint t = p; \ + *((uint32_t *) rgba) = (t >> 24) | (p << 8); \ + } while (0) +# else +# define READ_RGBA( rgba, _x, _y ) \ + do { \ + GLuint p = GET_VALUE(_x, _y); \ + rgba[0] = (p >> 8) & 0xff; \ + rgba[1] = (p >> 16) & 0xff; \ + rgba[2] = (p >> 24) & 0xff; \ + rgba[3] = (p >> 0) & 0xff; \ + } while (0) +# endif + +#elif (SPANTMP_PIXEL_FMT == GL_BGR) && (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_INT_8_8_8_8_REV) + +/** + ** GL_BGR, GL_UNSIGNED_INT_8_8_8_8_REV + ** + ** This is really for MESA_FORMAT_XRGB8888. The spantmp code needs to be + ** kicked to the curb, and we need to just code-gen this. + **/ + +#ifndef GET_VALUE +#ifndef GET_PTR +#define GET_PTR(_x, _y) ( buf + (_x) * 4 + (_y) * pitch) +#endif + +#define GET_VALUE(_x, _y) *(volatile GLuint *)(GET_PTR(_x, _y)) +#define PUT_VALUE(_x, _y, _v) *(volatile GLuint *)(GET_PTR(_x, _y)) = (_v) +#endif /* GET_VALUE */ + +# define INIT_MONO_PIXEL(p, color) \ + p = PACK_COLOR_8888(0xff, color[0], color[1], color[2]) + +# define WRITE_RGBA(_x, _y, r, g, b, a) \ + PUT_VALUE(_x, _y, ((r << 16) | \ + (g << 8) | \ + (b << 0) | \ + (0xff << 24))) + +#define WRITE_PIXEL(_x, _y, p) PUT_VALUE(_x, _y, p) + +# if defined( USE_X86_ASM ) +# define READ_RGBA(rgba, _x, _y) \ + do { \ + GLuint p = GET_VALUE(_x, _y); \ + __asm__ __volatile__( "bswap %0; rorl $8, %0" \ + : "=r" (p) : "0" (p) ); \ + ((GLuint *)rgba)[0] = p | 0xff000000; \ + } while (0) +# elif defined( MESA_BIG_ENDIAN ) + /* On PowerPC with GCC 3.4.2 the shift madness below becomes a single + * rotlwi instruction. It also produces good code on SPARC. + */ +# define READ_RGBA( rgba, _x, _y ) \ + do { \ + GLuint p = GET_VALUE(_x, _y); \ + *((uint32_t *) rgba) = (t << 8) | 0xff; \ + } while (0) +# else +# define READ_RGBA( rgba, _x, _y ) \ + do { \ + GLuint p = GET_VALUE(_x, _y); \ + rgba[0] = (p >> 16) & 0xff; \ + rgba[1] = (p >> 8) & 0xff; \ + rgba[2] = (p >> 0) & 0xff; \ + rgba[3] = 0xff; \ + } while (0) +# endif + #else #error SPANTMP_PIXEL_FMT must be set to a valid value! #endif diff --git a/src/mesa/drivers/dri/common/texmem.c b/src/mesa/drivers/dri/common/texmem.c index b64618a03c8..895139b55b8 100644 --- a/src/mesa/drivers/dri/common/texmem.c +++ b/src/mesa/drivers/dri/common/texmem.c @@ -42,14 +42,10 @@ * application. */ -#include "texmem.h" -#include "main/simple_list.h" #include "main/imports.h" #include "main/macros.h" -#include "main/texformat.h" - -#include <assert.h> - +#include "main/simple_list.h" +#include "texmem.h" static unsigned dummy_swap_counter; @@ -1302,46 +1298,44 @@ driCalculateTextureFirstLastLevel( driTextureObject * t ) /** - * \name DRI texture formats. Pointers initialized to either the big- or - * little-endian Mesa formats. + * \name DRI texture formats. These vars are initialized to either the + * big- or little-endian Mesa formats. */ /*@{*/ -const struct gl_texture_format *_dri_texformat_rgba8888 = NULL; -const struct gl_texture_format *_dri_texformat_argb8888 = NULL; -const struct gl_texture_format *_dri_texformat_rgb565 = NULL; -const struct gl_texture_format *_dri_texformat_argb4444 = NULL; -const struct gl_texture_format *_dri_texformat_argb1555 = NULL; -const struct gl_texture_format *_dri_texformat_al88 = NULL; -const struct gl_texture_format *_dri_texformat_a8 = &_mesa_texformat_a8; -const struct gl_texture_format *_dri_texformat_ci8 = &_mesa_texformat_ci8; -const struct gl_texture_format *_dri_texformat_i8 = &_mesa_texformat_i8; -const struct gl_texture_format *_dri_texformat_l8 = &_mesa_texformat_l8; +gl_format _dri_texformat_rgba8888 = MESA_FORMAT_NONE; +gl_format _dri_texformat_argb8888 = MESA_FORMAT_NONE; +gl_format _dri_texformat_rgb565 = MESA_FORMAT_NONE; +gl_format _dri_texformat_argb4444 = MESA_FORMAT_NONE; +gl_format _dri_texformat_argb1555 = MESA_FORMAT_NONE; +gl_format _dri_texformat_al88 = MESA_FORMAT_NONE; +gl_format _dri_texformat_a8 = MESA_FORMAT_A8; +gl_format _dri_texformat_ci8 = MESA_FORMAT_CI8; +gl_format _dri_texformat_i8 = MESA_FORMAT_I8; +gl_format _dri_texformat_l8 = MESA_FORMAT_L8; /*@}*/ /** - * Initialize little endian target, host byte order independent texture formats + * Initialize _dri_texformat_* vars according to whether we're on + * a big or little endian system. */ void driInitTextureFormats(void) { - const GLuint ui = 1; - const GLubyte littleEndian = *((const GLubyte *) &ui); - - if (littleEndian) { - _dri_texformat_rgba8888 = &_mesa_texformat_rgba8888; - _dri_texformat_argb8888 = &_mesa_texformat_argb8888; - _dri_texformat_rgb565 = &_mesa_texformat_rgb565; - _dri_texformat_argb4444 = &_mesa_texformat_argb4444; - _dri_texformat_argb1555 = &_mesa_texformat_argb1555; - _dri_texformat_al88 = &_mesa_texformat_al88; + if (_mesa_little_endian()) { + _dri_texformat_rgba8888 = MESA_FORMAT_RGBA8888; + _dri_texformat_argb8888 = MESA_FORMAT_ARGB8888; + _dri_texformat_rgb565 = MESA_FORMAT_RGB565; + _dri_texformat_argb4444 = MESA_FORMAT_ARGB4444; + _dri_texformat_argb1555 = MESA_FORMAT_ARGB1555; + _dri_texformat_al88 = MESA_FORMAT_AL88; } else { - _dri_texformat_rgba8888 = &_mesa_texformat_rgba8888_rev; - _dri_texformat_argb8888 = &_mesa_texformat_argb8888_rev; - _dri_texformat_rgb565 = &_mesa_texformat_rgb565_rev; - _dri_texformat_argb4444 = &_mesa_texformat_argb4444_rev; - _dri_texformat_argb1555 = &_mesa_texformat_argb1555_rev; - _dri_texformat_al88 = &_mesa_texformat_al88_rev; + _dri_texformat_rgba8888 = MESA_FORMAT_RGBA8888_REV; + _dri_texformat_argb8888 = MESA_FORMAT_ARGB8888_REV; + _dri_texformat_rgb565 = MESA_FORMAT_RGB565_REV; + _dri_texformat_argb4444 = MESA_FORMAT_ARGB4444_REV; + _dri_texformat_argb1555 = MESA_FORMAT_ARGB1555_REV; + _dri_texformat_al88 = MESA_FORMAT_AL88_REV; } } diff --git a/src/mesa/drivers/dri/common/texmem.h b/src/mesa/drivers/dri/common/texmem.h index 9c065da8b4f..725ba2e1196 100644 --- a/src/mesa/drivers/dri/common/texmem.h +++ b/src/mesa/drivers/dri/common/texmem.h @@ -39,6 +39,7 @@ #define DRI_TEXMEM_H #include "main/mtypes.h" +#include "main/formats.h" #include "main/mm.h" #include "xf86drm.h" @@ -317,16 +318,16 @@ GLboolean driValidateTextureHeaps( driTexHeap * const * texture_heaps, extern void driCalculateTextureFirstLastLevel( driTextureObject * t ); -extern const struct gl_texture_format *_dri_texformat_rgba8888; -extern const struct gl_texture_format *_dri_texformat_argb8888; -extern const struct gl_texture_format *_dri_texformat_rgb565; -extern const struct gl_texture_format *_dri_texformat_argb4444; -extern const struct gl_texture_format *_dri_texformat_argb1555; -extern const struct gl_texture_format *_dri_texformat_al88; -extern const struct gl_texture_format *_dri_texformat_a8; -extern const struct gl_texture_format *_dri_texformat_ci8; -extern const struct gl_texture_format *_dri_texformat_i8; -extern const struct gl_texture_format *_dri_texformat_l8; +extern gl_format _dri_texformat_rgba8888; +extern gl_format _dri_texformat_argb8888; +extern gl_format _dri_texformat_rgb565; +extern gl_format _dri_texformat_argb4444; +extern gl_format _dri_texformat_argb1555; +extern gl_format _dri_texformat_al88; +extern gl_format _dri_texformat_a8; +extern gl_format _dri_texformat_ci8; +extern gl_format _dri_texformat_i8; +extern gl_format _dri_texformat_l8; extern void driInitTextureFormats( void ); diff --git a/src/mesa/drivers/dri/common/utils.c b/src/mesa/drivers/dri/common/utils.c index 66f277c10b2..b272eb74eaf 100644 --- a/src/mesa/drivers/dri/common/utils.c +++ b/src/mesa/drivers/dri/common/utils.c @@ -38,9 +38,6 @@ #include "utils.h" -int driDispatchRemapTable[ driDispatchRemapTable_size ]; - - unsigned driParseDebugString( const char * debug, const struct dri_debug_control * control ) @@ -142,7 +139,7 @@ driGetRendererString( char * buffer, const char * hardware_name, #define need_GL_EXT_blend_func_separate #define need_GL_NV_vertex_program -#include "extension_helper.h" +#include "main/remap_helper.h" static const struct dri_extension all_mesa_extensions[] = { { "GL_ARB_draw_buffers", GL_ARB_draw_buffers_functions }, @@ -165,8 +162,12 @@ static const struct dri_extension all_mesa_extensions[] = { /** - * Enable extensions supported by the driver. + * Enable and map extensions supported by the driver. * + * When ctx is NULL, extensions are not enabled, but their functions + * are still mapped. When extensions_to_enable is NULL, all static + * functions known to mesa core are mapped. + * * \bug * ARB_imaging isn't handled properly. In Mesa, enabling ARB_imaging also * enables all the sub-extensions that are folded into it. This means that @@ -181,18 +182,23 @@ void driInitExtensions( GLcontext * ctx, unsigned i; if ( first_time ) { - for ( i = 0 ; i < driDispatchRemapTable_size ; i++ ) { - driDispatchRemapTable[i] = -1; - } - first_time = 0; - driInitExtensions( ctx, all_mesa_extensions, GL_FALSE ); + driInitExtensions( NULL, all_mesa_extensions, GL_FALSE ); } if ( (ctx != NULL) && enable_imaging ) { _mesa_enable_imaging_extensions( ctx ); } + /* The caller is too lazy to list any extension */ + if ( extensions_to_enable == NULL ) { + /* Map the static functions. Together with those mapped by remap + * table, this should cover everything mesa core knows. + */ + _mesa_map_static_functions(); + return; + } + for ( i = 0 ; extensions_to_enable[i].name != NULL ; i++ ) { driInitSingleExtension( ctx, & extensions_to_enable[i] ); } @@ -202,80 +208,18 @@ void driInitExtensions( GLcontext * ctx, /** - * Enable and add dispatch functions for a single extension + * Enable and map functions for a single extension * * \param ctx Context where extension is to be enabled. * \param ext Extension that is to be enabled. * - * \sa driInitExtensions, _mesa_enable_extension, _glapi_add_entrypoint - * - * \todo - * Determine if it would be better to use \c strlen instead of the hardcoded - * for-loops. + * \sa driInitExtensions, _mesa_enable_extension, _mesa_map_function_array */ void driInitSingleExtension( GLcontext * ctx, const struct dri_extension * ext ) { - unsigned i; - - if ( ext->functions != NULL ) { - for ( i = 0 ; ext->functions[i].strings != NULL ; i++ ) { - const char * functions[16]; - const char * parameter_signature; - const char * str = ext->functions[i].strings; - unsigned j; - unsigned offset; - - - /* Separate the parameter signature from the rest of the string. - * If the parameter signature is empty (i.e., the string starts - * with a NUL character), then the function has a void parameter - * list. - */ - parameter_signature = str; - while ( str[0] != '\0' ) { - str++; - } - str++; - - - /* Divide the string into the substrings that name each - * entry-point for the function. - */ - for ( j = 0 ; j < 16 ; j++ ) { - if ( str[0] == '\0' ) { - functions[j] = NULL; - break; - } - - functions[j] = str; - - while ( str[0] != '\0' ) { - str++; - } - str++; - } - - - /* Add each entry-point to the dispatch table. - */ - offset = _glapi_add_dispatch( functions, parameter_signature ); - if (offset == -1) { -#if 0 /* this causes noise with egl */ - fprintf(stderr, "DISPATCH ERROR! _glapi_add_dispatch failed " - "to add %s!\n", functions[0]); -#endif - } - else if (ext->functions[i].remap_index != -1) { - driDispatchRemapTable[ ext->functions[i].remap_index ] = - offset; - } - else if (ext->functions[i].offset != offset) { - fprintf(stderr, "DISPATCH ERROR! %s -> %u != %u\n", - functions[0], offset, ext->functions[i].offset); - } - } + _mesa_map_function_array(ext->functions); } if ( ctx != NULL ) { diff --git a/src/mesa/drivers/dri/common/utils.h b/src/mesa/drivers/dri/common/utils.h index 9e9e5bc224b..2aa6de66c18 100644 --- a/src/mesa/drivers/dri/common/utils.h +++ b/src/mesa/drivers/dri/common/utils.h @@ -31,6 +31,7 @@ #include <GL/gl.h> #include <GL/internal/dri_interface.h> #include "main/context.h" +#include "main/remap.h" typedef struct __DRIutilversionRec2 __DRIutilversion2; @@ -40,35 +41,6 @@ struct dri_debug_control { }; /** - * Description of the entry-points and parameters for an OpenGL function. - */ -struct dri_extension_function { - /** - * \brief - * Packed string describing the parameter signature and the entry-point - * names. - * - * The parameter signature and the names of the entry-points for this - * function are packed into a single string. The substrings are - * separated by NUL characters. The whole string is terminated by - * two consecutive NUL characters. - */ - const char * strings; - - - /** - * Location in the remap table where the dispatch offset should be - * stored. - */ - int remap_index; - - /** - * Offset of the function in the dispatch table. - */ - int offset; -}; - -/** * Description of the API for an extension to OpenGL. */ struct dri_extension { @@ -83,7 +55,7 @@ struct dri_extension { * is terminated by a structure with a \c NULL * \c dri_extension_function::strings pointer. */ - const struct dri_extension_function * functions; + const struct gl_function_remap * functions; }; /** diff --git a/src/mesa/drivers/dri/fb/Makefile b/src/mesa/drivers/dri/fb/Makefile index 309f50b95f8..cf9b3a85563 100644 --- a/src/mesa/drivers/dri/fb/Makefile +++ b/src/mesa/drivers/dri/fb/Makefile @@ -25,5 +25,3 @@ ASM_SOURCES = include ../Makefile.template -symlinks: - diff --git a/src/mesa/drivers/dri/fb/fb_dri.c b/src/mesa/drivers/dri/fb/fb_dri.c index 571b8922d5d..fd869b2fe7d 100644 --- a/src/mesa/drivers/dri/fb/fb_dri.c +++ b/src/mesa/drivers/dri/fb/fb_dri.c @@ -437,7 +437,7 @@ fbCreateBuffer( __DRIscreenPrivate *driScrnPriv, /* XXX double-check these parameters (bpp vs cpp, etc) */ { - driRenderbuffer *drb = driNewRenderbuffer(GL_RGBA, + driRenderbuffer *drb = driNewRenderbuffer(MESA_FORMAT_ARGB8888, driScrnPriv->pFB, driScrnPriv->fbBPP / 8, driScrnPriv->fbOrigin, @@ -451,7 +451,7 @@ fbCreateBuffer( __DRIscreenPrivate *driScrnPriv, /* XXX what are the correct origin/stride values? */ GLvoid *backBuf = _mesa_malloc(driScrnPriv->fbStride * driScrnPriv->fbHeight); - driRenderbuffer *drb = driNewRenderbuffer(GL_RGBA, + driRenderbuffer *drb = driNewRenderbuffer(MESA_FORMAT_ARGB8888, backBuf, driScrnPriv->fbBPP /8, driScrnPriv->fbOrigin, diff --git a/src/mesa/drivers/dri/fb/fb_egl.c b/src/mesa/drivers/dri/fb/fb_egl.c index 4e41860d8c8..eb7adf8224b 100644 --- a/src/mesa/drivers/dri/fb/fb_egl.c +++ b/src/mesa/drivers/dri/fb/fb_egl.c @@ -692,7 +692,7 @@ fbCreateScreenSurfaceMESA(_EGLDriver *drv, EGLDisplay dpy, EGLConfig cfg, /* front color renderbuffer */ { - driRenderbuffer *drb = driNewRenderbuffer(GL_RGBA, display->pFB, + driRenderbuffer *drb = driNewRenderbuffer(MESA_FORMAT_ARGB8888, display->pFB, bytesPerPixel, origin, stride, NULL); fbSetSpanFunctions(drb, &vis); @@ -703,7 +703,7 @@ fbCreateScreenSurfaceMESA(_EGLDriver *drv, EGLDisplay dpy, EGLConfig cfg, /* back color renderbuffer */ if (vis.doubleBufferMode) { GLubyte *backBuf = _mesa_malloc(stride * height); - driRenderbuffer *drb = driNewRenderbuffer(GL_RGBA, backBuf, + driRenderbuffer *drb = driNewRenderbuffer(MESA_FORMAT_ARGB8888, backBuf, bytesPerPixel, origin, stride, NULL); fbSetSpanFunctions(drb, &vis); diff --git a/src/mesa/drivers/dri/ffb/Makefile b/src/mesa/drivers/dri/ffb/Makefile index cb73238c030..e9da8f9066e 100644 --- a/src/mesa/drivers/dri/ffb/Makefile +++ b/src/mesa/drivers/dri/ffb/Makefile @@ -33,4 +33,3 @@ ASM_SOURCES = include ../Makefile.template -symlinks: diff --git a/src/mesa/drivers/dri/ffb/ffb_vbtmp.h b/src/mesa/drivers/dri/ffb/ffb_vbtmp.h index 0495d0e2767..c548ef3ad58 100644 --- a/src/mesa/drivers/dri/ffb/ffb_vbtmp.h +++ b/src/mesa/drivers/dri/ffb/ffb_vbtmp.h @@ -38,11 +38,11 @@ static void TAG(emit)(GLcontext *ctx, GLuint start, GLuint end) #endif #if (IND & (FFB_VB_RGBA_BIT)) - col0 = VB->ColorPtr[0]->data; - col0_stride = VB->ColorPtr[0]->stride; + col0 = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col0_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; #if (IND & (FFB_VB_TWOSIDE_BIT)) - col1 = VB->ColorPtr[1]->data; - col1_stride = VB->ColorPtr[1]->stride; + col1 = VB->BackfaceColorPtr->data; + col1_stride = VB->BackfaceColorPtr->stride; #endif #endif diff --git a/src/mesa/drivers/dri/ffb/ffb_xmesa.c b/src/mesa/drivers/dri/ffb/ffb_xmesa.c index 3b9f5c67591..09cc26d09eb 100644 --- a/src/mesa/drivers/dri/ffb/ffb_xmesa.c +++ b/src/mesa/drivers/dri/ffb/ffb_xmesa.c @@ -347,7 +347,7 @@ ffbCreateBuffer(__DRIscreenPrivate *driScrnPriv, { driRenderbuffer *frontRb - = driNewRenderbuffer(GL_RGBA, NULL, bpp, offset, bogusPitch, + = driNewRenderbuffer(MESA_FORMAT_ARGB8888, NULL, bpp, offset, bogusPitch, driDrawPriv); ffbSetSpanFunctions(frontRb, mesaVis); _mesa_add_renderbuffer(fb, BUFFER_FRONT_LEFT, &frontRb->Base); @@ -355,7 +355,7 @@ ffbCreateBuffer(__DRIscreenPrivate *driScrnPriv, if (mesaVis->doubleBufferMode) { driRenderbuffer *backRb - = driNewRenderbuffer(GL_RGBA, NULL, bpp, offset, bogusPitch, + = driNewRenderbuffer(MESA_FORMAT_ARGB8888, NULL, bpp, offset, bogusPitch, driDrawPriv); ffbSetSpanFunctions(backRb, mesaVis); _mesa_add_renderbuffer(fb, BUFFER_BACK_LEFT, &backRb->Base); @@ -363,7 +363,7 @@ ffbCreateBuffer(__DRIscreenPrivate *driScrnPriv, if (mesaVis->depthBits == 16) { driRenderbuffer *depthRb - = driNewRenderbuffer(GL_DEPTH_COMPONENT16, NULL, bpp, offset, + = driNewRenderbuffer(MESA_FORMAT_Z16, NULL, bpp, offset, bogusPitch, driDrawPriv); ffbSetDepthFunctions(depthRb, mesaVis); _mesa_add_renderbuffer(fb, BUFFER_DEPTH, &depthRb->Base); @@ -371,7 +371,7 @@ ffbCreateBuffer(__DRIscreenPrivate *driScrnPriv, if (mesaVis->stencilBits > 0 && !swStencil) { driRenderbuffer *stencilRb - = driNewRenderbuffer(GL_STENCIL_INDEX8_EXT, NULL, bpp, offset, + = driNewRenderbuffer(MESA_FORMAT_S8, NULL, bpp, offset, bogusPitch, driDrawPriv); ffbSetStencilFunctions(stencilRb, mesaVis); _mesa_add_renderbuffer(fb, BUFFER_STENCIL, &stencilRb->Base); diff --git a/src/mesa/drivers/dri/gamma/Makefile b/src/mesa/drivers/dri/gamma/Makefile index 250d3ac0891..09df1578fc7 100644 --- a/src/mesa/drivers/dri/gamma/Makefile +++ b/src/mesa/drivers/dri/gamma/Makefile @@ -32,4 +32,3 @@ ASM_SOURCES = include ../Makefile.template -symlinks: diff --git a/src/mesa/drivers/dri/gamma/gamma_render.c b/src/mesa/drivers/dri/gamma/gamma_render.c index 1b9fd169f46..a03a93d132c 100644 --- a/src/mesa/drivers/dri/gamma/gamma_render.c +++ b/src/mesa/drivers/dri/gamma/gamma_render.c @@ -53,13 +53,13 @@ static void gamma_emit( GLcontext *ctx, GLuint start, GLuint end) GLfloat (*tc0)[4] = 0; GLuint tc0_size = 0; - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; if (ctx->Texture.Unit[0]._ReallyEnabled) { - tc0_stride = VB->TexCoordPtr[0]->stride; - tc0 = VB->TexCoordPtr[0]->data; - tc0_size = VB->TexCoordPtr[0]->size; + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0]->stride; + tc0 = VB->AttribPtr[_TNL_ATTRIB_TEX0]->data; + tc0_size = VB->AttribPtr[_TNL_ATTRIB_TEX0]->size; coord = VB->ClipPtr->data; coord_stride = VB->ClipPtr->stride; } else { diff --git a/src/mesa/drivers/dri/gamma/gamma_tex.c b/src/mesa/drivers/dri/gamma/gamma_tex.c index 97797d47882..0dad250e4d8 100644 --- a/src/mesa/drivers/dri/gamma/gamma_tex.c +++ b/src/mesa/drivers/dri/gamma/gamma_tex.c @@ -1,21 +1,18 @@ -#include <stdlib.h> -#include <stdio.h> - #include "main/glheader.h" #include "main/mtypes.h" +#include "main/colormac.h" #include "main/imports.h" #include "main/simple_list.h" #include "main/enums.h" +#include "main/mm.h" #include "main/texstore.h" -#include "teximage.h" -#include "main/texformat.h" -#include "texobj.h" +#include "main/teximage.h" +#include "main/texobj.h" + #include "swrast/swrast.h" -#include "main/mm.h" #include "gammacontext.h" -#include "colormac.h" /* diff --git a/src/mesa/drivers/dri/glcore/Makefile b/src/mesa/drivers/dri/glcore/Makefile deleted file mode 100644 index ac7e1de9285..00000000000 --- a/src/mesa/drivers/dri/glcore/Makefile +++ /dev/null @@ -1,84 +0,0 @@ -# src/mesa/drivers/dri/glcore/Makefile - -TOP = ../../../../.. -include $(TOP)/configs/current - -LIBNAME = glcore_dri.so - -DRIVER_SOURCES = glcore_driver.c \ - $(TOP)/src/mesa/drivers/common/driverfuncs.c \ - ../common/dri_util.c - -C_SOURCES = \ - $(DRIVER_SOURCES) \ - $(DRI_SOURCES) - - -# Include directories -INCLUDE_DIRS = \ - -I. \ - -I../common \ - -I../dri_client \ - -I../dri_client/imports \ - -Iserver \ - -I$(TOP)/include \ - -I$(DRM_SOURCE_PATH)/shared-core \ - -I$(TOP)/src/mesa \ - -I$(TOP)/src/mesa/main \ - -I$(TOP)/src/mesa/glapi \ - -I$(TOP)/src/mesa/math \ - -I$(TOP)/src/mesa/transform \ - -I$(TOP)/src/mesa/shader \ - -I$(TOP)/src/mesa/swrast \ - -I$(TOP)/src/mesa/swrast_setup - -# Core Mesa objects -MESA_MODULES = $(TOP)/src/mesa/libmesa.a - -# Libraries that the driver shared lib depends on -LIB_DEPS = -lm -lpthread -lc -# LIB_DEPS = -lGL -lm -lpthread -lc - - -ASM_SOURCES = - -OBJECTS = $(C_SOURCES:.c=.o) \ - $(ASM_SOURCES:.S=.o) - - -##### RULES ##### - -.c.o: - $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $(DEFINES) $< -o $@ - -.S.o: - $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $(DEFINES) $< -o $@ - - -##### TARGETS ##### - -default: depend $(TOP)/$(LIB_DIR)/$(LIBNAME) - - -$(TOP)/$(LIB_DIR)/$(LIBNAME): $(OBJECTS) $(MESA_MODULES) $(WINOBJ) Makefile - CC="$(CC)" CXX="$(CXX)" $(TOP)/bin/mklib -o $(LIBNAME) -noprefix -install $(TOP)/$(LIB_DIR) \ - $(OBJECTS) $(WINLIB) $(LIB_DEPS) $(WINOBJ) $(MESA_MODULES) - - -depend: $(C_SOURCES) $(ASM_SOURCES) - rm -f depend - touch depend - $(MKDEP) $(MKDEP_OPTIONS) $(INCLUDE_DIRS) $(C_SOURCES) $(ASM_SOURCES) \ - > /dev/null - - -# Emacs tags -tags: - etags `find . -name \*.[ch]` `find ../include` - - -clean: - -rm -f *.o server/*.o - - -include depend diff --git a/src/mesa/drivers/dri/i810/Makefile b/src/mesa/drivers/dri/i810/Makefile index a7825b49b44..3874faee518 100644 --- a/src/mesa/drivers/dri/i810/Makefile +++ b/src/mesa/drivers/dri/i810/Makefile @@ -29,4 +29,3 @@ ASM_SOURCES = include ../Makefile.template -symlinks: diff --git a/src/mesa/drivers/dri/i810/i810context.c b/src/mesa/drivers/dri/i810/i810context.c index 67856556865..7311b2e7654 100644 --- a/src/mesa/drivers/dri/i810/i810context.c +++ b/src/mesa/drivers/dri/i810/i810context.c @@ -116,7 +116,7 @@ static void i810BufferSize(GLframebuffer *buffer, GLuint *width, GLuint *height) /* Extension strings exported by the i810 driver. */ -const struct dri_extension card_extensions[] = +static const struct dri_extension card_extensions[] = { { "GL_ARB_multitexture", NULL }, { "GL_ARB_texture_env_add", NULL }, diff --git a/src/mesa/drivers/dri/i810/i810screen.c b/src/mesa/drivers/dri/i810/i810screen.c index 6e49f3466c3..2f6b8631ffc 100644 --- a/src/mesa/drivers/dri/i810/i810screen.c +++ b/src/mesa/drivers/dri/i810/i810screen.c @@ -53,8 +53,6 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "GL/internal/dri_interface.h" -extern const struct dri_extension card_extensions[]; - static const __DRIconfig ** i810FillInModes( __DRIscreenPrivate *psp, unsigned pixel_bits, unsigned depth_bits, @@ -166,8 +164,6 @@ i810InitScreen(__DRIscreen *sPriv) return NULL; } - driInitExtensions( NULL, card_extensions, GL_TRUE ); - if (sPriv->devPrivSize != sizeof(I810DRIRec)) { fprintf(stderr,"\nERROR! sizeof(I810DRIRec) does not match passed size from device driver\n"); return GL_FALSE; @@ -293,7 +289,7 @@ i810CreateBuffer( __DRIscreenPrivate *driScrnPriv, { driRenderbuffer *frontRb - = driNewRenderbuffer(GL_RGBA, + = driNewRenderbuffer(MESA_FORMAT_ARGB8888, driScrnPriv->pFB, screen->cpp, /*screen->frontOffset*/0, screen->backPitch, @@ -304,7 +300,7 @@ i810CreateBuffer( __DRIscreenPrivate *driScrnPriv, if (mesaVis->doubleBufferMode) { driRenderbuffer *backRb - = driNewRenderbuffer(GL_RGBA, + = driNewRenderbuffer(MESA_FORMAT_ARGB8888, screen->back.map, screen->cpp, screen->backOffset, screen->backPitch, @@ -315,7 +311,7 @@ i810CreateBuffer( __DRIscreenPrivate *driScrnPriv, if (mesaVis->depthBits == 16) { driRenderbuffer *depthRb - = driNewRenderbuffer(GL_DEPTH_COMPONENT16, + = driNewRenderbuffer(MESA_FORMAT_Z16, screen->depth.map, screen->cpp, screen->depthOffset, screen->backPitch, diff --git a/src/mesa/drivers/dri/i810/i810tex.c b/src/mesa/drivers/dri/i810/i810tex.c index cd6e1a8e6e8..2f6978f5aae 100644 --- a/src/mesa/drivers/dri/i810/i810tex.c +++ b/src/mesa/drivers/dri/i810/i810tex.c @@ -28,7 +28,6 @@ #include "main/simple_list.h" #include "main/enums.h" #include "main/texstore.h" -#include "main/texformat.h" #include "main/teximage.h" #include "main/texobj.h" #include "main/colormac.h" @@ -440,7 +439,7 @@ static void i810DeleteTexture( GLcontext *ctx, struct gl_texture_object *tObj ) * The i810 only supports 5 texture modes that are useful to Mesa. That * makes this routine pretty simple. */ -static const struct gl_texture_format * +static gl_format i810ChooseTextureFormat( GLcontext *ctx, GLint internalFormat, GLenum format, GLenum type ) { @@ -458,9 +457,9 @@ i810ChooseTextureFormat( GLcontext *ctx, GLint internalFormat, if ( ((format == GL_BGRA) && (type == GL_UNSIGNED_SHORT_1_5_5_5_REV)) || ((format == GL_RGBA) && (type == GL_UNSIGNED_SHORT_5_5_5_1)) || (internalFormat == GL_RGB5_A1) ) { - return &_mesa_texformat_argb1555; + return MESA_FORMAT_ARGB1555; } - return &_mesa_texformat_argb4444; + return MESA_FORMAT_ARGB4444; case 3: case GL_RGB: @@ -472,7 +471,7 @@ i810ChooseTextureFormat( GLcontext *ctx, GLint internalFormat, case GL_RGB10: case GL_RGB12: case GL_RGB16: - return &_mesa_texformat_rgb565; + return MESA_FORMAT_RGB565; case GL_ALPHA: case GL_ALPHA4: @@ -502,21 +501,21 @@ i810ChooseTextureFormat( GLcontext *ctx, GLint internalFormat, case GL_INTENSITY12: case GL_INTENSITY16: case GL_COMPRESSED_INTENSITY: - return &_mesa_texformat_al88; + return MESA_FORMAT_AL88; case GL_YCBCR_MESA: if (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE) - return &_mesa_texformat_ycbcr; + return MESA_FORMAT_YCBCR; else - return &_mesa_texformat_ycbcr_rev; + return MESA_FORMAT_YCBCR_REV; default: fprintf(stderr, "unexpected texture format in %s\n", __FUNCTION__); - return NULL; + return MESA_FORMAT_NONE; } - return NULL; /* never get here */ + return MESA_FORMAT_NONE; /* never get here */ } /** diff --git a/src/mesa/drivers/dri/i810/i810texmem.c b/src/mesa/drivers/dri/i810/i810texmem.c index 5ad66dbf5cb..d93afbf9ef6 100644 --- a/src/mesa/drivers/dri/i810/i810texmem.c +++ b/src/mesa/drivers/dri/i810/i810texmem.c @@ -30,7 +30,6 @@ #include "main/enums.h" #include "main/colormac.h" #include "main/mm.h" -#include "main/texformat.h" #include "i810screen.h" #include "i810_dri.h" @@ -92,44 +91,47 @@ static void i810UploadTexLevel( i810ContextPtr imesa, { const struct gl_texture_image *image = t->image[hwlevel].image; int j; + GLuint texelBytes; if (!image || !image->Data) return; - if (image->Width * image->TexFormat->TexelBytes == t->Pitch) { + texelBytes = _mesa_get_format_bytes(image->TexFormat); + + if (image->Width * texelBytes == t->Pitch) { GLubyte *dst = (GLubyte *)(t->BufAddr + t->image[hwlevel].offset); GLubyte *src = (GLubyte *)image->Data; memcpy( dst, src, t->Pitch * image->Height ); } - else switch (image->TexFormat->TexelBytes) { - case 1: - { - GLubyte *dst = (GLubyte *)(t->BufAddr + t->image[hwlevel].offset); - GLubyte *src = (GLubyte *)image->Data; - - for (j = 0 ; j < image->Height ; j++, dst += t->Pitch) { - __memcpy(dst, src, image->Width ); - src += image->Width; - } + else { + switch (texelBytes) { + case 1: + { + GLubyte *dst = (GLubyte *)(t->BufAddr + t->image[hwlevel].offset); + GLubyte *src = (GLubyte *)image->Data; + + for (j = 0 ; j < image->Height ; j++, dst += t->Pitch) { + __memcpy(dst, src, image->Width ); + src += image->Width; + } + } + break; + case 2: + { + GLushort *dst = (GLushort *)(t->BufAddr + t->image[hwlevel].offset); + GLushort *src = (GLushort *)image->Data; + + for (j = 0 ; j < image->Height ; j++, dst += (t->Pitch/2)) { + __memcpy(dst, src, image->Width * 2 ); + src += image->Width; + } + } + break; + default: + fprintf(stderr, "%s: Not supported texel size %d\n", + __FUNCTION__, texelBytes); } - break; - - case 2: - { - GLushort *dst = (GLushort *)(t->BufAddr + t->image[hwlevel].offset); - GLushort *src = (GLushort *)image->Data; - - for (j = 0 ; j < image->Height ; j++, dst += (t->Pitch/2)) { - __memcpy(dst, src, image->Width * 2 ); - src += image->Width; - } - } - break; - - default: - fprintf(stderr, "%s: Not supported texel size %d\n", - __FUNCTION__, image->TexFormat->TexelBytes); } } diff --git a/src/mesa/drivers/dri/i810/i810texstate.c b/src/mesa/drivers/dri/i810/i810texstate.c index 0e09f54c41b..bff28c11c89 100644 --- a/src/mesa/drivers/dri/i810/i810texstate.c +++ b/src/mesa/drivers/dri/i810/i810texstate.c @@ -25,7 +25,6 @@ #include "main/glheader.h" #include "main/macros.h" #include "main/mtypes.h" -#include "main/texformat.h" #include "main/simple_list.h" #include "main/enums.h" #include "main/mm.h" @@ -53,7 +52,7 @@ static void i810SetTexImages( i810ContextPtr imesa, /* fprintf(stderr, "%s\n", __FUNCTION__); */ t->texelBytes = 2; - switch (baseImage->TexFormat->MesaFormat) { + switch (baseImage->TexFormat) { case MESA_FORMAT_ARGB1555: textureFormat = MI1_FMT_16BPP | MI1_PF_16BPP_ARGB1555; break; diff --git a/src/mesa/drivers/dri/i915/Makefile b/src/mesa/drivers/dri/i915/Makefile index 9d049dea8fe..37f15aa7671 100644 --- a/src/mesa/drivers/dri/i915/Makefile +++ b/src/mesa/drivers/dri/i915/Makefile @@ -19,7 +19,6 @@ DRIVER_SOURCES = \ intel_batchbuffer.c \ intel_clear.c \ intel_extensions.c \ - intel_generatemipmap.c \ intel_mipmap_tree.c \ intel_tex_layout.c \ intel_tex_image.c \ @@ -73,4 +72,3 @@ intel_decode.o: ../intel/intel_decode.c intel_tex_layout.o: ../intel/intel_tex_layout.c -symlinks: diff --git a/src/mesa/drivers/dri/i915/i830_texblend.c b/src/mesa/drivers/dri/i915/i830_texblend.c index 09f7f37e765..3f64be8c962 100644 --- a/src/mesa/drivers/dri/i915/i830_texblend.c +++ b/src/mesa/drivers/dri/i915/i830_texblend.c @@ -30,7 +30,6 @@ #include "main/mtypes.h" #include "main/simple_list.h" #include "main/enums.h" -#include "main/texformat.h" #include "main/texstore.h" #include "main/mm.h" diff --git a/src/mesa/drivers/dri/i915/i830_texstate.c b/src/mesa/drivers/dri/i915/i830_texstate.c index 6f998fa6f77..27c5aa1e085 100644 --- a/src/mesa/drivers/dri/i915/i830_texstate.c +++ b/src/mesa/drivers/dri/i915/i830_texstate.c @@ -27,7 +27,7 @@ #include "main/mtypes.h" #include "main/enums.h" -#include "main/texformat.h" +#include "main/colormac.h" #include "intel_mipmap_tree.h" #include "intel_tex.h" @@ -56,10 +56,9 @@ translate_texture_format(GLuint mesa_format, GLuint internal_format) case MESA_FORMAT_ARGB4444: return MAPSURF_16BIT | MT_16BIT_ARGB4444; case MESA_FORMAT_ARGB8888: - if (internal_format == GL_RGB) - return MAPSURF_32BIT | MT_32BIT_XRGB8888; - else - return MAPSURF_32BIT | MT_32BIT_ARGB8888; + return MAPSURF_32BIT | MT_32BIT_ARGB8888; + case MESA_FORMAT_XRGB8888: + return MAPSURF_32BIT | MT_32BIT_XRGB8888; case MESA_FORMAT_YCBCR_REV: return (MAPSURF_422 | MT_422_YCRCB_NORMAL); case MESA_FORMAT_YCBCR: @@ -160,13 +159,20 @@ i830_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3) pitch = intelObj->pitchOverride; } else { + GLuint dst_x, dst_y; + + intel_miptree_get_image_offset(intelObj->mt, intelObj->firstLevel, 0, 0, + &dst_x, &dst_y); + dri_bo_reference(intelObj->mt->region->buffer); i830->state.tex_buffer[unit] = intelObj->mt->region->buffer; - i830->state.tex_offset[unit] = intel_miptree_image_offset(intelObj->mt, - 0, intelObj-> - firstLevel); + /* XXX: This calculation is probably broken for tiled images with + * a non-page-aligned offset. + */ + i830->state.tex_offset[unit] = (dst_x + dst_y * intelObj->mt->pitch) * + intelObj->mt->cpp; - format = translate_texture_format(firstImage->TexFormat->MesaFormat, + format = translate_texture_format(firstImage->TexFormat, firstImage->InternalFormat); pitch = intelObj->mt->pitch * intelObj->mt->cpp; } @@ -303,11 +309,10 @@ i830_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3) CLAMPED_FLOAT_TO_UBYTE(border[2], tObj->BorderColor[2]); CLAMPED_FLOAT_TO_UBYTE(border[3], tObj->BorderColor[3]); - state[I830_TEXREG_TM0S4] = INTEL_PACKCOLOR8888(border[0], - border[1], - border[2], - border[3]); - + state[I830_TEXREG_TM0S4] = PACK_COLOR_8888(border[3], + border[0], + border[1], + border[2]); I830_ACTIVESTATE(i830, I830_UPLOAD_TEX(unit), GL_TRUE); /* memcmp was already disabled, but definitely won't work as the diff --git a/src/mesa/drivers/dri/i915/i830_vtbl.c b/src/mesa/drivers/dri/i915/i830_vtbl.c index 983f6724c98..1e3c8301d8d 100644 --- a/src/mesa/drivers/dri/i915/i830_vtbl.c +++ b/src/mesa/drivers/dri/i915/i830_vtbl.c @@ -26,7 +26,6 @@ **************************************************************************/ #include "glapi/glapi.h" -#include "main/texformat.h" #include "i830_context.h" #include "i830_reg.h" @@ -127,7 +126,7 @@ i830_render_start(struct intel_context *intel) for (i = 0; i < I830_TEX_UNITS; i++) { if (RENDERINPUTS_TEST(index_bitset, _TNL_ATTRIB_TEX(i))) { - GLuint sz = VB->TexCoordPtr[i]->size; + GLuint sz = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size; GLuint emit; GLuint mcs = (i830->state.Tex[i][I830_TEXREG_MCS] & ~TEXCOORDTYPE_MASK); @@ -646,8 +645,9 @@ i830_state_draw_region(struct intel_context *intel, DSTORG_VERT_BIAS(0x8) | DEPTH_IS_Z); /* .5 */ if (irb != NULL) { - switch (irb->texformat->MesaFormat) { + switch (irb->Base.Format) { case MESA_FORMAT_ARGB8888: + case MESA_FORMAT_XRGB8888: value |= DV_PF_8888; break; case MESA_FORMAT_RGB565: @@ -661,7 +661,7 @@ i830_state_draw_region(struct intel_context *intel, break; default: _mesa_problem(ctx, "Bad renderbuffer format: %d\n", - irb->texformat->MesaFormat); + irb->Base.Format); } } @@ -714,20 +714,8 @@ i830_new_batch(struct intel_context *intel) { struct i830_context *i830 = i830_context(&intel->ctx); i830->state.emitted = 0; - - /* Check that we didn't just wrap our batchbuffer at a bad time. */ - assert(!intel->no_batch_wrap); -} - - - -static GLuint -i830_flush_cmd(void) -{ - return MI_FLUSH | FLUSH_MAP_CACHE; } - static void i830_assert_not_dirty( struct intel_context *intel ) { @@ -753,7 +741,6 @@ i830InitVtbl(struct i830_context *i830) i830->intel.vtbl.reduced_primitive_state = i830_reduced_primitive_state; i830->intel.vtbl.set_draw_region = i830_set_draw_region; i830->intel.vtbl.update_texture_state = i830UpdateTextureState; - i830->intel.vtbl.flush_cmd = i830_flush_cmd; i830->intel.vtbl.render_start = i830_render_start; i830->intel.vtbl.render_prevalidate = i830_render_prevalidate; i830->intel.vtbl.assert_not_dirty = i830_assert_not_dirty; diff --git a/src/mesa/drivers/dri/i915/i915_context.c b/src/mesa/drivers/dri/i915/i915_context.c index 3ab7d682ee7..0485be2cc1f 100644 --- a/src/mesa/drivers/dri/i915/i915_context.c +++ b/src/mesa/drivers/dri/i915/i915_context.c @@ -40,6 +40,7 @@ #include "utils.h" #include "i915_reg.h" +#include "i915_program.h" #include "intel_regions.h" #include "intel_batchbuffer.h" @@ -80,6 +81,8 @@ i915InvalidateState(GLcontext * ctx, GLuint new_state) i915_update_stencil(ctx); if (new_state & (_NEW_LIGHT)) i915_update_provoking_vertex(ctx); + if (new_state & (_NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS)) + i915_update_program(ctx); } @@ -139,7 +142,10 @@ i915CreateContext(const __GLcontextModes * mesaVis, ctx->Const.MaxTextureUnits = I915_TEX_UNITS; ctx->Const.MaxTextureImageUnits = I915_TEX_UNITS; ctx->Const.MaxTextureCoordUnits = I915_TEX_UNITS; - + ctx->Const.MaxVarying = I915_TEX_UNITS; + ctx->Const.MaxCombinedTextureImageUnits = + ctx->Const.MaxVertexTextureImageUnits + + ctx->Const.MaxTextureImageUnits; /* Advertise the full hardware capabilities. The new memory * manager should cope much better with overload situations: diff --git a/src/mesa/drivers/dri/i915/i915_context.h b/src/mesa/drivers/dri/i915/i915_context.h index 8de4a9d0d36..25418d5f7aa 100644 --- a/src/mesa/drivers/dri/i915/i915_context.h +++ b/src/mesa/drivers/dri/i915/i915_context.h @@ -39,6 +39,7 @@ #define I915_FALLBACK_LOGICOP 0x20000 #define I915_FALLBACK_POLYGON_SMOOTH 0x40000 #define I915_FALLBACK_POINT_SMOOTH 0x80000 +#define I915_FALLBACK_POINT_SPRITE_COORD_ORIGIN 0x100000 #define I915_UPLOAD_CTX 0x1 #define I915_UPLOAD_BUFFERS 0x2 @@ -121,10 +122,14 @@ enum { #define I915_MAX_CONSTANT 32 #define I915_CONSTANT_SIZE (2+(4*I915_MAX_CONSTANT)) +#define I915_MAX_INSN (I915_MAX_DECL_INSN + \ + I915_MAX_TEX_INSN + \ + I915_MAX_ALU_INSN) -#define I915_PROGRAM_SIZE 192 - -#define I915_MAX_INSN (I915_MAX_TEX_INSN+I915_MAX_ALU_INSN) +/* Maximum size of the program packet, which matches the limits on + * decl, tex, and ALU instructions. + */ +#define I915_PROGRAM_SIZE (I915_MAX_INSN * 3 + 1) /* Hardware version of a parsed fragment program. "Derived" from the * mesa fragment_program struct. @@ -154,8 +159,9 @@ struct i915_fragment_program */ GLcontext *ctx; - GLuint declarations[I915_PROGRAM_SIZE]; - GLuint program[I915_PROGRAM_SIZE]; + /* declarations contains the packet header. */ + GLuint declarations[I915_MAX_DECL_INSN * 3 + 1]; + GLuint program[(I915_MAX_TEX_INSN + I915_MAX_ALU_INSN) * 3]; GLfloat constant[I915_MAX_CONSTANT][4]; GLuint constant_flags[I915_MAX_CONSTANT]; diff --git a/src/mesa/drivers/dri/i915/i915_debug.c b/src/mesa/drivers/dri/i915/i915_debug.c index f7bb7ea44c9..fecfac30339 100644 --- a/src/mesa/drivers/dri/i915/i915_debug.c +++ b/src/mesa/drivers/dri/i915/i915_debug.c @@ -806,6 +806,7 @@ static GLboolean i915_debug_packet( struct debug_stream *stream ) default: return debug(stream, "", 0); } + break; default: assert(0); return 0; diff --git a/src/mesa/drivers/dri/i915/i915_fragprog.c b/src/mesa/drivers/dri/i915/i915_fragprog.c index 2db10c60e99..9e4d3180360 100644 --- a/src/mesa/drivers/dri/i915/i915_fragprog.c +++ b/src/mesa/drivers/dri/i915/i915_fragprog.c @@ -89,7 +89,8 @@ src_vector(struct i915_fragment_program *p, */ case PROGRAM_TEMPORARY: if (source->Index >= I915_MAX_TEMPORARY) { - i915_program_error(p, "Exceeded max temporary reg"); + i915_program_error(p, "Exceeded max temporary reg: %d/%d", + source->Index, I915_MAX_TEMPORARY); return 0; } src = UREG(REG_TYPE_R, source->Index); @@ -121,10 +122,23 @@ src_vector(struct i915_fragment_program *p, src = i915_emit_decl(p, REG_TYPE_T, T_TEX0 + (source->Index - FRAG_ATTRIB_TEX0), D0_CHANNEL_ALL); + break; + + case FRAG_ATTRIB_VAR0: + case FRAG_ATTRIB_VAR0 + 1: + case FRAG_ATTRIB_VAR0 + 2: + case FRAG_ATTRIB_VAR0 + 3: + case FRAG_ATTRIB_VAR0 + 4: + case FRAG_ATTRIB_VAR0 + 5: + case FRAG_ATTRIB_VAR0 + 6: + case FRAG_ATTRIB_VAR0 + 7: + src = i915_emit_decl(p, REG_TYPE_T, + T_TEX0 + (source->Index - FRAG_ATTRIB_VAR0), + D0_CHANNEL_ALL); break; default: - i915_program_error(p, "Bad source->Index"); + i915_program_error(p, "Bad source->Index: %d", source->Index); return 0; } break; @@ -146,6 +160,7 @@ src_vector(struct i915_fragment_program *p, case PROGRAM_CONSTANT: case PROGRAM_STATE_VAR: case PROGRAM_NAMED_PARAM: + case PROGRAM_UNIFORM: src = i915_emit_param4fv(p, program->Base.Parameters->ParameterValues[source-> @@ -153,7 +168,7 @@ src_vector(struct i915_fragment_program *p, break; default: - i915_program_error(p, "Bad source->File"); + i915_program_error(p, "Bad source->File: %d", source->File); return 0; } @@ -186,13 +201,14 @@ get_result_vector(struct i915_fragment_program *p, p->depth_written = 1; return UREG(REG_TYPE_OD, 0); default: - i915_program_error(p, "Bad inst->DstReg.Index"); + i915_program_error(p, "Bad inst->DstReg.Index: %d", + inst->DstReg.Index); return 0; } case PROGRAM_TEMPORARY: return UREG(REG_TYPE_R, inst->DstReg.Index); default: - i915_program_error(p, "Bad inst->DstReg.File"); + i915_program_error(p, "Bad inst->DstReg.File: %d", inst->DstReg.File); return 0; } } @@ -231,7 +247,7 @@ translate_tex_src_target(struct i915_fragment_program *p, GLubyte bit) case TEXTURE_CUBE_INDEX: return D0_SAMPLE_TYPE_CUBE; default: - i915_program_error(p, "TexSrcBit"); + i915_program_error(p, "TexSrcBit: %d", bit); return 0; } } @@ -351,7 +367,7 @@ upload_program(struct i915_fragment_program *p) while (1) { GLuint src0, src1, src2, flags; - GLuint tmp = 0, consts0 = 0, consts1 = 0; + GLuint tmp = 0, dst, consts0 = 0, consts1 = 0; switch (inst->Opcode) { case OPCODE_ABS: @@ -503,6 +519,10 @@ upload_program(struct i915_fragment_program *p) EMIT_1ARG_ARITH(A0_FLR); break; + case OPCODE_TRUNC: + EMIT_1ARG_ARITH(A0_TRC); + break; + case OPCODE_FRC: EMIT_1ARG_ARITH(A0_FRC); break; @@ -516,6 +536,22 @@ upload_program(struct i915_fragment_program *p) 0, src0, T0_TEXKILL); break; + case OPCODE_KIL_NV: + if (inst->DstReg.CondMask == COND_TR) { + tmp = i915_get_utemp(p); + + i915_emit_texld(p, get_live_regs(p, inst), + tmp, A0_DEST_CHANNEL_ALL, + 0, /* use a dummy dest reg */ + swizzle(tmp, ONE, ONE, ONE, ONE), /* always */ + T0_TEXKILL); + } else { + p->error = 1; + i915_program_error(p, "Unsupported KIL_NV condition code: %d", + inst->DstReg.CondMask); + } + break; + case OPCODE_LG2: src0 = src_vector(p, &inst->SrcReg[0], program); @@ -615,6 +651,20 @@ upload_program(struct i915_fragment_program *p) EMIT_2ARG_ARITH(A0_MUL); break; + case OPCODE_NOISE1: + case OPCODE_NOISE2: + case OPCODE_NOISE3: + case OPCODE_NOISE4: + /* Don't implement noise because we just don't have the instructions + * to spare. We aren't the first vendor to do so. + */ + i915_program_error(p, "Stubbed-out noise functions"); + i915_emit_arith(p, + A0_MOV, + get_result_vector(p, inst), + get_result_flags(inst), 0, + swizzle(src0, ZERO, ZERO, ZERO, ZERO), 0, 0); + case OPCODE_POW: src0 = src_vector(p, &inst->SrcReg[0], program); src1 = src_vector(p, &inst->SrcReg[1], program); @@ -721,9 +771,38 @@ upload_program(struct i915_fragment_program *p) } break; - case OPCODE_SGE: - EMIT_2ARG_ARITH(A0_SGE); - break; + case OPCODE_SEQ: + tmp = i915_get_utemp(p); + flags = get_result_flags(inst); + dst = get_result_vector(p, inst); + + /* dst = src1 >= src2 */ + i915_emit_arith(p, + A0_SGE, + dst, + flags, 0, + src_vector(p, &inst->SrcReg[0], program), + src_vector(p, &inst->SrcReg[1], program), + 0); + /* tmp = src1 <= src2 */ + i915_emit_arith(p, + A0_SGE, + tmp, + flags, 0, + negate(src_vector(p, &inst->SrcReg[0], program), + 1, 1, 1, 1), + negate(src_vector(p, &inst->SrcReg[1], program), + 1, 1, 1, 1), + 0); + /* dst = tmp && dst */ + i915_emit_arith(p, + A0_MUL, + dst, + flags, 0, + dst, + tmp, + 0); + break; case OPCODE_SIN: src0 = src_vector(p, &inst->SrcReg[0], program); @@ -809,10 +888,71 @@ upload_program(struct i915_fragment_program *p) break; + case OPCODE_SGE: + EMIT_2ARG_ARITH(A0_SGE); + break; + + case OPCODE_SGT: + i915_emit_arith(p, + A0_SLT, + get_result_vector( p, inst ), + get_result_flags( inst ), 0, + negate(src_vector( p, &inst->SrcReg[0], program), + 1, 1, 1, 1), + negate(src_vector( p, &inst->SrcReg[1], program), + 1, 1, 1, 1), + 0); + break; + + case OPCODE_SLE: + i915_emit_arith(p, + A0_SGE, + get_result_vector( p, inst ), + get_result_flags( inst ), 0, + negate(src_vector( p, &inst->SrcReg[0], program), + 1, 1, 1, 1), + negate(src_vector( p, &inst->SrcReg[1], program), + 1, 1, 1, 1), + 0); + break; + case OPCODE_SLT: EMIT_2ARG_ARITH(A0_SLT); break; + case OPCODE_SNE: + tmp = i915_get_utemp(p); + flags = get_result_flags(inst); + dst = get_result_vector(p, inst); + + /* dst = src1 < src2 */ + i915_emit_arith(p, + A0_SLT, + dst, + flags, 0, + src_vector(p, &inst->SrcReg[0], program), + src_vector(p, &inst->SrcReg[1], program), + 0); + /* tmp = src1 > src2 */ + i915_emit_arith(p, + A0_SLT, + tmp, + flags, 0, + negate(src_vector(p, &inst->SrcReg[0], program), + 1, 1, 1, 1), + negate(src_vector(p, &inst->SrcReg[1], program), + 1, 1, 1, 1), + 0); + /* dst = tmp || dst */ + i915_emit_arith(p, + A0_ADD, + dst, + flags | A0_DEST_SATURATE, 0, + dst, + tmp, + 0); + break; + case OPCODE_SUB: src0 = src_vector(p, &inst->SrcReg[0], program); src1 = src_vector(p, &inst->SrcReg[1], program); @@ -869,8 +1009,39 @@ upload_program(struct i915_fragment_program *p) case OPCODE_END: return; + case OPCODE_BGNLOOP: + case OPCODE_BGNSUB: + case OPCODE_BRA: + case OPCODE_BRK: + case OPCODE_CAL: + case OPCODE_CONT: + case OPCODE_DDX: + case OPCODE_DDY: + case OPCODE_ELSE: + case OPCODE_ENDIF: + case OPCODE_ENDLOOP: + case OPCODE_ENDSUB: + case OPCODE_IF: + case OPCODE_RET: + p->error = 1; + i915_program_error(p, "Unsupported opcode: %s", + _mesa_opcode_string(inst->Opcode)); + return; + + case OPCODE_EXP: + case OPCODE_LOG: + /* These opcodes are claimed as GLSL, NV_vp, and ARB_vp in + * prog_instruction.h, but apparently GLSL doesn't ever emit them. + * Instead, it translates to EX2 or LG2. + */ + case OPCODE_TXD: + case OPCODE_TXL: + /* These opcodes are claimed by GLSL in prog_instruction.h, but + * only NV_vp/fp appears to emit them. + */ default: - i915_program_error(p, "bad opcode"); + i915_program_error(p, "bad opcode: %s", + _mesa_opcode_string(inst->Opcode)); return; } @@ -906,7 +1077,7 @@ check_wpos(struct i915_fragment_program *p) p->wpos_tex = -1; for (i = 0; i < p->ctx->Const.MaxTextureCoordUnits; i++) { - if (inputs & FRAG_BIT_TEX(i)) + if (inputs & (FRAG_BIT_TEX(i) | FRAG_BIT_VAR(i))) continue; else if (inputs & FRAG_BIT_WPOS) { p->wpos_tex = i; @@ -1055,6 +1226,28 @@ i915ProgramStringNotify(GLcontext * ctx, _tnl_program_string(ctx, target, prog); } +void +i915_update_program(GLcontext *ctx) +{ + struct intel_context *intel = intel_context(ctx); + struct i915_context *i915 = i915_context(&intel->ctx); + struct i915_fragment_program *fp = + (struct i915_fragment_program *) ctx->FragmentProgram._Current; + + if (i915->current_program != fp) { + if (i915->current_program) { + i915->current_program->on_hardware = 0; + i915->current_program->params_uptodate = 0; + } + + i915->current_program = fp; + } + + if (!fp->translated) + translate_program(fp); + + FALLBACK(&i915->intel, I915_FALLBACK_PROGRAM, fp->error); +} void i915ValidateFragmentProgram(struct i915_context *i915) @@ -1072,16 +1265,6 @@ i915ValidateFragmentProgram(struct i915_context *i915) GLuint s2 = S2_TEXCOORD_NONE; int i, offset = 0; - if (i915->current_program != p) { - if (i915->current_program) { - i915->current_program->on_hardware = 0; - i915->current_program->params_uptodate = 0; - } - - i915->current_program = p; - } - - /* Important: */ VB->AttribPtr[VERT_ATTRIB_POS] = VB->NdcPtr; @@ -1118,13 +1301,21 @@ i915ValidateFragmentProgram(struct i915_context *i915) for (i = 0; i < p->ctx->Const.MaxTextureCoordUnits; i++) { if (inputsRead & FRAG_BIT_TEX(i)) { - int sz = VB->TexCoordPtr[i]->size; + int sz = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size; s2 &= ~S2_TEXCOORD_FMT(i, S2_TEXCOORD_FMT0_MASK); s2 |= S2_TEXCOORD_FMT(i, SZ_TO_HW(sz)); EMIT_ATTR(_TNL_ATTRIB_TEX0 + i, EMIT_SZ(sz), 0, sz * 4); } + else if (inputsRead & FRAG_BIT_VAR(i)) { + int sz = VB->AttribPtr[_TNL_ATTRIB_GENERIC0 + i]->size; + + s2 &= ~S2_TEXCOORD_FMT(i, S2_TEXCOORD_FMT0_MASK); + s2 |= S2_TEXCOORD_FMT(i, SZ_TO_HW(sz)); + + EMIT_ATTR(_TNL_ATTRIB_GENERIC0 + i, EMIT_SZ(sz), 0, sz * 4); + } else if (i == p->wpos_tex) { /* If WPOS is required, duplicate the XYZ position data in an diff --git a/src/mesa/drivers/dri/i915/i915_program.c b/src/mesa/drivers/dri/i915/i915_program.c index e87700f8e0a..e7908bd48fc 100644 --- a/src/mesa/drivers/dri/i915/i915_program.c +++ b/src/mesa/drivers/dri/i915/i915_program.c @@ -130,6 +130,7 @@ i915_emit_decl(struct i915_fragment_program *p, *(p->decl++) = (D0_DCL | D0_DEST(reg) | d0_flags); *(p->decl++) = D1_MBZ; *(p->decl++) = D2_MBZ; + assert(p->decl <= p->declarations + ARRAY_SIZE(p->declarations)); p->nr_decl_insn++; return reg; @@ -186,6 +187,11 @@ i915_emit_arith(struct i915_fragment_program * p, p->utemp_flag = old_utemp_flag; /* restore */ } + if (p->csr >= p->program + ARRAY_SIZE(p->program)) { + i915_program_error(p, "Program contains too many instructions"); + return UREG_BAD; + } + *(p->csr++) = (op | A0_DEST(dest) | mask | saturate | A0_SRC0(src0)); *(p->csr++) = (A1_SRC0(src0) | A1_SRC1(src1)); *(p->csr++) = (A2_SRC1(src1) | A2_SRC2(src2)); @@ -270,6 +276,11 @@ GLuint i915_emit_texld( struct i915_fragment_program *p, p->register_phases[GET_UREG_NR(coord)] == p->nr_tex_indirect) p->nr_tex_indirect++; + if (p->csr >= p->program + ARRAY_SIZE(p->program)) { + i915_program_error(p, "Program contains too many instructions"); + return UREG_BAD; + } + *(p->csr++) = (op | T0_DEST( dest ) | T0_SAMPLER( sampler )); @@ -424,12 +435,21 @@ i915_emit_param4fv(struct i915_fragment_program * p, const GLfloat * values) return 0; } - - +/* Warning the user about program errors seems to be quite valuable, from + * our bug reports. It unfortunately means piglit reporting errors + * when we fall back to software due to an unsupportable program, though. + */ void -i915_program_error(struct i915_fragment_program *p, const char *msg) +i915_program_error(struct i915_fragment_program *p, const char *fmt, ...) { - _mesa_problem(NULL, "i915_program_error: %s", msg); + va_list args; + + fprintf(stderr, "i915_program_error: "); + va_start(args, fmt); + vfprintf(stderr, fmt, args); + va_end(args); + + fprintf(stderr, "\n"); p->error = 1; } @@ -511,7 +531,8 @@ i915_upload_program(struct i915_context *i915, GLuint program_size = p->csr - p->program; GLuint decl_size = p->decl - p->declarations; - FALLBACK(&i915->intel, I915_FALLBACK_PROGRAM, p->error); + if (p->error) + return; /* Could just go straight to the batchbuffer from here: */ diff --git a/src/mesa/drivers/dri/i915/i915_program.h b/src/mesa/drivers/dri/i915/i915_program.h index 14a3f08801f..0d17d048653 100644 --- a/src/mesa/drivers/dri/i915/i915_program.h +++ b/src/mesa/drivers/dri/i915/i915_program.h @@ -145,7 +145,7 @@ extern GLuint i915_emit_param4fv(struct i915_fragment_program *p, const GLfloat * values); extern void i915_program_error(struct i915_fragment_program *p, - const char *msg); + const char *fmt, ...); extern void i915_init_program(struct i915_context *i915, struct i915_fragment_program *p); @@ -155,7 +155,6 @@ extern void i915_upload_program(struct i915_context *i915, extern void i915_fini_program(struct i915_fragment_program *p); - - +extern void i915_update_program(GLcontext *ctx); #endif diff --git a/src/mesa/drivers/dri/i915/i915_reg.h b/src/mesa/drivers/dri/i915/i915_reg.h index b5fa7fddb96..7f31ff674f2 100644 --- a/src/mesa/drivers/dri/i915/i915_reg.h +++ b/src/mesa/drivers/dri/i915/i915_reg.h @@ -626,9 +626,9 @@ #define MT_32BIT_AWVU2101010 (0xA<<3) #define MT_32BIT_GR1616 (0xB<<3) #define MT_32BIT_VU1616 (0xC<<3) -#define MT_32BIT_xI824 (0xD<<3) -#define MT_32BIT_xA824 (0xE<<3) -#define MT_32BIT_xL824 (0xF<<3) +#define MT_32BIT_x8I24 (0xD<<3) +#define MT_32BIT_x8L24 (0xE<<3) +#define MT_32BIT_x8A24 (0xF<<3) #define MT_422_YCRCB_SWAPY (0<<3) /* SURFACE_422 */ #define MT_422_YCRCB_NORMAL (1<<3) #define MT_422_YCRCB_SWAPUV (2<<3) diff --git a/src/mesa/drivers/dri/i915/i915_state.c b/src/mesa/drivers/dri/i915/i915_state.c index b60efea75bd..cc98d125dbc 100644 --- a/src/mesa/drivers/dri/i915/i915_state.c +++ b/src/mesa/drivers/dri/i915/i915_state.c @@ -585,7 +585,7 @@ i915PointSize(GLcontext * ctx, GLfloat size) { struct i915_context *i915 = I915_CONTEXT(ctx); int lis4 = i915->state.Ctx[I915_CTXREG_LIS4] & ~S4_POINT_WIDTH_MASK; - GLint point_size = (int) size; + GLint point_size = (int) round(size); DBG("%s\n", __FUNCTION__); @@ -599,6 +599,24 @@ i915PointSize(GLcontext * ctx, GLfloat size) } +static void +i915PointParameterfv(GLcontext * ctx, GLenum pname, const GLfloat *params) +{ + struct i915_context *i915 = I915_CONTEXT(ctx); + + switch (pname) { + case GL_POINT_SPRITE_COORD_ORIGIN: + /* This could be supported, but it would require modifying the fragment + * program to invert the y component of the texture coordinate by + * inserting a 'SUB tc.y, {1.0}.xxxx, tc' instruction. + */ + FALLBACK(&i915->intel, I915_FALLBACK_POINT_SPRITE_COORD_ORIGIN, + (params[0] != GL_UPPER_LEFT)); + break; + } +} + + /* ============================================================= * Color masks */ @@ -939,6 +957,17 @@ i915Enable(GLcontext * ctx, GLenum cap, GLboolean state) case GL_POLYGON_SMOOTH: break; + case GL_POINT_SPRITE: + /* This state change is handled in i915_reduced_primitive_state because + * the hardware bit should only be set when rendering points. + */ + I915_STATECHANGE(i915, I915_UPLOAD_CTX); + if (state) + i915->state.Ctx[I915_CTXREG_LIS4] |= S4_SPRITE_POINT_ENABLE; + else + i915->state.Ctx[I915_CTXREG_LIS4] &= ~S4_SPRITE_POINT_ENABLE; + break; + case GL_POINT_SMOOTH: break; @@ -1108,6 +1137,7 @@ i915InitStateFunctions(struct dd_function_table *functions) functions->LineWidth = i915LineWidth; functions->LogicOpcode = i915LogicOp; functions->PointSize = i915PointSize; + functions->PointParameterfv = i915PointParameterfv; functions->PolygonStipple = i915PolygonStipple; functions->Scissor = i915Scissor; functions->ShadeModel = i915ShadeModel; diff --git a/src/mesa/drivers/dri/i915/i915_texstate.c b/src/mesa/drivers/dri/i915/i915_texstate.c index 32d4b30cf9a..2b03331a004 100644 --- a/src/mesa/drivers/dri/i915/i915_texstate.c +++ b/src/mesa/drivers/dri/i915/i915_texstate.c @@ -27,7 +27,8 @@ #include "main/mtypes.h" #include "main/enums.h" -#include "main/texformat.h" +#include "main/macros.h" +#include "main/colormac.h" #include "intel_mipmap_tree.h" #include "intel_tex.h" @@ -37,7 +38,7 @@ static GLuint -translate_texture_format(GLuint mesa_format, GLuint internal_format, +translate_texture_format(gl_format mesa_format, GLuint internal_format, GLenum DepthMode) { switch (mesa_format) { @@ -56,10 +57,9 @@ translate_texture_format(GLuint mesa_format, GLuint internal_format, case MESA_FORMAT_ARGB4444: return MAPSURF_16BIT | MT_16BIT_ARGB4444; case MESA_FORMAT_ARGB8888: - if (internal_format == GL_RGB) - return MAPSURF_32BIT | MT_32BIT_XRGB8888; - else - return MAPSURF_32BIT | MT_32BIT_ARGB8888; + return MAPSURF_32BIT | MT_32BIT_ARGB8888; + case MESA_FORMAT_XRGB8888: + return MAPSURF_32BIT | MT_32BIT_XRGB8888; case MESA_FORMAT_YCBCR_REV: return (MAPSURF_422 | MT_422_YCRCB_NORMAL); case MESA_FORMAT_YCBCR: @@ -82,7 +82,12 @@ translate_texture_format(GLuint mesa_format, GLuint internal_format, case MESA_FORMAT_RGBA_DXT5: return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT4_5); case MESA_FORMAT_S8_Z24: - return (MAPSURF_32BIT | MT_32BIT_xI824); + if (DepthMode == GL_ALPHA) + return (MAPSURF_32BIT | MT_32BIT_x8A24); + else if (DepthMode == GL_INTENSITY) + return (MAPSURF_32BIT | MT_32BIT_x8I24); + else + return (MAPSURF_32BIT | MT_32BIT_x8L24); default: fprintf(stderr, "%s: bad image format %x\n", __FUNCTION__, mesa_format); abort(); @@ -171,13 +176,20 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3) pitch = intelObj->pitchOverride; } else { + GLuint dst_x, dst_y; + + intel_miptree_get_image_offset(intelObj->mt, intelObj->firstLevel, 0, 0, + &dst_x, &dst_y); + dri_bo_reference(intelObj->mt->region->buffer); i915->state.tex_buffer[unit] = intelObj->mt->region->buffer; - i915->state.tex_offset[unit] = intel_miptree_image_offset(intelObj->mt, - 0, intelObj-> - firstLevel); + /* XXX: This calculation is probably broken for tiled images with + * a non-page-aligned offset. + */ + i915->state.tex_offset[unit] = (dst_x + dst_y * intelObj->mt->pitch) * + intelObj->mt->cpp; - format = translate_texture_format(firstImage->TexFormat->MesaFormat, + format = translate_texture_format(firstImage->TexFormat, firstImage->InternalFormat, tObj->DepthMode); pitch = intelObj->mt->pitch * intelObj->mt->cpp; @@ -194,10 +206,10 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3) } state[I915_TEXREG_MS4] = - ((((pitch / 4) - 1) << MS4_PITCH_SHIFT) | MS4_CUBE_FACE_ENA_MASK | - ((((intelObj->lastLevel - intelObj->firstLevel) * 4)) << - MS4_MAX_LOD_SHIFT) | ((firstImage->Depth - 1) << - MS4_VOLUME_DEPTH_SHIFT)); + ((((pitch / 4) - 1) << MS4_PITCH_SHIFT) | + MS4_CUBE_FACE_ENA_MASK | + (U_FIXED(CLAMP(tObj->MaxLod, 0.0, 11.0), 2) << MS4_MAX_LOD_SHIFT) | + ((firstImage->Depth - 1) << MS4_VOLUME_DEPTH_SHIFT)); { @@ -263,8 +275,8 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3) /* YUV conversion: */ - if (firstImage->TexFormat->MesaFormat == MESA_FORMAT_YCBCR || - firstImage->TexFormat->MesaFormat == MESA_FORMAT_YCBCR_REV) + if (firstImage->TexFormat == MESA_FORMAT_YCBCR || + firstImage->TexFormat == MESA_FORMAT_YCBCR_REV) state[I915_TEXREG_SS2] |= SS2_COLORSPACE_CONVERSION; /* Shadow: @@ -293,6 +305,12 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3) GLenum wt = tObj->WrapT; GLenum wr = tObj->WrapR; + /* We program 1D textures as 2D textures, so the 2D texcoord could + * result in sampling border values if we don't set the T wrap to + * repeat. + */ + if (tObj->Target == GL_TEXTURE_1D) + wt = GL_REPEAT; /* 3D textures don't seem to respect the border color. * Fallback if there's ever a danger that they might refer to @@ -327,6 +345,9 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3) (translate_wrap_mode(wr) << SS3_TCZ_ADDR_MODE_SHIFT)); state[I915_TEXREG_SS3] |= (unit << SS3_TEXTUREMAP_INDEX_SHIFT); + state[I915_TEXREG_SS3] |= (U_FIXED(CLAMP(tObj->MinLod, 0.0, 11.0), 4) << + SS3_MIN_LOD_SHIFT); + } /* convert border color from float to ubyte */ @@ -340,15 +361,15 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3) * R channel, while the hardware uses A. Spam R into all the channels * for safety. */ - state[I915_TEXREG_SS4] = INTEL_PACKCOLOR8888(border[0], - border[0], - border[0], - border[0]); + state[I915_TEXREG_SS4] = PACK_COLOR_8888(border[0], + border[0], + border[0], + border[0]); } else { - state[I915_TEXREG_SS4] = INTEL_PACKCOLOR8888(border[0], - border[1], - border[2], - border[3]); + state[I915_TEXREG_SS4] = PACK_COLOR_8888(border[3], + border[0], + border[1], + border[2]); } diff --git a/src/mesa/drivers/dri/i915/i915_vtbl.c b/src/mesa/drivers/dri/i915/i915_vtbl.c index 9a723d3cd73..ba6be9796e1 100644 --- a/src/mesa/drivers/dri/i915/i915_vtbl.c +++ b/src/mesa/drivers/dri/i915/i915_vtbl.c @@ -32,7 +32,6 @@ #include "main/imports.h" #include "main/macros.h" #include "main/colormac.h" -#include "main/texformat.h" #include "tnl/t_context.h" #include "tnl/t_vertex.h" @@ -54,8 +53,7 @@ i915_render_prevalidate(struct intel_context *intel) { struct i915_context *i915 = i915_context(&intel->ctx); - if (!intel->Fallback) - i915ValidateFragmentProgram(i915); + i915ValidateFragmentProgram(i915); } static void @@ -589,8 +587,9 @@ i915_state_draw_region(struct intel_context *intel, DSTORG_VERT_BIAS(0x8) | /* .5 */ LOD_PRECLAMP_OGL | TEX_DEFAULT_COLOR_OGL); if (irb != NULL) { - switch (irb->texformat->MesaFormat) { + switch (irb->Base.Format) { case MESA_FORMAT_ARGB8888: + case MESA_FORMAT_XRGB8888: value |= DV_PF_8888; break; case MESA_FORMAT_RGB565: @@ -604,7 +603,7 @@ i915_state_draw_region(struct intel_context *intel, break; default: _mesa_problem(ctx, "Bad renderbuffer format: %d\n", - irb->texformat->MesaFormat); + irb->Base.Format); } } @@ -668,15 +667,6 @@ i915_new_batch(struct intel_context *intel) * difficulties associated with them (physical address requirements). */ i915->state.emitted = 0; - - /* Check that we didn't just wrap our batchbuffer at a bad time. */ - assert(!intel->no_batch_wrap); -} - -static GLuint -i915_flush_cmd(void) -{ - return MI_FLUSH | FLUSH_MAP_CACHE; } static void @@ -700,7 +690,6 @@ i915InitVtbl(struct i915_context *i915) i915->intel.vtbl.render_prevalidate = i915_render_prevalidate; i915->intel.vtbl.set_draw_region = i915_set_draw_region; i915->intel.vtbl.update_texture_state = i915UpdateTextureState; - i915->intel.vtbl.flush_cmd = i915_flush_cmd; i915->intel.vtbl.assert_not_dirty = i915_assert_not_dirty; i915->intel.vtbl.finish_batch = intel_finish_vb; } diff --git a/src/mesa/drivers/dri/i915/intel_generatemipmap.c b/src/mesa/drivers/dri/i915/intel_generatemipmap.c deleted file mode 120000 index 4c6b37ada01..00000000000 --- a/src/mesa/drivers/dri/i915/intel_generatemipmap.c +++ /dev/null @@ -1 +0,0 @@ -../intel/intel_generatemipmap.c
\ No newline at end of file diff --git a/src/mesa/drivers/dri/i915/intel_tris.c b/src/mesa/drivers/dri/i915/intel_tris.c index a905455342d..8a3ab39bc28 100644 --- a/src/mesa/drivers/dri/i915/intel_tris.c +++ b/src/mesa/drivers/dri/i915/intel_tris.c @@ -1076,7 +1076,9 @@ intelRunPipeline(GLcontext * ctx) intel->NewGLState = 0; } + intel_map_vertex_shader_textures(ctx); _tnl_run_pipeline(ctx); + intel_unmap_vertex_shader_textures(ctx); _mesa_unlock_context_textures(ctx); } @@ -1086,6 +1088,7 @@ intelRenderStart(GLcontext * ctx) { struct intel_context *intel = intel_context(ctx); + intel_check_front_buffer_rendering(intel); intel->vtbl.render_start(intel_context(ctx)); intel->vtbl.emit_state(intel); } @@ -1194,12 +1197,16 @@ getFallbackString(GLuint bit) +/** + * Enable/disable a fallback flag. + * \param bit one of INTEL_FALLBACK_x flags. + */ void -intelFallback(struct intel_context *intel, GLuint bit, GLboolean mode) +intelFallback(struct intel_context *intel, GLbitfield bit, GLboolean mode) { GLcontext *ctx = &intel->ctx; TNLcontext *tnl = TNL_CONTEXT(ctx); - GLuint oldfallback = intel->Fallback; + const GLbitfield oldfallback = intel->Fallback; if (mode) { intel->Fallback |= bit; @@ -1243,81 +1250,6 @@ union fi GLint i; }; - -/**********************************************************************/ -/* Used only with the metaops callbacks. */ -/**********************************************************************/ -static void -intel_meta_draw_poly(struct intel_context *intel, - GLuint n, - GLfloat xy[][2], - GLfloat z, GLuint color, GLfloat tex[][2]) -{ - union fi *vb; - GLint i; - unsigned int saved_vertex_size = intel->vertex_size; - - LOCK_HARDWARE(intel); - - intel->vertex_size = 6; - - /* All 3d primitives should be emitted with LOOP_CLIPRECTS, - * otherwise the drawing origin (DR4) might not be set correctly. - */ - intel_set_prim(intel, PRIM3D_TRIFAN); - vb = (union fi *) intel_get_prim_space(intel, n); - - for (i = 0; i < n; i++) { - vb[0].f = xy[i][0]; - vb[1].f = xy[i][1]; - vb[2].f = z; - vb[3].i = color; - vb[4].f = tex[i][0]; - vb[5].f = tex[i][1]; - vb += 6; - } - - INTEL_FIREVERTICES(intel); - - intel->vertex_size = saved_vertex_size; - - UNLOCK_HARDWARE(intel); -} - -static void -intel_meta_draw_quad(struct intel_context *intel, - GLfloat x0, GLfloat x1, - GLfloat y0, GLfloat y1, - GLfloat z, - GLuint color, - GLfloat s0, GLfloat s1, GLfloat t0, GLfloat t1) -{ - GLfloat xy[4][2]; - GLfloat tex[4][2]; - - xy[0][0] = x0; - xy[0][1] = y0; - xy[1][0] = x1; - xy[1][1] = y0; - xy[2][0] = x1; - xy[2][1] = y1; - xy[3][0] = x0; - xy[3][1] = y1; - - tex[0][0] = s0; - tex[0][1] = t0; - tex[1][0] = s1; - tex[1][1] = t0; - tex[2][0] = s1; - tex[2][1] = t1; - tex[3][0] = s0; - tex[3][1] = t1; - - intel_meta_draw_poly(intel, 4, xy, z, color, tex); -} - - - /**********************************************************************/ /* Initialization. */ /**********************************************************************/ @@ -1326,7 +1258,6 @@ intel_meta_draw_quad(struct intel_context *intel, void intelInitTriFuncs(GLcontext * ctx) { - struct intel_context *intel = intel_context(ctx); TNLcontext *tnl = TNL_CONTEXT(ctx); static int firsttime = 1; @@ -1343,6 +1274,4 @@ intelInitTriFuncs(GLcontext * ctx) tnl->Driver.Render.BuildVertices = _tnl_build_vertices; tnl->Driver.Render.CopyPV = _tnl_copy_pv; tnl->Driver.Render.Interp = _tnl_interp; - - intel->vtbl.meta_draw_quad = intel_meta_draw_quad; } diff --git a/src/mesa/drivers/dri/i965/Makefile b/src/mesa/drivers/dri/i965/Makefile index 6e9a9a29a39..7a55333e896 100644 --- a/src/mesa/drivers/dri/i965/Makefile +++ b/src/mesa/drivers/dri/i965/Makefile @@ -14,7 +14,6 @@ DRIVER_SOURCES = \ intel_decode.c \ intel_extensions.c \ intel_fbo.c \ - intel_generatemipmap.c \ intel_mipmap_tree.c \ intel_regions.c \ intel_screen.c \ @@ -101,6 +100,5 @@ DRI_LIB_DEPS += -ldrm_intel include ../Makefile.template -symlinks: intel_decode.o: ../intel/intel_decode.c intel_tex_layout.o: ../intel/intel_tex_layout.c diff --git a/src/mesa/drivers/dri/i965/brw_cc.c b/src/mesa/drivers/dri/i965/brw_cc.c index 1088a7a6070..ab301b9a3a0 100644 --- a/src/mesa/drivers/dri/i965/brw_cc.c +++ b/src/mesa/drivers/dri/i965/brw_cc.c @@ -34,6 +34,7 @@ #include "brw_state.h" #include "brw_defines.h" #include "brw_util.h" +#include "intel_fbo.h" #include "main/macros.h" #include "main/enums.h" @@ -44,17 +45,24 @@ static void prepare_cc_vp( struct brw_context *brw ) memset(&ccv, 0, sizeof(ccv)); - /* _NEW_VIEWPORT */ - ccv.min_depth = ctx->Viewport.Near; - ccv.max_depth = ctx->Viewport.Far; + /* _NEW_TRANSOFORM */ + if (ctx->Transform.DepthClamp) { + /* _NEW_VIEWPORT */ + ccv.min_depth = MIN2(ctx->Viewport.Near, ctx->Viewport.Far); + ccv.max_depth = MAX2(ctx->Viewport.Near, ctx->Viewport.Far); + } else { + ccv.min_depth = 0.0; + ccv.max_depth = 1.0; + } dri_bo_unreference(brw->cc.vp_bo); - brw->cc.vp_bo = brw_cache_data( &brw->cache, BRW_CC_VP, &ccv, NULL, 0 ); + brw->cc.vp_bo = brw_cache_data(&brw->cache, BRW_CC_VP, &ccv, sizeof(ccv), + NULL, 0); } const struct brw_tracked_state brw_cc_vp = { .dirty = { - .mesa = _NEW_VIEWPORT, + .mesa = _NEW_VIEWPORT | _NEW_TRANSFORM, .brw = BRW_NEW_CONTEXT, .cache = 0 }, @@ -82,6 +90,28 @@ struct brw_cc_unit_key { GLenum depth_func; }; +/** + * Modify blend function to force destination alpha to 1.0 + * + * If \c function specifies a blend function that uses destination alpha, + * replace it with a function that hard-wires destination alpha to 1.0. This + * is used when rendering to xRGB targets. + */ +static GLenum +fix_xRGB_alpha(GLenum function) +{ + switch (function) { + case GL_DST_ALPHA: + return GL_ONE; + + case GL_ONE_MINUS_DST_ALPHA: + case GL_SRC_ALPHA_SATURATE: + return GL_ZERO; + } + + return function; +} + static void cc_unit_populate_key(struct brw_context *brw, struct brw_cc_unit_key *key) { @@ -125,6 +155,17 @@ cc_unit_populate_key(struct brw_context *brw, struct brw_cc_unit_key *key) key->blend_dst_rgb = ctx->Color.BlendDstRGB; key->blend_src_a = ctx->Color.BlendSrcA; key->blend_dst_a = ctx->Color.BlendDstA; + + /* If the renderbuffer is XRGB, we have to frob the blend function to + * force the destination alpha to 1.0. This means replacing GL_DST_ALPHA + * with GL_ONE and GL_ONE_MINUS_DST_ALPAH with GL_ZERO. + */ + if (ctx->Visual.alphaBits == 0) { + key->blend_src_rgb = fix_xRGB_alpha(key->blend_src_rgb); + key->blend_src_a = fix_xRGB_alpha(key->blend_src_a); + key->blend_dst_rgb = fix_xRGB_alpha(key->blend_dst_rgb); + key->blend_dst_a = fix_xRGB_alpha(key->blend_dst_a); + } } key->alpha_enabled = ctx->Color.AlphaEnabled; diff --git a/src/mesa/drivers/dri/i965/brw_clip.c b/src/mesa/drivers/dri/i965/brw_clip.c index 20a927cf386..dbd10a52974 100644 --- a/src/mesa/drivers/dri/i965/brw_clip.c +++ b/src/mesa/drivers/dri/i965/brw_clip.c @@ -78,7 +78,7 @@ static void compile_clip_prog( struct brw_context *brw, delta = REG_SIZE; for (i = 0; i < VERT_RESULT_MAX; i++) - if (c.key.attrs & (1<<i)) { + if (c.key.attrs & BITFIELD64_BIT(i)) { c.offset[i] = delta; delta += ATTR_SIZE; } @@ -156,6 +156,7 @@ static void upload_clip_prog(struct brw_context *brw) key.attrs = brw->vs.prog_data->outputs_written; /* _NEW_LIGHT */ key.do_flat_shading = (ctx->Light.ShadeModel == GL_FLAT); + key.pv_first = (ctx->Light.ProvokingVertex == GL_FIRST_VERTEX_CONVENTION); /* _NEW_TRANSFORM */ key.nr_userclip = brw_count_bits(ctx->Transform.ClipPlanesEnabled); diff --git a/src/mesa/drivers/dri/i965/brw_clip.h b/src/mesa/drivers/dri/i965/brw_clip.h index 957df441ab0..1c6825510a2 100644 --- a/src/mesa/drivers/dri/i965/brw_clip.h +++ b/src/mesa/drivers/dri/i965/brw_clip.h @@ -42,22 +42,21 @@ * up polygon offset and flatshading at this point: */ struct brw_clip_prog_key { - GLuint attrs:32; + GLbitfield64 attrs; GLuint primitive:4; GLuint nr_userclip:3; GLuint do_flat_shading:1; + GLuint pv_first:1; GLuint do_unfilled:1; GLuint fill_cw:2; /* includes cull information */ GLuint fill_ccw:2; /* includes cull information */ GLuint offset_cw:1; GLuint offset_ccw:1; - GLuint pad0:17; - GLuint copy_bfc_cw:1; GLuint copy_bfc_ccw:1; GLuint clip_mode:3; - GLuint pad1:27; - + GLuint pad0:11; + GLfloat offset_factor; GLfloat offset_units; }; diff --git a/src/mesa/drivers/dri/i965/brw_clip_line.c b/src/mesa/drivers/dri/i965/brw_clip_line.c index 048ca620fab..fa9648f50f3 100644 --- a/src/mesa/drivers/dri/i965/brw_clip_line.c +++ b/src/mesa/drivers/dri/i965/brw_clip_line.c @@ -269,8 +269,12 @@ void brw_emit_line_clip( struct brw_clip_compile *c ) brw_clip_line_alloc_regs(c); brw_clip_init_ff_sync(c); - if (c->key.do_flat_shading) - brw_clip_copy_colors(c, 0, 1); + if (c->key.do_flat_shading) { + if (c->key.pv_first) + brw_clip_copy_colors(c, 1, 0); + else + brw_clip_copy_colors(c, 0, 1); + } clip_and_emit_line(c); } diff --git a/src/mesa/drivers/dri/i965/brw_clip_tri.c b/src/mesa/drivers/dri/i965/brw_clip_tri.c index 0efd77225e2..cf79224be43 100644 --- a/src/mesa/drivers/dri/i965/brw_clip_tri.c +++ b/src/mesa/drivers/dri/i965/brw_clip_tri.c @@ -188,14 +188,20 @@ void brw_clip_tri_flat_shade( struct brw_clip_compile *c ) brw_imm_ud(_3DPRIM_POLYGON)); is_poly = brw_IF(p, BRW_EXECUTE_1); - { + { brw_clip_copy_colors(c, 1, 0); brw_clip_copy_colors(c, 2, 0); } is_poly = brw_ELSE(p, is_poly); { - brw_clip_copy_colors(c, 0, 2); - brw_clip_copy_colors(c, 1, 2); + if (c->key.pv_first) { + brw_clip_copy_colors(c, 1, 0); + brw_clip_copy_colors(c, 2, 0); + } + else { + brw_clip_copy_colors(c, 0, 2); + brw_clip_copy_colors(c, 1, 2); + } } brw_ENDIF(p, is_poly); } diff --git a/src/mesa/drivers/dri/i965/brw_context.c b/src/mesa/drivers/dri/i965/brw_context.c index 3c5b8483197..78bea829493 100644 --- a/src/mesa/drivers/dri/i965/brw_context.c +++ b/src/mesa/drivers/dri/i965/brw_context.c @@ -105,11 +105,15 @@ GLboolean brwCreateContext( const __GLcontextModes *mesaVis, TNL_CONTEXT(ctx)->Driver.RunPipeline = _tnl_run_pipeline; + ctx->Const.MaxDrawBuffers = BRW_MAX_DRAW_BUFFERS; ctx->Const.MaxTextureImageUnits = BRW_MAX_TEX_UNIT; ctx->Const.MaxTextureCoordUnits = 8; /* Mesa limit */ ctx->Const.MaxTextureUnits = MIN2(ctx->Const.MaxTextureCoordUnits, ctx->Const.MaxTextureImageUnits); ctx->Const.MaxVertexTextureImageUnits = 0; /* no vertex shader textures */ + ctx->Const.MaxCombinedTextureImageUnits = + ctx->Const.MaxVertexTextureImageUnits + + ctx->Const.MaxTextureImageUnits; /* Mesa limits textures to 4kx4k; it would be nice to fix that someday */ @@ -125,6 +129,7 @@ GLboolean brwCreateContext( const __GLcontextModes *mesaVis, /* We want the GLSL compiler to emit code that uses condition codes */ ctx->Shader.EmitCondCodes = GL_TRUE; + ctx->Shader.EmitNVTempInitialization = GL_TRUE; ctx->Const.VertexProgram.MaxNativeInstructions = (16 * 1024); ctx->Const.VertexProgram.MaxAluInstructions = 0; diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h index a5209ac41be..e73e21433c5 100644 --- a/src/mesa/drivers/dri/i965/brw_context.h +++ b/src/mesa/drivers/dri/i965/brw_context.h @@ -115,7 +115,7 @@ * Handles blending and (presumably) depth and stencil testing. */ -#define BRW_FALLBACK_TEXTURE 0x1 + #define BRW_MAX_CURBE (32*16) struct brw_context; @@ -172,8 +172,8 @@ struct brw_fragment_program { GLuint id; /**< serial no. to identify frag progs, never re-used */ GLboolean isGLSL; /**< really, any IF/LOOP/CONT/BREAK instructions */ - dri_bo *const_buffer; /** Program constant buffer/surface */ GLboolean use_const_buffer; + dri_bo *const_buffer; /** Program constant buffer/surface */ /** for debugging, which texture units are referenced */ GLbitfield tex_units_used; @@ -231,7 +231,7 @@ struct brw_vs_prog_data { GLuint curb_read_length; GLuint urb_read_length; GLuint total_grf; - GLuint outputs_written; + GLbitfield64 outputs_written; GLuint nr_params; /**< number of float params/constants */ GLuint inputs_read; @@ -252,20 +252,23 @@ struct brw_vs_ouput_sizes { /** Number of texture sampler units */ #define BRW_MAX_TEX_UNIT 16 +/** Max number of render targets in a shader */ +#define BRW_MAX_DRAW_BUFFERS 4 + /** * Size of our surface binding table for the WM. * This contains pointers to the drawing surfaces and current texture * objects and shader constant buffers (+2). */ -#define BRW_WM_MAX_SURF (MAX_DRAW_BUFFERS + BRW_MAX_TEX_UNIT + 1) +#define BRW_WM_MAX_SURF (BRW_MAX_DRAW_BUFFERS + BRW_MAX_TEX_UNIT + 1) /** * Helpers to convert drawing buffers, textures and constant buffers * to surface binding table indexes, for WM. */ #define SURF_INDEX_DRAW(d) (d) -#define SURF_INDEX_FRAG_CONST_BUFFER (MAX_DRAW_BUFFERS) -#define SURF_INDEX_TEXTURE(t) (MAX_DRAW_BUFFERS + 1 + (t)) +#define SURF_INDEX_FRAG_CONST_BUFFER (BRW_MAX_DRAW_BUFFERS) +#define SURF_INDEX_TEXTURE(t) (BRW_MAX_DRAW_BUFFERS + 1 + (t)) /** * Size of surface binding table for the VS. @@ -317,7 +320,6 @@ struct brw_cache_item { GLuint nr_reloc_bufs; dri_bo *bo; - GLuint data_size; struct brw_cache_item *next; }; @@ -330,7 +332,6 @@ struct brw_cache { struct brw_cache_item **items; GLuint size, n_items; - GLuint key_size[BRW_MAX_CACHE]; /* for fixed-size keys */ GLuint aux_size[BRW_MAX_CACHE]; char *name[BRW_MAX_CACHE]; @@ -410,23 +411,6 @@ struct brw_vertex_info { GLuint sizes[ATTRIB_BIT_DWORDS * 2]; /* sizes:2[VERT_ATTRIB_MAX] */ }; - - - -/* Cache for TNL programs. - */ -struct brw_tnl_cache_item { - GLuint hash; - void *key; - void *data; - struct brw_tnl_cache_item *next; -}; - -struct brw_tnl_cache { - struct brw_tnl_cache_item **items; - GLuint size, n_items; -}; - struct brw_query_object { struct gl_query_object Base; @@ -454,8 +438,6 @@ struct brw_context GLuint primitive; GLboolean emit_state_always; - GLboolean tmp_fallback; - GLboolean no_batch_wrap; struct { struct brw_state_flags dirty; @@ -705,10 +687,6 @@ void brw_debug_batch(struct intel_context *intel); /*====================================================================== * brw_tex.c */ -void brwUpdateTextureState( struct intel_context *intel ); -void brw_FrameBufferTexInit( struct brw_context *brw, - struct intel_region *region ); -void brw_FrameBufferTexDestroy( struct brw_context *brw ); void brw_validate_textures( struct brw_context *brw ); @@ -763,9 +741,5 @@ brw_fragment_program_const(const struct gl_fragment_program *p) return (const struct brw_fragment_program *) p; } - - -#define DO_SETUP_BITS ((1<<(FRAG_ATTRIB_MAX)) - 1) - #endif diff --git a/src/mesa/drivers/dri/i965/brw_curbe.c b/src/mesa/drivers/dri/i965/brw_curbe.c index 0b0e6931a06..aadcfbe2dac 100644 --- a/src/mesa/drivers/dri/i965/brw_curbe.c +++ b/src/mesa/drivers/dri/i965/brw_curbe.c @@ -130,7 +130,7 @@ static void calculate_curbe_offsets( struct brw_context *brw ) const struct brw_tracked_state brw_curbe_offsets = { .dirty = { .mesa = _NEW_TRANSFORM, - .brw = BRW_NEW_VERTEX_PROGRAM, + .brw = BRW_NEW_VERTEX_PROGRAM | BRW_NEW_CONTEXT, .cache = CACHE_NEW_WM_PROG }, .prepare = calculate_curbe_offsets @@ -248,6 +248,9 @@ static void prepare_constant_buffer(struct brw_context *brw) GLuint offset = brw->curbe.vs_start * 16; GLuint nr = brw->vs.prog_data->nr_params / 4; + if (brw->vertex_program->IsNVProgram) + _mesa_load_tracked_matrices(ctx); + /* Updates the ParamaterValues[i] pointers for all parameters of the * basic type of PROGRAM_STATE_VAR. */ diff --git a/src/mesa/drivers/dri/i965/brw_defines.h b/src/mesa/drivers/dri/i965/brw_defines.h index 78d457ad2be..c19510bbd42 100644 --- a/src/mesa/drivers/dri/i965/brw_defines.h +++ b/src/mesa/drivers/dri/i965/brw_defines.h @@ -673,18 +673,10 @@ #define BRW_SAMPLER_MESSAGE_SIMD8_LD 3 #define BRW_SAMPLER_MESSAGE_SIMD16_LD 3 -#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_IGDNG 0 -#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_IGDNG 0 -#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_IGDNG 0 -#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_BIAS_IGDNG 1 -#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_BIAS_IGDNG 1 -#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS_IGDNG 1 -#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_LOD_IGDNG 2 -#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_LOD_IGDNG 2 -#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_LOD_IGDNG 2 -#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_COMPARE_IGDNG 3 -#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_COMPARE_IGDNG 3 -#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE_IGDNG 3 +#define BRW_SAMPLER_MESSAGE_SAMPLE_IGDNG 0 +#define BRW_SAMPLER_MESSAGE_SAMPLE_BIAS_IGDNG 1 +#define BRW_SAMPLER_MESSAGE_SAMPLE_LOD_IGDNG 2 +#define BRW_SAMPLER_MESSAGE_SAMPLE_COMPARE_IGDNG 3 /* for IGDNG only */ #define BRW_SAMPLER_SIMD_MODE_SIMD4X2 0 diff --git a/src/mesa/drivers/dri/i965/brw_draw.c b/src/mesa/drivers/dri/i965/brw_draw.c index 44bb7bd5882..7ad860898fc 100644 --- a/src/mesa/drivers/dri/i965/brw_draw.c +++ b/src/mesa/drivers/dri/i965/brw_draw.c @@ -145,7 +145,7 @@ static void brw_emit_prim(struct brw_context *brw, prim_packet.base_vert_location = prim->basevertex; /* Can't wrap here, since we rely on the validated state. */ - brw->no_batch_wrap = GL_TRUE; + intel->no_batch_wrap = GL_TRUE; /* If we're set to always flush, do it before and after the primitive emit. * We want to catch both missed flushes that hurt instruction/state cache @@ -153,21 +153,17 @@ static void brw_emit_prim(struct brw_context *brw, * the besides the draw code. */ if (intel->always_flush_cache) { - BEGIN_BATCH(1, IGNORE_CLIPRECTS); - OUT_BATCH(intel->vtbl.flush_cmd()); - ADVANCE_BATCH(); + intel_batchbuffer_emit_mi_flush(intel->batch); } if (prim_packet.verts_per_instance) { intel_batchbuffer_data( brw->intel.batch, &prim_packet, sizeof(prim_packet), LOOP_CLIPRECTS); } if (intel->always_flush_cache) { - BEGIN_BATCH(1, IGNORE_CLIPRECTS); - OUT_BATCH(intel->vtbl.flush_cmd()); - ADVANCE_BATCH(); + intel_batchbuffer_emit_mi_flush(intel->batch); } - brw->no_batch_wrap = GL_FALSE; + intel->no_batch_wrap = GL_FALSE; } static void brw_merge_inputs( struct brw_context *brw, diff --git a/src/mesa/drivers/dri/i965/brw_draw_upload.c b/src/mesa/drivers/dri/i965/brw_draw_upload.c index 765ae5a2fe5..7c796dae936 100644 --- a/src/mesa/drivers/dri/i965/brw_draw_upload.c +++ b/src/mesa/drivers/dri/i965/brw_draw_upload.c @@ -243,14 +243,6 @@ static void wrap_buffers( struct brw_context *brw, dri_bo_unreference(brw->vb.upload.bo); brw->vb.upload.bo = dri_bo_alloc(brw->intel.bufmgr, "temporary VBO", size, 1); - - /* Set the internal VBO\ to no-backing-store. We only use them as a - * temporary within a brw_try_draw_prims while the lock is held. - */ - /* DON'T DO THIS AS IF WE HAVE TO RE-ORG MEMORY WE NEED SOMEWHERE WITH - FAKE TO PUSH THIS STUFF */ -// if (!brw->intel.ttm) -// dri_bo_fake_disable_backing_store(brw->vb.upload.bo, NULL, NULL); } static void get_space( struct brw_context *brw, @@ -375,7 +367,7 @@ static void brw_prepare_vertices(struct brw_context *brw) * isn't an issue at this point. */ if (brw->vb.nr_enabled >= BRW_VEP_MAX) { - intel->Fallback = 1; + intel->Fallback = GL_TRUE; /* boolean, not bitfield */ return; } @@ -427,7 +419,7 @@ static void brw_prepare_vertices(struct brw_context *brw) /* Position array not properly enabled: */ if (input->glarray->StrideB == 0) { - intel->Fallback = 1; + intel->Fallback = GL_TRUE; /* boolean, not bitfield */ return; } @@ -536,16 +528,9 @@ static void brw_emit_vertices(struct brw_context *brw) I915_GEM_DOMAIN_VERTEX, 0, input->offset); if (BRW_IS_IGDNG(brw)) { - if (input->stride) { - OUT_RELOC(input->bo, - I915_GEM_DOMAIN_VERTEX, 0, - input->offset + input->stride * input->count); - } else { - assert(input->count == 1); - OUT_RELOC(input->bo, - I915_GEM_DOMAIN_VERTEX, 0, - input->offset + input->element_size); - } + OUT_RELOC(input->bo, + I915_GEM_DOMAIN_VERTEX, 0, + input->bo->size - 1); } else OUT_BATCH(input->stride ? input->count : 0); OUT_BATCH(0); /* Instance data step rate */ @@ -726,7 +711,7 @@ static void brw_emit_index_buffer(struct brw_context *brw) brw->ib.offset); OUT_RELOC(brw->ib.bo, I915_GEM_DOMAIN_VERTEX, 0, - brw->ib.offset + brw->ib.size); + brw->ib.offset + brw->ib.size - 1); OUT_BATCH( 0 ); ADVANCE_BATCH(); } diff --git a/src/mesa/drivers/dri/i965/brw_eu.h b/src/mesa/drivers/dri/i965/brw_eu.h index 30603bdd0e6..39eb88d7c2b 100644 --- a/src/mesa/drivers/dri/i965/brw_eu.h +++ b/src/mesa/drivers/dri/i965/brw_eu.h @@ -170,11 +170,11 @@ static INLINE struct brw_reg brw_reg( GLuint file, GLuint writemask ) { struct brw_reg reg; - if (type == BRW_GENERAL_REGISTER_FILE) + if (file == BRW_GENERAL_REGISTER_FILE) assert(nr < BRW_MAX_GRF); - else if (type == BRW_MESSAGE_REGISTER_FILE) - assert(nr < BRW_MAX_MRF); - else if (type == BRW_ARCHITECTURE_REGISTER_FILE) + else if (file == BRW_MESSAGE_REGISTER_FILE) + assert((nr & ~(1 << 7)) < BRW_MAX_MRF); + else if (file == BRW_ARCHITECTURE_REGISTER_FILE) assert(nr <= BRW_ARF_IP); reg.type = type; @@ -538,7 +538,7 @@ static INLINE struct brw_reg brw_mask_reg( GLuint subnr ) static INLINE struct brw_reg brw_message_reg( GLuint nr ) { - assert(nr < BRW_MAX_MRF); + assert((nr & ~(1 << 7)) < BRW_MAX_MRF); return brw_vec8_reg(BRW_MESSAGE_REGISTER_FILE, nr, 0); diff --git a/src/mesa/drivers/dri/i965/brw_eu_emit.c b/src/mesa/drivers/dri/i965/brw_eu_emit.c index 241cdc33f86..7ceabba288a 100644 --- a/src/mesa/drivers/dri/i965/brw_eu_emit.c +++ b/src/mesa/drivers/dri/i965/brw_eu_emit.c @@ -55,7 +55,8 @@ static void guess_execution_size( struct brw_instruction *insn, static void brw_set_dest( struct brw_instruction *insn, struct brw_reg dest ) { - if (dest.type != BRW_ARCHITECTURE_REGISTER_FILE) + if (dest.file != BRW_ARCHITECTURE_REGISTER_FILE && + dest.file != BRW_MESSAGE_REGISTER_FILE) assert(dest.nr < 128); insn->bits1.da1.dest_reg_file = dest.file; diff --git a/src/mesa/drivers/dri/i965/brw_fallback.c b/src/mesa/drivers/dri/i965/brw_fallback.c index d27c6c24ca5..562a17844b4 100644 --- a/src/mesa/drivers/dri/i965/brw_fallback.c +++ b/src/mesa/drivers/dri/i965/brw_fallback.c @@ -133,7 +133,11 @@ const struct brw_tracked_state brw_check_fallback = { -/* Not used: +/** + * Called by the INTEL_FALLBACK() macro. + * NOTE: this is a no-op for the i965 driver. The brw->intel.Fallback + * field is treated as a boolean, not a bitmask. It's only set in a + * couple of places. */ void intelFallback( struct intel_context *intel, GLuint bit, GLboolean mode ) { diff --git a/src/mesa/drivers/dri/i965/brw_gs.c b/src/mesa/drivers/dri/i965/brw_gs.c index 48c2b9a41ce..610b6c35e2a 100644 --- a/src/mesa/drivers/dri/i965/brw_gs.c +++ b/src/mesa/drivers/dri/i965/brw_gs.c @@ -85,10 +85,10 @@ static void compile_gs_prog( struct brw_context *brw, */ switch (key->primitive) { case GL_QUADS: - brw_gs_quads( &c ); + brw_gs_quads( &c, key ); break; case GL_QUAD_STRIP: - brw_gs_quad_strip( &c ); + brw_gs_quad_strip( &c, key ); break; case GL_LINE_LOOP: brw_gs_lines( &c ); @@ -149,6 +149,7 @@ static const GLenum gs_prim[GL_POLYGON+1] = { static void populate_key( struct brw_context *brw, struct brw_gs_prog_key *key ) { + GLcontext *ctx = &brw->intel.ctx; memset(key, 0, sizeof(*key)); /* CACHE_NEW_VS_PROG */ @@ -158,6 +159,9 @@ static void populate_key( struct brw_context *brw, key->primitive = gs_prim[brw->primitive]; key->hint_gs_always = 0; /* debug code? */ + + /* _NEW_LIGHT */ + key->pv_first = (ctx->Light.ProvokingVertex == GL_FIRST_VERTEX_CONVENTION); key->need_gs_prog = (key->hint_gs_always || brw->primitive == GL_QUADS || @@ -193,7 +197,7 @@ static void prepare_gs_prog(struct brw_context *brw) const struct brw_tracked_state brw_gs_prog = { .dirty = { - .mesa = 0, + .mesa = _NEW_LIGHT, .brw = BRW_NEW_PRIMITIVE, .cache = CACHE_NEW_VS_PROG }, diff --git a/src/mesa/drivers/dri/i965/brw_gs.h b/src/mesa/drivers/dri/i965/brw_gs.h index bbb991ea2e5..010c1c23521 100644 --- a/src/mesa/drivers/dri/i965/brw_gs.h +++ b/src/mesa/drivers/dri/i965/brw_gs.h @@ -40,11 +40,12 @@ #define MAX_GS_VERTS (4) struct brw_gs_prog_key { - GLuint attrs:32; + GLbitfield64 attrs; GLuint primitive:4; GLuint hint_gs_always:1; + GLuint pv_first:1; GLuint need_gs_prog:1; - GLuint pad:26; + GLuint pad:25; }; struct brw_gs_compile { @@ -67,8 +68,8 @@ struct brw_gs_compile { #define ATTR_SIZE (4*4) -void brw_gs_quads( struct brw_gs_compile *c ); -void brw_gs_quad_strip( struct brw_gs_compile *c ); +void brw_gs_quads( struct brw_gs_compile *c, struct brw_gs_prog_key *key ); +void brw_gs_quad_strip( struct brw_gs_compile *c, struct brw_gs_prog_key *key ); void brw_gs_tris( struct brw_gs_compile *c ); void brw_gs_lines( struct brw_gs_compile *c ); void brw_gs_points( struct brw_gs_compile *c ); diff --git a/src/mesa/drivers/dri/i965/brw_gs_emit.c b/src/mesa/drivers/dri/i965/brw_gs_emit.c index a9b2aa2eace..0fc5b02c619 100644 --- a/src/mesa/drivers/dri/i965/brw_gs_emit.c +++ b/src/mesa/drivers/dri/i965/brw_gs_emit.c @@ -120,7 +120,7 @@ static void brw_gs_ff_sync(struct brw_gs_compile *c, int num_prim) } -void brw_gs_quads( struct brw_gs_compile *c ) +void brw_gs_quads( struct brw_gs_compile *c, struct brw_gs_prog_key *key ) { brw_gs_alloc_regs(c, 4); @@ -128,23 +128,39 @@ void brw_gs_quads( struct brw_gs_compile *c ) * is the PV for quads, but vertex 0 for polygons: */ if (c->need_ff_sync) - brw_gs_ff_sync(c, 1); - brw_gs_emit_vue(c, c->reg.vertex[3], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START)); - brw_gs_emit_vue(c, c->reg.vertex[0], 0, (_3DPRIM_POLYGON << 2)); - brw_gs_emit_vue(c, c->reg.vertex[1], 0, (_3DPRIM_POLYGON << 2)); - brw_gs_emit_vue(c, c->reg.vertex[2], 1, ((_3DPRIM_POLYGON << 2) | R02_PRIM_END)); + brw_gs_ff_sync(c, 1); + if (key->pv_first) { + brw_gs_emit_vue(c, c->reg.vertex[0], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START)); + brw_gs_emit_vue(c, c->reg.vertex[1], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[2], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[3], 1, ((_3DPRIM_POLYGON << 2) | R02_PRIM_END)); + } + else { + brw_gs_emit_vue(c, c->reg.vertex[3], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START)); + brw_gs_emit_vue(c, c->reg.vertex[0], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[1], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[2], 1, ((_3DPRIM_POLYGON << 2) | R02_PRIM_END)); + } } -void brw_gs_quad_strip( struct brw_gs_compile *c ) +void brw_gs_quad_strip( struct brw_gs_compile *c, struct brw_gs_prog_key *key ) { brw_gs_alloc_regs(c, 4); if (c->need_ff_sync) brw_gs_ff_sync(c, 1); - brw_gs_emit_vue(c, c->reg.vertex[2], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START)); - brw_gs_emit_vue(c, c->reg.vertex[3], 0, (_3DPRIM_POLYGON << 2)); - brw_gs_emit_vue(c, c->reg.vertex[0], 0, (_3DPRIM_POLYGON << 2)); - brw_gs_emit_vue(c, c->reg.vertex[1], 1, ((_3DPRIM_POLYGON << 2) | R02_PRIM_END)); + if (key->pv_first) { + brw_gs_emit_vue(c, c->reg.vertex[0], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START)); + brw_gs_emit_vue(c, c->reg.vertex[1], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[2], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[3], 1, ((_3DPRIM_POLYGON << 2) | R02_PRIM_END)); + } + else { + brw_gs_emit_vue(c, c->reg.vertex[2], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START)); + brw_gs_emit_vue(c, c->reg.vertex[3], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[0], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[1], 1, ((_3DPRIM_POLYGON << 2) | R02_PRIM_END)); + } } void brw_gs_tris( struct brw_gs_compile *c ) diff --git a/src/mesa/drivers/dri/i965/brw_misc_state.c b/src/mesa/drivers/dri/i965/brw_misc_state.c index ea718575484..4b0d5983360 100644 --- a/src/mesa/drivers/dri/i965/brw_misc_state.c +++ b/src/mesa/drivers/dri/i965/brw_misc_state.c @@ -66,7 +66,7 @@ static void upload_blend_constant_color(struct brw_context *brw) const struct brw_tracked_state brw_blend_constant_color = { .dirty = { .mesa = _NEW_COLOR, - .brw = 0, + .brw = BRW_NEW_CONTEXT, .cache = 0 }, .emit = upload_blend_constant_color @@ -93,7 +93,7 @@ static void upload_drawing_rect(struct brw_context *brw) const struct brw_tracked_state brw_drawing_rect = { .dirty = { .mesa = _NEW_BUFFERS, - .brw = 0, + .brw = BRW_NEW_CONTEXT, .cache = 0 }, .emit = upload_drawing_rect @@ -317,7 +317,7 @@ static void upload_polygon_stipple(struct brw_context *brw) const struct brw_tracked_state brw_polygon_stipple = { .dirty = { .mesa = _NEW_POLYGONSTIPPLE, - .brw = 0, + .brw = BRW_NEW_CONTEXT, .cache = 0 }, .emit = upload_polygon_stipple @@ -362,7 +362,7 @@ static void upload_polygon_stipple_offset(struct brw_context *brw) const struct brw_tracked_state brw_polygon_stipple_offset = { .dirty = { .mesa = _NEW_WINDOW_POS, - .brw = 0, + .brw = BRW_NEW_CONTEXT, .cache = 0 }, .emit = upload_polygon_stipple_offset @@ -425,7 +425,7 @@ static void upload_line_stipple(struct brw_context *brw) const struct brw_tracked_state brw_line_stipple = { .dirty = { .mesa = _NEW_LINE, - .brw = 0, + .brw = BRW_NEW_CONTEXT, .cache = 0 }, .emit = upload_line_stipple diff --git a/src/mesa/drivers/dri/i965/brw_sf.c b/src/mesa/drivers/dri/i965/brw_sf.c index e1c2c7777b5..968890f7fb1 100644 --- a/src/mesa/drivers/dri/i965/brw_sf.c +++ b/src/mesa/drivers/dri/i965/brw_sf.c @@ -61,7 +61,7 @@ static void compile_sf_prog( struct brw_context *brw, c.key = *key; c.nr_attrs = brw_count_bits(c.key.attrs); c.nr_attr_regs = (c.nr_attrs+1)/2; - c.nr_setup_attrs = brw_count_bits(c.key.attrs & DO_SETUP_BITS); + c.nr_setup_attrs = brw_count_bits(c.key.attrs); c.nr_setup_regs = (c.nr_setup_attrs+1)/2; c.prog_data.urb_read_length = c.nr_attr_regs; @@ -70,7 +70,7 @@ static void compile_sf_prog( struct brw_context *brw, /* Construct map from attribute number to position in the vertex. */ for (i = idx = 0; i < VERT_RESULT_MAX; i++) - if (c.key.attrs & (1<<i)) { + if (c.key.attrs & BITFIELD64_BIT(i)) { c.attr_to_idx[i] = idx; c.idx_to_attr[idx] = i; if (i >= VERT_RESULT_TEX0 && i <= VERT_RESULT_TEX7) { @@ -147,7 +147,7 @@ static void upload_sf_prog(struct brw_context *brw) * edgeflag testing here, it is already done in the clip * program. */ - if (key.attrs & (1<<VERT_RESULT_EDGE)) + if (key.attrs & BITFIELD64_BIT(VERT_RESULT_EDGE)) key.primitive = SF_UNFILLED_TRIS; else key.primitive = SF_TRIANGLES; @@ -161,7 +161,7 @@ static void upload_sf_prog(struct brw_context *brw) } key.do_point_sprite = ctx->Point.PointSprite; - key.SpriteOrigin = ctx->Point.SpriteOrigin; + key.sprite_origin_lower_left = (ctx->Point.SpriteOrigin == GL_LOWER_LEFT); /* _NEW_LIGHT */ key.do_flat_shading = (ctx->Light.ShadeModel == GL_FLAT); key.do_twoside_color = (ctx->Light.Enabled && ctx->Light.Model.TwoSide); diff --git a/src/mesa/drivers/dri/i965/brw_sf.h b/src/mesa/drivers/dri/i965/brw_sf.h index 6426b6df9ff..0ba731fac99 100644 --- a/src/mesa/drivers/dri/i965/brw_sf.h +++ b/src/mesa/drivers/dri/i965/brw_sf.h @@ -45,19 +45,19 @@ #define SF_UNFILLED_TRIS 3 struct brw_sf_prog_key { - GLuint attrs:32; + GLbitfield64 attrs; GLuint primitive:2; GLuint do_twoside_color:1; GLuint do_flat_shading:1; GLuint frontface_ccw:1; GLuint do_point_sprite:1; GLuint linear_color:1; /**< linear interp vs. perspective interp */ - GLuint pad:25; - GLenum SpriteOrigin; + GLuint sprite_origin_lower_left:1; + GLuint pad:24; }; struct brw_sf_point_tex { - GLboolean CoordReplace; + GLboolean CoordReplace; }; struct brw_sf_compile { diff --git a/src/mesa/drivers/dri/i965/brw_sf_emit.c b/src/mesa/drivers/dri/i965/brw_sf_emit.c index ca8f97f9f9e..3eae41ee742 100644 --- a/src/mesa/drivers/dri/i965/brw_sf_emit.c +++ b/src/mesa/drivers/dri/i965/brw_sf_emit.c @@ -56,7 +56,7 @@ static struct brw_reg get_vert_attr(struct brw_sf_compile *c, static GLboolean have_attr(struct brw_sf_compile *c, GLuint attr) { - return (c->key.attrs & (1<<attr)) ? 1 : 0; + return (c->key.attrs & BITFIELD64_BIT(attr)) ? 1 : 0; } /*********************************************************************** @@ -122,8 +122,8 @@ static void do_twoside_color( struct brw_sf_compile *c ) * Flat shading */ -#define VERT_RESULT_COLOR_BITS ((1<<VERT_RESULT_COL0) | \ - (1<<VERT_RESULT_COL1)) +#define VERT_RESULT_COLOR_BITS (BITFIELD64_BIT(VERT_RESULT_COL0) | \ + BITFIELD64_BIT(VERT_RESULT_COL1)) static void copy_colors( struct brw_sf_compile *c, struct brw_reg dst, @@ -312,8 +312,8 @@ static GLboolean calculate_masks( struct brw_sf_compile *c, GLushort *pc_linear) { GLboolean is_last_attr = (reg == c->nr_setup_regs - 1); - GLuint persp_mask; - GLuint linear_mask; + GLbitfield64 persp_mask; + GLbitfield64 linear_mask; if (c->key.do_flat_shading || c->key.linear_color) persp_mask = c->key.attrs & ~(FRAG_BIT_WPOS | @@ -331,10 +331,10 @@ static GLboolean calculate_masks( struct brw_sf_compile *c, *pc_linear = 0; *pc = 0xf; - if (persp_mask & (1 << c->idx_to_attr[reg*2])) + if (persp_mask & BITFIELD64_BIT(c->idx_to_attr[reg*2])) *pc_persp = 0xf; - if (linear_mask & (1 << c->idx_to_attr[reg*2])) + if (linear_mask & BITFIELD64_BIT(c->idx_to_attr[reg*2])) *pc_linear = 0xf; /* Maybe only processs one attribute on the final round: @@ -342,10 +342,10 @@ static GLboolean calculate_masks( struct brw_sf_compile *c, if (reg*2+1 < c->nr_setup_attrs) { *pc |= 0xf0; - if (persp_mask & (1 << c->idx_to_attr[reg*2+1])) + if (persp_mask & BITFIELD64_BIT(c->idx_to_attr[reg*2+1])) *pc_persp |= 0xf0; - if (linear_mask & (1 << c->idx_to_attr[reg*2+1])) + if (linear_mask & BITFIELD64_BIT(c->idx_to_attr[reg*2+1])) *pc_linear |= 0xf0; } @@ -551,7 +551,7 @@ void brw_emit_point_sprite_setup( struct brw_sf_compile *c, GLboolean allocate) BRW_MATH_DATA_SCALAR, BRW_MATH_PRECISION_FULL); - if (c->key.SpriteOrigin == GL_LOWER_LEFT) { + if (c->key.sprite_origin_lower_left) { brw_MUL(p, c->m1Cx, c->tmp, c->inv_w[0]); brw_MOV(p, vec1(suboffset(c->m1Cx, 1)), brw_imm_f(0.0)); brw_MUL(p, c->m2Cy, c->tmp, negate(c->inv_w[0])); @@ -570,7 +570,7 @@ void brw_emit_point_sprite_setup( struct brw_sf_compile *c, GLboolean allocate) { brw_set_predicate_control_flag_value(p, pc); if (tex->CoordReplace) { - if (c->key.SpriteOrigin == GL_LOWER_LEFT) { + if (c->key.sprite_origin_lower_left) { brw_MUL(p, c->m3C0, c->inv_w[0], brw_imm_f(1.0)); brw_MOV(p, vec1(suboffset(c->m3C0, 0)), brw_imm_f(0.0)); } diff --git a/src/mesa/drivers/dri/i965/brw_sf_state.c b/src/mesa/drivers/dri/i965/brw_sf_state.c index bc0f0760738..bb69435ec0a 100644 --- a/src/mesa/drivers/dri/i965/brw_sf_state.c +++ b/src/mesa/drivers/dri/i965/brw_sf_state.c @@ -93,7 +93,8 @@ static void upload_sf_vp(struct brw_context *brw) } dri_bo_unreference(brw->sf.vp_bo); - brw->sf.vp_bo = brw_cache_data( &brw->cache, BRW_SF_VP, &sfv, NULL, 0 ); + brw->sf.vp_bo = brw_cache_data(&brw->cache, BRW_SF_VP, &sfv, sizeof(sfv), + NULL, 0); } const struct brw_tracked_state brw_sf_vp = { @@ -113,7 +114,8 @@ struct brw_sf_unit_key { unsigned int nr_urb_entries, urb_size, sfsize; - GLenum front_face, cull_face, provoking_vertex; + GLenum front_face, cull_face; + unsigned pv_first:1; unsigned scissor:1; unsigned line_smooth:1; unsigned point_sprite:1; @@ -154,7 +156,7 @@ sf_unit_populate_key(struct brw_context *brw, struct brw_sf_unit_key *key) key->point_attenuated = ctx->Point._Attenuated; /* _NEW_LIGHT */ - key->provoking_vertex = ctx->Light.ProvokingVertex; + key->pv_first = (ctx->Light.ProvokingVertex == GL_FIRST_VERTEX_CONVENTION); key->render_to_fbo = brw->intel.ctx.DrawBuffer->Name != 0; } @@ -287,7 +289,7 @@ sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key, /* might be BRW_NEW_PRIMITIVE if we have to adjust pv for polygons: */ - if (key->provoking_vertex == GL_LAST_VERTEX_CONVENTION) { + if (!key->pv_first) { sf.sf7.trifan_pv = 2; sf.sf7.linestrip_pv = 1; sf.sf7.tristrip_pv = 2; diff --git a/src/mesa/drivers/dri/i965/brw_state.h b/src/mesa/drivers/dri/i965/brw_state.h index 78572356a3d..b129b1f1c3f 100644 --- a/src/mesa/drivers/dri/i965/brw_state.h +++ b/src/mesa/drivers/dri/i965/brw_state.h @@ -86,9 +86,6 @@ const struct brw_tracked_state brw_psp_urb_cbs; const struct brw_tracked_state brw_pipe_control; -const struct brw_tracked_state brw_clear_surface_cache; -const struct brw_tracked_state brw_clear_batch_cache; - const struct brw_tracked_state brw_drawing_rect; const struct brw_tracked_state brw_indices; const struct brw_tracked_state brw_vertices; @@ -115,6 +112,7 @@ void brw_validate_state(struct brw_context *brw); void brw_upload_state(struct brw_context *brw); void brw_init_state(struct brw_context *brw); void brw_destroy_state(struct brw_context *brw); +void brw_clear_validated_bos(struct brw_context *brw); /*********************************************************************** * brw_state_cache.c @@ -122,16 +120,10 @@ void brw_destroy_state(struct brw_context *brw); dri_bo *brw_cache_data(struct brw_cache *cache, enum brw_cache_id cache_id, const void *data, + GLuint size, dri_bo **reloc_bufs, GLuint nr_reloc_bufs); -dri_bo *brw_cache_data_sz(struct brw_cache *cache, - enum brw_cache_id cache_id, - const void *data, - GLuint data_size, - dri_bo **reloc_bufs, - GLuint nr_reloc_bufs); - dri_bo *brw_upload_cache( struct brw_cache *cache, enum brw_cache_id cache_id, const void *key, @@ -154,6 +146,7 @@ void brw_state_cache_check_size( struct brw_context *brw ); void brw_init_caches( struct brw_context *brw ); void brw_destroy_caches( struct brw_context *brw ); +void brw_state_cache_bo_delete(struct brw_cache *cache, dri_bo *bo); /*********************************************************************** * brw_state_batch.c @@ -165,7 +158,7 @@ GLboolean brw_cached_batch_struct( struct brw_context *brw, const void *data, GLuint sz ); void brw_destroy_batch_cache( struct brw_context *brw ); -void brw_clear_batch_cache_flush( struct brw_context *brw ); +void brw_clear_batch_cache( struct brw_context *brw ); /* brw_wm_surface_state.c */ dri_bo * diff --git a/src/mesa/drivers/dri/i965/brw_state_batch.c b/src/mesa/drivers/dri/i965/brw_state_batch.c index 811940edc05..7821898cf9b 100644 --- a/src/mesa/drivers/dri/i965/brw_state_batch.c +++ b/src/mesa/drivers/dri/i965/brw_state_batch.c @@ -79,7 +79,7 @@ GLboolean brw_cached_batch_struct( struct brw_context *brw, return GL_TRUE; } -static void clear_batch_cache( struct brw_context *brw ) +void brw_clear_batch_cache( struct brw_context *brw ) { struct brw_cached_batch_item *item = brw->cached_batch_items; @@ -93,18 +93,7 @@ static void clear_batch_cache( struct brw_context *brw ) brw->cached_batch_items = NULL; } -void brw_clear_batch_cache_flush( struct brw_context *brw ) -{ - clear_batch_cache(brw); - - brw->state.dirty.mesa |= ~0; - brw->state.dirty.brw |= ~0; - brw->state.dirty.cache |= ~0; -} - - - void brw_destroy_batch_cache( struct brw_context *brw ) { - clear_batch_cache(brw); + brw_clear_batch_cache(brw); } diff --git a/src/mesa/drivers/dri/i965/brw_state_cache.c b/src/mesa/drivers/dri/i965/brw_state_cache.c index e40d7a04164..e4c9ba7d870 100644 --- a/src/mesa/drivers/dri/i965/brw_state_cache.c +++ b/src/mesa/drivers/dri/i965/brw_state_cache.c @@ -245,7 +245,6 @@ brw_upload_cache( struct brw_cache *cache, item->bo = bo; dri_bo_reference(bo); - item->data_size = data_size; if (cache->n_items > cache->size * 1.5) rehash(cache); @@ -275,15 +274,22 @@ brw_upload_cache( struct brw_cache *cache, /** - * This doesn't really work with aux data. Use search/upload instead + * Wrapper around brw_cache_data_sz using the cache_id's canonical key size. + * + * If nr_reloc_bufs is nonzero, brw_search_cache()/brw_upload_cache() would be + * better to use, as the potentially changing offsets in the data-used-as-key + * will result in excessive cache misses. + * + * If aux data is involved, use search/upload instead. + */ dri_bo * -brw_cache_data_sz(struct brw_cache *cache, - enum brw_cache_id cache_id, - const void *data, - GLuint data_size, - dri_bo **reloc_bufs, - GLuint nr_reloc_bufs) +brw_cache_data(struct brw_cache *cache, + enum brw_cache_id cache_id, + const void *data, + GLuint data_size, + dri_bo **reloc_bufs, + GLuint nr_reloc_bufs) { dri_bo *bo; struct brw_cache_item *item; @@ -306,25 +312,6 @@ brw_cache_data_sz(struct brw_cache *cache, return bo; } - -/** - * Wrapper around brw_cache_data_sz using the cache_id's canonical key size. - * - * If nr_reloc_bufs is nonzero, brw_search_cache()/brw_upload_cache() would be - * better to use, as the potentially changing offsets in the data-used-as-key - * will result in excessive cache misses. - */ -dri_bo * -brw_cache_data(struct brw_cache *cache, - enum brw_cache_id cache_id, - const void *data, - dri_bo **reloc_bufs, - GLuint nr_reloc_bufs) -{ - return brw_cache_data_sz(cache, cache_id, data, cache->key_size[cache_id], - reloc_bufs, nr_reloc_bufs); -} - enum pool_type { DW_SURFACE_STATE, DW_GENERAL_STATE @@ -335,11 +322,9 @@ static void brw_init_cache_id(struct brw_cache *cache, const char *name, enum brw_cache_id id, - GLuint key_size, GLuint aux_size) { cache->name[id] = strdup(name); - cache->key_size[id] = key_size; cache->aux_size[id] = aux_size; } @@ -359,91 +344,76 @@ brw_init_non_surface_cache(struct brw_context *brw) brw_init_cache_id(cache, "CC_VP", BRW_CC_VP, - sizeof(struct brw_cc_viewport), 0); brw_init_cache_id(cache, "CC_UNIT", BRW_CC_UNIT, - sizeof(struct brw_cc_unit_state), 0); brw_init_cache_id(cache, "WM_PROG", BRW_WM_PROG, - sizeof(struct brw_wm_prog_key), sizeof(struct brw_wm_prog_data)); brw_init_cache_id(cache, "SAMPLER_DEFAULT_COLOR", BRW_SAMPLER_DEFAULT_COLOR, - sizeof(struct brw_sampler_default_color), 0); brw_init_cache_id(cache, "SAMPLER", BRW_SAMPLER, - 0, /* variable key/data size */ 0); brw_init_cache_id(cache, "WM_UNIT", BRW_WM_UNIT, - sizeof(struct brw_wm_unit_state), 0); brw_init_cache_id(cache, "SF_PROG", BRW_SF_PROG, - sizeof(struct brw_sf_prog_key), sizeof(struct brw_sf_prog_data)); brw_init_cache_id(cache, "SF_VP", BRW_SF_VP, - sizeof(struct brw_sf_viewport), 0); brw_init_cache_id(cache, "SF_UNIT", BRW_SF_UNIT, - sizeof(struct brw_sf_unit_state), 0); brw_init_cache_id(cache, "VS_UNIT", BRW_VS_UNIT, - sizeof(struct brw_vs_unit_state), 0); brw_init_cache_id(cache, "VS_PROG", BRW_VS_PROG, - sizeof(struct brw_vs_prog_key), sizeof(struct brw_vs_prog_data)); brw_init_cache_id(cache, "CLIP_UNIT", BRW_CLIP_UNIT, - sizeof(struct brw_clip_unit_state), 0); brw_init_cache_id(cache, "CLIP_PROG", BRW_CLIP_PROG, - sizeof(struct brw_clip_prog_key), sizeof(struct brw_clip_prog_data)); brw_init_cache_id(cache, "GS_UNIT", BRW_GS_UNIT, - sizeof(struct brw_gs_unit_state), 0); brw_init_cache_id(cache, "GS_PROG", BRW_GS_PROG, - sizeof(struct brw_gs_prog_key), sizeof(struct brw_gs_prog_data)); } @@ -463,13 +433,11 @@ brw_init_surface_cache(struct brw_context *brw) brw_init_cache_id(cache, "SS_SURFACE", BRW_SS_SURFACE, - sizeof(struct brw_surface_state), 0); brw_init_cache_id(cache, "SS_SURF_BIND", BRW_SS_SURF_BIND, - 0, 0); } @@ -517,6 +485,41 @@ brw_clear_cache(struct brw_context *brw, struct brw_cache *cache) brw->state.dirty.cache |= ~0; } +/* Clear all entries from the cache that point to the given bo. + * + * This lets us release memory for reuse earlier for known-dead buffers, + * at the cost of walking the entire hash table. + */ +void +brw_state_cache_bo_delete(struct brw_cache *cache, dri_bo *bo) +{ + struct brw_cache_item **prev; + GLuint i; + + if (INTEL_DEBUG & DEBUG_STATE) + _mesa_printf("%s\n", __FUNCTION__); + + for (i = 0; i < cache->size; i++) { + for (prev = &cache->items[i]; *prev;) { + struct brw_cache_item *c = *prev; + + if (drm_intel_bo_references(c->bo, bo)) { + int j; + + *prev = c->next; + + for (j = 0; j < c->nr_reloc_bufs; j++) + dri_bo_unreference(c->reloc_bufs[j]); + dri_bo_unreference(c->bo); + free((void *)c->key); + free(c); + cache->n_items--; + } else { + prev = &c->next; + } + } + } +} void brw_state_cache_check_size(struct brw_context *brw) diff --git a/src/mesa/drivers/dri/i965/brw_state_upload.c b/src/mesa/drivers/dri/i965/brw_state_upload.c index 414620d0b39..af8dfb4c15c 100644 --- a/src/mesa/drivers/dri/i965/brw_state_upload.c +++ b/src/mesa/drivers/dri/i965/brw_state_upload.c @@ -34,6 +34,7 @@ #include "brw_context.h" #include "brw_state.h" #include "intel_batchbuffer.h" +#include "intel_buffers.h" /* This is used to initialize brw->state.atoms[]. We could use this * list directly except for a single atom, brw_constant_buffer, which @@ -142,7 +143,7 @@ static void xor_states( struct brw_state_flags *result, result->cache = a->cache ^ b->cache; } -static void +void brw_clear_validated_bos(struct brw_context *brw) { int i; @@ -287,6 +288,7 @@ void brw_validate_state( struct brw_context *brw ) if (brw->emit_state_always) { state->mesa |= ~0; state->brw |= ~0; + state->cache |= ~0; } if (brw->fragment_program != ctx->FragmentProgram._Current) { @@ -305,9 +307,9 @@ void brw_validate_state( struct brw_context *brw ) return; if (brw->state.dirty.brw & BRW_NEW_CONTEXT) - brw_clear_batch_cache_flush(brw); + brw_clear_batch_cache(brw); - brw->intel.Fallback = 0; + brw->intel.Fallback = GL_FALSE; /* boolean, not bitfield */ /* do prepare stage for all atoms */ for (i = 0; i < Elements(atoms); i++) { @@ -323,6 +325,8 @@ void brw_validate_state( struct brw_context *brw ) } } + intel_check_front_buffer_rendering(intel); + /* Make sure that the textures which are referenced by the current * brw fragment program are actually present/valid. * If this fails, we can experience GPU lock-ups. diff --git a/src/mesa/drivers/dri/i965/brw_tex.c b/src/mesa/drivers/dri/i965/brw_tex.c index 71bff166dda..e911b105b23 100644 --- a/src/mesa/drivers/dri/i965/brw_tex.c +++ b/src/mesa/drivers/dri/i965/brw_tex.c @@ -39,38 +39,6 @@ #include "intel_tex.h" #include "brw_context.h" - -void brw_FrameBufferTexInit( struct brw_context *brw, - struct intel_region *region ) -{ - struct intel_context *intel = &brw->intel; - GLcontext *ctx = &intel->ctx; - struct gl_texture_object *obj; - struct gl_texture_image *img; - - intel->frame_buffer_texobj = obj = - ctx->Driver.NewTextureObject( ctx, (GLuint) -1, GL_TEXTURE_2D ); - - obj->MinFilter = GL_NEAREST; - obj->MagFilter = GL_NEAREST; - - img = ctx->Driver.NewTextureImage( ctx ); - - _mesa_init_teximage_fields( ctx, GL_TEXTURE_2D, img, - region->pitch, region->height, 1, 0, - region->cpp == 4 ? GL_RGBA : GL_RGB ); - - _mesa_set_tex_image( obj, GL_TEXTURE_2D, 0, img ); -} - -void brw_FrameBufferTexDestroy( struct brw_context *brw ) -{ - if (brw->intel.frame_buffer_texobj != NULL) - brw->intel.ctx.Driver.DeleteTexture( &brw->intel.ctx, - brw->intel.frame_buffer_texobj ); - brw->intel.frame_buffer_texobj = NULL; -} - /** * Finalizes all textures, completing any rendering that needs to be done * to prepare them. diff --git a/src/mesa/drivers/dri/i965/brw_tex_layout.c b/src/mesa/drivers/dri/i965/brw_tex_layout.c index 5986cbffade..e59e52ed861 100644 --- a/src/mesa/drivers/dri/i965/brw_tex_layout.c +++ b/src/mesa/drivers/dri/i965/brw_tex_layout.c @@ -86,10 +86,10 @@ GLboolean brw_miptree_layout(struct intel_context *intel, mt->pitch = intel_miptree_pitch_align(intel, mt, tiling, mt->pitch); if (mt->compressed) { - qpitch = (y_pitch + ALIGN(minify(y_pitch), align_h) + 11 * align_h) / 4 * mt->pitch * mt->cpp; + qpitch = (y_pitch + ALIGN(minify(y_pitch), align_h) + 11 * align_h) / 4; mt->total_height = (y_pitch + ALIGN(minify(y_pitch), align_h) + 11 * align_h) / 4 * 6; } else { - qpitch = (y_pitch + ALIGN(minify(y_pitch), align_h) + 11 * align_h) * mt->pitch * mt->cpp; + qpitch = (y_pitch + ALIGN(minify(y_pitch), align_h) + 11 * align_h); mt->total_height = (y_pitch + ALIGN(minify(y_pitch), align_h) + 11 * align_h) * 6; } @@ -102,7 +102,8 @@ GLboolean brw_miptree_layout(struct intel_context *intel, height, 1); for (q = 0; q < nr_images; q++) - intel_miptree_set_image_offset_ex(mt, level, q, x, y, q * qpitch); + intel_miptree_set_image_offset(mt, level, q, + x, y + q * qpitch); if (mt->compressed) img_height = MAX2(1, height/4); diff --git a/src/mesa/drivers/dri/i965/brw_util.c b/src/mesa/drivers/dri/i965/brw_util.c index ce21aa48695..bba9249d1b4 100644 --- a/src/mesa/drivers/dri/i965/brw_util.c +++ b/src/mesa/drivers/dri/i965/brw_util.c @@ -35,7 +35,7 @@ #include "brw_util.h" #include "brw_defines.h" -GLuint brw_count_bits( GLuint val ) +GLuint brw_count_bits(uint64_t val) { GLuint i; for (i = 0; val ; val >>= 1) diff --git a/src/mesa/drivers/dri/i965/brw_util.h b/src/mesa/drivers/dri/i965/brw_util.h index 33e7cd87e42..04f3175d3e1 100644 --- a/src/mesa/drivers/dri/i965/brw_util.h +++ b/src/mesa/drivers/dri/i965/brw_util.h @@ -35,7 +35,7 @@ #include "main/mtypes.h" -extern GLuint brw_count_bits( GLuint val ); +extern GLuint brw_count_bits(uint64_t val); extern GLuint brw_parameter_list_state_flags(struct gl_program_parameter_list *paramList); extern GLuint brw_translate_blend_factor( GLenum factor ); extern GLuint brw_translate_blend_equation( GLenum mode ); diff --git a/src/mesa/drivers/dri/i965/brw_vs.c b/src/mesa/drivers/dri/i965/brw_vs.c index e3111c66800..fd055e225e0 100644 --- a/src/mesa/drivers/dri/i965/brw_vs.c +++ b/src/mesa/drivers/dri/i965/brw_vs.c @@ -56,7 +56,7 @@ static void do_vs_prog( struct brw_context *brw, c.prog_data.inputs_read = vp->program.Base.InputsRead; if (c.key.copy_edgeflag) { - c.prog_data.outputs_written |= 1<<VERT_RESULT_EDGE; + c.prog_data.outputs_written |= BITFIELD64_BIT(VERT_RESULT_EDGE); c.prog_data.inputs_read |= 1<<VERT_ATTRIB_EDGEFLAG; } @@ -90,8 +90,6 @@ static void brw_upload_vs_prog(struct brw_context *brw) struct brw_vertex_program *vp = (struct brw_vertex_program *)brw->vertex_program; - assert (vp && !vp->program.IsNVProgram); - memset(&key, 0, sizeof(key)); /* Just upload the program verbatim for now. Always send it all diff --git a/src/mesa/drivers/dri/i965/brw_vs_emit.c b/src/mesa/drivers/dri/i965/brw_vs_emit.c index 1638ef81115..00efd3443d1 100644 --- a/src/mesa/drivers/dri/i965/brw_vs_emit.c +++ b/src/mesa/drivers/dri/i965/brw_vs_emit.c @@ -147,7 +147,7 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c ) mrf = 4; for (i = 0; i < VERT_RESULT_MAX; i++) { - if (c->prog_data.outputs_written & (1 << i)) { + if (c->prog_data.outputs_written & BITFIELD64_BIT(i)) { c->nr_outputs++; assert(i < Elements(c->regs[PROGRAM_OUTPUT])); if (i == VERT_RESULT_HPOS) { @@ -331,63 +331,65 @@ static void unalias3( struct brw_vs_compile *c, } } -static void emit_sop( struct brw_compile *p, +static void emit_sop( struct brw_vs_compile *c, struct brw_reg dst, struct brw_reg arg0, struct brw_reg arg1, GLuint cond) { + struct brw_compile *p = &c->func; + brw_MOV(p, dst, brw_imm_f(0.0f)); brw_CMP(p, brw_null_reg(), cond, arg0, arg1); brw_MOV(p, dst, brw_imm_f(1.0f)); brw_set_predicate_control_flag_value(p, 0xff); } -static void emit_seq( struct brw_compile *p, +static void emit_seq( struct brw_vs_compile *c, struct brw_reg dst, struct brw_reg arg0, struct brw_reg arg1 ) { - emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_EQ); + emit_sop(c, dst, arg0, arg1, BRW_CONDITIONAL_EQ); } -static void emit_sne( struct brw_compile *p, +static void emit_sne( struct brw_vs_compile *c, struct brw_reg dst, struct brw_reg arg0, struct brw_reg arg1 ) { - emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_NEQ); + emit_sop(c, dst, arg0, arg1, BRW_CONDITIONAL_NEQ); } -static void emit_slt( struct brw_compile *p, +static void emit_slt( struct brw_vs_compile *c, struct brw_reg dst, struct brw_reg arg0, struct brw_reg arg1 ) { - emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_L); + emit_sop(c, dst, arg0, arg1, BRW_CONDITIONAL_L); } -static void emit_sle( struct brw_compile *p, +static void emit_sle( struct brw_vs_compile *c, struct brw_reg dst, struct brw_reg arg0, struct brw_reg arg1 ) { - emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_LE); + emit_sop(c, dst, arg0, arg1, BRW_CONDITIONAL_LE); } -static void emit_sgt( struct brw_compile *p, +static void emit_sgt( struct brw_vs_compile *c, struct brw_reg dst, struct brw_reg arg0, struct brw_reg arg1 ) { - emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_G); + emit_sop(c, dst, arg0, arg1, BRW_CONDITIONAL_G); } -static void emit_sge( struct brw_compile *p, +static void emit_sge( struct brw_vs_compile *c, struct brw_reg dst, struct brw_reg arg0, struct brw_reg arg1 ) { - emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_GE); + emit_sop(c, dst, arg0, arg1, BRW_CONDITIONAL_GE); } static void emit_max( struct brw_compile *p, @@ -912,6 +914,7 @@ get_src_reg( struct brw_vs_compile *c, case PROGRAM_CONSTANT: case PROGRAM_UNIFORM: case PROGRAM_ENV_PARAM: + case PROGRAM_LOCAL_PARAM: if (c->vp->use_const_buffer) { return get_constant(c, inst, argIndex); } @@ -930,7 +933,6 @@ get_src_reg( struct brw_vs_compile *c, /* this is a normal case since we loop over all three src args */ return brw_null_reg(); - case PROGRAM_LOCAL_PARAM: case PROGRAM_WRITE_ONLY: default: assert(0); @@ -1122,7 +1124,7 @@ static void emit_vertex_write( struct brw_vs_compile *c) /* Update the header for point size, user clipping flags, and -ve rhw * workaround. */ - if ((c->prog_data.outputs_written & (1<<VERT_RESULT_PSIZ)) || + if ((c->prog_data.outputs_written & BITFIELD64_BIT(VERT_RESULT_PSIZ)) || c->key.nr_userclip || BRW_IS_965(p->brw)) { struct brw_reg header1 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD); @@ -1132,7 +1134,7 @@ static void emit_vertex_write( struct brw_vs_compile *c) brw_set_access_mode(p, BRW_ALIGN_16); - if (c->prog_data.outputs_written & (1<<VERT_RESULT_PSIZ)) { + if (c->prog_data.outputs_written & BITFIELD64_BIT(VERT_RESULT_PSIZ)) { struct brw_reg psiz = c->regs[PROGRAM_OUTPUT][VERT_RESULT_PSIZ]; brw_MUL(p, brw_writemask(header1, WRITEMASK_W), brw_swizzle1(psiz, 0), brw_imm_f(1<<11)); brw_AND(p, brw_writemask(header1, WRITEMASK_W), header1, brw_imm_ud(0x7ff<<8)); @@ -1222,7 +1224,7 @@ static void emit_vertex_write( struct brw_vs_compile *c) */ GLuint i, mrf = 0; for (i = c->first_overflow_output; i < VERT_RESULT_MAX; i++) { - if (c->prog_data.outputs_written & (1 << i)) { + if (c->prog_data.outputs_written & BITFIELD64_BIT(i)) { /* move from GRF to MRF */ brw_MOV(p, brw_message_reg(4+mrf), c->regs[PROGRAM_OUTPUT][i]); mrf++; @@ -1269,6 +1271,38 @@ post_vs_emit( struct brw_vs_compile *c, } } +static GLboolean +accumulator_contains(struct brw_vs_compile *c, struct brw_reg val) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *prev_insn = &p->store[p->nr_insn - 1]; + + if (p->nr_insn == 0) + return GL_FALSE; + + if (val.address_mode != BRW_ADDRESS_DIRECT) + return GL_FALSE; + + switch (prev_insn->header.opcode) { + case BRW_OPCODE_MOV: + case BRW_OPCODE_MAC: + case BRW_OPCODE_MUL: + if (prev_insn->header.access_mode == BRW_ALIGN_16 && + prev_insn->header.execution_size == val.width && + prev_insn->bits1.da1.dest_reg_file == val.file && + prev_insn->bits1.da1.dest_reg_type == val.type && + prev_insn->bits1.da1.dest_address_mode == val.address_mode && + prev_insn->bits1.da1.dest_reg_nr == val.nr && + prev_insn->bits1.da16.dest_subreg_nr == val.subnr / 16 && + prev_insn->bits1.da16.dest_writemask == 0xf) + return GL_TRUE; + else + return GL_FALSE; + default: + return GL_FALSE; + } +} + static uint32_t get_predicate(const struct prog_instruction *inst) { @@ -1447,7 +1481,8 @@ void brw_vs_emit(struct brw_vs_compile *c ) unalias3(c, dst, args[0], args[1], args[2], emit_lrp_noalias); break; case OPCODE_MAD: - brw_MOV(p, brw_acc_reg(), args[2]); + if (!accumulator_contains(c, args[2])) + brw_MOV(p, brw_acc_reg(), args[2]); brw_MAC(p, dst, args[0], args[1]); break; case OPCODE_MAX: @@ -1473,25 +1508,25 @@ void brw_vs_emit(struct brw_vs_compile *c ) break; case OPCODE_SEQ: - emit_seq(p, dst, args[0], args[1]); + unalias2(c, dst, args[0], args[1], emit_seq); break; case OPCODE_SIN: emit_math1(c, BRW_MATH_FUNCTION_SIN, dst, args[0], BRW_MATH_PRECISION_FULL); break; case OPCODE_SNE: - emit_sne(p, dst, args[0], args[1]); + unalias2(c, dst, args[0], args[1], emit_sne); break; case OPCODE_SGE: - emit_sge(p, dst, args[0], args[1]); + unalias2(c, dst, args[0], args[1], emit_sge); break; case OPCODE_SGT: - emit_sgt(p, dst, args[0], args[1]); + unalias2(c, dst, args[0], args[1], emit_sgt); break; case OPCODE_SLT: - emit_slt(p, dst, args[0], args[1]); + unalias2(c, dst, args[0], args[1], emit_slt); break; case OPCODE_SLE: - emit_sle(p, dst, args[0], args[1]); + unalias2(c, dst, args[0], args[1], emit_sle); break; case OPCODE_SUB: brw_ADD(p, dst, args[0], negate(args[1])); diff --git a/src/mesa/drivers/dri/i965/brw_vs_state.c b/src/mesa/drivers/dri/i965/brw_vs_state.c index d790ab65553..7285466645b 100644 --- a/src/mesa/drivers/dri/i965/brw_vs_state.c +++ b/src/mesa/drivers/dri/i965/brw_vs_state.c @@ -109,10 +109,39 @@ vs_unit_create_from_key(struct brw_context *brw, struct brw_vs_unit_key *key) vs.thread3.urb_entry_read_offset = 0; vs.thread3.const_urb_entry_read_offset = key->curbe_offset * 2; - if (BRW_IS_IGDNG(brw)) - vs.thread4.nr_urb_entries = key->nr_urb_entries >> 2; - else - vs.thread4.nr_urb_entries = key->nr_urb_entries; + if (BRW_IS_IGDNG(brw)) { + switch (key->nr_urb_entries) { + case 8: + case 12: + case 16: + case 32: + case 64: + case 96: + case 128: + case 168: + case 192: + case 224: + case 256: + vs.thread4.nr_urb_entries = key->nr_urb_entries >> 2; + break; + default: + assert(0); + } + } else { + switch (key->nr_urb_entries) { + case 8: + case 12: + case 16: + case 32: + break; + case 64: + assert(BRW_IS_G4X(brw)); + break; + default: + assert(0); + } + vs.thread4.nr_urb_entries = key->nr_urb_entries; + } vs.thread4.urb_entry_allocation_size = key->urb_size - 1; diff --git a/src/mesa/drivers/dri/i965/brw_vs_surface_state.c b/src/mesa/drivers/dri/i965/brw_vs_surface_state.c index 89f47522a1c..3bc9840a971 100644 --- a/src/mesa/drivers/dri/i965/brw_vs_surface_state.c +++ b/src/mesa/drivers/dri/i965/brw_vs_surface_state.c @@ -30,7 +30,6 @@ */ #include "main/mtypes.h" -#include "main/texformat.h" #include "main/texstore.h" #include "shader/prog_parameter.h" @@ -53,6 +52,7 @@ brw_vs_update_constant_buffer(struct brw_context *brw) const struct gl_program_parameter_list *params = vp->program.Base.Parameters; const int size = params->NumParameters * 4 * sizeof(GLfloat); drm_intel_bo *const_buffer; + int i; /* BRW_NEW_VERTEX_PROGRAM */ if (!vp->use_const_buffer) @@ -62,7 +62,19 @@ brw_vs_update_constant_buffer(struct brw_context *brw) size, 64); /* _NEW_PROGRAM_CONSTANTS */ - dri_bo_subdata(const_buffer, 0, size, params->ParameterValues); + + /* Updates the ParamaterValues[i] pointers for all parameters of the + * basic type of PROGRAM_STATE_VAR. + */ + _mesa_load_state_parameters(&brw->intel.ctx, vp->program.Base.Parameters); + + intel_bo_map_gtt_preferred(intel, const_buffer, GL_TRUE); + for (i = 0; i < params->NumParameters; i++) { + memcpy(const_buffer->virtual + i * 4 * sizeof(float), + params->ParameterValues[i], + 4 * sizeof(float)); + } + intel_bo_unmap_gtt_preferred(intel, const_buffer); return const_buffer; } diff --git a/src/mesa/drivers/dri/i965/brw_vtbl.c b/src/mesa/drivers/dri/i965/brw_vtbl.c index ac117901515..72749b3859d 100644 --- a/src/mesa/drivers/dri/i965/brw_vtbl.c +++ b/src/mesa/drivers/dri/i965/brw_vtbl.c @@ -46,7 +46,7 @@ #include "brw_state.h" #include "brw_fallback.h" #include "brw_vs.h" - +#include "brw_wm.h" static void dri_bo_release(dri_bo **bo) @@ -66,10 +66,14 @@ static void brw_destroy_context( struct intel_context *intel ) brw_destroy_state(brw); brw_draw_destroy( brw ); - - _mesa_free(brw->wm.compile_data); - - brw_FrameBufferTexDestroy( brw ); + brw_clear_validated_bos(brw); + if (brw->wm.compile_data) { + _mesa_free(brw->wm.compile_data->instruction); + _mesa_free(brw->wm.compile_data->vreg); + _mesa_free(brw->wm.compile_data->refs); + _mesa_free(brw->wm.compile_data->prog_instructions); + _mesa_free(brw->wm.compile_data); + } for (i = 0; i < brw->state.nr_color_regions; i++) intel_region_release(&brw->state.color_regions[i]); @@ -146,9 +150,6 @@ static void brw_new_batch( struct intel_context *intel ) { struct brw_context *brw = brw_context(&intel->ctx); - /* Check that we didn't just wrap our batchbuffer at a bad time. */ - assert(!brw->no_batch_wrap); - brw->curbe.need_new_bo = GL_TRUE; /* Mark all context state as needing to be re-emitted. @@ -177,20 +178,6 @@ static void brw_note_fence( struct intel_context *intel, GLuint fence ) brw_context(&intel->ctx)->state.dirty.brw |= BRW_NEW_FENCE; } -/* called from intelWaitForIdle() and intelFlush() - * - * For now, just flush everything. Could be smarter later. - */ -static GLuint brw_flush_cmd( void ) -{ - struct brw_mi_flush flush; - flush.opcode = CMD_MI_FLUSH; - flush.pad = 0; - flush.flags = BRW_FLUSH_STATE_CACHE; - return *(GLuint *)&flush; -} - - static void brw_invalidate_state( struct intel_context *intel, GLuint new_state ) { /* nothing */ @@ -211,6 +198,5 @@ void brwInitVtbl( struct brw_context *brw ) brw->intel.vtbl.finish_batch = brw_finish_batch; brw->intel.vtbl.destroy = brw_destroy_context; brw->intel.vtbl.set_draw_region = brw_set_draw_region; - brw->intel.vtbl.flush_cmd = brw_flush_cmd; brw->intel.vtbl.debug_batch = brw_debug_batch; } diff --git a/src/mesa/drivers/dri/i965/brw_wm.c b/src/mesa/drivers/dri/i965/brw_wm.c index 2292de94c44..6895f644104 100644 --- a/src/mesa/drivers/dri/i965/brw_wm.c +++ b/src/mesa/drivers/dri/i965/brw_wm.c @@ -29,7 +29,6 @@ * Keith Whitwell <[email protected]> */ -#include "main/texformat.h" #include "brw_context.h" #include "brw_util.h" #include "brw_wm.h" @@ -153,8 +152,21 @@ static void do_wm_prog( struct brw_context *brw, */ return; } + c->instruction = _mesa_calloc(BRW_WM_MAX_INSN * sizeof(*c->instruction)); + c->prog_instructions = _mesa_calloc(BRW_WM_MAX_INSN * + sizeof(*c->prog_instructions)); + c->vreg = _mesa_calloc(BRW_WM_MAX_VREG * sizeof(*c->vreg)); + c->refs = _mesa_calloc(BRW_WM_MAX_REF * sizeof(*c->refs)); } else { + void *instruction = c->instruction; + void *prog_instructions = c->prog_instructions; + void *vreg = c->vreg; + void *refs = c->refs; memset(c, 0, sizeof(*brw->wm.compile_data)); + c->instruction = instruction; + c->prog_instructions = prog_instructions; + c->vreg = vreg; + c->refs = refs; } memcpy(&c->key, key, sizeof(*key)); @@ -218,7 +230,7 @@ static void brw_wm_populate_key( struct brw_context *brw, ctx->Color.AlphaEnabled) lookup |= IZ_PS_KILL_ALPHATEST_BIT; - if (fp->program.Base.OutputsWritten & (1<<FRAG_RESULT_DEPTH)) + if (fp->program.Base.OutputsWritten & BITFIELD64_BIT(FRAG_RESULT_DEPTH)) lookup |= IZ_PS_COMPUTES_DEPTH_BIT; /* _NEW_DEPTH */ @@ -288,7 +300,7 @@ static void brw_wm_populate_key( struct brw_context *brw, const struct gl_texture_image *img = t->Image[0][t->BaseLevel]; if (img->InternalFormat == GL_YCBCR_MESA) { key->yuvtex_mask |= 1 << i; - if (img->TexFormat->MesaFormat == MESA_FORMAT_YCBCR) + if (img->TexFormat == MESA_FORMAT_YCBCR) key->yuvtex_swap_mask |= 1 << i; } @@ -309,6 +321,9 @@ static void brw_wm_populate_key( struct brw_context *brw, * from the incoming screen origin relative position we get as part of our * payload. * + * This is only needed for the WM_WPOSXY opcode when the fragment program + * uses the gl_FragCoord input. + * * We could avoid recompiling by including this as a constant referenced by * our program, but if we were to do that it would also be nice to handle * getting that constant updated at batchbuffer submit time (when we @@ -317,17 +332,21 @@ static void brw_wm_populate_key( struct brw_context *brw, * just avoid using this as key data if the program doesn't use * fragment.position. * - * This pretty much becomes moot with DRI2 and redirected buffers anyway, - * as our origins will always be zero then. + * For DRI2 the origin_x/y will always be (0,0) but we still need the + * drawable height in order to invert the Y axis. */ - if (brw->intel.driDrawable != NULL) { - key->origin_x = brw->intel.driDrawable->x; - key->origin_y = brw->intel.driDrawable->y; - key->drawable_height = brw->intel.driDrawable->h; + if (fp->program.Base.InputsRead & FRAG_BIT_WPOS) { + if (brw->intel.driDrawable != NULL) { + key->origin_x = brw->intel.driDrawable->x; + key->origin_y = brw->intel.driDrawable->y; + key->drawable_height = brw->intel.driDrawable->h; + } } + key->nr_color_regions = brw->state.nr_color_regions; + /* CACHE_NEW_VS_PROG */ - key->vp_outputs_written = brw->vs.prog_data->outputs_written & DO_SETUP_BITS; + key->vp_outputs_written = brw->vs.prog_data->outputs_written; /* The unique fragment program ID */ key->program_string_id = fp->id; diff --git a/src/mesa/drivers/dri/i965/brw_wm.h b/src/mesa/drivers/dri/i965/brw_wm.h index 872b1f3ecf4..b9b987ea70a 100644 --- a/src/mesa/drivers/dri/i965/brw_wm.h +++ b/src/mesa/drivers/dri/i965/brw_wm.h @@ -67,18 +67,19 @@ struct brw_wm_prog_key { GLuint flat_shade:1; GLuint linear_color:1; /**< linear interpolation vs perspective interp */ GLuint runtime_check_aads_emit:1; + GLuint nr_color_regions:2; GLbitfield proj_attrib_mask; /**< one bit per fragment program attribute */ GLuint shadowtex_mask:16; GLuint yuvtex_mask:16; GLuint yuvtex_swap_mask:16; /* UV swaped */ - GLuint tex_swizzles[BRW_MAX_TEX_UNIT]; + GLushort tex_swizzles[BRW_MAX_TEX_UNIT]; + GLushort origin_x, origin_y; + GLushort drawable_height; + GLbitfield64 vp_outputs_written; GLuint program_string_id:32; - GLuint origin_x, origin_y; - GLuint drawable_height; - GLuint vp_outputs_written; }; @@ -153,15 +154,16 @@ struct brw_wm_instruction { }; -#define BRW_WM_MAX_INSN (MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS*3 + FRAG_ATTRIB_MAX + 3) +#define BRW_WM_MAX_INSN (MAX_PROGRAM_INSTRUCTIONS*3 + FRAG_ATTRIB_MAX + 3) #define BRW_WM_MAX_GRF 128 /* hardware limit */ #define BRW_WM_MAX_VREG (BRW_WM_MAX_INSN * 4) #define BRW_WM_MAX_REF (BRW_WM_MAX_INSN * 12) #define BRW_WM_MAX_PARAM 256 #define BRW_WM_MAX_CONST 256 -#define BRW_WM_MAX_KILLS MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS #define BRW_WM_MAX_SUBROUTINE 16 +/* used in masks next to WRITEMASK_*. */ +#define SATURATE (1<<5) /* New opcodes to track internal operations required for WM unit. @@ -200,7 +202,7 @@ struct brw_wm_compile { * simplifying and adding instructions for interpolation and * framebuffer writes. */ - struct prog_instruction prog_instructions[BRW_WM_MAX_INSN]; + struct prog_instruction *prog_instructions; GLuint nr_fp_insns; GLuint fp_temp; GLuint fp_interp_emitted; @@ -211,7 +213,7 @@ struct brw_wm_compile { struct prog_src_register pixel_w; - struct brw_wm_value vreg[BRW_WM_MAX_VREG]; + struct brw_wm_value *vreg; GLuint nr_vreg; struct brw_wm_value creg[BRW_WM_MAX_PARAM]; @@ -228,10 +230,10 @@ struct brw_wm_compile { struct brw_wm_ref undef_ref; struct brw_wm_value undef_value; - struct brw_wm_ref refs[BRW_WM_MAX_REF]; + struct brw_wm_ref *refs; GLuint nr_refs; - struct brw_wm_instruction instruction[BRW_WM_MAX_INSN]; + struct brw_wm_instruction *instruction; GLuint nr_insns; struct brw_wm_constref constref[BRW_WM_MAX_CONST]; @@ -271,6 +273,12 @@ struct brw_wm_compile { }; +/** Bits for prog_instruction::Aux field */ +#define INST_AUX_EOT 0x1 +#define INST_AUX_TARGET(T) (T << 1) +#define INST_AUX_GET_TARGET(AUX) ((AUX) >> 1) + + GLuint brw_wm_nr_args( GLuint opcode ); GLuint brw_wm_is_scalar_result( GLuint opcode ); @@ -300,10 +308,141 @@ void brw_wm_lookup_iz( GLuint line_aa, GLboolean brw_wm_is_glsl(const struct gl_fragment_program *fp); void brw_wm_glsl_emit(struct brw_context *brw, struct brw_wm_compile *c); +/* brw_wm_emit.c */ +void emit_alu1(struct brw_compile *p, + struct brw_instruction *(*func)(struct brw_compile *, + struct brw_reg, + struct brw_reg), + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0); +void emit_alu2(struct brw_compile *p, + struct brw_instruction *(*func)(struct brw_compile *, + struct brw_reg, + struct brw_reg, + struct brw_reg), + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1); +void emit_cinterp(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0); void emit_ddxy(struct brw_compile *p, const struct brw_reg *dst, GLuint mask, GLboolean is_ddx, const struct brw_reg *arg0); +void emit_delta_xy(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0); +void emit_dp3(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1); +void emit_dp4(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1); +void emit_dph(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1); +void emit_fb_write(struct brw_wm_compile *c, + struct brw_reg *arg0, + struct brw_reg *arg1, + struct brw_reg *arg2, + GLuint target, + GLuint eot); +void emit_frontfacing(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask); +void emit_linterp(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *deltas); +void emit_lrp(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1, + const struct brw_reg *arg2); +void emit_mad(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1, + const struct brw_reg *arg2); +void emit_math1(struct brw_wm_compile *c, + GLuint function, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0); +void emit_math2(struct brw_wm_compile *c, + GLuint function, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1); +void emit_min(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1); +void emit_max(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1); +void emit_pinterp(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *deltas, + const struct brw_reg *w); +void emit_pixel_xy(struct brw_wm_compile *c, + const struct brw_reg *dst, + GLuint mask); +void emit_pixel_w(struct brw_wm_compile *c, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *deltas); +void emit_sop(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + GLuint cond, + const struct brw_reg *arg0, + const struct brw_reg *arg1); +void emit_tex(struct brw_wm_compile *c, + struct brw_reg *dst, + GLuint dst_flags, + struct brw_reg *arg, + struct brw_reg depth_payload, + GLuint tex_idx, + GLuint sampler, + GLboolean shadow); +void emit_txb(struct brw_wm_compile *c, + struct brw_reg *dst, + GLuint dst_flags, + struct brw_reg *arg, + struct brw_reg depth_payload, + GLuint tex_idx, + GLuint sampler); +void emit_wpos_xy(struct brw_wm_compile *c, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0); +void emit_xpd(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1); #endif diff --git a/src/mesa/drivers/dri/i965/brw_wm_emit.c b/src/mesa/drivers/dri/i965/brw_wm_emit.c index bf80a2942a4..5390fd25849 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_emit.c +++ b/src/mesa/drivers/dri/i965/brw_wm_emit.c @@ -44,6 +44,7 @@ static INLINE struct brw_reg sechalf( struct brw_reg reg ) return reg; } + /* Payload R0: * * R0.0 -- pixel mask, one bit for each of 4 pixels in 4 tiles, @@ -60,42 +61,50 @@ static INLINE struct brw_reg sechalf( struct brw_reg reg ) * R1.8 -- ? */ - -static void emit_pixel_xy(struct brw_compile *p, - const struct brw_reg *dst, - GLuint mask) +void emit_pixel_xy(struct brw_wm_compile *c, + const struct brw_reg *dst, + GLuint mask) { + struct brw_compile *p = &c->func; struct brw_reg r1 = brw_vec1_grf(1, 0); struct brw_reg r1_uw = retype(r1, BRW_REGISTER_TYPE_UW); + struct brw_reg dst0_uw, dst1_uw; + brw_push_insn_state(p); brw_set_compression_control(p, BRW_COMPRESSION_NONE); + if (c->dispatch_width == 16) { + dst0_uw = vec16(retype(dst[0], BRW_REGISTER_TYPE_UW)); + dst1_uw = vec16(retype(dst[1], BRW_REGISTER_TYPE_UW)); + } else { + dst0_uw = vec8(retype(dst[0], BRW_REGISTER_TYPE_UW)); + dst1_uw = vec8(retype(dst[1], BRW_REGISTER_TYPE_UW)); + } + /* Calculate pixel centers by adding 1 or 0 to each of the * micro-tile coordinates passed in r1. */ if (mask & WRITEMASK_X) { brw_ADD(p, - vec16(retype(dst[0], BRW_REGISTER_TYPE_UW)), + dst0_uw, stride(suboffset(r1_uw, 4), 2, 4, 0), brw_imm_v(0x10101010)); } if (mask & WRITEMASK_Y) { brw_ADD(p, - vec16(retype(dst[1], BRW_REGISTER_TYPE_UW)), + dst1_uw, stride(suboffset(r1_uw,5), 2, 4, 0), brw_imm_v(0x11001100)); } - - brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED); + brw_pop_insn_state(p); } - -static void emit_delta_xy(struct brw_compile *p, - const struct brw_reg *dst, - GLuint mask, - const struct brw_reg *arg0) +void emit_delta_xy(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0) { struct brw_reg r1 = brw_vec1_grf(1, 0); @@ -118,10 +127,10 @@ static void emit_delta_xy(struct brw_compile *p, } } -static void emit_wpos_xy(struct brw_wm_compile *c, - const struct brw_reg *dst, - GLuint mask, - const struct brw_reg *arg0) +void emit_wpos_xy(struct brw_wm_compile *c, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0) { struct brw_compile *p = &c->func; @@ -146,12 +155,14 @@ static void emit_wpos_xy(struct brw_wm_compile *c, } -static void emit_pixel_w( struct brw_compile *p, - const struct brw_reg *dst, - GLuint mask, - const struct brw_reg *arg0, - const struct brw_reg *deltas) +void emit_pixel_w(struct brw_wm_compile *c, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *deltas) { + struct brw_compile *p = &c->func; + /* Don't need this if all you are doing is interpolating color, for * instance. */ @@ -165,21 +176,29 @@ static void emit_pixel_w( struct brw_compile *p, brw_MAC(p, brw_message_reg(2), suboffset(interp3, 1), deltas[1]); /* Calc w */ - brw_math_16( p, dst[3], - BRW_MATH_FUNCTION_INV, - BRW_MATH_SATURATE_NONE, - 2, brw_null_reg(), - BRW_MATH_PRECISION_FULL); + if (c->dispatch_width == 16) { + brw_math_16(p, dst[3], + BRW_MATH_FUNCTION_INV, + BRW_MATH_SATURATE_NONE, + 2, brw_null_reg(), + BRW_MATH_PRECISION_FULL); + } else { + brw_math(p, dst[3], + BRW_MATH_FUNCTION_INV, + BRW_MATH_SATURATE_NONE, + 2, brw_null_reg(), + BRW_MATH_DATA_VECTOR, + BRW_MATH_PRECISION_FULL); + } } } - -static void emit_linterp( struct brw_compile *p, - const struct brw_reg *dst, - GLuint mask, - const struct brw_reg *arg0, - const struct brw_reg *deltas ) +void emit_linterp(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *deltas) { struct brw_reg interp[4]; GLuint nr = arg0[0].nr; @@ -199,12 +218,12 @@ static void emit_linterp( struct brw_compile *p, } -static void emit_pinterp( struct brw_compile *p, - const struct brw_reg *dst, - GLuint mask, - const struct brw_reg *arg0, - const struct brw_reg *deltas, - const struct brw_reg *w) +void emit_pinterp(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *deltas, + const struct brw_reg *w) { struct brw_reg interp[4]; GLuint nr = arg0[0].nr; @@ -229,10 +248,10 @@ static void emit_pinterp( struct brw_compile *p, } -static void emit_cinterp( struct brw_compile *p, - const struct brw_reg *dst, - GLuint mask, - const struct brw_reg *arg0 ) +void emit_cinterp(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0) { struct brw_reg interp[4]; GLuint nr = arg0[0].nr; @@ -251,9 +270,9 @@ static void emit_cinterp( struct brw_compile *p, } /* Sets the destination channels to 1.0 or 0.0 according to glFrontFacing. */ -static void emit_frontfacing( struct brw_compile *p, - const struct brw_reg *dst, - GLuint mask ) +void emit_frontfacing(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask) { struct brw_reg r1_6ud = retype(brw_vec1_grf(1, 6), BRW_REGISTER_TYPE_UD); GLuint i; @@ -352,13 +371,13 @@ void emit_ddxy(struct brw_compile *p, brw_set_saturate(p, 0); } -static void emit_alu1( struct brw_compile *p, - struct brw_instruction *(*func)(struct brw_compile *, - struct brw_reg, - struct brw_reg), - const struct brw_reg *dst, - GLuint mask, - const struct brw_reg *arg0 ) +void emit_alu1(struct brw_compile *p, + struct brw_instruction *(*func)(struct brw_compile *, + struct brw_reg, + struct brw_reg), + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0) { GLuint i; @@ -376,15 +395,15 @@ static void emit_alu1( struct brw_compile *p, } -static void emit_alu2( struct brw_compile *p, - struct brw_instruction *(*func)(struct brw_compile *, - struct brw_reg, - struct brw_reg, - struct brw_reg), - const struct brw_reg *dst, - GLuint mask, - const struct brw_reg *arg0, - const struct brw_reg *arg1 ) +void emit_alu2(struct brw_compile *p, + struct brw_instruction *(*func)(struct brw_compile *, + struct brw_reg, + struct brw_reg, + struct brw_reg), + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1) { GLuint i; @@ -402,12 +421,12 @@ static void emit_alu2( struct brw_compile *p, } -static void emit_mad( struct brw_compile *p, - const struct brw_reg *dst, - GLuint mask, - const struct brw_reg *arg0, - const struct brw_reg *arg1, - const struct brw_reg *arg2 ) +void emit_mad(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1, + const struct brw_reg *arg2) { GLuint i; @@ -422,26 +441,12 @@ static void emit_mad( struct brw_compile *p, } } -static void emit_trunc( struct brw_compile *p, - const struct brw_reg *dst, - GLuint mask, - const struct brw_reg *arg0) -{ - GLuint i; - - for (i = 0; i < 4; i++) { - if (mask & (1<<i)) { - brw_RNDZ(p, dst[i], arg0[i]); - } - } -} - -static void emit_lrp( struct brw_compile *p, - const struct brw_reg *dst, - GLuint mask, - const struct brw_reg *arg0, - const struct brw_reg *arg1, - const struct brw_reg *arg2 ) +void emit_lrp(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1, + const struct brw_reg *arg2) { GLuint i; @@ -461,21 +466,24 @@ static void emit_lrp( struct brw_compile *p, } } -static void emit_sop( struct brw_compile *p, - const struct brw_reg *dst, - GLuint mask, - GLuint cond, - const struct brw_reg *arg0, - const struct brw_reg *arg1 ) +void emit_sop(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + GLuint cond, + const struct brw_reg *arg0, + const struct brw_reg *arg1) { GLuint i; for (i = 0; i < 4; i++) { if (mask & (1<<i)) { - brw_MOV(p, dst[i], brw_imm_f(0)); + brw_push_insn_state(p); brw_CMP(p, brw_null_reg(), cond, arg0[i], arg1[i]); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_MOV(p, dst[i], brw_imm_f(0)); + brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); brw_MOV(p, dst[i], brw_imm_f(1.0)); - brw_set_predicate_control_flag_value(p, 0xff); + brw_pop_insn_state(p); } } } @@ -559,11 +567,11 @@ static void emit_cmp( struct brw_compile *p, } } -static void emit_max( struct brw_compile *p, - const struct brw_reg *dst, - GLuint mask, - const struct brw_reg *arg0, - const struct brw_reg *arg1 ) +void emit_max(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1) { GLuint i; @@ -583,11 +591,11 @@ static void emit_max( struct brw_compile *p, } } -static void emit_min( struct brw_compile *p, - const struct brw_reg *dst, - GLuint mask, - const struct brw_reg *arg0, - const struct brw_reg *arg1 ) +void emit_min(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1) { GLuint i; @@ -608,11 +616,11 @@ static void emit_min( struct brw_compile *p, } -static void emit_dp3( struct brw_compile *p, - const struct brw_reg *dst, - GLuint mask, - const struct brw_reg *arg0, - const struct brw_reg *arg1 ) +void emit_dp3(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1) { int dst_chan = _mesa_ffs(mask & WRITEMASK_XYZW) - 1; @@ -630,11 +638,11 @@ static void emit_dp3( struct brw_compile *p, } -static void emit_dp4( struct brw_compile *p, - const struct brw_reg *dst, - GLuint mask, - const struct brw_reg *arg0, - const struct brw_reg *arg1 ) +void emit_dp4(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1) { int dst_chan = _mesa_ffs(mask & WRITEMASK_XYZW) - 1; @@ -653,11 +661,11 @@ static void emit_dp4( struct brw_compile *p, } -static void emit_dph( struct brw_compile *p, - const struct brw_reg *dst, - GLuint mask, - const struct brw_reg *arg0, - const struct brw_reg *arg1 ) +void emit_dph(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1) { const int dst_chan = _mesa_ffs(mask & WRITEMASK_XYZW) - 1; @@ -676,11 +684,11 @@ static void emit_dph( struct brw_compile *p, } -static void emit_xpd( struct brw_compile *p, - const struct brw_reg *dst, - GLuint mask, - const struct brw_reg *arg0, - const struct brw_reg *arg1 ) +void emit_xpd(struct brw_compile *p, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1) { GLuint i; @@ -701,41 +709,68 @@ static void emit_xpd( struct brw_compile *p, } -static void emit_math1( struct brw_compile *p, - GLuint function, - const struct brw_reg *dst, - GLuint mask, - const struct brw_reg *arg0 ) +void emit_math1(struct brw_wm_compile *c, + GLuint function, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0) { + struct brw_compile *p = &c->func; int dst_chan = _mesa_ffs(mask & WRITEMASK_XYZW) - 1; + GLuint saturate = ((mask & SATURATE) ? + BRW_MATH_SATURATE_SATURATE : + BRW_MATH_SATURATE_NONE); if (!(mask & WRITEMASK_XYZW)) return; /* Do not emit dead code */ assert(is_power_of_two(mask & WRITEMASK_XYZW)); + /* If compressed, this will write message reg 2,3 from arg0.x's 16 + * channels. + */ brw_MOV(p, brw_message_reg(2), arg0[0]); /* Send two messages to perform all 16 operations: */ - brw_math_16(p, - dst[dst_chan], + brw_push_insn_state(p); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_math(p, + dst[dst_chan], + function, + saturate, + 2, + brw_null_reg(), + BRW_MATH_DATA_VECTOR, + BRW_MATH_PRECISION_FULL); + + if (c->dispatch_width == 16) { + brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF); + brw_math(p, + offset(dst[dst_chan],1), function, - (mask & SATURATE) ? BRW_MATH_SATURATE_SATURATE : BRW_MATH_SATURATE_NONE, - 2, + saturate, + 3, brw_null_reg(), + BRW_MATH_DATA_VECTOR, BRW_MATH_PRECISION_FULL); + } + brw_pop_insn_state(p); } -static void emit_math2( struct brw_compile *p, - GLuint function, - const struct brw_reg *dst, - GLuint mask, - const struct brw_reg *arg0, - const struct brw_reg *arg1) +void emit_math2(struct brw_wm_compile *c, + GLuint function, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1) { + struct brw_compile *p = &c->func; int dst_chan = _mesa_ffs(mask & WRITEMASK_XYZW) - 1; + GLuint saturate = ((mask & SATURATE) ? + BRW_MATH_SATURATE_SATURATE : + BRW_MATH_SATURATE_NONE); if (!(mask & WRITEMASK_XYZW)) return; /* Do not emit dead code */ @@ -746,173 +781,231 @@ static void emit_math2( struct brw_compile *p, brw_set_compression_control(p, BRW_COMPRESSION_NONE); brw_MOV(p, brw_message_reg(2), arg0[0]); - brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF); - brw_MOV(p, brw_message_reg(4), sechalf(arg0[0])); + if (c->dispatch_width == 16) { + brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF); + brw_MOV(p, brw_message_reg(4), sechalf(arg0[0])); + } brw_set_compression_control(p, BRW_COMPRESSION_NONE); brw_MOV(p, brw_message_reg(3), arg1[0]); - brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF); - brw_MOV(p, brw_message_reg(5), sechalf(arg1[0])); + if (c->dispatch_width == 16) { + brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF); + brw_MOV(p, brw_message_reg(5), sechalf(arg1[0])); + } - - /* Send two messages to perform all 16 operations: - */ brw_set_compression_control(p, BRW_COMPRESSION_NONE); brw_math(p, dst[dst_chan], function, - (mask & SATURATE) ? BRW_MATH_SATURATE_SATURATE : BRW_MATH_SATURATE_NONE, + saturate, 2, brw_null_reg(), BRW_MATH_DATA_VECTOR, BRW_MATH_PRECISION_FULL); - brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF); - brw_math(p, - offset(dst[dst_chan],1), - function, - (mask & SATURATE) ? BRW_MATH_SATURATE_SATURATE : BRW_MATH_SATURATE_NONE, - 4, - brw_null_reg(), - BRW_MATH_DATA_VECTOR, - BRW_MATH_PRECISION_FULL); - + /* Send two messages to perform all 16 operations: + */ + if (c->dispatch_width == 16) { + brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF); + brw_math(p, + offset(dst[dst_chan],1), + function, + saturate, + 4, + brw_null_reg(), + BRW_MATH_DATA_VECTOR, + BRW_MATH_PRECISION_FULL); + } brw_pop_insn_state(p); } - -static void emit_tex( struct brw_wm_compile *c, - const struct brw_wm_instruction *inst, - struct brw_reg *dst, - GLuint dst_flags, - struct brw_reg *arg ) +void emit_tex(struct brw_wm_compile *c, + struct brw_reg *dst, + GLuint dst_flags, + struct brw_reg *arg, + struct brw_reg depth_payload, + GLuint tex_idx, + GLuint sampler, + GLboolean shadow) { struct brw_compile *p = &c->func; - GLuint msgLength, responseLength; - GLuint i, nr; + struct brw_reg dst_retyped; + GLuint cur_mrf = 2, response_length; + GLuint i, nr_texcoords; GLuint emit; GLuint msg_type; + GLuint mrf_per_channel; + GLuint simd_mode; + + if (c->dispatch_width == 16) { + mrf_per_channel = 2; + response_length = 8; + dst_retyped = retype(vec16(dst[0]), BRW_REGISTER_TYPE_UW); + simd_mode = BRW_SAMPLER_SIMD_MODE_SIMD16; + } else { + mrf_per_channel = 1; + response_length = 4; + dst_retyped = retype(vec8(dst[0]), BRW_REGISTER_TYPE_UW); + simd_mode = BRW_SAMPLER_SIMD_MODE_SIMD8; + } /* How many input regs are there? */ - switch (inst->tex_idx) { + switch (tex_idx) { case TEXTURE_1D_INDEX: emit = WRITEMASK_X; - nr = 1; + nr_texcoords = 1; break; case TEXTURE_2D_INDEX: case TEXTURE_RECT_INDEX: emit = WRITEMASK_XY; - nr = 2; + nr_texcoords = 2; break; case TEXTURE_3D_INDEX: case TEXTURE_CUBE_INDEX: emit = WRITEMASK_XYZ; - nr = 3; + nr_texcoords = 3; break; default: /* unexpected target */ abort(); } - if (inst->tex_shadow) { - nr = 4; - emit |= WRITEMASK_W; - } + /* Pre-Ironlake, the 8-wide sampler always took u,v,r. */ + if (!BRW_IS_IGDNG(p->brw) && c->dispatch_width == 8) + nr_texcoords = 3; - msgLength = 1; + /* For shadow comparisons, we have to supply u,v,r. */ + if (shadow) + nr_texcoords = 3; - for (i = 0; i < nr; i++) { - static const GLuint swz[4] = {0,1,2,2}; - if (emit & (1<<i)) - brw_MOV(p, brw_message_reg(msgLength+1), arg[swz[i]]); + /* Emit the texcoords. */ + for (i = 0; i < nr_texcoords; i++) { + if (emit & (1<<i)) + brw_MOV(p, brw_message_reg(cur_mrf), arg[i]); else - brw_MOV(p, brw_message_reg(msgLength+1), brw_imm_f(0)); - msgLength += 2; + brw_MOV(p, brw_message_reg(cur_mrf), brw_imm_f(0)); + cur_mrf += mrf_per_channel; } - responseLength = 8; /* always */ + /* Fill in the shadow comparison reference value. */ + if (shadow) { + if (BRW_IS_IGDNG(p->brw)) { + /* Fill in the cube map array index value. */ + brw_MOV(p, brw_message_reg(cur_mrf), brw_imm_f(0)); + cur_mrf += mrf_per_channel; + } else if (c->dispatch_width == 8) { + /* Fill in the LOD bias value. */ + brw_MOV(p, brw_message_reg(cur_mrf), brw_imm_f(0)); + cur_mrf += mrf_per_channel; + } + brw_MOV(p, brw_message_reg(cur_mrf), arg[2]); + cur_mrf += mrf_per_channel; + } if (BRW_IS_IGDNG(p->brw)) { - if (inst->tex_shadow) - msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE_IGDNG; - else - msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_IGDNG; + if (shadow) + msg_type = BRW_SAMPLER_MESSAGE_SAMPLE_COMPARE_IGDNG; + else + msg_type = BRW_SAMPLER_MESSAGE_SAMPLE_IGDNG; } else { - if (inst->tex_shadow) - msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE; - else - msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE; + /* Note that G45 and older determines shadow compare and dispatch width + * from message length for most messages. + */ + if (c->dispatch_width == 16 && shadow) + msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE; + else + msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE; } - brw_SAMPLE(p, - retype(vec16(dst[0]), BRW_REGISTER_TYPE_UW), + brw_SAMPLE(p, + dst_retyped, 1, - retype(c->payload.depth[0].hw_reg, BRW_REGISTER_TYPE_UW), - SURF_INDEX_TEXTURE(inst->tex_unit), - inst->tex_unit, /* sampler */ - inst->writemask, - msg_type, - responseLength, - msgLength, - 0, + retype(depth_payload, BRW_REGISTER_TYPE_UW), + SURF_INDEX_TEXTURE(sampler), + sampler, + dst_flags & WRITEMASK_XYZW, + msg_type, + response_length, + cur_mrf - 1, + 0, 1, - BRW_SAMPLER_SIMD_MODE_SIMD16); + simd_mode); } -static void emit_txb( struct brw_wm_compile *c, - const struct brw_wm_instruction *inst, - struct brw_reg *dst, - GLuint dst_flags, - struct brw_reg *arg ) +void emit_txb(struct brw_wm_compile *c, + struct brw_reg *dst, + GLuint dst_flags, + struct brw_reg *arg, + struct brw_reg depth_payload, + GLuint tex_idx, + GLuint sampler) { struct brw_compile *p = &c->func; GLuint msgLength; GLuint msg_type; - /* Shadow ignored for txb. + GLuint mrf_per_channel; + GLuint response_length; + struct brw_reg dst_retyped; + + /* The G45 and older chipsets don't support 8-wide dispatch for LOD biased + * samples, so we'll use the 16-wide instruction, leave the second halves + * undefined, and trust the execution mask to keep the undefined pixels + * from mattering. */ - switch (inst->tex_idx) { + if (c->dispatch_width == 16 || !BRW_IS_IGDNG(p->brw)) { + if (BRW_IS_IGDNG(p->brw)) + msg_type = BRW_SAMPLER_MESSAGE_SAMPLE_BIAS_IGDNG; + else + msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS; + mrf_per_channel = 2; + dst_retyped = retype(vec16(dst[0]), BRW_REGISTER_TYPE_UW); + response_length = 8; + } else { + msg_type = BRW_SAMPLER_MESSAGE_SAMPLE_BIAS_IGDNG; + mrf_per_channel = 1; + dst_retyped = retype(vec8(dst[0]), BRW_REGISTER_TYPE_UW); + response_length = 4; + } + + /* Shadow ignored for txb. */ + switch (tex_idx) { case TEXTURE_1D_INDEX: - brw_MOV(p, brw_message_reg(2), arg[0]); - brw_MOV(p, brw_message_reg(4), brw_imm_f(0)); - brw_MOV(p, brw_message_reg(6), brw_imm_f(0)); + brw_MOV(p, brw_message_reg(2 + 0 * mrf_per_channel), arg[0]); + brw_MOV(p, brw_message_reg(2 + 1 * mrf_per_channel), brw_imm_f(0)); + brw_MOV(p, brw_message_reg(2 + 2 * mrf_per_channel), brw_imm_f(0)); break; case TEXTURE_2D_INDEX: case TEXTURE_RECT_INDEX: - brw_MOV(p, brw_message_reg(2), arg[0]); - brw_MOV(p, brw_message_reg(4), arg[1]); - brw_MOV(p, brw_message_reg(6), brw_imm_f(0)); + brw_MOV(p, brw_message_reg(2 + 0 * mrf_per_channel), arg[0]); + brw_MOV(p, brw_message_reg(2 + 1 * mrf_per_channel), arg[1]); + brw_MOV(p, brw_message_reg(2 + 2 * mrf_per_channel), brw_imm_f(0)); break; case TEXTURE_3D_INDEX: case TEXTURE_CUBE_INDEX: - brw_MOV(p, brw_message_reg(2), arg[0]); - brw_MOV(p, brw_message_reg(4), arg[1]); - brw_MOV(p, brw_message_reg(6), arg[2]); + brw_MOV(p, brw_message_reg(2 + 0 * mrf_per_channel), arg[0]); + brw_MOV(p, brw_message_reg(2 + 1 * mrf_per_channel), arg[1]); + brw_MOV(p, brw_message_reg(2 + 2 * mrf_per_channel), arg[2]); break; default: /* unexpected target */ abort(); } - brw_MOV(p, brw_message_reg(8), arg[3]); - msgLength = 9; - - if (BRW_IS_IGDNG(p->brw)) - msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS_IGDNG; - else - msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS; + brw_MOV(p, brw_message_reg(2 + 3 * mrf_per_channel), arg[3]); + msgLength = 2 + 4 * mrf_per_channel - 1; brw_SAMPLE(p, - retype(vec16(dst[0]), BRW_REGISTER_TYPE_UW), + dst_retyped, 1, - retype(c->payload.depth[0].hw_reg, BRW_REGISTER_TYPE_UW), - SURF_INDEX_TEXTURE(inst->tex_unit), - inst->tex_unit, /* sampler */ - inst->writemask, + retype(depth_payload, BRW_REGISTER_TYPE_UW), + SURF_INDEX_TEXTURE(sampler), + sampler, + dst_flags & WRITEMASK_XYZW, msg_type, - 8, /* responseLength */ + response_length, msgLength, 0, 1, @@ -920,11 +1013,13 @@ static void emit_txb( struct brw_wm_compile *c, } -static void emit_lit( struct brw_compile *p, - const struct brw_reg *dst, - GLuint mask, - const struct brw_reg *arg0 ) +static void emit_lit(struct brw_wm_compile *c, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0) { + struct brw_compile *p = &c->func; + assert((mask & WRITEMASK_XW) == 0); if (mask & WRITEMASK_Y) { @@ -934,7 +1029,7 @@ static void emit_lit( struct brw_compile *p, } if (mask & WRITEMASK_Z) { - emit_math2(p, BRW_MATH_FUNCTION_POW, + emit_math2(c, BRW_MATH_FUNCTION_POW, &dst[2], WRITEMASK_X | (mask & SATURATE), &arg0[1], @@ -1001,7 +1096,13 @@ static void fire_fb_write( struct brw_wm_compile *c, GLuint eot ) { struct brw_compile *p = &c->func; - + struct brw_reg dst; + + if (c->dispatch_width == 16) + dst = retype(vec16(brw_null_reg()), BRW_REGISTER_TYPE_UW); + else + dst = retype(vec8(brw_null_reg()), BRW_REGISTER_TYPE_UW); + /* Pass through control information: */ /* mov (8) m1.0<1>:ud r1.0<8;8,1>:ud { Align1 NoMask } */ @@ -1018,7 +1119,7 @@ static void fire_fb_write( struct brw_wm_compile *c, /* Send framebuffer write message: */ /* send (16) null.0<1>:uw m0 r0.0<8;8,1>:uw 0x85a04000:ud { Align1 EOT } */ brw_fb_WRITE(p, - retype(vec16(brw_null_reg()), BRW_REGISTER_TYPE_UW), + dst, base_reg, retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UW), target, @@ -1050,14 +1151,15 @@ static void emit_aa( struct brw_wm_compile *c, * \param arg1 the pass-through depth value * \param arg2 the shader-computed depth value */ -static void emit_fb_write( struct brw_wm_compile *c, - struct brw_reg *arg0, - struct brw_reg *arg1, - struct brw_reg *arg2, - GLuint target, - GLuint eot) +void emit_fb_write(struct brw_wm_compile *c, + struct brw_reg *arg0, + struct brw_reg *arg1, + struct brw_reg *arg2, + GLuint target, + GLuint eot) { struct brw_compile *p = &c->func; + struct brw_context *brw = p->brw; GLuint nr = 2; GLuint channel; @@ -1069,30 +1171,37 @@ static void emit_fb_write( struct brw_wm_compile *c, /* I don't really understand how this achieves the color interleave * (ie RGBARGBA) in the result: [Do the saturation here] */ - { - brw_push_insn_state(p); - - for (channel = 0; channel < 4; channel++) { + brw_push_insn_state(p); + + for (channel = 0; channel < 4; channel++) { + if (c->dispatch_width == 16 && (BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw))) { + /* By setting the high bit of the MRF register number, we indicate + * that we want COMPR4 mode - instead of doing the usual destination + * + 1 for the second half we get destination + 4. + */ + brw_MOV(p, + brw_message_reg(nr + channel + (1 << 7)), + arg0[channel]); + } else { /* mov (8) m2.0<1>:ud r28.0<8;8,1>:ud { Align1 } */ /* mov (8) m6.0<1>:ud r29.0<8;8,1>:ud { Align1 SecHalf } */ - brw_set_compression_control(p, BRW_COMPRESSION_NONE); brw_MOV(p, brw_message_reg(nr + channel), arg0[channel]); - - brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF); - brw_MOV(p, - brw_message_reg(nr + channel + 4), - sechalf(arg0[channel])); - } - /* skip over the regs populated above: - */ - nr += 8; - - brw_pop_insn_state(p); + if (c->dispatch_width == 16) { + brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF); + brw_MOV(p, + brw_message_reg(nr + channel + 4), + sechalf(arg0[channel])); + } + } } + /* skip over the regs populated above: + */ + nr += 8; + brw_pop_insn_state(p); if (c->key.source_depth_to_render_target) { @@ -1142,7 +1251,7 @@ static void emit_fb_write( struct brw_wm_compile *c, get_element_ud(brw_vec8_grf(1,0), 6), brw_imm_ud(1<<26)); - jmp = brw_JMPI(p, ip, ip, brw_imm_d(0)); + jmp = brw_JMPI(p, ip, ip, brw_imm_w(0)); { emit_aa(c, arg1, 2); fire_fb_write(c, 0, nr, target, eot); @@ -1156,7 +1265,6 @@ static void emit_fb_write( struct brw_wm_compile *c, } } - /** * Move a GPR to scratch memory. */ @@ -1294,7 +1402,7 @@ void brw_wm_emit( struct brw_wm_compile *c ) /* Generated instructions for calculating triangle interpolants: */ case WM_PIXELXY: - emit_pixel_xy(p, dst, dst_flags); + emit_pixel_xy(c, dst, dst_flags); break; case WM_DELTAXY: @@ -1306,7 +1414,7 @@ void brw_wm_emit( struct brw_wm_compile *c ) break; case WM_PIXELW: - emit_pixel_w(p, dst, dst_flags, args[0], args[1]); + emit_pixel_w(c, dst, dst_flags, args[0], args[1]); break; case WM_LINTERP: @@ -1364,7 +1472,7 @@ void brw_wm_emit( struct brw_wm_compile *c ) break; case OPCODE_TRUNC: - emit_trunc(p, dst, dst_flags, args[0]); + emit_alu1(p, brw_RNDZ, dst, dst_flags, args[0]); break; case OPCODE_LRP: @@ -1391,27 +1499,27 @@ void brw_wm_emit( struct brw_wm_compile *c ) /* Higher math functions: */ case OPCODE_RCP: - emit_math1(p, BRW_MATH_FUNCTION_INV, dst, dst_flags, args[0]); + emit_math1(c, BRW_MATH_FUNCTION_INV, dst, dst_flags, args[0]); break; case OPCODE_RSQ: - emit_math1(p, BRW_MATH_FUNCTION_RSQ, dst, dst_flags, args[0]); + emit_math1(c, BRW_MATH_FUNCTION_RSQ, dst, dst_flags, args[0]); break; case OPCODE_SIN: - emit_math1(p, BRW_MATH_FUNCTION_SIN, dst, dst_flags, args[0]); + emit_math1(c, BRW_MATH_FUNCTION_SIN, dst, dst_flags, args[0]); break; case OPCODE_COS: - emit_math1(p, BRW_MATH_FUNCTION_COS, dst, dst_flags, args[0]); + emit_math1(c, BRW_MATH_FUNCTION_COS, dst, dst_flags, args[0]); break; case OPCODE_EX2: - emit_math1(p, BRW_MATH_FUNCTION_EXP, dst, dst_flags, args[0]); + emit_math1(c, BRW_MATH_FUNCTION_EXP, dst, dst_flags, args[0]); break; case OPCODE_LG2: - emit_math1(p, BRW_MATH_FUNCTION_LOG, dst, dst_flags, args[0]); + emit_math1(c, BRW_MATH_FUNCTION_LOG, dst, dst_flags, args[0]); break; case OPCODE_SCS: @@ -1419,13 +1527,13 @@ void brw_wm_emit( struct brw_wm_compile *c ) * fixup for 16-element execution. */ if (dst_flags & WRITEMASK_X) - emit_math1(p, BRW_MATH_FUNCTION_COS, dst, (dst_flags&SATURATE)|WRITEMASK_X, args[0]); + emit_math1(c, BRW_MATH_FUNCTION_COS, dst, (dst_flags&SATURATE)|WRITEMASK_X, args[0]); if (dst_flags & WRITEMASK_Y) - emit_math1(p, BRW_MATH_FUNCTION_SIN, dst+1, (dst_flags&SATURATE)|WRITEMASK_X, args[0]); + emit_math1(c, BRW_MATH_FUNCTION_SIN, dst+1, (dst_flags&SATURATE)|WRITEMASK_X, args[0]); break; case OPCODE_POW: - emit_math2(p, BRW_MATH_FUNCTION_POW, dst, dst_flags, args[0], args[1]); + emit_math2(c, BRW_MATH_FUNCTION_POW, dst, dst_flags, args[0], args[1]); break; /* Comparisons: @@ -1463,17 +1571,20 @@ void brw_wm_emit( struct brw_wm_compile *c ) break; case OPCODE_LIT: - emit_lit(p, dst, dst_flags, args[0]); + emit_lit(c, dst, dst_flags, args[0]); break; /* Texturing operations: */ case OPCODE_TEX: - emit_tex(c, inst, dst, dst_flags, args[0]); + emit_tex(c, dst, dst_flags, args[0], c->payload.depth[0].hw_reg, + inst->tex_idx, inst->tex_unit, + inst->tex_shadow); break; case OPCODE_TXB: - emit_txb(c, inst, dst, dst_flags, args[0]); + emit_txb(c, dst, dst_flags, args[0], c->payload.depth[0].hw_reg, + inst->tex_idx, inst->tex_unit); break; case OPCODE_KIL: diff --git a/src/mesa/drivers/dri/i965/brw_wm_fp.c b/src/mesa/drivers/dri/i965/brw_wm_fp.c index 4e3edfbbffa..3737faf26fb 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_fp.c +++ b/src/mesa/drivers/dri/i965/brw_wm_fp.c @@ -138,7 +138,6 @@ static struct prog_dst_register dst_reg(GLuint file, GLuint idx) reg.CondMask = COND_TR; reg.CondSwizzle = 0; reg.CondSrc = 0; - reg.pad = 0; return reg; } @@ -181,6 +180,9 @@ static void release_temp( struct brw_wm_compile *c, struct prog_dst_register tem static struct prog_instruction *get_fp_inst(struct brw_wm_compile *c) { + assert(c->nr_fp_insns < BRW_WM_MAX_INSN); + memset(&c->prog_instructions[c->nr_fp_insns], 0, + sizeof(*c->prog_instructions)); return &c->prog_instructions[c->nr_fp_insns++]; } @@ -447,7 +449,6 @@ static void emit_interp( struct brw_wm_compile *c, break; case FRAG_ATTRIB_FACE: - /* XXX review/test this case */ emit_op(c, WM_FRONTFACING, dst_mask(dst, WRITEMASK_X), @@ -956,7 +957,7 @@ static void precalc_txp( struct brw_wm_compile *c, -static void emit_fb_write( struct brw_wm_compile *c ) +static void emit_render_target_writes( struct brw_wm_compile *c ) { struct prog_src_register payload_r0_depth = src_reg(PROGRAM_PAYLOAD, PAYLOAD_DEPTH); struct prog_src_register outdepth = src_reg(PROGRAM_OUTPUT, FRAG_RESULT_DEPTH); @@ -964,36 +965,34 @@ static void emit_fb_write( struct brw_wm_compile *c ) GLuint i; struct prog_instruction *inst, *last_inst; - struct brw_context *brw = c->func.brw; /* The inst->Aux field is used for FB write target and the EOT marker */ - if (brw->state.nr_color_regions > 1) { - for (i = 0 ; i < brw->state.nr_color_regions; i++) { + if (c->key.nr_color_regions > 1) { + for (i = 0 ; i < c->key.nr_color_regions; i++) { outcolor = src_reg(PROGRAM_OUTPUT, FRAG_RESULT_DATA0 + i); - last_inst = inst = emit_op(c, - WM_FB_WRITE, dst_mask(dst_undef(),0), 0, - outcolor, payload_r0_depth, outdepth); - inst->Aux = (i<<1); + last_inst = inst = emit_op(c, WM_FB_WRITE, dst_mask(dst_undef(), 0), + 0, outcolor, payload_r0_depth, outdepth); + inst->Aux = INST_AUX_TARGET(i); if (c->fp_fragcolor_emitted) { outcolor = src_reg(PROGRAM_OUTPUT, FRAG_RESULT_COLOR); - last_inst = inst = emit_op(c, WM_FB_WRITE, dst_mask(dst_undef(),0), + last_inst = inst = emit_op(c, WM_FB_WRITE, dst_mask(dst_undef(), 0), 0, outcolor, payload_r0_depth, outdepth); - inst->Aux = (i<<1); + inst->Aux = INST_AUX_TARGET(i); } } - last_inst->Aux |= 1; //eot + last_inst->Aux |= INST_AUX_EOT; } else { /* if gl_FragData[0] is written, use it, else use gl_FragColor */ - if (c->fp->program.Base.OutputsWritten & (1 << FRAG_RESULT_DATA0)) + if (c->fp->program.Base.OutputsWritten & BITFIELD64_BIT(FRAG_RESULT_DATA0)) outcolor = src_reg(PROGRAM_OUTPUT, FRAG_RESULT_DATA0); else outcolor = src_reg(PROGRAM_OUTPUT, FRAG_RESULT_COLOR); inst = emit_op(c, WM_FB_WRITE, dst_mask(dst_undef(),0), 0, outcolor, payload_r0_depth, outdepth); - inst->Aux = 1|(0<<1); + inst->Aux = INST_AUX_EOT | INST_AUX_TARGET(0); } } @@ -1155,7 +1154,7 @@ void brw_wm_pass_fp( struct brw_wm_compile *c ) out->DstReg.WriteMask = 0; break; case OPCODE_END: - emit_fb_write(c); + emit_render_target_writes(c); break; case OPCODE_PRINT: break; diff --git a/src/mesa/drivers/dri/i965/brw_wm_glsl.c b/src/mesa/drivers/dri/i965/brw_wm_glsl.c index c9fe1dd8ad2..e8c2cb66ec3 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_glsl.c +++ b/src/mesa/drivers/dri/i965/brw_wm_glsl.c @@ -371,7 +371,7 @@ static void prealloc_reg(struct brw_wm_compile *c) for (j = 0; j < 4; j++) set_reg(c, PROGRAM_PAYLOAD, fp_input, j, reg); } - if (c->key.vp_outputs_written & (1 << i)) { + if (c->key.vp_outputs_written & BITFIELD64_BIT(i)) { reg_index += 2; } } @@ -550,42 +550,6 @@ static struct brw_reg get_src_reg(struct brw_wm_compile *c, } } - -/** - * Same as \sa get_src_reg() but if the register is a literal, emit - * a brw_reg encoding the literal. - * Note that a brw instruction only allows one src operand to be a literal. - * For instructions with more than one operand, only the second can be a - * literal. This means that we treat some literals as constants/uniforms - * (which why PROGRAM_CONSTANT is checked in fetch_constants()). - * - */ -static struct brw_reg get_src_reg_imm(struct brw_wm_compile *c, - const struct prog_instruction *inst, - GLuint srcRegIndex, GLuint channel) -{ - const struct prog_src_register *src = &inst->SrcReg[srcRegIndex]; - if (src->File == PROGRAM_CONSTANT) { - /* a literal */ - const int component = GET_SWZ(src->Swizzle, channel); - const GLfloat *param = - c->fp->program.Base.Parameters->ParameterValues[src->Index]; - GLfloat value = param[component]; - if (src->Negate & (1 << channel)) - value = -value; - if (src->Abs) - value = FABSF(value); -#if 0 - printf(" form immed value %f for chan %d\n", value, channel); -#endif - return brw_imm_f(value); - } - else { - return get_src_reg(c, inst, srcRegIndex, channel); - } -} - - /** * Subroutines are minimal support for resusable instruction sequences. * They are implemented as simply as possible to minimise overhead: there @@ -650,542 +614,110 @@ static void invoke_subroutine( struct brw_wm_compile *c, } } -static void emit_trunc( struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - int i; - struct brw_compile *p = &c->func; - GLuint mask = inst->DstReg.WriteMask; - brw_set_saturate(p, inst->SaturateMode != SATURATE_OFF); - for (i = 0; i < 4; i++) { - if (mask & (1<<i)) { - struct brw_reg src, dst; - dst = get_dst_reg(c, inst, i); - src = get_src_reg(c, inst, 0, i); - brw_RNDZ(p, dst, src); - } - } - brw_set_saturate(p, 0); -} - -static void emit_mov( struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - int i; - struct brw_compile *p = &c->func; - GLuint mask = inst->DstReg.WriteMask; - brw_set_saturate(p, inst->SaturateMode != SATURATE_OFF); - for (i = 0; i < 4; i++) { - if (mask & (1<<i)) { - struct brw_reg src, dst; - dst = get_dst_reg(c, inst, i); - /* XXX some moves from immediate value don't work reliably!!! */ - /*src = get_src_reg_imm(c, inst, 0, i);*/ - src = get_src_reg(c, inst, 0, i); - brw_MOV(p, dst, src); - } - } - brw_set_saturate(p, 0); -} - -static void emit_pixel_xy(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - struct brw_reg r1 = brw_vec1_grf(1, 0); - struct brw_reg r1_uw = retype(r1, BRW_REGISTER_TYPE_UW); - - struct brw_reg dst0, dst1; - struct brw_compile *p = &c->func; - GLuint mask = inst->DstReg.WriteMask; - - dst0 = get_dst_reg(c, inst, 0); - dst1 = get_dst_reg(c, inst, 1); - /* Calculate pixel centers by adding 1 or 0 to each of the - * micro-tile coordinates passed in r1. - */ - if (mask & WRITEMASK_X) { - brw_ADD(p, - vec8(retype(dst0, BRW_REGISTER_TYPE_UW)), - stride(suboffset(r1_uw, 4), 2, 4, 0), - brw_imm_v(0x10101010)); - } - - if (mask & WRITEMASK_Y) { - brw_ADD(p, - vec8(retype(dst1, BRW_REGISTER_TYPE_UW)), - stride(suboffset(r1_uw, 5), 2, 4, 0), - brw_imm_v(0x11001100)); - } -} - -static void emit_delta_xy(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - struct brw_reg r1 = brw_vec1_grf(1, 0); - struct brw_reg dst0, dst1, src0, src1; - struct brw_compile *p = &c->func; - GLuint mask = inst->DstReg.WriteMask; - - dst0 = get_dst_reg(c, inst, 0); - dst1 = get_dst_reg(c, inst, 1); - src0 = get_src_reg(c, inst, 0, 0); - src1 = get_src_reg(c, inst, 0, 1); - /* Calc delta X,Y by subtracting origin in r1 from the pixel - * centers. - */ - if (mask & WRITEMASK_X) { - brw_ADD(p, - dst0, - retype(src0, BRW_REGISTER_TYPE_UW), - negate(r1)); - } - - if (mask & WRITEMASK_Y) { - brw_ADD(p, - dst1, - retype(src1, BRW_REGISTER_TYPE_UW), - negate(suboffset(r1,1))); - - } -} - -static void fire_fb_write( struct brw_wm_compile *c, - GLuint base_reg, - GLuint nr, - GLuint target, - GLuint eot) -{ - struct brw_compile *p = &c->func; - /* Pass through control information: - */ - /* mov (8) m1.0<1>:ud r1.0<8;8,1>:ud { Align1 NoMask } */ - { - brw_push_insn_state(p); - brw_set_mask_control(p, BRW_MASK_DISABLE); /* ? */ - brw_MOV(p, - brw_message_reg(base_reg + 1), - brw_vec8_grf(1, 0)); - brw_pop_insn_state(p); - } - /* Send framebuffer write message: */ - brw_fb_WRITE(p, - retype(vec8(brw_null_reg()), BRW_REGISTER_TYPE_UW), - base_reg, - retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UW), - target, - nr, - 0, - eot); -} - -static void emit_fb_write(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - int nr = 2; - int channel; - GLuint target, eot; - struct brw_reg src0; - - /* Reserve a space for AA - may not be needed: - */ - if (c->key.aa_dest_stencil_reg) - nr += 1; - - brw_push_insn_state(p); - for (channel = 0; channel < 4; channel++) { - src0 = get_src_reg(c, inst, 0, channel); - /* mov (8) m2.0<1>:ud r28.0<8;8,1>:ud { Align1 } */ - /* mov (8) m6.0<1>:ud r29.0<8;8,1>:ud { Align1 SecHalf } */ - brw_MOV(p, brw_message_reg(nr + channel), src0); - } - /* skip over the regs populated above: */ - nr += 8; - brw_pop_insn_state(p); - - if (c->key.source_depth_to_render_target) { - if (c->key.computes_depth) { - src0 = get_src_reg(c, inst, 2, 2); - brw_MOV(p, brw_message_reg(nr), src0); - } - else { - src0 = get_src_reg(c, inst, 1, 1); - brw_MOV(p, brw_message_reg(nr), src0); - } - - nr += 2; - } - - if (c->key.dest_depth_reg) { - const GLuint comp = c->key.dest_depth_reg / 2; - const GLuint off = c->key.dest_depth_reg % 2; - - if (off != 0) { - /* XXX this code needs review/testing */ - struct brw_reg arg1_0 = get_src_reg(c, inst, 1, comp); - struct brw_reg arg1_1 = get_src_reg(c, inst, 1, comp+1); - - brw_push_insn_state(p); - brw_set_compression_control(p, BRW_COMPRESSION_NONE); - - brw_MOV(p, brw_message_reg(nr), offset(arg1_0, 1)); - /* 2nd half? */ - brw_MOV(p, brw_message_reg(nr+1), arg1_1); - brw_pop_insn_state(p); - } - else - { - struct brw_reg src = get_src_reg(c, inst, 1, 1); - brw_MOV(p, brw_message_reg(nr), src); - } - nr += 2; - } - - target = inst->Aux >> 1; - eot = inst->Aux & 1; - fire_fb_write(c, 0, nr, target, eot); -} - -static void emit_pixel_w( struct brw_wm_compile *c, - const struct prog_instruction *inst) +/* Workaround for using brw_wm_emit.c's emit functions, which expect + * destination regs to be uniquely written. Moves arguments out to + * temporaries as necessary for instructions which use their destination as + * a temporary. + */ +static void +unalias3(struct brw_wm_compile *c, + void (*func)(struct brw_compile *c, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1, + const struct brw_reg *arg2), + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1, + const struct brw_reg *arg2) { struct brw_compile *p = &c->func; - GLuint mask = inst->DstReg.WriteMask; - if (mask & WRITEMASK_W) { - struct brw_reg dst, src0, delta0, delta1; - struct brw_reg interp3; - - dst = get_dst_reg(c, inst, 3); - src0 = get_src_reg(c, inst, 0, 0); - delta0 = get_src_reg(c, inst, 1, 0); - delta1 = get_src_reg(c, inst, 1, 1); - - interp3 = brw_vec1_grf(src0.nr+1, 4); - /* Calc 1/w - just linterp wpos[3] optimized by putting the - * result straight into a message reg. - */ - brw_LINE(p, brw_null_reg(), interp3, delta0); - brw_MAC(p, brw_message_reg(2), suboffset(interp3, 1), delta1); - - /* Calc w */ - brw_math_16( p, dst, - BRW_MATH_FUNCTION_INV, - BRW_MATH_SATURATE_NONE, - 2, brw_null_reg(), - BRW_MATH_PRECISION_FULL); - } -} + struct brw_reg tmp_arg0[4], tmp_arg1[4], tmp_arg2[4]; + int i, j; + int mark = mark_tmps(c); -static void emit_linterp(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - GLuint mask = inst->DstReg.WriteMask; - struct brw_reg interp[4]; - struct brw_reg dst, delta0, delta1; - struct brw_reg src0; - GLuint nr, i; - - src0 = get_src_reg(c, inst, 0, 0); - delta0 = get_src_reg(c, inst, 1, 0); - delta1 = get_src_reg(c, inst, 1, 1); - nr = src0.nr; - - interp[0] = brw_vec1_grf(nr, 0); - interp[1] = brw_vec1_grf(nr, 4); - interp[2] = brw_vec1_grf(nr+1, 0); - interp[3] = brw_vec1_grf(nr+1, 4); - - for(i = 0; i < 4; i++ ) { - if (mask & (1<<i)) { - dst = get_dst_reg(c, inst, i); - brw_LINE(p, brw_null_reg(), interp[i], delta0); - brw_MAC(p, dst, suboffset(interp[i],1), delta1); - } + for (j = 0; j < 4; j++) { + tmp_arg0[j] = arg0[j]; + tmp_arg1[j] = arg1[j]; + tmp_arg2[j] = arg2[j]; } -} - -static void emit_cinterp(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - GLuint mask = inst->DstReg.WriteMask; - struct brw_reg interp[4]; - struct brw_reg dst, src0; - GLuint nr, i; - - src0 = get_src_reg(c, inst, 0, 0); - nr = src0.nr; - - interp[0] = brw_vec1_grf(nr, 0); - interp[1] = brw_vec1_grf(nr, 4); - interp[2] = brw_vec1_grf(nr+1, 0); - interp[3] = brw_vec1_grf(nr+1, 4); - - for(i = 0; i < 4; i++ ) { + for (i = 0; i < 4; i++) { if (mask & (1<<i)) { - dst = get_dst_reg(c, inst, i); - brw_MOV(p, dst, suboffset(interp[i],3)); + for (j = 0; j < 4; j++) { + if (arg0[j].file == dst[i].file && + dst[i].nr == arg0[j].nr) { + tmp_arg0[j] = alloc_tmp(c); + brw_MOV(p, tmp_arg0[j], arg0[j]); + } + if (arg1[j].file == dst[i].file && + dst[i].nr == arg1[j].nr) { + tmp_arg1[j] = alloc_tmp(c); + brw_MOV(p, tmp_arg1[j], arg1[j]); + } + if (arg2[j].file == dst[i].file && + dst[i].nr == arg2[j].nr) { + tmp_arg2[j] = alloc_tmp(c); + brw_MOV(p, tmp_arg2[j], arg2[j]); + } + } } } -} - -static void emit_pinterp(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - GLuint mask = inst->DstReg.WriteMask; - - struct brw_reg interp[4]; - struct brw_reg dst, delta0, delta1; - struct brw_reg src0, w; - GLuint nr, i; - src0 = get_src_reg(c, inst, 0, 0); - delta0 = get_src_reg(c, inst, 1, 0); - delta1 = get_src_reg(c, inst, 1, 1); - w = get_src_reg(c, inst, 2, 3); - nr = src0.nr; + func(p, dst, mask, tmp_arg0, tmp_arg1, tmp_arg2); - interp[0] = brw_vec1_grf(nr, 0); - interp[1] = brw_vec1_grf(nr, 4); - interp[2] = brw_vec1_grf(nr+1, 0); - interp[3] = brw_vec1_grf(nr+1, 4); - - for(i = 0; i < 4; i++ ) { - if (mask & (1<<i)) { - dst = get_dst_reg(c, inst, i); - brw_LINE(p, brw_null_reg(), interp[i], delta0); - brw_MAC(p, dst, suboffset(interp[i],1), - delta1); - brw_MUL(p, dst, dst, w); - } - } + release_tmps(c, mark); } -/* Sets the destination channels to 1.0 or 0.0 according to glFrontFacing. */ -static void emit_frontfacing(struct brw_wm_compile *c, - const struct prog_instruction *inst) +/* Workaround for using brw_wm_emit.c's emit functions, which expect + * destination regs to be uniquely written. Moves arguments out to + * temporaries as necessary for instructions which use their destination as + * a temporary. + */ +static void +unalias2(struct brw_wm_compile *c, + void (*func)(struct brw_compile *c, + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1), + const struct brw_reg *dst, + GLuint mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1) { struct brw_compile *p = &c->func; - struct brw_reg r1_6ud = retype(brw_vec1_grf(1, 6), BRW_REGISTER_TYPE_UD); - struct brw_reg dst; - GLuint mask = inst->DstReg.WriteMask; - int i; + struct brw_reg tmp_arg0[4], tmp_arg1[4]; + int i, j; + int mark = mark_tmps(c); - for (i = 0; i < 4; i++) { - if (mask & (1<<i)) { - dst = get_dst_reg(c, inst, i); - brw_MOV(p, dst, brw_imm_f(0.0)); - } + for (j = 0; j < 4; j++) { + tmp_arg0[j] = arg0[j]; + tmp_arg1[j] = arg1[j]; } - /* bit 31 is "primitive is back face", so checking < (1 << 31) gives - * us front face - */ - brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, r1_6ud, brw_imm_ud(1 << 31)); for (i = 0; i < 4; i++) { if (mask & (1<<i)) { - dst = get_dst_reg(c, inst, i); - brw_MOV(p, dst, brw_imm_f(1.0)); + for (j = 0; j < 4; j++) { + if (arg0[j].file == dst[i].file && + dst[i].nr == arg0[j].nr) { + tmp_arg0[j] = alloc_tmp(c); + brw_MOV(p, tmp_arg0[j], arg0[j]); + } + if (arg1[j].file == dst[i].file && + dst[i].nr == arg1[j].nr) { + tmp_arg1[j] = alloc_tmp(c); + brw_MOV(p, tmp_arg1[j], arg1[j]); + } + } } } - brw_set_predicate_control_flag_value(p, 0xff); -} -static void emit_xpd(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - int i; - struct brw_compile *p = &c->func; - GLuint mask = inst->DstReg.WriteMask; - for (i = 0; i < 4; i++) { - GLuint i2 = (i+2)%3; - GLuint i1 = (i+1)%3; - if (mask & (1<<i)) { - struct brw_reg src0, src1, dst; - dst = get_dst_reg(c, inst, i); - src0 = negate(get_src_reg(c, inst, 0, i2)); - src1 = get_src_reg_imm(c, inst, 1, i1); - brw_MUL(p, brw_null_reg(), src0, src1); - src0 = get_src_reg(c, inst, 0, i1); - src1 = get_src_reg_imm(c, inst, 1, i2); - brw_set_saturate(p, inst->SaturateMode != SATURATE_OFF); - brw_MAC(p, dst, src0, src1); - brw_set_saturate(p, 0); - } - } - brw_set_saturate(p, 0); -} + func(p, dst, mask, tmp_arg0, tmp_arg1); -static void emit_dp3(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - struct brw_reg src0[3], src1[3], dst; - int i; - struct brw_compile *p = &c->func; - GLuint mask = inst->DstReg.WriteMask; - int dst_chan = _mesa_ffs(mask & WRITEMASK_XYZW) - 1; - - if (!(mask & WRITEMASK_XYZW)) - return; - - assert(is_power_of_two(mask & WRITEMASK_XYZW)); - - for (i = 0; i < 3; i++) { - src0[i] = get_src_reg(c, inst, 0, i); - src1[i] = get_src_reg_imm(c, inst, 1, i); - } - - dst = get_dst_reg(c, inst, dst_chan); - brw_MUL(p, brw_null_reg(), src0[0], src1[0]); - brw_MAC(p, brw_null_reg(), src0[1], src1[1]); - brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); - brw_MAC(p, dst, src0[2], src1[2]); - brw_set_saturate(p, 0); -} - -static void emit_dp4(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - struct brw_reg src0[4], src1[4], dst; - int i; - struct brw_compile *p = &c->func; - GLuint mask = inst->DstReg.WriteMask; - int dst_chan = _mesa_ffs(mask & WRITEMASK_XYZW) - 1; - - if (!(mask & WRITEMASK_XYZW)) - return; - - assert(is_power_of_two(mask & WRITEMASK_XYZW)); - - for (i = 0; i < 4; i++) { - src0[i] = get_src_reg(c, inst, 0, i); - src1[i] = get_src_reg_imm(c, inst, 1, i); - } - dst = get_dst_reg(c, inst, dst_chan); - brw_MUL(p, brw_null_reg(), src0[0], src1[0]); - brw_MAC(p, brw_null_reg(), src0[1], src1[1]); - brw_MAC(p, brw_null_reg(), src0[2], src1[2]); - brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); - brw_MAC(p, dst, src0[3], src1[3]); - brw_set_saturate(p, 0); -} - -static void emit_dph(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - struct brw_reg src0[4], src1[4], dst; - int i; - struct brw_compile *p = &c->func; - GLuint mask = inst->DstReg.WriteMask; - int dst_chan = _mesa_ffs(mask & WRITEMASK_XYZW) - 1; - - if (!(mask & WRITEMASK_XYZW)) - return; - - assert(is_power_of_two(mask & WRITEMASK_XYZW)); - - for (i = 0; i < 4; i++) { - src0[i] = get_src_reg(c, inst, 0, i); - src1[i] = get_src_reg_imm(c, inst, 1, i); - } - dst = get_dst_reg(c, inst, dst_chan); - brw_MUL(p, brw_null_reg(), src0[0], src1[0]); - brw_MAC(p, brw_null_reg(), src0[1], src1[1]); - brw_MAC(p, dst, src0[2], src1[2]); - brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); - brw_ADD(p, dst, dst, src1[3]); - brw_set_saturate(p, 0); -} - -/** - * Emit a scalar instruction, like RCP, RSQ, LOG, EXP. - * Note that the result of the function is smeared across the dest - * register's X, Y, Z and W channels (subject to writemasking of course). - */ -static void emit_math1(struct brw_wm_compile *c, - const struct prog_instruction *inst, GLuint func) -{ - struct brw_compile *p = &c->func; - struct brw_reg src0, dst; - GLuint mask = inst->DstReg.WriteMask; - int dst_chan = _mesa_ffs(mask & WRITEMASK_XYZW) - 1; - - if (!(mask & WRITEMASK_XYZW)) - return; - - assert(is_power_of_two(mask & WRITEMASK_XYZW)); - - /* Get first component of source register */ - dst = get_dst_reg(c, inst, dst_chan); - src0 = get_src_reg(c, inst, 0, 0); - - brw_MOV(p, brw_message_reg(2), src0); - brw_math(p, - dst, - func, - (inst->SaturateMode != SATURATE_OFF) ? BRW_MATH_SATURATE_SATURATE : BRW_MATH_SATURATE_NONE, - 2, - brw_null_reg(), - BRW_MATH_DATA_VECTOR, - BRW_MATH_PRECISION_FULL); -} - -static void emit_rcp(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - emit_math1(c, inst, BRW_MATH_FUNCTION_INV); -} - -static void emit_rsq(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - emit_math1(c, inst, BRW_MATH_FUNCTION_RSQ); -} - -static void emit_sin(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - emit_math1(c, inst, BRW_MATH_FUNCTION_SIN); -} - -static void emit_cos(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - emit_math1(c, inst, BRW_MATH_FUNCTION_COS); -} - -static void emit_ex2(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - emit_math1(c, inst, BRW_MATH_FUNCTION_EXP); -} - -static void emit_lg2(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - emit_math1(c, inst, BRW_MATH_FUNCTION_LOG); -} - -static void emit_add(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - struct brw_reg src0, src1, dst; - GLuint mask = inst->DstReg.WriteMask; - int i; - brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); - for (i = 0 ; i < 4; i++) { - if (mask & (1<<i)) { - dst = get_dst_reg(c, inst, i); - src0 = get_src_reg(c, inst, 0, i); - src1 = get_src_reg_imm(c, inst, 1, i); - brw_ADD(p, dst, src0, src1); - } - } - brw_set_saturate(p, 0); + release_tmps(c, mark); } static void emit_arl(struct brw_wm_compile *c, @@ -1201,180 +733,6 @@ static void emit_arl(struct brw_wm_compile *c, brw_set_saturate(p, 0); } - -static void emit_mul(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - struct brw_reg src0, src1, dst; - GLuint mask = inst->DstReg.WriteMask; - int i; - brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); - for (i = 0 ; i < 4; i++) { - if (mask & (1<<i)) { - dst = get_dst_reg(c, inst, i); - src0 = get_src_reg(c, inst, 0, i); - src1 = get_src_reg_imm(c, inst, 1, i); - brw_MUL(p, dst, src0, src1); - } - } - brw_set_saturate(p, 0); -} - -static void emit_frc(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - struct brw_reg src0, dst; - GLuint mask = inst->DstReg.WriteMask; - int i; - brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); - for (i = 0 ; i < 4; i++) { - if (mask & (1<<i)) { - dst = get_dst_reg(c, inst, i); - src0 = get_src_reg_imm(c, inst, 0, i); - brw_FRC(p, dst, src0); - } - } - if (inst->SaturateMode != SATURATE_OFF) - brw_set_saturate(p, 0); -} - -static void emit_flr(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - struct brw_reg src0, dst; - GLuint mask = inst->DstReg.WriteMask; - int i; - brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); - for (i = 0 ; i < 4; i++) { - if (mask & (1<<i)) { - dst = get_dst_reg(c, inst, i); - src0 = get_src_reg_imm(c, inst, 0, i); - brw_RNDD(p, dst, src0); - } - } - brw_set_saturate(p, 0); -} - - -static void emit_min_max(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - const GLuint mask = inst->DstReg.WriteMask; - const int mark = mark_tmps(c); - int i; - brw_push_insn_state(p); - for (i = 0; i < 4; i++) { - if (mask & (1<<i)) { - struct brw_reg real_dst = get_dst_reg(c, inst, i); - struct brw_reg src0 = get_src_reg(c, inst, 0, i); - struct brw_reg src1 = get_src_reg(c, inst, 1, i); - struct brw_reg dst; - /* if dst==src0 or dst==src1 we need to use a temp reg */ - GLboolean use_temp = brw_same_reg(dst, src0) || - brw_same_reg(dst, src1); - if (use_temp) - dst = alloc_tmp(c); - else - dst = real_dst; - - /* - printf(" Min/max: dst %d src0 %d src1 %d\n", - dst.nr, src0.nr, src1.nr); - */ - brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); - brw_MOV(p, dst, src0); - brw_set_saturate(p, 0); - - if (inst->Opcode == OPCODE_MIN) - brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, src1, src0); - else - brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_G, src1, src0); - - brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); - brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); - brw_MOV(p, dst, src1); - brw_set_saturate(p, 0); - brw_set_predicate_control_flag_value(p, 0xff); - if (use_temp) - brw_MOV(p, real_dst, dst); - } - } - brw_pop_insn_state(p); - release_tmps(c, mark); -} - -static void emit_pow(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - struct brw_reg dst, src0, src1; - GLuint mask = inst->DstReg.WriteMask; - int dst_chan = _mesa_ffs(mask & WRITEMASK_XYZW) - 1; - - if (!(mask & WRITEMASK_XYZW)) - return; - - assert(is_power_of_two(mask & WRITEMASK_XYZW)); - - dst = get_dst_reg(c, inst, dst_chan); - src0 = get_src_reg_imm(c, inst, 0, 0); - src1 = get_src_reg_imm(c, inst, 1, 0); - - brw_MOV(p, brw_message_reg(2), src0); - brw_MOV(p, brw_message_reg(3), src1); - - brw_math(p, - dst, - BRW_MATH_FUNCTION_POW, - (inst->SaturateMode != SATURATE_OFF) ? BRW_MATH_SATURATE_SATURATE : BRW_MATH_SATURATE_NONE, - 2, - brw_null_reg(), - BRW_MATH_DATA_VECTOR, - BRW_MATH_PRECISION_FULL); -} - -static void emit_lrp(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - GLuint mask = inst->DstReg.WriteMask; - struct brw_reg dst, tmp1, tmp2, src0, src1, src2; - int i; - int mark = mark_tmps(c); - for (i = 0; i < 4; i++) { - if (mask & (1<<i)) { - dst = get_dst_reg(c, inst, i); - src0 = get_src_reg(c, inst, 0, i); - - src1 = get_src_reg_imm(c, inst, 1, i); - - if (src1.nr == dst.nr) { - tmp1 = alloc_tmp(c); - brw_MOV(p, tmp1, src1); - } else - tmp1 = src1; - - src2 = get_src_reg(c, inst, 2, i); - if (src2.nr == dst.nr) { - tmp2 = alloc_tmp(c); - brw_MOV(p, tmp2, src2); - } else - tmp2 = src2; - - brw_ADD(p, dst, negate(src0), brw_imm_f(1.0)); - brw_MUL(p, brw_null_reg(), dst, tmp2); - brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); - brw_MAC(p, dst, src0, tmp1); - brw_set_saturate(p, 0); - } - release_tmps(c, mark); - } -} - /** * For GLSL shaders, this KIL will be unconditional. * It may be contained inside an IF/ENDIF structure of course. @@ -1390,89 +748,6 @@ static void emit_kil(struct brw_wm_compile *c) brw_pop_insn_state(p); } -static void emit_mad(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - GLuint mask = inst->DstReg.WriteMask; - struct brw_reg dst, src0, src1, src2; - int i; - - for (i = 0; i < 4; i++) { - if (mask & (1<<i)) { - dst = get_dst_reg(c, inst, i); - src0 = get_src_reg(c, inst, 0, i); - src1 = get_src_reg_imm(c, inst, 1, i); - src2 = get_src_reg_imm(c, inst, 2, i); - brw_MUL(p, dst, src0, src1); - - brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); - brw_ADD(p, dst, dst, src2); - brw_set_saturate(p, 0); - } - } -} - -static void emit_sop(struct brw_wm_compile *c, - const struct prog_instruction *inst, GLuint cond) -{ - struct brw_compile *p = &c->func; - GLuint mask = inst->DstReg.WriteMask; - struct brw_reg dst, src0, src1; - int i; - - for (i = 0; i < 4; i++) { - if (mask & (1<<i)) { - dst = get_dst_reg(c, inst, i); - src0 = get_src_reg(c, inst, 0, i); - src1 = get_src_reg_imm(c, inst, 1, i); - brw_push_insn_state(p); - brw_CMP(p, brw_null_reg(), cond, src0, src1); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - brw_MOV(p, dst, brw_imm_f(0.0)); - brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); - brw_MOV(p, dst, brw_imm_f(1.0)); - brw_pop_insn_state(p); - } - } -} - -static void emit_slt(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - emit_sop(c, inst, BRW_CONDITIONAL_L); -} - -static void emit_sle(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - emit_sop(c, inst, BRW_CONDITIONAL_LE); -} - -static void emit_sgt(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - emit_sop(c, inst, BRW_CONDITIONAL_G); -} - -static void emit_sge(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - emit_sop(c, inst, BRW_CONDITIONAL_GE); -} - -static void emit_seq(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - emit_sop(c, inst, BRW_CONDITIONAL_EQ); -} - -static void emit_sne(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - emit_sop(c, inst, BRW_CONDITIONAL_NEQ); -} - static INLINE struct brw_reg high_words( struct brw_reg reg ) { return stride( suboffset( retype( reg, BRW_REGISTER_TYPE_W ), 1 ), @@ -2525,196 +1800,6 @@ static void emit_noise4( struct brw_wm_compile *c, release_tmps( c, mark ); } - -static void emit_wpos_xy(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - GLuint mask = inst->DstReg.WriteMask; - struct brw_reg src0[2], dst[2]; - - dst[0] = get_dst_reg(c, inst, 0); - dst[1] = get_dst_reg(c, inst, 1); - - src0[0] = get_src_reg(c, inst, 0, 0); - src0[1] = get_src_reg(c, inst, 0, 1); - - /* Calculate the pixel offset from window bottom left into destination - * X and Y channels. - */ - if (mask & WRITEMASK_X) { - /* X' = X - origin_x */ - brw_ADD(p, - dst[0], - retype(src0[0], BRW_REGISTER_TYPE_W), - brw_imm_d(0 - c->key.origin_x)); - } - - if (mask & WRITEMASK_Y) { - /* Y' = height - (Y - origin_y) = height + origin_y - Y */ - brw_ADD(p, - dst[1], - negate(retype(src0[1], BRW_REGISTER_TYPE_W)), - brw_imm_d(c->key.origin_y + c->key.drawable_height - 1)); - } -} - -/* TODO - BIAS on SIMD8 not working yet... - */ -static void emit_txb(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - struct brw_reg dst[4], src[4], payload_reg; - /* Note: TexSrcUnit was already looked up through SamplerTextures[] */ - const GLuint unit = inst->TexSrcUnit; - GLuint i; - GLuint msg_type; - - assert(unit < BRW_MAX_TEX_UNIT); - - payload_reg = get_reg(c, PROGRAM_PAYLOAD, PAYLOAD_DEPTH, 0, 1, 0, 0); - - for (i = 0; i < 4; i++) - dst[i] = get_dst_reg(c, inst, i); - for (i = 0; i < 4; i++) - src[i] = get_src_reg(c, inst, 0, i); - - switch (inst->TexSrcTarget) { - case TEXTURE_1D_INDEX: - brw_MOV(p, brw_message_reg(2), src[0]); /* s coord */ - brw_MOV(p, brw_message_reg(3), brw_imm_f(0)); /* t coord */ - brw_MOV(p, brw_message_reg(4), brw_imm_f(0)); /* r coord */ - break; - case TEXTURE_2D_INDEX: - case TEXTURE_RECT_INDEX: - brw_MOV(p, brw_message_reg(2), src[0]); - brw_MOV(p, brw_message_reg(3), src[1]); - brw_MOV(p, brw_message_reg(4), brw_imm_f(0)); - break; - case TEXTURE_3D_INDEX: - case TEXTURE_CUBE_INDEX: - brw_MOV(p, brw_message_reg(2), src[0]); - brw_MOV(p, brw_message_reg(3), src[1]); - brw_MOV(p, brw_message_reg(4), src[2]); - break; - default: - /* invalid target */ - abort(); - } - brw_MOV(p, brw_message_reg(5), src[3]); /* bias */ - brw_MOV(p, brw_message_reg(6), brw_imm_f(0)); /* ref (unused?) */ - - if (BRW_IS_IGDNG(p->brw)) { - msg_type = BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_BIAS_IGDNG; - } else { - /* Does it work well on SIMD8? */ - msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS; - } - - brw_SAMPLE(p, - retype(vec8(dst[0]), BRW_REGISTER_TYPE_UW), /* dest */ - 1, /* msg_reg_nr */ - retype(payload_reg, BRW_REGISTER_TYPE_UW), /* src0 */ - SURF_INDEX_TEXTURE(unit), - unit, /* sampler */ - inst->DstReg.WriteMask, /* writemask */ - msg_type, /* msg_type */ - 4, /* response_length */ - 4, /* msg_length */ - 0, /* eot */ - 1, - BRW_SAMPLER_SIMD_MODE_SIMD8); -} - - -static void emit_tex(struct brw_wm_compile *c, - const struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - struct brw_reg dst[4], src[4], payload_reg; - /* Note: TexSrcUnit was already looked up through SamplerTextures[] */ - const GLuint unit = inst->TexSrcUnit; - GLuint msg_len; - GLuint i, nr; - GLuint emit; - GLboolean shadow = (c->key.shadowtex_mask & (1<<unit)) ? 1 : 0; - GLuint msg_type; - - assert(unit < BRW_MAX_TEX_UNIT); - - payload_reg = get_reg(c, PROGRAM_PAYLOAD, PAYLOAD_DEPTH, 0, 1, 0, 0); - - for (i = 0; i < 4; i++) - dst[i] = get_dst_reg(c, inst, i); - for (i = 0; i < 4; i++) - src[i] = get_src_reg(c, inst, 0, i); - - switch (inst->TexSrcTarget) { - case TEXTURE_1D_INDEX: - emit = WRITEMASK_X; - nr = 1; - break; - case TEXTURE_2D_INDEX: - case TEXTURE_RECT_INDEX: - emit = WRITEMASK_XY; - nr = 2; - break; - case TEXTURE_3D_INDEX: - case TEXTURE_CUBE_INDEX: - emit = WRITEMASK_XYZ; - nr = 3; - break; - default: - /* invalid target */ - abort(); - } - msg_len = 1; - - /* move/load S, T, R coords */ - for (i = 0; i < nr; i++) { - static const GLuint swz[4] = {0,1,2,2}; - if (emit & (1<<i)) - brw_MOV(p, brw_message_reg(msg_len+1), src[swz[i]]); - else - brw_MOV(p, brw_message_reg(msg_len+1), brw_imm_f(0)); - msg_len += 1; - } - - if (shadow) { - brw_MOV(p, brw_message_reg(5), brw_imm_f(0)); /* lod / bias */ - brw_MOV(p, brw_message_reg(6), src[2]); /* ref value / R coord */ - } - - if (BRW_IS_IGDNG(p->brw)) { - if (shadow) - msg_type = BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_COMPARE_IGDNG; - else - msg_type = BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_IGDNG; - } else { - /* Does it work for shadow on SIMD8 ? */ - msg_type = BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE; - } - - brw_SAMPLE(p, - retype(vec8(dst[0]), BRW_REGISTER_TYPE_UW), /* dest */ - 1, /* msg_reg_nr */ - retype(payload_reg, BRW_REGISTER_TYPE_UW), /* src0 */ - SURF_INDEX_TEXTURE(unit), - unit, /* sampler */ - inst->DstReg.WriteMask, /* writemask */ - msg_type, /* msg_type */ - 4, /* response_length */ - shadow ? 6 : 4, /* msg_length */ - 0, /* eot */ - 1, - BRW_SAMPLER_SIMD_MODE_SIMD8); - - if (shadow) - brw_MOV(p, dst[3], brw_imm_f(1.0)); -} - /** * Resolve subroutine calls after code emit is done. @@ -2771,6 +1856,21 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c) if (c->fp->use_const_buffer) fetch_constants(c, inst); + if (inst->Opcode != OPCODE_ARL) { + for (j = 0; j < 4; j++) { + if (inst->DstReg.WriteMask & (1 << j)) + dst[j] = get_dst_reg(c, inst, j); + else + dst[j] = brw_null_reg(); + } + } + for (j = 0; j < brw_wm_nr_args(inst->Opcode); j++) + get_argument_regs(c, inst, j, args[j], WRITEMASK_XYZW); + + dst_flags = inst->DstReg.WriteMask; + if (inst->SaturateMode == SATURATE_ZERO_ONE) + dst_flags |= SATURATE; + if (inst->CondUpdate) brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); else @@ -2782,126 +1882,131 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c) switch (inst->Opcode) { case WM_PIXELXY: - emit_pixel_xy(c, inst); + emit_pixel_xy(c, dst, dst_flags); break; case WM_DELTAXY: - emit_delta_xy(c, inst); + emit_delta_xy(p, dst, dst_flags, args[0]); break; case WM_PIXELW: - emit_pixel_w(c, inst); + emit_pixel_w(c, dst, dst_flags, args[0], args[1]); break; case WM_LINTERP: - emit_linterp(c, inst); + emit_linterp(p, dst, dst_flags, args[0], args[1]); break; case WM_PINTERP: - emit_pinterp(c, inst); + emit_pinterp(p, dst, dst_flags, args[0], args[1], args[2]); break; case WM_CINTERP: - emit_cinterp(c, inst); + emit_cinterp(p, dst, dst_flags, args[0]); break; case WM_WPOSXY: - emit_wpos_xy(c, inst); + emit_wpos_xy(c, dst, dst_flags, args[0]); break; case WM_FB_WRITE: - emit_fb_write(c, inst); + emit_fb_write(c, args[0], args[1], args[2], + INST_AUX_GET_TARGET(inst->Aux), + inst->Aux & INST_AUX_EOT); break; case WM_FRONTFACING: - emit_frontfacing(c, inst); + emit_frontfacing(p, dst, dst_flags); break; case OPCODE_ADD: - emit_add(c, inst); + emit_alu2(p, brw_ADD, dst, dst_flags, args[0], args[1]); break; case OPCODE_ARL: emit_arl(c, inst); break; case OPCODE_FRC: - emit_frc(c, inst); + emit_alu1(p, brw_FRC, dst, dst_flags, args[0]); break; case OPCODE_FLR: - emit_flr(c, inst); + emit_alu1(p, brw_RNDD, dst, dst_flags, args[0]); break; case OPCODE_LRP: - emit_lrp(c, inst); + unalias3(c, emit_lrp, + dst, dst_flags, args[0], args[1], args[2]); break; case OPCODE_TRUNC: - emit_trunc(c, inst); + emit_alu1(p, brw_RNDZ, dst, dst_flags, args[0]); break; case OPCODE_MOV: case OPCODE_SWZ: - emit_mov(c, inst); + emit_alu1(p, brw_MOV, dst, dst_flags, args[0]); break; case OPCODE_DP3: - emit_dp3(c, inst); + emit_dp3(p, dst, dst_flags, args[0], args[1]); break; case OPCODE_DP4: - emit_dp4(c, inst); + emit_dp4(p, dst, dst_flags, args[0], args[1]); break; case OPCODE_XPD: - emit_xpd(c, inst); + emit_xpd(p, dst, dst_flags, args[0], args[1]); break; case OPCODE_DPH: - emit_dph(c, inst); + emit_dph(p, dst, dst_flags, args[0], args[1]); break; case OPCODE_RCP: - emit_rcp(c, inst); + emit_math1(c, BRW_MATH_FUNCTION_INV, dst, dst_flags, args[0]); break; case OPCODE_RSQ: - emit_rsq(c, inst); + emit_math1(c, BRW_MATH_FUNCTION_RSQ, dst, dst_flags, args[0]); break; case OPCODE_SIN: - emit_sin(c, inst); + emit_math1(c, BRW_MATH_FUNCTION_SIN, dst, dst_flags, args[0]); break; case OPCODE_COS: - emit_cos(c, inst); + emit_math1(c, BRW_MATH_FUNCTION_COS, dst, dst_flags, args[0]); break; case OPCODE_EX2: - emit_ex2(c, inst); + emit_math1(c, BRW_MATH_FUNCTION_EXP, dst, dst_flags, args[0]); break; case OPCODE_LG2: - emit_lg2(c, inst); + emit_math1(c, BRW_MATH_FUNCTION_LOG, dst, dst_flags, args[0]); break; case OPCODE_MIN: + unalias2(c, emit_min, dst, dst_flags, args[0], args[1]); + break; case OPCODE_MAX: - emit_min_max(c, inst); + unalias2(c, emit_max, dst, dst_flags, args[0], args[1]); break; case OPCODE_DDX: case OPCODE_DDY: - for (j = 0; j < 4; j++) { - if (inst->DstReg.WriteMask & (1 << j)) - dst[j] = get_dst_reg(c, inst, j); - else - dst[j] = brw_null_reg(); - } - get_argument_regs(c, inst, 0, args[0], WRITEMASK_XYZW); emit_ddxy(p, dst, dst_flags, (inst->Opcode == OPCODE_DDX), args[0]); break; case OPCODE_SLT: - emit_slt(c, inst); + emit_sop(p, dst, dst_flags, + BRW_CONDITIONAL_L, args[0], args[1]); break; case OPCODE_SLE: - emit_sle(c, inst); + emit_sop(p, dst, dst_flags, + BRW_CONDITIONAL_LE, args[0], args[1]); break; case OPCODE_SGT: - emit_sgt(c, inst); + emit_sop(p, dst, dst_flags, + BRW_CONDITIONAL_G, args[0], args[1]); break; case OPCODE_SGE: - emit_sge(c, inst); + emit_sop(p, dst, dst_flags, + BRW_CONDITIONAL_GE, args[0], args[1]); break; case OPCODE_SEQ: - emit_seq(c, inst); + emit_sop(p, dst, dst_flags, + BRW_CONDITIONAL_EQ, args[0], args[1]); break; case OPCODE_SNE: - emit_sne(c, inst); + emit_sop(p, dst, dst_flags, + BRW_CONDITIONAL_NEQ, args[0], args[1]); break; case OPCODE_MUL: - emit_mul(c, inst); + emit_alu2(p, brw_MUL, dst, dst_flags, args[0], args[1]); break; case OPCODE_POW: - emit_pow(c, inst); + emit_math2(c, BRW_MATH_FUNCTION_POW, + dst, dst_flags, args[0], args[1]); break; case OPCODE_MAD: - emit_mad(c, inst); + emit_mad(p, dst, dst_flags, args[0], args[1], args[2]); break; case OPCODE_NOISE1: emit_noise1(c, inst); @@ -2916,10 +2021,19 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c) emit_noise4(c, inst); break; case OPCODE_TEX: - emit_tex(c, inst); + emit_tex(c, dst, dst_flags, args[0], + get_reg(c, PROGRAM_PAYLOAD, PAYLOAD_DEPTH, + 0, 1, 0, 0), + inst->TexSrcTarget, + inst->TexSrcUnit, + (c->key.shadowtex_mask & (1 << inst->TexSrcUnit)) != 0); break; case OPCODE_TXB: - emit_txb(c, inst); + emit_txb(c, dst, dst_flags, args[0], + get_reg(c, PROGRAM_PAYLOAD, PAYLOAD_DEPTH, + 0, 1, 0, 0), + inst->TexSrcTarget, + c->fp->program.Base.SamplerUnits[inst->TexSrcUnit]); break; case OPCODE_KIL_NV: emit_kil(c); diff --git a/src/mesa/drivers/dri/i965/brw_wm_pass0.c b/src/mesa/drivers/dri/i965/brw_wm_pass0.c index 62792583396..ff4c082d5e8 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_pass0.c +++ b/src/mesa/drivers/dri/i965/brw_wm_pass0.c @@ -42,12 +42,14 @@ static struct brw_wm_ref *get_ref( struct brw_wm_compile *c ) { assert(c->nr_refs < BRW_WM_MAX_REF); + memset(&c->refs[c->nr_refs], 0, sizeof(*c->refs)); return &c->refs[c->nr_refs++]; } static struct brw_wm_value *get_value( struct brw_wm_compile *c) { assert(c->nr_refs < BRW_WM_MAX_VREG); + memset(&c->vreg[c->nr_vreg], 0, sizeof(*c->vreg)); return &c->vreg[c->nr_vreg++]; } @@ -55,6 +57,7 @@ static struct brw_wm_value *get_value( struct brw_wm_compile *c) static struct brw_wm_instruction *get_instruction( struct brw_wm_compile *c ) { assert(c->nr_insns < BRW_WM_MAX_INSN); + memset(&c->instruction[c->nr_insns], 0, sizeof(*c->instruction)); return &c->instruction[c->nr_insns++]; } @@ -322,8 +325,8 @@ translate_insn(struct brw_wm_compile *c, out->tex_unit = inst->TexSrcUnit; out->tex_idx = inst->TexSrcTarget; out->tex_shadow = inst->TexShadow; - out->eot = inst->Aux & 1; - out->target = inst->Aux >> 1; + out->eot = inst->Aux & INST_AUX_EOT; + out->target = INST_AUX_GET_TARGET(inst->Aux); /* Args: */ diff --git a/src/mesa/drivers/dri/i965/brw_wm_pass2.c b/src/mesa/drivers/dri/i965/brw_wm_pass2.c index 6faea018fbc..31303febf09 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_pass2.c +++ b/src/mesa/drivers/dri/i965/brw_wm_pass2.c @@ -82,8 +82,8 @@ static void init_registers( struct brw_wm_compile *c ) for (j = 0; j < c->nr_creg; j++) prealloc_reg(c, &c->creg[j], i++); - for (j = 0; j < FRAG_ATTRIB_MAX; j++) { - if (c->key.vp_outputs_written & (1<<j)) { + for (j = 0; j < VERT_RESULT_MAX; j++) { + if (c->key.vp_outputs_written & BITFIELD64_BIT(j)) { int fp_index; if (j >= VERT_RESULT_VAR0) diff --git a/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c b/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c index dff466587a8..aa2e5195881 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c +++ b/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c @@ -66,19 +66,6 @@ static GLuint translate_wrap_mode( GLenum wrap ) } } - -static GLuint U_FIXED(GLfloat value, GLuint frac_bits) -{ - value *= (1<<frac_bits); - return value < 0 ? 0 : value; -} - -static GLint S_FIXED(GLfloat value, GLuint frac_bits) -{ - return value * (1<<frac_bits); -} - - static dri_bo *upload_default_color( struct brw_context *brw, const GLfloat *color ) { @@ -86,8 +73,8 @@ static dri_bo *upload_default_color( struct brw_context *brw, COPY_4V(sdc.color, color); - return brw_cache_data( &brw->cache, BRW_SAMPLER_DEFAULT_COLOR, &sdc, - NULL, 0 ); + return brw_cache_data(&brw->cache, BRW_SAMPLER_DEFAULT_COLOR, + &sdc, sizeof(sdc), NULL, 0); } @@ -228,8 +215,8 @@ static void brw_update_sampler_state(struct wm_sampler_entry *key, */ sampler->ss0.base_level = U_FIXED(0, 1); - sampler->ss1.max_lod = U_FIXED(MIN2(MAX2(key->maxlod, 0), 13), 6); - sampler->ss1.min_lod = U_FIXED(MIN2(MAX2(key->minlod, 0), 13), 6); + sampler->ss1.max_lod = U_FIXED(CLAMP(key->maxlod, 0, 13), 6); + sampler->ss1.min_lod = U_FIXED(CLAMP(key->minlod, 0, 13), 6); sampler->ss2.default_color_pointer = sdc_bo->offset >> 5; /* reloc */ } diff --git a/src/mesa/drivers/dri/i965/brw_wm_state.c b/src/mesa/drivers/dri/i965/brw_wm_state.c index 361f91292be..f89ed9bce70 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_state.c +++ b/src/mesa/drivers/dri/i965/brw_wm_state.c @@ -106,7 +106,7 @@ wm_unit_populate_key(struct brw_context *brw, struct brw_wm_unit_key *key) /* as far as we can tell */ key->computes_depth = - (fp->Base.OutputsWritten & (1 << FRAG_RESULT_DEPTH)) != 0; + (fp->Base.OutputsWritten & BITFIELD64_BIT(FRAG_RESULT_DEPTH)) != 0; /* BRW_NEW_DEPTH_BUFFER * Override for NULL depthbuffer case, required by the Pixel Shader Computed * Depth field. diff --git a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c index 51539ac1e73..3f9b1fbfdc6 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c +++ b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c @@ -31,7 +31,6 @@ #include "main/mtypes.h" -#include "main/texformat.h" #include "main/texstore.h" #include "shader/prog_parameter.h" @@ -70,7 +69,8 @@ static GLuint translate_tex_target( GLenum target ) } -static GLuint translate_tex_format( GLuint mesa_format, GLenum internal_format, +static GLuint translate_tex_format( gl_format mesa_format, + GLenum internal_format, GLenum depth_mode ) { switch( mesa_format ) { @@ -86,21 +86,22 @@ static GLuint translate_tex_format( GLuint mesa_format, GLenum internal_format, case MESA_FORMAT_AL88: return BRW_SURFACEFORMAT_L8A8_UNORM; + case MESA_FORMAT_AL1616: + return BRW_SURFACEFORMAT_L16A16_UNORM; + case MESA_FORMAT_RGB888: assert(0); /* not supported for sampling */ return BRW_SURFACEFORMAT_R8G8B8_UNORM; case MESA_FORMAT_ARGB8888: - if (internal_format == GL_RGB) - return BRW_SURFACEFORMAT_B8G8R8X8_UNORM; - else - return BRW_SURFACEFORMAT_B8G8R8A8_UNORM; + return BRW_SURFACEFORMAT_B8G8R8A8_UNORM; + + case MESA_FORMAT_XRGB8888: + return BRW_SURFACEFORMAT_B8G8R8X8_UNORM; case MESA_FORMAT_RGBA8888_REV: - if (internal_format == GL_RGB) - return BRW_SURFACEFORMAT_R8G8B8X8_UNORM; - else - return BRW_SURFACEFORMAT_R8G8B8A8_UNORM; + _mesa_problem(NULL, "unexpected format in i965:translate_tex_format()"); + return BRW_SURFACEFORMAT_R8G8B8A8_UNORM; case MESA_FORMAT_RGB565: return BRW_SURFACEFORMAT_B5G6R5_UNORM; @@ -287,7 +288,7 @@ brw_update_texture_surface( GLcontext *ctx, GLuint unit ) key.bo = NULL; key.offset = intelObj->textureOffset; } else { - key.format = firstImage->TexFormat->MesaFormat; + key.format = firstImage->TexFormat; key.internal_format = firstImage->InternalFormat; key.pitch = intelObj->mt->pitch; key.depth = firstImage->Depth; @@ -354,7 +355,10 @@ brw_create_constant_surface( struct brw_context *brw, NULL, NULL); if (key->bo) { - /* Emit relocation to surface contents */ + /* Emit relocation to surface contents. Section 5.1.1 of the gen4 + * bspec ("Data Cache") says that the data cache does not exist as + * a separate cache and is just the sampler cache. + */ dri_bo_emit_reloc(bo, I915_GEM_DOMAIN_SAMPLER, 0, 0, @@ -527,8 +531,15 @@ brw_update_renderbuffer_surface(struct brw_context *brw, region_bo = region->buffer; key.surface_type = BRW_SURFACE_2D; - switch (irb->texformat->MesaFormat) { + switch (irb->Base.Format) { + /* XRGB and ARGB are treated the same here because the chips in this + * family cannot render to XRGB targets. This means that we have to + * mask writes to alpha (ala glColorMask) and reconfigure the alpha + * blending hardware to use GL_ONE (or GL_ZERO) for cases where + * GL_DST_ALPHA (or GL_ONE_MINUS_DST_ALPHA) is used. + */ case MESA_FORMAT_ARGB8888: + case MESA_FORMAT_XRGB8888: key.surface_format = BRW_SURFACEFORMAT_B8G8R8A8_UNORM; break; case MESA_FORMAT_RGB565: @@ -541,8 +552,7 @@ brw_update_renderbuffer_surface(struct brw_context *brw, key.surface_format = BRW_SURFACEFORMAT_B4G4R4A4_UNORM; break; default: - _mesa_problem(ctx, "Bad renderbuffer format: %d\n", - irb->texformat->MesaFormat); + _mesa_problem(ctx, "Bad renderbuffer format: %d\n", irb->Base.Format); } key.tiling = region->tiling; if (brw->intel.intelScreen->driScrnPriv->dri2.enabled) { @@ -564,8 +574,16 @@ brw_update_renderbuffer_surface(struct brw_context *brw, key.cpp = 4; key.draw_offset = 0; } + /* _NEW_COLOR */ memcpy(key.color_mask, ctx->Color.ColorMask, sizeof(key.color_mask)); + + /* As mentioned above, disable writes to the alpha component when the + * renderbuffer is XRGB. + */ + if (ctx->Visual.alphaBits == 0) + key.color_mask[3] = GL_FALSE; + key.color_blend = (!ctx->Color._LogicOpEnabled && ctx->Color.BlendEnabled); @@ -660,7 +678,7 @@ brw_wm_get_binding_table(struct brw_context *brw) if (bind_bo == NULL) { GLuint data_size = brw->wm.nr_surfaces * sizeof(GLuint); - uint32_t *data = malloc(data_size); + uint32_t data[BRW_WM_MAX_SURF]; int i; for (i = 0; i < brw->wm.nr_surfaces; i++) @@ -685,8 +703,6 @@ brw_wm_get_binding_table(struct brw_context *brw) brw->wm.surf_bo[i]); } } - - free(data); } return bind_bo; @@ -695,11 +711,10 @@ brw_wm_get_binding_table(struct brw_context *brw) static void prepare_wm_surfaces(struct brw_context *brw ) { GLcontext *ctx = &brw->intel.ctx; - struct intel_context *intel = &brw->intel; GLuint i; int old_nr_surfaces; - /* _NEW_BUFFERS */ + /* _NEW_BUFFERS | _NEW_COLOR */ /* Update surfaces for drawing buffers */ if (ctx->DrawBuffer->_NumColorDrawBuffers >= 1) { for (i = 0; i < ctx->DrawBuffer->_NumColorDrawBuffers; i++) { @@ -712,7 +727,7 @@ static void prepare_wm_surfaces(struct brw_context *brw ) } old_nr_surfaces = brw->wm.nr_surfaces; - brw->wm.nr_surfaces = MAX_DRAW_BUFFERS; + brw->wm.nr_surfaces = BRW_MAX_DRAW_BUFFERS; if (brw->wm.surf_bo[SURF_INDEX_FRAG_CONST_BUFFER] != NULL) brw->wm.nr_surfaces = SURF_INDEX_FRAG_CONST_BUFFER + 1; @@ -724,17 +739,8 @@ static void prepare_wm_surfaces(struct brw_context *brw ) /* _NEW_TEXTURE, BRW_NEW_TEXDATA */ if (texUnit->_ReallyEnabled) { - if (texUnit->_Current == intel->frame_buffer_texobj) { - /* render to texture */ - dri_bo_unreference(brw->wm.surf_bo[surf]); - brw->wm.surf_bo[surf] = brw->wm.surf_bo[0]; - dri_bo_reference(brw->wm.surf_bo[surf]); - brw->wm.nr_surfaces = surf + 1; - } else { - /* regular texture */ - brw_update_texture_surface(ctx, i); - brw->wm.nr_surfaces = surf + 1; - } + brw_update_texture_surface(ctx, i); + brw->wm.nr_surfaces = surf + 1; } else { dri_bo_unreference(brw->wm.surf_bo[surf]); brw->wm.surf_bo[surf] = NULL; diff --git a/src/mesa/drivers/dri/i965/intel_generatemipmap.c b/src/mesa/drivers/dri/i965/intel_generatemipmap.c deleted file mode 120000 index 4c6b37ada01..00000000000 --- a/src/mesa/drivers/dri/i965/intel_generatemipmap.c +++ /dev/null @@ -1 +0,0 @@ -../intel/intel_generatemipmap.c
\ No newline at end of file diff --git a/src/mesa/drivers/dri/intel/intel_batchbuffer.c b/src/mesa/drivers/dri/intel/intel_batchbuffer.c index 6aa36d10b12..2eae9b66d87 100644 --- a/src/mesa/drivers/dri/intel/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/intel/intel_batchbuffer.c @@ -80,7 +80,7 @@ intel_batchbuffer_reset(struct intel_batchbuffer *batch) batch->buf = NULL; } - if (!batch->buffer && intel->ttm == GL_TRUE) + if (!batch->buffer) batch->buffer = malloc (intel->maxBatchSize); batch->buf = dri_bo_alloc(intel->bufmgr, "batchbuffer", @@ -201,11 +201,6 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file, drm_intel_bo_reference(intel->first_post_swapbuffers_batch); } - if (intel->first_post_swapbuffers_batch == NULL) { - intel->first_post_swapbuffers_batch = intel->batch->buf; - drm_intel_bo_reference(intel->first_post_swapbuffers_batch); - } - if (used == 0) { batch->cliprect_mode = IGNORE_CLIPRECTS; return; @@ -215,10 +210,10 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file, fprintf(stderr, "%s:%d: Batchbuffer flush with %db used\n", file, line, used); + batch->reserved_space = 0; /* Emit a flush if the bufmgr doesn't do it for us. */ - if (intel->always_flush_cache || !intel->ttm) { - *(GLuint *) (batch->ptr) = intel->vtbl.flush_cmd(); - batch->ptr += 4; + if (intel->always_flush_cache) { + intel_batchbuffer_emit_mi_flush(batch); used = batch->ptr - batch->map; } @@ -249,6 +244,11 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file, if (intel->vtbl.finish_batch) intel->vtbl.finish_batch(intel); + /* Check that we didn't just wrap our batchbuffer at a bad time. */ + assert(!intel->no_batch_wrap); + + batch->reserved_space = BATCH_RESERVED; + /* TODO: Just pass the relocation list and dma buffer up to the * kernel. */ @@ -304,3 +304,31 @@ intel_batchbuffer_data(struct intel_batchbuffer *batch, __memcpy(batch->ptr, data, bytes); batch->ptr += bytes; } + +/* Emit a pipelined flush to either flush render and texture cache for + * reading from a FBO-drawn texture, or flush so that frontbuffer + * render appears on the screen in DRI1. + * + * This is also used for the always_flush_cache driconf debug option. + */ +void +intel_batchbuffer_emit_mi_flush(struct intel_batchbuffer *batch) +{ + struct intel_context *intel = batch->intel; + + if (intel->gen >= 4) { + BEGIN_BATCH(4, IGNORE_CLIPRECTS); + OUT_BATCH(_3DSTATE_PIPE_CONTROL | + PIPE_CONTROL_INSTRUCTION_FLUSH | + PIPE_CONTROL_WRITE_FLUSH | + PIPE_CONTROL_NO_WRITE); + OUT_BATCH(0); /* write address */ + OUT_BATCH(0); /* write data */ + OUT_BATCH(0); /* write data */ + ADVANCE_BATCH(); + } else { + BEGIN_BATCH(1, IGNORE_CLIPRECTS); + OUT_BATCH(MI_FLUSH); + ADVANCE_BATCH(); + } +} diff --git a/src/mesa/drivers/dri/intel/intel_batchbuffer.h b/src/mesa/drivers/dri/intel/intel_batchbuffer.h index 51579df09e7..d4a94454dd5 100644 --- a/src/mesa/drivers/dri/intel/intel_batchbuffer.h +++ b/src/mesa/drivers/dri/intel/intel_batchbuffer.h @@ -62,6 +62,7 @@ struct intel_batchbuffer } emit; GLuint dirty_state; + GLuint reserved_space; }; struct intel_batchbuffer *intel_batchbuffer_alloc(struct intel_context @@ -95,6 +96,7 @@ GLboolean intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, uint32_t read_domains, uint32_t write_domain, uint32_t offset); +void intel_batchbuffer_emit_mi_flush(struct intel_batchbuffer *batch); /* Inline functions - might actually be better off with these * non-inlined. Certainly better off switching all command packets to @@ -104,7 +106,7 @@ GLboolean intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, static INLINE GLint intel_batchbuffer_space(struct intel_batchbuffer *batch) { - return (batch->size - BATCH_RESERVED) - (batch->ptr - batch->map); + return (batch->size - batch->reserved_space) - (batch->ptr - batch->map); } @@ -157,7 +159,7 @@ intel_batchbuffer_require_space(struct intel_batchbuffer *batch, #define OUT_BATCH(d) intel_batchbuffer_emit_dword(intel->batch, d) #define OUT_RELOC(buf, read_domains, write_domain, delta) do { \ - assert((delta) >= 0); \ + assert((unsigned) (delta) < buf->size); \ intel_batchbuffer_emit_reloc(intel->batch, buf, \ read_domains, write_domain, delta); \ } while (0) @@ -173,12 +175,4 @@ intel_batchbuffer_require_space(struct intel_batchbuffer *batch, intel->batch->emit.start_ptr = NULL; \ } while(0) - -static INLINE void -intel_batchbuffer_emit_mi_flush(struct intel_batchbuffer *batch) -{ - intel_batchbuffer_require_space(batch, 4, IGNORE_CLIPRECTS); - intel_batchbuffer_emit_dword(batch, MI_FLUSH); -} - #endif diff --git a/src/mesa/drivers/dri/intel/intel_blit.c b/src/mesa/drivers/dri/intel/intel_blit.c index 43141c509c7..cdf1408cd33 100644 --- a/src/mesa/drivers/dri/intel/intel_blit.c +++ b/src/mesa/drivers/dri/intel/intel_blit.c @@ -29,7 +29,6 @@ #include "main/mtypes.h" #include "main/context.h" #include "main/enums.h" -#include "main/texformat.h" #include "main/colormac.h" #include "intel_blit.h" @@ -371,8 +370,6 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask) skipBuffers = BUFFER_BIT_STENCIL; } - /* XXX Move this flush/lock into the following conditional? */ - intelFlush(&intel->ctx); LOCK_HARDWARE(intel); intel_get_cliprects(intel, &cliprects, &num_cliprects, &x_off, &y_off); @@ -438,6 +435,10 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask) intel_region_buffer(intel, irb->region, all ? INTEL_WRITE_FULL : INTEL_WRITE_PART); + int x1 = b.x1 + irb->region->draw_x; + int y1 = b.y1 + irb->region->draw_y; + int x2 = b.x2 + irb->region->draw_x; + int y2 = b.y2 + irb->region->draw_y; GLuint clearVal; GLint pitch, cpp; @@ -446,11 +447,10 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask) pitch = irb->region->pitch; cpp = irb->region->cpp; - DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", + DBG("%s dst:buf(%p)/%d %d,%d sz:%dx%d\n", __FUNCTION__, irb->region->buffer, (pitch * cpp), - irb->region->draw_offset, - b.x1, b.y1, b.x2 - b.x1, b.y2 - b.y1); + x1, y1, x2 - x1, y2 - y1); BR13 = 0xf0 << 16; CMD = XY_COLOR_BLT_CMD; @@ -496,12 +496,14 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask) CLAMPED_FLOAT_TO_UBYTE(clear[2], color[2]); CLAMPED_FLOAT_TO_UBYTE(clear[3], color[3]); - switch (irb->texformat->MesaFormat) { + switch (irb->Base.Format) { case MESA_FORMAT_ARGB8888: - clearVal = intel->ClearColor8888; + case MESA_FORMAT_XRGB8888: + clearVal = PACK_COLOR_8888(clear[3], clear[0], + clear[1], clear[2]); break; case MESA_FORMAT_RGB565: - clearVal = intel->ClearColor565; + clearVal = PACK_COLOR_565(clear[0], clear[1], clear[2]); break; case MESA_FORMAT_ARGB4444: clearVal = PACK_COLOR_4444(clear[3], clear[0], @@ -513,7 +515,7 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask) break; default: _mesa_problem(ctx, "Unexpected renderbuffer format: %d\n", - irb->texformat->MesaFormat); + irb->Base.Format); clearVal = 0; } } @@ -523,17 +525,17 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask) buf, irb->Base.Name); */ - assert(b.x1 < b.x2); - assert(b.y1 < b.y2); + assert(x1 < x2); + assert(y1 < y2); BEGIN_BATCH(6, REFERENCES_CLIPRECTS); OUT_BATCH(CMD); OUT_BATCH(BR13); - OUT_BATCH((b.y1 << 16) | b.x1); - OUT_BATCH((b.y2 << 16) | b.x2); + OUT_BATCH((y1 << 16) | x1); + OUT_BATCH((y2 << 16) | x2); OUT_RELOC(write_buffer, I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER, - irb->region->draw_offset); + 0); OUT_BATCH(clearVal); ADVANCE_BATCH(); clearMask &= ~bufBit; /* turn off bit, for faster loop exit */ diff --git a/src/mesa/drivers/dri/intel/intel_buffer_objects.c b/src/mesa/drivers/dri/intel/intel_buffer_objects.c index c55c5c426e0..3b7015b5ad3 100644 --- a/src/mesa/drivers/dri/intel/intel_buffer_objects.c +++ b/src/mesa/drivers/dri/intel/intel_buffer_objects.c @@ -207,8 +207,26 @@ intel_bufferobj_subdata(GLcontext * ctx, if (intel_obj->sys_buffer) memcpy((char *)intel_obj->sys_buffer + offset, data, size); - else - dri_bo_subdata(intel_obj->buffer, offset, size, data); + else { + /* Flush any existing batchbuffer that might reference this data. */ + if (drm_intel_bo_busy(intel_obj->buffer) || + drm_intel_bo_references(intel->batch->buf, intel_obj->buffer)) { + drm_intel_bo *temp_bo; + + temp_bo = drm_intel_bo_alloc(intel->bufmgr, "subdata temp", size, 64); + + drm_intel_bo_subdata(temp_bo, 0, size, data); + + intel_emit_linear_blit(intel, + intel_obj->buffer, offset, + temp_bo, 0, + size); + + drm_intel_bo_unreference(temp_bo); + } else { + dri_bo_subdata(intel_obj->buffer, offset, size, data); + } + } } @@ -225,7 +243,10 @@ intel_bufferobj_get_subdata(GLcontext * ctx, struct intel_buffer_object *intel_obj = intel_buffer_object(obj); assert(intel_obj); - dri_bo_get_subdata(intel_obj->buffer, offset, size, data); + if (intel_obj->sys_buffer) + memcpy(data, (char *)intel_obj->sys_buffer + offset, size); + else + dri_bo_get_subdata(intel_obj->buffer, offset, size, data); } @@ -247,13 +268,14 @@ intel_bufferobj_map(GLcontext * ctx, if (intel_obj->sys_buffer) { obj->Pointer = intel_obj->sys_buffer; + obj->Length = obj->Size; + obj->Offset = 0; return obj->Pointer; } - /* Flush any existing batchbuffer that might have written to this - * buffer. - */ - intelFlush(ctx); + /* Flush any existing batchbuffer that might reference this data. */ + if (drm_intel_bo_references(intel->batch->buf, intel_obj->buffer)) + intelFlush(ctx); if (intel_obj->region) intel_bufferobj_cow(intel, intel_obj); @@ -323,7 +345,8 @@ intel_bufferobj_map_range(GLcontext * ctx, * the batchbuffer so that GEM knows about the buffer access for later * syncing. */ - if (!(access & GL_MAP_UNSYNCHRONIZED_BIT)) + if (!(access & GL_MAP_UNSYNCHRONIZED_BIT) && + drm_intel_bo_references(intel->batch->buf, intel_obj->buffer)) intelFlush(ctx); if (intel_obj->buffer == NULL) { diff --git a/src/mesa/drivers/dri/intel/intel_buffers.c b/src/mesa/drivers/dri/intel/intel_buffers.c index e7357e78c53..6b12d484d85 100644 --- a/src/mesa/drivers/dri/intel/intel_buffers.c +++ b/src/mesa/drivers/dri/intel/intel_buffers.c @@ -133,6 +133,25 @@ intel_get_cliprects(struct intel_context *intel, /** + * Check if we're about to draw into the front color buffer. + * If so, set the intel->front_buffer_dirty field to true. + */ +void +intel_check_front_buffer_rendering(struct intel_context *intel) +{ + const struct gl_framebuffer *fb = intel->ctx.DrawBuffer; + if (fb->Name == 0) { + /* drawing to window system buffer */ + if (fb->_NumColorDrawBuffers > 0) { + if (fb->_ColorDrawBufferIndexes[0] == BUFFER_FRONT_LEFT) { + intel->front_buffer_dirty = GL_TRUE; + } + } + } +} + + +/** * Update the hardware state for drawing into a window or framebuffer object. * * Called by glDrawBuffer, glBindFramebufferEXT, MakeCurrent, and other @@ -192,7 +211,7 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) } else { /* Get the intel_renderbuffer for the single colorbuffer we're drawing - * into, and set up cliprects if it's . + * into, and set up cliprects if it's a DRI1 window front buffer. */ if (fb->Name == 0) { intel->constant_cliprect = intel->driScreen->dri2.enabled; @@ -202,14 +221,12 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) intel_batchbuffer_flush(intel->batch); intel->front_cliprects = GL_TRUE; colorRegions[0] = intel_get_rb_region(fb, BUFFER_FRONT_LEFT); - - intel->front_buffer_dirty = GL_TRUE; } else { if (!intel->constant_cliprect && intel->front_cliprects) intel_batchbuffer_flush(intel->batch); intel->front_cliprects = GL_FALSE; - colorRegions[0]= intel_get_rb_region(fb, BUFFER_BACK_LEFT); + colorRegions[0] = intel_get_rb_region(fb, BUFFER_BACK_LEFT); } } else { @@ -257,7 +274,7 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) if (fb->_StencilBuffer && fb->_StencilBuffer->Wrapped) { irbStencil = intel_renderbuffer(fb->_StencilBuffer->Wrapped); if (irbStencil && irbStencil->region) { - ASSERT(irbStencil->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + ASSERT(irbStencil->Base.Format == MESA_FORMAT_S8_Z24); FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_FALSE); } else { diff --git a/src/mesa/drivers/dri/intel/intel_buffers.h b/src/mesa/drivers/dri/intel/intel_buffers.h index 6069d38e9eb..d7800f2ca2f 100644 --- a/src/mesa/drivers/dri/intel/intel_buffers.h +++ b/src/mesa/drivers/dri/intel/intel_buffers.h @@ -45,6 +45,8 @@ extern struct intel_region *intel_readbuf_region(struct intel_context *intel); extern struct intel_region *intel_drawbuf_region(struct intel_context *intel); +extern void intel_check_front_buffer_rendering(struct intel_context *intel); + extern void intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb); extern void intelInitBufferFuncs(struct dd_function_table *functions); diff --git a/src/mesa/drivers/dri/intel/intel_clear.c b/src/mesa/drivers/dri/intel/intel_clear.c index 9efe6a277c7..f682ee3de5b 100644 --- a/src/mesa/drivers/dri/intel/intel_clear.c +++ b/src/mesa/drivers/dri/intel/intel_clear.c @@ -38,6 +38,7 @@ #include "intel_fbo.h" #include "intel_pixel.h" #include "intel_regions.h" +#include "intel_batchbuffer.h" #define FILE_DEBUG_FLAG DEBUG_BLIT @@ -74,6 +75,10 @@ intelClear(GLcontext *ctx, GLbitfield mask) struct gl_framebuffer *fb = ctx->DrawBuffer; GLuint i; + if (mask & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_FRONT_RIGHT)) { + intel->front_buffer_dirty = GL_TRUE; + } + if (0) fprintf(stderr, "%s\n", __FUNCTION__); @@ -170,8 +175,8 @@ intelClear(GLcontext *ctx, GLbitfield mask) } DBG("\n"); } - intelFlush(&intel->ctx); - _mesa_meta_clear(&intel->ctx, tri_mask); + + _mesa_meta_Clear(&intel->ctx, tri_mask); } if (swrast_mask) { diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c index 03db8b1c68e..1434ae530be 100644 --- a/src/mesa/drivers/dri/intel/intel_context.c +++ b/src/mesa/drivers/dri/intel/intel_context.c @@ -176,9 +176,7 @@ intelGetString(GLcontext * ctx, GLenum name) break; } - (void) driGetRendererString(buffer, chipset, - (intel->ttm) ? DRIVER_DATE_GEM : DRIVER_DATE, - 0); + (void) driGetRendererString(buffer, chipset, DRIVER_DATE_GEM, 0); return (GLubyte *) buffer; default: @@ -189,19 +187,7 @@ intelGetString(GLcontext * ctx, GLenum name) static unsigned intel_bits_per_pixel(const struct intel_renderbuffer *rb) { - switch (rb->Base._ActualFormat) { - case GL_RGB5: - case GL_DEPTH_COMPONENT16: - return 16; - case GL_RGB8: - case GL_RGBA8: - case GL_DEPTH_COMPONENT24: - case GL_DEPTH24_STENCIL8_EXT: - case GL_STENCIL_INDEX8_EXT: - return 32; - default: - return 0; - } + return _mesa_get_format_bytes(rb->Base.Format) * 8; } void @@ -489,14 +475,14 @@ intel_flush(GLcontext *ctx, GLboolean needs_mi_flush) if (intel->Fallback) _swrast_flush(ctx); - if (!IS_965(intel->intelScreen->deviceID)) + if (intel->gen < 4) INTEL_FIREVERTICES(intel); /* Emit a flush so that any frontbuffer rendering that might have occurred * lands onscreen in a timely manner, even if the X Server doesn't trigger * a flush for us. */ - if (needs_mi_flush) + if (!intel->driScreen->dri2.enabled && needs_mi_flush) intel_batchbuffer_emit_mi_flush(intel->batch); if (intel->batch->map != intel->batch->ptr) @@ -507,7 +493,8 @@ intel_flush(GLcontext *ctx, GLboolean needs_mi_flush) if (screen->dri2.loader && (screen->dri2.loader->base.version >= 2) - && (screen->dri2.loader->flushFrontBuffer != NULL)) { + && (screen->dri2.loader->flushFrontBuffer != NULL) && + intel->driDrawable && intel->driDrawable->loaderPrivate) { (*screen->dri2.loader->flushFrontBuffer)(intel->driDrawable, intel->driDrawable->loaderPrivate); @@ -612,6 +599,7 @@ intelInitContext(struct intel_context *intel, __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; int fthrottle_mode; + int bo_reuse_mode; if (!_mesa_initialize_context(&intel->ctx, mesaVis, shareCtx, functions, (void *) intel)) { @@ -625,6 +613,13 @@ intelInitContext(struct intel_context *intel, intel->sarea = intelScreen->sarea; intel->driContext = driContextPriv; + if (IS_965(intel->intelScreen->deviceID)) + intel->gen = 4; + else if (IS_9XX(intel->intelScreen->deviceID)) + intel->gen = 3; + else + intel->gen = 2; + /* Dri stuff */ intel->hHWContext = driContextPriv->hHWContext; intel->driFd = sPriv->fd; @@ -632,25 +627,21 @@ intelInitContext(struct intel_context *intel, driParseConfigFiles(&intel->optionCache, &intelScreen->optionCache, intel->driScreen->myNum, - IS_965(intelScreen->deviceID) ? "i965" : "i915"); + (intel->gen >= 4) ? "i965" : "i915"); if (intelScreen->deviceID == PCI_CHIP_I865_G) intel->maxBatchSize = 4096; else intel->maxBatchSize = BATCH_SZ; intel->bufmgr = intelScreen->bufmgr; - intel->ttm = intelScreen->ttm; - if (intel->ttm) { - int bo_reuse_mode; - bo_reuse_mode = driQueryOptioni(&intel->optionCache, "bo_reuse"); - switch (bo_reuse_mode) { - case DRI_CONF_BO_REUSE_DISABLED: - break; - case DRI_CONF_BO_REUSE_ALL: - intel_bufmgr_gem_enable_reuse(intel->bufmgr); - break; - } + bo_reuse_mode = driQueryOptioni(&intel->optionCache, "bo_reuse"); + switch (bo_reuse_mode) { + case DRI_CONF_BO_REUSE_DISABLED: + break; + case DRI_CONF_BO_REUSE_ALL: + intel_bufmgr_gem_enable_reuse(intel->bufmgr); + break; } /* This doesn't yet catch all non-conformant rendering, but it's a @@ -694,7 +685,7 @@ intelInitContext(struct intel_context *intel, meta_init_metaops(ctx, &intel->meta); ctx->Const.MaxColorAttachments = 4; /* XXX FBO: review this */ - if (IS_965(intelScreen->deviceID)) { + if (intel->gen >= 4) { if (MAX_WIDTH > 8192) ctx->Const.MaxRenderbufferSize = 8192; } else { @@ -731,25 +722,19 @@ intelInitContext(struct intel_context *intel, break; } - if (IS_965(intelScreen->deviceID)) + if (intel->gen >= 4) intel->polygon_offset_scale /= 0xffff; intel->RenderIndex = ~0; fthrottle_mode = driQueryOptioni(&intel->optionCache, "fthrottle_mode"); - intel->irqsEmitted = 0; - - intel->do_irqs = (intel->intelScreen->irq_active && - fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS); - - intel->do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS); - if (IS_965(intelScreen->deviceID) && !intel->intelScreen->irq_active) { + if (intel->gen >= 4 && !intel->intelScreen->irq_active) { _mesa_printf("IRQs not active. Exiting\n"); exit(1); } - intelInitExtensions(ctx, GL_FALSE); + intelInitExtensions(ctx); INTEL_DEBUG = driParseDebugString(getenv("INTEL_DEBUG"), debug_control); if (INTEL_DEBUG & DEBUG_BUFMGR) @@ -829,7 +814,7 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) _vbo_DestroyContext(&intel->ctx); _swrast_DestroyContext(&intel->ctx); - intel->Fallback = 0; /* don't call _swrast_Flush later */ + intel->Fallback = 0x0; /* don't call _swrast_Flush later */ intel_batchbuffer_free(intel->batch); intel->batch = NULL; @@ -934,10 +919,23 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, __DRIdrawablePrivate * driReadPriv) { __DRIscreenPrivate *psp = driDrawPriv->driScreenPriv; + struct intel_context *intel; + GET_CURRENT_CONTEXT(curCtx); + + if (driContextPriv) + intel = (struct intel_context *) driContextPriv->driverPrivate; + else + intel = NULL; + + /* According to the glXMakeCurrent() man page: "Pending commands to + * the previous context, if any, are flushed before it is released." + * But only flush if we're actually changing contexts. + */ + if (intel_context(curCtx) && intel_context(curCtx) != intel) { + _mesa_flush(curCtx); + } if (driContextPriv) { - struct intel_context *intel = - (struct intel_context *) driContextPriv->driverPrivate; struct intel_framebuffer *intel_fb = (struct intel_framebuffer *) driDrawPriv->driverPrivate; GLframebuffer *readFb = (GLframebuffer *) driReadPriv->driverPrivate; @@ -983,41 +981,35 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, _mesa_make_current(&intel->ctx, &intel_fb->Base, readFb); - /* The drawbuffer won't always be updated by _mesa_make_current: - */ - if (intel->ctx.DrawBuffer == &intel_fb->Base) { - - if (intel->driReadDrawable != driReadPriv) - intel->driReadDrawable = driReadPriv; - - if (intel->driDrawable != driDrawPriv) { - if (driDrawPriv->swap_interval == (unsigned)-1) { - int i; - - driDrawPriv->vblFlags = (intel->intelScreen->irq_active != 0) - ? driGetDefaultVBlankFlags(&intel->optionCache) - : VBLANK_FLAG_NO_IRQ; - - /* Prevent error printf if one crtc is disabled, this will - * be properly calculated in intelWindowMoved() next. - */ - driDrawPriv->vblFlags = intelFixupVblank(intel, driDrawPriv); - - (*psp->systemTime->getUST) (&intel_fb->swap_ust); - driDrawableInitVBlank(driDrawPriv); - intel_fb->vbl_waited = driDrawPriv->vblSeq; - - for (i = 0; i < 2; i++) { - if (intel_fb->color_rb[i]) - intel_fb->color_rb[i]->vbl_pending = driDrawPriv->vblSeq; - } - } - intel->driDrawable = driDrawPriv; - intelWindowMoved(intel); - } + intel->driReadDrawable = driReadPriv; + + if (intel->driDrawable != driDrawPriv) { + if (driDrawPriv->swap_interval == (unsigned)-1) { + int i; + + driDrawPriv->vblFlags = (intel->intelScreen->irq_active != 0) + ? driGetDefaultVBlankFlags(&intel->optionCache) + : VBLANK_FLAG_NO_IRQ; + + /* Prevent error printf if one crtc is disabled, this will + * be properly calculated in intelWindowMoved() next. + */ + driDrawPriv->vblFlags = intelFixupVblank(intel, driDrawPriv); - intel_draw_buffer(&intel->ctx, &intel_fb->Base); + (*psp->systemTime->getUST) (&intel_fb->swap_ust); + driDrawableInitVBlank(driDrawPriv); + intel_fb->vbl_waited = driDrawPriv->vblSeq; + + for (i = 0; i < 2; i++) { + if (intel_fb->color_rb[i]) + intel_fb->color_rb[i]->vbl_pending = driDrawPriv->vblSeq; + } + } + intel->driDrawable = driDrawPriv; + intelWindowMoved(intel); } + + intel_draw_buffer(&intel->ctx, &intel_fb->Base); } else { _mesa_make_current(NULL, NULL, NULL); @@ -1055,21 +1047,6 @@ intelContendedLock(struct intel_context *intel, GLuint flags) sarea->ctxOwner = me; } - /* If the last consumer of the texture memory wasn't us, notify the fake - * bufmgr and record the new owner. We should have the memory shared - * between contexts of a single fake bufmgr, but this will at least make - * things correct for now. - */ - if (!intel->ttm && sarea->texAge != intel->hHWContext) { - sarea->texAge = intel->hHWContext; - intel_bufmgr_fake_contended_lock_take(intel->bufmgr); - if (INTEL_DEBUG & DEBUG_BATCH) - intel_decode_context_reset(); - if (INTEL_DEBUG & DEBUG_BUFMGR) - fprintf(stderr, "Lost Textures: sarea->texAge %x hw context %x\n", - sarea->ctxOwner, intel->hHWContext); - } - /* Drawable changed? */ if (dPriv && intel->lastStamp != dPriv->lastStamp) { diff --git a/src/mesa/drivers/dri/intel/intel_context.h b/src/mesa/drivers/dri/intel/intel_context.h index 03e7cf39d68..481202c971b 100644 --- a/src/mesa/drivers/dri/intel/intel_context.h +++ b/src/mesa/drivers/dri/intel/intel_context.h @@ -61,6 +61,10 @@ typedef void (*intel_line_func) (struct intel_context *, intelVertex *, intelVertex *); typedef void (*intel_point_func) (struct intel_context *, intelVertex *); +/** + * Bits for intel->Fallback field + */ +/*@{*/ #define INTEL_FALLBACK_DRAW_BUFFER 0x1 #define INTEL_FALLBACK_READ_BUFFER 0x2 #define INTEL_FALLBACK_DEPTH_BUFFER 0x4 @@ -68,8 +72,10 @@ typedef void (*intel_point_func) (struct intel_context *, intelVertex *); #define INTEL_FALLBACK_USER 0x10 #define INTEL_FALLBACK_RENDERMODE 0x20 #define INTEL_FALLBACK_TEXTURE 0x40 +#define INTEL_FALLBACK_DRIVER 0x1000 /**< first for drivers */ +/*@}*/ -extern void intelFallback(struct intel_context *intel, GLuint bit, +extern void intelFallback(struct intel_context *intel, GLbitfield bit, GLboolean mode); #define FALLBACK( intel, bit, mode ) intelFallback( intel, bit, mode ) @@ -111,8 +117,6 @@ struct intel_context struct intel_region * depth_region, GLuint num_regions); - GLuint (*flush_cmd) (void); - void (*reduced_primitive_state) (struct intel_context * intel, GLenum rprim); @@ -131,14 +135,6 @@ struct intel_context struct intel_region * draw_region, struct intel_region * depth_region); - void (*meta_draw_quad)(struct intel_context *intel, - GLfloat x0, GLfloat x1, - GLfloat y0, GLfloat y1, - GLfloat z, - GLuint color, /* ARGB32 */ - GLfloat s0, GLfloat s1, - GLfloat t0, GLfloat t1); - void (*meta_color_mask) (struct intel_context * intel, GLboolean); void (*meta_stencil_replace) (struct intel_context * intel, @@ -170,27 +166,24 @@ struct intel_context struct dri_metaops meta; - GLint refcount; - GLuint Fallback; + GLbitfield Fallback; /**< mask of INTEL_FALLBACK_x bits */ GLuint NewGLState; dri_bufmgr *bufmgr; unsigned int maxBatchSize; + /** + * Generation number of the hardware: 2 is 8xx, 3 is 9xx pre-965, 4 is 965. + */ + int gen; + struct intel_region *front_region; struct intel_region *back_region; struct intel_region *depth_region; - /** - * This value indicates that the kernel memory manager is being used - * instead of the fake client-side memory manager. - */ - GLboolean ttm; - struct intel_batchbuffer *batch; drm_intel_bo *first_post_swapbuffers_batch; GLboolean no_batch_wrap; - unsigned batch_id; struct { @@ -210,10 +203,6 @@ struct intel_context char *prevLockFile; int prevLockLine; - GLuint ClearColor565; - GLuint ClearColor8888; - - /* Offsets of fields within the current vertex: */ GLuint coloroffset; @@ -230,6 +219,7 @@ struct intel_context GLboolean hw_stipple; GLboolean depth_buffer_is_float; GLboolean no_rast; + GLboolean no_hw; GLboolean always_flush_batch; GLboolean always_flush_cache; @@ -254,9 +244,6 @@ struct intel_context intel_line_func draw_line; intel_tri_func draw_tri; - /* These refer to the current drawing buffer: - */ - struct gl_texture_object *frame_buffer_texobj; /** * Set to true if a single constant cliprect should be used in the * batchbuffer. Otherwise, cliprects must be calculated at batchbuffer @@ -296,16 +283,8 @@ struct intel_context GLboolean use_texture_tiling; GLboolean use_early_z; - drm_clip_rect_t fboRect; /**< cliprect for FBO rendering */ - int perf_boxes; - - GLuint do_usleeps; - int do_irqs; - GLuint irqsEmitted; - - GLboolean scissor; drm_clip_rect_t draw_rect; drm_clip_rect_t scissor_rect; @@ -322,8 +301,6 @@ struct intel_context GLuint lastStamp; - GLboolean no_hw; - /** * Configuration cache */ @@ -351,6 +328,19 @@ extern char *__progname; #define ALIGN(value, alignment) ((value + alignment - 1) & ~(alignment - 1)) #define IS_POWER_OF_TWO(val) (((val) & (val - 1)) == 0) +static inline uint32_t +U_FIXED(float value, uint32_t frac_bits) +{ + value *= (1 << frac_bits); + return value < 0 ? 0 : value; +} + +static inline uint32_t +S_FIXED(float value, uint32_t frac_bits) +{ + return value * (1 << frac_bits); +} + #define INTEL_FIREVERTICES(intel) \ do { \ if ((intel)->prim.flush) \ @@ -358,29 +348,6 @@ do { \ } while (0) /* ================================================================ - * Color packing: - */ - -#define INTEL_PACKCOLOR4444(r,g,b,a) \ - ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4)) - -#define INTEL_PACKCOLOR1555(r,g,b,a) \ - ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) | \ - ((a) ? 0x8000 : 0)) - -#define INTEL_PACKCOLOR565(r,g,b) \ - ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3)) - -#define INTEL_PACKCOLOR8888(r,g,b,a) \ - ((a<<24) | (r<<16) | (g<<8) | b) - -#define INTEL_PACKCOLOR(format, r, g, b, a) \ -(format == DV_PF_555 ? INTEL_PACKCOLOR1555(r,g,b,a) : \ - (format == DV_PF_565 ? INTEL_PACKCOLOR565(r,g,b) : \ - (format == DV_PF_8888 ? INTEL_PACKCOLOR8888(r,g,b,a) : \ - 0))) - -/* ================================================================ * From linux kernel i386 header files, copes with odd sizes better * than COPY_DWORDS would: * XXX Put this in src/mesa/main/imports.h ??? @@ -572,4 +539,25 @@ is_power_of_two(uint32_t value) return (value & (value - 1)) == 0; } +static inline void +intel_bo_map_gtt_preferred(struct intel_context *intel, + drm_intel_bo *bo, + GLboolean write) +{ + if (intel->intelScreen->kernel_exec_fencing) + drm_intel_gem_bo_map_gtt(bo); + else + drm_intel_bo_map(bo, write); +} + +static inline void +intel_bo_unmap_gtt_preferred(struct intel_context *intel, + drm_intel_bo *bo) +{ + if (intel->intelScreen->kernel_exec_fencing) + drm_intel_gem_bo_unmap_gtt(bo); + else + drm_intel_bo_unmap(bo); +} + #endif diff --git a/src/mesa/drivers/dri/intel/intel_depthtmp.h b/src/mesa/drivers/dri/intel/intel_depthtmp.h index 16d77084530..a9c75d44cf3 100644 --- a/src/mesa/drivers/dri/intel/intel_depthtmp.h +++ b/src/mesa/drivers/dri/intel/intel_depthtmp.h @@ -31,6 +31,16 @@ */ #define VALUE_TYPE INTEL_VALUE_TYPE +#define WRITE_DEPTH(_x, _y, d) \ + (*(INTEL_VALUE_TYPE *)(irb->region->buffer->virtual + \ + NO_TILE(_x, _y)) = d) +#define READ_DEPTH(d, _x, _y) \ + d = *(INTEL_VALUE_TYPE *)(irb->region->buffer->virtual + \ + NO_TILE(_x, _y)) +#define TAG(x) INTEL_TAG(intel_gttmap_##x) +#include "depthtmp.h" + +#define VALUE_TYPE INTEL_VALUE_TYPE #define WRITE_DEPTH(_x, _y, d) INTEL_WRITE_DEPTH(NO_TILE(_x, _y), d) #define READ_DEPTH(d, _x, _y) d = INTEL_READ_DEPTH(NO_TILE(_x, _y)) #define TAG(x) INTEL_TAG(intel##x) diff --git a/src/mesa/drivers/dri/intel/intel_extensions.c b/src/mesa/drivers/dri/intel/intel_extensions.c index 5431cf90a1b..86dc42cc510 100644 --- a/src/mesa/drivers/dri/intel/intel_extensions.c +++ b/src/mesa/drivers/dri/intel/intel_extensions.c @@ -28,6 +28,7 @@ #include "intel_chipset.h" #include "intel_context.h" #include "intel_extensions.h" +#include "utils.h" #define need_GL_ARB_copy_buffer @@ -63,7 +64,7 @@ #define need_GL_VERSION_2_0 #define need_GL_VERSION_2_1 -#include "extension_helper.h" +#include "main/remap_helper.h" /** @@ -78,8 +79,12 @@ static const struct dri_extension card_extensions[] = { { "GL_ARB_half_float_pixel", NULL }, { "GL_ARB_map_buffer_range", GL_ARB_map_buffer_range_functions }, { "GL_ARB_multitexture", NULL }, + { "GL_ARB_pixel_buffer_object", NULL }, { "GL_ARB_point_parameters", GL_ARB_point_parameters_functions }, { "GL_ARB_point_sprite", NULL }, + { "GL_ARB_shader_objects", GL_ARB_shader_objects_functions }, + { "GL_ARB_shading_language_100", GL_VERSION_2_0_functions }, + { "GL_ARB_shading_language_120", GL_VERSION_2_1_functions }, { "GL_ARB_sync", GL_ARB_sync_functions }, { "GL_ARB_texture_border_clamp", NULL }, { "GL_ARB_texture_cube_map", NULL }, @@ -91,6 +96,7 @@ static const struct dri_extension card_extensions[] = { { "GL_ARB_texture_rectangle", NULL }, { "GL_ARB_vertex_array_object", GL_ARB_vertex_array_object_functions}, { "GL_ARB_vertex_program", GL_ARB_vertex_program_functions }, + { "GL_ARB_vertex_shader", GL_ARB_vertex_shader_functions }, { "GL_ARB_window_pos", GL_ARB_window_pos_functions }, { "GL_EXT_blend_color", GL_EXT_blend_color_functions }, { "GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions }, @@ -99,6 +105,8 @@ static const struct dri_extension card_extensions[] = { { "GL_EXT_blend_logic_op", NULL }, { "GL_EXT_blend_subtract", NULL }, { "GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions }, + { "GL_EXT_framebuffer_blit", GL_EXT_framebuffer_blit_functions }, + { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions }, { "GL_EXT_fog_coord", GL_EXT_fog_coord_functions }, { "GL_EXT_gpu_program_parameters", GL_EXT_gpu_program_parameters_functions }, { "GL_EXT_packed_depth_stencil", NULL }, @@ -116,7 +124,6 @@ static const struct dri_extension card_extensions[] = { { "GL_MESA_pack_invert", NULL }, { "GL_MESA_ycbcr_texture", NULL }, { "GL_NV_blend_square", NULL }, - { "GL_NV_point_sprite", GL_NV_point_sprite_functions }, { "GL_NV_vertex_program", GL_NV_vertex_program_functions }, { "GL_NV_vertex_program1_1", NULL }, { "GL_SGIS_generate_mipmap", NULL }, @@ -150,13 +157,9 @@ static const struct dri_extension brw_extensions[] = { { "GL_ARB_occlusion_query", GL_ARB_occlusion_query_functions }, { "GL_ARB_point_sprite", NULL }, { "GL_ARB_seamless_cube_map", NULL }, - { "GL_ARB_shader_objects", GL_ARB_shader_objects_functions }, - { "GL_ARB_shading_language_100", GL_VERSION_2_0_functions }, - { "GL_ARB_shading_language_120", GL_VERSION_2_1_functions }, { "GL_ARB_shadow", NULL }, { "GL_MESA_texture_signed_rgba", NULL }, { "GL_ARB_texture_non_power_of_two", NULL }, - { "GL_ARB_vertex_shader", GL_ARB_vertex_shader_functions }, { "GL_EXT_shadow_funcs", NULL }, { "GL_EXT_stencil_two_side", GL_EXT_stencil_two_side_functions }, { "GL_EXT_texture_sRGB", NULL }, @@ -171,40 +174,39 @@ static const struct dri_extension brw_extensions[] = { static const struct dri_extension arb_oq_extensions[] = { + { "GL_ARB_occlusion_query", GL_ARB_occlusion_query_functions }, { NULL, NULL } }; - -static const struct dri_extension ttm_extensions[] = { - { "GL_ARB_pixel_buffer_object", NULL }, - { "GL_EXT_framebuffer_blit", GL_EXT_framebuffer_blit_functions }, - { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions }, +static const struct dri_extension fragment_shader_extensions[] = { + { "GL_ARB_fragment_shader", NULL }, { NULL, NULL } }; - /** * Initializes potential list of extensions if ctx == NULL, or actually enables * extensions for a context. */ void -intelInitExtensions(GLcontext *ctx, GLboolean enable_imaging) +intelInitExtensions(GLcontext *ctx) { - struct intel_context *intel = ctx?intel_context(ctx):NULL; + struct intel_context *intel = intel_context(ctx); /* Disable imaging extension until convolution is working in teximage paths. */ - enable_imaging = GL_FALSE; - - driInitExtensions(ctx, card_extensions, enable_imaging); - - if (intel == NULL || intel->ttm) - driInitExtensions(ctx, ttm_extensions, GL_FALSE); + driInitExtensions(ctx, card_extensions, GL_FALSE); - if (intel == NULL || IS_965(intel->intelScreen->deviceID)) + if (IS_965(intel->intelScreen->deviceID)) driInitExtensions(ctx, brw_extensions, GL_FALSE); - if (intel == NULL || IS_915(intel->intelScreen->deviceID) - || IS_945(intel->intelScreen->deviceID)) + if (IS_915(intel->intelScreen->deviceID) + || IS_945(intel->intelScreen->deviceID)) { driInitExtensions(ctx, i915_extensions, GL_FALSE); + + if (driQueryOptionb(&intel->optionCache, "fragment_shader")) + driInitExtensions(ctx, fragment_shader_extensions, GL_FALSE); + + if (driQueryOptionb(&intel->optionCache, "stub_occlusion_query")) + driInitExtensions(ctx, arb_oq_extensions, GL_FALSE); + } } diff --git a/src/mesa/drivers/dri/intel/intel_extensions.h b/src/mesa/drivers/dri/intel/intel_extensions.h index 97147ecdb0f..1d1c97a4a9b 100644 --- a/src/mesa/drivers/dri/intel/intel_extensions.h +++ b/src/mesa/drivers/dri/intel/intel_extensions.h @@ -30,7 +30,7 @@ extern void -intelInitExtensions(GLcontext *ctx, GLboolean enable_imaging); +intelInitExtensions(GLcontext *ctx); #endif diff --git a/src/mesa/drivers/dri/intel/intel_fbo.c b/src/mesa/drivers/dri/intel/intel_fbo.c index 804c0348401..608f75b8240 100644 --- a/src/mesa/drivers/dri/intel/intel_fbo.c +++ b/src/mesa/drivers/dri/intel/intel_fbo.c @@ -33,11 +33,11 @@ #include "main/framebuffer.h" #include "main/renderbuffer.h" #include "main/context.h" -#include "main/texformat.h" #include "main/texrender.h" #include "drivers/common/meta.h" #include "intel_context.h" +#include "intel_batchbuffer.h" #include "intel_buffers.h" #include "intel_fbo.h" #include "intel_mipmap_tree.h" @@ -106,8 +106,8 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, { struct intel_context *intel = intel_context(ctx); struct intel_renderbuffer *irb = intel_renderbuffer(rb); - GLboolean softwareBuffer = GL_FALSE; int cpp; + GLuint pitch; ASSERT(rb->Name != 0); @@ -115,27 +115,16 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, case GL_R3_G3_B2: case GL_RGB4: case GL_RGB5: - rb->_ActualFormat = GL_RGB5; + rb->Format = MESA_FORMAT_RGB565; rb->DataType = GL_UNSIGNED_BYTE; - rb->RedBits = 5; - rb->GreenBits = 6; - rb->BlueBits = 5; - irb->texformat = &_mesa_texformat_rgb565; - cpp = 2; break; case GL_RGB: case GL_RGB8: case GL_RGB10: case GL_RGB12: case GL_RGB16: - rb->_ActualFormat = GL_RGB8; + rb->Format = MESA_FORMAT_XRGB8888; rb->DataType = GL_UNSIGNED_BYTE; - rb->RedBits = 8; - rb->GreenBits = 8; - rb->BlueBits = 8; - rb->AlphaBits = 0; - irb->texformat = &_mesa_texformat_argb8888; /* XXX: Need xrgb8888 */ - cpp = 4; break; case GL_RGBA: case GL_RGBA2: @@ -145,14 +134,8 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, case GL_RGB10_A2: case GL_RGBA12: case GL_RGBA16: - rb->_ActualFormat = GL_RGBA8; + rb->Format = MESA_FORMAT_ARGB8888; rb->DataType = GL_UNSIGNED_BYTE; - rb->RedBits = 8; - rb->GreenBits = 8; - rb->BlueBits = 8; - rb->AlphaBits = 8; - irb->texformat = &_mesa_texformat_argb8888; - cpp = 4; break; case GL_STENCIL_INDEX: case GL_STENCIL_INDEX1_EXT: @@ -160,36 +143,23 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, case GL_STENCIL_INDEX8_EXT: case GL_STENCIL_INDEX16_EXT: /* alloc a depth+stencil buffer */ - rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT; + rb->Format = MESA_FORMAT_S8_Z24; rb->DataType = GL_UNSIGNED_INT_24_8_EXT; - rb->StencilBits = 8; - cpp = 4; - irb->texformat = &_mesa_texformat_s8_z24; break; case GL_DEPTH_COMPONENT16: - rb->_ActualFormat = GL_DEPTH_COMPONENT16; + rb->Format = MESA_FORMAT_Z16; rb->DataType = GL_UNSIGNED_SHORT; - rb->DepthBits = 16; - cpp = 2; - irb->texformat = &_mesa_texformat_z16; break; case GL_DEPTH_COMPONENT: case GL_DEPTH_COMPONENT24: case GL_DEPTH_COMPONENT32: - rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT; + rb->Format = MESA_FORMAT_S8_Z24; rb->DataType = GL_UNSIGNED_INT_24_8_EXT; - rb->DepthBits = 24; - cpp = 4; - irb->texformat = &_mesa_texformat_s8_z24; break; case GL_DEPTH_STENCIL_EXT: case GL_DEPTH24_STENCIL8_EXT: - rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT; + rb->Format = MESA_FORMAT_S8_Z24; rb->DataType = GL_UNSIGNED_INT_24_8_EXT; - rb->DepthBits = 24; - rb->StencilBits = 8; - cpp = 4; - irb->texformat = &_mesa_texformat_s8_z24; break; default: _mesa_problem(ctx, @@ -197,6 +167,9 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, return GL_FALSE; } + rb->_BaseFormat = _mesa_base_fbo_format(ctx, internalFormat); + cpp = _mesa_get_format_bytes(rb->Format); + intelFlush(ctx); /* free old region */ @@ -205,32 +178,25 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, } /* allocate new memory region/renderbuffer */ - if (softwareBuffer) { - return _mesa_soft_renderbuffer_storage(ctx, rb, internalFormat, - width, height); - } - else { - /* Choose a pitch to match hardware requirements: - */ - GLuint pitch = ((cpp * width + 63) & ~63) / cpp; - /* alloc hardware renderbuffer */ - DBG("Allocating %d x %d Intel RBO (pitch %d)\n", width, - height, pitch); + /* Choose a pitch to match hardware requirements: + */ + pitch = ((cpp * width + 63) & ~63) / cpp; - irb->region = intel_region_alloc(intel, I915_TILING_NONE, - cpp, width, height, pitch, - GL_TRUE); - if (!irb->region) - return GL_FALSE; /* out of memory? */ + /* alloc hardware renderbuffer */ + DBG("Allocating %d x %d Intel RBO (pitch %d)\n", width, height, pitch); - ASSERT(irb->region->buffer); + irb->region = intel_region_alloc(intel, I915_TILING_NONE, cpp, + width, height, pitch, GL_TRUE); + if (!irb->region) + return GL_FALSE; /* out of memory? */ - rb->Width = width; - rb->Height = height; + ASSERT(irb->region->buffer); - return GL_TRUE; - } + rb->Width = width; + rb->Height = height; + + return GL_TRUE; } @@ -246,7 +212,7 @@ intel_alloc_window_storage(GLcontext * ctx, struct gl_renderbuffer *rb, ASSERT(rb->Name == 0); rb->Width = width; rb->Height = height; - rb->_ActualFormat = internalFormat; + rb->InternalFormat = internalFormat; return GL_TRUE; } @@ -307,12 +273,11 @@ intel_renderbuffer_set_region(struct intel_renderbuffer *rb, * not a user-created renderbuffer. */ struct intel_renderbuffer * -intel_create_renderbuffer(GLenum intFormat) +intel_create_renderbuffer(gl_format format) { GET_CURRENT_CONTEXT(ctx); struct intel_renderbuffer *irb; - const GLuint name = 0; irb = CALLOC_STRUCT(intel_renderbuffer); if (!irb) { @@ -320,75 +285,43 @@ intel_create_renderbuffer(GLenum intFormat) return NULL; } - _mesa_init_renderbuffer(&irb->Base, name); + _mesa_init_renderbuffer(&irb->Base, 0); irb->Base.ClassID = INTEL_RB_CLASS; - switch (intFormat) { - case GL_RGB5: - irb->Base._ActualFormat = GL_RGB5; - irb->Base._BaseFormat = GL_RGBA; - irb->Base.RedBits = 5; - irb->Base.GreenBits = 6; - irb->Base.BlueBits = 5; + switch (format) { + case MESA_FORMAT_RGB565: + irb->Base._BaseFormat = GL_RGB; irb->Base.DataType = GL_UNSIGNED_BYTE; - irb->texformat = &_mesa_texformat_rgb565; break; - case GL_RGB8: - irb->Base._ActualFormat = GL_RGB8; + case MESA_FORMAT_XRGB8888: irb->Base._BaseFormat = GL_RGB; - irb->Base.RedBits = 8; - irb->Base.GreenBits = 8; - irb->Base.BlueBits = 8; - irb->Base.AlphaBits = 0; irb->Base.DataType = GL_UNSIGNED_BYTE; - irb->texformat = &_mesa_texformat_argb8888; /* XXX: Need xrgb8888 */ break; - case GL_RGBA8: - irb->Base._ActualFormat = GL_RGBA8; + case MESA_FORMAT_ARGB8888: irb->Base._BaseFormat = GL_RGBA; - irb->Base.RedBits = 8; - irb->Base.GreenBits = 8; - irb->Base.BlueBits = 8; - irb->Base.AlphaBits = 8; - irb->Base.DataType = GL_UNSIGNED_BYTE; - irb->texformat = &_mesa_texformat_argb8888; - break; - case GL_STENCIL_INDEX8_EXT: - irb->Base._ActualFormat = GL_STENCIL_INDEX8_EXT; - irb->Base._BaseFormat = GL_STENCIL_INDEX; - irb->Base.StencilBits = 8; irb->Base.DataType = GL_UNSIGNED_BYTE; - irb->texformat = &_mesa_texformat_s8_z24; break; - case GL_DEPTH_COMPONENT16: - irb->Base._ActualFormat = GL_DEPTH_COMPONENT16; + case MESA_FORMAT_Z16: irb->Base._BaseFormat = GL_DEPTH_COMPONENT; - irb->Base.DepthBits = 16; irb->Base.DataType = GL_UNSIGNED_SHORT; - irb->texformat = &_mesa_texformat_z16; break; - case GL_DEPTH_COMPONENT24: - irb->Base._ActualFormat = GL_DEPTH24_STENCIL8_EXT; + case MESA_FORMAT_X8_Z24: irb->Base._BaseFormat = GL_DEPTH_COMPONENT; - irb->Base.DepthBits = 24; irb->Base.DataType = GL_UNSIGNED_INT; - irb->texformat = &_mesa_texformat_s8_z24; break; - case GL_DEPTH24_STENCIL8_EXT: - irb->Base._ActualFormat = GL_DEPTH24_STENCIL8_EXT; - irb->Base._BaseFormat = GL_DEPTH_STENCIL_EXT; - irb->Base.DepthBits = 24; - irb->Base.StencilBits = 8; + case MESA_FORMAT_S8_Z24: + irb->Base._BaseFormat = GL_DEPTH_STENCIL; irb->Base.DataType = GL_UNSIGNED_INT_24_8_EXT; - irb->texformat = &_mesa_texformat_s8_z24; break; default: _mesa_problem(NULL, "Unexpected intFormat in intel_create_renderbuffer"); + _mesa_free(irb); return NULL; } - irb->Base.InternalFormat = intFormat; + irb->Base.Format = format; + irb->Base.InternalFormat = irb->Base._BaseFormat; /* intel-specific methods */ irb->Base.Delete = intel_delete_renderbuffer; @@ -465,59 +398,49 @@ static GLboolean intel_update_wrapper(GLcontext *ctx, struct intel_renderbuffer *irb, struct gl_texture_image *texImage) { - irb->texformat = texImage->TexFormat; + gl_format texFormat; - if (texImage->TexFormat == &_mesa_texformat_argb8888) { - irb->Base._ActualFormat = GL_RGBA8; - irb->Base._BaseFormat = GL_RGBA; + if (texImage->TexFormat == MESA_FORMAT_ARGB8888) { irb->Base.DataType = GL_UNSIGNED_BYTE; DBG("Render to RGBA8 texture OK\n"); } - else if (texImage->TexFormat == &_mesa_texformat_rgb565) { - irb->Base._ActualFormat = GL_RGB5; - irb->Base._BaseFormat = GL_RGB; + else if (texImage->TexFormat == MESA_FORMAT_XRGB8888) { + irb->Base.DataType = GL_UNSIGNED_BYTE; + DBG("Render to XGBA8 texture OK\n"); + } + else if (texImage->TexFormat == MESA_FORMAT_RGB565) { irb->Base.DataType = GL_UNSIGNED_BYTE; DBG("Render to RGB5 texture OK\n"); } - else if (texImage->TexFormat == &_mesa_texformat_argb1555) { - irb->Base._ActualFormat = GL_RGB5_A1; - irb->Base._BaseFormat = GL_RGBA; + else if (texImage->TexFormat == MESA_FORMAT_ARGB1555) { irb->Base.DataType = GL_UNSIGNED_BYTE; DBG("Render to ARGB1555 texture OK\n"); } - else if (texImage->TexFormat == &_mesa_texformat_argb4444) { - irb->Base._ActualFormat = GL_RGBA4; - irb->Base._BaseFormat = GL_RGBA; + else if (texImage->TexFormat == MESA_FORMAT_ARGB4444) { irb->Base.DataType = GL_UNSIGNED_BYTE; DBG("Render to ARGB4444 texture OK\n"); } - else if (texImage->TexFormat == &_mesa_texformat_z16) { - irb->Base._ActualFormat = GL_DEPTH_COMPONENT16; - irb->Base._BaseFormat = GL_DEPTH_COMPONENT; + else if (texImage->TexFormat == MESA_FORMAT_Z16) { irb->Base.DataType = GL_UNSIGNED_SHORT; DBG("Render to DEPTH16 texture OK\n"); } - else if (texImage->TexFormat == &_mesa_texformat_s8_z24) { - irb->Base._ActualFormat = GL_DEPTH24_STENCIL8_EXT; - irb->Base._BaseFormat = GL_DEPTH_STENCIL_EXT; + else if (texImage->TexFormat == MESA_FORMAT_S8_Z24) { irb->Base.DataType = GL_UNSIGNED_INT_24_8_EXT; DBG("Render to DEPTH_STENCIL texture OK\n"); } else { - DBG("Render to texture BAD FORMAT %d\n", - texImage->TexFormat->MesaFormat); + DBG("Render to texture BAD FORMAT %d\n", texImage->TexFormat); return GL_FALSE; } - irb->Base.InternalFormat = irb->Base._ActualFormat; + irb->Base.Format = texImage->TexFormat; + + texFormat = texImage->TexFormat; + + irb->Base.InternalFormat = texImage->InternalFormat; + irb->Base._BaseFormat = _mesa_base_fbo_format(ctx, irb->Base.InternalFormat); irb->Base.Width = texImage->Width; irb->Base.Height = texImage->Height; - irb->Base.RedBits = texImage->TexFormat->RedBits; - irb->Base.GreenBits = texImage->TexFormat->GreenBits; - irb->Base.BlueBits = texImage->TexFormat->BlueBits; - irb->Base.AlphaBits = texImage->TexFormat->AlphaBits; - irb->Base.DepthBits = texImage->TexFormat->DepthBits; - irb->Base.StencilBits = texImage->TexFormat->StencilBits; irb->Base.Delete = intel_delete_renderbuffer; irb->Base.AllocStorage = intel_nop_alloc_storage; @@ -571,7 +494,7 @@ intel_render_texture(GLcontext * ctx, = att->Texture->Image[att->CubeMapFace][att->TextureLevel]; struct intel_renderbuffer *irb = intel_renderbuffer(att->Renderbuffer); struct intel_texture_image *intel_image; - GLuint imageOffset; + GLuint dst_x, dst_y; (void) fb; @@ -618,18 +541,17 @@ intel_render_texture(GLcontext * ctx, } /* compute offset of the particular 2D image within the texture region */ - imageOffset = intel_miptree_image_offset(intel_image->mt, - att->CubeMapFace, - att->TextureLevel); - - if (att->Texture->Target == GL_TEXTURE_3D) { - const GLuint *offsets = intel_miptree_depth_offsets(intel_image->mt, - att->TextureLevel); - imageOffset += offsets[att->Zoffset]; - } - - /* store that offset in the region */ - intel_image->mt->region->draw_offset = imageOffset; + intel_miptree_get_image_offset(intel_image->mt, + att->TextureLevel, + att->CubeMapFace, + att->Zoffset, + &dst_x, &dst_y); + + intel_image->mt->region->draw_offset = (dst_y * intel_image->mt->pitch + + dst_x) * intel_image->mt->cpp; + intel_image->mt->region->draw_x = dst_x; + intel_image->mt->region->draw_y = dst_y; + intel_image->used_as_render_target = GL_TRUE; /* update drawing region, etc */ intel_draw_buffer(ctx, fb); @@ -643,19 +565,23 @@ static void intel_finish_render_texture(GLcontext * ctx, struct gl_renderbuffer_attachment *att) { - /* no-op - * Previously we released the renderbuffer's intel_region but - * that's not necessary and actually caused problems when trying - * to do a glRead/CopyPixels from the renderbuffer later. - * The region will be released later if the texture is replaced - * or the renderbuffer deleted. - * - * The intention of this driver hook is more of a "done rendering - * to texture, please re-twiddle/etc if necessary". + struct intel_context *intel = intel_context(ctx); + struct gl_texture_object *tex_obj = att->Texture; + struct gl_texture_image *image = + tex_obj->Image[att->CubeMapFace][att->TextureLevel]; + struct intel_texture_image *intel_image = intel_texture_image(image); + + /* Flag that this image may now be validated into the object's miptree. */ + intel_image->used_as_render_target = GL_FALSE; + + /* Since we've (probably) rendered to the texture and will (likely) use + * it in the texture domain later on in this batchbuffer, flush the + * batch. Once again, we wish for a domain tracker in libdrm to cover + * usage inside of a batchbuffer like GEM does in the kernel. */ + intel_batchbuffer_emit_mi_flush(intel->batch); } - /** * Do additional "completeness" testing of a framebuffer object. */ @@ -687,8 +613,9 @@ intel_validate_framebuffer(GLcontext *ctx, struct gl_framebuffer *fb) continue; } - switch (irb->texformat->MesaFormat) { + switch (irb->Base.Format) { case MESA_FORMAT_ARGB8888: + case MESA_FORMAT_XRGB8888: case MESA_FORMAT_RGB565: case MESA_FORMAT_ARGB1555: case MESA_FORMAT_ARGB4444: @@ -715,5 +642,5 @@ intel_fbo_init(struct intel_context *intel) intel->ctx.Driver.FinishRenderTexture = intel_finish_render_texture; intel->ctx.Driver.ResizeBuffers = intel_resize_buffers; intel->ctx.Driver.ValidateFramebuffer = intel_validate_framebuffer; - intel->ctx.Driver.BlitFramebuffer = _mesa_meta_blit_framebuffer; + intel->ctx.Driver.BlitFramebuffer = _mesa_meta_BlitFramebuffer; } diff --git a/src/mesa/drivers/dri/intel/intel_fbo.h b/src/mesa/drivers/dri/intel/intel_fbo.h index f0665af482e..fa43077d6a7 100644 --- a/src/mesa/drivers/dri/intel/intel_fbo.h +++ b/src/mesa/drivers/dri/intel/intel_fbo.h @@ -28,6 +28,7 @@ #ifndef INTEL_FBO_H #define INTEL_FBO_H +#include "main/formats.h" #include "intel_screen.h" struct intel_context; @@ -61,8 +62,6 @@ struct intel_renderbuffer struct gl_renderbuffer Base; struct intel_region *region; - const struct gl_texture_format *texformat; - GLuint vbl_pending; /**< vblank sequence number of pending flip */ uint8_t *span_cache; @@ -114,7 +113,7 @@ intel_renderbuffer_set_region(struct intel_renderbuffer *irb, extern struct intel_renderbuffer * -intel_create_renderbuffer(GLenum intFormat); +intel_create_renderbuffer(gl_format format); extern void diff --git a/src/mesa/drivers/dri/intel/intel_generatemipmap.c b/src/mesa/drivers/dri/intel/intel_generatemipmap.c deleted file mode 100644 index 237754d4695..00000000000 --- a/src/mesa/drivers/dri/intel/intel_generatemipmap.c +++ /dev/null @@ -1,279 +0,0 @@ -/* - * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. - * Copyright © 2009 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS - * IN THE SOFTWARE. - * - * Authors: - * Eric Anholt <[email protected]> - * - */ - -#include "main/glheader.h" -#include "main/enums.h" -#include "main/image.h" -#include "main/mtypes.h" -#include "main/macros.h" -#include "main/bufferobj.h" -#include "main/teximage.h" -#include "main/texenv.h" -#include "main/texobj.h" -#include "main/texstate.h" -#include "main/texparam.h" -#include "main/varray.h" -#include "main/attrib.h" -#include "main/enable.h" -#include "main/buffers.h" -#include "main/fbobject.h" -#include "main/framebuffer.h" -#include "main/renderbuffer.h" -#include "main/depth.h" -#include "main/hash.h" -#include "main/mipmap.h" -#include "main/blend.h" -#include "glapi/dispatch.h" -#include "swrast/swrast.h" - -#include "intel_screen.h" -#include "intel_context.h" -#include "intel_batchbuffer.h" -#include "intel_pixel.h" -#include "intel_tex.h" -#include "intel_mipmap_tree.h" - -static const char *intel_fp_tex2d = - "!!ARBfp1.0\n" - "TEX result.color, fragment.texcoord[0], texture[0], 2D;\n" - "END\n"; - -static GLboolean -intel_generate_mipmap_level(GLcontext *ctx, GLuint tex_name, - int level, int width, int height) -{ - struct intel_context *intel = intel_context(ctx); - GLfloat vertices[4][2]; - GLint status; - - /* Set to source from the previous level */ - _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_BASE_LEVEL, level - 1); - _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAX_LEVEL, level - 1); - - /* Set to draw into the current level */ - _mesa_FramebufferTexture2DEXT(GL_FRAMEBUFFER_EXT, - GL_COLOR_ATTACHMENT0_EXT, - GL_TEXTURE_2D, - tex_name, - level); - /* Choose to render to the color attachment. */ - _mesa_DrawBuffer(GL_COLOR_ATTACHMENT0_EXT); - - status = _mesa_CheckFramebufferStatusEXT (GL_FRAMEBUFFER_EXT); - if (status != GL_FRAMEBUFFER_COMPLETE_EXT) - return GL_FALSE; - - meta_set_passthrough_transform(&intel->meta); - - /* XXX: Doing it right would involve setting up the transformation to do - * 0-1 mapping or something, and not changing the vertex data. - */ - vertices[0][0] = 0; - vertices[0][1] = 0; - vertices[1][0] = width; - vertices[1][1] = 0; - vertices[2][0] = width; - vertices[2][1] = height; - vertices[3][0] = 0; - vertices[3][1] = height; - - _mesa_VertexPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &vertices); - _mesa_Enable(GL_VERTEX_ARRAY); - meta_set_default_texrect(&intel->meta); - - _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4); - - meta_restore_texcoords(&intel->meta); - meta_restore_transform(&intel->meta); - - return GL_TRUE; -} - -static GLboolean -intel_generate_mipmap_2d(GLcontext *ctx, - GLenum target, - struct gl_texture_object *texObj) -{ - struct intel_context *intel = intel_context(ctx); - GLint old_active_texture; - int level, max_levels, start_level, end_level; - GLuint fb_name; - GLboolean success = GL_FALSE; - struct gl_framebuffer *saved_fbo = NULL; - - _mesa_PushAttrib(GL_ENABLE_BIT | GL_TEXTURE_BIT | - GL_CURRENT_BIT | GL_COLOR_BUFFER_BIT | - GL_DEPTH_BUFFER_BIT); - _mesa_PushClientAttrib(GL_CLIENT_VERTEX_ARRAY_BIT); - old_active_texture = ctx->Texture.CurrentUnit; - _mesa_reference_framebuffer(&saved_fbo, ctx->DrawBuffer); - - _mesa_Disable(GL_POLYGON_STIPPLE); - _mesa_Disable(GL_DEPTH_TEST); - _mesa_Disable(GL_STENCIL_TEST); - _mesa_ColorMask(GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE); - _mesa_DepthMask(GL_FALSE); - - /* Bind the given texture to GL_TEXTURE_2D with linear filtering for our - * minification. - */ - _mesa_ActiveTextureARB(GL_TEXTURE0_ARB); - _mesa_Enable(GL_TEXTURE_2D); - _mesa_BindTexture(GL_TEXTURE_2D, texObj->Name); - _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, - GL_LINEAR_MIPMAP_NEAREST); - _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR); - - /* Bind the new renderbuffer to the color attachment point. */ - _mesa_GenFramebuffersEXT(1, &fb_name); - _mesa_BindFramebufferEXT(GL_FRAMEBUFFER_EXT, fb_name); - - meta_set_fragment_program(&intel->meta, &intel->meta.tex2d_fp, - intel_fp_tex2d); - meta_set_passthrough_vertex_program(&intel->meta); - - max_levels = _mesa_max_texture_levels(ctx, texObj->Target); - start_level = texObj->BaseLevel; - end_level = texObj->MaxLevel; - - /* Loop generating level+1 from level. */ - for (level = start_level; level < end_level && level < max_levels - 1; level++) { - const struct gl_texture_image *srcImage; - int width, height; - - srcImage = _mesa_select_tex_image(ctx, texObj, target, level); - if (srcImage->Border != 0) - goto fail; - - width = srcImage->Width / 2; - if (width < 1) - width = 1; - height = srcImage->Height / 2; - if (height < 1) - height = 1; - - if (width == srcImage->Width && - height == srcImage->Height) { - /* Neither _mesa_max_texture_levels nor texObj->MaxLevel are the - * maximum texture level for the object, so break out when we've gone - * over the edge. - */ - break; - } - - /* Make sure that there's space allocated for the target level. - * We could skip this if there's already space allocated and save some - * time. - */ - _mesa_TexImage2D(GL_TEXTURE_2D, level + 1, srcImage->InternalFormat, - width, height, 0, - GL_RGBA, GL_UNSIGNED_INT, NULL); - - if (!intel_generate_mipmap_level(ctx, texObj->Name, level + 1, - width, height)) - goto fail; - } - - success = GL_TRUE; - -fail: - meta_restore_fragment_program(&intel->meta); - meta_restore_vertex_program(&intel->meta); - - _mesa_DeleteFramebuffersEXT(1, &fb_name); - _mesa_ActiveTextureARB(GL_TEXTURE0_ARB + old_active_texture); - if (saved_fbo) - _mesa_BindFramebufferEXT(GL_FRAMEBUFFER_EXT, saved_fbo->Name); - _mesa_reference_framebuffer(&saved_fbo, NULL); - _mesa_PopClientAttrib(); - _mesa_PopAttrib(); - - return success; -} - - -/** - * Generate new mipmap data from BASE+1 to BASE+p (the minimally-sized mipmap - * level). - * - * The texture object's miptree must be mapped. - * - * This function should also include an accelerated path. - */ -void -intel_generate_mipmap(GLcontext *ctx, GLenum target, - struct gl_texture_object *texObj) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_texture_object *intelObj = intel_texture_object(texObj); - GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; - int face, i; - - /* HW path */ - if (target == GL_TEXTURE_2D && - ctx->Extensions.EXT_framebuffer_object && - ctx->Extensions.ARB_fragment_program && - ctx->Extensions.ARB_vertex_program) { - GLboolean success; - - /* We'll be accessing this texture using GL entrypoints, which should - * be resilient against other access to this texture. - */ - _mesa_unlock_texture(ctx, texObj); - success = intel_generate_mipmap_2d(ctx, target, texObj); - _mesa_lock_texture(ctx, texObj); - - if (success) - return; - } - - /* SW path */ - intel_tex_map_level_images(intel, intelObj, texObj->BaseLevel); - _mesa_generate_mipmap(ctx, target, texObj); - intel_tex_unmap_level_images(intel, intelObj, texObj->BaseLevel); - - /* Update the level information in our private data in the new images, since - * it didn't get set as part of a normal TexImage path. - */ - for (face = 0; face < nr_faces; face++) { - for (i = texObj->BaseLevel + 1; i < texObj->MaxLevel; i++) { - struct intel_texture_image *intelImage; - - intelImage = intel_texture_image(texObj->Image[face][i]); - if (intelImage == NULL) - break; - - intelImage->level = i; - intelImage->face = face; - /* Unreference the miptree to signal that the new Data is a bare - * pointer from mesa. - */ - intel_miptree_release(intel, &intelImage->mt); - } - } -} diff --git a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c index c985da5aa25..6a565f80cf4 100644 --- a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c +++ b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c @@ -28,11 +28,16 @@ #include "intel_context.h" #include "intel_mipmap_tree.h" #include "intel_regions.h" +#include "intel_tex_layout.h" #include "intel_chipset.h" +#ifndef I915 +#include "brw_state.h" +#endif #include "main/enums.h" #define FILE_DEBUG_FLAG DEBUG_MIPTREE + static GLenum target_to_target(GLenum target) { @@ -49,6 +54,7 @@ target_to_target(GLenum target) } } + static struct intel_mipmap_tree * intel_miptree_create_internal(struct intel_context *intel, GLenum target, @@ -98,6 +104,7 @@ intel_miptree_create_internal(struct intel_context *intel, return mt; } + struct intel_mipmap_tree * intel_miptree_create(struct intel_context *intel, GLenum target, @@ -115,7 +122,7 @@ intel_miptree_create(struct intel_context *intel, if (intel->use_texture_tiling && compress_byte == 0 && intel->intelScreen->kernel_exec_fencing) { - if (IS_965(intel->intelScreen->deviceID) && + if (intel->gen >= 4 && (base_format == GL_DEPTH_COMPONENT || base_format == GL_DEPTH_STENCIL_EXT)) tiling = I915_TILING_Y; @@ -131,8 +138,10 @@ intel_miptree_create(struct intel_context *intel, /* * pitch == 0 || height == 0 indicates the null texture */ - if (!mt || !mt->pitch || !mt->total_height) + if (!mt || !mt->pitch || !mt->total_height) { + free(mt); return NULL; + } mt->region = intel_region_alloc(intel, tiling, @@ -150,6 +159,7 @@ intel_miptree_create(struct intel_context *intel, return mt; } + struct intel_mipmap_tree * intel_miptree_create_for_region(struct intel_context *intel, GLenum target, @@ -187,7 +197,8 @@ intel_miptree_create_for_region(struct intel_context *intel, intel_region_reference(&mt->region, region); return mt; - } +} + /** * intel_miptree_pitch_align: @@ -201,7 +212,6 @@ intel_miptree_create_for_region(struct intel_context *intel, * Given @pitch, compute a larger value which accounts for * any necessary alignment required by the device */ - int intel_miptree_pitch_align (struct intel_context *intel, struct intel_mipmap_tree *mt, uint32_t tiling, @@ -214,16 +224,12 @@ int intel_miptree_pitch_align (struct intel_context *intel, if (!mt->compressed) { int pitch_align; - if (intel->ttm) { - /* XXX: Align pitch to multiple of 64 bytes for now to allow - * render-to-texture to work in all cases. This should probably be - * replaced at some point by some scheme to only do this when really - * necessary. - */ - pitch_align = 64; - } else { - pitch_align = 4; - } + /* XXX: Align pitch to multiple of 64 bytes for now to allow + * render-to-texture to work in all cases. This should probably be + * replaced at some point by some scheme to only do this when really + * necessary. + */ + pitch_align = 64; if (tiling == I915_TILING_X) pitch_align = 512; @@ -247,6 +253,7 @@ int intel_miptree_pitch_align (struct intel_context *intel, return pitch; } + void intel_miptree_reference(struct intel_mipmap_tree **dst, struct intel_mipmap_tree *src) @@ -256,6 +263,7 @@ intel_miptree_reference(struct intel_mipmap_tree **dst, DBG("%s %p refcount now %d\n", __FUNCTION__, src, src->refcount); } + void intel_miptree_release(struct intel_context *intel, struct intel_mipmap_tree **mt) @@ -269,11 +277,25 @@ intel_miptree_release(struct intel_context *intel, DBG("%s deleting %p\n", __FUNCTION__, *mt); +#ifndef I915 + /* Free up cached binding tables holding a reference on our buffer, to + * avoid excessive memory consumption. + * + * This isn't as aggressive as we could be, as we'd like to do + * it from any time we free the last ref on a region. But intel_region.c + * is context-agnostic. Perhaps our constant state cache should be, as + * well. + */ + brw_state_cache_bo_delete(&brw_context(&intel->ctx)->surface_cache, + (*mt)->region->buffer); +#endif + intel_region_release(&((*mt)->region)); - for (i = 0; i < MAX_TEXTURE_LEVELS; i++) - if ((*mt)->level[i].image_offset) - free((*mt)->level[i].image_offset); + for (i = 0; i < MAX_TEXTURE_LEVELS; i++) { + free((*mt)->level[i].x_offset); + free((*mt)->level[i].y_offset); + } free(*mt); } @@ -281,33 +303,31 @@ intel_miptree_release(struct intel_context *intel, } - - -/* Can the image be pulled into a unified mipmap tree. This mirrors +/** + * Can the image be pulled into a unified mipmap tree? This mirrors * the completeness test in a lot of ways. * * Not sure whether I want to pass gl_texture_image here. */ GLboolean intel_miptree_match_image(struct intel_mipmap_tree *mt, - struct gl_texture_image *image, - GLuint face, GLuint level) + struct gl_texture_image *image) { - /* Images with borders are never pulled into mipmap trees. - */ - if (image->Border || - ((image->_BaseFormat == GL_DEPTH_COMPONENT) && - ((image->TexObject->WrapS == GL_CLAMP_TO_BORDER) || - (image->TexObject->WrapT == GL_CLAMP_TO_BORDER)))) + GLboolean isCompressed = _mesa_is_format_compressed(image->TexFormat); + struct intel_texture_image *intelImage = intel_texture_image(image); + GLuint level = intelImage->level; + + /* Images with borders are never pulled into mipmap trees. */ + if (image->Border) return GL_FALSE; if (image->InternalFormat != mt->internal_format || - image->IsCompressed != mt->compressed) + isCompressed != mt->compressed) return GL_FALSE; - if (!image->IsCompressed && + if (!isCompressed && !mt->compressed && - image->TexFormat->TexelBytes != mt->cpp) + _mesa_get_format_bytes(image->TexFormat) != mt->cpp) return GL_FALSE; /* Test image dimensions against the base level image adjusted for @@ -334,82 +354,59 @@ intel_miptree_set_level_info(struct intel_mipmap_tree *mt, mt->level[level].height = h; mt->level[level].depth = d; mt->level[level].level_offset = (x + y * mt->pitch) * mt->cpp; + mt->level[level].level_x = x; + mt->level[level].level_y = y; mt->level[level].nr_images = nr_images; DBG("%s level %d size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, level, w, h, d, x, y, mt->level[level].level_offset); - /* Not sure when this would happen, but anyway: - */ - if (mt->level[level].image_offset) { - free(mt->level[level].image_offset); - mt->level[level].image_offset = NULL; - } - assert(nr_images); + assert(!mt->level[level].x_offset); - mt->level[level].image_offset = malloc(nr_images * sizeof(GLuint)); - mt->level[level].image_offset[0] = 0; + mt->level[level].x_offset = malloc(nr_images * sizeof(GLuint)); + mt->level[level].x_offset[0] = mt->level[level].level_x; + mt->level[level].y_offset = malloc(nr_images * sizeof(GLuint)); + mt->level[level].y_offset[0] = mt->level[level].level_y; } void -intel_miptree_set_image_offset_ex(struct intel_mipmap_tree *mt, - GLuint level, GLuint img, - GLuint x, GLuint y, - GLuint offset) -{ - if (img == 0 && level == 0) - assert(x == 0 && y == 0); - - assert(img < mt->level[level].nr_images); - - mt->level[level].image_offset[img] = (x + y * mt->pitch) * mt->cpp + offset; - - DBG("%s level %d img %d pos %d,%d image_offset %x\n", - __FUNCTION__, level, img, x, y, mt->level[level].image_offset[img]); -} - -void intel_miptree_set_image_offset(struct intel_mipmap_tree *mt, GLuint level, GLuint img, GLuint x, GLuint y) { - intel_miptree_set_image_offset_ex(mt, level, img, x, y, 0); -} + if (img == 0 && level == 0) + assert(x == 0 && y == 0); + assert(img < mt->level[level].nr_images); -/* Although we use the image_offset[] array to store relative offsets - * to cube faces, Mesa doesn't know anything about this and expects - * each cube face to be treated as a separate image. - * - * These functions present that view to mesa: - */ -const GLuint * -intel_miptree_depth_offsets(struct intel_mipmap_tree *mt, GLuint level) -{ - static const GLuint zero = 0; + mt->level[level].x_offset[img] = mt->level[level].level_x + x; + mt->level[level].y_offset[img] = mt->level[level].level_y + y; - if (mt->target != GL_TEXTURE_3D || mt->level[level].nr_images == 1) - return &zero; - else - return mt->level[level].image_offset; + DBG("%s level %d img %d pos %d,%d\n", + __FUNCTION__, level, img, + mt->level[level].x_offset[img], mt->level[level].y_offset[img]); } -GLuint -intel_miptree_image_offset(struct intel_mipmap_tree *mt, - GLuint face, GLuint level) +void +intel_miptree_get_image_offset(struct intel_mipmap_tree *mt, + GLuint level, GLuint face, GLuint depth, + GLuint *x, GLuint *y) { - if (mt->target == GL_TEXTURE_CUBE_MAP_ARB) - return (mt->level[level].level_offset + - mt->level[level].image_offset[face]); - else - return mt->level[level].level_offset; + if (mt->target == GL_TEXTURE_CUBE_MAP_ARB) { + *x = mt->level[level].x_offset[face]; + *y = mt->level[level].y_offset[face]; + } else if (mt->target == GL_TEXTURE_3D) { + *x = mt->level[level].x_offset[depth]; + *y = mt->level[level].y_offset[depth]; + } else { + *x = mt->level[level].x_offset[0]; + *y = mt->level[level].y_offset[0]; + } } - - /** * Map a teximage in a mipmap tree. * \param row_stride returns row stride in bytes @@ -425,6 +422,7 @@ intel_miptree_image_map(struct intel_context * intel, GLuint level, GLuint * row_stride, GLuint * image_offsets) { + GLuint x, y; DBG("%s \n", __FUNCTION__); if (row_stride) @@ -433,19 +431,26 @@ intel_miptree_image_map(struct intel_context * intel, if (mt->target == GL_TEXTURE_3D) { int i; - for (i = 0; i < mt->level[level].depth; i++) - image_offsets[i] = mt->level[level].image_offset[i] / mt->cpp; + for (i = 0; i < mt->level[level].depth; i++) { + + intel_miptree_get_image_offset(mt, level, face, i, + &x, &y); + image_offsets[i] = x + y * mt->pitch; + } + + return intel_region_map(intel, mt->region); } else { assert(mt->level[level].depth == 1); - assert(mt->target == GL_TEXTURE_CUBE_MAP || - mt->level[level].image_offset[0] == 0); + intel_miptree_get_image_offset(mt, level, face, 0, + &x, &y); image_offsets[0] = 0; - } - return (intel_region_map(intel, mt->region) + - intel_miptree_image_offset(mt, face, level)); + return intel_region_map(intel, mt->region) + + (x + y * mt->pitch) * mt->cpp; + } } + void intel_miptree_image_unmap(struct intel_context *intel, struct intel_mipmap_tree *mt) @@ -455,8 +460,8 @@ intel_miptree_image_unmap(struct intel_context *intel, } - -/* Upload data for a particular image. +/** + * Upload data for a particular image. */ void intel_miptree_image_data(struct intel_context *intel, @@ -467,21 +472,21 @@ intel_miptree_image_data(struct intel_context *intel, GLuint src_row_pitch, GLuint src_image_pitch) { - GLuint depth = dst->level[level].depth; - GLuint dst_offset = intel_miptree_image_offset(dst, face, level); - const GLuint *dst_depth_offset = intel_miptree_depth_offsets(dst, level); + const GLuint depth = dst->level[level].depth; GLuint i; - GLuint height = 0; DBG("%s: %d/%d\n", __FUNCTION__, face, level); for (i = 0; i < depth; i++) { + GLuint dst_x, dst_y, height; + + intel_miptree_get_image_offset(dst, level, face, i, &dst_x, &dst_y); + height = dst->level[level].height; if(dst->compressed) height = (height + 3) / 4; + intel_region_data(intel, - dst->region, - dst_offset + dst_depth_offset[i], /* dst_offset */ - 0, 0, /* dstx, dsty */ + dst->region, 0, dst_x, dst_y, src, src_row_pitch, 0, 0, /* source x, y */ @@ -491,8 +496,9 @@ intel_miptree_image_data(struct intel_context *intel, } } -extern void intel_get_texture_alignment_unit(GLenum, GLuint *, GLuint *); -/* Copy mipmap image between trees + +/** + * Copy mipmap image between trees */ void intel_miptree_image_copy(struct intel_context *intel, @@ -503,38 +509,37 @@ intel_miptree_image_copy(struct intel_context *intel, GLuint width = src->level[level].width; GLuint height = src->level[level].height; GLuint depth = src->level[level].depth; - GLuint dst_offset = intel_miptree_image_offset(dst, face, level); - GLuint src_offset = intel_miptree_image_offset(src, face, level); - const GLuint *dst_depth_offset = intel_miptree_depth_offsets(dst, level); - const GLuint *src_depth_offset = intel_miptree_depth_offsets(src, level); + GLuint src_x, src_y, dst_x, dst_y; GLuint i; GLboolean success; if (dst->compressed) { GLuint align_w, align_h; - intel_get_texture_alignment_unit(dst->internal_format, &align_w, &align_h); + intel_get_texture_alignment_unit(dst->internal_format, + &align_w, &align_h); height = (height + 3) / 4; width = ALIGN(width, align_w); } for (i = 0; i < depth; i++) { + intel_miptree_get_image_offset(src, level, face, i, &src_x, &src_y); + intel_miptree_get_image_offset(dst, level, face, i, &dst_x, &dst_y); success = intel_region_copy(intel, - dst->region, dst_offset + dst_depth_offset[i], - 0, 0, - src->region, src_offset + src_depth_offset[i], - 0, 0, width, height, GL_COPY); + dst->region, 0, dst_x, dst_y, + src->region, 0, src_x, src_y, width, height, + GL_COPY); if (!success) { GLubyte *src_ptr, *dst_ptr; src_ptr = intel_region_map(intel, src->region); dst_ptr = intel_region_map(intel, dst->region); - _mesa_copy_rect(dst_ptr + dst_offset + dst_depth_offset[i], + _mesa_copy_rect(dst_ptr + dst->cpp * (dst_x + dst_y * dst->pitch), dst->cpp, dst->pitch, 0, 0, width, height, - src_ptr + src_offset + src_depth_offset[i], + src_ptr + src->cpp * (src_x + src_y * src->pitch), src->pitch, 0, 0); intel_region_unmap(intel, src->region); diff --git a/src/mesa/drivers/dri/intel/intel_mipmap_tree.h b/src/mesa/drivers/dri/intel/intel_mipmap_tree.h index c890b2a0d04..b19c548def3 100644 --- a/src/mesa/drivers/dri/intel/intel_mipmap_tree.h +++ b/src/mesa/drivers/dri/intel/intel_mipmap_tree.h @@ -70,6 +70,10 @@ struct intel_mipmap_level * always zero in that case. */ GLuint level_offset; + /** Offset to this miptree level, used in computing x_offset. */ + GLuint level_x; + /** Offset to this miptree level, used in computing y_offset. */ + GLuint level_y; GLuint width; GLuint height; /** Depth of the mipmap at this level: 1 for 1D/2D/CUBE, n for 3D. */ @@ -86,7 +90,7 @@ struct intel_mipmap_level * compute the offsets of depth/cube images within a mipmap level, * so have to store them as a lookup table. */ - GLuint *image_offset; + GLuint *x_offset, *y_offset; }; struct intel_mipmap_tree @@ -161,8 +165,7 @@ void intel_miptree_release(struct intel_context *intel, /* Check if an image fits an existing mipmap tree layout */ GLboolean intel_miptree_match_image(struct intel_mipmap_tree *mt, - struct gl_texture_image *image, - GLuint face, GLuint level); + struct gl_texture_image *image); /* Return a pointer to an image within a tree. Return image stride as * well. @@ -176,19 +179,10 @@ GLubyte *intel_miptree_image_map(struct intel_context *intel, void intel_miptree_image_unmap(struct intel_context *intel, struct intel_mipmap_tree *mt); - -/* Return the linear offset of an image relative to the start of the - * tree: - */ -GLuint intel_miptree_image_offset(struct intel_mipmap_tree *mt, - GLuint face, GLuint level); - -/* Return pointers to each 2d slice within an image. Indexed by depth - * value. - */ -const GLuint *intel_miptree_depth_offsets(struct intel_mipmap_tree *mt, - GLuint level); - +void +intel_miptree_get_image_offset(struct intel_mipmap_tree *mt, + GLuint level, GLuint face, GLuint depth, + GLuint *x, GLuint *y); void intel_miptree_set_level_info(struct intel_mipmap_tree *mt, GLuint level, @@ -196,16 +190,10 @@ void intel_miptree_set_level_info(struct intel_mipmap_tree *mt, GLuint x, GLuint y, GLuint w, GLuint h, GLuint d); -void intel_miptree_set_image_offset_ex(struct intel_mipmap_tree *mt, - GLuint level, - GLuint img, GLuint x, GLuint y, - GLuint offset); - void intel_miptree_set_image_offset(struct intel_mipmap_tree *mt, GLuint level, GLuint img, GLuint x, GLuint y); - /* Upload an image into a tree */ void intel_miptree_image_data(struct intel_context *intel, diff --git a/src/mesa/drivers/dri/intel/intel_pixel.c b/src/mesa/drivers/dri/intel/intel_pixel.c index a3001416559..993e427a992 100644 --- a/src/mesa/drivers/dri/intel/intel_pixel.c +++ b/src/mesa/drivers/dri/intel/intel_pixel.c @@ -129,20 +129,6 @@ intel_check_blit_fragment_ops(GLcontext * ctx, GLboolean src_alpha_is_one) return GL_TRUE; } - -GLboolean -intel_check_meta_tex_fragment_ops(GLcontext * ctx) -{ - if (ctx->NewState) - _mesa_update_state(ctx); - - /* Some of _ImageTransferState (scale, bias) could be done with - * fragment programs on i915. - */ - return !(ctx->_ImageTransferState || ctx->Fog.Enabled || /* not done yet */ - ctx->Texture._EnabledUnits || ctx->FragmentProgram._Enabled); -} - /* The intel_region struct doesn't really do enough to capture the * format of the pixels in the region. For now this code assumes that * the region is a display surface and hence is either ARGB8888 or diff --git a/src/mesa/drivers/dri/intel/intel_pixel.h b/src/mesa/drivers/dri/intel/intel_pixel.h index 96a6dd17b25..743b6497c52 100644 --- a/src/mesa/drivers/dri/intel/intel_pixel.h +++ b/src/mesa/drivers/dri/intel/intel_pixel.h @@ -34,8 +34,6 @@ void intelInitPixelFuncs(struct dd_function_table *functions); GLboolean intel_check_blit_fragment_ops(GLcontext * ctx, GLboolean src_alpha_is_one); -GLboolean intel_check_meta_tex_fragment_ops(GLcontext * ctx); - GLboolean intel_check_blit_format(struct intel_region *region, GLenum format, GLenum type); diff --git a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c index 9a0bcc07a5b..668697cb5eb 100644 --- a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c +++ b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c @@ -33,6 +33,7 @@ #include "main/macros.h" #include "main/bufferobj.h" #include "main/pixelstore.h" +#include "main/polygon.h" #include "main/state.h" #include "main/teximage.h" #include "main/texenv.h" @@ -227,10 +228,9 @@ do_blit_bitmap( GLcontext *ctx, UNCLAMPED_FLOAT_TO_UBYTE(ubcolor[3], tmpColor[3]); if (dst->cpp == 2) - color = INTEL_PACKCOLOR565(ubcolor[0], ubcolor[1], ubcolor[2]); + color = PACK_COLOR_565(ubcolor[0], ubcolor[1], ubcolor[2]); else - color = INTEL_PACKCOLOR8888(ubcolor[0], ubcolor[1], - ubcolor[2], ubcolor[3]); + color = PACK_COLOR_8888(ubcolor[3], ubcolor[0], ubcolor[1], ubcolor[2]); if (!intel_check_blit_fragment_ops(ctx, tmpColor[3] == 1.0F)) return GL_FALSE; @@ -335,6 +335,8 @@ out: unpack->BufferObj); } + intel_check_front_buffer_rendering(intel); + return GL_TRUE; } @@ -435,13 +437,14 @@ intel_texture_bitmap(GLcontext * ctx, } /* Save GL state before we start setting up our drawing */ - _mesa_PushAttrib(GL_ENABLE_BIT | GL_CURRENT_BIT | - GL_VIEWPORT_BIT); + _mesa_PushAttrib(GL_ENABLE_BIT | GL_CURRENT_BIT | GL_POLYGON_BIT | + GL_TEXTURE_BIT | GL_VIEWPORT_BIT); _mesa_PushClientAttrib(GL_CLIENT_VERTEX_ARRAY_BIT | GL_CLIENT_PIXEL_STORE_BIT); old_active_texture = ctx->Texture.CurrentUnit; _mesa_Disable(GL_POLYGON_STIPPLE); + _mesa_PolygonMode(GL_FRONT_AND_BACK, GL_FILL); /* Upload our bitmap data to an alpha texture */ _mesa_ActiveTextureARB(GL_TEXTURE0_ARB); @@ -501,8 +504,6 @@ intel_texture_bitmap(GLcontext * ctx, meta_restore_vertex_program(&intel->meta); _mesa_PopClientAttrib(); - _mesa_Disable(GL_TEXTURE_2D); /* asserted that it was disabled at entry */ - _mesa_ActiveTextureARB(GL_TEXTURE0_ARB + old_active_texture); _mesa_PopAttrib(); _mesa_DeleteTextures(1, &texname); diff --git a/src/mesa/drivers/dri/intel/intel_pixel_copy.c b/src/mesa/drivers/dri/intel/intel_pixel_copy.c index 07ca8f7ddb5..622aaa22d67 100644 --- a/src/mesa/drivers/dri/intel/intel_pixel_copy.c +++ b/src/mesa/drivers/dri/intel/intel_pixel_copy.c @@ -222,6 +222,8 @@ do_blit_copypixels(GLcontext * ctx, out: UNLOCK_HARDWARE(intel); + intel_check_front_buffer_rendering(intel); + DBG("%s: success\n", __FUNCTION__); return GL_TRUE; } @@ -240,5 +242,5 @@ intelCopyPixels(GLcontext * ctx, return; /* this will use swrast if needed */ - _mesa_meta_copy_pixels(ctx, srcx, srcy, width, height, destx, desty, type); + _mesa_meta_CopyPixels(ctx, srcx, srcy, width, height, destx, desty, type); } diff --git a/src/mesa/drivers/dri/intel/intel_pixel_draw.c b/src/mesa/drivers/dri/intel/intel_pixel_draw.c index 7fbb89fd6aa..9b382e36225 100644 --- a/src/mesa/drivers/dri/intel/intel_pixel_draw.c +++ b/src/mesa/drivers/dri/intel/intel_pixel_draw.c @@ -54,7 +54,7 @@ #include "intel_fbo.h" -/** XXX compare perf of this vs. _mesa_meta_draw_pixels(STENCIL) */ +/** XXX compare perf of this vs. _mesa_meta_DrawPixels(STENCIL) */ static GLboolean intel_stencil_drawpixels(GLcontext * ctx, GLint x, GLint y, @@ -169,7 +169,7 @@ intel_stencil_drawpixels(GLcontext * ctx, * buffer. */ depth_irb = intel_get_renderbuffer(ctx->DrawBuffer, BUFFER_DEPTH); - irb = intel_create_renderbuffer(GL_RGBA8); + irb = intel_create_renderbuffer(MESA_FORMAT_ARGB8888); rb = &irb->Base; irb->Base.Width = depth_irb->Base.Width; irb->Base.Height = depth_irb->Base.Height; @@ -265,7 +265,7 @@ intelDrawPixels(GLcontext * ctx, /* XXX this function doesn't seem to work reliably even when all * the pre-requisite conditions are met. * Note that this function is never hit with conform. - * Fall back to swrast because even the _mesa_meta_draw_pixels() approach + * Fall back to swrast because even the _mesa_meta_DrawPixels() approach * isn't working because of an apparent stencil bug. */ if (intel_stencil_drawpixels(ctx, x, y, width, height, format, type, @@ -280,6 +280,6 @@ intelDrawPixels(GLcontext * ctx, } #endif - _mesa_meta_draw_pixels(ctx, x, y, width, height, format, type, - unpack, pixels); + _mesa_meta_DrawPixels(ctx, x, y, width, height, format, type, + unpack, pixels); } diff --git a/src/mesa/drivers/dri/intel/intel_pixel_read.c b/src/mesa/drivers/dri/intel/intel_pixel_read.c index bc67f6242a1..20424e2e589 100644 --- a/src/mesa/drivers/dri/intel/intel_pixel_read.c +++ b/src/mesa/drivers/dri/intel/intel_pixel_read.c @@ -216,9 +216,8 @@ do_blit_readpixels(GLcontext * ctx, rowLength = -rowLength; } - /* XXX 64-bit cast? */ - dst_offset = (GLuint) _mesa_image_address(2, pack, pixels, width, height, - format, type, 0, 0, 0); + dst_offset = (GLintptr) _mesa_image_address(2, pack, pixels, width, height, + format, type, 0, 0, 0); /* Although the blits go on the command buffer, need to do this and @@ -227,14 +226,14 @@ do_blit_readpixels(GLcontext * ctx, intelFlush(&intel->ctx); LOCK_HARDWARE(intel); - if (intel->driDrawable->numClipRects) { + if (intel->driReadDrawable->numClipRects) { GLboolean all = (width * height * src->cpp == dst->Base.Size && x == 0 && dst_offset == 0); dri_bo *dst_buffer = intel_bufferobj_buffer(intel, dst, all ? INTEL_WRITE_FULL : INTEL_WRITE_PART); - __DRIdrawablePrivate *dPriv = intel->driDrawable; + __DRIdrawablePrivate *dPriv = intel->driReadDrawable; int nbox = dPriv->numClipRects; drm_clip_rect_t *box = dPriv->pClipRects; drm_clip_rect_t rect; @@ -286,11 +285,11 @@ intelReadPixels(GLcontext * ctx, intelFlush(ctx); -#ifdef I915 if (do_blit_readpixels (ctx, x, y, width, height, format, type, pack, pixels)) return; +#ifdef I915 if (do_texture_readpixels (ctx, x, y, width, height, format, type, pack, pixels)) return; diff --git a/src/mesa/drivers/dri/intel/intel_regions.c b/src/mesa/drivers/dri/intel/intel_regions.c index a86c66a8443..d6b9dc44466 100644 --- a/src/mesa/drivers/dri/intel/intel_regions.c +++ b/src/mesa/drivers/dri/intel/intel_regions.c @@ -542,56 +542,18 @@ intel_recreate_static(struct intel_context *intel, region->buffer = NULL; } - if (intel->ttm) { - assert(region_desc->bo_handle != -1); - region->buffer = intel_bo_gem_create_from_name(intel->bufmgr, - name, - region_desc->bo_handle); - - ret = dri_bo_get_tiling(region->buffer, ®ion->tiling, - ®ion->bit_6_swizzle); - if (ret != 0) { - fprintf(stderr, "Couldn't get tiling of buffer %d (%s): %s\n", - region_desc->bo_handle, name, strerror(-ret)); - intel_region_release(®ion); - return NULL; - } - } else { - if (region->classic_map != NULL) { - drmUnmap(region->classic_map, - region->pitch * region->cpp * region->height); - region->classic_map = NULL; - } - ret = drmMap(intel->driFd, region_desc->handle, - region->pitch * region->cpp * region->height, - ®ion->classic_map); - if (ret != 0) { - fprintf(stderr, "Failed to drmMap %s buffer\n", name); - free(region); - return NULL; - } - - region->buffer = intel_bo_fake_alloc_static(intel->bufmgr, + assert(region_desc->bo_handle != -1); + region->buffer = intel_bo_gem_create_from_name(intel->bufmgr, name, - region_desc->offset, - region->pitch * region->cpp * - region->height, - region->classic_map); - - /* The sarea just gives us a boolean for whether it's tiled or not, - * instead of which tiling mode it is. Guess. - */ - if (region_desc->tiled) { - if (IS_965(intel->intelScreen->deviceID) && - region_desc == &intelScreen->depth) - region->tiling = I915_TILING_Y; - else - region->tiling = I915_TILING_X; - } else { - region->tiling = I915_TILING_NONE; - } + region_desc->bo_handle); - region->bit_6_swizzle = I915_BIT_6_SWIZZLE_NONE; + ret = dri_bo_get_tiling(region->buffer, ®ion->tiling, + ®ion->bit_6_swizzle); + if (ret != 0) { + fprintf(stderr, "Couldn't get tiling of buffer %d (%s): %s\n", + region_desc->bo_handle, name, strerror(-ret)); + intel_region_release(®ion); + return NULL; } assert(region->buffer != NULL); diff --git a/src/mesa/drivers/dri/intel/intel_regions.h b/src/mesa/drivers/dri/intel/intel_regions.h index 0d379bdc6e2..535fcd7be0b 100644 --- a/src/mesa/drivers/dri/intel/intel_regions.h +++ b/src/mesa/drivers/dri/intel/intel_regions.h @@ -62,6 +62,8 @@ struct intel_region GLuint map_refcount; /**< Reference count for mapping */ GLuint draw_offset; /**< Offset of drawing address within the region */ + GLuint draw_x, draw_y; /**< Offset of drawing within the region */ + uint32_t tiling; /**< Which tiling mode the region is in */ uint32_t bit_6_swizzle; /**< GEM flag for address swizzling requirement */ drmAddress classic_map; /**< drmMap of the region when not in GEM mode */ diff --git a/src/mesa/drivers/dri/intel/intel_screen.c b/src/mesa/drivers/dri/intel/intel_screen.c index 1b8c56e68d6..2c5a884a9b4 100644 --- a/src/mesa/drivers/dri/intel/intel_screen.c +++ b/src/mesa/drivers/dri/intel/intel_screen.c @@ -79,6 +79,10 @@ PUBLIC const char __driConfigOptions[] = DRI_CONF_DESC(en, "Enable early Z in classic mode (unstable, 945-only).") DRI_CONF_OPT_END + DRI_CONF_OPT_BEGIN(fragment_shader, bool, false) + DRI_CONF_DESC(en, "Enable limited ARB_fragment_shader support on 915/945.") + DRI_CONF_OPT_END + DRI_CONF_SECTION_END DRI_CONF_SECTION_QUALITY DRI_CONF_FORCE_S3TC_ENABLE(false) @@ -88,10 +92,14 @@ PUBLIC const char __driConfigOptions[] = DRI_CONF_NO_RAST(false) DRI_CONF_ALWAYS_FLUSH_BATCH(false) DRI_CONF_ALWAYS_FLUSH_CACHE(false) + + DRI_CONF_OPT_BEGIN(stub_occlusion_query, bool, false) + DRI_CONF_DESC(en, "Enable stub ARB_occlusion_query support on 915/945.") + DRI_CONF_OPT_END DRI_CONF_SECTION_END DRI_CONF_END; -const GLuint __driNConfigOptions = 10; +const GLuint __driNConfigOptions = 12; #ifdef USE_NEW_INTERFACE static PFNGLXCREATECONTEXTMODES create_context_modes = NULL; @@ -341,7 +349,7 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, else { GLboolean swStencil = (mesaVis->stencilBits > 0 && mesaVis->depthBits != 24); - GLenum rgbFormat; + gl_format rgbFormat; struct intel_framebuffer *intel_fb = CALLOC_STRUCT(intel_framebuffer); @@ -351,11 +359,11 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, _mesa_initialize_framebuffer(&intel_fb->Base, mesaVis); if (mesaVis->redBits == 5) - rgbFormat = GL_RGB5; + rgbFormat = MESA_FORMAT_RGB565; else if (mesaVis->alphaBits == 0) - rgbFormat = GL_RGB8; + rgbFormat = MESA_FORMAT_XRGB8888; else - rgbFormat = GL_RGBA8; + rgbFormat = MESA_FORMAT_ARGB8888; /* setup the hardware-based renderbuffers */ intel_fb->color_rb[0] = intel_create_renderbuffer(rgbFormat); @@ -374,7 +382,7 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, if (mesaVis->stencilBits == 8) { /* combined depth/stencil buffer */ struct intel_renderbuffer *depthStencilRb - = intel_create_renderbuffer(GL_DEPTH24_STENCIL8_EXT); + = intel_create_renderbuffer(MESA_FORMAT_S8_Z24); /* note: bind RB to two attachment points */ _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, &depthStencilRb->Base); @@ -382,7 +390,7 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, &depthStencilRb->Base); } else { struct intel_renderbuffer *depthRb - = intel_create_renderbuffer(GL_DEPTH_COMPONENT24); + = intel_create_renderbuffer(MESA_FORMAT_X8_Z24); _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, &depthRb->Base); } @@ -390,7 +398,7 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, else if (mesaVis->depthBits == 16) { /* just 16-bit depth buffer, no hw stencil */ struct intel_renderbuffer *depthRb - = intel_create_renderbuffer(GL_DEPTH_COMPONENT16); + = intel_create_renderbuffer(MESA_FORMAT_Z16); _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, &depthRb->Base); } @@ -597,7 +605,6 @@ intelFillInModes(__DRIscreenPrivate *psp, static GLboolean intel_init_bufmgr(intelScreenPrivate *intelScreen) { - GLboolean gem_disable = getenv("INTEL_NO_GEM") != NULL; int gem_kernel = 0; GLboolean gem_supported; struct drm_i915_getparam gp; @@ -614,43 +621,24 @@ intel_init_bufmgr(intelScreenPrivate *intelScreen) /* If we've got a new enough DDX that's initializing GEM and giving us * object handles for the shared buffers, use that. */ - intelScreen->ttm = GL_FALSE; if (intelScreen->driScrnPriv->dri2.enabled) gem_supported = GL_TRUE; else if (intelScreen->driScrnPriv->ddx_version.minor >= 9 && gem_kernel && intelScreen->front.bo_handle != -1) gem_supported = GL_TRUE; - else - gem_supported = GL_FALSE; - - if (!gem_disable && gem_supported) { - intelScreen->bufmgr = intel_bufmgr_gem_init(spriv->fd, BATCH_SZ); - if (intelScreen->bufmgr != NULL) - intelScreen->ttm = GL_TRUE; + else { + fprintf(stderr, "[%s:%u] Error initializing GEM.\n", + __func__, __LINE__); + return GL_FALSE; } + + intelScreen->bufmgr = intel_bufmgr_gem_init(spriv->fd, BATCH_SZ); /* Otherwise, use the classic buffer manager. */ if (intelScreen->bufmgr == NULL) { - if (gem_disable) { - _mesa_warning(NULL, "GEM disabled. Using classic."); - } else { - _mesa_warning(NULL, - "Failed to initialize GEM. Falling back to classic."); - } - - if (intelScreen->tex.size == 0) { - fprintf(stderr, "[%s:%u] Error initializing buffer manager.\n", - __func__, __LINE__); - return GL_FALSE; - } - - intelScreen->bufmgr = - intel_bufmgr_fake_init(spriv->fd, - intelScreen->tex.offset, - intelScreen->tex.map, - intelScreen->tex.size, - (unsigned int * volatile) - &intelScreen->sarea->last_dispatch); + fprintf(stderr, "[%s:%u] Error initializing buffer manager.\n", + __func__, __LINE__); + return GL_FALSE; } if (intel_get_param(spriv, I915_PARAM_NUM_FENCES_AVAIL, &num_fences)) @@ -688,18 +676,6 @@ static const __DRIconfig **intelInitScreen(__DRIscreenPrivate *psp) return NULL; } - /* Calling driInitExtensions here, with a NULL context pointer, - * does not actually enable the extensions. It just makes sure - * that all the dispatch offsets for all the extensions that - * *might* be enables are known. This is needed because the - * dispatch offsets need to be known when _mesa_context_create is - * called, but we can't enable the extensions until we have a - * context pointer. - * - * Hello chicken. Hello egg. How are you two today? - */ - intelInitExtensions(NULL, GL_TRUE); - if (!intelInitDriver(psp)) return NULL; @@ -752,18 +728,6 @@ __DRIconfig **intelInitScreen2(__DRIscreenPrivate *psp) int color; __DRIconfig **configs = NULL; - /* Calling driInitExtensions here, with a NULL context pointer, - * does not actually enable the extensions. It just makes sure - * that all the dispatch offsets for all the extensions that - * *might* be enables are known. This is needed because the - * dispatch offsets need to be known when _mesa_context_create is - * called, but we can't enable the extensions until we have a - * context pointer. - * - * Hello chicken. Hello egg. How are you two today? - */ - intelInitExtensions(NULL, GL_TRUE); - /* Allocate the private area */ intelScreen = (intelScreenPrivate *) CALLOC(sizeof(intelScreenPrivate)); if (!intelScreen) { diff --git a/src/mesa/drivers/dri/intel/intel_screen.h b/src/mesa/drivers/dri/intel/intel_screen.h index a9b9e109a6a..14ca0903b6c 100644 --- a/src/mesa/drivers/dri/intel/intel_screen.h +++ b/src/mesa/drivers/dri/intel/intel_screen.h @@ -77,7 +77,6 @@ typedef struct GLboolean no_hw; GLboolean no_vbo; - int ttm; dri_bufmgr *bufmgr; GLboolean kernel_exec_fencing; diff --git a/src/mesa/drivers/dri/intel/intel_span.c b/src/mesa/drivers/dri/intel/intel_span.c index 28eabbc0054..d1681e9088a 100644 --- a/src/mesa/drivers/dri/intel/intel_span.c +++ b/src/mesa/drivers/dri/intel/intel_span.c @@ -29,7 +29,6 @@ #include "main/macros.h" #include "main/mtypes.h" #include "main/colormac.h" -#include "main/texformat.h" #include "intel_buffers.h" #include "intel_fbo.h" @@ -132,18 +131,6 @@ pwrite_8(struct intel_renderbuffer *irb, uint32_t offset, uint8_t val) dri_bo_subdata(irb->region->buffer, offset, 1, &val); } -static uint32_t -z24s8_to_s8z24(uint32_t val) -{ - return (val << 24) | (val >> 8); -} - -static uint32_t -s8z24_to_z24s8(uint32_t val) -{ - return (val >> 24) | (val << 8); -} - static uint32_t no_tile_swizzle(struct intel_renderbuffer *irb, int x, int y) { @@ -163,6 +150,9 @@ static uint32_t x_tile_swizzle(struct intel_renderbuffer *irb, int x_tile_number, y_tile_number; int tile_off, tile_base; + x += irb->region->draw_x; + y += irb->region->draw_y; + tile_stride = (irb->region->pitch * irb->region->cpp) << 3; xbyte = x * irb->region->cpp; @@ -218,6 +208,9 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb, int x_tile_number, y_tile_number; int tile_off, tile_base; + x += irb->region->draw_x; + y += irb->region->draw_y; + tile_stride = (irb->region->pitch * irb->region->cpp) << 5; xbyte = x * irb->region->cpp; @@ -273,8 +266,11 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb, unsigned int num_cliprects; \ struct drm_clip_rect *cliprects; \ int x_off, y_off; \ + int pitch = irb->region->pitch * irb->region->cpp; \ + void *buf = irb->region->buffer->virtual; \ GLuint p; \ (void) p; \ + (void)buf; (void)pitch; /* unused for non-gttmap. */ \ intel_get_cliprects(intel, &cliprects, &num_cliprects, &x_off, &y_off); /* XXX FBO: this is identical to the macro in spantmp2.h except we get @@ -296,7 +292,6 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb, #define Y_FLIP(_y) ((_y) * yScale + yBias) -/* XXX with GEM, these need to tell the kernel */ #define HW_LOCK() #define HW_UNLOCK() @@ -339,7 +334,7 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb, #include "intel_spantmp.h" /* x8r8g8b8 color span and pixel functions */ -#define INTEL_PIXEL_FMT GL_BGRA +#define INTEL_PIXEL_FMT GL_BGR #define INTEL_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV #define INTEL_READ_VALUE(offset) pread_xrgb8888(irb, offset) #define INTEL_WRITE_VALUE(offset, v) pwrite_xrgb8888(irb, offset, v) @@ -354,6 +349,9 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb, unsigned int num_cliprects; \ struct drm_clip_rect *cliprects; \ int x_off, y_off; \ + int pitch = irb->region->pitch * irb->region->cpp; \ + void *buf = irb->region->buffer->virtual; \ + (void)buf; (void)pitch; /* unused for non-gttmap. */ \ intel_get_cliprects(intel, &cliprects, &num_cliprects, &x_off, &y_off); @@ -366,20 +364,22 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb, #define INTEL_TAG(name) name##_z16 #include "intel_depthtmp.h" -/* z24 depthbuffer functions. */ +/* z24x8 depthbuffer functions. */ #define INTEL_VALUE_TYPE GLuint #define INTEL_WRITE_DEPTH(offset, d) pwrite_32(irb, offset, d) #define INTEL_READ_DEPTH(offset) pread_32(irb, offset) -#define INTEL_TAG(name) name##_z24 +#define INTEL_TAG(name) name##_z24_x8 #include "intel_depthtmp.h" -/* z24s8 depthbuffer functions. */ -#define INTEL_VALUE_TYPE GLuint -#define INTEL_WRITE_DEPTH(offset, d) pwrite_32(irb, offset, z24s8_to_s8z24(d)) -#define INTEL_READ_DEPTH(offset) s8z24_to_z24s8(pread_32(irb, offset)) -#define INTEL_TAG(name) name##_z24_s8 -#include "intel_depthtmp.h" +/** + ** 8-bit stencil function (XXX FBO: This is obsolete) + **/ +/* XXX */ +#define WRITE_STENCIL(_x, _y, d) pwrite_8(irb, NO_TILE(_x, _y) + 3, d) +#define READ_STENCIL(d, _x, _y) d = pread_8(irb, NO_TILE(_x, _y) + 3); +#define TAG(x) intel_gttmap_##x##_z24_s8 +#include "stenciltmp.h" /** ** 8-bit stencil function (XXX FBO: This is obsolete) @@ -413,6 +413,9 @@ intel_renderbuffer_map(struct intel_context *intel, struct gl_renderbuffer *rb) if (irb == NULL || irb->region == NULL) return; + if (intel->intelScreen->kernel_exec_fencing) + drm_intel_gem_bo_map_gtt(irb->region->buffer); + intel_set_span_functions(intel, rb); } @@ -425,7 +428,10 @@ intel_renderbuffer_unmap(struct intel_context *intel, if (irb == NULL || irb->region == NULL) return; - clear_span_cache(irb); + if (intel->intelScreen->kernel_exec_fencing) + drm_intel_gem_bo_unmap_gtt(irb->region->buffer); + else + clear_span_cache(irb); rb->GetRow = NULL; rb->PutRow = NULL; @@ -494,6 +500,8 @@ intel_map_unmap_framebuffer(struct intel_context *intel, else intel_renderbuffer_unmap(intel, fb->_StencilBuffer->Wrapped); } + + intel_check_front_buffer_rendering(intel); } /** @@ -558,6 +566,43 @@ intelInitSpanFuncs(GLcontext * ctx) swdd->SpanRenderFinish = intelSpanRenderFinish; } +void +intel_map_vertex_shader_textures(GLcontext *ctx) +{ + struct intel_context *intel = intel_context(ctx); + int i; + + if (ctx->VertexProgram._Current == NULL) + return; + + for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) { + if (ctx->Texture.Unit[i]._ReallyEnabled && + ctx->VertexProgram._Current->Base.TexturesUsed[i] != 0) { + struct gl_texture_object *texObj = ctx->Texture.Unit[i]._Current; + + intel_tex_map_images(intel, intel_texture_object(texObj)); + } + } +} + +void +intel_unmap_vertex_shader_textures(GLcontext *ctx) +{ + struct intel_context *intel = intel_context(ctx); + int i; + + if (ctx->VertexProgram._Current == NULL) + return; + + for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) { + if (ctx->Texture.Unit[i]._ReallyEnabled && + ctx->VertexProgram._Current->Base.TexturesUsed[i] != 0) { + struct gl_texture_object *texObj = ctx->Texture.Unit[i]._Current; + + intel_tex_unmap_images(intel, intel_texture_object(texObj)); + } + } +} /** * Plug in appropriate span read/write functions for the given renderbuffer. @@ -568,17 +613,56 @@ intel_set_span_functions(struct intel_context *intel, struct gl_renderbuffer *rb) { struct intel_renderbuffer *irb = (struct intel_renderbuffer *) rb; - uint32_t tiling; + uint32_t tiling = irb->region->tiling; + + if (intel->intelScreen->kernel_exec_fencing) { + switch (irb->Base.Format) { + case MESA_FORMAT_RGB565: + intel_gttmap_InitPointers_RGB565(rb); + break; + case MESA_FORMAT_ARGB4444: + intel_gttmap_InitPointers_ARGB4444(rb); + break; + case MESA_FORMAT_ARGB1555: + intel_gttmap_InitPointers_ARGB1555(rb); + break; + case MESA_FORMAT_XRGB8888: + intel_gttmap_InitPointers_xRGB8888(rb); + break; + case MESA_FORMAT_ARGB8888: + intel_gttmap_InitPointers_ARGB8888(rb); + break; + case MESA_FORMAT_Z16: + intel_gttmap_InitDepthPointers_z16(rb); + break; + case MESA_FORMAT_X8_Z24: + intel_gttmap_InitDepthPointers_z24_x8(rb); + break; + case MESA_FORMAT_S8_Z24: + /* There are a few different ways SW asks us to access the S8Z24 data: + * Z24 depth-only depth reads + * S8Z24 depth reads + * S8Z24 stencil reads. + */ + if (rb->Format == MESA_FORMAT_S8_Z24) { + intel_gttmap_InitDepthPointers_z24_x8(rb); + } else if (rb->Format == MESA_FORMAT_S8) { + intel_gttmap_InitStencilPointers_z24_s8(rb); + } + break; + default: + _mesa_problem(NULL, + "Unexpected MesaFormat %d in intelSetSpanFunctions", + irb->Base.Format); + break; + } + return; + } /* If in GEM mode, we need to do the tile address swizzling ourselves, * instead of the fence registers handling it. */ - if (intel->ttm) - tiling = irb->region->tiling; - else - tiling = I915_TILING_NONE; - - switch (irb->texformat->MesaFormat) { + switch (irb->Base.Format) { case MESA_FORMAT_RGB565: switch (tiling) { case I915_TILING_NONE: @@ -621,35 +705,33 @@ intel_set_span_functions(struct intel_context *intel, break; } break; + case MESA_FORMAT_XRGB8888: + switch (tiling) { + case I915_TILING_NONE: + default: + intelInitPointers_xRGB8888(rb); + break; + case I915_TILING_X: + intel_XTile_InitPointers_xRGB8888(rb); + break; + case I915_TILING_Y: + intel_YTile_InitPointers_xRGB8888(rb); + break; + } + break; case MESA_FORMAT_ARGB8888: - if (rb->AlphaBits == 0) { /* XXX: Need xRGB8888 Mesa format */ - /* 8888 RGBx */ - switch (tiling) { - case I915_TILING_NONE: - default: - intelInitPointers_xRGB8888(rb); - break; - case I915_TILING_X: - intel_XTile_InitPointers_xRGB8888(rb); - break; - case I915_TILING_Y: - intel_YTile_InitPointers_xRGB8888(rb); - break; - } - } else { - /* 8888 RGBA */ - switch (tiling) { - case I915_TILING_NONE: - default: - intelInitPointers_ARGB8888(rb); - break; - case I915_TILING_X: - intel_XTile_InitPointers_ARGB8888(rb); - break; - case I915_TILING_Y: - intel_YTile_InitPointers_ARGB8888(rb); - break; - } + /* 8888 RGBA */ + switch (tiling) { + case I915_TILING_NONE: + default: + intelInitPointers_ARGB8888(rb); + break; + case I915_TILING_X: + intel_XTile_InitPointers_ARGB8888(rb); + break; + case I915_TILING_Y: + intel_YTile_InitPointers_ARGB8888(rb); + break; } break; case MESA_FORMAT_Z16: @@ -666,39 +748,27 @@ intel_set_span_functions(struct intel_context *intel, break; } break; + case MESA_FORMAT_X8_Z24: case MESA_FORMAT_S8_Z24: /* There are a few different ways SW asks us to access the S8Z24 data: * Z24 depth-only depth reads * S8Z24 depth reads * S8Z24 stencil reads. */ - if (rb->_ActualFormat == GL_DEPTH_COMPONENT24) { - switch (tiling) { - case I915_TILING_NONE: - default: - intelInitDepthPointers_z24(rb); - break; - case I915_TILING_X: - intel_XTile_InitDepthPointers_z24(rb); - break; - case I915_TILING_Y: - intel_YTile_InitDepthPointers_z24(rb); - break; - } - } else if (rb->_ActualFormat == GL_DEPTH24_STENCIL8_EXT) { + if (rb->Format == MESA_FORMAT_S8_Z24) { switch (tiling) { case I915_TILING_NONE: default: - intelInitDepthPointers_z24_s8(rb); + intelInitDepthPointers_z24_x8(rb); break; case I915_TILING_X: - intel_XTile_InitDepthPointers_z24_s8(rb); + intel_XTile_InitDepthPointers_z24_x8(rb); break; case I915_TILING_Y: - intel_YTile_InitDepthPointers_z24_s8(rb); + intel_YTile_InitDepthPointers_z24_x8(rb); break; } - } else if (rb->_ActualFormat == GL_STENCIL_INDEX8_EXT) { + } else if (rb->Format == MESA_FORMAT_S8) { switch (tiling) { case I915_TILING_NONE: default: diff --git a/src/mesa/drivers/dri/intel/intel_span.h b/src/mesa/drivers/dri/intel/intel_span.h index acbeb4abe1c..bffe109aa5b 100644 --- a/src/mesa/drivers/dri/intel/intel_span.h +++ b/src/mesa/drivers/dri/intel/intel_span.h @@ -36,5 +36,7 @@ void intel_renderbuffer_map(struct intel_context *intel, struct gl_renderbuffer *rb); void intel_renderbuffer_unmap(struct intel_context *intel, struct gl_renderbuffer *rb); +void intel_map_vertex_shader_textures(GLcontext *ctx); +void intel_unmap_vertex_shader_textures(GLcontext *ctx); #endif diff --git a/src/mesa/drivers/dri/intel/intel_spantmp.h b/src/mesa/drivers/dri/intel/intel_spantmp.h index ead0b1c1683..bad03398f68 100644 --- a/src/mesa/drivers/dri/intel/intel_spantmp.h +++ b/src/mesa/drivers/dri/intel/intel_spantmp.h @@ -32,6 +32,12 @@ #define SPANTMP_PIXEL_FMT INTEL_PIXEL_FMT #define SPANTMP_PIXEL_TYPE INTEL_PIXEL_TYPE +#define TAG(x) INTEL_TAG(intel_gttmap_##x) +#define TAG2(x, y) INTEL_TAG(intel_gttmap_##x##y) +#include "spantmp2.h" + +#define SPANTMP_PIXEL_FMT INTEL_PIXEL_FMT +#define SPANTMP_PIXEL_TYPE INTEL_PIXEL_TYPE #define PUT_VALUE(_x, _y, v) INTEL_WRITE_VALUE(NO_TILE(_x, _y), v) #define GET_VALUE(_x, _y) INTEL_READ_VALUE(NO_TILE(_x, _y)) #define TAG(x) INTEL_TAG(intel##x) @@ -48,8 +54,8 @@ #define SPANTMP_PIXEL_FMT INTEL_PIXEL_FMT #define SPANTMP_PIXEL_TYPE INTEL_PIXEL_TYPE -#define PUT_VALUE(_x, _y, v) INTEL_WRITE_VALUE(X_TILE(_x, _y), v) -#define GET_VALUE(_x, _y) INTEL_READ_VALUE(X_TILE(_x, _y)) +#define PUT_VALUE(_x, _y, v) INTEL_WRITE_VALUE(Y_TILE(_x, _y), v) +#define GET_VALUE(_x, _y) INTEL_READ_VALUE(Y_TILE(_x, _y)) #define TAG(x) INTEL_TAG(intel_YTile_##x) #define TAG2(x, y) INTEL_TAG(intel_YTile_##x)##y #include "spantmp2.h" diff --git a/src/mesa/drivers/dri/intel/intel_state.c b/src/mesa/drivers/dri/intel/intel_state.c index 4ee742377d6..aefae53eb20 100644 --- a/src/mesa/drivers/dri/intel/intel_state.c +++ b/src/mesa/drivers/dri/intel/intel_state.c @@ -196,25 +196,6 @@ intel_translate_logic_op(GLenum opcode) } } - -static void -intelClearColor(GLcontext *ctx, const GLfloat color[4]) -{ - struct intel_context *intel = intel_context(ctx); - GLubyte clear[4]; - - CLAMPED_FLOAT_TO_UBYTE(clear[0], color[0]); - CLAMPED_FLOAT_TO_UBYTE(clear[1], color[1]); - CLAMPED_FLOAT_TO_UBYTE(clear[2], color[2]); - CLAMPED_FLOAT_TO_UBYTE(clear[3], color[3]); - - /* compute both 32 and 16-bit clear values */ - intel->ClearColor8888 = INTEL_PACKCOLOR8888(clear[0], clear[1], - clear[2], clear[3]); - intel->ClearColor565 = INTEL_PACKCOLOR565(clear[0], clear[1], clear[2]); -} - - /* Fallback to swrast for select and feedback. */ static void @@ -229,5 +210,4 @@ void intelInitStateFuncs(struct dd_function_table *functions) { functions->RenderMode = intelRenderMode; - functions->ClearColor = intelClearColor; } diff --git a/src/mesa/drivers/dri/intel/intel_syncobj.c b/src/mesa/drivers/dri/intel/intel_syncobj.c index 1286fe929bc..0d7889d3c25 100644 --- a/src/mesa/drivers/dri/intel/intel_syncobj.c +++ b/src/mesa/drivers/dri/intel/intel_syncobj.c @@ -114,7 +114,7 @@ static void intel_check_sync(GLcontext *ctx, struct gl_sync_object *s) { struct intel_sync_object *sync = (struct intel_sync_object *)s; - if (sync->bo && drm_intel_bo_busy(sync->bo)) { + if (sync->bo && !drm_intel_bo_busy(sync->bo)) { drm_intel_bo_unreference(sync->bo); sync->bo = NULL; s->StatusFlag = 1; diff --git a/src/mesa/drivers/dri/intel/intel_tex.c b/src/mesa/drivers/dri/intel/intel_tex.c index df63f29a42c..215a534a5c5 100644 --- a/src/mesa/drivers/dri/intel/intel_tex.c +++ b/src/mesa/drivers/dri/intel/intel_tex.c @@ -2,6 +2,7 @@ #include "main/texobj.h" #include "main/teximage.h" #include "main/mipmap.h" +#include "drivers/common/meta.h" #include "intel_context.h" #include "intel_mipmap_tree.h" #include "intel_tex.h" @@ -158,11 +159,58 @@ timed_memcpy(void *dest, const void *src, size_t n) } #endif /* DO_DEBUG */ + +/** + * Called via ctx->Driver.GenerateMipmap() + * This is basically a wrapper for _mesa_meta_GenerateMipmap() which checks + * if we'll be using software mipmap generation. In that case, we need to + * map/unmap the base level texture image. + */ +static void +intelGenerateMipmap(GLcontext *ctx, GLenum target, + struct gl_texture_object *texObj) +{ + if (_mesa_meta_check_generate_mipmap_fallback(ctx, target, texObj)) { + /* sw path: need to map texture images */ + struct intel_context *intel = intel_context(ctx); + struct intel_texture_object *intelObj = intel_texture_object(texObj); + intel_tex_map_level_images(intel, intelObj, texObj->BaseLevel); + _mesa_generate_mipmap(ctx, target, texObj); + intel_tex_unmap_level_images(intel, intelObj, texObj->BaseLevel); + + { + GLuint nr_faces = (texObj->Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; + GLuint face, i; + /* Update the level information in our private data in the new images, + * since it didn't get set as part of a normal TexImage path. + */ + for (face = 0; face < nr_faces; face++) { + for (i = texObj->BaseLevel + 1; i < texObj->MaxLevel; i++) { + struct intel_texture_image *intelImage = + intel_texture_image(texObj->Image[face][i]); + if (!intelImage) + break; + intelImage->level = i; + intelImage->face = face; + /* Unreference the miptree to signal that the new Data is a + * bare pointer from mesa. + */ + intel_miptree_release(intel, &intelImage->mt); + } + } + } + } + else { + _mesa_meta_GenerateMipmap(ctx, target, texObj); + } +} + + void intelInitTextureFuncs(struct dd_function_table *functions) { functions->ChooseTextureFormat = intelChooseTextureFormat; - functions->GenerateMipmap = intel_generate_mipmap; + functions->GenerateMipmap = intelGenerateMipmap; functions->NewTextureObject = intelNewTextureObject; functions->NewTextureImage = intelNewTextureImage; diff --git a/src/mesa/drivers/dri/intel/intel_tex.h b/src/mesa/drivers/dri/intel/intel_tex.h index 471aa2a240b..f3cc0fff5c8 100644 --- a/src/mesa/drivers/dri/intel/intel_tex.h +++ b/src/mesa/drivers/dri/intel/intel_tex.h @@ -29,6 +29,7 @@ #define INTELTEX_INC #include "main/mtypes.h" +#include "main/formats.h" #include "intel_context.h" #include "texmem.h" @@ -41,10 +42,8 @@ void intelInitTextureSubImageFuncs(struct dd_function_table *functions); void intelInitTextureCopyImageFuncs(struct dd_function_table *functions); -const struct gl_texture_format *intelChooseTextureFormat(GLcontext * ctx, - GLint internalFormat, - GLenum format, - GLenum type); +gl_format intelChooseTextureFormat(GLcontext *ctx, GLint internalFormat, + GLenum format, GLenum type); void intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname, unsigned long long offset, GLint depth, GLuint pitch); @@ -71,7 +70,4 @@ void intel_tex_unmap_images(struct intel_context *intel, int intel_compressed_num_bytes(GLuint mesaFormat); -void intel_generate_mipmap(GLcontext *ctx, GLenum target, - struct gl_texture_object *texObj); - #endif diff --git a/src/mesa/drivers/dri/intel/intel_tex_copy.c b/src/mesa/drivers/dri/intel/intel_tex_copy.c index 148e8c048c7..767d04d2f4b 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_copy.c +++ b/src/mesa/drivers/dri/intel/intel_tex_copy.c @@ -29,8 +29,10 @@ #include "main/enums.h" #include "main/image.h" #include "main/teximage.h" +#include "main/texstate.h" #include "main/mipmap.h" -#include "swrast/swrast.h" + +#include "drivers/common/meta.h" #include "intel_screen.h" #include "intel_context.h" @@ -74,6 +76,7 @@ get_teximage_source(struct intel_context *intel, GLenum internalFormat) case GL_RGBA: case GL_RGBA8: case GL_RGB: + case GL_RGB8: return intel_readbuf_region(intel); default: return NULL; @@ -90,8 +93,7 @@ do_copy_texsubimage(struct intel_context *intel, GLint x, GLint y, GLsizei width, GLsizei height) { GLcontext *ctx = &intel->ctx; - const struct intel_region *src = - get_teximage_source(intel, internalFormat); + const struct intel_region *src = get_teximage_source(intel, internalFormat); if (!intelImage->mt || !src) { if (INTEL_DEBUG & DEBUG_FALLBACKS) @@ -107,33 +109,36 @@ do_copy_texsubimage(struct intel_context *intel, return GL_FALSE; } - intelFlush(ctx); + // intelFlush(ctx); LOCK_HARDWARE(intel); { drm_intel_bo *dst_bo = intel_region_buffer(intel, intelImage->mt->region, INTEL_WRITE_PART); - GLuint image_offset = intel_miptree_image_offset(intelImage->mt, - intelImage->face, - intelImage->level); const GLint orig_x = x; const GLint orig_y = y; + GLuint image_x, image_y; GLshort src_pitch; + /* get dest x/y in destination texture */ + intel_miptree_get_image_offset(intelImage->mt, + intelImage->level, + intelImage->face, + 0, + &image_x, &image_y); /* Update dst for clipped src. Need to also clip the source rect. */ dstx += x - orig_x; dsty += y - orig_y; /* Can't blit to tiled buffers with non-tile-aligned offset. */ - if (intelImage->mt->region->tiling != I915_TILING_NONE && - (image_offset & 4095) != 0) { + if (intelImage->mt->region->tiling == I915_TILING_Y) { UNLOCK_HARDWARE(intel); return GL_FALSE; } if (ctx->ReadBuffer->Name == 0) { /* reading from a window, adjust x, y */ - __DRIdrawablePrivate *dPriv = intel->driDrawable; + const __DRIdrawablePrivate *dPriv = intel->driReadDrawable; y = dPriv->y + (dPriv->h - (y + height)); x += dPriv->x; @@ -150,17 +155,19 @@ do_copy_texsubimage(struct intel_context *intel, src_pitch = src->pitch; } + /* blit from src buffer to texture */ if (!intelEmitCopyBlit(intel, intelImage->mt->cpp, src_pitch, src->buffer, - 0, + src->draw_offset, src->tiling, intelImage->mt->pitch, dst_bo, - image_offset, + 0, intelImage->mt->region->tiling, - x, y, dstx, dsty, width, height, + x, y, image_x + dstx, image_y + dsty, + width, height, GL_COPY)) { UNLOCK_HARDWARE(intel); return GL_FALSE; @@ -178,8 +185,7 @@ intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, GLenum internalFormat, GLint x, GLint y, GLsizei width, GLint border) { - struct gl_texture_unit *texUnit = - &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + struct gl_texture_unit *texUnit = _mesa_get_current_tex_unit(ctx); struct gl_texture_object *texObj = _mesa_select_tex_object(ctx, texUnit, target); struct gl_texture_image *texImage = @@ -215,8 +221,10 @@ intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, return; fail: - _swrast_copy_teximage1d(ctx, target, level, internalFormat, x, y, - width, border); + if (INTEL_DEBUG & DEBUG_FALLBACKS) + fprintf(stderr, "%s - fallback to swrast\n", __FUNCTION__); + _mesa_meta_CopyTexImage1D(ctx, target, level, internalFormat, x, y, + width, border); } @@ -226,8 +234,7 @@ intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, GLint x, GLint y, GLsizei width, GLsizei height, GLint border) { - struct gl_texture_unit *texUnit = - &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + struct gl_texture_unit *texUnit = _mesa_get_current_tex_unit(ctx); struct gl_texture_object *texObj = _mesa_select_tex_object(ctx, texUnit, target); struct gl_texture_image *texImage = @@ -242,7 +249,7 @@ intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, */ ctx->Driver.TexImage2D(ctx, target, level, internalFormat, width, height, border, - GL_RGBA, CHAN_TYPE, NULL, + GL_RGBA, GL_UNSIGNED_BYTE, NULL, &ctx->DefaultPacking, texObj, texImage); srcx = x; @@ -263,8 +270,10 @@ intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, return; fail: - _swrast_copy_teximage2d(ctx, target, level, internalFormat, x, y, - width, height, border); + if (INTEL_DEBUG & DEBUG_FALLBACKS) + fprintf(stderr, "%s - fallback to swrast\n", __FUNCTION__); + _mesa_meta_CopyTexImage2D(ctx, target, level, internalFormat, x, y, + width, height, border); } @@ -272,8 +281,7 @@ static void intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, GLint xoffset, GLint x, GLint y, GLsizei width) { - struct gl_texture_unit *texUnit = - &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + struct gl_texture_unit *texUnit = _mesa_get_current_tex_unit(ctx); struct gl_texture_object *texObj = _mesa_select_tex_object(ctx, texUnit, target); struct gl_texture_image *texImage = @@ -288,7 +296,9 @@ intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, if (!do_copy_texsubimage(intel_context(ctx), target, intel_texture_image(texImage), internalFormat, xoffset, 0, x, y, width, 1)) { - _swrast_copy_texsubimage1d(ctx, target, level, xoffset, x, y, width); + if (INTEL_DEBUG & DEBUG_FALLBACKS) + fprintf(stderr, "%s - fallback to swrast\n", __FUNCTION__); + _mesa_meta_CopyTexSubImage1D(ctx, target, level, xoffset, x, y, width); } } @@ -298,8 +308,7 @@ intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint x, GLint y, GLsizei width, GLsizei height) { - struct gl_texture_unit *texUnit = - &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + struct gl_texture_unit *texUnit = _mesa_get_current_tex_unit(ctx); struct gl_texture_object *texObj = _mesa_select_tex_object(ctx, texUnit, target); struct gl_texture_image *texImage = @@ -314,10 +323,10 @@ intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, internalFormat, xoffset, yoffset, x, y, width, height)) { - DBG("%s - fallback to swrast\n", __FUNCTION__); - - _swrast_copy_texsubimage2d(ctx, target, level, - xoffset, yoffset, x, y, width, height); + if (INTEL_DEBUG & DEBUG_FALLBACKS) + fprintf(stderr, "%s - fallback to swrast\n", __FUNCTION__); + _mesa_meta_CopyTexSubImage2D(ctx, target, level, + xoffset, yoffset, x, y, width, height); } } diff --git a/src/mesa/drivers/dri/intel/intel_tex_format.c b/src/mesa/drivers/dri/intel/intel_tex_format.c index 3322a711307..87efb72cc51 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_format.c +++ b/src/mesa/drivers/dri/intel/intel_tex_format.c @@ -1,7 +1,6 @@ #include "intel_context.h" #include "intel_tex.h" #include "intel_chipset.h" -#include "main/texformat.h" #include "main/enums.h" @@ -16,7 +15,7 @@ * these if we take the step of simply swizzling the colors * immediately after sampling... */ -const struct gl_texture_format * +gl_format intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, GLenum format, GLenum type) { @@ -34,48 +33,48 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, case GL_COMPRESSED_RGBA: if (format == GL_BGRA) { if (type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV) { - return &_mesa_texformat_argb8888; + return MESA_FORMAT_ARGB8888; } else if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV) { - return &_mesa_texformat_argb4444; + return MESA_FORMAT_ARGB4444; } else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV) { - return &_mesa_texformat_argb1555; + return MESA_FORMAT_ARGB1555; } } - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + return do32bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB4444; case 3: case GL_RGB: case GL_COMPRESSED_RGB: if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) { - return &_mesa_texformat_rgb565; + return MESA_FORMAT_RGB565; } - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565; + return do32bpt ? MESA_FORMAT_XRGB8888 : MESA_FORMAT_RGB565; case GL_RGBA8: case GL_RGB10_A2: case GL_RGBA12: case GL_RGBA16: - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + return do32bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB4444; case GL_RGBA4: case GL_RGBA2: - return &_mesa_texformat_argb4444; + return MESA_FORMAT_ARGB4444; case GL_RGB5_A1: - return &_mesa_texformat_argb1555; + return MESA_FORMAT_ARGB1555; case GL_RGB8: case GL_RGB10: case GL_RGB12: case GL_RGB16: - return &_mesa_texformat_argb8888; + return MESA_FORMAT_XRGB8888; case GL_RGB5: case GL_RGB4: case GL_R3_G3_B2: - return &_mesa_texformat_rgb565; + return MESA_FORMAT_RGB565; case GL_ALPHA: case GL_ALPHA4: @@ -83,7 +82,7 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, case GL_ALPHA12: case GL_ALPHA16: case GL_COMPRESSED_ALPHA: - return &_mesa_texformat_a8; + return MESA_FORMAT_A8; case 1: case GL_LUMINANCE: @@ -92,18 +91,24 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, case GL_LUMINANCE12: case GL_LUMINANCE16: case GL_COMPRESSED_LUMINANCE: - return &_mesa_texformat_l8; + return MESA_FORMAT_L8; + + case GL_LUMINANCE12_ALPHA4: + case GL_LUMINANCE12_ALPHA12: + case GL_LUMINANCE16_ALPHA16: +#ifndef I915 + return MESA_FORMAT_AL1616; +#else + /* FALLTHROUGH */ +#endif case 2: case GL_LUMINANCE_ALPHA: case GL_LUMINANCE4_ALPHA4: case GL_LUMINANCE6_ALPHA2: case GL_LUMINANCE8_ALPHA8: - case GL_LUMINANCE12_ALPHA4: - case GL_LUMINANCE12_ALPHA12: - case GL_LUMINANCE16_ALPHA16: case GL_COMPRESSED_LUMINANCE_ALPHA: - return &_mesa_texformat_al88; + return MESA_FORMAT_AL88; case GL_INTENSITY: case GL_INTENSITY4: @@ -111,41 +116,41 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, case GL_INTENSITY12: case GL_INTENSITY16: case GL_COMPRESSED_INTENSITY: - return &_mesa_texformat_i8; + return MESA_FORMAT_I8; case GL_YCBCR_MESA: if (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE) - return &_mesa_texformat_ycbcr; + return MESA_FORMAT_YCBCR; else - return &_mesa_texformat_ycbcr_rev; + return MESA_FORMAT_YCBCR_REV; case GL_COMPRESSED_RGB_FXT1_3DFX: - return &_mesa_texformat_rgb_fxt1; + return MESA_FORMAT_RGB_FXT1; case GL_COMPRESSED_RGBA_FXT1_3DFX: - return &_mesa_texformat_rgba_fxt1; + return MESA_FORMAT_RGBA_FXT1; case GL_RGB_S3TC: case GL_RGB4_S3TC: case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: - return &_mesa_texformat_rgb_dxt1; + return MESA_FORMAT_RGB_DXT1; case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: - return &_mesa_texformat_rgba_dxt1; + return MESA_FORMAT_RGBA_DXT1; case GL_RGBA_S3TC: case GL_RGBA4_S3TC: case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: - return &_mesa_texformat_rgba_dxt3; + return MESA_FORMAT_RGBA_DXT3; case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: - return &_mesa_texformat_rgba_dxt5; + return MESA_FORMAT_RGBA_DXT5; case GL_DEPTH_COMPONENT: case GL_DEPTH_COMPONENT16: case GL_DEPTH_COMPONENT24: case GL_DEPTH_COMPONENT32: #if 0 - return &_mesa_texformat_z16; + return MESA_FORMAT_Z16; #else /* fall-through. * 16bpp depth texture can't be paired with a stencil buffer so @@ -154,7 +159,7 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, #endif case GL_DEPTH_STENCIL_EXT: case GL_DEPTH24_STENCIL8_EXT: - return &_mesa_texformat_s8_z24; + return MESA_FORMAT_S8_Z24; #ifndef I915 case GL_SRGB_EXT: @@ -165,41 +170,41 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, case GL_COMPRESSED_SRGB_ALPHA_EXT: case GL_COMPRESSED_SLUMINANCE_EXT: case GL_COMPRESSED_SLUMINANCE_ALPHA_EXT: - return &_mesa_texformat_sargb8; + return MESA_FORMAT_SARGB8; case GL_SLUMINANCE_EXT: case GL_SLUMINANCE8_EXT: if (IS_G4X(intel->intelScreen->deviceID)) - return &_mesa_texformat_sl8; + return MESA_FORMAT_SL8; else - return &_mesa_texformat_sargb8; + return MESA_FORMAT_SARGB8; case GL_SLUMINANCE_ALPHA_EXT: case GL_SLUMINANCE8_ALPHA8_EXT: if (IS_G4X(intel->intelScreen->deviceID)) - return &_mesa_texformat_sla8; + return MESA_FORMAT_SLA8; else - return &_mesa_texformat_sargb8; + return MESA_FORMAT_SARGB8; case GL_COMPRESSED_SRGB_S3TC_DXT1_EXT: case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT: case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT: case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT: - return &_mesa_texformat_srgb_dxt1; + return MESA_FORMAT_SRGB_DXT1; /* i915 could also do this */ case GL_DUDV_ATI: case GL_DU8DV8_ATI: - return &_mesa_texformat_dudv8; + return MESA_FORMAT_DUDV8; case GL_RGBA_SNORM: case GL_RGBA8_SNORM: - return &_mesa_texformat_signed_rgba8888_rev; + return MESA_FORMAT_SIGNED_RGBA8888_REV; #endif default: fprintf(stderr, "unexpected texture format %s in %s\n", _mesa_lookup_enum_by_nr(internalFormat), __FUNCTION__); - return NULL; + return MESA_FORMAT_NONE; } - return NULL; /* never get here */ + return MESA_FORMAT_NONE; /* never get here */ } int intel_compressed_num_bytes(GLuint mesaFormat) diff --git a/src/mesa/drivers/dri/intel/intel_tex_image.c b/src/mesa/drivers/dri/intel/intel_tex_image.c index 66201b1f465..66d61f93ea9 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_image.c +++ b/src/mesa/drivers/dri/intel/intel_tex_image.c @@ -6,8 +6,10 @@ #include "main/bufferobj.h" #include "main/convolve.h" #include "main/context.h" +#include "main/formats.h" +#include "main/image.h" #include "main/texcompress.h" -#include "main/texformat.h" +#include "main/texstore.h" #include "main/texgetimage.h" #include "main/texobj.h" #include "main/texstore.h" @@ -69,6 +71,7 @@ guess_and_alloc_mipmap_tree(struct intel_context *intel, GLuint depth = intelImage->base.Depth; GLuint l2width, l2height, l2depth; GLuint i, comp_byte = 0; + GLuint texelBytes; DBG("%s\n", __FUNCTION__); @@ -112,7 +115,8 @@ guess_and_alloc_mipmap_tree(struct intel_context *intel, */ if ((intelObj->base.MinFilter == GL_NEAREST || intelObj->base.MinFilter == GL_LINEAR) && - intelImage->level == firstLevel) { + intelImage->level == firstLevel && + (intel->gen < 4 || firstLevel == 0)) { lastLevel = firstLevel; } else { @@ -123,8 +127,11 @@ guess_and_alloc_mipmap_tree(struct intel_context *intel, } assert(!intelObj->mt); - if (intelImage->base.IsCompressed) - comp_byte = intel_compressed_num_bytes(intelImage->base.TexFormat->MesaFormat); + if (_mesa_is_format_compressed(intelImage->base.TexFormat)) + comp_byte = intel_compressed_num_bytes(intelImage->base.TexFormat); + + texelBytes = _mesa_get_format_bytes(intelImage->base.TexFormat); + intelObj->mt = intel_miptree_create(intel, intelObj->base.Target, intelImage->base._BaseFormat, @@ -134,7 +141,7 @@ guess_and_alloc_mipmap_tree(struct intel_context *intel, width, height, depth, - intelImage->base.TexFormat->TexelBytes, + texelBytes, comp_byte, expect_accelerated_upload); @@ -166,7 +173,7 @@ target_to_face(GLenum target) static GLboolean check_pbo_format(GLint internalFormat, GLenum format, GLenum type, - const struct gl_texture_format *mesa_format) + gl_format mesa_format) { switch (internalFormat) { case 4: @@ -174,12 +181,12 @@ check_pbo_format(GLint internalFormat, return (format == GL_BGRA && (type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV) && - mesa_format == &_mesa_texformat_argb8888); + mesa_format == MESA_FORMAT_ARGB8888); case 3: case GL_RGB: return (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5 && - mesa_format == &_mesa_texformat_rgb565); + mesa_format == MESA_FORMAT_RGB565); case GL_YCBCR_MESA: return (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE); default: @@ -200,7 +207,10 @@ try_pbo_upload(struct intel_context *intel, { struct intel_buffer_object *pbo = intel_buffer_object(unpack->BufferObj); GLuint src_offset, src_stride; - GLuint dst_offset, dst_stride; + GLuint dst_x, dst_y, dst_stride; + dri_bo *dst_buffer = intel_region_buffer(intel, + intelImage->mt->region, + INTEL_WRITE_FULL); if (!_mesa_is_bufferobj(unpack->BufferObj) || intel->ctx._ImageTransferState || @@ -217,26 +227,23 @@ try_pbo_upload(struct intel_context *intel, else src_stride = width; - dst_offset = intel_miptree_image_offset(intelImage->mt, - intelImage->face, - intelImage->level); + intel_miptree_get_image_offset(intelImage->mt, intelImage->level, + intelImage->face, 0, + &dst_x, &dst_y); dst_stride = intelImage->mt->pitch; - intelFlush(&intel->ctx); + if (drm_intel_bo_references(intel->batch->buf, dst_buffer)) + intelFlush(&intel->ctx); LOCK_HARDWARE(intel); { dri_bo *src_buffer = intel_bufferobj_buffer(intel, pbo, INTEL_READ); - dri_bo *dst_buffer = intel_region_buffer(intel, - intelImage->mt->region, - INTEL_WRITE_FULL); - if (!intelEmitCopyBlit(intel, intelImage->mt->cpp, src_stride, src_buffer, src_offset, GL_FALSE, - dst_stride, dst_buffer, dst_offset, GL_FALSE, - 0, 0, 0, 0, width, height, + dst_stride, dst_buffer, 0, GL_FALSE, + 0, 0, dst_x, dst_y, width, height, GL_COPY)) { UNLOCK_HARDWARE(intel); return GL_FALSE; @@ -258,7 +265,7 @@ try_pbo_zcopy(struct intel_context *intel, { struct intel_buffer_object *pbo = intel_buffer_object(unpack->BufferObj); GLuint src_offset, src_stride; - GLuint dst_offset, dst_stride; + GLuint dst_x, dst_y, dst_stride; if (!_mesa_is_bufferobj(unpack->BufferObj) || intel->ctx._ImageTransferState || @@ -275,13 +282,14 @@ try_pbo_zcopy(struct intel_context *intel, else src_stride = width; - dst_offset = intel_miptree_image_offset(intelImage->mt, - intelImage->face, - intelImage->level); + intel_miptree_get_image_offset(intelImage->mt, intelImage->level, + intelImage->face, 0, + &dst_x, &dst_y); dst_stride = intelImage->mt->pitch; - if (src_stride != dst_stride || dst_offset != 0 || src_offset != 0) { + if (src_stride != dst_stride || dst_x != 0 || dst_y != 0 || + src_offset != 0) { DBG("%s: failure 2\n", __FUNCTION__); return GL_FALSE; } @@ -316,8 +324,6 @@ intelTexImage(GLcontext * ctx, DBG("%s target %s level %d %dx%dx%d border %d\n", __FUNCTION__, _mesa_lookup_enum_by_nr(target), level, width, height, depth, border); - intelFlush(ctx); - intelImage->face = target_to_face(target); intelImage->level = level; @@ -326,22 +332,11 @@ intelTexImage(GLcontext * ctx, &postConvHeight); } - /* choose the texture format */ - texImage->TexFormat = intelChooseTextureFormat(ctx, internalFormat, - format, type); - - _mesa_set_fetch_functions(texImage, dims); - - if (texImage->TexFormat->TexelBytes == 0) { - /* must be a compressed format */ + if (_mesa_is_format_compressed(texImage->TexFormat)) { texelBytes = 0; - texImage->IsCompressed = GL_TRUE; - texImage->CompressedSize = - ctx->Driver.CompressedTextureSize(ctx, texImage->Width, - texImage->Height, texImage->Depth, - texImage->TexFormat->MesaFormat); - } else { - texelBytes = texImage->TexFormat->TexelBytes; + } + else { + texelBytes = _mesa_get_format_bytes(texImage->TexFormat); /* Minimum pitch of 32 bytes */ if (postConvWidth * texelBytes < 32) { @@ -374,8 +369,7 @@ intelTexImage(GLcontext * ctx, intelObj->mt->first_level == level && intelObj->mt->last_level == level && intelObj->mt->target != GL_TEXTURE_CUBE_MAP_ARB && - !intel_miptree_match_image(intelObj->mt, &intelImage->base, - intelImage->face, intelImage->level)) { + !intel_miptree_match_image(intelObj->mt, &intelImage->base)) { DBG("release it\n"); intel_miptree_release(intel, &intelObj->mt); @@ -392,17 +386,17 @@ intelTexImage(GLcontext * ctx, assert(!intelImage->mt); if (intelObj->mt && - intel_miptree_match_image(intelObj->mt, &intelImage->base, - intelImage->face, intelImage->level)) { + intel_miptree_match_image(intelObj->mt, &intelImage->base)) { intel_miptree_reference(&intelImage->mt, intelObj->mt); assert(intelImage->mt); } else if (intelImage->base.Border == 0) { int comp_byte = 0; - - if (intelImage->base.IsCompressed) { + GLuint texelBytes = _mesa_get_format_bytes(intelImage->base.TexFormat); + GLenum baseFormat = _mesa_get_format_base_format(intelImage->base.TexFormat); + if (_mesa_is_format_compressed(intelImage->base.TexFormat)) { comp_byte = - intel_compressed_num_bytes(intelImage->base.TexFormat->MesaFormat); + intel_compressed_num_bytes(intelImage->base.TexFormat); } /* Didn't fit in the object miptree, but it's suitable for inclusion in @@ -410,11 +404,11 @@ intelTexImage(GLcontext * ctx, * It'll get moved into the object miptree at validate time. */ intelImage->mt = intel_miptree_create(intel, target, - intelImage->base.TexFormat->BaseFormat, + baseFormat, internalFormat, level, level, width, height, depth, - intelImage->base.TexFormat->TexelBytes, + texelBytes, comp_byte, pixels == NULL); } @@ -478,21 +472,31 @@ intelTexImage(GLcontext * ctx, LOCK_HARDWARE(intel); if (intelImage->mt) { - if (pixels != NULL) + if (pixels != NULL) { + /* Flush any queued rendering with the texture before mapping. */ + if (drm_intel_bo_references(intel->batch->buf, + intelImage->mt->region->buffer)) { + intelFlush(ctx); + } texImage->Data = intel_miptree_image_map(intel, intelImage->mt, intelImage->face, intelImage->level, &dstRowStride, intelImage->base.ImageOffsets); + } + texImage->RowStride = dstRowStride / intelImage->mt->cpp; } else { /* Allocate regular memory and store the image there temporarily. */ - if (texImage->IsCompressed) { - sizeInBytes = texImage->CompressedSize; + if (_mesa_is_format_compressed(texImage->TexFormat)) { + sizeInBytes = _mesa_format_image_size(texImage->TexFormat, + texImage->Width, + texImage->Height, + texImage->Depth); dstRowStride = - _mesa_compressed_row_stride(texImage->TexFormat->MesaFormat, width); + _mesa_format_row_stride(texImage->TexFormat, width); assert(dims != 3); } else { @@ -523,17 +527,20 @@ intelTexImage(GLcontext * ctx, pixels, srcRowStride, 0, 0); - } else + } + else { memcpy(texImage->Data, pixels, imageSize); - } else if (!texImage->TexFormat->StoreImage(ctx, dims, - texImage->_BaseFormat, - texImage->TexFormat, - texImage->Data, 0, 0, 0, /* dstX/Y/Zoffset */ - dstRowStride, - texImage->ImageOffsets, - width, height, depth, - format, type, pixels, unpack)) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); + } + } + else if (!_mesa_texstore(ctx, dims, + texImage->_BaseFormat, + texImage->TexFormat, + texImage->Data, 0, 0, 0, /* dstX/Y/Zoffset */ + dstRowStride, + texImage->ImageOffsets, + width, height, depth, + format, type, pixels, unpack)) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); } } @@ -661,9 +668,10 @@ intel_get_tex_image(GLcontext * ctx, GLenum target, GLint level, if (compressed) { _mesa_get_compressed_teximage(ctx, target, level, pixels, texObj, texImage); - } else { + } + else { _mesa_get_teximage(ctx, target, level, format, type, pixels, - texObj, texImage); + texObj, texImage); } @@ -726,17 +734,16 @@ intelSetTexBuffer2(__DRIcontext *pDRICtx, GLint target, { struct intel_framebuffer *intel_fb = dPriv->driverPrivate; struct intel_context *intel = pDRICtx->driverPrivate; + GLcontext *ctx = &intel->ctx; struct intel_texture_object *intelObj; struct intel_texture_image *intelImage; struct intel_mipmap_tree *mt; struct intel_renderbuffer *rb; - struct gl_texture_unit *texUnit; struct gl_texture_object *texObj; struct gl_texture_image *texImage; - int level = 0, type, format, internalFormat; + int level = 0, internalFormat; - texUnit = &intel->ctx.Texture.Unit[intel->ctx.Texture.CurrentUnit]; - texObj = _mesa_select_tex_object(&intel->ctx, texUnit, target); + texObj = _mesa_get_current_tex_object(ctx, target); intelObj = intel_texture_object(texObj); if (!intelObj) @@ -751,8 +758,6 @@ intelSetTexBuffer2(__DRIcontext *pDRICtx, GLint target, if (rb->region == NULL) return; - type = GL_BGRA; - format = GL_UNSIGNED_BYTE; if (glx_texture_format == GLX_TEXTURE_FORMAT_RGB_EXT) internalFormat = GL_RGB; else @@ -783,14 +788,14 @@ intelSetTexBuffer2(__DRIcontext *pDRICtx, GLint target, intelImage->face = target_to_face(target); intelImage->level = level; - texImage->TexFormat = intelChooseTextureFormat(&intel->ctx, internalFormat, - type, format); - _mesa_set_fetch_functions(texImage, 2); + if (glx_texture_format == GLX_TEXTURE_FORMAT_RGB_EXT) + texImage->TexFormat = MESA_FORMAT_XRGB8888; + else + texImage->TexFormat = MESA_FORMAT_ARGB8888; texImage->RowStride = rb->region->pitch; intel_miptree_reference(&intelImage->mt, intelObj->mt); - if (!intel_miptree_match_image(intelObj->mt, &intelImage->base, - intelImage->face, intelImage->level)) { + if (!intel_miptree_match_image(intelObj->mt, &intelImage->base)) { fprintf(stderr, "miptree doesn't match image\n"); } diff --git a/src/mesa/drivers/dri/intel/intel_tex_layout.h b/src/mesa/drivers/dri/intel/intel_tex_layout.h index c9de9b56784..a9ac9e7eb48 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_layout.h +++ b/src/mesa/drivers/dri/intel/intel_tex_layout.h @@ -33,7 +33,7 @@ #include "main/macros.h" -static GLuint minify( GLuint d ) +static INLINE GLuint minify( GLuint d ) { return MAX2(1, d>>1); } diff --git a/src/mesa/drivers/dri/intel/intel_tex_obj.h b/src/mesa/drivers/dri/intel/intel_tex_obj.h index 5a934615259..3ad10d3d238 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_obj.h +++ b/src/mesa/drivers/dri/intel/intel_tex_obj.h @@ -66,6 +66,7 @@ struct intel_texture_image * Else there is no image data. */ struct intel_mipmap_tree *mt; + GLboolean used_as_render_target; }; static INLINE struct intel_texture_object * diff --git a/src/mesa/drivers/dri/intel/intel_tex_subimage.c b/src/mesa/drivers/dri/intel/intel_tex_subimage.c index 751ec2c98c2..1f682082660 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_subimage.c +++ b/src/mesa/drivers/dri/intel/intel_tex_subimage.c @@ -85,13 +85,13 @@ intelTexSubimage(GLcontext * ctx, &dstRowStride, texImage->ImageOffsets); else { - if (texImage->IsCompressed) { + if (_mesa_is_format_compressed(texImage->TexFormat)) { dstRowStride = - _mesa_compressed_row_stride(texImage->TexFormat->MesaFormat, width); + _mesa_format_row_stride(texImage->TexFormat, width); assert(dims != 3); } else { - dstRowStride = texImage->RowStride * texImage->TexFormat->TexelBytes; + dstRowStride = texImage->RowStride * _mesa_get_format_bytes(texImage->TexFormat); } } @@ -105,18 +105,20 @@ intelTexSubimage(GLcontext * ctx, xoffset, yoffset / 4, (width + 3) & ~3, (height + 3) / 4, pixels, (width + 3) & ~3, 0, 0); - } else + } + else { memcpy(texImage->Data, pixels, imageSize); + } } else { - if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat, - texImage->TexFormat, - texImage->Data, - xoffset, yoffset, zoffset, - dstRowStride, - texImage->ImageOffsets, - width, height, depth, - format, type, pixels, packing)) { + if (!_mesa_texstore(ctx, dims, texImage->_BaseFormat, + texImage->TexFormat, + texImage->Data, + xoffset, yoffset, zoffset, + dstRowStride, + texImage->ImageOffsets, + width, height, depth, + format, type, pixels, packing)) { _mesa_error(ctx, GL_OUT_OF_MEMORY, "intelTexSubImage"); } } diff --git a/src/mesa/drivers/dri/intel/intel_tex_validate.c b/src/mesa/drivers/dri/intel/intel_tex_validate.c index a284d5475f4..c9a24ac3983 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_validate.c +++ b/src/mesa/drivers/dri/intel/intel_tex_validate.c @@ -5,6 +5,7 @@ #include "intel_batchbuffer.h" #include "intel_mipmap_tree.h" #include "intel_tex.h" +#include "intel_chipset.h" #define FILE_DEBUG_FLAG DEBUG_TEXTURE @@ -14,7 +15,8 @@ * GL_TEXTURE_MAX_LOD, GL_TEXTURE_BASE_LEVEL, and GL_TEXTURE_MAX_LEVEL. */ static void -intel_calculate_first_last_level(struct intel_texture_object *intelObj) +intel_calculate_first_last_level(struct intel_context *intel, + struct intel_texture_object *intelObj) { struct gl_texture_object *tObj = &intelObj->base; const struct gl_texture_image *const baseImage = @@ -40,27 +42,27 @@ intel_calculate_first_last_level(struct intel_texture_object *intelObj) firstLevel = lastLevel = tObj->BaseLevel; } else { -#ifdef I915 - firstLevel = tObj->BaseLevel + (GLint) (tObj->MinLod + 0.5); - firstLevel = MAX2(firstLevel, tObj->BaseLevel); - firstLevel = MIN2(firstLevel, tObj->BaseLevel + baseImage->MaxLog2); - lastLevel = tObj->BaseLevel + (GLint) (tObj->MaxLod + 0.5); - lastLevel = MAX2(lastLevel, tObj->BaseLevel); - lastLevel = MIN2(lastLevel, tObj->BaseLevel + baseImage->MaxLog2); - lastLevel = MIN2(lastLevel, tObj->MaxLevel); - lastLevel = MAX2(firstLevel, lastLevel); /* need at least one level */ -#else - /* Currently not taking min/max lod into account here, those - * values are programmed as sampler state elsewhere and we - * upload the same mipmap levels regardless. Not sure if - * this makes sense as it means it isn't possible for the app - * to use min/max lod to reduce texture memory pressure: - */ - firstLevel = tObj->BaseLevel; - lastLevel = MIN2(tObj->BaseLevel + baseImage->MaxLog2, - tObj->MaxLevel); - lastLevel = MAX2(firstLevel, lastLevel); /* need at least one level */ -#endif + if (intel->gen == 2) { + firstLevel = tObj->BaseLevel + (GLint) (tObj->MinLod + 0.5); + firstLevel = MAX2(firstLevel, tObj->BaseLevel); + firstLevel = MIN2(firstLevel, tObj->BaseLevel + baseImage->MaxLog2); + lastLevel = tObj->BaseLevel + (GLint) (tObj->MaxLod + 0.5); + lastLevel = MAX2(lastLevel, tObj->BaseLevel); + lastLevel = MIN2(lastLevel, tObj->BaseLevel + baseImage->MaxLog2); + lastLevel = MIN2(lastLevel, tObj->MaxLevel); + lastLevel = MAX2(firstLevel, lastLevel); /* need at least one level */ + } else { + /* Min/max LOD are taken into account in sampler state. We don't + * want to re-layout textures just because clamping has been applied + * since it means a bunch of blitting around and probably no memory + * savings (since we have to keep the other levels around anyway). + */ + firstLevel = tObj->BaseLevel; + lastLevel = MIN2(tObj->BaseLevel + baseImage->MaxLog2, + tObj->MaxLevel); + /* need at least one level */ + lastLevel = MAX2(firstLevel, lastLevel); + } } break; case GL_TEXTURE_RECTANGLE_NV: @@ -135,9 +137,8 @@ intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit) /* What levels must the tree include at a minimum? */ - intel_calculate_first_last_level(intelObj); - firstImage = - intel_texture_image(intelObj->base.Image[0][intelObj->firstLevel]); + intel_calculate_first_last_level(intel, intelObj); + firstImage = intel_texture_image(tObj->Image[0][intelObj->firstLevel]); /* Fallback case: */ @@ -165,11 +166,12 @@ intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit) intel_miptree_reference(&intelObj->mt, firstImage->mt); } - if (firstImage->base.IsCompressed) { - comp_byte = intel_compressed_num_bytes(firstImage->base.TexFormat->MesaFormat); + if (_mesa_is_format_compressed(firstImage->base.TexFormat)) { + comp_byte = intel_compressed_num_bytes(firstImage->base.TexFormat); cpp = comp_byte; } - else cpp = firstImage->base.TexFormat->TexelBytes; + else + cpp = _mesa_get_format_bytes(firstImage->base.TexFormat); /* Check tree can hold all active levels. Check tree matches * target, imageFormat, etc. @@ -189,7 +191,7 @@ intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit) intelObj->mt->height0 != firstImage->base.Height || intelObj->mt->depth0 != firstImage->base.Depth || intelObj->mt->cpp != cpp || - intelObj->mt->compressed != firstImage->base.IsCompressed)) { + intelObj->mt->compressed != _mesa_is_format_compressed(firstImage->base.TexFormat))) { intel_miptree_release(intel, &intelObj->mt); } @@ -220,8 +222,13 @@ intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit) intel_texture_image(intelObj->base.Image[face][i]); /* Need to import images in main memory or held in other trees. + * If it's a render target, then its data isn't needed to be in + * the object tree (otherwise we'd be FBO incomplete), and we need + * to keep track of the image's MT as needing to be pulled in still, + * or we'll lose the rendering that's done to it. */ - if (intelObj->mt != intelImage->mt) { + if (intelObj->mt != intelImage->mt && + !intelImage->used_as_render_target) { copy_image_data_to_tree(intel, intelObj, intelImage); } } diff --git a/src/mesa/drivers/dri/mach64/Makefile b/src/mesa/drivers/dri/mach64/Makefile index 7246d51f5d7..a8f463e9fd0 100644 --- a/src/mesa/drivers/dri/mach64/Makefile +++ b/src/mesa/drivers/dri/mach64/Makefile @@ -30,4 +30,3 @@ ASM_SOURCES = include ../Makefile.template -symlinks: diff --git a/src/mesa/drivers/dri/mach64/mach64_context.c b/src/mesa/drivers/dri/mach64/mach64_context.c index 9c7f513c6f5..2bca293b3c2 100644 --- a/src/mesa/drivers/dri/mach64/mach64_context.c +++ b/src/mesa/drivers/dri/mach64/mach64_context.c @@ -76,7 +76,7 @@ static const struct dri_debug_control debug_control[] = { NULL, 0 } }; -const struct dri_extension card_extensions[] = +static const struct dri_extension card_extensions[] = { { "GL_ARB_multitexture", NULL }, { "GL_EXT_texture_edge_clamp", NULL }, diff --git a/src/mesa/drivers/dri/mach64/mach64_native_vb.c b/src/mesa/drivers/dri/mach64/mach64_native_vb.c index 99f1a14e171..816682ec5f1 100644 --- a/src/mesa/drivers/dri/mach64/mach64_native_vb.c +++ b/src/mesa/drivers/dri/mach64/mach64_native_vb.c @@ -207,19 +207,19 @@ INTERP_QUALIFIER void TAG(interp_extras)( GLcontext *ctx, LOCALVARS struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - if (VB->ColorPtr[1]) { - assert(VB->ColorPtr[1]->stride == 4 * sizeof(GLfloat)); + if (VB->BackfaceColorPtr) { + assert(VB->BackfaceColorPtr->stride == 4 * sizeof(GLfloat)); INTERP_4F( t, - GET_COLOR(VB->ColorPtr[1], dst), - GET_COLOR(VB->ColorPtr[1], out), - GET_COLOR(VB->ColorPtr[1], in) ); + GET_COLOR(VB->BackfaceColorPtr, dst), + GET_COLOR(VB->BackfaceColorPtr, out), + GET_COLOR(VB->BackfaceColorPtr, in) ); - if (VB->SecondaryColorPtr[1]) { + if (VB->BackfaceSecondaryColorPtr) { INTERP_3F( t, - GET_COLOR(VB->SecondaryColorPtr[1], dst), - GET_COLOR(VB->SecondaryColorPtr[1], out), - GET_COLOR(VB->SecondaryColorPtr[1], in) ); + GET_COLOR(VB->BackfaceSecondaryColorPtr, dst), + GET_COLOR(VB->BackfaceSecondaryColorPtr, out), + GET_COLOR(VB->BackfaceSecondaryColorPtr, in) ); } } @@ -236,13 +236,13 @@ INTERP_QUALIFIER void TAG(copy_pv_extras)( GLcontext *ctx, LOCALVARS struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - if (VB->ColorPtr[1]) { - COPY_4FV( GET_COLOR(VB->ColorPtr[1], dst), - GET_COLOR(VB->ColorPtr[1], src) ); + if (VB->BackfaceColorPtr) { + COPY_4FV( GET_COLOR(VB->BackfaceColorPtr, dst), + GET_COLOR(VB->BackfaceColorPtr, src) ); - if (VB->SecondaryColorPtr[1]) { - COPY_4FV( GET_COLOR(VB->SecondaryColorPtr[1], dst), - GET_COLOR(VB->SecondaryColorPtr[1], src) ); + if (VB->BackfaceSecondaryColorPtr) { + COPY_4FV( GET_COLOR(VB->BackfaceSecondaryColorPtr, dst), + GET_COLOR(VB->BackfaceSecondaryColorPtr, src) ); } } diff --git a/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h b/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h index 684f2acc890..6e5fa3520e1 100644 --- a/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h +++ b/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h @@ -103,10 +103,10 @@ static void TAG(emit)( GLcontext *ctx, #if DO_TEX1 { const GLuint t1 = GET_TEXSOURCE(1); - tc1 = VB->TexCoordPtr[t1]->data; - tc1_stride = VB->TexCoordPtr[t1]->stride; + tc1 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->data; + tc1_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->stride; #if DO_PTEX - tc1_size = VB->TexCoordPtr[t1]->size; + tc1_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->size; #endif } #endif @@ -114,18 +114,18 @@ static void TAG(emit)( GLcontext *ctx, #if DO_TEX0 { const GLuint t0 = GET_TEXSOURCE(0); - tc0 = VB->TexCoordPtr[t0]->data; - tc0_stride = VB->TexCoordPtr[t0]->stride; + tc0 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->data; + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->stride; #if DO_PTEX - tc0_size = VB->TexCoordPtr[t0]->size; + tc0_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->size; #endif } #endif #if DO_SPEC - if (VB->SecondaryColorPtr[0]) { - spec = VB->SecondaryColorPtr[0]->data; - spec_stride = VB->SecondaryColorPtr[0]->stride; + if (VB->AttribPtr[_TNL_ATTRIB_COLOR1]) { + spec = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->data; + spec_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->stride; } else { spec = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_COLOR1]; spec_stride = 0; @@ -133,9 +133,9 @@ static void TAG(emit)( GLcontext *ctx, #endif #if DO_FOG - if (VB->FogCoordPtr) { - fog = VB->FogCoordPtr->data; - fog_stride = VB->FogCoordPtr->stride; + if (VB->AttribPtr[_TNL_ATTRIB_FOG]) { + fog = VB->AttribPtr[_TNL_ATTRIB_FOG]->data; + fog_stride = VB->AttribPtr[_TNL_ATTRIB_FOG]->stride; } else { static GLfloat tmp[4] = {0, 0, 0, 0}; fog = &tmp; @@ -144,8 +144,8 @@ static void TAG(emit)( GLcontext *ctx, #endif #if DO_RGBA - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; #endif coord = VB->NdcPtr->data; @@ -319,8 +319,8 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) /* Force 'missing' texcoords to something valid. */ - if (DO_TEX1 && VB->TexCoordPtr[0] == 0) - VB->TexCoordPtr[0] = VB->TexCoordPtr[1]; + if (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0] = VB->AttribPtr[_TNL_ATTRIB_TEX1]; if (DO_PTEX) return GL_TRUE; @@ -328,12 +328,12 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) /* No hardware support for projective texture. Can fake it for * TEX0 only. */ - if ((DO_TEX1 && VB->TexCoordPtr[GET_TEXSOURCE(1)]->size == 4)) { + if ((DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(1)]->size == 4)) { PTEX_FALLBACK(); return GL_FALSE; } - if (DO_TEX0 && VB->TexCoordPtr[GET_TEXSOURCE(0)]->size == 4) { + if (DO_TEX0 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(0)]->size == 4) { if (DO_TEX1) { PTEX_FALLBACK(); } diff --git a/src/mesa/drivers/dri/mach64/mach64_screen.c b/src/mesa/drivers/dri/mach64/mach64_screen.c index 6440027ca4a..3b19cf53333 100644 --- a/src/mesa/drivers/dri/mach64/mach64_screen.c +++ b/src/mesa/drivers/dri/mach64/mach64_screen.c @@ -67,8 +67,6 @@ static const GLuint __driNConfigOptions = 3; static const GLuint __driNConfigOptions = 2; #endif -extern const struct dri_extension card_extensions[]; - static const __DRIconfig ** mach64FillInModes( __DRIscreenPrivate *psp, unsigned pixel_bits, unsigned depth_bits, @@ -316,7 +314,7 @@ mach64CreateBuffer( __DRIscreenPrivate *driScrnPriv, { driRenderbuffer *frontRb - = driNewRenderbuffer(GL_RGBA, + = driNewRenderbuffer(MESA_FORMAT_ARGB8888, NULL, screen->cpp, screen->frontOffset, screen->frontPitch, @@ -327,7 +325,7 @@ mach64CreateBuffer( __DRIscreenPrivate *driScrnPriv, if (mesaVis->doubleBufferMode) { driRenderbuffer *backRb - = driNewRenderbuffer(GL_RGBA, + = driNewRenderbuffer(MESA_FORMAT_ARGB8888, NULL, screen->cpp, screen->backOffset, screen->backPitch, @@ -338,7 +336,7 @@ mach64CreateBuffer( __DRIscreenPrivate *driScrnPriv, if (mesaVis->depthBits == 16) { driRenderbuffer *depthRb - = driNewRenderbuffer(GL_DEPTH_COMPONENT16, + = driNewRenderbuffer(MESA_FORMAT_Z16, NULL, screen->cpp, screen->depthOffset, screen->depthPitch, driDrawPriv); @@ -348,7 +346,7 @@ mach64CreateBuffer( __DRIscreenPrivate *driScrnPriv, else if (mesaVis->depthBits == 24) { /* XXX I don't think 24-bit Z is supported - so this isn't used */ driRenderbuffer *depthRb - = driNewRenderbuffer(GL_DEPTH_COMPONENT24, + = driNewRenderbuffer(MESA_FORMAT_Z24_S8, NULL, screen->cpp, screen->depthOffset, screen->depthPitch, @@ -436,18 +434,6 @@ mach64InitScreen(__DRIscreenPrivate *psp) return NULL; } - /* Calling driInitExtensions here, with a NULL context pointer, - * does not actually enable the extensions. It just makes sure - * that all the dispatch offsets for all the extensions that - * *might* be enables are known. This is needed because the - * dispatch offsets need to be known when _mesa_context_create is - * called, but we can't enable the extensions until we have a - * context pointer. - * - * Hello chicken. Hello egg. How are you two today? - */ - driInitExtensions( NULL, card_extensions, GL_FALSE ); - if (!mach64InitDriver(psp)) return NULL; diff --git a/src/mesa/drivers/dri/mach64/mach64_span.c b/src/mesa/drivers/dri/mach64/mach64_span.c index 91d46ce32eb..500319e0e31 100644 --- a/src/mesa/drivers/dri/mach64/mach64_span.c +++ b/src/mesa/drivers/dri/mach64/mach64_span.c @@ -157,15 +157,13 @@ void mach64DDInitSpanFuncs( GLcontext *ctx ) void mach64SetSpanFunctions(driRenderbuffer *drb, const GLvisual *vis) { - if (drb->Base.InternalFormat == GL_RGBA) { - if (vis->redBits == 5 && vis->greenBits == 6 && vis->blueBits == 5) { - mach64InitPointers_RGB565(&drb->Base); - } - else { - mach64InitPointers_ARGB8888(&drb->Base); - } + if (drb->Base.Format == MESA_FORMAT_RGB565) { + mach64InitPointers_RGB565(&drb->Base); } - else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT16) { + else if (drb->Base.Format == MESA_FORMAT_ARGB8888) { + mach64InitPointers_ARGB8888(&drb->Base); + } + else if (drb->Base.Format == MESA_FORMAT_Z16) { mach64InitDepthPointers_z16(&drb->Base); } } diff --git a/src/mesa/drivers/dri/mach64/mach64_tex.c b/src/mesa/drivers/dri/mach64/mach64_tex.c index 225d23179e1..72917ee13bf 100644 --- a/src/mesa/drivers/dri/mach64/mach64_tex.c +++ b/src/mesa/drivers/dri/mach64/mach64_tex.c @@ -41,7 +41,6 @@ #include "main/simple_list.h" #include "main/enums.h" #include "main/texstore.h" -#include "main/texformat.h" #include "main/teximage.h" #include "main/texobj.h" #include "main/imports.h" @@ -138,7 +137,7 @@ mach64AllocTexObj( struct gl_texture_object *texObj ) /* Called by the _mesa_store_teximage[123]d() functions. */ -static const struct gl_texture_format * +static gl_format mach64ChooseTextureFormat( GLcontext *ctx, GLint internalFormat, GLenum format, GLenum type ) { @@ -167,15 +166,15 @@ mach64ChooseTextureFormat( GLcontext *ctx, GLint internalFormat, case GL_RGBA2: case GL_COMPRESSED_RGBA: if (mmesa->mach64Screen->cpp == 4) - return &_mesa_texformat_argb8888; + return MESA_FORMAT_ARGB8888; else - return &_mesa_texformat_argb4444; + return MESA_FORMAT_ARGB4444; case GL_RGB5_A1: if (mmesa->mach64Screen->cpp == 4) - return &_mesa_texformat_argb8888; + return MESA_FORMAT_ARGB8888; else - return &_mesa_texformat_argb1555; + return MESA_FORMAT_ARGB1555; case GL_RGBA8: case GL_RGB10_A2: @@ -183,9 +182,9 @@ mach64ChooseTextureFormat( GLcontext *ctx, GLint internalFormat, case GL_RGBA16: case GL_RGBA4: if (mmesa->mach64Screen->cpp == 4) - return &_mesa_texformat_argb8888; + return MESA_FORMAT_ARGB8888; else - return &_mesa_texformat_argb4444; + return MESA_FORMAT_ARGB4444; case 3: case GL_RGB: @@ -198,9 +197,9 @@ mach64ChooseTextureFormat( GLcontext *ctx, GLint internalFormat, case GL_RGB16: case GL_COMPRESSED_RGB: if (mmesa->mach64Screen->cpp == 4) - return &_mesa_texformat_argb8888; + return MESA_FORMAT_ARGB8888; else - return &_mesa_texformat_rgb565; + return MESA_FORMAT_RGB565; case 1: case GL_LUMINANCE: @@ -210,9 +209,9 @@ mach64ChooseTextureFormat( GLcontext *ctx, GLint internalFormat, case GL_LUMINANCE16: case GL_COMPRESSED_LUMINANCE: if (mmesa->mach64Screen->cpp == 4) - return &_mesa_texformat_argb8888; /* inefficient but accurate */ + return MESA_FORMAT_ARGB8888; /* inefficient but accurate */ else - return &_mesa_texformat_argb1555; + return MESA_FORMAT_ARGB1555; case GL_INTENSITY4: case GL_INTENSITY: @@ -221,9 +220,9 @@ mach64ChooseTextureFormat( GLcontext *ctx, GLint internalFormat, case GL_INTENSITY16: case GL_COMPRESSED_INTENSITY: if (mmesa->mach64Screen->cpp == 4) - return &_mesa_texformat_argb8888; /* inefficient but accurate */ + return MESA_FORMAT_ARGB8888; /* inefficient but accurate */ else - return &_mesa_texformat_argb4444; + return MESA_FORMAT_ARGB4444; case GL_COLOR_INDEX: case GL_COLOR_INDEX1_EXT: @@ -232,18 +231,18 @@ mach64ChooseTextureFormat( GLcontext *ctx, GLint internalFormat, case GL_COLOR_INDEX8_EXT: case GL_COLOR_INDEX12_EXT: case GL_COLOR_INDEX16_EXT: - return &_mesa_texformat_ci8; + return MESA_FORMAT_CI8; case GL_YCBCR_MESA: if (type == GL_UNSIGNED_SHORT_8_8_APPLE || type == GL_UNSIGNED_BYTE) - return &_mesa_texformat_ycbcr; + return MESA_FORMAT_YCBCR; else - return &_mesa_texformat_ycbcr_rev; + return MESA_FORMAT_YCBCR_REV; default: _mesa_problem( ctx, "unexpected format in %s", __FUNCTION__ ); - return NULL; + return MESA_FORMAT_NONE; } } @@ -566,8 +565,6 @@ void mach64InitTextureFuncs( struct dd_function_table *functions ) functions->IsTextureResident = driIsTextureResident; functions->UpdateTexturePalette = NULL; - functions->ActiveTexture = NULL; - functions->PrioritizeTexture = NULL; driInitTextureFormats(); } diff --git a/src/mesa/drivers/dri/mach64/mach64_texmem.c b/src/mesa/drivers/dri/mach64/mach64_texmem.c index 734e547952e..b97e9eec25b 100644 --- a/src/mesa/drivers/dri/mach64/mach64_texmem.c +++ b/src/mesa/drivers/dri/mach64/mach64_texmem.c @@ -31,6 +31,11 @@ * Jose Fonseca <[email protected]> */ +#include "main/context.h" +#include "main/macros.h" +#include "main/simple_list.h" +#include "main/imports.h" + #include "mach64_context.h" #include "mach64_state.h" #include "mach64_ioctl.h" @@ -38,12 +43,6 @@ #include "mach64_tris.h" #include "mach64_tex.h" -#include "main/context.h" -#include "main/macros.h" -#include "main/simple_list.h" -#include "main/texformat.h" -#include "main/imports.h" - /* Destroy hardware state associated with texture `t'. */ @@ -76,6 +75,7 @@ static void mach64UploadAGPSubImage( mach64ContextPtr mmesa, struct gl_texture_image *image; int texelsPerDword = 0; int dwords; + GLuint texelBytes; /* Ensure we have a valid texture to upload */ if ( ( level < 0 ) || ( level > mmesa->glCtx->Const.MaxTextureLevels ) ) @@ -85,7 +85,9 @@ static void mach64UploadAGPSubImage( mach64ContextPtr mmesa, if ( !image ) return; - switch ( image->TexFormat->TexelBytes ) { + texelBytes = _mesa_get_format_bytes(image->TexFormat); + + switch ( texelBytes ) { case 1: texelsPerDword = 4; break; case 2: texelsPerDword = 2; break; case 4: texelsPerDword = 1; break; @@ -118,8 +120,8 @@ static void mach64UploadAGPSubImage( mach64ContextPtr mmesa, { CARD32 *dst = (CARD32 *)((char *)mach64Screen->agpTextures.map + t->base.memBlock->ofs); const GLubyte *src = (const GLubyte *) image->Data + - (y * image->Width + x) * image->TexFormat->TexelBytes; - const GLuint bytes = width * height * image->TexFormat->TexelBytes; + (y * image->Width + x) * texelBytes; + const GLuint bytes = width * height * texelBytes; memcpy(dst, src, bytes); } @@ -140,6 +142,7 @@ static void mach64UploadLocalSubImage( mach64ContextPtr mmesa, const int maxdwords = (MACH64_BUFFER_MAX_DWORDS - (MACH64_HOSTDATA_BLIT_OFFSET / 4)); CARD32 pitch, offset; int i; + GLuint texelBytes; /* Ensure we have a valid texture to upload */ if ( ( level < 0 ) || ( level > mmesa->glCtx->Const.MaxTextureLevels ) ) @@ -149,7 +152,9 @@ static void mach64UploadLocalSubImage( mach64ContextPtr mmesa, if ( !image ) return; - switch ( image->TexFormat->TexelBytes ) { + texelBytes = _mesa_get_format_bytes(image->TexFormat); + + switch ( texelBytes ) { case 1: texelsPerDword = 4; break; case 2: texelsPerDword = 2; break; case 4: texelsPerDword = 1; break; @@ -259,7 +264,7 @@ static void mach64UploadLocalSubImage( mach64ContextPtr mmesa, { const GLubyte *src = (const GLubyte *) image->Data + - (y * image->Width + x) * image->TexFormat->TexelBytes; + (y * image->Width + x) * texelBytes; mach64FireBlitLocked( mmesa, (void *)src, offset, pitch, format, x, y, width, height ); diff --git a/src/mesa/drivers/dri/mach64/mach64_texstate.c b/src/mesa/drivers/dri/mach64/mach64_texstate.c index fd2369dd882..df0a09a5c12 100644 --- a/src/mesa/drivers/dri/mach64/mach64_texstate.c +++ b/src/mesa/drivers/dri/mach64/mach64_texstate.c @@ -33,7 +33,6 @@ #include "main/imports.h" #include "main/context.h" #include "main/macros.h" -#include "main/texformat.h" #include "mach64_context.h" #include "mach64_ioctl.h" @@ -55,7 +54,7 @@ static void mach64SetTexImages( mach64ContextPtr mmesa, if ( MACH64_DEBUG & DEBUG_VERBOSE_API ) fprintf( stderr, "%s( %p )\n", __FUNCTION__, tObj ); - switch (baseImage->TexFormat->MesaFormat) { + switch (baseImage->TexFormat) { case MESA_FORMAT_ARGB8888: t->textureFormat = MACH64_DATATYPE_ARGB8888; break; @@ -89,7 +88,7 @@ static void mach64SetTexImages( mach64ContextPtr mmesa, totalSize = ( baseImage->Height * baseImage->Width * - baseImage->TexFormat->TexelBytes ); + _mesa_get_format_bytes(baseImage->TexFormat) ); totalSize = (totalSize + 31) & ~31; diff --git a/src/mesa/drivers/dri/mach64/mach64_vbtmp.h b/src/mesa/drivers/dri/mach64/mach64_vbtmp.h index 938804af9ec..60bfab8f6dc 100644 --- a/src/mesa/drivers/dri/mach64/mach64_vbtmp.h +++ b/src/mesa/drivers/dri/mach64/mach64_vbtmp.h @@ -156,53 +156,53 @@ static void TAG(emit)( GLcontext *ctx, if (DO_TEX3) { const GLuint t3 = GET_TEXSOURCE(3); - tc3 = VB->TexCoordPtr[t3]->data; - tc3_stride = VB->TexCoordPtr[t3]->stride; + tc3 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t3]->data; + tc3_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t3]->stride; if (DO_PTEX) - tc3_size = VB->TexCoordPtr[t3]->size; + tc3_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t3]->size; } if (DO_TEX2) { const GLuint t2 = GET_TEXSOURCE(2); - tc2 = VB->TexCoordPtr[t2]->data; - tc2_stride = VB->TexCoordPtr[t2]->stride; + tc2 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->data; + tc2_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->stride; if (DO_PTEX) - tc2_size = VB->TexCoordPtr[t2]->size; + tc2_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->size; } if (DO_TEX1) { const GLuint t1 = GET_TEXSOURCE(1); - tc1 = VB->TexCoordPtr[t1]->data; - tc1_stride = VB->TexCoordPtr[t1]->stride; + tc1 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->data; + tc1_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->stride; if (DO_PTEX) - tc1_size = VB->TexCoordPtr[t1]->size; + tc1_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->size; } if (DO_TEX0) { const GLuint t0 = GET_TEXSOURCE(0); - tc0_stride = VB->TexCoordPtr[t0]->stride; - tc0 = VB->TexCoordPtr[t0]->data; + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->stride; + tc0 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->data; if (DO_PTEX) - tc0_size = VB->TexCoordPtr[t0]->size; + tc0_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->size; } if (DO_RGBA) { - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; } if (DO_SPEC) { - spec = VB->SecondaryColorPtr[0]->data; - spec_stride = VB->SecondaryColorPtr[0]->stride; + spec = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->data; + spec_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->stride; } else { spec = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_COLOR1]; spec_stride = 0; } if (DO_FOG) { - if (VB->FogCoordPtr) { - fog = VB->FogCoordPtr->data; - fog_stride = VB->FogCoordPtr->stride; + if (VB->AttribPtr[_TNL_ATTRIB_FOG]) { + fog = VB->AttribPtr[_TNL_ATTRIB_FOG]->data; + fog_stride = VB->AttribPtr[_TNL_ATTRIB_FOG]->stride; } else { static GLfloat tmp[4] = {0, 0, 0, 0}; fog = &tmp; @@ -384,8 +384,8 @@ static void TAG(emit)( GLcontext *ctx, GLuint start, GLuint end, ASSERT(stride == 4); - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; /* Pack what's left into a 4-dword vertex. Color is in a different * place, and there is no 'w' coordinate. @@ -432,8 +432,8 @@ static void TAG(emit)( GLcontext *ctx, GLuint start, GLuint end, GLfloat *v = (GLfloat *)dest; int i; - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; if (start) STRIDE_4F(col, col_stride * start); @@ -473,22 +473,22 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) /* Force 'missing' texcoords to something valid. */ - if (DO_TEX3 && VB->TexCoordPtr[2] == 0) - VB->TexCoordPtr[2] = VB->TexCoordPtr[3]; + if (DO_TEX3 && VB->AttribPtr[_TNL_ATTRIB_TEX2] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX2] = VB->AttribPtr[_TNL_ATTRIB_TEX3]; - if (DO_TEX2 && VB->TexCoordPtr[1] == 0) - VB->TexCoordPtr[1] = VB->TexCoordPtr[2]; + if (DO_TEX2 && VB->AttribPtr[_TNL_ATTRIB_TEX1] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX1] = VB->AttribPtr[_TNL_ATTRIB_TEX2]; - if (DO_TEX1 && VB->TexCoordPtr[0] == 0) - VB->TexCoordPtr[0] = VB->TexCoordPtr[1]; + if (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0] = VB->AttribPtr[_TNL_ATTRIB_TEX1]; if (DO_PTEX) return GL_TRUE; - if ((DO_TEX3 && VB->TexCoordPtr[GET_TEXSOURCE(3)]->size == 4) || - (DO_TEX2 && VB->TexCoordPtr[GET_TEXSOURCE(2)]->size == 4) || - (DO_TEX1 && VB->TexCoordPtr[GET_TEXSOURCE(1)]->size == 4) || - (DO_TEX0 && VB->TexCoordPtr[GET_TEXSOURCE(0)]->size == 4)) + if ((DO_TEX3 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(3)]->size == 4) || + (DO_TEX2 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(2)]->size == 4) || + (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(1)]->size == 4) || + (DO_TEX0 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(0)]->size == 4)) return GL_FALSE; return GL_TRUE; @@ -501,14 +501,14 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) /* Force 'missing' texcoords to something valid. */ - if (DO_TEX3 && VB->TexCoordPtr[2] == 0) - VB->TexCoordPtr[2] = VB->TexCoordPtr[3]; + if (DO_TEX3 && VB->AttribPtr[_TNL_ATTRIB_TEX2] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX2] = VB->AttribPtr[_TNL_ATTRIB_TEX3]; - if (DO_TEX2 && VB->TexCoordPtr[1] == 0) - VB->TexCoordPtr[1] = VB->TexCoordPtr[2]; + if (DO_TEX2 && VB->AttribPtr[_TNL_ATTRIB_TEX1] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX1] = VB->AttribPtr[_TNL_ATTRIB_TEX2]; - if (DO_TEX1 && VB->TexCoordPtr[0] == 0) - VB->TexCoordPtr[0] = VB->TexCoordPtr[1]; + if (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0] = VB->AttribPtr[_TNL_ATTRIB_TEX1]; if (DO_PTEX) return GL_TRUE; @@ -516,14 +516,14 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) /* No hardware support for projective texture. Can fake it for * TEX0 only. */ - if ((DO_TEX3 && VB->TexCoordPtr[GET_TEXSOURCE(3)]->size == 4) || - (DO_TEX2 && VB->TexCoordPtr[GET_TEXSOURCE(2)]->size == 4) || - (DO_TEX1 && VB->TexCoordPtr[GET_TEXSOURCE(1)]->size == 4)) { + if ((DO_TEX3 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(3)]->size == 4) || + (DO_TEX2 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(2)]->size == 4) || + (DO_TEX1 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(1)]->size == 4)) { PTEX_FALLBACK(); return GL_FALSE; } - if (DO_TEX0 && VB->TexCoordPtr[GET_TEXSOURCE(0)]->size == 4) { + if (DO_TEX0 && VB->AttribPtr[_TNL_ATTRIB_TEX0 + GET_TEXSOURCE(0)]->size == 4) { if (DO_TEX1 || DO_TEX2 || DO_TEX3) { PTEX_FALLBACK(); } diff --git a/src/mesa/drivers/dri/mga/Makefile b/src/mesa/drivers/dri/mga/Makefile index a871064c629..0cc329fb22d 100644 --- a/src/mesa/drivers/dri/mga/Makefile +++ b/src/mesa/drivers/dri/mga/Makefile @@ -31,4 +31,3 @@ ASM_SOURCES = include ../Makefile.template -symlinks: diff --git a/src/mesa/drivers/dri/mga/mga_texstate.c b/src/mesa/drivers/dri/mga/mga_texstate.c index ad765d1dd76..54eda62a96a 100644 --- a/src/mesa/drivers/dri/mga/mga_texstate.c +++ b/src/mesa/drivers/dri/mga/mga_texstate.c @@ -27,22 +27,19 @@ * Keith Whitwell <[email protected]> */ -#include <stdlib.h> +#include "main/context.h" +#include "main/enums.h" +#include "main/macros.h" #include "main/mm.h" +#include "main/imports.h" +#include "main/simple_list.h" + #include "mgacontext.h" #include "mgatex.h" #include "mgaregs.h" #include "mgatris.h" #include "mgaioctl.h" -#include "main/context.h" -#include "main/enums.h" -#include "main/macros.h" -#include "main/imports.h" - -#include "main/simple_list.h" -#include "main/texformat.h" - #define MGA_USE_TABLE_FOR_FORMAT #ifdef MGA_USE_TABLE_FOR_FORMAT #define TMC_nr_tformat (MESA_FORMAT_YCBCR_REV + 1) @@ -94,14 +91,14 @@ mgaSetTexImages( mgaContextPtr mmesa, return; } #else - if ( (baseImage->TexFormat->MesaFormat >= TMC_nr_tformat) - || (TMC_tformat[ baseImage->TexFormat->MesaFormat ] == 0) ) + if ( (baseImage->TexFormat >= TMC_nr_tformat) + || (TMC_tformat[ baseImage->TexFormat ] == 0) ) { _mesa_problem(NULL, "unexpected texture format in %s", __FUNCTION__); return; } - txformat = TMC_tformat[ baseImage->TexFormat->MesaFormat ]; + txformat = TMC_tformat[ baseImage->TexFormat ]; #endif /* MGA_USE_TABLE_FOR_FORMAT */ @@ -131,7 +128,7 @@ mgaSetTexImages( mgaContextPtr mmesa, break; size = texImage->Width * texImage->Height * - baseImage->TexFormat->TexelBytes; + _mesa_get_format_bytes(baseImage->TexFormat); t->offsets[i] = totalSize; t->base.dirty_images[0] |= (1<<i); diff --git a/src/mesa/drivers/dri/mga/mga_xmesa.c b/src/mesa/drivers/dri/mga/mga_xmesa.c index 0dc76fea50a..2c7f50c498e 100644 --- a/src/mesa/drivers/dri/mga/mga_xmesa.c +++ b/src/mesa/drivers/dri/mga/mga_xmesa.c @@ -78,7 +78,7 @@ #endif #define need_GL_APPLE_vertex_array_object #define need_GL_NV_vertex_program -#include "extension_helper.h" +#include "main/remap_helper.h" /* MGA configuration */ @@ -723,7 +723,7 @@ mgaCreateBuffer( __DRIscreenPrivate *driScrnPriv, { driRenderbuffer *frontRb - = driNewRenderbuffer(GL_RGBA, + = driNewRenderbuffer(MESA_FORMAT_ARGB8888, NULL, screen->cpp, screen->frontOffset, screen->frontPitch, @@ -734,7 +734,7 @@ mgaCreateBuffer( __DRIscreenPrivate *driScrnPriv, if (mesaVis->doubleBufferMode) { driRenderbuffer *backRb - = driNewRenderbuffer(GL_RGBA, + = driNewRenderbuffer(MESA_FORMAT_ARGB8888, NULL, screen->cpp, screen->backOffset, screen->backPitch, @@ -745,7 +745,7 @@ mgaCreateBuffer( __DRIscreenPrivate *driScrnPriv, if (mesaVis->depthBits == 16) { driRenderbuffer *depthRb - = driNewRenderbuffer(GL_DEPTH_COMPONENT16, + = driNewRenderbuffer(MESA_FORMAT_Z16, NULL, screen->cpp, screen->depthOffset, screen->depthPitch, @@ -757,7 +757,7 @@ mgaCreateBuffer( __DRIscreenPrivate *driScrnPriv, /* XXX is this right? */ if (mesaVis->stencilBits) { driRenderbuffer *depthRb - = driNewRenderbuffer(GL_DEPTH_COMPONENT24, + = driNewRenderbuffer(MESA_FORMAT_Z24_S8, NULL, screen->cpp, screen->depthOffset, screen->depthPitch, @@ -767,7 +767,7 @@ mgaCreateBuffer( __DRIscreenPrivate *driScrnPriv, } else { driRenderbuffer *depthRb - = driNewRenderbuffer(GL_DEPTH_COMPONENT32, + = driNewRenderbuffer(MESA_FORMAT_Z32, NULL, screen->cpp, screen->depthOffset, screen->depthPitch, @@ -778,7 +778,7 @@ mgaCreateBuffer( __DRIscreenPrivate *driScrnPriv, } else if (mesaVis->depthBits == 32) { driRenderbuffer *depthRb - = driNewRenderbuffer(GL_DEPTH_COMPONENT32, + = driNewRenderbuffer(MESA_FORMAT_Z32, NULL, screen->cpp, screen->depthOffset, screen->depthPitch, @@ -789,7 +789,7 @@ mgaCreateBuffer( __DRIscreenPrivate *driScrnPriv, if (mesaVis->stencilBits > 0 && !swStencil) { driRenderbuffer *stencilRb - = driNewRenderbuffer(GL_STENCIL_INDEX8_EXT, + = driNewRenderbuffer(MESA_FORMAT_S8, NULL, screen->cpp, screen->depthOffset, screen->depthPitch, @@ -945,22 +945,6 @@ static const __DRIconfig **mgaInitScreen(__DRIscreen *psp) return NULL; - /* Calling driInitExtensions here, with a NULL context pointer, - * does not actually enable the extensions. It just makes sure - * that all the dispatch offsets for all the extensions that - * *might* be enables are known. This is needed because the - * dispatch offsets need to be known when _mesa_context_create is - * called, but we can't enable the extensions until we have a - * context pointer. - * - * Hello chicken. Hello egg. How are you two today? - */ - - driInitExtensions( NULL, card_extensions, GL_FALSE ); - driInitExtensions( NULL, g400_extensions, GL_FALSE ); - driInitExtensions(NULL, ARB_vp_extensions, GL_FALSE); - driInitExtensions( NULL, NV_vp_extensions, GL_FALSE ); - if (!mgaInitDriver(psp)) return NULL; diff --git a/src/mesa/drivers/dri/mga/mgaspan.c b/src/mesa/drivers/dri/mga/mgaspan.c index 5b6d323ca98..2ff1cac8e24 100644 --- a/src/mesa/drivers/dri/mga/mgaspan.c +++ b/src/mesa/drivers/dri/mga/mgaspan.c @@ -206,24 +206,22 @@ void mgaDDInitSpanFuncs( GLcontext *ctx ) void mgaSetSpanFunctions(driRenderbuffer *drb, const GLvisual *vis) { - if (drb->Base.InternalFormat == GL_RGBA) { - if (vis->redBits == 5 && vis->greenBits == 6 && vis->blueBits == 5) { - mgaInitPointers_565(&drb->Base); - } - else { - mgaInitPointers_8888(&drb->Base); - } + if (drb->Base.Format == MESA_FORMAT_RGB565) { + mgaInitPointers_565(&drb->Base); } - else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT16) { + else if (drb->Base.Format == MESA_FORMAT_ARGB8888) { + mgaInitPointers_8888(&drb->Base); + } + else if (drb->Base.Format == MESA_FORMAT_Z16) { mgaInitDepthPointers_z16(&drb->Base); } - else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT24) { + else if (drb->Base.Format == MESA_FORMAT_Z24_S8) { mgaInitDepthPointers_z24_s8(&drb->Base); } - else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT32) { + else if (drb->Base.Format == MESA_FORMAT_Z32) { mgaInitDepthPointers_z32(&drb->Base); } - else if (drb->Base.InternalFormat == GL_STENCIL_INDEX8_EXT) { + else if (drb->Base.Format == MESA_FORMAT_S8) { mgaInitStencilPointers_z24_s8(&drb->Base); } } diff --git a/src/mesa/drivers/dri/mga/mgatex.c b/src/mesa/drivers/dri/mga/mgatex.c index 33eb0be4496..9163371b332 100644 --- a/src/mesa/drivers/dri/mga/mgatex.c +++ b/src/mesa/drivers/dri/mga/mgatex.c @@ -27,23 +27,22 @@ #include "main/glheader.h" #include "main/mm.h" -#include "mgacontext.h" -#include "mgatex.h" -#include "mgaregs.h" -#include "mgatris.h" -#include "mgaioctl.h" - #include "main/colormac.h" #include "main/context.h" #include "main/enums.h" #include "main/simple_list.h" #include "main/imports.h" #include "main/macros.h" -#include "main/texformat.h" #include "main/texstore.h" #include "main/teximage.h" #include "main/texobj.h" +#include "mgacontext.h" +#include "mgatex.h" +#include "mgaregs.h" +#include "mgatris.h" +#include "mgaioctl.h" + #include "swrast/swrast.h" #include "xmlpool.h" @@ -164,7 +163,7 @@ static void mgaSetTexBorderColor(mgaTextureObjectPtr t, const GLfloat color[4]) } -static const struct gl_texture_format * +static gl_format mgaChooseTextureFormat( GLcontext *ctx, GLint internalFormat, GLenum format, GLenum type ) { @@ -182,15 +181,15 @@ mgaChooseTextureFormat( GLcontext *ctx, GLint internalFormat, switch ( type ) { case GL_UNSIGNED_INT_10_10_10_2: case GL_UNSIGNED_INT_2_10_10_10_REV: - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb1555; + return do32bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB1555; case GL_UNSIGNED_SHORT_4_4_4_4: case GL_UNSIGNED_SHORT_4_4_4_4_REV: - return &_mesa_texformat_argb4444; + return MESA_FORMAT_ARGB4444; case GL_UNSIGNED_SHORT_5_5_5_1: case GL_UNSIGNED_SHORT_1_5_5_5_REV: - return &_mesa_texformat_argb1555; + return MESA_FORMAT_ARGB1555; default: - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + return do32bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB4444; } case 3: @@ -199,15 +198,15 @@ mgaChooseTextureFormat( GLcontext *ctx, GLint internalFormat, switch ( type ) { case GL_UNSIGNED_SHORT_4_4_4_4: case GL_UNSIGNED_SHORT_4_4_4_4_REV: - return &_mesa_texformat_argb4444; + return MESA_FORMAT_ARGB4444; case GL_UNSIGNED_SHORT_5_5_5_1: case GL_UNSIGNED_SHORT_1_5_5_5_REV: - return &_mesa_texformat_argb1555; + return MESA_FORMAT_ARGB1555; case GL_UNSIGNED_SHORT_5_6_5: case GL_UNSIGNED_SHORT_5_6_5_REV: - return &_mesa_texformat_rgb565; + return MESA_FORMAT_RGB565; default: - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565; + return do32bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_RGB565; } case GL_RGBA8: @@ -215,25 +214,25 @@ mgaChooseTextureFormat( GLcontext *ctx, GLint internalFormat, case GL_RGBA12: case GL_RGBA16: return !force16bpt ? - &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB4444; case GL_RGBA4: case GL_RGBA2: - return &_mesa_texformat_argb4444; + return MESA_FORMAT_ARGB4444; case GL_RGB5_A1: - return &_mesa_texformat_argb1555; + return MESA_FORMAT_ARGB1555; case GL_RGB8: case GL_RGB10: case GL_RGB12: case GL_RGB16: - return !force16bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565; + return !force16bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_RGB565; case GL_RGB5: case GL_RGB4: case GL_R3_G3_B2: - return &_mesa_texformat_rgb565; + return MESA_FORMAT_RGB565; case GL_ALPHA: case GL_ALPHA4: @@ -242,7 +241,7 @@ mgaChooseTextureFormat( GLcontext *ctx, GLint internalFormat, case GL_ALPHA16: case GL_COMPRESSED_ALPHA: /* FIXME: This will report incorrect component sizes... */ - return MGA_IS_G400(mmesa) ? &_mesa_texformat_al88 : &_mesa_texformat_argb4444; + return MGA_IS_G400(mmesa) ? MESA_FORMAT_AL88 : MESA_FORMAT_ARGB4444; case 1: case GL_LUMINANCE: @@ -252,7 +251,7 @@ mgaChooseTextureFormat( GLcontext *ctx, GLint internalFormat, case GL_LUMINANCE16: case GL_COMPRESSED_LUMINANCE: /* FIXME: This will report incorrect component sizes... */ - return MGA_IS_G400(mmesa) ? &_mesa_texformat_al88 : &_mesa_texformat_rgb565; + return MGA_IS_G400(mmesa) ? MESA_FORMAT_AL88 : MESA_FORMAT_RGB565; case 2: case GL_LUMINANCE_ALPHA: @@ -264,7 +263,7 @@ mgaChooseTextureFormat( GLcontext *ctx, GLint internalFormat, case GL_LUMINANCE16_ALPHA16: case GL_COMPRESSED_LUMINANCE_ALPHA: /* FIXME: This will report incorrect component sizes... */ - return MGA_IS_G400(mmesa) ? &_mesa_texformat_al88 : &_mesa_texformat_argb4444; + return MGA_IS_G400(mmesa) ? MESA_FORMAT_AL88 : MESA_FORMAT_ARGB4444; case GL_INTENSITY: case GL_INTENSITY4: @@ -273,15 +272,15 @@ mgaChooseTextureFormat( GLcontext *ctx, GLint internalFormat, case GL_INTENSITY16: case GL_COMPRESSED_INTENSITY: /* FIXME: This will report incorrect component sizes... */ - return MGA_IS_G400(mmesa) ? &_mesa_texformat_i8 : &_mesa_texformat_argb4444; + return MGA_IS_G400(mmesa) ? MESA_FORMAT_I8 : MESA_FORMAT_ARGB4444; case GL_YCBCR_MESA: if (MGA_IS_G400(mmesa) && (type == GL_UNSIGNED_SHORT_8_8_APPLE || type == GL_UNSIGNED_BYTE)) - return &_mesa_texformat_ycbcr; + return MESA_FORMAT_YCBCR; else - return &_mesa_texformat_ycbcr_rev; + return MESA_FORMAT_YCBCR_REV; case GL_COLOR_INDEX: case GL_COLOR_INDEX1_EXT: @@ -290,14 +289,14 @@ mgaChooseTextureFormat( GLcontext *ctx, GLint internalFormat, case GL_COLOR_INDEX8_EXT: case GL_COLOR_INDEX12_EXT: case GL_COLOR_INDEX16_EXT: - return &_mesa_texformat_ci8; + return MESA_FORMAT_CI8; default: _mesa_problem( ctx, "unexpected texture format in %s", __FUNCTION__ ); - return NULL; + return MESA_FORMAT_NONE; } - return NULL; /* never get here */ + return MESA_FORMAT_NONE; /* never get here */ } diff --git a/src/mesa/drivers/dri/mga/mgatexmem.c b/src/mesa/drivers/dri/mga/mgatexmem.c index 9a2d62b53b7..47be6f3c185 100644 --- a/src/mesa/drivers/dri/mga/mgatexmem.c +++ b/src/mesa/drivers/dri/mga/mgatexmem.c @@ -137,7 +137,7 @@ static void mgaUploadSubImage( mgaContextPtr mmesa, * directly used by the hardware for texturing. */ - texelBytes = texImage->TexFormat->TexelBytes; + texelBytes = _mesa_get_format_bytes(texImage->TexFormat); length = texImage->Width * texImage->Height * texelBytes; if ( t->base.heap->heapId == MGA_CARD_HEAP ) { unsigned tex_offset = 0; diff --git a/src/mesa/drivers/dri/r128/Makefile b/src/mesa/drivers/dri/r128/Makefile index 796dfbc5167..52c5a38a705 100644 --- a/src/mesa/drivers/dri/r128/Makefile +++ b/src/mesa/drivers/dri/r128/Makefile @@ -29,4 +29,3 @@ ASM_SOURCES = include ../Makefile.template -symlinks: diff --git a/src/mesa/drivers/dri/r128/r128_context.c b/src/mesa/drivers/dri/r128/r128_context.c index f511a67bada..0b250876c54 100644 --- a/src/mesa/drivers/dri/r128/r128_context.c +++ b/src/mesa/drivers/dri/r128/r128_context.c @@ -68,9 +68,9 @@ int R128_DEBUG = 0; #define need_GL_EXT_blend_minmax #define need_GL_EXT_fog_coord #define need_GL_EXT_secondary_color -#include "extension_helper.h" +#include "main/remap_helper.h" -const struct dri_extension card_extensions[] = +static const struct dri_extension card_extensions[] = { { "GL_ARB_multitexture", NULL }, { "GL_ARB_texture_env_add", NULL }, diff --git a/src/mesa/drivers/dri/r128/r128_screen.c b/src/mesa/drivers/dri/r128/r128_screen.c index f5bcc2f2906..9da3b5fb73d 100644 --- a/src/mesa/drivers/dri/r128/r128_screen.c +++ b/src/mesa/drivers/dri/r128/r128_screen.c @@ -74,8 +74,6 @@ static const GLuint __driNConfigOptions = 4; static const GLuint __driNConfigOptions = 3; #endif -extern const struct dri_extension card_extensions[]; - #if 1 /* Including xf86PciInfo.h introduces a bunch of errors... */ @@ -284,7 +282,7 @@ r128CreateBuffer( __DRIscreenPrivate *driScrnPriv, { driRenderbuffer *frontRb - = driNewRenderbuffer(GL_RGBA, + = driNewRenderbuffer(MESA_FORMAT_ARGB8888, NULL, screen->cpp, screen->frontOffset, screen->frontPitch, @@ -295,7 +293,7 @@ r128CreateBuffer( __DRIscreenPrivate *driScrnPriv, if (mesaVis->doubleBufferMode) { driRenderbuffer *backRb - = driNewRenderbuffer(GL_RGBA, + = driNewRenderbuffer(MESA_FORMAT_ARGB8888, NULL, screen->cpp, screen->backOffset, screen->backPitch, @@ -306,7 +304,7 @@ r128CreateBuffer( __DRIscreenPrivate *driScrnPriv, if (mesaVis->depthBits == 16) { driRenderbuffer *depthRb - = driNewRenderbuffer(GL_DEPTH_COMPONENT16, + = driNewRenderbuffer(MESA_FORMAT_Z16, NULL, screen->cpp, screen->depthOffset, screen->depthPitch, @@ -316,7 +314,7 @@ r128CreateBuffer( __DRIscreenPrivate *driScrnPriv, } else if (mesaVis->depthBits == 24) { driRenderbuffer *depthRb - = driNewRenderbuffer(GL_DEPTH_COMPONENT24, + = driNewRenderbuffer(MESA_FORMAT_S8_Z24, NULL, screen->cpp, screen->depthOffset, screen->depthPitch, @@ -327,7 +325,7 @@ r128CreateBuffer( __DRIscreenPrivate *driScrnPriv, if (mesaVis->stencilBits > 0 && !swStencil) { driRenderbuffer *stencilRb - = driNewRenderbuffer(GL_STENCIL_INDEX8_EXT, + = driNewRenderbuffer(MESA_FORMAT_S8, NULL, screen->cpp, screen->depthOffset, screen->depthPitch, @@ -493,18 +491,6 @@ r128InitScreen(__DRIscreenPrivate *psp) &psp->drm_version, & drm_expected ) ) return NULL; - /* Calling driInitExtensions here, with a NULL context pointer, - * does not actually enable the extensions. It just makes sure - * that all the dispatch offsets for all the extensions that - * *might* be enables are known. This is needed because the - * dispatch offsets need to be known when _mesa_context_create is - * called, but we can't enable the extensions until we have a - * context pointer. - * - * Hello chicken. Hello egg. How are you two today? - */ - driInitExtensions( NULL, card_extensions, GL_FALSE ); - if (!r128InitDriver(psp)) return NULL; diff --git a/src/mesa/drivers/dri/r128/r128_span.c b/src/mesa/drivers/dri/r128/r128_span.c index dd177e0def5..d238cc3c941 100644 --- a/src/mesa/drivers/dri/r128/r128_span.c +++ b/src/mesa/drivers/dri/r128/r128_span.c @@ -433,21 +433,19 @@ void r128DDInitSpanFuncs( GLcontext *ctx ) void r128SetSpanFunctions(driRenderbuffer *drb, const GLvisual *vis) { - if (drb->Base.InternalFormat == GL_RGBA) { - if (vis->redBits == 5 && vis->greenBits == 6 && vis->blueBits == 5) { - r128InitPointers_RGB565(&drb->Base); - } - else { - r128InitPointers_ARGB8888(&drb->Base); - } + if (drb->Base.Format == MESA_FORMAT_RGB565) { + r128InitPointers_RGB565(&drb->Base); } - else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT16) { + else if (drb->Base.Format == MESA_FORMAT_ARGB8888) { + r128InitPointers_ARGB8888(&drb->Base); + } + else if (drb->Base.Format == MESA_FORMAT_Z16) { r128InitDepthPointers_z16(&drb->Base); } - else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT24) { + else if (drb->Base.Format == MESA_FORMAT_S8_Z24) { r128InitDepthPointers_z24_s8(&drb->Base); } - else if (drb->Base.InternalFormat == GL_STENCIL_INDEX8_EXT) { + else if (drb->Base.Format == MESA_FORMAT_S8) { radeonInitStencilPointers_z24_s8(&drb->Base); } } diff --git a/src/mesa/drivers/dri/r128/r128_tex.c b/src/mesa/drivers/dri/r128/r128_tex.c index 0920270d7b6..0a1207fb89f 100644 --- a/src/mesa/drivers/dri/r128/r128_tex.c +++ b/src/mesa/drivers/dri/r128/r128_tex.c @@ -44,7 +44,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/simple_list.h" #include "main/enums.h" #include "main/texstore.h" -#include "main/texformat.h" #include "main/teximage.h" #include "main/texobj.h" #include "main/imports.h" @@ -178,7 +177,7 @@ static r128TexObjPtr r128AllocTexObj( struct gl_texture_object *texObj ) /* Called by the _mesa_store_teximage[123]d() functions. */ -static const struct gl_texture_format * +static gl_format r128ChooseTextureFormat( GLcontext *ctx, GLint internalFormat, GLenum format, GLenum type ) { @@ -282,13 +281,13 @@ r128ChooseTextureFormat( GLcontext *ctx, GLint internalFormat, case GL_YCBCR_MESA: if (type == GL_UNSIGNED_SHORT_8_8_APPLE || type == GL_UNSIGNED_BYTE) - return &_mesa_texformat_ycbcr; + return MESA_FORMAT_YCBCR; else - return &_mesa_texformat_ycbcr_rev; + return MESA_FORMAT_YCBCR_REV; default: _mesa_problem( ctx, "unexpected format in %s", __FUNCTION__ ); - return NULL; + return MESA_FORMAT_NONE; } } diff --git a/src/mesa/drivers/dri/r128/r128_texmem.c b/src/mesa/drivers/dri/r128/r128_texmem.c index 111fe1fd742..4ddcb86bcdc 100644 --- a/src/mesa/drivers/dri/r128/r128_texmem.c +++ b/src/mesa/drivers/dri/r128/r128_texmem.c @@ -41,7 +41,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/context.h" #include "main/macros.h" #include "main/simple_list.h" -#include "main/texformat.h" #include "main/imports.h" #define TEX_0 1 @@ -95,7 +94,7 @@ static void uploadSubImage( r128ContextPtr rmesa, r128TexObjPtr t, if ( !image ) return; - switch ( image->TexFormat->TexelBytes ) { + switch ( _mesa_get_format_bytes(image->TexFormat) ) { case 1: texelsPerDword = 4; break; case 2: texelsPerDword = 2; break; case 4: texelsPerDword = 1; break; @@ -215,9 +214,11 @@ static void uploadSubImage( r128ContextPtr rmesa, r128TexObjPtr t, /* Copy the next chunck of the texture image into the blit buffer */ { + const GLuint texelBytes = + _mesa_get_format_bytes(image->TexFormat); const GLubyte *src = (const GLubyte *) image->Data + - (y * image->Width + x) * image->TexFormat->TexelBytes; - const GLuint bytes = width * height * image->TexFormat->TexelBytes; + (y * image->Width + x) * texelBytes; + const GLuint bytes = width * height * texelBytes; memcpy(dst, src, bytes); } diff --git a/src/mesa/drivers/dri/r128/r128_texstate.c b/src/mesa/drivers/dri/r128/r128_texstate.c index a9c9568003b..cb2b5f95360 100644 --- a/src/mesa/drivers/dri/r128/r128_texstate.c +++ b/src/mesa/drivers/dri/r128/r128_texstate.c @@ -36,7 +36,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/imports.h" #include "main/context.h" #include "main/macros.h" -#include "main/texformat.h" #include "r128_context.h" #include "r128_state.h" @@ -61,7 +60,7 @@ static void r128SetTexImages( r128ContextPtr rmesa, if ( R128_DEBUG & DEBUG_VERBOSE_API ) fprintf( stderr, "%s( %p )\n", __FUNCTION__, (void *) tObj ); - switch (baseImage->TexFormat->MesaFormat) { + switch (baseImage->TexFormat) { case MESA_FORMAT_ARGB8888: case MESA_FORMAT_ARGB8888_REV: t->textureFormat = R128_DATATYPE_ARGB8888; @@ -123,7 +122,7 @@ static void r128SetTexImages( r128ContextPtr rmesa, totalSize += (tObj->Image[0][i]->Height * tObj->Image[0][i]->Width * - tObj->Image[0][i]->TexFormat->TexelBytes); + _mesa_get_format_bytes(tObj->Image[0][i]->TexFormat)); /* Offsets must be 32-byte aligned for host data blits and tiling */ totalSize = (totalSize + 31) & ~31; diff --git a/src/mesa/drivers/dri/r128/r128_tris.c b/src/mesa/drivers/dri/r128/r128_tris.c index 5b91271d740..448e34e047c 100644 --- a/src/mesa/drivers/dri/r128/r128_tris.c +++ b/src/mesa/drivers/dri/r128/r128_tris.c @@ -650,12 +650,12 @@ static void r128RenderStart( GLcontext *ctx ) } if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(rmesa->tmu_source[0]) )) { - if ( VB->TexCoordPtr[rmesa->tmu_source[0]]->size > 2 ) + if ( VB->AttribPtr[_TNL_ATTRIB_TEX0 + rmesa->tmu_source[0]]->size > 2 ) fallback_projtex = GL_TRUE; EMIT_ATTR( _TNL_ATTRIB_TEX0, EMIT_2F, R128_CCE_VC_FRMT_S_T, 8 ); } if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(rmesa->tmu_source[1]) )) { - if ( VB->TexCoordPtr[rmesa->tmu_source[1]]->size > 2 ) + if ( VB->AttribPtr[_TNL_ATTRIB_TEX0 + rmesa->tmu_source[1]]->size > 2 ) fallback_projtex = GL_TRUE; EMIT_ATTR( _TNL_ATTRIB_TEX1, EMIT_2F, R128_CCE_VC_FRMT_S2_T2, 8 ); } diff --git a/src/mesa/drivers/dri/r200/Makefile b/src/mesa/drivers/dri/r200/Makefile index fbce70c37bc..776f1e3f3f7 100644 --- a/src/mesa/drivers/dri/r200/Makefile +++ b/src/mesa/drivers/dri/r200/Makefile @@ -66,4 +66,3 @@ include ../Makefile.template #INCLUDES += -I../radeon/server -symlinks: diff --git a/src/mesa/drivers/dri/r200/r200_context.c b/src/mesa/drivers/dri/r200/r200_context.c index 3ddb5bf7d60..5f985d624d8 100644 --- a/src/mesa/drivers/dri/r200/r200_context.c +++ b/src/mesa/drivers/dri/r200/r200_context.c @@ -75,7 +75,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define need_GL_NV_vertex_program #define need_GL_ARB_point_parameters #define need_GL_EXT_framebuffer_object -#include "extension_helper.h" +#include "main/remap_helper.h" #define DRIVER_DATE "20060602" @@ -115,7 +115,7 @@ static const GLubyte *r200GetString( GLcontext *ctx, GLenum name ) /* Extension strings exported by the R200 driver. */ -const struct dri_extension card_extensions[] = +static const struct dri_extension card_extensions[] = { { "GL_ARB_multitexture", NULL }, { "GL_ARB_occlusion_query", GL_ARB_occlusion_query_functions}, @@ -146,31 +146,31 @@ const struct dri_extension card_extensions[] = { NULL, NULL } }; -const struct dri_extension blend_extensions[] = { +static const struct dri_extension blend_extensions[] = { { "GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions }, { "GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions }, { NULL, NULL } }; -const struct dri_extension ARB_vp_extension[] = { +static const struct dri_extension ARB_vp_extension[] = { { "GL_ARB_vertex_program", GL_ARB_vertex_program_functions } }; -const struct dri_extension NV_vp_extension[] = { +static const struct dri_extension NV_vp_extension[] = { { "GL_NV_vertex_program", GL_NV_vertex_program_functions } }; -const struct dri_extension ATI_fs_extension[] = { +static const struct dri_extension ATI_fs_extension[] = { { "GL_ATI_fragment_shader", GL_ATI_fragment_shader_functions } }; -const struct dri_extension point_extensions[] = { +static const struct dri_extension point_extensions[] = { { "GL_ARB_point_sprite", NULL }, { "GL_ARB_point_parameters", GL_ARB_point_parameters_functions }, { NULL, NULL } }; -const struct dri_extension mm_extensions[] = { +static const struct dri_extension mm_extensions[] = { { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions }, { NULL, NULL } }; @@ -325,9 +325,9 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual, _mesa_init_driver_functions(&functions); r200InitDriverFuncs(&functions); r200InitIoctlFuncs(&functions); - r200InitStateFuncs(&functions, screen->kernel_mm); + r200InitStateFuncs(&functions); r200InitTextureFuncs(&functions); - r200InitShaderFuncs(&functions); + r200InitShaderFuncs(&functions); radeonInitQueryObjFunctions(&functions); if (!radeonInitContext(&rmesa->radeon, &functions, diff --git a/src/mesa/drivers/dri/r200/r200_maos_arrays.c b/src/mesa/drivers/dri/r200/r200_maos_arrays.c index 383a0c4b0d3..249c0bbc11d 100644 --- a/src/mesa/drivers/dri/r200/r200_maos_arrays.c +++ b/src/mesa/drivers/dri/r200/r200_maos_arrays.c @@ -90,12 +90,14 @@ static void r200_emit_vecfog(GLcontext *ctx, struct radeon_aos *aos, aos->components = size; aos->count = count; + radeon_bo_map(aos->bo, 1); out = (uint32_t*)((char*)aos->bo->ptr + aos->offset); for (i = 0; i < count; i++) { out[0] = r200ComputeFogBlendFactor( ctx, *(GLfloat *)data ); out++; data += stride; } + radeon_bo_unmap(aos->bo); } /* Emit any changed arrays to new GART memory, re-emit a packet to diff --git a/src/mesa/drivers/dri/r200/r200_state.c b/src/mesa/drivers/dri/r200/r200_state.c index 76852e315c1..6d99c039ded 100644 --- a/src/mesa/drivers/dri/r200/r200_state.c +++ b/src/mesa/drivers/dri/r200/r200_state.c @@ -1578,13 +1578,6 @@ static void r200ClearStencil( GLcontext *ctx, GLint s ) * Window position and viewport transformation */ -/* - * To correctly position primitives: - */ -#define SUBPIXEL_X 0.125 -#define SUBPIXEL_Y 0.125 - - /** * Called when window size or position changes or viewport or depth range * state is changed. We update the hardware viewport state here. @@ -1609,9 +1602,9 @@ void r200UpdateWindow( GLcontext *ctx ) } float_ui32_type sx = { v[MAT_SX] }; - float_ui32_type tx = { v[MAT_TX] + xoffset + SUBPIXEL_X }; + float_ui32_type tx = { v[MAT_TX] + xoffset }; float_ui32_type sy = { v[MAT_SY] * y_scale }; - float_ui32_type ty = { (v[MAT_TY] * y_scale) + y_bias + SUBPIXEL_Y }; + float_ui32_type ty = { (v[MAT_TY] * y_scale) + y_bias }; float_ui32_type sz = { v[MAT_SZ] * depthScale }; float_ui32_type tz = { v[MAT_TZ] * depthScale }; @@ -1680,8 +1673,8 @@ void r200UpdateViewportOffset( GLcontext *ctx ) float_ui32_type tx; float_ui32_type ty; - tx.f = v[MAT_TX] + xoffset + SUBPIXEL_X; - ty.f = (- v[MAT_TY]) + yoffset + SUBPIXEL_Y; + tx.f = v[MAT_TX] + xoffset; + ty.f = (- v[MAT_TY]) + yoffset; if ( rmesa->hw.vpt.cmd[VPT_SE_VPORT_XOFFSET] != tx.ui32 || rmesa->hw.vpt.cmd[VPT_SE_VPORT_YOFFSET] != ty.ui32 ) @@ -2483,7 +2476,7 @@ static void r200PolygonStipple( GLcontext *ctx, const GLubyte *mask ) } /* Initialize the driver's state functions. */ -void r200InitStateFuncs( struct dd_function_table *functions, GLboolean dri2 ) +void r200InitStateFuncs( struct dd_function_table *functions ) { functions->UpdateState = r200InvalidateState; functions->LightingSpaceChange = r200LightingSpaceChange; @@ -2517,10 +2510,7 @@ void r200InitStateFuncs( struct dd_function_table *functions, GLboolean dri2 ) functions->LogicOpcode = r200LogicOpCode; functions->PolygonMode = r200PolygonMode; functions->PolygonOffset = r200PolygonOffset; - if (dri2) - functions->PolygonStipple = r200PolygonStipple; - else - functions->PolygonStipple = radeonPolygonStipplePreKMS; + functions->PolygonStipple = r200PolygonStipple; functions->PointParameterfv = r200PointParameter; functions->PointSize = r200PointSize; functions->RenderMode = r200RenderMode; diff --git a/src/mesa/drivers/dri/r200/r200_state.h b/src/mesa/drivers/dri/r200/r200_state.h index 9c62f0a6446..7b9b0c106aa 100644 --- a/src/mesa/drivers/dri/r200/r200_state.h +++ b/src/mesa/drivers/dri/r200/r200_state.h @@ -38,7 +38,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r200_context.h" extern void r200InitState( r200ContextPtr rmesa ); -extern void r200InitStateFuncs( struct dd_function_table *functions, GLboolean dri2 ); +extern void r200InitStateFuncs( struct dd_function_table *functions ); extern void r200InitTnlFuncs( GLcontext *ctx ); extern void r200UpdateMaterial( GLcontext *ctx ); diff --git a/src/mesa/drivers/dri/r200/r200_state_init.c b/src/mesa/drivers/dri/r200/r200_state_init.c index 7697306d889..6c5a0b79eed 100644 --- a/src/mesa/drivers/dri/r200/r200_state_init.c +++ b/src/mesa/drivers/dri/r200/r200_state_init.c @@ -529,16 +529,18 @@ static void ctx_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom) atom->cmd[CTX_RB3D_CNTL] &= ~(0xf << 10); if (rrb->cpp == 4) atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB8888; - else switch (rrb->base._ActualFormat) { - case GL_RGB5: + else switch (rrb->base.Format) { + case MESA_FORMAT_RGB565: atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_RGB565; break; - case GL_RGBA4: + case MESA_FORMAT_ARGB4444: atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB4444; break; - case GL_RGB5_A1: + case MESA_FORMAT_ARGB1555: atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB1555; break; + default: + _mesa_problem(ctx, "Unexpected format in ctx_emit_cs"); } cbpitch = (rrb->pitch / rrb->cpp); @@ -638,7 +640,7 @@ static void tex_emit(GLcontext *ctx, struct radeon_state_atom *atom) OUT_BATCH_TABLE(atom->cmd, 10); if (t && t->mt && !t->image_override) { - OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, 0, + OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, get_base_teximage_offset(t), RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0); } else if (!t) { /* workaround for old CS mechanism */ @@ -885,10 +887,8 @@ void r200InitState( r200ContextPtr rmesa ) } } } - /* polygon stipple is done with irq for non-kms */ - if (rmesa->radeon.radeonScreen->kernel_mm) { - ALLOC_STATE( stp, always, STP_STATE_SIZE, "STP/stp", 0 ); - } + + ALLOC_STATE( stp, always, STP_STATE_SIZE, "STP/stp", 0 ); for (i = 0; i < 6; i++) if (rmesa->radeon.radeonScreen->kernel_mm) @@ -1120,12 +1120,11 @@ void r200InitState( r200ContextPtr rmesa ) rmesa->hw.sci.cmd[SCI_CMD_1] = CP_PACKET0(R200_RE_TOP_LEFT, 0); rmesa->hw.sci.cmd[SCI_CMD_2] = CP_PACKET0(R200_RE_WIDTH_HEIGHT, 0); - if (rmesa->radeon.radeonScreen->kernel_mm) { - - rmesa->hw.stp.cmd[STP_CMD_0] = CP_PACKET0(RADEON_RE_STIPPLE_ADDR, 0); - rmesa->hw.stp.cmd[STP_DATA_0] = 0; - rmesa->hw.stp.cmd[STP_CMD_1] = CP_PACKET0_ONE(RADEON_RE_STIPPLE_DATA, 31); + rmesa->hw.stp.cmd[STP_CMD_0] = CP_PACKET0(RADEON_RE_STIPPLE_ADDR, 0); + rmesa->hw.stp.cmd[STP_DATA_0] = 0; + rmesa->hw.stp.cmd[STP_CMD_1] = CP_PACKET0_ONE(RADEON_RE_STIPPLE_DATA, 31); + if (rmesa->radeon.radeonScreen->kernel_mm) { rmesa->hw.mtl[0].emit = mtl_emit; rmesa->hw.mtl[1].emit = mtl_emit; diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.c b/src/mesa/drivers/dri/r200/r200_swtcl.c index 240fb450780..4596912ddc7 100644 --- a/src/mesa/drivers/dri/r200/r200_swtcl.c +++ b/src/mesa/drivers/dri/r200/r200_swtcl.c @@ -168,7 +168,7 @@ static void r200SetVertexFormat( GLcontext *ctx ) for (i = 0; i < ctx->Const.MaxTextureUnits; i++) { if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(i) )) { - GLuint sz = VB->TexCoordPtr[i]->size; + GLuint sz = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size; fmt_1 |= sz << (3 * i); EMIT_ATTR( _TNL_ATTRIB_TEX0+i, EMIT_1F + sz - 1, 0 ); @@ -297,7 +297,7 @@ void r200_swtcl_flush(GLcontext *ctx, uint32_t current_offset) radeonEmitState(&rmesa->radeon); r200EmitVertexAOS( rmesa, rmesa->radeon.swtcl.vertex_size, - first_elem(&rmesa->radeon.dma.reserved)->bo, + rmesa->radeon.swtcl.bo, current_offset); diff --git a/src/mesa/drivers/dri/r200/r200_tcl.c b/src/mesa/drivers/dri/r200/r200_tcl.c index c702910ef21..e7d48a7f29f 100644 --- a/src/mesa/drivers/dri/r200/r200_tcl.c +++ b/src/mesa/drivers/dri/r200/r200_tcl.c @@ -509,25 +509,26 @@ static GLboolean r200_run_tcl_render( GLcontext *ctx, prog to a not enabled output however, so just don't mess with it. We only need to change compsel. */ GLuint out_compsel = 0; - GLuint vp_out = rmesa->curr_vp_hw->mesa_program.Base.OutputsWritten; + const GLbitfield64 vp_out = + rmesa->curr_vp_hw->mesa_program.Base.OutputsWritten; vimap_rev = &rmesa->curr_vp_hw->inputmap_rev[0]; - assert(vp_out & (1 << VERT_RESULT_HPOS)); + assert(vp_out & BITFIELD64_BIT(VERT_RESULT_HPOS)); out_compsel = R200_OUTPUT_XYZW; - if (vp_out & (1 << VERT_RESULT_COL0)) { + if (vp_out & BITFIELD64_BIT(VERT_RESULT_COL0)) { out_compsel |= R200_OUTPUT_COLOR_0; } - if (vp_out & (1 << VERT_RESULT_COL1)) { + if (vp_out & BITFIELD64_BIT(VERT_RESULT_COL1)) { out_compsel |= R200_OUTPUT_COLOR_1; } - if (vp_out & (1 << VERT_RESULT_FOGC)) { + if (vp_out & BITFIELD64_BIT(VERT_RESULT_FOGC)) { out_compsel |= R200_OUTPUT_DISCRETE_FOG; } - if (vp_out & (1 << VERT_RESULT_PSIZ)) { + if (vp_out & BITFIELD64_BIT(VERT_RESULT_PSIZ)) { out_compsel |= R200_OUTPUT_PT_SIZE; } for (i = VERT_RESULT_TEX0; i < VERT_RESULT_TEX6; i++) { - if (vp_out & (1 << i)) { + if (vp_out & BITFIELD64_BIT(i)) { out_compsel |= R200_OUTPUT_TEX_0 << (i - VERT_RESULT_TEX0); } } diff --git a/src/mesa/drivers/dri/r200/r200_tex.c b/src/mesa/drivers/dri/r200/r200_tex.c index 36d9e37d875..a4177215537 100644 --- a/src/mesa/drivers/dri/r200/r200_tex.c +++ b/src/mesa/drivers/dri/r200/r200_tex.c @@ -38,7 +38,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/enums.h" #include "main/image.h" #include "main/simple_list.h" -#include "main/texformat.h" #include "main/texstore.h" #include "main/teximage.h" #include "main/texobj.h" @@ -386,16 +385,7 @@ static void r200TexParameter( GLcontext *ctx, GLenum target, case GL_TEXTURE_MAX_LEVEL: case GL_TEXTURE_MIN_LOD: case GL_TEXTURE_MAX_LOD: - /* This isn't the most efficient solution but there doesn't appear to - * be a nice alternative. Since there's no LOD clamping, - * we just have to rely on loading the right subset of mipmap levels - * to simulate a clamped LOD. - */ - if (t->mt) { - radeon_miptree_unreference(t->mt); - t->mt = 0; - t->validated = GL_FALSE; - } + t->validated = GL_FALSE; break; default: @@ -414,7 +404,7 @@ static void r200DeleteTexture(GLcontext * ctx, struct gl_texture_object *texObj) (void *)texObj, _mesa_lookup_enum_by_nr(texObj->Target)); } - + if (rmesa) { int i; radeon_firevertices(&rmesa->radeon); @@ -426,11 +416,9 @@ static void r200DeleteTexture(GLcontext * ctx, struct gl_texture_object *texObj) } } } - - if (t->mt) { - radeon_miptree_unreference(t->mt); - t->mt = 0; - } + + radeon_miptree_unreference(&t->mt); + _mesa_delete_texture_object(ctx, texObj); } diff --git a/src/mesa/drivers/dri/r200/r200_texstate.c b/src/mesa/drivers/dri/r200/r200_texstate.c index c94834752ed..e2f9cf0ea86 100644 --- a/src/mesa/drivers/dri/r200/r200_texstate.c +++ b/src/mesa/drivers/dri/r200/r200_texstate.c @@ -36,7 +36,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/imports.h" #include "main/context.h" #include "main/macros.h" -#include "main/texformat.h" #include "main/teximage.h" #include "main/texobj.h" #include "main/enums.h" @@ -798,24 +797,13 @@ void r200SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_fo return; } - radeon_update_renderbuffers(pDRICtx, dPriv); - /* back & depth buffer are useless free them right away */ - rb = (void*)rfb->base.Attachment[BUFFER_DEPTH].Renderbuffer; - if (rb && rb->bo) { - radeon_bo_unref(rb->bo); - rb->bo = NULL; - } - rb = (void*)rfb->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer; - if (rb && rb->bo) { - radeon_bo_unref(rb->bo); - rb->bo = NULL; - } + radeon_update_renderbuffers(pDRICtx, dPriv, GL_TRUE); rb = rfb->color_rb[0]; if (rb->bo == NULL) { /* Failed to BO for the buffer */ return; } - + _mesa_lock_texture(radeon->glCtx, texObj); if (t->bo) { radeon_bo_unref(t->bo); @@ -825,20 +813,14 @@ void r200SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_fo radeon_bo_unref(rImage->bo); rImage->bo = NULL; } - if (t->mt) { - radeon_miptree_unreference(t->mt); - t->mt = NULL; - } - if (rImage->mt) { - radeon_miptree_unreference(rImage->mt); - rImage->mt = NULL; - } + + radeon_miptree_unreference(&t->mt); + radeon_miptree_unreference(&rImage->mt); + _mesa_init_teximage_fields(radeon->glCtx, target, texImage, rb->base.Width, rb->base.Height, 1, 0, rb->cpp); texImage->RowStride = rb->pitch / rb->cpp; - texImage->TexFormat = radeonChooseTextureFormat(radeon->glCtx, - internalFormat, - type, format, 0); + rImage->bo = rb->bo; radeon_bo_ref(rImage->bo); t->bo = rb->bo; @@ -1426,10 +1408,9 @@ void set_re_cntl_d3d( GLcontext *ctx, int unit, GLboolean use_d3d ) */ static void setup_hardware_state(r200ContextPtr rmesa, radeonTexObj *t) { - int firstlevel = t->mt ? t->mt->firstLevel : 0; - const struct gl_texture_image *firstImage = t->base.Image[0][firstlevel]; + const struct gl_texture_image *firstImage = t->base.Image[0][t->minLod]; GLint log2Width, log2Height, log2Depth, texelBytes; - + if ( t->bo ) { return; } @@ -1437,11 +1418,11 @@ static void setup_hardware_state(r200ContextPtr rmesa, radeonTexObj *t) log2Width = firstImage->WidthLog2; log2Height = firstImage->HeightLog2; log2Depth = firstImage->DepthLog2; - texelBytes = firstImage->TexFormat->TexelBytes; + texelBytes = _mesa_get_format_bytes(firstImage->TexFormat); if (!t->image_override) { - if (VALID_FORMAT(firstImage->TexFormat->MesaFormat)) { + if (VALID_FORMAT(firstImage->TexFormat)) { const struct tx_table *table = _mesa_little_endian() ? tx_table_le : tx_table_be; @@ -1449,17 +1430,17 @@ static void setup_hardware_state(r200ContextPtr rmesa, radeonTexObj *t) R200_TXFORMAT_ALPHA_IN_MAP); t->pp_txfilter &= ~R200_YUV_TO_RGB; - t->pp_txformat |= table[ firstImage->TexFormat->MesaFormat ].format; - t->pp_txfilter |= table[ firstImage->TexFormat->MesaFormat ].filter; + t->pp_txformat |= table[ firstImage->TexFormat ].format; + t->pp_txfilter |= table[ firstImage->TexFormat ].filter; } else { _mesa_problem(NULL, "unexpected texture format in %s", __FUNCTION__); return; } } - + t->pp_txfilter &= ~R200_MAX_MIP_LEVEL_MASK; - t->pp_txfilter |= (t->mt->lastLevel - t->mt->firstLevel) << R200_MAX_MIP_LEVEL_SHIFT; + t->pp_txfilter |= (t->maxLod - t->minLod) << R200_MAX_MIP_LEVEL_SHIFT; t->pp_txformat &= ~(R200_TXFORMAT_WIDTH_MASK | R200_TXFORMAT_HEIGHT_MASK | @@ -1504,7 +1485,7 @@ static void setup_hardware_state(r200ContextPtr rmesa, radeonTexObj *t) | ((firstImage->Height - 1) << R200_PP_TX_HEIGHTMASK_SHIFT)); if ( !t->image_override ) { - if (firstImage->IsCompressed) + if (_mesa_is_format_compressed(firstImage->TexFormat)) t->pp_txpitch = (firstImage->Width + 63) & ~(63); else t->pp_txpitch = ((firstImage->Width * texelBytes) + 63) & ~(63); diff --git a/src/mesa/drivers/dri/r300/Makefile b/src/mesa/drivers/dri/r300/Makefile index fe775eac99a..409d126ab2b 100644 --- a/src/mesa/drivers/dri/r300/Makefile +++ b/src/mesa/drivers/dri/r300/Makefile @@ -43,17 +43,19 @@ RADEON_COMMON_SOURCES = \ DRIVER_SOURCES = \ radeon_screen.c \ + r300_blit.c \ r300_context.c \ r300_draw.c \ - r300_ioctl.c \ r300_cmdbuf.c \ r300_state.c \ r300_render.c \ r300_tex.c \ + r300_texcopy.c \ r300_texstate.c \ r300_vertprog.c \ r300_fragprog_common.c \ r300_shader.c \ + radeon_mesa_to_rc.c \ r300_emit.c \ r300_swtcl.c \ $(RADEON_COMMON_SOURCES) \ @@ -68,16 +70,12 @@ DRIVER_DEFINES = -DRADEON_R300 DRI_LIB_DEPS += $(RADEON_LDFLAGS) -PIPE_DRIVERS = compiler/libr300compiler.a +SUBDIRS = compiler -##### TARGETS ##### +EXTRA_MODULES = compiler/libr300compiler.a -include ../Makefile.template -symlinks: +##### TARGETS ##### -# Mark the archive phony so that we always check for recompilation -.PHONY : compiler/libr300compiler.a +include ../Makefile.template -compiler/libr300compiler.a: - cd compiler && $(MAKE) diff --git a/src/mesa/drivers/dri/r300/compiler/Makefile b/src/mesa/drivers/dri/r300/compiler/Makefile index d9738441928..d83888d90a3 100644 --- a/src/mesa/drivers/dri/r300/compiler/Makefile +++ b/src/mesa/drivers/dri/r300/compiler/Makefile @@ -8,10 +8,17 @@ LIBNAME = r300compiler C_SOURCES = \ radeon_code.c \ radeon_compiler.c \ - radeon_nqssadce.c \ radeon_program.c \ + radeon_program_print.c \ + radeon_opcodes.c \ radeon_program_alu.c \ radeon_program_pair.c \ + radeon_pair_translate.c \ + radeon_pair_schedule.c \ + radeon_pair_regalloc.c \ + radeon_dataflow.c \ + radeon_dataflow_deadcode.c \ + radeon_dataflow_swizzles.c \ r3xx_fragprog.c \ r300_fragprog.c \ r300_fragprog_swizzle.c \ diff --git a/src/mesa/drivers/dri/r300/compiler/SConscript b/src/mesa/drivers/dri/r300/compiler/SConscript new file mode 100755 index 00000000000..46075a8aee9 --- /dev/null +++ b/src/mesa/drivers/dri/r300/compiler/SConscript @@ -0,0 +1,37 @@ +Import('*') + +env = env.Clone() +env.Append(CPPPATH = '#/include') +env.Append(CPPPATH = '#/src/mesa') + +# temporary fix +env['CFLAGS'] = str(env['CFLAGS']).replace('-Werror=declaration-after-statement', '') + +r300compiler = env.ConvenienceLibrary( + target = 'r300compiler', + source = [ + 'radeon_code.c', + 'radeon_compiler.c', + 'radeon_program.c', + 'radeon_program_print.c', + 'radeon_opcodes.c', + 'radeon_program_alu.c', + 'radeon_program_pair.c', + 'radeon_pair_translate.c', + 'radeon_pair_schedule.c', + 'radeon_pair_regalloc.c', + 'radeon_dataflow.c', + 'radeon_dataflow_deadcode.c', + 'radeon_dataflow_swizzles.c', + 'r3xx_fragprog.c', + 'r300_fragprog.c', + 'r300_fragprog_swizzle.c', + 'r300_fragprog_emit.c', + 'r500_fragprog.c', + 'r500_fragprog_emit.c', + 'r3xx_vertprog.c', + 'r3xx_vertprog_dump.c', + 'memory_pool.c', + ]) + +Return('r300compiler') diff --git a/src/mesa/drivers/dri/r300/compiler/r300_fragprog.c b/src/mesa/drivers/dri/r300/compiler/r300_fragprog.c index 6c9fba49146..aa69b0fc72b 100644 --- a/src/mesa/drivers/dri/r300/compiler/r300_fragprog.c +++ b/src/mesa/drivers/dri/r300/compiler/r300_fragprog.c @@ -27,17 +27,17 @@ #include "r300_fragprog.h" -#include "shader/prog_parameter.h" +#include <stdio.h> #include "../r300_reg.h" -static struct prog_src_register shadow_ambient(struct radeon_compiler * c, int tmu) +static struct rc_src_register shadow_ambient(struct radeon_compiler * c, int tmu) { - struct prog_src_register reg = { 0, }; + struct rc_src_register reg = { 0, }; - reg.File = PROGRAM_STATE_VAR; + reg.File = RC_FILE_CONSTANT; reg.Index = rc_constants_add_state(&c->Program.Constants, RC_STATE_SHADOW_AMBIENT, tmu); - reg.Swizzle = SWIZZLE_WWWW; + reg.Swizzle = RC_SWIZZLE_WWWW; return reg; } @@ -47,7 +47,7 @@ static struct prog_src_register shadow_ambient(struct radeon_compiler * c, int t * - extract operand swizzles * - introduce a temporary register when write masks are needed */ -GLboolean r300_transform_TEX( +int r300_transform_TEX( struct radeon_compiler * c, struct rc_instruction* inst, void* data) @@ -55,77 +55,77 @@ GLboolean r300_transform_TEX( struct r300_fragment_program_compiler *compiler = (struct r300_fragment_program_compiler*)data; - if (inst->I.Opcode != OPCODE_TEX && - inst->I.Opcode != OPCODE_TXB && - inst->I.Opcode != OPCODE_TXP && - inst->I.Opcode != OPCODE_KIL) - return GL_FALSE; + if (inst->U.I.Opcode != RC_OPCODE_TEX && + inst->U.I.Opcode != RC_OPCODE_TXB && + inst->U.I.Opcode != RC_OPCODE_TXP && + inst->U.I.Opcode != RC_OPCODE_KIL) + return 0; /* ARB_shadow & EXT_shadow_funcs */ - if (inst->I.Opcode != OPCODE_KIL && - c->Program.ShadowSamplers & (1 << inst->I.TexSrcUnit)) { - GLuint comparefunc = GL_NEVER + compiler->state.unit[inst->I.TexSrcUnit].texture_compare_func; + if (inst->U.I.Opcode != RC_OPCODE_KIL && + c->Program.ShadowSamplers & (1 << inst->U.I.TexSrcUnit)) { + rc_compare_func comparefunc = compiler->state.unit[inst->U.I.TexSrcUnit].texture_compare_func; - if (comparefunc == GL_NEVER || comparefunc == GL_ALWAYS) { - inst->I.Opcode = OPCODE_MOV; + if (comparefunc == RC_COMPARE_FUNC_NEVER || comparefunc == RC_COMPARE_FUNC_ALWAYS) { + inst->U.I.Opcode = RC_OPCODE_MOV; - if (comparefunc == GL_ALWAYS) { - inst->I.SrcReg[0].File = PROGRAM_BUILTIN; - inst->I.SrcReg[0].Swizzle = SWIZZLE_1111; + if (comparefunc == RC_COMPARE_FUNC_ALWAYS) { + inst->U.I.SrcReg[0].File = RC_FILE_NONE; + inst->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_1111; } else { - inst->I.SrcReg[0] = shadow_ambient(c, inst->I.TexSrcUnit); + inst->U.I.SrcReg[0] = shadow_ambient(c, inst->U.I.TexSrcUnit); } - return GL_TRUE; + return 1; } else { - GLuint comparefunc = GL_NEVER + compiler->state.unit[inst->I.TexSrcUnit].texture_compare_func; - GLuint depthmode = compiler->state.unit[inst->I.TexSrcUnit].depth_texture_mode; + rc_compare_func comparefunc = compiler->state.unit[inst->U.I.TexSrcUnit].texture_compare_func; + unsigned int depthmode = compiler->state.unit[inst->U.I.TexSrcUnit].depth_texture_mode; struct rc_instruction * inst_rcp = rc_insert_new_instruction(c, inst); struct rc_instruction * inst_mad = rc_insert_new_instruction(c, inst_rcp); struct rc_instruction * inst_cmp = rc_insert_new_instruction(c, inst_mad); int pass, fail; - inst_rcp->I.Opcode = OPCODE_RCP; - inst_rcp->I.DstReg.File = PROGRAM_TEMPORARY; - inst_rcp->I.DstReg.Index = rc_find_free_temporary(c); - inst_rcp->I.DstReg.WriteMask = WRITEMASK_W; - inst_rcp->I.SrcReg[0] = inst->I.SrcReg[0]; - inst_rcp->I.SrcReg[0].Swizzle = SWIZZLE_WWWW; - - inst_cmp->I.DstReg = inst->I.DstReg; - inst->I.DstReg.File = PROGRAM_TEMPORARY; - inst->I.DstReg.Index = rc_find_free_temporary(c); - inst->I.DstReg.WriteMask = WRITEMASK_XYZW; - - inst_mad->I.Opcode = OPCODE_MAD; - inst_mad->I.DstReg.File = PROGRAM_TEMPORARY; - inst_mad->I.DstReg.Index = rc_find_free_temporary(c); - inst_mad->I.SrcReg[0] = inst->I.SrcReg[0]; - inst_mad->I.SrcReg[0].Swizzle = SWIZZLE_ZZZZ; - inst_mad->I.SrcReg[1].File = PROGRAM_TEMPORARY; - inst_mad->I.SrcReg[1].Index = inst_rcp->I.DstReg.Index; - inst_mad->I.SrcReg[1].Swizzle = SWIZZLE_WWWW; - inst_mad->I.SrcReg[2].File = PROGRAM_TEMPORARY; - inst_mad->I.SrcReg[2].Index = inst->I.DstReg.Index; + inst_rcp->U.I.Opcode = RC_OPCODE_RCP; + inst_rcp->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst_rcp->U.I.DstReg.Index = rc_find_free_temporary(c); + inst_rcp->U.I.DstReg.WriteMask = RC_MASK_W; + inst_rcp->U.I.SrcReg[0] = inst->U.I.SrcReg[0]; + inst_rcp->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_WWWW; + + inst_cmp->U.I.DstReg = inst->U.I.DstReg; + inst->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst->U.I.DstReg.Index = rc_find_free_temporary(c); + inst->U.I.DstReg.WriteMask = RC_MASK_XYZW; + + inst_mad->U.I.Opcode = RC_OPCODE_MAD; + inst_mad->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst_mad->U.I.DstReg.Index = rc_find_free_temporary(c); + inst_mad->U.I.SrcReg[0] = inst->U.I.SrcReg[0]; + inst_mad->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_ZZZZ; + inst_mad->U.I.SrcReg[1].File = RC_FILE_TEMPORARY; + inst_mad->U.I.SrcReg[1].Index = inst_rcp->U.I.DstReg.Index; + inst_mad->U.I.SrcReg[1].Swizzle = RC_SWIZZLE_WWWW; + inst_mad->U.I.SrcReg[2].File = RC_FILE_TEMPORARY; + inst_mad->U.I.SrcReg[2].Index = inst->U.I.DstReg.Index; if (depthmode == 0) /* GL_LUMINANCE */ - inst_mad->I.SrcReg[2].Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_Z); + inst_mad->U.I.SrcReg[2].Swizzle = RC_MAKE_SWIZZLE(RC_SWIZZLE_X, RC_SWIZZLE_Y, RC_SWIZZLE_Z, RC_SWIZZLE_Z); else if (depthmode == 2) /* GL_ALPHA */ - inst_mad->I.SrcReg[2].Swizzle = SWIZZLE_WWWW; + inst_mad->U.I.SrcReg[2].Swizzle = RC_SWIZZLE_WWWW; /* Recall that SrcReg[0] is tex, SrcReg[2] is r and: * r < tex <=> -tex+r < 0 * r >= tex <=> not (-tex+r < 0 */ - if (comparefunc == GL_LESS || comparefunc == GL_GEQUAL) - inst_mad->I.SrcReg[2].Negate = inst_mad->I.SrcReg[2].Negate ^ NEGATE_XYZW; + if (comparefunc == RC_COMPARE_FUNC_LESS || comparefunc == RC_COMPARE_FUNC_GEQUAL) + inst_mad->U.I.SrcReg[2].Negate = inst_mad->U.I.SrcReg[2].Negate ^ RC_MASK_XYZW; else - inst_mad->I.SrcReg[0].Negate = inst_mad->I.SrcReg[0].Negate ^ NEGATE_XYZW; + inst_mad->U.I.SrcReg[0].Negate = inst_mad->U.I.SrcReg[0].Negate ^ RC_MASK_XYZW; - inst_cmp->I.Opcode = OPCODE_CMP; + inst_cmp->U.I.Opcode = RC_OPCODE_CMP; /* DstReg has been filled out above */ - inst_cmp->I.SrcReg[0].File = PROGRAM_TEMPORARY; - inst_cmp->I.SrcReg[0].Index = inst_mad->I.DstReg.Index; + inst_cmp->U.I.SrcReg[0].File = RC_FILE_TEMPORARY; + inst_cmp->U.I.SrcReg[0].Index = inst_mad->U.I.DstReg.Index; - if (comparefunc == GL_LESS || comparefunc == GL_GREATER) { + if (comparefunc == RC_COMPARE_FUNC_LESS || comparefunc == RC_COMPARE_FUNC_GREATER) { pass = 1; fail = 2; } else { @@ -133,9 +133,9 @@ GLboolean r300_transform_TEX( fail = 1; } - inst_cmp->I.SrcReg[pass].File = PROGRAM_BUILTIN; - inst_cmp->I.SrcReg[pass].Swizzle = SWIZZLE_1111; - inst_cmp->I.SrcReg[fail] = shadow_ambient(c, inst->I.TexSrcUnit); + inst_cmp->U.I.SrcReg[pass].File = RC_FILE_NONE; + inst_cmp->U.I.SrcReg[pass].Swizzle = RC_SWIZZLE_1111; + inst_cmp->U.I.SrcReg[fail] = shadow_ambient(c, inst->U.I.TexSrcUnit); } } @@ -143,52 +143,52 @@ GLboolean r300_transform_TEX( * instead of [0..Width]x[0..Height]. * Add a scaling instruction. */ - if (inst->I.Opcode != OPCODE_KIL && inst->I.TexSrcTarget == TEXTURE_RECT_INDEX) { + if (inst->U.I.Opcode != RC_OPCODE_KIL && inst->U.I.TexSrcTarget == RC_TEXTURE_RECT) { struct rc_instruction * inst_mul = rc_insert_new_instruction(c, inst->Prev); - inst_mul->I.Opcode = OPCODE_MUL; - inst_mul->I.DstReg.File = PROGRAM_TEMPORARY; - inst_mul->I.DstReg.Index = rc_find_free_temporary(c); - inst_mul->I.SrcReg[0] = inst->I.SrcReg[0]; - inst_mul->I.SrcReg[1].File = PROGRAM_STATE_VAR; - inst_mul->I.SrcReg[1].Index = rc_constants_add_state(&c->Program.Constants, RC_STATE_R300_TEXRECT_FACTOR, inst->I.TexSrcUnit); + inst_mul->U.I.Opcode = RC_OPCODE_MUL; + inst_mul->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst_mul->U.I.DstReg.Index = rc_find_free_temporary(c); + inst_mul->U.I.SrcReg[0] = inst->U.I.SrcReg[0]; + inst_mul->U.I.SrcReg[1].File = RC_FILE_CONSTANT; + inst_mul->U.I.SrcReg[1].Index = rc_constants_add_state(&c->Program.Constants, RC_STATE_R300_TEXRECT_FACTOR, inst->U.I.TexSrcUnit); - reset_srcreg(&inst->I.SrcReg[0]); - inst->I.SrcReg[0].File = PROGRAM_TEMPORARY; - inst->I.SrcReg[0].Index = inst_mul->I.DstReg.Index; + reset_srcreg(&inst->U.I.SrcReg[0]); + inst->U.I.SrcReg[0].File = RC_FILE_TEMPORARY; + inst->U.I.SrcReg[0].Index = inst_mul->U.I.DstReg.Index; } /* Cannot write texture to output registers or with masks */ - if (inst->I.Opcode != OPCODE_KIL && - (inst->I.DstReg.File != PROGRAM_TEMPORARY || inst->I.DstReg.WriteMask != WRITEMASK_XYZW)) { + if (inst->U.I.Opcode != RC_OPCODE_KIL && + (inst->U.I.DstReg.File != RC_FILE_TEMPORARY || inst->U.I.DstReg.WriteMask != RC_MASK_XYZW)) { struct rc_instruction * inst_mov = rc_insert_new_instruction(c, inst); - inst_mov->I.Opcode = OPCODE_MOV; - inst_mov->I.DstReg = inst->I.DstReg; - inst_mov->I.SrcReg[0].File = PROGRAM_TEMPORARY; - inst_mov->I.SrcReg[0].Index = rc_find_free_temporary(c); + inst_mov->U.I.Opcode = RC_OPCODE_MOV; + inst_mov->U.I.DstReg = inst->U.I.DstReg; + inst_mov->U.I.SrcReg[0].File = RC_FILE_TEMPORARY; + inst_mov->U.I.SrcReg[0].Index = rc_find_free_temporary(c); - inst->I.DstReg.File = PROGRAM_TEMPORARY; - inst->I.DstReg.Index = inst_mov->I.SrcReg[0].Index; - inst->I.DstReg.WriteMask = WRITEMASK_XYZW; + inst->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst->U.I.DstReg.Index = inst_mov->U.I.SrcReg[0].Index; + inst->U.I.DstReg.WriteMask = RC_MASK_XYZW; } /* Cannot read texture coordinate from constants file */ - if (inst->I.SrcReg[0].File != PROGRAM_TEMPORARY && inst->I.SrcReg[0].File != PROGRAM_INPUT) { + if (inst->U.I.SrcReg[0].File != RC_FILE_TEMPORARY && inst->U.I.SrcReg[0].File != RC_FILE_INPUT) { struct rc_instruction * inst_mov = rc_insert_new_instruction(c, inst->Prev); - inst_mov->I.Opcode = OPCODE_MOV; - inst_mov->I.DstReg.File = PROGRAM_TEMPORARY; - inst_mov->I.DstReg.Index = rc_find_free_temporary(c); - inst_mov->I.SrcReg[0] = inst->I.SrcReg[0]; + inst_mov->U.I.Opcode = RC_OPCODE_MOV; + inst_mov->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst_mov->U.I.DstReg.Index = rc_find_free_temporary(c); + inst_mov->U.I.SrcReg[0] = inst->U.I.SrcReg[0]; - reset_srcreg(&inst->I.SrcReg[0]); - inst->I.SrcReg[0].File = PROGRAM_TEMPORARY; - inst->I.SrcReg[0].Index = inst_mov->I.DstReg.Index; + reset_srcreg(&inst->U.I.SrcReg[0]); + inst->U.I.SrcReg[0].File = RC_FILE_TEMPORARY; + inst->U.I.SrcReg[0].Index = inst_mov->U.I.DstReg.Index; } - return GL_TRUE; + return 1; } /* just some random things... */ diff --git a/src/mesa/drivers/dri/r300/compiler/r300_fragprog.h b/src/mesa/drivers/dri/r300/compiler/r300_fragprog.h index 0ac46dbd9cb..418df36c936 100644 --- a/src/mesa/drivers/dri/r300/compiler/r300_fragprog.h +++ b/src/mesa/drivers/dri/r300/compiler/r300_fragprog.h @@ -33,9 +33,6 @@ #ifndef __R300_FRAGPROG_H_ #define __R300_FRAGPROG_H_ -#include "shader/program.h" -#include "shader/prog_instruction.h" - #include "radeon_compiler.h" #include "radeon_program.h" @@ -44,6 +41,6 @@ extern void r300BuildFragmentProgramHwCode(struct r300_fragment_program_compiler extern void r300FragmentProgramDump(struct rX00_fragment_program_code *c); -extern GLboolean r300_transform_TEX(struct radeon_compiler * c, struct rc_instruction* inst, void* data); +extern int r300_transform_TEX(struct radeon_compiler * c, struct rc_instruction* inst, void* data); #endif diff --git a/src/mesa/drivers/dri/r300/compiler/r300_fragprog_emit.c b/src/mesa/drivers/dri/r300/compiler/r300_fragprog_emit.c index 305dc074ee8..375838d98e7 100644 --- a/src/mesa/drivers/dri/r300/compiler/r300_fragprog_emit.c +++ b/src/mesa/drivers/dri/r300/compiler/r300_fragprog_emit.c @@ -56,7 +56,6 @@ struct r300_emit_state { }; #define PROG_CODE \ - struct r300_emit_state * emit = (struct r300_emit_state*)data; \ struct r300_fragment_program_compiler *c = emit->compiler; \ struct r300_fragment_program_code *code = &c->code->code.r300 @@ -69,64 +68,76 @@ struct r300_emit_state { /** * Mark a temporary register as used. */ -static void use_temporary(struct r300_fragment_program_code *code, GLuint index) +static void use_temporary(struct r300_fragment_program_code *code, unsigned int index) { if (index > code->pixsize) code->pixsize = index; } +static unsigned int use_source(struct r300_fragment_program_code* code, struct radeon_pair_instruction_source src) +{ + if (src.File == RC_FILE_CONSTANT) { + return src.Index | (1 << 5); + } else if (src.File == RC_FILE_TEMPORARY) { + use_temporary(code, src.Index); + return src.Index; + } + + return 0; +} + -static GLuint translate_rgb_opcode(struct r300_fragment_program_compiler * c, GLuint opcode) +static unsigned int translate_rgb_opcode(struct r300_fragment_program_compiler * c, rc_opcode opcode) { switch(opcode) { - case OPCODE_CMP: return R300_ALU_OUTC_CMP; - case OPCODE_DP3: return R300_ALU_OUTC_DP3; - case OPCODE_DP4: return R300_ALU_OUTC_DP4; - case OPCODE_FRC: return R300_ALU_OUTC_FRC; + case RC_OPCODE_CMP: return R300_ALU_OUTC_CMP; + case RC_OPCODE_DP3: return R300_ALU_OUTC_DP3; + case RC_OPCODE_DP4: return R300_ALU_OUTC_DP4; + case RC_OPCODE_FRC: return R300_ALU_OUTC_FRC; default: error("translate_rgb_opcode(%i): Unknown opcode", opcode); /* fall through */ - case OPCODE_NOP: + case RC_OPCODE_NOP: /* fall through */ - case OPCODE_MAD: return R300_ALU_OUTC_MAD; - case OPCODE_MAX: return R300_ALU_OUTC_MAX; - case OPCODE_MIN: return R300_ALU_OUTC_MIN; - case OPCODE_REPL_ALPHA: return R300_ALU_OUTC_REPL_ALPHA; + case RC_OPCODE_MAD: return R300_ALU_OUTC_MAD; + case RC_OPCODE_MAX: return R300_ALU_OUTC_MAX; + case RC_OPCODE_MIN: return R300_ALU_OUTC_MIN; + case RC_OPCODE_REPL_ALPHA: return R300_ALU_OUTC_REPL_ALPHA; } } -static GLuint translate_alpha_opcode(struct r300_fragment_program_compiler * c, GLuint opcode) +static unsigned int translate_alpha_opcode(struct r300_fragment_program_compiler * c, rc_opcode opcode) { switch(opcode) { - case OPCODE_CMP: return R300_ALU_OUTA_CMP; - case OPCODE_DP3: return R300_ALU_OUTA_DP4; - case OPCODE_DP4: return R300_ALU_OUTA_DP4; - case OPCODE_EX2: return R300_ALU_OUTA_EX2; - case OPCODE_FRC: return R300_ALU_OUTA_FRC; - case OPCODE_LG2: return R300_ALU_OUTA_LG2; + case RC_OPCODE_CMP: return R300_ALU_OUTA_CMP; + case RC_OPCODE_DP3: return R300_ALU_OUTA_DP4; + case RC_OPCODE_DP4: return R300_ALU_OUTA_DP4; + case RC_OPCODE_EX2: return R300_ALU_OUTA_EX2; + case RC_OPCODE_FRC: return R300_ALU_OUTA_FRC; + case RC_OPCODE_LG2: return R300_ALU_OUTA_LG2; default: error("translate_rgb_opcode(%i): Unknown opcode", opcode); /* fall through */ - case OPCODE_NOP: + case RC_OPCODE_NOP: /* fall through */ - case OPCODE_MAD: return R300_ALU_OUTA_MAD; - case OPCODE_MAX: return R300_ALU_OUTA_MAX; - case OPCODE_MIN: return R300_ALU_OUTA_MIN; - case OPCODE_RCP: return R300_ALU_OUTA_RCP; - case OPCODE_RSQ: return R300_ALU_OUTA_RSQ; + case RC_OPCODE_MAD: return R300_ALU_OUTA_MAD; + case RC_OPCODE_MAX: return R300_ALU_OUTA_MAX; + case RC_OPCODE_MIN: return R300_ALU_OUTA_MIN; + case RC_OPCODE_RCP: return R300_ALU_OUTA_RCP; + case RC_OPCODE_RSQ: return R300_ALU_OUTA_RSQ; } } /** * Emit one paired ALU instruction. */ -static GLboolean emit_alu(void* data, struct radeon_pair_instruction* inst) +static int emit_alu(struct r300_emit_state * emit, struct rc_pair_instruction* inst) { PROG_CODE; if (code->alu.length >= R300_PFS_MAX_ALU_INST) { error("Too many ALU instructions"); - return GL_FALSE; + return 0; } int ip = code->alu.length++; @@ -136,17 +147,13 @@ static GLboolean emit_alu(void* data, struct radeon_pair_instruction* inst) code->alu.inst[ip].alpha_inst = translate_alpha_opcode(c, inst->Alpha.Opcode); for(j = 0; j < 3; ++j) { - GLuint src = inst->RGB.Src[j].Index | (inst->RGB.Src[j].Constant << 5); - if (!inst->RGB.Src[j].Constant) - use_temporary(code, inst->RGB.Src[j].Index); + unsigned int src = use_source(code, inst->RGB.Src[j]); code->alu.inst[ip].rgb_addr |= src << (6*j); - src = inst->Alpha.Src[j].Index | (inst->Alpha.Src[j].Constant << 5); - if (!inst->Alpha.Src[j].Constant) - use_temporary(code, inst->Alpha.Src[j].Index); + src = use_source(code, inst->Alpha.Src[j]); code->alu.inst[ip].alpha_addr |= src << (6*j); - GLuint arg = r300FPTranslateRGBSwizzle(inst->RGB.Arg[j].Source, inst->RGB.Arg[j].Swizzle); + unsigned int arg = r300FPTranslateRGBSwizzle(inst->RGB.Arg[j].Source, inst->RGB.Arg[j].Swizzle); arg |= inst->RGB.Arg[j].Abs << 6; arg |= inst->RGB.Arg[j].Negate << 5; code->alu.inst[ip].rgb_inst |= arg << (7*j); @@ -186,27 +193,27 @@ static GLboolean emit_alu(void* data, struct radeon_pair_instruction* inst) if (inst->Alpha.DepthWriteMask) { code->alu.inst[ip].alpha_addr |= R300_ALU_DSTA_DEPTH; emit->node_flags |= R300_W_OUT; - c->code->writes_depth = GL_TRUE; + c->code->writes_depth = 1; } - return GL_TRUE; + return 1; } /** * Finish the current node without advancing to the next one. */ -static GLboolean finish_node(struct r300_emit_state * emit) +static int finish_node(struct r300_emit_state * emit) { struct r300_fragment_program_compiler * c = emit->compiler; struct r300_fragment_program_code *code = &emit->compiler->code->code.r300; if (code->alu.length == emit->node_first_alu) { /* Generate a single NOP for this node */ - struct radeon_pair_instruction inst; - _mesa_bzero(&inst, sizeof(inst)); + struct rc_pair_instruction inst; + memset(&inst, 0, sizeof(inst)); if (!emit_alu(emit, &inst)) - return GL_FALSE; + return 0; } unsigned alu_offset = emit->node_first_alu; @@ -217,7 +224,7 @@ static GLboolean finish_node(struct r300_emit_state * emit) if (code->tex.length == emit->node_first_tex) { if (emit->current_node > 0) { error("Node %i has no TEX instructions", emit->current_node); - return GL_FALSE; + return 0; } tex_end = 0; @@ -240,7 +247,7 @@ static GLboolean finish_node(struct r300_emit_state * emit) (tex_end << R300_TEX_SIZE_SHIFT) | emit->node_flags; - return GL_TRUE; + return 1; } @@ -248,79 +255,72 @@ static GLboolean finish_node(struct r300_emit_state * emit) * Begin a block of texture instructions. * Create the necessary indirection. */ -static GLboolean begin_tex(void* data) +static int begin_tex(struct r300_emit_state * emit) { PROG_CODE; if (code->alu.length == emit->node_first_alu && code->tex.length == emit->node_first_tex) { - return GL_TRUE; + return 1; } if (emit->current_node == 3) { error("Too many texture indirections"); - return GL_FALSE; + return 0; } if (!finish_node(emit)) - return GL_FALSE; + return 0; emit->current_node++; emit->node_first_tex = code->tex.length; emit->node_first_alu = code->alu.length; emit->node_flags = 0; - return GL_TRUE; + return 1; } -static GLboolean emit_tex(void* data, struct radeon_pair_texture_instruction* inst) +static int emit_tex(struct r300_emit_state * emit, struct rc_instruction * inst) { PROG_CODE; if (code->tex.length >= R300_PFS_MAX_TEX_INST) { error("Too many TEX instructions"); - return GL_FALSE; + return 0; } - GLuint unit = inst->TexSrcUnit; - GLuint dest = inst->DestIndex; - GLuint opcode; + unsigned int unit = inst->U.I.TexSrcUnit; + unsigned int dest = inst->U.I.DstReg.Index; + unsigned int opcode; - switch(inst->Opcode) { - case RADEON_OPCODE_KIL: opcode = R300_TEX_OP_KIL; break; - case RADEON_OPCODE_TEX: opcode = R300_TEX_OP_LD; break; - case RADEON_OPCODE_TXB: opcode = R300_TEX_OP_TXB; break; - case RADEON_OPCODE_TXP: opcode = R300_TEX_OP_TXP; break; + switch(inst->U.I.Opcode) { + case RC_OPCODE_KIL: opcode = R300_TEX_OP_KIL; break; + case RC_OPCODE_TEX: opcode = R300_TEX_OP_LD; break; + case RC_OPCODE_TXB: opcode = R300_TEX_OP_TXB; break; + case RC_OPCODE_TXP: opcode = R300_TEX_OP_TXP; break; default: - error("Unknown texture opcode %i", inst->Opcode); - return GL_FALSE; + error("Unknown texture opcode %i", inst->U.I.Opcode); + return 0; } - if (inst->Opcode == RADEON_OPCODE_KIL) { + if (inst->U.I.Opcode == RC_OPCODE_KIL) { unit = 0; dest = 0; } else { use_temporary(code, dest); } - use_temporary(code, inst->SrcIndex); + use_temporary(code, inst->U.I.SrcReg[0].Index); code->tex.inst[code->tex.length++] = - (inst->SrcIndex << R300_SRC_ADDR_SHIFT) | + (inst->U.I.SrcReg[0].Index << R300_SRC_ADDR_SHIFT) | (dest << R300_DST_ADDR_SHIFT) | (unit << R300_TEX_ID_SHIFT) | (opcode << R300_TEX_INST_SHIFT); - return GL_TRUE; + return 1; } -static const struct radeon_pair_handler pair_handler = { - .EmitPaired = &emit_alu, - .EmitTex = &emit_tex, - .BeginTexBlock = &begin_tex, - .MaxHwTemps = R300_PFS_NUM_TEMP_REGS -}; - /** * Final compilation step: Turn the intermediate radeon_program into * machine-readable instructions. @@ -333,9 +333,26 @@ void r300BuildFragmentProgramHwCode(struct r300_fragment_program_compiler *compi memset(&emit, 0, sizeof(emit)); emit.compiler = compiler; - _mesa_bzero(code, sizeof(struct r300_fragment_program_code)); + memset(code, 0, sizeof(struct r300_fragment_program_code)); + + for(struct rc_instruction * inst = compiler->Base.Program.Instructions.Next; + inst != &compiler->Base.Program.Instructions && !compiler->Base.Error; + inst = inst->Next) { + if (inst->Type == RC_INSTRUCTION_NORMAL) { + if (inst->U.I.Opcode == RC_OPCODE_BEGIN_TEX) { + begin_tex(&emit); + continue; + } + + emit_tex(&emit, inst); + } else { + emit_alu(&emit, &inst->U.P); + } + } + + if (code->pixsize >= R300_PFS_NUM_TEMP_REGS) + rc_error(&compiler->Base, "Too many hardware temporaries used.\n"); - radeonPairProgram(compiler, &pair_handler, &emit); if (compiler->Base.Error) return; @@ -352,7 +369,7 @@ void r300BuildFragmentProgramHwCode(struct r300_fragment_program_compiler *compi if (emit.current_node < 3) { int shift = 3 - emit.current_node; int i; - for(i = 0; i <= emit.current_node; ++i) + for(i = emit.current_node; i >= 0; --i) code->code_addr[shift + i] = code->code_addr[i]; for(i = 0; i < shift; ++i) code->code_addr[i] = 0; diff --git a/src/mesa/drivers/dri/r300/compiler/r300_fragprog_swizzle.c b/src/mesa/drivers/dri/r300/compiler/r300_fragprog_swizzle.c index 1b14cc3888b..cfa48a59e3a 100644 --- a/src/mesa/drivers/dri/r300/compiler/r300_fragprog_swizzle.c +++ b/src/mesa/drivers/dri/r300/compiler/r300_fragprog_swizzle.c @@ -33,16 +33,17 @@ #include "r300_fragprog_swizzle.h" +#include <stdio.h> + #include "../r300_reg.h" -#include "radeon_nqssadce.h" #include "radeon_compiler.h" -#define MAKE_SWZ3(x, y, z) (MAKE_SWIZZLE4(SWIZZLE_##x, SWIZZLE_##y, SWIZZLE_##z, SWIZZLE_ZERO)) +#define MAKE_SWZ3(x, y, z) (RC_MAKE_SWIZZLE(RC_SWIZZLE_##x, RC_SWIZZLE_##y, RC_SWIZZLE_##z, RC_SWIZZLE_ZERO)) struct swizzle_data { - GLuint hash; /**< swizzle value this matches */ - GLuint base; /**< base value for hw swizzle */ - GLuint stride; /**< difference in base between arg0/1/2 */ + unsigned int hash; /**< swizzle value this matches */ + unsigned int base; /**< base value for hw swizzle */ + unsigned int stride; /**< difference in base between arg0/1/2 */ }; static const struct swizzle_data native_swizzles[] = { @@ -65,15 +66,15 @@ static const int num_native_swizzles = sizeof(native_swizzles)/sizeof(native_swi * Find a native RGB swizzle that matches the given swizzle. * Returns 0 if none found. */ -static const struct swizzle_data* lookup_native_swizzle(GLuint swizzle) +static const struct swizzle_data* lookup_native_swizzle(unsigned int swizzle) { int i, comp; for(i = 0; i < num_native_swizzles; ++i) { const struct swizzle_data* sd = &native_swizzles[i]; for(comp = 0; comp < 3; ++comp) { - GLuint swz = GET_SWZ(swizzle, comp); - if (swz == SWIZZLE_NIL) + unsigned int swz = GET_SWZ(swizzle, comp); + if (swz == RC_SWIZZLE_UNUSED) continue; if (swz != GET_SWZ(sd->hash, comp)) break; @@ -90,71 +91,72 @@ static const struct swizzle_data* lookup_native_swizzle(GLuint swizzle) * Check whether the given instruction supports the swizzle and negate * combinations in the given source register. */ -GLboolean r300FPIsNativeSwizzle(GLuint opcode, struct prog_src_register reg) +static int r300_swizzle_is_native(rc_opcode opcode, struct rc_src_register reg) { if (reg.Abs) - reg.Negate = NEGATE_NONE; + reg.Negate = RC_MASK_NONE; - if (opcode == OPCODE_KIL || - opcode == OPCODE_TEX || - opcode == OPCODE_TXB || - opcode == OPCODE_TXP) { + if (opcode == RC_OPCODE_KIL || + opcode == RC_OPCODE_TEX || + opcode == RC_OPCODE_TXB || + opcode == RC_OPCODE_TXP) { int j; if (reg.Abs || reg.Negate) - return GL_FALSE; + return 0; for(j = 0; j < 4; ++j) { - GLuint swz = GET_SWZ(reg.Swizzle, j); - if (swz == SWIZZLE_NIL) + unsigned int swz = GET_SWZ(reg.Swizzle, j); + if (swz == RC_SWIZZLE_UNUSED) continue; if (swz != j) - return GL_FALSE; + return 0; } - return GL_TRUE; + return 1; } - GLuint relevant = 0; + unsigned int relevant = 0; int j; for(j = 0; j < 3; ++j) - if (GET_SWZ(reg.Swizzle, j) != SWIZZLE_NIL) + if (GET_SWZ(reg.Swizzle, j) != RC_SWIZZLE_UNUSED) relevant |= 1 << j; if ((reg.Negate & relevant) && ((reg.Negate & relevant) != relevant)) - return GL_FALSE; + return 0; if (!lookup_native_swizzle(reg.Swizzle)) - return GL_FALSE; + return 0; - return GL_TRUE; + return 1; } -/** - * Generate MOV dst, src using only native swizzles. - */ -void r300FPBuildSwizzle(struct nqssadce_state *s, struct prog_dst_register dst, struct prog_src_register src) +static void r300_swizzle_split( + struct rc_src_register src, unsigned int mask, + struct rc_swizzle_split * split) { if (src.Abs) - src.Negate = NEGATE_NONE; + src.Negate = RC_MASK_NONE; + + split->NumPhases = 0; - while(dst.WriteMask) { + while(mask) { const struct swizzle_data *best_swizzle = 0; - GLuint best_matchcount = 0; - GLuint best_matchmask = 0; + unsigned int best_matchcount = 0; + unsigned int best_matchmask = 0; int i, comp; for(i = 0; i < num_native_swizzles; ++i) { const struct swizzle_data *sd = &native_swizzles[i]; - GLuint matchcount = 0; - GLuint matchmask = 0; + unsigned int matchcount = 0; + unsigned int matchmask = 0; for(comp = 0; comp < 3; ++comp) { - if (!GET_BIT(dst.WriteMask, comp)) + if (!GET_BIT(mask, comp)) continue; - GLuint swz = GET_SWZ(src.Swizzle, comp); - if (swz == SWIZZLE_NIL) + unsigned int swz = GET_SWZ(src.Swizzle, comp); + if (swz == RC_SWIZZLE_UNUSED) continue; if (swz == GET_SWZ(sd->hash, comp)) { /* check if the negate bit of current component @@ -170,34 +172,35 @@ void r300FPBuildSwizzle(struct nqssadce_state *s, struct prog_dst_register dst, best_swizzle = sd; best_matchcount = matchcount; best_matchmask = matchmask; - if (matchmask == (dst.WriteMask & WRITEMASK_XYZ)) + if (matchmask == (mask & RC_MASK_XYZ)) break; } } - struct rc_instruction *inst = rc_insert_new_instruction(s->Compiler, s->IP->Prev); - inst->I.Opcode = OPCODE_MOV; - inst->I.DstReg = dst; - inst->I.DstReg.WriteMask &= (best_matchmask | WRITEMASK_W); - inst->I.SrcReg[0] = src; - inst->I.SrcReg[0].Negate = (best_matchmask & src.Negate) ? NEGATE_XYZW : NEGATE_NONE; - /* Note: We rely on NqSSA/DCE to set unused swizzle components to NIL */ + if (mask & RC_MASK_W) + best_matchmask |= RC_MASK_W; - dst.WriteMask &= ~inst->I.DstReg.WriteMask; + split->Phase[split->NumPhases++] = best_matchmask; + mask &= ~best_matchmask; } } +struct rc_swizzle_caps r300_swizzle_caps = { + .IsNative = r300_swizzle_is_native, + .Split = r300_swizzle_split +}; + /** * Translate an RGB (XYZ) swizzle into the hardware code for the given * instruction source. */ -GLuint r300FPTranslateRGBSwizzle(GLuint src, GLuint swizzle) +unsigned int r300FPTranslateRGBSwizzle(unsigned int src, unsigned int swizzle) { const struct swizzle_data* sd = lookup_native_swizzle(swizzle); if (!sd) { - _mesa_printf("Not a native swizzle: %08x\n", swizzle); + fprintf(stderr, "Not a native swizzle: %08x\n", swizzle); return 0; } @@ -209,15 +212,15 @@ GLuint r300FPTranslateRGBSwizzle(GLuint src, GLuint swizzle) * Translate an Alpha (W) swizzle into the hardware code for the given * instruction source. */ -GLuint r300FPTranslateAlphaSwizzle(GLuint src, GLuint swizzle) +unsigned int r300FPTranslateAlphaSwizzle(unsigned int src, unsigned int swizzle) { if (swizzle < 3) return swizzle + 3*src; switch(swizzle) { - case SWIZZLE_W: return R300_ALU_ARGA_SRC0A + src; - case SWIZZLE_ONE: return R300_ALU_ARGA_ONE; - case SWIZZLE_ZERO: return R300_ALU_ARGA_ZERO; + case RC_SWIZZLE_W: return R300_ALU_ARGA_SRC0A + src; + case RC_SWIZZLE_ONE: return R300_ALU_ARGA_ONE; + case RC_SWIZZLE_ZERO: return R300_ALU_ARGA_ZERO; default: return R300_ALU_ARGA_ONE; } } diff --git a/src/mesa/drivers/dri/r300/compiler/r300_fragprog_swizzle.h b/src/mesa/drivers/dri/r300/compiler/r300_fragprog_swizzle.h index 231bf4eef5f..118476af132 100644 --- a/src/mesa/drivers/dri/r300/compiler/r300_fragprog_swizzle.h +++ b/src/mesa/drivers/dri/r300/compiler/r300_fragprog_swizzle.h @@ -28,15 +28,11 @@ #ifndef __R300_FRAGPROG_SWIZZLE_H_ #define __R300_FRAGPROG_SWIZZLE_H_ -#include "main/glheader.h" -#include "shader/prog_instruction.h" +#include "radeon_swizzle.h" -struct nqssadce_state; +extern struct rc_swizzle_caps r300_swizzle_caps; -GLboolean r300FPIsNativeSwizzle(GLuint opcode, struct prog_src_register reg); -void r300FPBuildSwizzle(struct nqssadce_state*, struct prog_dst_register dst, struct prog_src_register src); - -GLuint r300FPTranslateRGBSwizzle(GLuint src, GLuint swizzle); -GLuint r300FPTranslateAlphaSwizzle(GLuint src, GLuint swizzle); +unsigned int r300FPTranslateRGBSwizzle(unsigned int src, unsigned int swizzle); +unsigned int r300FPTranslateAlphaSwizzle(unsigned int src, unsigned int swizzle); #endif /* __R300_FRAGPROG_SWIZZLE_H_ */ diff --git a/src/mesa/drivers/dri/r300/compiler/r3xx_fragprog.c b/src/mesa/drivers/dri/r300/compiler/r3xx_fragprog.c index 76c3a7ecfd2..5581f25352d 100644 --- a/src/mesa/drivers/dri/r300/compiler/r3xx_fragprog.c +++ b/src/mesa/drivers/dri/r300/compiler/r3xx_fragprog.c @@ -22,22 +22,21 @@ #include "radeon_compiler.h" -#include "shader/prog_parameter.h" -#include "shader/prog_print.h" -#include "shader/prog_statevars.h" +#include <stdio.h> -#include "radeon_nqssadce.h" +#include "radeon_dataflow.h" #include "radeon_program_alu.h" #include "r300_fragprog.h" #include "r300_fragprog_swizzle.h" #include "r500_fragprog.h" -static void nqssadce_init(struct nqssadce_state* s) +static void dataflow_outputs_mark_use(void * userdata, void * data, + void (*callback)(void *, unsigned int, unsigned int)) { - struct r300_fragment_program_compiler * c = s->UserData; - s->Outputs[c->OutputColor].Sourced = WRITEMASK_XYZW; - s->Outputs[c->OutputDepth].Sourced = WRITEMASK_W; + struct r300_fragment_program_compiler * c = userdata; + callback(data, c->OutputColor, RC_MASK_XYZW); + callback(data, c->OutputDepth, RC_MASK_W); } static void rewrite_depth_out(struct r300_fragment_program_compiler * c) @@ -45,35 +44,35 @@ static void rewrite_depth_out(struct r300_fragment_program_compiler * c) struct rc_instruction *rci; for (rci = c->Base.Program.Instructions.Next; rci != &c->Base.Program.Instructions; rci = rci->Next) { - struct prog_instruction * inst = &rci->I; + struct rc_sub_instruction * inst = &rci->U.I; - if (inst->DstReg.File != PROGRAM_OUTPUT || inst->DstReg.Index != c->OutputDepth) + if (inst->DstReg.File != RC_FILE_OUTPUT || inst->DstReg.Index != c->OutputDepth) continue; - if (inst->DstReg.WriteMask & WRITEMASK_Z) { - inst->DstReg.WriteMask = WRITEMASK_W; + if (inst->DstReg.WriteMask & RC_MASK_Z) { + inst->DstReg.WriteMask = RC_MASK_W; } else { inst->DstReg.WriteMask = 0; continue; } switch (inst->Opcode) { - case OPCODE_FRC: - case OPCODE_MOV: - inst->SrcReg[0] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[0]); + case RC_OPCODE_FRC: + case RC_OPCODE_MOV: + inst->SrcReg[0] = lmul_swizzle(RC_SWIZZLE_ZZZZ, inst->SrcReg[0]); break; - case OPCODE_ADD: - case OPCODE_MAX: - case OPCODE_MIN: - case OPCODE_MUL: - inst->SrcReg[0] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[0]); - inst->SrcReg[1] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[1]); + case RC_OPCODE_ADD: + case RC_OPCODE_MAX: + case RC_OPCODE_MIN: + case RC_OPCODE_MUL: + inst->SrcReg[0] = lmul_swizzle(RC_SWIZZLE_ZZZZ, inst->SrcReg[0]); + inst->SrcReg[1] = lmul_swizzle(RC_SWIZZLE_ZZZZ, inst->SrcReg[1]); break; - case OPCODE_CMP: - case OPCODE_MAD: - inst->SrcReg[0] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[0]); - inst->SrcReg[1] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[1]); - inst->SrcReg[2] = lmul_swizzle(SWIZZLE_ZZZZ, inst->SrcReg[2]); + case RC_OPCODE_CMP: + case RC_OPCODE_MAD: + inst->SrcReg[0] = lmul_swizzle(RC_SWIZZLE_ZZZZ, inst->SrcReg[0]); + inst->SrcReg[1] = lmul_swizzle(RC_SWIZZLE_ZZZZ, inst->SrcReg[1]); + inst->SrcReg[2] = lmul_swizzle(RC_SWIZZLE_ZZZZ, inst->SrcReg[2]); break; default: // Scalar instructions needn't be reswizzled @@ -89,11 +88,14 @@ void r3xx_compile_fragment_program(struct r300_fragment_program_compiler* c) if (c->is_r500) { struct radeon_program_transformation transformations[] = { { &r500_transform_TEX, c }, + { &r500_transform_IF, 0 }, { &radeonTransformALU, 0 }, { &radeonTransformDeriv, 0 }, { &radeonTransformTrigScale, 0 } }; - radeonLocalTransform(&c->Base, 4, transformations); + radeonLocalTransform(&c->Base, 5, transformations); + + c->Base.SwizzleCaps = &r500_swizzle_caps; } else { struct radeon_program_transformation transformations[] = { { &r300_transform_TEX, c }, @@ -101,32 +103,66 @@ void r3xx_compile_fragment_program(struct r300_fragment_program_compiler* c) { &radeonTransformTrigSimple, 0 } }; radeonLocalTransform(&c->Base, 3, transformations); + + c->Base.SwizzleCaps = &r300_swizzle_caps; } if (c->Base.Debug) { - _mesa_printf("Fragment Program: After native rewrite:\n"); + fprintf(stderr, "Fragment Program: After native rewrite:\n"); rc_print_program(&c->Base.Program); fflush(stderr); } - if (c->is_r500) { - struct radeon_nqssadce_descr nqssadce = { - .Init = &nqssadce_init, - .IsNativeSwizzle = &r500FPIsNativeSwizzle, - .BuildSwizzle = &r500FPBuildSwizzle - }; - radeonNqssaDce(&c->Base, &nqssadce, c); - } else { - struct radeon_nqssadce_descr nqssadce = { - .Init = &nqssadce_init, - .IsNativeSwizzle = &r300FPIsNativeSwizzle, - .BuildSwizzle = &r300FPBuildSwizzle - }; - radeonNqssaDce(&c->Base, &nqssadce, c); + rc_dataflow_deadcode(&c->Base, &dataflow_outputs_mark_use, c); + if (c->Base.Error) + return; + + if (c->Base.Debug) { + fprintf(stderr, "Fragment Program: After deadcode:\n"); + rc_print_program(&c->Base.Program); + fflush(stderr); + } + + rc_dataflow_swizzles(&c->Base); + if (c->Base.Error) + return; + + if (c->Base.Debug) { + fprintf(stderr, "Compiler: after dataflow passes:\n"); + rc_print_program(&c->Base.Program); + fflush(stderr); + } + + rc_pair_translate(c); + if (c->Base.Error) + return; + + if (c->Base.Debug) { + fprintf(stderr, "Compiler: after pair translate:\n"); + rc_print_program(&c->Base.Program); + fflush(stderr); } + rc_pair_schedule(c); + if (c->Base.Error) + return; + + if (c->Base.Debug) { + fprintf(stderr, "Compiler: after pair scheduling:\n"); + rc_print_program(&c->Base.Program); + fflush(stderr); + } + + if (c->is_r500) + rc_pair_regalloc(c, 128); + else + rc_pair_regalloc(c, R300_PFS_NUM_TEMP_REGS); + + if (c->Base.Error) + return; + if (c->Base.Debug) { - _mesa_printf("Compiler: after NqSSA-DCE:\n"); + fprintf(stderr, "Compiler: after pair register allocation:\n"); rc_print_program(&c->Base.Program); fflush(stderr); } diff --git a/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog.c b/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog.c index 93a516105e6..1b2cb8dde7d 100644 --- a/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog.c +++ b/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog.c @@ -22,13 +22,13 @@ #include "radeon_compiler.h" +#include <stdio.h> + #include "../r300_reg.h" -#include "radeon_nqssadce.h" -#include "radeon_program.h" +#include "radeon_dataflow.h" #include "radeon_program_alu.h" - -#include "shader/prog_print.h" +#include "radeon_swizzle.h" /* @@ -42,103 +42,83 @@ t_swizzle(y), \ t_swizzle(y), \ t_src_class(vpi->SrcReg[x].File), \ - NEGATE_NONE) | (vpi->SrcReg[x].RelAddr << 4)) + RC_MASK_NONE) | (vpi->SrcReg[x].RelAddr << 4)) -static unsigned long t_dst_mask(GLuint mask) +static unsigned long t_dst_mask(unsigned int mask) { - /* WRITEMASK_* is equivalent to VSF_FLAG_* */ - return mask & WRITEMASK_XYZW; + /* RC_MASK_* is equivalent to VSF_FLAG_* */ + return mask & RC_MASK_XYZW; } -static unsigned long t_dst_class(gl_register_file file) +static unsigned long t_dst_class(rc_register_file file) { - switch (file) { - case PROGRAM_TEMPORARY: + default: + fprintf(stderr, "%s: Bad register file %i\n", __FUNCTION__, file); + /* fall-through */ + case RC_FILE_TEMPORARY: return PVS_DST_REG_TEMPORARY; - case PROGRAM_OUTPUT: + case RC_FILE_OUTPUT: return PVS_DST_REG_OUT; - case PROGRAM_ADDRESS: + case RC_FILE_ADDRESS: return PVS_DST_REG_A0; - /* - case PROGRAM_INPUT: - case PROGRAM_LOCAL_PARAM: - case PROGRAM_ENV_PARAM: - case PROGRAM_NAMED_PARAM: - case PROGRAM_STATE_VAR: - case PROGRAM_WRITE_ONLY: - case PROGRAM_ADDRESS: - */ - default: - fprintf(stderr, "problem in %s", __FUNCTION__); - _mesa_exit(-1); - return -1; } } static unsigned long t_dst_index(struct r300_vertex_program_code *vp, - struct prog_dst_register *dst) + struct rc_dst_register *dst) { - if (dst->File == PROGRAM_OUTPUT) + if (dst->File == RC_FILE_OUTPUT) return vp->outputs[dst->Index]; return dst->Index; } -static unsigned long t_src_class(gl_register_file file) +static unsigned long t_src_class(rc_register_file file) { switch (file) { - case PROGRAM_TEMPORARY: + default: + fprintf(stderr, "%s: Bad register file %i\n", __FUNCTION__, file); + /* fall-through */ + case RC_FILE_NONE: + case RC_FILE_TEMPORARY: return PVS_SRC_REG_TEMPORARY; - case PROGRAM_INPUT: + case RC_FILE_INPUT: return PVS_SRC_REG_INPUT; - case PROGRAM_LOCAL_PARAM: - case PROGRAM_ENV_PARAM: - case PROGRAM_NAMED_PARAM: - case PROGRAM_CONSTANT: - case PROGRAM_STATE_VAR: + case RC_FILE_CONSTANT: return PVS_SRC_REG_CONSTANT; - /* - case PROGRAM_OUTPUT: - case PROGRAM_WRITE_ONLY: - case PROGRAM_ADDRESS: - */ - default: - fprintf(stderr, "problem in %s", __FUNCTION__); - _mesa_exit(-1); - return -1; } } -static GLboolean t_src_conflict(struct prog_src_register a, struct prog_src_register b) +static int t_src_conflict(struct rc_src_register a, struct rc_src_register b) { unsigned long aclass = t_src_class(a.File); unsigned long bclass = t_src_class(b.File); if (aclass != bclass) - return GL_FALSE; + return 0; if (aclass == PVS_SRC_REG_TEMPORARY) - return GL_FALSE; + return 0; if (a.RelAddr || b.RelAddr) - return GL_TRUE; + return 1; if (a.Index != b.Index) - return GL_TRUE; + return 1; - return GL_FALSE; + return 0; } -static INLINE unsigned long t_swizzle(GLubyte swizzle) +static inline unsigned long t_swizzle(unsigned int swizzle) { - /* this is in fact a NOP as the Mesa SWIZZLE_* are all identical to VSF_IN_COMPONENT_* */ + /* this is in fact a NOP as the Mesa RC_SWIZZLE_* are all identical to VSF_IN_COMPONENT_* */ return swizzle; } static unsigned long t_src_index(struct r300_vertex_program_code *vp, - struct prog_src_register *src) + struct rc_src_register *src) { - if (src->File == PROGRAM_INPUT) { + if (src->File == RC_FILE_INPUT) { assert(vp->inputs[src->Index] != -1); return vp->inputs[src->Index]; } else { @@ -154,9 +134,9 @@ static unsigned long t_src_index(struct r300_vertex_program_code *vp, /* these two functions should probably be merged... */ static unsigned long t_src(struct r300_vertex_program_code *vp, - struct prog_src_register *src) + struct rc_src_register *src) { - /* src->Negate uses the NEGATE_ flags from program_instruction.h, + /* src->Negate uses the RC_MASK_ flags from program_instruction.h, * which equal our VSF_FLAGS_ values, so it's safe to just pass it here. */ return PVS_SRC_OPERAND(t_src_index(vp, src), @@ -169,9 +149,9 @@ static unsigned long t_src(struct r300_vertex_program_code *vp, } static unsigned long t_src_scalar(struct r300_vertex_program_code *vp, - struct prog_src_register *src) + struct rc_src_register *src) { - /* src->Negate uses the NEGATE_ flags from program_instruction.h, + /* src->Negate uses the RC_MASK_ flags from program_instruction.h, * which equal our VSF_FLAGS_ values, so it's safe to just pass it here. */ return PVS_SRC_OPERAND(t_src_index(vp, src), @@ -180,79 +160,79 @@ static unsigned long t_src_scalar(struct r300_vertex_program_code *vp, t_swizzle(GET_SWZ(src->Swizzle, 0)), t_swizzle(GET_SWZ(src->Swizzle, 0)), t_src_class(src->File), - src->Negate ? NEGATE_XYZW : NEGATE_NONE) | + src->Negate ? RC_MASK_XYZW : RC_MASK_NONE) | (src->RelAddr << 4); } -static GLboolean valid_dst(struct r300_vertex_program_code *vp, - struct prog_dst_register *dst) +static int valid_dst(struct r300_vertex_program_code *vp, + struct rc_dst_register *dst) { - if (dst->File == PROGRAM_OUTPUT && vp->outputs[dst->Index] == -1) { - return GL_FALSE; - } else if (dst->File == PROGRAM_ADDRESS) { + if (dst->File == RC_FILE_OUTPUT && vp->outputs[dst->Index] == -1) { + return 0; + } else if (dst->File == RC_FILE_ADDRESS) { assert(dst->Index == 0); } - return GL_TRUE; + return 1; } static void ei_vector1(struct r300_vertex_program_code *vp, - GLuint hw_opcode, - struct prog_instruction *vpi, - GLuint * inst) + unsigned int hw_opcode, + struct rc_sub_instruction *vpi, + unsigned int * inst) { inst[0] = PVS_OP_DST_OPERAND(hw_opcode, - GL_FALSE, - GL_FALSE, + 0, + 0, t_dst_index(vp, &vpi->DstReg), t_dst_mask(vpi->DstReg.WriteMask), t_dst_class(vpi->DstReg.File)); inst[1] = t_src(vp, &vpi->SrcReg[0]); - inst[2] = __CONST(0, SWIZZLE_ZERO); - inst[3] = __CONST(0, SWIZZLE_ZERO); + inst[2] = __CONST(0, RC_SWIZZLE_ZERO); + inst[3] = __CONST(0, RC_SWIZZLE_ZERO); } static void ei_vector2(struct r300_vertex_program_code *vp, - GLuint hw_opcode, - struct prog_instruction *vpi, - GLuint * inst) + unsigned int hw_opcode, + struct rc_sub_instruction *vpi, + unsigned int * inst) { inst[0] = PVS_OP_DST_OPERAND(hw_opcode, - GL_FALSE, - GL_FALSE, + 0, + 0, t_dst_index(vp, &vpi->DstReg), t_dst_mask(vpi->DstReg.WriteMask), t_dst_class(vpi->DstReg.File)); inst[1] = t_src(vp, &vpi->SrcReg[0]); inst[2] = t_src(vp, &vpi->SrcReg[1]); - inst[3] = __CONST(1, SWIZZLE_ZERO); + inst[3] = __CONST(1, RC_SWIZZLE_ZERO); } static void ei_math1(struct r300_vertex_program_code *vp, - GLuint hw_opcode, - struct prog_instruction *vpi, - GLuint * inst) + unsigned int hw_opcode, + struct rc_sub_instruction *vpi, + unsigned int * inst) { inst[0] = PVS_OP_DST_OPERAND(hw_opcode, - GL_TRUE, - GL_FALSE, + 1, + 0, t_dst_index(vp, &vpi->DstReg), t_dst_mask(vpi->DstReg.WriteMask), t_dst_class(vpi->DstReg.File)); inst[1] = t_src_scalar(vp, &vpi->SrcReg[0]); - inst[2] = __CONST(0, SWIZZLE_ZERO); - inst[3] = __CONST(0, SWIZZLE_ZERO); + inst[2] = __CONST(0, RC_SWIZZLE_ZERO); + inst[3] = __CONST(0, RC_SWIZZLE_ZERO); } static void ei_lit(struct r300_vertex_program_code *vp, - struct prog_instruction *vpi, - GLuint * inst) + struct rc_sub_instruction *vpi, + unsigned int * inst) { //LIT TMP 1.Y Z TMP 1{} {X W Z Y} TMP 1{} {Y W Z X} TMP 1{} {Y X Z W} inst[0] = PVS_OP_DST_OPERAND(ME_LIGHT_COEFF_DX, - GL_TRUE, - GL_FALSE, + 1, + 0, t_dst_index(vp, &vpi->DstReg), t_dst_mask(vpi->DstReg.WriteMask), t_dst_class(vpi->DstReg.File)); @@ -262,27 +242,27 @@ static void ei_lit(struct r300_vertex_program_code *vp, PVS_SRC_SELECT_FORCE_0, // Z t_swizzle(GET_SWZ(vpi->SrcReg[0].Swizzle, 1)), // Y t_src_class(vpi->SrcReg[0].File), - vpi->SrcReg[0].Negate ? NEGATE_XYZW : NEGATE_NONE) | + vpi->SrcReg[0].Negate ? RC_MASK_XYZW : RC_MASK_NONE) | (vpi->SrcReg[0].RelAddr << 4); inst[2] = PVS_SRC_OPERAND(t_src_index(vp, &vpi->SrcReg[0]), t_swizzle(GET_SWZ(vpi->SrcReg[0].Swizzle, 1)), // Y t_swizzle(GET_SWZ(vpi->SrcReg[0].Swizzle, 3)), // W PVS_SRC_SELECT_FORCE_0, // Z t_swizzle(GET_SWZ(vpi->SrcReg[0].Swizzle, 0)), // X t_src_class(vpi->SrcReg[0].File), - vpi->SrcReg[0].Negate ? NEGATE_XYZW : NEGATE_NONE) | + vpi->SrcReg[0].Negate ? RC_MASK_XYZW : RC_MASK_NONE) | (vpi->SrcReg[0].RelAddr << 4); inst[3] = PVS_SRC_OPERAND(t_src_index(vp, &vpi->SrcReg[0]), t_swizzle(GET_SWZ(vpi->SrcReg[0].Swizzle, 1)), // Y t_swizzle(GET_SWZ(vpi->SrcReg[0].Swizzle, 0)), // X PVS_SRC_SELECT_FORCE_0, // Z t_swizzle(GET_SWZ(vpi->SrcReg[0].Swizzle, 3)), // W t_src_class(vpi->SrcReg[0].File), - vpi->SrcReg[0].Negate ? NEGATE_XYZW : NEGATE_NONE) | + vpi->SrcReg[0].Negate ? RC_MASK_XYZW : RC_MASK_NONE) | (vpi->SrcReg[0].RelAddr << 4); } static void ei_mad(struct r300_vertex_program_code *vp, - struct prog_instruction *vpi, - GLuint * inst) + struct rc_sub_instruction *vpi, + unsigned int * inst) { /* Remarks about hardware limitations of MAD * (please preserve this comment, as this information is _NOT_ @@ -310,22 +290,22 @@ static void ei_mad(struct r300_vertex_program_code *vp, * according to AMD docs, this should improve performance by one clock * as a nice side bonus. */ - if (vpi->SrcReg[0].File == PROGRAM_TEMPORARY && - vpi->SrcReg[1].File == PROGRAM_TEMPORARY && - vpi->SrcReg[2].File == PROGRAM_TEMPORARY && + if (vpi->SrcReg[0].File == RC_FILE_TEMPORARY && + vpi->SrcReg[1].File == RC_FILE_TEMPORARY && + vpi->SrcReg[2].File == RC_FILE_TEMPORARY && vpi->SrcReg[0].Index != vpi->SrcReg[1].Index && vpi->SrcReg[0].Index != vpi->SrcReg[2].Index && vpi->SrcReg[1].Index != vpi->SrcReg[2].Index) { inst[0] = PVS_OP_DST_OPERAND(PVS_MACRO_OP_2CLK_MADD, - GL_FALSE, - GL_TRUE, + 0, + 1, t_dst_index(vp, &vpi->DstReg), t_dst_mask(vpi->DstReg.WriteMask), t_dst_class(vpi->DstReg.File)); } else { inst[0] = PVS_OP_DST_OPERAND(VE_MULTIPLY_ADD, - GL_FALSE, - GL_FALSE, + 0, + 0, t_dst_index(vp, &vpi->DstReg), t_dst_mask(vpi->DstReg.WriteMask), t_dst_class(vpi->DstReg.File)); @@ -336,17 +316,17 @@ static void ei_mad(struct r300_vertex_program_code *vp, } static void ei_pow(struct r300_vertex_program_code *vp, - struct prog_instruction *vpi, - GLuint * inst) + struct rc_sub_instruction *vpi, + unsigned int * inst) { inst[0] = PVS_OP_DST_OPERAND(ME_POWER_FUNC_FF, - GL_TRUE, - GL_FALSE, + 1, + 0, t_dst_index(vp, &vpi->DstReg), t_dst_mask(vpi->DstReg.WriteMask), t_dst_class(vpi->DstReg.File)); inst[1] = t_src_scalar(vp, &vpi->SrcReg[0]); - inst[2] = __CONST(0, SWIZZLE_ZERO); + inst[2] = __CONST(0, RC_SWIZZLE_ZERO); inst[3] = t_src_scalar(vp, &vpi->SrcReg[1]); } @@ -361,8 +341,8 @@ static void translate_vertex_program(struct r300_vertex_program_compiler * compi compiler->SetHwInputOutput(compiler); for(rci = compiler->Base.Program.Instructions.Next; rci != &compiler->Base.Program.Instructions; rci = rci->Next) { - struct prog_instruction *vpi = &rci->I; - GLuint *inst = compiler->code->body.d + compiler->code->length; + struct rc_sub_instruction *vpi = &rci->U.I; + unsigned int *inst = compiler->code->body.d + compiler->code->length; /* Skip instructions writing to non-existing destination */ if (!valid_dst(compiler->code, &vpi->DstReg)) @@ -374,26 +354,26 @@ static void translate_vertex_program(struct r300_vertex_program_compiler * compi } switch (vpi->Opcode) { - case OPCODE_ADD: ei_vector2(compiler->code, VE_ADD, vpi, inst); break; - case OPCODE_ARL: ei_vector1(compiler->code, VE_FLT2FIX_DX, vpi, inst); break; - case OPCODE_DP4: ei_vector2(compiler->code, VE_DOT_PRODUCT, vpi, inst); break; - case OPCODE_DST: ei_vector2(compiler->code, VE_DISTANCE_VECTOR, vpi, inst); break; - case OPCODE_EX2: ei_math1(compiler->code, ME_EXP_BASE2_FULL_DX, vpi, inst); break; - case OPCODE_EXP: ei_math1(compiler->code, ME_EXP_BASE2_DX, vpi, inst); break; - case OPCODE_FRC: ei_vector1(compiler->code, VE_FRACTION, vpi, inst); break; - case OPCODE_LG2: ei_math1(compiler->code, ME_LOG_BASE2_FULL_DX, vpi, inst); break; - case OPCODE_LIT: ei_lit(compiler->code, vpi, inst); break; - case OPCODE_LOG: ei_math1(compiler->code, ME_LOG_BASE2_DX, vpi, inst); break; - case OPCODE_MAD: ei_mad(compiler->code, vpi, inst); break; - case OPCODE_MAX: ei_vector2(compiler->code, VE_MAXIMUM, vpi, inst); break; - case OPCODE_MIN: ei_vector2(compiler->code, VE_MINIMUM, vpi, inst); break; - case OPCODE_MOV: ei_vector1(compiler->code, VE_ADD, vpi, inst); break; - case OPCODE_MUL: ei_vector2(compiler->code, VE_MULTIPLY, vpi, inst); break; - case OPCODE_POW: ei_pow(compiler->code, vpi, inst); break; - case OPCODE_RCP: ei_math1(compiler->code, ME_RECIP_DX, vpi, inst); break; - case OPCODE_RSQ: ei_math1(compiler->code, ME_RECIP_SQRT_DX, vpi, inst); break; - case OPCODE_SGE: ei_vector2(compiler->code, VE_SET_GREATER_THAN_EQUAL, vpi, inst); break; - case OPCODE_SLT: ei_vector2(compiler->code, VE_SET_LESS_THAN, vpi, inst); break; + case RC_OPCODE_ADD: ei_vector2(compiler->code, VE_ADD, vpi, inst); break; + case RC_OPCODE_ARL: ei_vector1(compiler->code, VE_FLT2FIX_DX, vpi, inst); break; + case RC_OPCODE_DP4: ei_vector2(compiler->code, VE_DOT_PRODUCT, vpi, inst); break; + case RC_OPCODE_DST: ei_vector2(compiler->code, VE_DISTANCE_VECTOR, vpi, inst); break; + case RC_OPCODE_EX2: ei_math1(compiler->code, ME_EXP_BASE2_FULL_DX, vpi, inst); break; + case RC_OPCODE_EXP: ei_math1(compiler->code, ME_EXP_BASE2_DX, vpi, inst); break; + case RC_OPCODE_FRC: ei_vector1(compiler->code, VE_FRACTION, vpi, inst); break; + case RC_OPCODE_LG2: ei_math1(compiler->code, ME_LOG_BASE2_FULL_DX, vpi, inst); break; + case RC_OPCODE_LIT: ei_lit(compiler->code, vpi, inst); break; + case RC_OPCODE_LOG: ei_math1(compiler->code, ME_LOG_BASE2_DX, vpi, inst); break; + case RC_OPCODE_MAD: ei_mad(compiler->code, vpi, inst); break; + case RC_OPCODE_MAX: ei_vector2(compiler->code, VE_MAXIMUM, vpi, inst); break; + case RC_OPCODE_MIN: ei_vector2(compiler->code, VE_MINIMUM, vpi, inst); break; + case RC_OPCODE_MOV: ei_vector1(compiler->code, VE_ADD, vpi, inst); break; + case RC_OPCODE_MUL: ei_vector2(compiler->code, VE_MULTIPLY, vpi, inst); break; + case RC_OPCODE_POW: ei_pow(compiler->code, vpi, inst); break; + case RC_OPCODE_RCP: ei_math1(compiler->code, ME_RECIP_DX, vpi, inst); break; + case RC_OPCODE_RSQ: ei_math1(compiler->code, ME_RECIP_SQRT_DX, vpi, inst); break; + case RC_OPCODE_SGE: ei_vector2(compiler->code, VE_SET_GREATER_THAN_EQUAL, vpi, inst); break; + case RC_OPCODE_SLT: ei_vector2(compiler->code, VE_SET_LESS_THAN, vpi, inst); break; default: rc_error(&compiler->Base, "Unknown opcode %i\n", vpi->Opcode); return; @@ -407,38 +387,37 @@ static void translate_vertex_program(struct r300_vertex_program_compiler * compi } struct temporary_allocation { - GLuint Allocated:1; - GLuint HwTemp:15; + unsigned int Allocated:1; + unsigned int HwTemp:15; struct rc_instruction * LastRead; }; static void allocate_temporary_registers(struct r300_vertex_program_compiler * compiler) { struct rc_instruction *inst; - GLuint num_orig_temps = 0; - GLboolean hwtemps[VSF_MAX_FRAGMENT_TEMPS]; + unsigned int num_orig_temps = 0; + char hwtemps[VSF_MAX_FRAGMENT_TEMPS]; struct temporary_allocation * ta; - GLuint i, j; + unsigned int i, j; compiler->code->num_temporaries = 0; memset(hwtemps, 0, sizeof(hwtemps)); /* Pass 1: Count original temporaries and allocate structures */ for(inst = compiler->Base.Program.Instructions.Next; inst != &compiler->Base.Program.Instructions; inst = inst->Next) { - GLuint numsrcs = _mesa_num_inst_src_regs(inst->I.Opcode); - GLuint numdsts = _mesa_num_inst_dst_regs(inst->I.Opcode); + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->U.I.Opcode); - for (i = 0; i < numsrcs; ++i) { - if (inst->I.SrcReg[i].File == PROGRAM_TEMPORARY) { - if (inst->I.SrcReg[i].Index >= num_orig_temps) - num_orig_temps = inst->I.SrcReg[i].Index + 1; + for (i = 0; i < opcode->NumSrcRegs; ++i) { + if (inst->U.I.SrcReg[i].File == RC_FILE_TEMPORARY) { + if (inst->U.I.SrcReg[i].Index >= num_orig_temps) + num_orig_temps = inst->U.I.SrcReg[i].Index + 1; } } - if (numdsts) { - if (inst->I.DstReg.File == PROGRAM_TEMPORARY) { - if (inst->I.DstReg.Index >= num_orig_temps) - num_orig_temps = inst->I.DstReg.Index + 1; + if (opcode->HasDstReg) { + if (inst->U.I.DstReg.File == RC_FILE_TEMPORARY) { + if (inst->U.I.DstReg.Index >= num_orig_temps) + num_orig_temps = inst->U.I.DstReg.Index + 1; } } } @@ -449,32 +428,31 @@ static void allocate_temporary_registers(struct r300_vertex_program_compiler * c /* Pass 2: Determine original temporary lifetimes */ for(inst = compiler->Base.Program.Instructions.Next; inst != &compiler->Base.Program.Instructions; inst = inst->Next) { - GLuint numsrcs = _mesa_num_inst_src_regs(inst->I.Opcode); + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->U.I.Opcode); - for (i = 0; i < numsrcs; ++i) { - if (inst->I.SrcReg[i].File == PROGRAM_TEMPORARY) - ta[inst->I.SrcReg[i].Index].LastRead = inst; + for (i = 0; i < opcode->NumSrcRegs; ++i) { + if (inst->U.I.SrcReg[i].File == RC_FILE_TEMPORARY) + ta[inst->U.I.SrcReg[i].Index].LastRead = inst; } } /* Pass 3: Register allocation */ for(inst = compiler->Base.Program.Instructions.Next; inst != &compiler->Base.Program.Instructions; inst = inst->Next) { - GLuint numsrcs = _mesa_num_inst_src_regs(inst->I.Opcode); - GLuint numdsts = _mesa_num_inst_dst_regs(inst->I.Opcode); + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->U.I.Opcode); - for (i = 0; i < numsrcs; ++i) { - if (inst->I.SrcReg[i].File == PROGRAM_TEMPORARY) { - GLuint orig = inst->I.SrcReg[i].Index; - inst->I.SrcReg[i].Index = ta[orig].HwTemp; + for (i = 0; i < opcode->NumSrcRegs; ++i) { + if (inst->U.I.SrcReg[i].File == RC_FILE_TEMPORARY) { + unsigned int orig = inst->U.I.SrcReg[i].Index; + inst->U.I.SrcReg[i].Index = ta[orig].HwTemp; if (ta[orig].Allocated && inst == ta[orig].LastRead) - hwtemps[ta[orig].HwTemp] = GL_FALSE; + hwtemps[ta[orig].HwTemp] = 0; } } - if (numdsts) { - if (inst->I.DstReg.File == PROGRAM_TEMPORARY) { - GLuint orig = inst->I.DstReg.Index; + if (opcode->HasDstReg) { + if (inst->U.I.DstReg.File == RC_FILE_TEMPORARY) { + unsigned int orig = inst->U.I.DstReg.Index; if (!ta[orig].Allocated) { for(j = 0; j < VSF_MAX_FRAGMENT_TEMPS; ++j) { @@ -484,16 +462,16 @@ static void allocate_temporary_registers(struct r300_vertex_program_compiler * c if (j >= VSF_MAX_FRAGMENT_TEMPS) { fprintf(stderr, "Out of hw temporaries\n"); } else { - ta[orig].Allocated = GL_TRUE; + ta[orig].Allocated = 1; ta[orig].HwTemp = j; - hwtemps[j] = GL_TRUE; + hwtemps[j] = 1; if (j >= compiler->code->num_temporaries) compiler->code->num_temporaries = j + 1; } } - inst->I.DstReg.Index = ta[orig].HwTemp; + inst->U.I.DstReg.Index = ta[orig].HwTemp; } } } @@ -504,45 +482,45 @@ static void allocate_temporary_registers(struct r300_vertex_program_compiler * c * Vertex engine cannot read two inputs or two constants at the same time. * Introduce intermediate MOVs to temporary registers to account for this. */ -static GLboolean transform_source_conflicts( +static int transform_source_conflicts( struct radeon_compiler *c, struct rc_instruction* inst, void* unused) { - GLuint num_operands = _mesa_num_inst_src_regs(inst->I.Opcode); + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->U.I.Opcode); - if (num_operands == 3) { - if (t_src_conflict(inst->I.SrcReg[1], inst->I.SrcReg[2]) - || t_src_conflict(inst->I.SrcReg[0], inst->I.SrcReg[2])) { + if (opcode->NumSrcRegs == 3) { + if (t_src_conflict(inst->U.I.SrcReg[1], inst->U.I.SrcReg[2]) + || t_src_conflict(inst->U.I.SrcReg[0], inst->U.I.SrcReg[2])) { int tmpreg = rc_find_free_temporary(c); struct rc_instruction * inst_mov = rc_insert_new_instruction(c, inst->Prev); - inst_mov->I.Opcode = OPCODE_MOV; - inst_mov->I.DstReg.File = PROGRAM_TEMPORARY; - inst_mov->I.DstReg.Index = tmpreg; - inst_mov->I.SrcReg[0] = inst->I.SrcReg[2]; - - reset_srcreg(&inst->I.SrcReg[2]); - inst->I.SrcReg[2].File = PROGRAM_TEMPORARY; - inst->I.SrcReg[2].Index = tmpreg; + inst_mov->U.I.Opcode = RC_OPCODE_MOV; + inst_mov->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst_mov->U.I.DstReg.Index = tmpreg; + inst_mov->U.I.SrcReg[0] = inst->U.I.SrcReg[2]; + + reset_srcreg(&inst->U.I.SrcReg[2]); + inst->U.I.SrcReg[2].File = RC_FILE_TEMPORARY; + inst->U.I.SrcReg[2].Index = tmpreg; } } - if (num_operands >= 2) { - if (t_src_conflict(inst->I.SrcReg[1], inst->I.SrcReg[0])) { + if (opcode->NumSrcRegs >= 2) { + if (t_src_conflict(inst->U.I.SrcReg[1], inst->U.I.SrcReg[0])) { int tmpreg = rc_find_free_temporary(c); struct rc_instruction * inst_mov = rc_insert_new_instruction(c, inst->Prev); - inst_mov->I.Opcode = OPCODE_MOV; - inst_mov->I.DstReg.File = PROGRAM_TEMPORARY; - inst_mov->I.DstReg.Index = tmpreg; - inst_mov->I.SrcReg[0] = inst->I.SrcReg[1]; - - reset_srcreg(&inst->I.SrcReg[1]); - inst->I.SrcReg[1].File = PROGRAM_TEMPORARY; - inst->I.SrcReg[1].Index = tmpreg; + inst_mov->U.I.Opcode = RC_OPCODE_MOV; + inst_mov->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst_mov->U.I.DstReg.Index = tmpreg; + inst_mov->U.I.SrcReg[0] = inst->U.I.SrcReg[1]; + + reset_srcreg(&inst->U.I.SrcReg[1]); + inst->U.I.SrcReg[1].File = RC_FILE_TEMPORARY; + inst->U.I.SrcReg[1].Index = tmpreg; } } - return GL_TRUE; + return 1; } static void addArtificialOutputs(struct r300_vertex_program_compiler * compiler) @@ -553,44 +531,52 @@ static void addArtificialOutputs(struct r300_vertex_program_compiler * compiler) if ((compiler->RequiredOutputs & (1 << i)) && !(compiler->Base.Program.OutputsWritten & (1 << i))) { struct rc_instruction * inst = rc_insert_new_instruction(&compiler->Base, compiler->Base.Program.Instructions.Prev); - inst->I.Opcode = OPCODE_MOV; + inst->U.I.Opcode = RC_OPCODE_MOV; - inst->I.DstReg.File = PROGRAM_OUTPUT; - inst->I.DstReg.Index = i; - inst->I.DstReg.WriteMask = WRITEMASK_XYZW; + inst->U.I.DstReg.File = RC_FILE_OUTPUT; + inst->U.I.DstReg.Index = i; + inst->U.I.DstReg.WriteMask = RC_MASK_XYZW; - inst->I.SrcReg[0].File = PROGRAM_CONSTANT; - inst->I.SrcReg[0].Index = 0; - inst->I.SrcReg[0].Swizzle = SWIZZLE_XYZW; + inst->U.I.SrcReg[0].File = RC_FILE_CONSTANT; + inst->U.I.SrcReg[0].Index = 0; + inst->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_XYZW; compiler->Base.Program.OutputsWritten |= 1 << i; } } } -static void nqssadceInit(struct nqssadce_state* s) +static void dataflow_outputs_mark_used(void * userdata, void * data, + void (*callback)(void *, unsigned int, unsigned int)) { - struct r300_vertex_program_compiler * compiler = s->UserData; + struct r300_vertex_program_compiler * c = userdata; int i; - for(i = 0; i < VERT_RESULT_MAX; ++i) { - if (compiler->RequiredOutputs & (1 << i)) - s->Outputs[i].Sourced = WRITEMASK_XYZW; + for(i = 0; i < 32; ++i) { + if (c->RequiredOutputs & (1 << i)) + callback(data, i, RC_MASK_XYZW); } } -static GLboolean swizzleIsNative(GLuint opcode, struct prog_src_register reg) +static int swizzle_is_native(rc_opcode opcode, struct rc_src_register reg) { (void) opcode; (void) reg; - return GL_TRUE; + return 1; } +static struct rc_swizzle_caps r300_vertprog_swizzle_caps = { + .IsNative = &swizzle_is_native, + .Split = 0 /* should never be called */ +}; + void r3xx_compile_vertex_program(struct r300_vertex_program_compiler* compiler) { + compiler->Base.SwizzleCaps = &r300_vertprog_swizzle_caps; + addArtificialOutputs(compiler); { @@ -623,22 +609,22 @@ void r3xx_compile_vertex_program(struct r300_vertex_program_compiler* compiler) fflush(stderr); } - { - struct radeon_nqssadce_descr nqssadce = { - .Init = &nqssadceInit, - .IsNativeSwizzle = &swizzleIsNative, - .BuildSwizzle = NULL - }; - radeonNqssaDce(&compiler->Base, &nqssadce, compiler); + rc_dataflow_deadcode(&compiler->Base, &dataflow_outputs_mark_used, compiler); + + if (compiler->Base.Debug) { + fprintf(stderr, "Vertex program after deadcode:\n"); + rc_print_program(&compiler->Base.Program); + fflush(stderr); + } - /* We need this step for reusing temporary registers */ - allocate_temporary_registers(compiler); + rc_dataflow_swizzles(&compiler->Base); - if (compiler->Base.Debug) { - fprintf(stderr, "Vertex program after NQSSADCE:\n"); - rc_print_program(&compiler->Base.Program); - fflush(stderr); - } + allocate_temporary_registers(compiler); + + if (compiler->Base.Debug) { + fprintf(stderr, "Vertex program after dataflow:\n"); + rc_print_program(&compiler->Base.Program); + fflush(stderr); } translate_vertex_program(compiler); diff --git a/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog_dump.c b/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog_dump.c index 980ef3eaea1..66f9b0529f2 100644 --- a/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog_dump.c +++ b/src/mesa/drivers/dri/r300/compiler/r3xx_vertprog_dump.c @@ -146,7 +146,7 @@ static void r300_vs_op_dump(uint32_t op) static void r300_vs_src_dump(uint32_t src) { fprintf(stderr, " reg: %d%s swiz: %s%s/%s%s/%s%s/%s%s\n", - (src >> 5) & 0x7f, r300_vs_src_debug[src & 0x3], + (src >> 5) & 0xff, r300_vs_src_debug[src & 0x3], src & (1 << 25) ? "-" : " ", r300_vs_swiz_debug[(src >> 13) & 0x7], src & (1 << 26) ? "-" : " ", diff --git a/src/mesa/drivers/dri/r300/compiler/r500_fragprog.c b/src/mesa/drivers/dri/r300/compiler/r500_fragprog.c index 7e2faed6908..d87acecdabe 100644 --- a/src/mesa/drivers/dri/r300/compiler/r500_fragprog.c +++ b/src/mesa/drivers/dri/r300/compiler/r500_fragprog.c @@ -27,15 +27,17 @@ #include "r500_fragprog.h" +#include <stdio.h> + #include "../r300_reg.h" -static struct prog_src_register shadow_ambient(struct radeon_compiler * c, int tmu) +static struct rc_src_register shadow_ambient(struct radeon_compiler * c, int tmu) { - struct prog_src_register reg = { 0, }; + struct rc_src_register reg = { 0, }; - reg.File = PROGRAM_STATE_VAR; + reg.File = RC_FILE_CONSTANT; reg.Index = rc_constants_add_state(&c->Program.Constants, RC_STATE_SHADOW_AMBIENT, tmu); - reg.Swizzle = SWIZZLE_WWWW; + reg.Swizzle = RC_SWIZZLE_WWWW; return reg; } @@ -44,7 +46,7 @@ static struct prog_src_register shadow_ambient(struct radeon_compiler * c, int t * - implement texture compare (shadow extensions) * - extract non-native source / destination operands */ -GLboolean r500_transform_TEX( +int r500_transform_TEX( struct radeon_compiler * c, struct rc_instruction * inst, void* data) @@ -52,77 +54,77 @@ GLboolean r500_transform_TEX( struct r300_fragment_program_compiler *compiler = (struct r300_fragment_program_compiler*)data; - if (inst->I.Opcode != OPCODE_TEX && - inst->I.Opcode != OPCODE_TXB && - inst->I.Opcode != OPCODE_TXP && - inst->I.Opcode != OPCODE_KIL) - return GL_FALSE; + if (inst->U.I.Opcode != RC_OPCODE_TEX && + inst->U.I.Opcode != RC_OPCODE_TXB && + inst->U.I.Opcode != RC_OPCODE_TXP && + inst->U.I.Opcode != RC_OPCODE_KIL) + return 0; /* ARB_shadow & EXT_shadow_funcs */ - if (inst->I.Opcode != OPCODE_KIL && - c->Program.ShadowSamplers & (1 << inst->I.TexSrcUnit)) { - GLuint comparefunc = GL_NEVER + compiler->state.unit[inst->I.TexSrcUnit].texture_compare_func; + if (inst->U.I.Opcode != RC_OPCODE_KIL && + c->Program.ShadowSamplers & (1 << inst->U.I.TexSrcUnit)) { + rc_compare_func comparefunc = compiler->state.unit[inst->U.I.TexSrcUnit].texture_compare_func; - if (comparefunc == GL_NEVER || comparefunc == GL_ALWAYS) { - inst->I.Opcode = OPCODE_MOV; + if (comparefunc == RC_COMPARE_FUNC_NEVER || comparefunc == RC_COMPARE_FUNC_ALWAYS) { + inst->U.I.Opcode = RC_OPCODE_MOV; - if (comparefunc == GL_ALWAYS) { - inst->I.SrcReg[0].File = PROGRAM_BUILTIN; - inst->I.SrcReg[0].Swizzle = SWIZZLE_1111; + if (comparefunc == RC_COMPARE_FUNC_ALWAYS) { + inst->U.I.SrcReg[0].File = RC_FILE_NONE; + inst->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_1111; } else { - inst->I.SrcReg[0] = shadow_ambient(c, inst->I.TexSrcUnit); + inst->U.I.SrcReg[0] = shadow_ambient(c, inst->U.I.TexSrcUnit); } - return GL_TRUE; + return 1; } else { - GLuint comparefunc = GL_NEVER + compiler->state.unit[inst->I.TexSrcUnit].texture_compare_func; - GLuint depthmode = compiler->state.unit[inst->I.TexSrcUnit].depth_texture_mode; + rc_compare_func comparefunc = compiler->state.unit[inst->U.I.TexSrcUnit].texture_compare_func; + unsigned int depthmode = compiler->state.unit[inst->U.I.TexSrcUnit].depth_texture_mode; struct rc_instruction * inst_rcp = rc_insert_new_instruction(c, inst); struct rc_instruction * inst_mad = rc_insert_new_instruction(c, inst_rcp); struct rc_instruction * inst_cmp = rc_insert_new_instruction(c, inst_mad); int pass, fail; - inst_rcp->I.Opcode = OPCODE_RCP; - inst_rcp->I.DstReg.File = PROGRAM_TEMPORARY; - inst_rcp->I.DstReg.Index = rc_find_free_temporary(c); - inst_rcp->I.DstReg.WriteMask = WRITEMASK_W; - inst_rcp->I.SrcReg[0] = inst->I.SrcReg[0]; - inst_rcp->I.SrcReg[0].Swizzle = SWIZZLE_WWWW; - - inst_cmp->I.DstReg = inst->I.DstReg; - inst->I.DstReg.File = PROGRAM_TEMPORARY; - inst->I.DstReg.Index = rc_find_free_temporary(c); - inst->I.DstReg.WriteMask = WRITEMASK_XYZW; - - inst_mad->I.Opcode = OPCODE_MAD; - inst_mad->I.DstReg.File = PROGRAM_TEMPORARY; - inst_mad->I.DstReg.Index = rc_find_free_temporary(c); - inst_mad->I.SrcReg[0] = inst->I.SrcReg[0]; - inst_mad->I.SrcReg[0].Swizzle = SWIZZLE_ZZZZ; - inst_mad->I.SrcReg[1].File = PROGRAM_TEMPORARY; - inst_mad->I.SrcReg[1].Index = inst_rcp->I.DstReg.Index; - inst_mad->I.SrcReg[1].Swizzle = SWIZZLE_WWWW; - inst_mad->I.SrcReg[2].File = PROGRAM_TEMPORARY; - inst_mad->I.SrcReg[2].Index = inst->I.DstReg.Index; + inst_rcp->U.I.Opcode = RC_OPCODE_RCP; + inst_rcp->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst_rcp->U.I.DstReg.Index = rc_find_free_temporary(c); + inst_rcp->U.I.DstReg.WriteMask = RC_MASK_W; + inst_rcp->U.I.SrcReg[0] = inst->U.I.SrcReg[0]; + inst_rcp->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_WWWW; + + inst_cmp->U.I.DstReg = inst->U.I.DstReg; + inst->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst->U.I.DstReg.Index = rc_find_free_temporary(c); + inst->U.I.DstReg.WriteMask = RC_MASK_XYZW; + + inst_mad->U.I.Opcode = RC_OPCODE_MAD; + inst_mad->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst_mad->U.I.DstReg.Index = rc_find_free_temporary(c); + inst_mad->U.I.SrcReg[0] = inst->U.I.SrcReg[0]; + inst_mad->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_ZZZZ; + inst_mad->U.I.SrcReg[1].File = RC_FILE_TEMPORARY; + inst_mad->U.I.SrcReg[1].Index = inst_rcp->U.I.DstReg.Index; + inst_mad->U.I.SrcReg[1].Swizzle = RC_SWIZZLE_WWWW; + inst_mad->U.I.SrcReg[2].File = RC_FILE_TEMPORARY; + inst_mad->U.I.SrcReg[2].Index = inst->U.I.DstReg.Index; if (depthmode == 0) /* GL_LUMINANCE */ - inst_mad->I.SrcReg[2].Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_Z); + inst_mad->U.I.SrcReg[2].Swizzle = RC_MAKE_SWIZZLE(RC_SWIZZLE_X, RC_SWIZZLE_Y, RC_SWIZZLE_Z, RC_SWIZZLE_Z); else if (depthmode == 2) /* GL_ALPHA */ - inst_mad->I.SrcReg[2].Swizzle = SWIZZLE_WWWW; + inst_mad->U.I.SrcReg[2].Swizzle = RC_SWIZZLE_WWWW; /* Recall that SrcReg[0] is tex, SrcReg[2] is r and: * r < tex <=> -tex+r < 0 * r >= tex <=> not (-tex+r < 0 */ - if (comparefunc == GL_LESS || comparefunc == GL_GEQUAL) - inst_mad->I.SrcReg[2].Negate = inst_mad->I.SrcReg[2].Negate ^ NEGATE_XYZW; + if (comparefunc == RC_COMPARE_FUNC_LESS || comparefunc == RC_COMPARE_FUNC_GEQUAL) + inst_mad->U.I.SrcReg[2].Negate = inst_mad->U.I.SrcReg[2].Negate ^ RC_MASK_XYZW; else - inst_mad->I.SrcReg[0].Negate = inst_mad->I.SrcReg[0].Negate ^ NEGATE_XYZW; + inst_mad->U.I.SrcReg[0].Negate = inst_mad->U.I.SrcReg[0].Negate ^ RC_MASK_XYZW; - inst_cmp->I.Opcode = OPCODE_CMP; + inst_cmp->U.I.Opcode = RC_OPCODE_CMP; /* DstReg has been filled out above */ - inst_cmp->I.SrcReg[0].File = PROGRAM_TEMPORARY; - inst_cmp->I.SrcReg[0].Index = inst_mad->I.DstReg.Index; + inst_cmp->U.I.SrcReg[0].File = RC_FILE_TEMPORARY; + inst_cmp->U.I.SrcReg[0].Index = inst_mad->U.I.DstReg.Index; - if (comparefunc == GL_LESS || comparefunc == GL_GREATER) { + if (comparefunc == RC_COMPARE_FUNC_LESS || comparefunc == RC_COMPARE_FUNC_GREATER) { pass = 1; fail = 2; } else { @@ -130,131 +132,161 @@ GLboolean r500_transform_TEX( fail = 1; } - inst_cmp->I.SrcReg[pass].File = PROGRAM_BUILTIN; - inst_cmp->I.SrcReg[pass].Swizzle = SWIZZLE_1111; - inst_cmp->I.SrcReg[fail] = shadow_ambient(c, inst->I.TexSrcUnit); + inst_cmp->U.I.SrcReg[pass].File = RC_FILE_NONE; + inst_cmp->U.I.SrcReg[pass].Swizzle = RC_SWIZZLE_1111; + inst_cmp->U.I.SrcReg[fail] = shadow_ambient(c, inst->U.I.TexSrcUnit); } } /* Cannot write texture to output registers */ - if (inst->I.Opcode != OPCODE_KIL && inst->I.DstReg.File != PROGRAM_TEMPORARY) { + if (inst->U.I.Opcode != RC_OPCODE_KIL && inst->U.I.DstReg.File != RC_FILE_TEMPORARY) { struct rc_instruction * inst_mov = rc_insert_new_instruction(c, inst); - inst_mov->I.Opcode = OPCODE_MOV; - inst_mov->I.DstReg = inst->I.DstReg; - inst_mov->I.SrcReg[0].File = PROGRAM_TEMPORARY; - inst_mov->I.SrcReg[0].Index = rc_find_free_temporary(c); + inst_mov->U.I.Opcode = RC_OPCODE_MOV; + inst_mov->U.I.DstReg = inst->U.I.DstReg; + inst_mov->U.I.SrcReg[0].File = RC_FILE_TEMPORARY; + inst_mov->U.I.SrcReg[0].Index = rc_find_free_temporary(c); - inst->I.DstReg.File = PROGRAM_TEMPORARY; - inst->I.DstReg.Index = inst_mov->I.SrcReg[0].Index; - inst->I.DstReg.WriteMask = WRITEMASK_XYZW; + inst->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst->U.I.DstReg.Index = inst_mov->U.I.SrcReg[0].Index; + inst->U.I.DstReg.WriteMask = RC_MASK_XYZW; } /* Cannot read texture coordinate from constants file */ - if (inst->I.SrcReg[0].File != PROGRAM_TEMPORARY && inst->I.SrcReg[0].File != PROGRAM_INPUT) { + if (inst->U.I.SrcReg[0].File != RC_FILE_TEMPORARY && inst->U.I.SrcReg[0].File != RC_FILE_INPUT) { struct rc_instruction * inst_mov = rc_insert_new_instruction(c, inst->Prev); - inst_mov->I.Opcode = OPCODE_MOV; - inst_mov->I.DstReg.File = PROGRAM_TEMPORARY; - inst_mov->I.DstReg.Index = rc_find_free_temporary(c); - inst_mov->I.SrcReg[0] = inst->I.SrcReg[0]; + inst_mov->U.I.Opcode = RC_OPCODE_MOV; + inst_mov->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst_mov->U.I.DstReg.Index = rc_find_free_temporary(c); + inst_mov->U.I.SrcReg[0] = inst->U.I.SrcReg[0]; - reset_srcreg(&inst->I.SrcReg[0]); - inst->I.SrcReg[0].File = PROGRAM_TEMPORARY; - inst->I.SrcReg[0].Index = inst_mov->I.DstReg.Index; + reset_srcreg(&inst->U.I.SrcReg[0]); + inst->U.I.SrcReg[0].File = RC_FILE_TEMPORARY; + inst->U.I.SrcReg[0].Index = inst_mov->U.I.DstReg.Index; } - return GL_TRUE; + return 1; +} + +/** + * Rewrite IF instructions to use the ALU result special register. + */ +int r500_transform_IF( + struct radeon_compiler * c, + struct rc_instruction * inst, + void* data) +{ + if (inst->U.I.Opcode != RC_OPCODE_IF) + return 0; + + struct rc_instruction * inst_mov = rc_insert_new_instruction(c, inst->Prev); + inst_mov->U.I.Opcode = RC_OPCODE_MOV; + inst_mov->U.I.DstReg.WriteMask = 0; + inst_mov->U.I.WriteALUResult = RC_ALURESULT_W; + inst_mov->U.I.ALUResultCompare = RC_COMPARE_FUNC_NOTEQUAL; + inst_mov->U.I.SrcReg[0] = inst->U.I.SrcReg[0]; + inst_mov->U.I.SrcReg[0].Swizzle = combine_swizzles4(inst_mov->U.I.SrcReg[0].Swizzle, + RC_SWIZZLE_UNUSED, RC_SWIZZLE_UNUSED, RC_SWIZZLE_UNUSED, RC_SWIZZLE_X); + + inst->U.I.SrcReg[0].File = RC_FILE_SPECIAL; + inst->U.I.SrcReg[0].Index = RC_SPECIAL_ALU_RESULT; + inst->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_XYZW; + inst->U.I.SrcReg[0].Negate = 0; + + return 1; } -GLboolean r500FPIsNativeSwizzle(GLuint opcode, struct prog_src_register reg) +static int r500_swizzle_is_native(rc_opcode opcode, struct rc_src_register reg) { - GLuint relevant; + unsigned int relevant; int i; - if (opcode == OPCODE_TEX || - opcode == OPCODE_TXB || - opcode == OPCODE_TXP || - opcode == OPCODE_KIL) { + if (opcode == RC_OPCODE_TEX || + opcode == RC_OPCODE_TXB || + opcode == RC_OPCODE_TXP || + opcode == RC_OPCODE_KIL) { if (reg.Abs) - return GL_FALSE; + return 0; - if (opcode == OPCODE_KIL && (reg.Swizzle != SWIZZLE_NOOP || reg.Negate != NEGATE_NONE)) - return GL_FALSE; + if (opcode == RC_OPCODE_KIL && (reg.Swizzle != RC_SWIZZLE_XYZW || reg.Negate != RC_MASK_NONE)) + return 0; if (reg.Negate) - reg.Negate ^= NEGATE_XYZW; + reg.Negate ^= RC_MASK_XYZW; for(i = 0; i < 4; ++i) { - GLuint swz = GET_SWZ(reg.Swizzle, i); - if (swz == SWIZZLE_NIL) { + unsigned int swz = GET_SWZ(reg.Swizzle, i); + if (swz == RC_SWIZZLE_UNUSED) { reg.Negate &= ~(1 << i); continue; } if (swz >= 4) - return GL_FALSE; + return 0; } if (reg.Negate) - return GL_FALSE; + return 0; - return GL_TRUE; - } else if (opcode == OPCODE_DDX || opcode == OPCODE_DDY) { + return 1; + } else if (opcode == RC_OPCODE_DDX || opcode == RC_OPCODE_DDY) { /* DDX/MDH and DDY/MDV explicitly ignore incoming swizzles; * if it doesn't fit perfectly into a .xyzw case... */ - if (reg.Swizzle == SWIZZLE_NOOP && !reg.Abs && !reg.Negate) - return GL_TRUE; + if (reg.Swizzle == RC_SWIZZLE_XYZW && !reg.Abs && !reg.Negate) + return 1; - return GL_FALSE; + return 0; } else { /* ALU instructions support almost everything */ if (reg.Abs) - return GL_TRUE; + return 1; relevant = 0; for(i = 0; i < 3; ++i) { - GLuint swz = GET_SWZ(reg.Swizzle, i); - if (swz != SWIZZLE_NIL && swz != SWIZZLE_ZERO) + unsigned int swz = GET_SWZ(reg.Swizzle, i); + if (swz != RC_SWIZZLE_UNUSED && swz != RC_SWIZZLE_ZERO) relevant |= 1 << i; } if ((reg.Negate & relevant) && ((reg.Negate & relevant) != relevant)) - return GL_FALSE; + return 0; - return GL_TRUE; + return 1; } } /** - * Implement a MOV with a potentially non-native swizzle. + * Split source register access. * * The only thing we *cannot* do in an ALU instruction is per-component - * negation. Therefore, we split the MOV into two instructions when necessary. + * negation. */ -void r500FPBuildSwizzle(struct nqssadce_state *s, struct prog_dst_register dst, struct prog_src_register src) +static void r500_swizzle_split(struct rc_src_register src, unsigned int usemask, + struct rc_swizzle_split * split) { - GLuint negatebase[2] = { 0, 0 }; + unsigned int negatebase[2] = { 0, 0 }; int i; for(i = 0; i < 4; ++i) { - GLuint swz = GET_SWZ(src.Swizzle, i); - if (swz == SWIZZLE_NIL) + unsigned int swz = GET_SWZ(src.Swizzle, i); + if (swz == RC_SWIZZLE_UNUSED || !GET_BIT(usemask, i)) continue; negatebase[GET_BIT(src.Negate, i)] |= 1 << i; } + split->NumPhases = 0; + for(i = 0; i <= 1; ++i) { if (!negatebase[i]) continue; - struct rc_instruction *inst = rc_insert_new_instruction(s->Compiler, s->IP->Prev); - inst->I.Opcode = OPCODE_MOV; - inst->I.DstReg = dst; - inst->I.DstReg.WriteMask = negatebase[i]; - inst->I.SrcReg[0] = src; - inst->I.SrcReg[0].Negate = (i == 0) ? NEGATE_NONE : NEGATE_XYZW; + split->Phase[split->NumPhases++] = negatebase[i]; } } +struct rc_swizzle_caps r500_swizzle_caps = { + .IsNative = r500_swizzle_is_native, + .Split = r500_swizzle_split +}; static char *toswiz(int swiz_val) { switch(swiz_val) { diff --git a/src/mesa/drivers/dri/r300/compiler/r500_fragprog.h b/src/mesa/drivers/dri/r300/compiler/r500_fragprog.h index 9091f65cd20..0918cdf518b 100644 --- a/src/mesa/drivers/dri/r300/compiler/r500_fragprog.h +++ b/src/mesa/drivers/dri/r300/compiler/r500_fragprog.h @@ -33,21 +33,21 @@ #ifndef __R500_FRAGPROG_H_ #define __R500_FRAGPROG_H_ -#include "shader/prog_parameter.h" -#include "shader/prog_instruction.h" - #include "radeon_compiler.h" -#include "radeon_nqssadce.h" +#include "radeon_swizzle.h" extern void r500BuildFragmentProgramHwCode(struct r300_fragment_program_compiler *compiler); extern void r500FragmentProgramDump(struct rX00_fragment_program_code *c); -extern GLboolean r500FPIsNativeSwizzle(GLuint opcode, struct prog_src_register reg); +extern struct rc_swizzle_caps r500_swizzle_caps; -extern void r500FPBuildSwizzle(struct nqssadce_state *s, struct prog_dst_register dst, struct prog_src_register src); +extern int r500_transform_TEX( + struct radeon_compiler * c, + struct rc_instruction * inst, + void* data); -extern GLboolean r500_transform_TEX( +extern int r500_transform_IF( struct radeon_compiler * c, struct rc_instruction * inst, void* data); diff --git a/src/mesa/drivers/dri/r300/compiler/r500_fragprog_emit.c b/src/mesa/drivers/dri/r300/compiler/r500_fragprog_emit.c index d694725c9bb..b1b14394b6e 100644 --- a/src/mesa/drivers/dri/r300/compiler/r500_fragprog_emit.c +++ b/src/mesa/drivers/dri/r300/compiler/r500_fragprog_emit.c @@ -37,10 +37,6 @@ * * \author Corbin Simpson <[email protected]> * - * \todo Depth write, WPOS/FOGC inputs - * - * \todo FogOption - * */ #include "r500_fragprog.h" @@ -51,7 +47,6 @@ #define PROG_CODE \ - struct r300_fragment_program_compiler *c = (struct r300_fragment_program_compiler*)data; \ struct r500_fragment_program_code *code = &c->code->code.r500 #define error(fmt, args...) do { \ @@ -60,63 +55,80 @@ } while(0) -static GLuint translate_rgb_op(struct r300_fragment_program_compiler *c, GLuint opcode) +struct branch_info { + int If; + int Else; + int Endif; +}; + +struct emit_state { + struct radeon_compiler * C; + struct r500_fragment_program_code * Code; + + struct branch_info * Branches; + unsigned int CurrentBranchDepth; + unsigned int BranchesReserved; + + unsigned int MaxBranchDepth; +}; + +static unsigned int translate_rgb_op(struct r300_fragment_program_compiler *c, rc_opcode opcode) { switch(opcode) { - case OPCODE_CMP: return R500_ALU_RGBA_OP_CMP; - case OPCODE_DDX: return R500_ALU_RGBA_OP_MDH; - case OPCODE_DDY: return R500_ALU_RGBA_OP_MDV; - case OPCODE_DP3: return R500_ALU_RGBA_OP_DP3; - case OPCODE_DP4: return R500_ALU_RGBA_OP_DP4; - case OPCODE_FRC: return R500_ALU_RGBA_OP_FRC; + case RC_OPCODE_CMP: return R500_ALU_RGBA_OP_CMP; + case RC_OPCODE_DDX: return R500_ALU_RGBA_OP_MDH; + case RC_OPCODE_DDY: return R500_ALU_RGBA_OP_MDV; + case RC_OPCODE_DP3: return R500_ALU_RGBA_OP_DP3; + case RC_OPCODE_DP4: return R500_ALU_RGBA_OP_DP4; + case RC_OPCODE_FRC: return R500_ALU_RGBA_OP_FRC; default: error("translate_rgb_op(%d): unknown opcode\n", opcode); /* fall through */ - case OPCODE_NOP: + case RC_OPCODE_NOP: /* fall through */ - case OPCODE_MAD: return R500_ALU_RGBA_OP_MAD; - case OPCODE_MAX: return R500_ALU_RGBA_OP_MAX; - case OPCODE_MIN: return R500_ALU_RGBA_OP_MIN; - case OPCODE_REPL_ALPHA: return R500_ALU_RGBA_OP_SOP; + case RC_OPCODE_MAD: return R500_ALU_RGBA_OP_MAD; + case RC_OPCODE_MAX: return R500_ALU_RGBA_OP_MAX; + case RC_OPCODE_MIN: return R500_ALU_RGBA_OP_MIN; + case RC_OPCODE_REPL_ALPHA: return R500_ALU_RGBA_OP_SOP; } } -static GLuint translate_alpha_op(struct r300_fragment_program_compiler *c, GLuint opcode) +static unsigned int translate_alpha_op(struct r300_fragment_program_compiler *c, rc_opcode opcode) { switch(opcode) { - case OPCODE_CMP: return R500_ALPHA_OP_CMP; - case OPCODE_COS: return R500_ALPHA_OP_COS; - case OPCODE_DDX: return R500_ALPHA_OP_MDH; - case OPCODE_DDY: return R500_ALPHA_OP_MDV; - case OPCODE_DP3: return R500_ALPHA_OP_DP; - case OPCODE_DP4: return R500_ALPHA_OP_DP; - case OPCODE_EX2: return R500_ALPHA_OP_EX2; - case OPCODE_FRC: return R500_ALPHA_OP_FRC; - case OPCODE_LG2: return R500_ALPHA_OP_LN2; + case RC_OPCODE_CMP: return R500_ALPHA_OP_CMP; + case RC_OPCODE_COS: return R500_ALPHA_OP_COS; + case RC_OPCODE_DDX: return R500_ALPHA_OP_MDH; + case RC_OPCODE_DDY: return R500_ALPHA_OP_MDV; + case RC_OPCODE_DP3: return R500_ALPHA_OP_DP; + case RC_OPCODE_DP4: return R500_ALPHA_OP_DP; + case RC_OPCODE_EX2: return R500_ALPHA_OP_EX2; + case RC_OPCODE_FRC: return R500_ALPHA_OP_FRC; + case RC_OPCODE_LG2: return R500_ALPHA_OP_LN2; default: error("translate_alpha_op(%d): unknown opcode\n", opcode); /* fall through */ - case OPCODE_NOP: + case RC_OPCODE_NOP: /* fall through */ - case OPCODE_MAD: return R500_ALPHA_OP_MAD; - case OPCODE_MAX: return R500_ALPHA_OP_MAX; - case OPCODE_MIN: return R500_ALPHA_OP_MIN; - case OPCODE_RCP: return R500_ALPHA_OP_RCP; - case OPCODE_RSQ: return R500_ALPHA_OP_RSQ; - case OPCODE_SIN: return R500_ALPHA_OP_SIN; + case RC_OPCODE_MAD: return R500_ALPHA_OP_MAD; + case RC_OPCODE_MAX: return R500_ALPHA_OP_MAX; + case RC_OPCODE_MIN: return R500_ALPHA_OP_MIN; + case RC_OPCODE_RCP: return R500_ALPHA_OP_RCP; + case RC_OPCODE_RSQ: return R500_ALPHA_OP_RSQ; + case RC_OPCODE_SIN: return R500_ALPHA_OP_SIN; } } -static GLuint fix_hw_swizzle(GLuint swz) +static unsigned int fix_hw_swizzle(unsigned int swz) { if (swz == 5) swz = 6; - if (swz == SWIZZLE_NIL) swz = 4; + if (swz == RC_SWIZZLE_UNUSED) swz = 4; return swz; } -static GLuint translate_arg_rgb(struct radeon_pair_instruction *inst, int arg) +static unsigned int translate_arg_rgb(struct rc_pair_instruction *inst, int arg) { - GLuint t = inst->RGB.Arg[arg].Source; + unsigned int t = inst->RGB.Arg[arg].Source; int comp; t |= inst->RGB.Arg[arg].Negate << 11; t |= inst->RGB.Arg[arg].Abs << 12; @@ -127,39 +139,57 @@ static GLuint translate_arg_rgb(struct radeon_pair_instruction *inst, int arg) return t; } -static GLuint translate_arg_alpha(struct radeon_pair_instruction *inst, int i) +static unsigned int translate_arg_alpha(struct rc_pair_instruction *inst, int i) { - GLuint t = inst->Alpha.Arg[i].Source; + unsigned int t = inst->Alpha.Arg[i].Source; t |= fix_hw_swizzle(inst->Alpha.Arg[i].Swizzle) << 2; t |= inst->Alpha.Arg[i].Negate << 5; t |= inst->Alpha.Arg[i].Abs << 6; return t; } -static void use_temporary(struct r500_fragment_program_code* code, GLuint index) +static uint32_t translate_alu_result_op(struct r300_fragment_program_compiler * c, rc_compare_func func) +{ + switch(func) { + case RC_COMPARE_FUNC_EQUAL: return R500_INST_ALU_RESULT_OP_EQ; + case RC_COMPARE_FUNC_LESS: return R500_INST_ALU_RESULT_OP_LT; + case RC_COMPARE_FUNC_GEQUAL: return R500_INST_ALU_RESULT_OP_GE; + case RC_COMPARE_FUNC_NOTEQUAL: return R500_INST_ALU_RESULT_OP_NE; + default: + rc_error(&c->Base, "%s: unsupported compare func %i\n", __FUNCTION__, func); + return 0; + } +} + +static void use_temporary(struct r500_fragment_program_code* code, unsigned int index) { if (index > code->max_temp_idx) code->max_temp_idx = index; } -static GLuint use_source(struct r500_fragment_program_code* code, struct radeon_pair_instruction_source src) +static unsigned int use_source(struct r500_fragment_program_code* code, struct radeon_pair_instruction_source src) { - if (!src.Constant) + if (src.File == RC_FILE_CONSTANT) { + return src.Index | 0x100; + } else if (src.File == RC_FILE_TEMPORARY) { use_temporary(code, src.Index); - return src.Index | src.Constant << 8; + return src.Index; + } + + return 0; } /** * Emit a paired ALU instruction. */ -static GLboolean emit_paired(void *data, struct radeon_pair_instruction *inst) +static void emit_paired(struct r300_fragment_program_compiler *c, struct rc_pair_instruction *inst) { PROG_CODE; if (code->inst_end >= 511) { error("emit_alu: Too many instructions"); - return GL_FALSE; + return; } int ip = ++code->inst_end; @@ -167,17 +197,22 @@ static GLboolean emit_paired(void *data, struct radeon_pair_instruction *inst) code->inst[ip].inst5 = translate_rgb_op(c, inst->RGB.Opcode); code->inst[ip].inst4 = translate_alpha_op(c, inst->Alpha.Opcode); - if (inst->RGB.OutputWriteMask || inst->Alpha.OutputWriteMask || inst->Alpha.DepthWriteMask) + if (inst->RGB.OutputWriteMask || inst->Alpha.OutputWriteMask || inst->Alpha.DepthWriteMask) { code->inst[ip].inst0 = R500_INST_TYPE_OUT; - else + if (inst->WriteALUResult) { + error("%s: cannot write output and ALU result at the same time"); + return; + } + } else { code->inst[ip].inst0 = R500_INST_TYPE_ALU; + } code->inst[ip].inst0 |= R500_INST_TEX_SEM_WAIT; code->inst[ip].inst0 |= (inst->RGB.WriteMask << 11) | (inst->Alpha.WriteMask << 14); code->inst[ip].inst0 |= (inst->RGB.OutputWriteMask << 15) | (inst->Alpha.OutputWriteMask << 18); if (inst->Alpha.DepthWriteMask) { code->inst[ip].inst4 |= R500_ALPHA_W_OMASK; - c->code->writes_depth = GL_TRUE; + c->code->writes_depth = 1; } code->inst[ip].inst4 |= R500_ALPHA_ADDRD(inst->Alpha.DestIndex); @@ -206,12 +241,21 @@ static GLboolean emit_paired(void *data, struct radeon_pair_instruction *inst) code->inst[ip].inst4 |= translate_arg_alpha(inst, 1) << R500_ALPHA_SEL_B_SHIFT; code->inst[ip].inst5 |= translate_arg_alpha(inst, 2) << R500_ALU_RGBA_ALPHA_SEL_C_SHIFT; - return GL_TRUE; + if (inst->WriteALUResult) { + code->inst[ip].inst3 |= R500_ALU_RGB_WMASK; + + if (inst->WriteALUResult == RC_ALURESULT_X) + code->inst[ip].inst0 |= R500_INST_ALU_RESULT_SEL_RED; + else + code->inst[ip].inst0 |= R500_INST_ALU_RESULT_SEL_ALPHA; + + code->inst[ip].inst0 |= translate_alu_result_op(c, inst->ALUResultCompare); + } } -static GLuint translate_strq_swizzle(GLuint swizzle) +static unsigned int translate_strq_swizzle(unsigned int swizzle) { - GLuint swiz = 0; + unsigned int swiz = 0; int i; for (i = 0; i < 4; i++) swiz |= (GET_SWZ(swizzle, i) & 0x3) << i*2; @@ -221,67 +265,193 @@ static GLuint translate_strq_swizzle(GLuint swizzle) /** * Emit a single TEX instruction */ -static GLboolean emit_tex(void *data, struct radeon_pair_texture_instruction *inst) +static int emit_tex(struct r300_fragment_program_compiler *c, struct rc_sub_instruction *inst) { PROG_CODE; if (code->inst_end >= 511) { error("emit_tex: Too many instructions"); - return GL_FALSE; + return 0; } int ip = ++code->inst_end; code->inst[ip].inst0 = R500_INST_TYPE_TEX - | (inst->WriteMask << 11) + | (inst->DstReg.WriteMask << 11) | R500_INST_TEX_SEM_WAIT; code->inst[ip].inst1 = R500_TEX_ID(inst->TexSrcUnit) | R500_TEX_SEM_ACQUIRE | R500_TEX_IGNORE_UNCOVERED; - if (inst->TexSrcTarget == TEXTURE_RECT_INDEX) - code->inst[ip].inst1 |= R500_TEX_UNSCALED; + if (inst->TexSrcTarget == RC_TEXTURE_RECT) + code->inst[ip].inst1 |= R500_TEX_UNSCALED; switch (inst->Opcode) { - case RADEON_OPCODE_KIL: + case RC_OPCODE_KIL: code->inst[ip].inst1 |= R500_TEX_INST_TEXKILL; break; - case RADEON_OPCODE_TEX: + case RC_OPCODE_TEX: code->inst[ip].inst1 |= R500_TEX_INST_LD; break; - case RADEON_OPCODE_TXB: + case RC_OPCODE_TXB: code->inst[ip].inst1 |= R500_TEX_INST_LODBIAS; break; - case RADEON_OPCODE_TXP: + case RC_OPCODE_TXP: code->inst[ip].inst1 |= R500_TEX_INST_PROJ; break; default: error("emit_tex can't handle opcode %x\n", inst->Opcode); } - code->inst[ip].inst2 = R500_TEX_SRC_ADDR(inst->SrcIndex) - | (translate_strq_swizzle(inst->SrcSwizzle) << 8) - | R500_TEX_DST_ADDR(inst->DestIndex) + use_temporary(code, inst->SrcReg[0].Index); + if (inst->Opcode != RC_OPCODE_KIL) + use_temporary(code, inst->DstReg.Index); + + code->inst[ip].inst2 = R500_TEX_SRC_ADDR(inst->SrcReg[0].Index) + | (translate_strq_swizzle(inst->SrcReg[0].Swizzle) << 8) + | R500_TEX_DST_ADDR(inst->DstReg.Index) | R500_TEX_DST_R_SWIZ_R | R500_TEX_DST_G_SWIZ_G | R500_TEX_DST_B_SWIZ_B | R500_TEX_DST_A_SWIZ_A; - return GL_TRUE; + return 1; } -static const struct radeon_pair_handler pair_handler = { - .EmitPaired = emit_paired, - .EmitTex = emit_tex, - .MaxHwTemps = 128 -}; +static void grow_branches(struct emit_state * s) +{ + unsigned int newreserved = s->BranchesReserved * 2; + struct branch_info * newbranches; + + if (!newreserved) + newreserved = 4; + + newbranches = memory_pool_malloc(&s->C->Pool, newreserved*sizeof(struct branch_info)); + memcpy(newbranches, s->Branches, s->CurrentBranchDepth*sizeof(struct branch_info)); + + s->Branches = newbranches; + s->BranchesReserved = newreserved; +} + +static void emit_flowcontrol(struct emit_state * s, struct rc_instruction * inst) +{ + if (s->Code->inst_end >= 511) { + rc_error(s->C, "emit_tex: Too many instructions"); + return; + } + + unsigned int newip = ++s->Code->inst_end; + + s->Code->inst[newip].inst0 = R500_INST_TYPE_FC | R500_INST_ALU_WAIT; + + if (inst->U.I.Opcode == RC_OPCODE_IF) { + if (s->CurrentBranchDepth >= 32) { + rc_error(s->C, "Branch depth exceeds hardware limit"); + return; + } + + if (s->CurrentBranchDepth >= s->BranchesReserved) + grow_branches(s); + + struct branch_info * branch = &s->Branches[s->CurrentBranchDepth++]; + branch->If = newip; + branch->Else = -1; + branch->Endif = -1; + + if (s->CurrentBranchDepth > s->MaxBranchDepth) + s->MaxBranchDepth = s->CurrentBranchDepth; + + /* actual instruction is filled in at ENDIF time */ + } else if (inst->U.I.Opcode == RC_OPCODE_ELSE) { + if (!s->CurrentBranchDepth) { + rc_error(s->C, "%s: got ELSE outside a branch", __FUNCTION__); + return; + } + + struct branch_info * branch = &s->Branches[s->CurrentBranchDepth - 1]; + branch->Else = newip; + + /* actual instruction is filled in at ENDIF time */ + } else if (inst->U.I.Opcode == RC_OPCODE_ENDIF) { + if (!s->CurrentBranchDepth) { + rc_error(s->C, "%s: got ELSE outside a branch", __FUNCTION__); + return; + } + + struct branch_info * branch = &s->Branches[s->CurrentBranchDepth - 1]; + branch->Endif = newip; + + s->Code->inst[branch->If].inst2 = R500_FC_OP_JUMP + | R500_FC_A_OP_NONE /* no address stack */ + | R500_FC_JUMP_FUNC(0x0f) /* jump if ALU result is false */ + | R500_FC_B_OP0_INCR /* increment branch counter if stay */ + ; + + if (branch->Else >= 0) { + /* increment branch counter also if jump */ + s->Code->inst[branch->If].inst2 |= R500_FC_B_OP1_INCR; + s->Code->inst[branch->If].inst3 = R500_FC_JUMP_ADDR(branch->Else + 1); + + s->Code->inst[branch->Else].inst2 = R500_FC_OP_JUMP + | R500_FC_A_OP_NONE /* no address stack */ + | R500_FC_B_ELSE /* all active pixels want to jump */ + | R500_FC_B_OP0_NONE /* no counter op if stay */ + | R500_FC_B_OP1_DECR /* decrement branch counter if jump */ + | R500_FC_B_POP_CNT(1) + ; + s->Code->inst[branch->Else].inst3 = R500_FC_JUMP_ADDR(branch->Endif + 1); + } else { + /* don't touch branch counter on jump */ + s->Code->inst[branch->If].inst2 |= R500_FC_B_OP1_NONE; + s->Code->inst[branch->If].inst3 = R500_FC_JUMP_ADDR(branch->Endif + 1); + } + + s->Code->inst[branch->Endif].inst2 = R500_FC_OP_JUMP + | R500_FC_A_OP_NONE /* no address stack */ + | R500_FC_JUMP_ANY /* docs says set this, but I don't understand why */ + | R500_FC_B_OP0_DECR /* decrement branch counter if stay */ + | R500_FC_B_OP1_NONE /* no branch counter if stay */ + | R500_FC_B_POP_CNT(1) + ; + s->Code->inst[branch->Endif].inst3 = R500_FC_JUMP_ADDR(branch->Endif + 1); + + s->CurrentBranchDepth--; + } else { + rc_error(s->C, "%s: unknown opcode %i\n", __FUNCTION__, inst->U.I.Opcode); + } +} void r500BuildFragmentProgramHwCode(struct r300_fragment_program_compiler *compiler) { + struct emit_state s; struct r500_fragment_program_code *code = &compiler->code->code.r500; - _mesa_bzero(code, sizeof(*code)); + memset(&s, 0, sizeof(s)); + s.C = &compiler->Base; + s.Code = code; + + memset(code, 0, sizeof(*code)); code->max_temp_idx = 1; code->inst_end = -1; - radeonPairProgram(compiler, &pair_handler, compiler); + for(struct rc_instruction * inst = compiler->Base.Program.Instructions.Next; + inst != &compiler->Base.Program.Instructions && !compiler->Base.Error; + inst = inst->Next) { + if (inst->Type == RC_INSTRUCTION_NORMAL) { + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->U.I.Opcode); + + if (opcode->IsFlowControl) { + emit_flowcontrol(&s, inst); + } else if (inst->U.I.Opcode == RC_OPCODE_BEGIN_TEX) { + continue; + } else { + emit_tex(compiler, &inst->U.I); + } + } else { + emit_paired(compiler, &inst->U.P); + } + } + + if (code->max_temp_idx >= 128) + rc_error(&compiler->Base, "Too many hardware temporaries used"); + if (compiler->Base.Error) return; @@ -296,4 +466,11 @@ void r500BuildFragmentProgramHwCode(struct r300_fragment_program_compiler *compi int ip = ++code->inst_end; code->inst[ip].inst0 = R500_INST_TYPE_OUT | R500_INST_TEX_SEM_WAIT; } + + if (s.MaxBranchDepth >= 4) { + if (code->max_temp_idx < 1) + code->max_temp_idx = 1; + + code->us_fc_ctrl |= R500_FC_FULL_FC_EN; + } } diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_code.c b/src/mesa/drivers/dri/r300/compiler/radeon_code.c index c7923004df9..1a3d8bb6412 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_code.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_code.c @@ -25,11 +25,13 @@ * */ -#include "main/mtypes.h" -#include "shader/prog_instruction.h" - #include "radeon_code.h" +#include <stdlib.h> +#include <string.h> + +#include "radeon_program.h" + void rc_constants_init(struct rc_constant_list * c) { memset(c, 0, sizeof(*c)); @@ -143,7 +145,7 @@ unsigned rc_constants_add_immediate_scalar(struct rc_constant_list * c, float da if (c->Constants[index].Type == RC_CONSTANT_IMMEDIATE) { for(unsigned comp = 0; comp < c->Constants[index].Size; ++comp) { if (c->Constants[index].u.Immediate[comp] == data) { - *swizzle = MAKE_SWIZZLE4(comp, comp, comp, comp); + *swizzle = RC_MAKE_SWIZZLE(comp, comp, comp, comp); return index; } } @@ -156,7 +158,7 @@ unsigned rc_constants_add_immediate_scalar(struct rc_constant_list * c, float da if (free_index >= 0) { unsigned comp = c->Constants[free_index].Size++; c->Constants[free_index].u.Immediate[comp] = data; - *swizzle = MAKE_SWIZZLE4(comp, comp, comp, comp); + *swizzle = RC_MAKE_SWIZZLE(comp, comp, comp, comp); return free_index; } @@ -164,7 +166,7 @@ unsigned rc_constants_add_immediate_scalar(struct rc_constant_list * c, float da constant.Type = RC_CONSTANT_IMMEDIATE; constant.Size = 1; constant.u.Immediate[0] = data; - *swizzle = SWIZZLE_XXXX; + *swizzle = RC_SWIZZLE_XXXX; return rc_constants_add(c, &constant); } diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_code.h b/src/mesa/drivers/dri/r300/compiler/radeon_code.h index 3e88554ba19..902b7cfa53b 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_code.h +++ b/src/mesa/drivers/dri/r300/compiler/radeon_code.h @@ -89,6 +89,23 @@ unsigned rc_constants_add_immediate_vec4(struct rc_constant_list * c, const floa unsigned rc_constants_add_immediate_scalar(struct rc_constant_list * c, float data, unsigned * swizzle); /** + * Compare functions. + * + * \note By design, RC_COMPARE_FUNC_xxx + GL_NEVER gives you + * the correct GL compare function. + */ +typedef enum { + RC_COMPARE_FUNC_NEVER = 0, + RC_COMPARE_FUNC_LESS, + RC_COMPARE_FUNC_EQUAL, + RC_COMPARE_FUNC_LEQUAL, + RC_COMPARE_FUNC_GREATER, + RC_COMPARE_FUNC_NOTEQUAL, + RC_COMPARE_FUNC_GEQUAL, + RC_COMPARE_FUNC_ALWAYS +} rc_compare_func; + +/** * Stores state that influences the compilation of a fragment program. */ struct r300_fragment_program_external_state { @@ -105,10 +122,12 @@ struct r300_fragment_program_external_state { /** * If the sampler is used as a shadow sampler, - * this field is (texture_compare_func - GL_NEVER). - * [e.g. if compare function is GL_LEQUAL, this field is 3] + * this field specifies the compare function. + * + * Otherwise, this field is \ref RC_COMPARE_FUNC_NEVER (aka 0). * * Otherwise, this field is 0. + * \sa rc_compare_func */ unsigned texture_compare_func : 3; } unit[16]; @@ -163,6 +182,8 @@ struct r500_fragment_program_code { int inst_end; /* Number of instructions - 1; also, last instruction to be executed */ int max_temp_idx; + + uint32_t us_fc_ctrl; }; struct rX00_fragment_program_code { diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_compiler.c b/src/mesa/drivers/dri/r300/compiler/radeon_compiler.c index da950d52894..c0e7a7f7a02 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_compiler.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_compiler.c @@ -23,6 +23,8 @@ #include "radeon_compiler.h" #include <stdarg.h> +#include <stdio.h> +#include <stdlib.h> #include "radeon_program.h" @@ -34,7 +36,7 @@ void rc_init(struct radeon_compiler * c) memory_pool_init(&c->Pool); c->Program.Instructions.Prev = &c->Program.Instructions; c->Program.Instructions.Next = &c->Program.Instructions; - c->Program.Instructions.I.Opcode = OPCODE_END; + c->Program.Instructions.U.I.Opcode = RC_OPCODE_ILLEGAL_OPCODE; } void rc_destroy(struct radeon_compiler * c) @@ -60,7 +62,7 @@ void rc_error(struct radeon_compiler * c, const char * fmt, ...) { va_list ap; - c->Error = GL_TRUE; + c->Error = 1; if (!c->ErrorMsg) { /* Only remember the first error */ @@ -91,28 +93,63 @@ void rc_error(struct radeon_compiler * c, const char * fmt, ...) } } +int rc_if_fail_helper(struct radeon_compiler * c, const char * file, int line, const char * assertion) +{ + rc_error(c, "ICE at %s:%i: assertion failed: %s\n", file, line, assertion); + return 1; +} + +/** + * Recompute c->Program.InputsRead and c->Program.OutputsWritten + * based on which inputs and outputs are actually referenced + * in program instructions. + */ +void rc_calculate_inputs_outputs(struct radeon_compiler * c) +{ + struct rc_instruction *inst; + + c->Program.InputsRead = 0; + c->Program.OutputsWritten = 0; + + for(inst = c->Program.Instructions.Next; inst != &c->Program.Instructions; inst = inst->Next) + { + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->U.I.Opcode); + int i; + + for (i = 0; i < opcode->NumSrcRegs; ++i) { + if (inst->U.I.SrcReg[i].File == RC_FILE_INPUT) + c->Program.InputsRead |= 1 << inst->U.I.SrcReg[i].Index; + } + + if (opcode->HasDstReg) { + if (inst->U.I.DstReg.File == RC_FILE_OUTPUT) + c->Program.OutputsWritten |= 1 << inst->U.I.DstReg.Index; + } + } +} + /** * Rewrite the program such that everything that source the given input * register will source new_input instead. */ -void rc_move_input(struct radeon_compiler * c, unsigned input, struct prog_src_register new_input) +void rc_move_input(struct radeon_compiler * c, unsigned input, struct rc_src_register new_input) { struct rc_instruction * inst; c->Program.InputsRead &= ~(1 << input); for(inst = c->Program.Instructions.Next; inst != &c->Program.Instructions; inst = inst->Next) { - const unsigned numsrcs = _mesa_num_inst_src_regs(inst->I.Opcode); + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->U.I.Opcode); unsigned i; - for(i = 0; i < numsrcs; ++i) { - if (inst->I.SrcReg[i].File == PROGRAM_INPUT && inst->I.SrcReg[i].Index == input) { - inst->I.SrcReg[i].File = new_input.File; - inst->I.SrcReg[i].Index = new_input.Index; - inst->I.SrcReg[i].Swizzle = combine_swizzles(new_input.Swizzle, inst->I.SrcReg[i].Swizzle); - if (!inst->I.SrcReg[i].Abs) { - inst->I.SrcReg[i].Negate ^= new_input.Negate; - inst->I.SrcReg[i].Abs = new_input.Abs; + for(i = 0; i < opcode->NumSrcRegs; ++i) { + if (inst->U.I.SrcReg[i].File == RC_FILE_INPUT && inst->U.I.SrcReg[i].Index == input) { + inst->U.I.SrcReg[i].File = new_input.File; + inst->U.I.SrcReg[i].Index = new_input.Index; + inst->U.I.SrcReg[i].Swizzle = combine_swizzles(new_input.Swizzle, inst->U.I.SrcReg[i].Swizzle); + if (!inst->U.I.SrcReg[i].Abs) { + inst->U.I.SrcReg[i].Negate ^= new_input.Negate; + inst->U.I.SrcReg[i].Abs = new_input.Abs; } c->Program.InputsRead |= 1 << new_input.Index; @@ -134,12 +171,12 @@ void rc_move_output(struct radeon_compiler * c, unsigned output, unsigned new_ou c->Program.OutputsWritten &= ~(1 << output); for(inst = c->Program.Instructions.Next; inst != &c->Program.Instructions; inst = inst->Next) { - const unsigned numdsts = _mesa_num_inst_dst_regs(inst->I.Opcode); + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->U.I.Opcode); - if (numdsts) { - if (inst->I.DstReg.File == PROGRAM_OUTPUT && inst->I.DstReg.Index == output) { - inst->I.DstReg.Index = new_output; - inst->I.DstReg.WriteMask &= writemask; + if (opcode->HasDstReg) { + if (inst->U.I.DstReg.File == RC_FILE_OUTPUT && inst->U.I.DstReg.Index == output) { + inst->U.I.DstReg.Index = new_output; + inst->U.I.DstReg.WriteMask &= writemask; c->Program.OutputsWritten |= 1 << new_output; } @@ -157,33 +194,33 @@ void rc_copy_output(struct radeon_compiler * c, unsigned output, unsigned dup_ou struct rc_instruction * inst; for(inst = c->Program.Instructions.Next; inst != &c->Program.Instructions; inst = inst->Next) { - const unsigned numdsts = _mesa_num_inst_dst_regs(inst->I.Opcode); + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->U.I.Opcode); - if (numdsts) { - if (inst->I.DstReg.File == PROGRAM_OUTPUT && inst->I.DstReg.Index == output) { - inst->I.DstReg.File = PROGRAM_TEMPORARY; - inst->I.DstReg.Index = tempreg; + if (opcode->HasDstReg) { + if (inst->U.I.DstReg.File == RC_FILE_OUTPUT && inst->U.I.DstReg.Index == output) { + inst->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst->U.I.DstReg.Index = tempreg; } } } inst = rc_insert_new_instruction(c, c->Program.Instructions.Prev); - inst->I.Opcode = OPCODE_MOV; - inst->I.DstReg.File = PROGRAM_OUTPUT; - inst->I.DstReg.Index = output; + inst->U.I.Opcode = RC_OPCODE_MOV; + inst->U.I.DstReg.File = RC_FILE_OUTPUT; + inst->U.I.DstReg.Index = output; - inst->I.SrcReg[0].File = PROGRAM_TEMPORARY; - inst->I.SrcReg[0].Index = tempreg; - inst->I.SrcReg[0].Swizzle = SWIZZLE_XYZW; + inst->U.I.SrcReg[0].File = RC_FILE_TEMPORARY; + inst->U.I.SrcReg[0].Index = tempreg; + inst->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_XYZW; inst = rc_insert_new_instruction(c, c->Program.Instructions.Prev); - inst->I.Opcode = OPCODE_MOV; - inst->I.DstReg.File = PROGRAM_OUTPUT; - inst->I.DstReg.Index = dup_output; + inst->U.I.Opcode = RC_OPCODE_MOV; + inst->U.I.DstReg.File = RC_FILE_OUTPUT; + inst->U.I.DstReg.Index = dup_output; - inst->I.SrcReg[0].File = PROGRAM_TEMPORARY; - inst->I.SrcReg[0].Index = tempreg; - inst->I.SrcReg[0].Swizzle = SWIZZLE_XYZW; + inst->U.I.SrcReg[0].File = RC_FILE_TEMPORARY; + inst->U.I.SrcReg[0].Index = tempreg; + inst->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_XYZW; c->Program.OutputsWritten |= 1 << dup_output; } @@ -201,60 +238,60 @@ void rc_transform_fragment_wpos(struct radeon_compiler * c, unsigned wpos, unsig /* perspective divide */ struct rc_instruction * inst_rcp = rc_insert_new_instruction(c, &c->Program.Instructions); - inst_rcp->I.Opcode = OPCODE_RCP; + inst_rcp->U.I.Opcode = RC_OPCODE_RCP; - inst_rcp->I.DstReg.File = PROGRAM_TEMPORARY; - inst_rcp->I.DstReg.Index = tempregi; - inst_rcp->I.DstReg.WriteMask = WRITEMASK_W; + inst_rcp->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst_rcp->U.I.DstReg.Index = tempregi; + inst_rcp->U.I.DstReg.WriteMask = RC_MASK_W; - inst_rcp->I.SrcReg[0].File = PROGRAM_INPUT; - inst_rcp->I.SrcReg[0].Index = new_input; - inst_rcp->I.SrcReg[0].Swizzle = SWIZZLE_WWWW; + inst_rcp->U.I.SrcReg[0].File = RC_FILE_INPUT; + inst_rcp->U.I.SrcReg[0].Index = new_input; + inst_rcp->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_WWWW; struct rc_instruction * inst_mul = rc_insert_new_instruction(c, inst_rcp); - inst_mul->I.Opcode = OPCODE_MUL; + inst_mul->U.I.Opcode = RC_OPCODE_MUL; - inst_mul->I.DstReg.File = PROGRAM_TEMPORARY; - inst_mul->I.DstReg.Index = tempregi; - inst_mul->I.DstReg.WriteMask = WRITEMASK_XYZ; + inst_mul->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst_mul->U.I.DstReg.Index = tempregi; + inst_mul->U.I.DstReg.WriteMask = RC_MASK_XYZ; - inst_mul->I.SrcReg[0].File = PROGRAM_INPUT; - inst_mul->I.SrcReg[0].Index = new_input; + inst_mul->U.I.SrcReg[0].File = RC_FILE_INPUT; + inst_mul->U.I.SrcReg[0].Index = new_input; - inst_mul->I.SrcReg[1].File = PROGRAM_TEMPORARY; - inst_mul->I.SrcReg[1].Index = tempregi; - inst_mul->I.SrcReg[1].Swizzle = SWIZZLE_WWWW; + inst_mul->U.I.SrcReg[1].File = RC_FILE_TEMPORARY; + inst_mul->U.I.SrcReg[1].Index = tempregi; + inst_mul->U.I.SrcReg[1].Swizzle = RC_SWIZZLE_WWWW; /* viewport transformation */ struct rc_instruction * inst_mad = rc_insert_new_instruction(c, inst_mul); - inst_mad->I.Opcode = OPCODE_MAD; + inst_mad->U.I.Opcode = RC_OPCODE_MAD; - inst_mad->I.DstReg.File = PROGRAM_TEMPORARY; - inst_mad->I.DstReg.Index = tempregi; - inst_mad->I.DstReg.WriteMask = WRITEMASK_XYZ; + inst_mad->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst_mad->U.I.DstReg.Index = tempregi; + inst_mad->U.I.DstReg.WriteMask = RC_MASK_XYZ; - inst_mad->I.SrcReg[0].File = PROGRAM_TEMPORARY; - inst_mad->I.SrcReg[0].Index = tempregi; - inst_mad->I.SrcReg[0].Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO); + inst_mad->U.I.SrcReg[0].File = RC_FILE_TEMPORARY; + inst_mad->U.I.SrcReg[0].Index = tempregi; + inst_mad->U.I.SrcReg[0].Swizzle = RC_MAKE_SWIZZLE(RC_SWIZZLE_X, RC_SWIZZLE_Y, RC_SWIZZLE_Z, RC_SWIZZLE_ZERO); - inst_mad->I.SrcReg[1].File = PROGRAM_STATE_VAR; - inst_mad->I.SrcReg[1].Index = rc_constants_add_state(&c->Program.Constants, RC_STATE_R300_WINDOW_DIMENSION, 0); - inst_mad->I.SrcReg[1].Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO); + inst_mad->U.I.SrcReg[1].File = RC_FILE_CONSTANT; + inst_mad->U.I.SrcReg[1].Index = rc_constants_add_state(&c->Program.Constants, RC_STATE_R300_WINDOW_DIMENSION, 0); + inst_mad->U.I.SrcReg[1].Swizzle = RC_MAKE_SWIZZLE(RC_SWIZZLE_X, RC_SWIZZLE_Y, RC_SWIZZLE_Z, RC_SWIZZLE_ZERO); - inst_mad->I.SrcReg[2].File = PROGRAM_STATE_VAR; - inst_mad->I.SrcReg[2].Index = inst_mad->I.SrcReg[1].Index; - inst_mad->I.SrcReg[2].Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO); + inst_mad->U.I.SrcReg[2].File = RC_FILE_CONSTANT; + inst_mad->U.I.SrcReg[2].Index = inst_mad->U.I.SrcReg[1].Index; + inst_mad->U.I.SrcReg[2].Swizzle = RC_MAKE_SWIZZLE(RC_SWIZZLE_X, RC_SWIZZLE_Y, RC_SWIZZLE_Z, RC_SWIZZLE_ZERO); struct rc_instruction * inst; for (inst = inst_mad->Next; inst != &c->Program.Instructions; inst = inst->Next) { - const unsigned numsrcs = _mesa_num_inst_src_regs(inst->I.Opcode); + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->U.I.Opcode); unsigned i; - for(i = 0; i < numsrcs; i++) { - if (inst->I.SrcReg[i].File == PROGRAM_INPUT && - inst->I.SrcReg[i].Index == wpos) { - inst->I.SrcReg[i].File = PROGRAM_TEMPORARY; - inst->I.SrcReg[i].Index = tempregi; + for(i = 0; i < opcode->NumSrcRegs; i++) { + if (inst->U.I.SrcReg[i].File == RC_FILE_INPUT && + inst->U.I.SrcReg[i].Index == wpos) { + inst->U.I.SrcReg[i].File = RC_FILE_TEMPORARY; + inst->U.I.SrcReg[i].Index = tempregi; } } } diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_compiler.h b/src/mesa/drivers/dri/r300/compiler/radeon_compiler.h index e63ab8840ab..87a732cd90d 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_compiler.h +++ b/src/mesa/drivers/dri/r300/compiler/radeon_compiler.h @@ -23,35 +23,11 @@ #ifndef RADEON_COMPILER_H #define RADEON_COMPILER_H -#include "main/mtypes.h" -#include "shader/prog_instruction.h" - #include "memory_pool.h" #include "radeon_code.h" +#include "radeon_program.h" - -struct rc_instruction { - struct rc_instruction * Prev; - struct rc_instruction * Next; - struct prog_instruction I; -}; - -struct rc_program { - /** - * Instructions.Next points to the first instruction, - * Instructions.Prev points to the last instruction. - */ - struct rc_instruction Instructions; - - /* Long term, we should probably remove InputsRead & OutputsWritten, - * since updating dependent state can be fragile, and they aren't - * actually used very often. */ - uint32_t InputsRead; - uint32_t OutputsWritten; - uint32_t ShadowSamplers; /**< Texture units used for shadow sampling. */ - - struct rc_constant_list Constants; -}; +struct rc_swizzle_caps; struct radeon_compiler { struct memory_pool Pool; @@ -59,6 +35,14 @@ struct radeon_compiler { unsigned Debug:1; unsigned Error:1; char * ErrorMsg; + + /** + * Variables used internally, not be touched by callers + * of the compiler + */ + /*@{*/ + struct rc_swizzle_caps * SwizzleCaps; + /*@}*/ }; void rc_init(struct radeon_compiler * c); @@ -67,11 +51,26 @@ void rc_destroy(struct radeon_compiler * c); void rc_debug(struct radeon_compiler * c, const char * fmt, ...); void rc_error(struct radeon_compiler * c, const char * fmt, ...); -void rc_mesa_to_rc_program(struct radeon_compiler * c, struct gl_program * program); +int rc_if_fail_helper(struct radeon_compiler * c, const char * file, int line, const char * assertion); + +/** + * This macro acts like an if-statement that can be used to implement + * non-aborting assertions in the compiler. + * + * It checks whether \p cond is true. If not, an internal compiler error is + * flagged and the if-clause is run. + * + * A typical use-case would be: + * + * if (rc_assert(c, condition-that-must-be-true)) + * return; + */ +#define rc_assert(c, cond) \ + (!(cond) && rc_if_fail_helper(c, __FILE__, __LINE__, #cond)) void rc_calculate_inputs_outputs(struct radeon_compiler * c); -void rc_move_input(struct radeon_compiler * c, unsigned input, struct prog_src_register new_input); +void rc_move_input(struct radeon_compiler * c, unsigned input, struct rc_src_register new_input); void rc_move_output(struct radeon_compiler * c, unsigned output, unsigned new_output, unsigned writemask); void rc_copy_output(struct radeon_compiler * c, unsigned output, unsigned dup_output); void rc_transform_fragment_wpos(struct radeon_compiler * c, unsigned wpos, unsigned new_input); @@ -97,7 +96,7 @@ void r3xx_compile_fragment_program(struct r300_fragment_program_compiler* c); struct r300_vertex_program_compiler { struct radeon_compiler Base; struct r300_vertex_program_code *code; - GLbitfield RequiredOutputs; + uint32_t RequiredOutputs; void * UserData; void (*SetHwInputOutput)(struct r300_vertex_program_compiler * c); diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_dataflow.c b/src/mesa/drivers/dri/r300/compiler/radeon_dataflow.c new file mode 100644 index 00000000000..cce9166e644 --- /dev/null +++ b/src/mesa/drivers/dri/r300/compiler/radeon_dataflow.c @@ -0,0 +1,162 @@ +/* + * Copyright (C) 2009 Nicolai Haehnle. + * + * 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, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "radeon_dataflow.h" + +#include "radeon_program.h" + + +static void reads_normal(struct rc_instruction * fullinst, rc_read_write_fn cb, void * userdata) +{ + struct rc_sub_instruction * inst = &fullinst->U.I; + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->Opcode); + + for(unsigned int src = 0; src < opcode->NumSrcRegs; ++src) { + unsigned int refmask = 0; + + if (inst->SrcReg[src].File == RC_FILE_NONE) + return; + + for(unsigned int chan = 0; chan < 4; ++chan) + refmask |= 1 << GET_SWZ(inst->SrcReg[src].Swizzle, chan); + + refmask &= RC_MASK_XYZW; + + for(unsigned int chan = 0; chan < 4; ++chan) { + if (GET_BIT(refmask, chan)) { + cb(userdata, fullinst, inst->SrcReg[src].File, inst->SrcReg[src].Index, chan); + } + } + + if (refmask && inst->SrcReg[src].RelAddr) + cb(userdata, fullinst, RC_FILE_ADDRESS, 0, RC_MASK_X); + } +} + +static void reads_pair(struct rc_instruction * fullinst, rc_read_write_fn cb, void * userdata) +{ + struct rc_pair_instruction * inst = &fullinst->U.P; + unsigned int refmasks[3] = { 0, 0, 0 }; + + if (inst->RGB.Opcode != RC_OPCODE_NOP) { + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->RGB.Opcode); + + for(unsigned int arg = 0; arg < opcode->NumSrcRegs; ++arg) { + for(unsigned int chan = 0; chan < 3; ++chan) { + unsigned int swz = GET_SWZ(inst->RGB.Arg[arg].Swizzle, chan); + if (swz < 4) + refmasks[inst->RGB.Arg[arg].Source] |= 1 << swz; + } + } + } + + if (inst->Alpha.Opcode != RC_OPCODE_NOP) { + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->Alpha.Opcode); + + for(unsigned int arg = 0; arg < opcode->NumSrcRegs; ++arg) { + if (inst->Alpha.Arg[arg].Swizzle < 4) + refmasks[inst->Alpha.Arg[arg].Source] |= 1 << inst->Alpha.Arg[arg].Swizzle; + } + } + + for(unsigned int src = 0; src < 3; ++src) { + if (inst->RGB.Src[src].Used) { + for(unsigned int chan = 0; chan < 3; ++chan) { + if (GET_BIT(refmasks[src], chan)) + cb(userdata, fullinst, inst->RGB.Src[src].File, inst->RGB.Src[src].Index, chan); + } + } + + if (inst->Alpha.Src[src].Used) { + if (GET_BIT(refmasks[src], 3)) + cb(userdata, fullinst, inst->Alpha.Src[src].File, inst->Alpha.Src[src].Index, 3); + } + } +} + +/** + * Calls a callback function for all sourced register channels. + * + * This is conservative, i.e. channels may be called multiple times, + * and the writemask of the instruction is not taken into account. + */ +void rc_for_all_reads(struct rc_instruction * inst, rc_read_write_fn cb, void * userdata) +{ + if (inst->Type == RC_INSTRUCTION_NORMAL) { + reads_normal(inst, cb, userdata); + } else { + reads_pair(inst, cb, userdata); + } +} + + + +static void writes_normal(struct rc_instruction * fullinst, rc_read_write_fn cb, void * userdata) +{ + struct rc_sub_instruction * inst = &fullinst->U.I; + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->Opcode); + + if (opcode->HasDstReg) { + for(unsigned int chan = 0; chan < 4; ++chan) { + if (GET_BIT(inst->DstReg.WriteMask, chan)) + cb(userdata, fullinst, inst->DstReg.File, inst->DstReg.Index, chan); + } + } + + if (inst->WriteALUResult) + cb(userdata, fullinst, RC_FILE_SPECIAL, RC_SPECIAL_ALU_RESULT, 0); +} + +static void writes_pair(struct rc_instruction * fullinst, rc_read_write_fn cb, void * userdata) +{ + struct rc_pair_instruction * inst = &fullinst->U.P; + + for(unsigned int chan = 0; chan < 3; ++chan) { + if (GET_BIT(inst->RGB.WriteMask, chan)) + cb(userdata, fullinst, RC_FILE_TEMPORARY, inst->RGB.DestIndex, chan); + } + + if (inst->Alpha.WriteMask) + cb(userdata, fullinst, RC_FILE_TEMPORARY, inst->Alpha.DestIndex, 3); + + if (inst->WriteALUResult) + cb(userdata, fullinst, RC_FILE_SPECIAL, RC_SPECIAL_ALU_RESULT, 0); +} + +/** + * Calls a callback function for all written register channels. + * + * \warning Does not report output registers for paired instructions! + */ +void rc_for_all_writes(struct rc_instruction * inst, rc_read_write_fn cb, void * userdata) +{ + if (inst->Type == RC_INSTRUCTION_NORMAL) { + writes_normal(inst, cb, userdata); + } else { + writes_pair(inst, cb, userdata); + } +} diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_dataflow.h b/src/mesa/drivers/dri/r300/compiler/radeon_dataflow.h new file mode 100644 index 00000000000..5aa4cb64f3d --- /dev/null +++ b/src/mesa/drivers/dri/r300/compiler/radeon_dataflow.h @@ -0,0 +1,59 @@ +/* + * Copyright (C) 2009 Nicolai Haehnle. + * + * 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, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef RADEON_DATAFLOW_H +#define RADEON_DATAFLOW_H + +#include "radeon_program_constants.h" + +struct radeon_compiler; +struct rc_instruction; +struct rc_swizzle_caps; + + +/** + * Help analyze the register accesses of instructions. + */ +/*@{*/ +typedef void (*rc_read_write_fn)(void * userdata, struct rc_instruction * inst, + rc_register_file file, unsigned int index, unsigned int chan); +void rc_for_all_reads(struct rc_instruction * inst, rc_read_write_fn cb, void * userdata); +void rc_for_all_writes(struct rc_instruction * inst, rc_read_write_fn cb, void * userdata); +/*@}*/ + + +/** + * Compiler passes based on dataflow analysis. + */ +/*@{*/ +typedef void (*rc_dataflow_mark_outputs_fn)(void * userdata, void * data, + void (*mark_fn)(void * data, unsigned int index, unsigned int mask)); +void rc_dataflow_deadcode(struct radeon_compiler * c, rc_dataflow_mark_outputs_fn dce, void * userdata); +void rc_dataflow_swizzles(struct radeon_compiler * c); +/*@}*/ + +#endif /* RADEON_DATAFLOW_H */ diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_dataflow_deadcode.c b/src/mesa/drivers/dri/r300/compiler/radeon_dataflow_deadcode.c new file mode 100644 index 00000000000..e0c66c4aeb0 --- /dev/null +++ b/src/mesa/drivers/dri/r300/compiler/radeon_dataflow_deadcode.c @@ -0,0 +1,295 @@ +/* + * Copyright (C) 2009 Nicolai Haehnle. + * + * 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, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "radeon_dataflow.h" + +#include "radeon_compiler.h" + + +struct updatemask_state { + unsigned char Output[RC_REGISTER_MAX_INDEX]; + unsigned char Temporary[RC_REGISTER_MAX_INDEX]; + unsigned char Address; + unsigned char Special[RC_NUM_SPECIAL_REGISTERS]; +}; + +struct instruction_state { + unsigned char WriteMask:4; + unsigned char WriteALUResult:1; + unsigned char SrcReg[3]; +}; + +struct branchinfo { + unsigned int HaveElse:1; + + struct updatemask_state StoreEndif; + struct updatemask_state StoreElse; +}; + +struct deadcode_state { + struct radeon_compiler * C; + struct instruction_state * Instructions; + + struct updatemask_state R; + + struct branchinfo * BranchStack; + unsigned int BranchStackSize; + unsigned int BranchStackReserved; +}; + + +static void or_updatemasks( + struct updatemask_state * dst, + struct updatemask_state * a, + struct updatemask_state * b) +{ + for(unsigned int i = 0; i < RC_REGISTER_MAX_INDEX; ++i) { + dst->Output[i] = a->Output[i] | b->Output[i]; + dst->Temporary[i] = a->Temporary[i] | b->Temporary[i]; + } + + for(unsigned int i = 0; i < RC_NUM_SPECIAL_REGISTERS; ++i) + dst->Special[i] = a->Special[i] | b->Special[i]; + + dst->Address = a->Address | b->Address; +} + +static void push_branch(struct deadcode_state * s) +{ + if (s->BranchStackSize >= s->BranchStackReserved) { + unsigned int new_reserve = 2 * s->BranchStackReserved; + struct branchinfo * new_stack; + + if (!new_reserve) + new_reserve = 4; + + new_stack = memory_pool_malloc(&s->C->Pool, new_reserve * sizeof(struct branchinfo)); + memcpy(new_stack, s->BranchStack, s->BranchStackSize * sizeof(struct branchinfo)); + + s->BranchStack = new_stack; + s->BranchStackReserved = new_reserve; + } + + struct branchinfo * branch = &s->BranchStack[s->BranchStackSize++]; + branch->HaveElse = 0; + memcpy(&branch->StoreEndif, &s->R, sizeof(s->R)); +} + +static unsigned char * get_used_ptr(struct deadcode_state *s, rc_register_file file, unsigned int index) +{ + if (file == RC_FILE_OUTPUT || file == RC_FILE_TEMPORARY) { + if (index >= RC_REGISTER_MAX_INDEX) { + rc_error(s->C, "%s: index %i is out of bounds for file %i\n", __FUNCTION__, index, file); + return 0; + } + + if (file == RC_FILE_OUTPUT) + return &s->R.Output[index]; + else + return &s->R.Temporary[index]; + } else if (file == RC_FILE_ADDRESS) { + return &s->R.Address; + } else if (file == RC_FILE_SPECIAL) { + if (index >= RC_NUM_SPECIAL_REGISTERS) { + rc_error(s->C, "%s: special file index %i out of bounds\n", __FUNCTION__, index); + return 0; + } + + return &s->R.Special[index]; + } + + return 0; +} + +static void mark_used(struct deadcode_state * s, rc_register_file file, unsigned int index, unsigned int mask) +{ + unsigned char * pused = get_used_ptr(s, file, index); + if (pused) + *pused |= mask; +} + +static void update_instruction(struct deadcode_state * s, struct rc_instruction * inst) +{ + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->U.I.Opcode); + struct instruction_state * insts = &s->Instructions[inst->IP]; + unsigned int usedmask = 0; + + if (opcode->HasDstReg) { + unsigned char * pused = get_used_ptr(s, inst->U.I.DstReg.File, inst->U.I.DstReg.Index); + if (pused) { + usedmask = *pused & inst->U.I.DstReg.WriteMask; + *pused &= ~usedmask; + } + } + + insts->WriteMask |= usedmask; + + if (inst->U.I.WriteALUResult) { + unsigned char * pused = get_used_ptr(s, RC_FILE_SPECIAL, RC_SPECIAL_ALU_RESULT); + if (pused && *pused) { + if (inst->U.I.WriteALUResult == RC_ALURESULT_X) + usedmask |= RC_MASK_X; + else if (inst->U.I.WriteALUResult == RC_ALURESULT_W) + usedmask |= RC_MASK_W; + + *pused = 0; + insts->WriteALUResult = 1; + } + } + + unsigned int srcmasks[3]; + rc_compute_sources_for_writemask(opcode, usedmask, srcmasks); + + for(unsigned int src = 0; src < opcode->NumSrcRegs; ++src) { + unsigned int refmask = 0; + unsigned int newsrcmask = srcmasks[src] & ~insts->SrcReg[src]; + insts->SrcReg[src] |= newsrcmask; + + for(unsigned int chan = 0; chan < 4; ++chan) { + if (GET_BIT(newsrcmask, chan)) + refmask |= 1 << GET_SWZ(inst->U.I.SrcReg[src].Swizzle, chan); + } + + /* get rid of spurious bits from ZERO, ONE, etc. swizzles */ + refmask &= RC_MASK_XYZW; + + if (!refmask) + continue; + + mark_used(s, inst->U.I.SrcReg[src].File, inst->U.I.SrcReg[src].Index, refmask); + + if (inst->U.I.SrcReg[src].RelAddr) + mark_used(s, RC_FILE_ADDRESS, 0, RC_MASK_X); + } +} + +static void mark_output_use(void * data, unsigned int index, unsigned int mask) +{ + struct deadcode_state * s = data; + + mark_used(s, RC_FILE_OUTPUT, index, mask); +} + +void rc_dataflow_deadcode(struct radeon_compiler * c, rc_dataflow_mark_outputs_fn dce, void * userdata) +{ + struct deadcode_state s; + unsigned int nr_instructions; + + memset(&s, 0, sizeof(s)); + s.C = c; + + nr_instructions = rc_recompute_ips(c); + s.Instructions = memory_pool_malloc(&c->Pool, sizeof(struct instruction_state)*nr_instructions); + memset(s.Instructions, 0, sizeof(struct instruction_state)*nr_instructions); + + dce(userdata, &s, &mark_output_use); + + for(struct rc_instruction * inst = c->Program.Instructions.Prev; + inst != &c->Program.Instructions; + inst = inst->Prev) { + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->U.I.Opcode); + + if (opcode->IsFlowControl) { + if (opcode->Opcode == RC_OPCODE_ENDIF) { + push_branch(&s); + } else { + if (s.BranchStackSize) { + struct branchinfo * branch = &s.BranchStack[s.BranchStackSize-1]; + + if (opcode->Opcode == RC_OPCODE_IF) { + or_updatemasks(&s.R, + &s.R, + branch->HaveElse ? &branch->StoreElse : &branch->StoreEndif); + + s.BranchStackSize--; + } else if (opcode->Opcode == RC_OPCODE_ELSE) { + if (branch->HaveElse) { + rc_error(c, "%s: Multiple ELSE for one IF/ENDIF\n", __FUNCTION__); + } else { + memcpy(&branch->StoreElse, &s.R, sizeof(s.R)); + memcpy(&s.R, &branch->StoreEndif, sizeof(s.R)); + branch->HaveElse = 1; + } + } else { + rc_error(c, "%s: Unhandled control flow instruction %s\n", __FUNCTION__, opcode->Name); + } + } else { + rc_error(c, "%s: Unexpected control flow instruction\n", __FUNCTION__); + } + } + } + + update_instruction(&s, inst); + } + + unsigned int ip = 0; + for(struct rc_instruction * inst = c->Program.Instructions.Next; + inst != &c->Program.Instructions; + inst = inst->Next, ++ip) { + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->U.I.Opcode);\ + int dead = 1; + + if (!opcode->HasDstReg) { + dead = 0; + } else { + inst->U.I.DstReg.WriteMask = s.Instructions[ip].WriteMask; + if (s.Instructions[ip].WriteMask) + dead = 0; + + if (s.Instructions[ip].WriteALUResult) + dead = 0; + else + inst->U.I.WriteALUResult = RC_ALURESULT_NONE; + } + + if (dead) { + struct rc_instruction * todelete = inst; + inst = inst->Prev; + rc_remove_instruction(todelete); + continue; + } + + unsigned int srcmasks[3]; + unsigned int usemask = s.Instructions[ip].WriteMask; + + if (inst->U.I.WriteALUResult == RC_ALURESULT_X) + usemask |= RC_MASK_X; + else if (inst->U.I.WriteALUResult == RC_ALURESULT_W) + usemask |= RC_MASK_W; + + rc_compute_sources_for_writemask(opcode, usemask, srcmasks); + + for(unsigned int src = 0; src < 3; ++src) { + for(unsigned int chan = 0; chan < 4; ++chan) { + if (!GET_BIT(srcmasks[src], chan)) + SET_SWZ(inst->U.I.SrcReg[src].Swizzle, chan, RC_SWIZZLE_UNUSED); + } + } + } + + rc_calculate_inputs_outputs(c); +} diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_dataflow_swizzles.c b/src/mesa/drivers/dri/r300/compiler/radeon_dataflow_swizzles.c new file mode 100644 index 00000000000..33acbd30f44 --- /dev/null +++ b/src/mesa/drivers/dri/r300/compiler/radeon_dataflow_swizzles.c @@ -0,0 +1,102 @@ +/* + * Copyright (C) 2009 Nicolai Haehnle. + * + * 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, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "radeon_dataflow.h" + +#include "radeon_compiler.h" +#include "radeon_swizzle.h" + + +static void rewrite_source(struct radeon_compiler * c, + struct rc_instruction * inst, unsigned src) +{ + struct rc_swizzle_split split; + unsigned int tempreg = rc_find_free_temporary(c); + unsigned int usemask; + + usemask = 0; + for(unsigned int chan = 0; chan < 4; ++chan) { + if (GET_SWZ(inst->U.I.SrcReg[src].Swizzle, chan) != RC_SWIZZLE_UNUSED) + usemask |= 1 << chan; + } + + c->SwizzleCaps->Split(inst->U.I.SrcReg[src], usemask, &split); + + for(unsigned int phase = 0; phase < split.NumPhases; ++phase) { + struct rc_instruction * mov = rc_insert_new_instruction(c, inst->Prev); + unsigned int phase_refmask; + unsigned int masked_negate; + + mov->U.I.Opcode = RC_OPCODE_MOV; + mov->U.I.DstReg.File = RC_FILE_TEMPORARY; + mov->U.I.DstReg.Index = tempreg; + mov->U.I.DstReg.WriteMask = split.Phase[phase]; + mov->U.I.SrcReg[0] = inst->U.I.SrcReg[src]; + + phase_refmask = 0; + for(unsigned int chan = 0; chan < 4; ++chan) { + if (!GET_BIT(split.Phase[phase], chan)) + SET_SWZ(mov->U.I.SrcReg[0].Swizzle, chan, RC_SWIZZLE_UNUSED); + else + phase_refmask |= 1 << GET_SWZ(mov->U.I.SrcReg[0].Swizzle, chan); + } + + phase_refmask &= RC_MASK_XYZW; + + masked_negate = split.Phase[phase] & mov->U.I.SrcReg[0].Negate; + if (masked_negate == 0) + mov->U.I.SrcReg[0].Negate = 0; + else if (masked_negate == split.Phase[phase]) + mov->U.I.SrcReg[0].Negate = RC_MASK_XYZW; + + } + + inst->U.I.SrcReg[src].File = RC_FILE_TEMPORARY; + inst->U.I.SrcReg[src].Index = tempreg; + inst->U.I.SrcReg[src].Swizzle = 0; + inst->U.I.SrcReg[src].Negate = RC_MASK_NONE; + inst->U.I.SrcReg[src].Abs = 0; + for(unsigned int chan = 0; chan < 4; ++chan) { + SET_SWZ(inst->U.I.SrcReg[src].Swizzle, chan, + GET_BIT(usemask, chan) ? chan : RC_SWIZZLE_UNUSED); + } +} + +void rc_dataflow_swizzles(struct radeon_compiler * c) +{ + struct rc_instruction * inst; + + for(inst = c->Program.Instructions.Next; inst != &c->Program.Instructions; inst = inst->Next) { + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->U.I.Opcode); + unsigned int src; + + for(src = 0; src < opcode->NumSrcRegs; ++src) { + if (!c->SwizzleCaps->IsNative(inst->U.I.Opcode, inst->U.I.SrcReg[src])) + rewrite_source(c, inst, src); + } + } +} diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_nqssadce.c b/src/mesa/drivers/dri/r300/compiler/radeon_nqssadce.c deleted file mode 100644 index aaaa50ad1f9..00000000000 --- a/src/mesa/drivers/dri/r300/compiler/radeon_nqssadce.c +++ /dev/null @@ -1,294 +0,0 @@ -/* - * Copyright (C) 2008 Nicolai Haehnle. - * - * 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, sublicense, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial - * portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) 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. - * - */ - -/** - * @file - * - * "Not-quite SSA" and Dead-Code Elimination. - * - * @note This code uses SWIZZLE_NIL in a source register to indicate that - * the corresponding component is ignored by the corresponding instruction. - */ - -#include "radeon_nqssadce.h" - -#include "radeon_compiler.h" - - -/** - * Return the @ref register_state for the given register (or 0 for untracked - * registers, i.e. constants). - */ -static struct register_state *get_reg_state(struct nqssadce_state* s, GLuint file, GLuint index) -{ - switch(file) { - case PROGRAM_TEMPORARY: return &s->Temps[index]; - case PROGRAM_OUTPUT: return &s->Outputs[index]; - case PROGRAM_ADDRESS: return &s->Address; - default: return 0; - } -} - - -/** - * Left multiplication of a register with a swizzle - * - * @note Works correctly only for X, Y, Z, W swizzles, not for constant swizzles. - */ -struct prog_src_register lmul_swizzle(GLuint swizzle, struct prog_src_register srcreg) -{ - struct prog_src_register tmp = srcreg; - int i; - tmp.Swizzle = 0; - tmp.Negate = NEGATE_NONE; - for(i = 0; i < 4; ++i) { - GLuint swz = GET_SWZ(swizzle, i); - if (swz < 4) { - tmp.Swizzle |= GET_SWZ(srcreg.Swizzle, swz) << (i*3); - tmp.Negate |= GET_BIT(srcreg.Negate, swz) << i; - } else { - tmp.Swizzle |= swz << (i*3); - } - } - return tmp; -} - - -static void track_used_srcreg(struct nqssadce_state* s, - GLint src, GLuint sourced) -{ - struct prog_instruction * inst = &s->IP->I; - int i; - GLuint deswz_source = 0; - - for(i = 0; i < 4; ++i) { - if (GET_BIT(sourced, i)) { - GLuint swz = GET_SWZ(inst->SrcReg[src].Swizzle, i); - deswz_source |= 1 << swz; - } else { - inst->SrcReg[src].Swizzle &= ~(7 << (3*i)); - inst->SrcReg[src].Swizzle |= SWIZZLE_NIL << (3*i); - } - } - - if (!s->Descr->IsNativeSwizzle(inst->Opcode, inst->SrcReg[src])) { - struct prog_dst_register dstreg = inst->DstReg; - dstreg.File = PROGRAM_TEMPORARY; - dstreg.Index = rc_find_free_temporary(s->Compiler); - dstreg.WriteMask = sourced; - - s->Descr->BuildSwizzle(s, dstreg, inst->SrcReg[src]); - - inst->SrcReg[src].File = PROGRAM_TEMPORARY; - inst->SrcReg[src].Index = dstreg.Index; - inst->SrcReg[src].Swizzle = 0; - inst->SrcReg[src].Negate = NEGATE_NONE; - inst->SrcReg[src].Abs = 0; - for(i = 0; i < 4; ++i) { - if (GET_BIT(sourced, i)) - inst->SrcReg[src].Swizzle |= i << (3*i); - else - inst->SrcReg[src].Swizzle |= SWIZZLE_NIL << (3*i); - } - deswz_source = sourced; - } - - struct register_state *regstate; - - if (inst->SrcReg[src].RelAddr) { - regstate = get_reg_state(s, PROGRAM_ADDRESS, 0); - if (regstate) - regstate->Sourced |= WRITEMASK_X; - } else { - regstate = get_reg_state(s, inst->SrcReg[src].File, inst->SrcReg[src].Index); - if (regstate) - regstate->Sourced |= deswz_source & 0xf; - } -} - -static void unalias_srcregs(struct rc_instruction *inst, GLuint oldindex, GLuint newindex) -{ - int nsrc = _mesa_num_inst_src_regs(inst->I.Opcode); - int i; - for(i = 0; i < nsrc; ++i) - if (inst->I.SrcReg[i].File == PROGRAM_TEMPORARY && inst->I.SrcReg[i].Index == oldindex) - inst->I.SrcReg[i].Index = newindex; -} - -static void unalias_temporary(struct nqssadce_state* s, GLuint oldindex) -{ - GLuint newindex = rc_find_free_temporary(s->Compiler); - struct rc_instruction * inst; - for(inst = s->Compiler->Program.Instructions.Next; inst != s->IP; inst = inst->Next) { - if (inst->I.DstReg.File == PROGRAM_TEMPORARY && inst->I.DstReg.Index == oldindex) - inst->I.DstReg.Index = newindex; - unalias_srcregs(inst, oldindex, newindex); - } - unalias_srcregs(s->IP, oldindex, newindex); -} - - -/** - * Handle one instruction. - */ -static void process_instruction(struct nqssadce_state* s) -{ - struct prog_instruction *inst = &s->IP->I; - GLuint WriteMask; - - if (inst->Opcode == OPCODE_END) - return; - - if (inst->Opcode != OPCODE_KIL) { - struct register_state *regstate = get_reg_state(s, inst->DstReg.File, inst->DstReg.Index); - if (!regstate) { - rc_error(s->Compiler, "NqssaDce: bad destination register (%i[%i])\n", - inst->DstReg.File, inst->DstReg.Index); - return; - } - - inst->DstReg.WriteMask &= regstate->Sourced; - regstate->Sourced &= ~inst->DstReg.WriteMask; - - if (inst->DstReg.WriteMask == 0) { - struct rc_instruction * inst_remove = s->IP; - s->IP = s->IP->Prev; - rc_remove_instruction(inst_remove); - return; - } - - if (inst->DstReg.File == PROGRAM_TEMPORARY && !regstate->Sourced) - unalias_temporary(s, inst->DstReg.Index); - } - - WriteMask = inst->DstReg.WriteMask; - - switch (inst->Opcode) { - case OPCODE_ARL: - case OPCODE_DDX: - case OPCODE_DDY: - case OPCODE_FRC: - case OPCODE_MOV: - track_used_srcreg(s, 0, WriteMask); - break; - case OPCODE_ADD: - case OPCODE_MAX: - case OPCODE_MIN: - case OPCODE_MUL: - case OPCODE_SGE: - case OPCODE_SLT: - track_used_srcreg(s, 0, WriteMask); - track_used_srcreg(s, 1, WriteMask); - break; - case OPCODE_CMP: - case OPCODE_MAD: - track_used_srcreg(s, 0, WriteMask); - track_used_srcreg(s, 1, WriteMask); - track_used_srcreg(s, 2, WriteMask); - break; - case OPCODE_COS: - case OPCODE_EX2: - case OPCODE_LG2: - case OPCODE_RCP: - case OPCODE_RSQ: - case OPCODE_SIN: - track_used_srcreg(s, 0, 0x1); - break; - case OPCODE_DP3: - track_used_srcreg(s, 0, 0x7); - track_used_srcreg(s, 1, 0x7); - break; - case OPCODE_DP4: - track_used_srcreg(s, 0, 0xf); - track_used_srcreg(s, 1, 0xf); - break; - case OPCODE_KIL: - case OPCODE_TEX: - case OPCODE_TXB: - case OPCODE_TXP: - track_used_srcreg(s, 0, 0xf); - break; - case OPCODE_DST: - track_used_srcreg(s, 0, 0x6); - track_used_srcreg(s, 1, 0xa); - break; - case OPCODE_EXP: - case OPCODE_LOG: - case OPCODE_POW: - track_used_srcreg(s, 0, 0x3); - break; - case OPCODE_LIT: - track_used_srcreg(s, 0, 0xb); - break; - default: - rc_error(s->Compiler, "NqssaDce: Unknown opcode %d\n", inst->Opcode); - return; - } - - s->IP = s->IP->Prev; -} - -void rc_calculate_inputs_outputs(struct radeon_compiler * c) -{ - struct rc_instruction *inst; - - c->Program.InputsRead = 0; - c->Program.OutputsWritten = 0; - - for(inst = c->Program.Instructions.Next; inst != &c->Program.Instructions; inst = inst->Next) - { - int i; - int num_src_regs = _mesa_num_inst_src_regs(inst->I.Opcode); - - for (i = 0; i < num_src_regs; ++i) { - if (inst->I.SrcReg[i].File == PROGRAM_INPUT) - c->Program.InputsRead |= 1 << inst->I.SrcReg[i].Index; - } - - if (_mesa_num_inst_dst_regs(inst->I.Opcode)) { - if (inst->I.DstReg.File == PROGRAM_OUTPUT) - c->Program.OutputsWritten |= 1 << inst->I.DstReg.Index; - } - } -} - -void radeonNqssaDce(struct radeon_compiler * c, struct radeon_nqssadce_descr* descr, void * data) -{ - struct nqssadce_state s; - - _mesa_bzero(&s, sizeof(s)); - s.Compiler = c; - s.Descr = descr; - s.UserData = data; - s.Descr->Init(&s); - s.IP = c->Program.Instructions.Prev; - - while(s.IP != &c->Program.Instructions && !c->Error) - process_instruction(&s); - - rc_calculate_inputs_outputs(c); -} diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_nqssadce.h b/src/mesa/drivers/dri/r300/compiler/radeon_nqssadce.h deleted file mode 100644 index b3fc77a35a6..00000000000 --- a/src/mesa/drivers/dri/r300/compiler/radeon_nqssadce.h +++ /dev/null @@ -1,91 +0,0 @@ -/* - * Copyright (C) 2008 Nicolai Haehnle. - * - * 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, sublicense, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial - * portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - */ - -#ifndef __RADEON_PROGRAM_NQSSADCE_H_ -#define __RADEON_PROGRAM_NQSSADCE_H_ - -#include "radeon_program.h" - -struct register_state { - /** - * Bitmask indicating which components of the register are sourced - * by later instructions. - */ - GLuint Sourced : 4; -}; - -/** - * Maintain state such as which registers are used, which registers are - * read from, etc. - */ -struct nqssadce_state { - struct radeon_compiler *Compiler; - struct radeon_nqssadce_descr *Descr; - - /** - * All instructions after this instruction pointer have been dealt with. - */ - struct rc_instruction * IP; - - /** - * Which registers are read by subsequent instructions? - */ - struct register_state Temps[MAX_PROGRAM_TEMPS]; - struct register_state Outputs[VERT_RESULT_MAX]; - struct register_state Address; - - void * UserData; -}; - - -/** - * This structure contains a description of the hardware in-so-far as - * it is required for the NqSSA-DCE pass. - */ -struct radeon_nqssadce_descr { - /** - * Fill in which outputs - */ - void (*Init)(struct nqssadce_state *); - - /** - * Check whether the given swizzle, absolute and negate combination - * can be implemented natively by the hardware for this opcode. - */ - GLboolean (*IsNativeSwizzle)(GLuint opcode, struct prog_src_register reg); - - /** - * Emit (at the current IP) the instruction MOV dst, src; - * The transformation will work recursively on the emitted instruction(s). - */ - void (*BuildSwizzle)(struct nqssadce_state*, struct prog_dst_register dst, struct prog_src_register src); -}; - -void radeonNqssaDce(struct radeon_compiler * c, struct radeon_nqssadce_descr* descr, void * data); -struct prog_src_register lmul_swizzle(GLuint swizzle, struct prog_src_register srcreg); - -#endif /* __RADEON_PROGRAM_NQSSADCE_H_ */ diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_opcodes.c b/src/mesa/drivers/dri/r300/compiler/radeon_opcodes.c new file mode 100644 index 00000000000..c1c0181fac1 --- /dev/null +++ b/src/mesa/drivers/dri/r300/compiler/radeon_opcodes.c @@ -0,0 +1,429 @@ +/* + * Copyright (C) 2009 Nicolai Haehnle. + * + * 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, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "radeon_opcodes.h" + +#include "radeon_program_constants.h" + +struct rc_opcode_info rc_opcodes[MAX_RC_OPCODE] = { + { + .Opcode = RC_OPCODE_NOP, + .Name = "NOP" + }, + { + .Opcode = RC_OPCODE_ILLEGAL_OPCODE, + .Name = "ILLEGAL OPCODE" + }, + { + .Opcode = RC_OPCODE_ABS, + .Name = "ABS", + .NumSrcRegs = 1, + .HasDstReg = 1, + .IsComponentwise = 1 + }, + { + .Opcode = RC_OPCODE_ADD, + .Name = "ADD", + .NumSrcRegs = 2, + .HasDstReg = 1, + .IsComponentwise = 1 + }, + { + .Opcode = RC_OPCODE_ARL, + .Name = "ARL", + .NumSrcRegs = 1, + .HasDstReg = 1 + }, + { + .Opcode = RC_OPCODE_CMP, + .Name = "CMP", + .NumSrcRegs = 3, + .HasDstReg = 1, + .IsComponentwise = 1 + }, + { + .Opcode = RC_OPCODE_COS, + .Name = "COS", + .NumSrcRegs = 1, + .HasDstReg = 1, + .IsStandardScalar = 1 + }, + { + .Opcode = RC_OPCODE_DDX, + .Name = "DDX", + .NumSrcRegs = 1, + .HasDstReg = 1, + .IsComponentwise = 1 + }, + { + .Opcode = RC_OPCODE_DDY, + .Name = "DDY", + .NumSrcRegs = 1, + .HasDstReg = 1, + .IsComponentwise = 1 + }, + { + .Opcode = RC_OPCODE_DP3, + .Name = "DP3", + .NumSrcRegs = 2, + .HasDstReg = 1 + }, + { + .Opcode = RC_OPCODE_DP4, + .Name = "DP4", + .NumSrcRegs = 2, + .HasDstReg = 1 + }, + { + .Opcode = RC_OPCODE_DPH, + .Name = "DPH", + .NumSrcRegs = 2, + .HasDstReg = 1 + }, + { + .Opcode = RC_OPCODE_DST, + .Name = "DST", + .NumSrcRegs = 2, + .HasDstReg = 1 + }, + { + .Opcode = RC_OPCODE_EX2, + .Name = "EX2", + .NumSrcRegs = 1, + .HasDstReg = 1, + .IsStandardScalar = 1 + }, + { + .Opcode = RC_OPCODE_EXP, + .Name = "EXP", + .NumSrcRegs = 1, + .HasDstReg = 1 + }, + { + .Opcode = RC_OPCODE_FLR, + .Name = "FLR", + .NumSrcRegs = 1, + .HasDstReg = 1, + .IsComponentwise = 1 + }, + { + .Opcode = RC_OPCODE_FRC, + .Name = "FRC", + .NumSrcRegs = 1, + .HasDstReg = 1, + .IsComponentwise = 1 + }, + { + .Opcode = RC_OPCODE_KIL, + .Name = "KIL", + .NumSrcRegs = 1 + }, + { + .Opcode = RC_OPCODE_LG2, + .Name = "LG2", + .NumSrcRegs = 1, + .HasDstReg = 1, + .IsStandardScalar = 1 + }, + { + .Opcode = RC_OPCODE_LIT, + .Name = "LIT", + .NumSrcRegs = 1, + .HasDstReg = 1 + }, + { + .Opcode = RC_OPCODE_LOG, + .Name = "LOG", + .NumSrcRegs = 1, + .HasDstReg = 1 + }, + { + .Opcode = RC_OPCODE_LRP, + .Name = "LRP", + .NumSrcRegs = 3, + .HasDstReg = 1, + .IsComponentwise = 1 + }, + { + .Opcode = RC_OPCODE_MAD, + .Name = "MAD", + .NumSrcRegs = 3, + .HasDstReg = 1, + .IsComponentwise = 1 + }, + { + .Opcode = RC_OPCODE_MAX, + .Name = "MAX", + .NumSrcRegs = 2, + .HasDstReg = 1, + .IsComponentwise = 1 + }, + { + .Opcode = RC_OPCODE_MIN, + .Name = "MIN", + .NumSrcRegs = 2, + .HasDstReg = 1, + .IsComponentwise = 1 + }, + { + .Opcode = RC_OPCODE_MOV, + .Name = "MOV", + .NumSrcRegs = 1, + .HasDstReg = 1, + .IsComponentwise = 1 + }, + { + .Opcode = RC_OPCODE_MUL, + .Name = "MUL", + .NumSrcRegs = 2, + .HasDstReg = 1, + .IsComponentwise = 1 + }, + { + .Opcode = RC_OPCODE_POW, + .Name = "POW", + .NumSrcRegs = 2, + .HasDstReg = 1, + .IsStandardScalar = 1 + }, + { + .Opcode = RC_OPCODE_RCP, + .Name = "RCP", + .NumSrcRegs = 1, + .HasDstReg = 1, + .IsStandardScalar = 1 + }, + { + .Opcode = RC_OPCODE_RSQ, + .Name = "RSQ", + .NumSrcRegs = 1, + .HasDstReg = 1, + .IsStandardScalar = 1 + }, + { + .Opcode = RC_OPCODE_SCS, + .Name = "SCS", + .NumSrcRegs = 1, + .HasDstReg = 1 + }, + { + .Opcode = RC_OPCODE_SEQ, + .Name = "SEQ", + .NumSrcRegs = 2, + .HasDstReg = 1, + .IsComponentwise = 1 + }, + { + .Opcode = RC_OPCODE_SFL, + .Name = "SFL", + .NumSrcRegs = 0, + .HasDstReg = 1, + .IsComponentwise = 1 + }, + { + .Opcode = RC_OPCODE_SGE, + .Name = "SGE", + .NumSrcRegs = 2, + .HasDstReg = 1, + .IsComponentwise = 1 + }, + { + .Opcode = RC_OPCODE_SGT, + .Name = "SGT", + .NumSrcRegs = 2, + .HasDstReg = 1, + .IsComponentwise = 1 + }, + { + .Opcode = RC_OPCODE_SIN, + .Name = "SIN", + .NumSrcRegs = 1, + .HasDstReg = 1, + .IsStandardScalar = 1 + }, + { + .Opcode = RC_OPCODE_SLE, + .Name = "SLE", + .NumSrcRegs = 2, + .HasDstReg = 1, + .IsComponentwise = 1 + }, + { + .Opcode = RC_OPCODE_SLT, + .Name = "SLT", + .NumSrcRegs = 2, + .HasDstReg = 1, + .IsComponentwise = 1 + }, + { + .Opcode = RC_OPCODE_SNE, + .Name = "SNE", + .NumSrcRegs = 2, + .HasDstReg = 1, + .IsComponentwise = 1 + }, + { + .Opcode = RC_OPCODE_SUB, + .Name = "SUB", + .NumSrcRegs = 2, + .HasDstReg = 1, + .IsComponentwise = 1 + }, + { + .Opcode = RC_OPCODE_SWZ, + .Name = "SWZ", + .NumSrcRegs = 1, + .HasDstReg = 1, + .IsComponentwise = 1 + }, + { + .Opcode = RC_OPCODE_XPD, + .Name = "XPD", + .NumSrcRegs = 2, + .HasDstReg = 1 + }, + { + .Opcode = RC_OPCODE_TEX, + .Name = "TEX", + .HasTexture = 1, + .NumSrcRegs = 1, + .HasDstReg = 1 + }, + { + .Opcode = RC_OPCODE_TXB, + .Name = "TXB", + .HasTexture = 1, + .NumSrcRegs = 1, + .HasDstReg = 1 + }, + { + .Opcode = RC_OPCODE_TXD, + .Name = "TXD", + .HasTexture = 1, + .NumSrcRegs = 3, + .HasDstReg = 1 + }, + { + .Opcode = RC_OPCODE_TXL, + .Name = "TXL", + .HasTexture = 1, + .NumSrcRegs = 1, + .HasDstReg = 1 + }, + { + .Opcode = RC_OPCODE_TXP, + .Name = "TXP", + .HasTexture = 1, + .NumSrcRegs = 1, + .HasDstReg = 1 + }, + { + .Opcode = RC_OPCODE_IF, + .Name = "IF", + .IsFlowControl = 1, + .NumSrcRegs = 1 + }, + { + .Opcode = RC_OPCODE_ELSE, + .Name = "ELSE", + .IsFlowControl = 1, + .NumSrcRegs = 0 + }, + { + .Opcode = RC_OPCODE_ENDIF, + .Name = "ENDIF", + .IsFlowControl = 1, + .NumSrcRegs = 0 + }, + { + .Opcode = RC_OPCODE_REPL_ALPHA, + .Name = "REPL_ALPHA", + .HasDstReg = 1 + }, + { + .Opcode = RC_OPCODE_BEGIN_TEX, + .Name = "BEGIN_TEX" + } +}; + +void rc_compute_sources_for_writemask( + const struct rc_opcode_info * opcode, + unsigned int writemask, + unsigned int *srcmasks) +{ + srcmasks[0] = 0; + srcmasks[1] = 0; + srcmasks[2] = 0; + + if (opcode->Opcode == RC_OPCODE_KIL) + srcmasks[0] |= RC_MASK_XYZW; + else if (opcode->Opcode == RC_OPCODE_IF) + srcmasks[0] |= RC_MASK_X; + + if (!writemask) + return; + + if (opcode->IsComponentwise) { + for(unsigned int src = 0; src < opcode->NumSrcRegs; ++src) + srcmasks[src] |= writemask; + } else if (opcode->IsStandardScalar) { + for(unsigned int src = 0; src < opcode->NumSrcRegs; ++src) + srcmasks[src] |= RC_MASK_X; + } else { + switch(opcode->Opcode) { + case RC_OPCODE_ARL: + srcmasks[0] |= RC_MASK_X; + break; + case RC_OPCODE_DP3: + srcmasks[0] |= RC_MASK_XYZ; + srcmasks[1] |= RC_MASK_XYZ; + break; + case RC_OPCODE_DP4: + srcmasks[0] |= RC_MASK_XYZW; + srcmasks[1] |= RC_MASK_XYZW; + break; + case RC_OPCODE_TEX: + case RC_OPCODE_TXB: + case RC_OPCODE_TXP: + srcmasks[0] |= RC_MASK_XYZW; + break; + case RC_OPCODE_DST: + srcmasks[0] |= 0x6; + srcmasks[1] |= 0xa; + break; + case RC_OPCODE_EXP: + case RC_OPCODE_LOG: + srcmasks[0] |= RC_MASK_XY; + break; + case RC_OPCODE_LIT: + srcmasks[0] |= 0xb; + break; + default: + break; + } + } +} diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_opcodes.h b/src/mesa/drivers/dri/r300/compiler/radeon_opcodes.h new file mode 100644 index 00000000000..a3c5b869546 --- /dev/null +++ b/src/mesa/drivers/dri/r300/compiler/radeon_opcodes.h @@ -0,0 +1,235 @@ +/* + * Copyright (C) 2009 Nicolai Haehnle. + * + * 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, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef RADEON_OPCODES_H +#define RADEON_OPCODES_H + +#include <assert.h> + +/** + * Opcodes understood by the Radeon compiler. + */ +typedef enum { + RC_OPCODE_NOP = 0, + RC_OPCODE_ILLEGAL_OPCODE, + + /** vec4 instruction: dst.c = abs(src0.c); */ + RC_OPCODE_ABS, + + /** vec4 instruction: dst.c = src0.c + src1.c; */ + RC_OPCODE_ADD, + + /** special instruction: load address register + * dst.x = floor(src.x), where dst must be an address register */ + RC_OPCODE_ARL, + + /** vec4 instruction: dst.c = src0.c < 0.0 ? src1.c : src2.c */ + RC_OPCODE_CMP, + + /** scalar instruction: dst = cos(src0.x) */ + RC_OPCODE_COS, + + /** special instruction: take vec4 partial derivative in X direction + * dst.c = d src0.c / dx */ + RC_OPCODE_DDX, + + /** special instruction: take vec4 partial derivative in Y direction + * dst.c = d src0.c / dy */ + RC_OPCODE_DDY, + + /** scalar instruction: dst = src0.x*src1.x + src0.y*src1.y + src0.z*src1.z */ + RC_OPCODE_DP3, + + /** scalar instruction: dst = src0.x*src1.x + src0.y*src1.y + src0.z*src1.z + src0.w*src1.w */ + RC_OPCODE_DP4, + + /** scalar instruction: dst = src0.x*src1.x + src0.y*src1.y + src0.z*src1.z + src1.w */ + RC_OPCODE_DPH, + + /** special instruction, see ARB_fragment_program */ + RC_OPCODE_DST, + + /** scalar instruction: dst = 2**src0.x */ + RC_OPCODE_EX2, + + /** special instruction, see ARB_vertex_program */ + RC_OPCODE_EXP, + + /** vec4 instruction: dst.c = floor(src0.c) */ + RC_OPCODE_FLR, + + /** vec4 instruction: dst.c = src0.c - floor(src0.c) */ + RC_OPCODE_FRC, + + /** special instruction: stop execution if any component of src0 is negative */ + RC_OPCODE_KIL, + + /** scalar instruction: dst = log_2(src0.x) */ + RC_OPCODE_LG2, + + /** special instruction, see ARB_vertex_program */ + RC_OPCODE_LIT, + + /** special instruction, see ARB_vertex_program */ + RC_OPCODE_LOG, + + /** vec4 instruction: dst.c = src0.c*src1.c + (1 - src0.c)*src2.c */ + RC_OPCODE_LRP, + + /** vec4 instruction: dst.c = src0.c*src1.c + src2.c */ + RC_OPCODE_MAD, + + /** vec4 instruction: dst.c = max(src0.c, src1.c) */ + RC_OPCODE_MAX, + + /** vec4 instruction: dst.c = min(src0.c, src1.c) */ + RC_OPCODE_MIN, + + /** vec4 instruction: dst.c = src0.c */ + RC_OPCODE_MOV, + + /** vec4 instruction: dst.c = src0.c*src1.c */ + RC_OPCODE_MUL, + + /** scalar instruction: dst = src0.x ** src1.x */ + RC_OPCODE_POW, + + /** scalar instruction: dst = 1 / src0.x */ + RC_OPCODE_RCP, + + /** scalar instruction: dst = 1 / sqrt(src0.x) */ + RC_OPCODE_RSQ, + + /** special instruction, see ARB_fragment_program */ + RC_OPCODE_SCS, + + /** vec4 instruction: dst.c = (src0.c == src1.c) ? 1.0 : 0.0 */ + RC_OPCODE_SEQ, + + /** vec4 instruction: dst.c = 0.0 */ + RC_OPCODE_SFL, + + /** vec4 instruction: dst.c = (src0.c >= src1.c) ? 1.0 : 0.0 */ + RC_OPCODE_SGE, + + /** vec4 instruction: dst.c = (src0.c > src1.c) ? 1.0 : 0.0 */ + RC_OPCODE_SGT, + + /** scalar instruction: dst = sin(src0.x) */ + RC_OPCODE_SIN, + + /** vec4 instruction: dst.c = (src0.c <= src1.c) ? 1.0 : 0.0 */ + RC_OPCODE_SLE, + + /** vec4 instruction: dst.c = (src0.c < src1.c) ? 1.0 : 0.0 */ + RC_OPCODE_SLT, + + /** vec4 instruction: dst.c = (src0.c != src1.c) ? 1.0 : 0.0 */ + RC_OPCODE_SNE, + + /** vec4 instruction: dst.c = src0.c - src1.c */ + RC_OPCODE_SUB, + + /** vec4 instruction: dst.c = src0.c */ + RC_OPCODE_SWZ, + + /** special instruction, see ARB_fragment_program */ + RC_OPCODE_XPD, + + RC_OPCODE_TEX, + RC_OPCODE_TXB, + RC_OPCODE_TXD, + RC_OPCODE_TXL, + RC_OPCODE_TXP, + + /** branch instruction: + * If src0.x != 0.0, continue with the next instruction; + * otherwise, jump to matching RC_OPCODE_ELSE or RC_OPCODE_ENDIF. + */ + RC_OPCODE_IF, + + /** branch instruction: jump to matching RC_OPCODE_ENDIF */ + RC_OPCODE_ELSE, + + /** branch instruction: has no effect */ + RC_OPCODE_ENDIF, + + /** special instruction, used in R300-R500 fragment program pair instructions + * indicates that the result of the alpha operation shall be replicated + * across all other channels */ + RC_OPCODE_REPL_ALPHA, + + /** special instruction, used in R300-R500 fragment programs + * to indicate the start of a block of texture instructions that + * can run simultaneously. */ + RC_OPCODE_BEGIN_TEX, + + MAX_RC_OPCODE +} rc_opcode; + + +struct rc_opcode_info { + rc_opcode Opcode; + const char * Name; + + /** true if the instruction reads from a texture. + * + * \note This is false for the KIL instruction, even though KIL is + * a texture instruction from a hardware point of view. */ + unsigned int HasTexture:1; + + unsigned int NumSrcRegs:2; + unsigned int HasDstReg:1; + + /** true if this instruction affects control flow */ + unsigned int IsFlowControl:1; + + /** true if this is a vector instruction that operates on components in parallel + * without any cross-component interaction */ + unsigned int IsComponentwise:1; + + /** true if this instruction sources only its operands X components + * to compute one result which is smeared across all output channels */ + unsigned int IsStandardScalar:1; +}; + +extern struct rc_opcode_info rc_opcodes[MAX_RC_OPCODE]; + +static inline const struct rc_opcode_info * rc_get_opcode_info(rc_opcode opcode) +{ + assert((unsigned int)opcode < MAX_RC_OPCODE); + assert(rc_opcodes[opcode].Opcode == opcode); + + return &rc_opcodes[opcode]; +} + +void rc_compute_sources_for_writemask( + const struct rc_opcode_info * opcode, + unsigned int writemask, + unsigned int *srcmasks); + +#endif /* RADEON_OPCODES_H */ diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_pair_regalloc.c b/src/mesa/drivers/dri/r300/compiler/radeon_pair_regalloc.c new file mode 100644 index 00000000000..828d0c8e28e --- /dev/null +++ b/src/mesa/drivers/dri/r300/compiler/radeon_pair_regalloc.c @@ -0,0 +1,350 @@ +/* + * Copyright (C) 2009 Nicolai Haehnle. + * + * 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, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "radeon_program_pair.h" + +#include <stdio.h> + +#include "radeon_compiler.h" +#include "radeon_dataflow.h" + + +#define VERBOSE 0 + +#define DBG(...) do { if (VERBOSE) fprintf(stderr, __VA_ARGS__); } while(0) + + +struct live_intervals { + int Start; + int End; + struct live_intervals * Next; +}; + +struct register_info { + struct live_intervals Live; + + unsigned int Used:1; + unsigned int Allocated:1; + rc_register_file File:3; + unsigned int Index:RC_REGISTER_INDEX_BITS; +}; + +struct hardware_register { + struct live_intervals * Used; +}; + +struct regalloc_state { + struct radeon_compiler * C; + + struct register_info Input[RC_REGISTER_MAX_INDEX]; + struct register_info Temporary[RC_REGISTER_MAX_INDEX]; + + struct hardware_register * HwTemporary; + unsigned int NumHwTemporaries; +}; + +static void print_live_intervals(struct live_intervals * src) +{ + if (!src) { + DBG("(null)"); + return; + } + + while(src) { + DBG("(%i,%i)", src->Start, src->End); + src = src->Next; + } +} + +static void add_live_intervals(struct regalloc_state * s, + struct live_intervals ** dst, struct live_intervals * src) +{ + struct live_intervals ** dst_backup = dst; + + if (VERBOSE) { + DBG("add_live_intervals: "); + print_live_intervals(*dst); + DBG(" to "); + print_live_intervals(src); + DBG("\n"); + } + + while(src) { + if (*dst && (*dst)->End < src->Start) { + dst = &(*dst)->Next; + } else if (!*dst || (*dst)->Start > src->End) { + struct live_intervals * li = memory_pool_malloc(&s->C->Pool, sizeof(*li)); + li->Start = src->Start; + li->End = src->End; + li->Next = *dst; + *dst = li; + src = src->Next; + } else { + if (src->End > (*dst)->End) + (*dst)->End = src->End; + if (src->Start < (*dst)->Start) + (*dst)->Start = src->Start; + src = src->Next; + } + } + + if (VERBOSE) { + DBG(" result: "); + print_live_intervals(*dst_backup); + DBG("\n"); + } +} + +static int overlap_live_intervals(struct live_intervals * dst, struct live_intervals * src) +{ + if (VERBOSE) { + DBG("overlap_live_intervals: "); + print_live_intervals(dst); + DBG(" to "); + print_live_intervals(src); + DBG("\n"); + } + + while(src && dst) { + if (dst->End <= src->Start) { + dst = dst->Next; + } else if (dst->End <= src->End) { + DBG(" overlap\n"); + return 1; + } else if (dst->Start < src->End) { + DBG(" overlap\n"); + return 1; + } else { + src = src->Next; + } + } + + DBG(" no overlap\n"); + + return 0; +} + +static int try_add_live_intervals(struct regalloc_state * s, + struct live_intervals ** dst, struct live_intervals * src) +{ + if (overlap_live_intervals(*dst, src)) + return 0; + + add_live_intervals(s, dst, src); + return 1; +} + +static void scan_callback(void * data, struct rc_instruction * inst, + rc_register_file file, unsigned int index, unsigned int chan) +{ + struct regalloc_state * s = data; + struct register_info * reg; + + if (file == RC_FILE_TEMPORARY) + reg = &s->Temporary[index]; + else if (file == RC_FILE_INPUT) + reg = &s->Input[index]; + else + return; + + if (!reg->Used) { + reg->Used = 1; + if (file == RC_FILE_INPUT) + reg->Live.Start = -1; + else + reg->Live.Start = inst->IP; + reg->Live.End = inst->IP; + } else { + if (inst->IP > reg->Live.End) + reg->Live.End = inst->IP; + } +} + +static void compute_live_intervals(struct regalloc_state * s) +{ + rc_recompute_ips(s->C); + + for(struct rc_instruction * inst = s->C->Program.Instructions.Next; + inst != &s->C->Program.Instructions; + inst = inst->Next) { + rc_for_all_reads(inst, scan_callback, s); + rc_for_all_writes(inst, scan_callback, s); + } +} + +static void rewrite_register(struct regalloc_state * s, + rc_register_file * file, unsigned int * index) +{ + const struct register_info * reg; + + if (*file == RC_FILE_TEMPORARY) + reg = &s->Temporary[*index]; + else if (*file == RC_FILE_INPUT) + reg = &s->Input[*index]; + else + return; + + if (reg->Allocated) { + *file = reg->File; + *index = reg->Index; + } +} + +static void rewrite_normal_instruction(struct regalloc_state * s, struct rc_sub_instruction * inst) +{ + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->Opcode); + + if (opcode->HasDstReg) { + rc_register_file file = inst->DstReg.File; + unsigned int index = inst->DstReg.Index; + + rewrite_register(s, &file, &index); + + inst->DstReg.File = file; + inst->DstReg.Index = index; + } + + for(unsigned int src = 0; src < opcode->NumSrcRegs; ++src) { + rc_register_file file = inst->SrcReg[src].File; + unsigned int index = inst->SrcReg[src].Index; + + rewrite_register(s, &file, &index); + + inst->SrcReg[src].File = file; + inst->SrcReg[src].Index = index; + } +} + +static void rewrite_pair_instruction(struct regalloc_state * s, struct rc_pair_instruction * inst) +{ + if (inst->RGB.WriteMask) { + rc_register_file file = RC_FILE_TEMPORARY; + unsigned int index = inst->RGB.DestIndex; + + rewrite_register(s, &file, &index); + + inst->RGB.DestIndex = index; + } + + if (inst->Alpha.WriteMask) { + rc_register_file file = RC_FILE_TEMPORARY; + unsigned int index = inst->Alpha.DestIndex; + + rewrite_register(s, &file, &index); + + inst->Alpha.DestIndex = index; + } + + for(unsigned int src = 0; src < 3; ++src) { + if (inst->RGB.Src[src].Used) { + rc_register_file file = inst->RGB.Src[src].File; + unsigned int index = inst->RGB.Src[src].Index; + + rewrite_register(s, &file, &index); + + inst->RGB.Src[src].File = file; + inst->RGB.Src[src].Index = index; + } + + if (inst->Alpha.Src[src].Used) { + rc_register_file file = inst->Alpha.Src[src].File; + unsigned int index = inst->Alpha.Src[src].Index; + + rewrite_register(s, &file, &index); + + inst->Alpha.Src[src].File = file; + inst->Alpha.Src[src].Index = index; + } + } +} + +static void do_regalloc(struct regalloc_state * s) +{ + /* Simple and stupid greedy register allocation */ + for(unsigned int index = 0; index < RC_REGISTER_MAX_INDEX; ++index) { + struct register_info * reg = &s->Temporary[index]; + + if (!reg->Used) + continue; + + for(unsigned int hwreg = 0; hwreg < s->NumHwTemporaries; ++hwreg) { + if (try_add_live_intervals(s, &s->HwTemporary[hwreg].Used, ®->Live)) { + reg->Allocated = 1; + reg->File = RC_FILE_TEMPORARY; + reg->Index = hwreg; + goto success; + } + } + + rc_error(s->C, "Ran out of hardware temporaries\n"); + return; + + success:; + } + + /* Rewrite all instructions based on the translation table we built */ + for(struct rc_instruction * inst = s->C->Program.Instructions.Next; + inst != &s->C->Program.Instructions; + inst = inst->Next) { + if (inst->Type == RC_INSTRUCTION_NORMAL) + rewrite_normal_instruction(s, &inst->U.I); + else + rewrite_pair_instruction(s, &inst->U.P); + } +} + +static void alloc_input(void * data, unsigned int input, unsigned int hwreg) +{ + struct regalloc_state * s = data; + + if (!s->Input[input].Used) + return; + + add_live_intervals(s, &s->HwTemporary[hwreg].Used, &s->Input[input].Live); + + s->Input[input].Allocated = 1; + s->Input[input].File = RC_FILE_TEMPORARY; + s->Input[input].Index = hwreg; + +} + +void rc_pair_regalloc(struct r300_fragment_program_compiler *c, unsigned maxtemps) +{ + struct regalloc_state s; + + memset(&s, 0, sizeof(s)); + s.C = &c->Base; + s.NumHwTemporaries = maxtemps; + s.HwTemporary = memory_pool_malloc(&s.C->Pool, maxtemps*sizeof(struct hardware_register)); + memset(s.HwTemporary, 0, maxtemps*sizeof(struct hardware_register)); + + compute_live_intervals(&s); + + c->AllocateHwInputs(c, &alloc_input, &s); + + do_regalloc(&s); +} diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_pair_schedule.c b/src/mesa/drivers/dri/r300/compiler/radeon_pair_schedule.c new file mode 100644 index 00000000000..df67aafe028 --- /dev/null +++ b/src/mesa/drivers/dri/r300/compiler/radeon_pair_schedule.c @@ -0,0 +1,501 @@ +/* + * Copyright (C) 2009 Nicolai Haehnle. + * + * 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, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "radeon_program_pair.h" + +#include <stdio.h> + +#include "radeon_compiler.h" +#include "radeon_dataflow.h" + + +#define VERBOSE 0 + +#define DBG(...) do { if (VERBOSE) fprintf(stderr, __VA_ARGS__); } while(0) + +struct schedule_instruction { + struct rc_instruction * Instruction; + + /** Next instruction in the linked list of ready instructions. */ + struct schedule_instruction *NextReady; + + /** Values that this instruction reads and writes */ + struct reg_value * WriteValues[4]; + struct reg_value * ReadValues[12]; + unsigned int NumWriteValues:3; + unsigned int NumReadValues:4; + + /** + * Number of (read and write) dependencies that must be resolved before + * this instruction can be scheduled. + */ + unsigned int NumDependencies:5; +}; + + +/** + * Used to keep track of which instructions read a value. + */ +struct reg_value_reader { + struct schedule_instruction *Reader; + struct reg_value_reader *Next; +}; + +/** + * Used to keep track which values are stored in each component of a + * RC_FILE_TEMPORARY. + */ +struct reg_value { + struct schedule_instruction * Writer; + + /** + * Unordered linked list of instructions that read from this value. + * When this value becomes available, we increase all readers' + * dependency count. + */ + struct reg_value_reader *Readers; + + /** + * Number of readers of this value. This is decremented each time + * a reader of the value is committed. + * When the reader cound reaches zero, the dependency count + * of the instruction writing \ref Next is decremented. + */ + unsigned int NumReaders; + + struct reg_value *Next; /**< Pointer to the next value to be written to the same register */ +}; + +struct register_state { + struct reg_value * Values[4]; +}; + +struct schedule_state { + struct radeon_compiler * C; + struct schedule_instruction * Current; + + struct register_state Temporary[RC_REGISTER_MAX_INDEX]; + + /** + * Linked lists of instructions that can be scheduled right now, + * based on which ALU/TEX resources they require. + */ + /*@{*/ + struct schedule_instruction *ReadyFullALU; + struct schedule_instruction *ReadyRGB; + struct schedule_instruction *ReadyAlpha; + struct schedule_instruction *ReadyTEX; + /*@}*/ +}; + +static struct reg_value ** get_reg_valuep(struct schedule_state * s, + rc_register_file file, unsigned int index, unsigned int chan) +{ + if (file != RC_FILE_TEMPORARY) + return 0; + + if (index >= RC_REGISTER_MAX_INDEX) { + rc_error(s->C, "%s: index %i out of bounds\n", __FUNCTION__, index); + return 0; + } + + return &s->Temporary[index].Values[chan]; +} + +static struct reg_value * get_reg_value(struct schedule_state * s, + rc_register_file file, unsigned int index, unsigned int chan) +{ + struct reg_value ** pv = get_reg_valuep(s, file, index, chan); + if (!pv) + return 0; + return *pv; +} + +static void add_inst_to_list(struct schedule_instruction ** list, struct schedule_instruction * inst) +{ + inst->NextReady = *list; + *list = inst; +} + +static void instruction_ready(struct schedule_state * s, struct schedule_instruction * sinst) +{ + DBG("%i is now ready\n", sinst->Instruction->IP); + + if (sinst->Instruction->Type == RC_INSTRUCTION_NORMAL) + add_inst_to_list(&s->ReadyTEX, sinst); + else if (sinst->Instruction->U.P.Alpha.Opcode == RC_OPCODE_NOP) + add_inst_to_list(&s->ReadyRGB, sinst); + else if (sinst->Instruction->U.P.RGB.Opcode == RC_OPCODE_NOP) + add_inst_to_list(&s->ReadyAlpha, sinst); + else + add_inst_to_list(&s->ReadyFullALU, sinst); +} + +static void decrease_dependencies(struct schedule_state * s, struct schedule_instruction * sinst) +{ + assert(sinst->NumDependencies > 0); + sinst->NumDependencies--; + if (!sinst->NumDependencies) + instruction_ready(s, sinst); +} + +static void commit_instruction(struct schedule_state * s, struct schedule_instruction * sinst) +{ + DBG("%i: commit\n", sinst->Instruction->IP); + + for(unsigned int i = 0; i < sinst->NumReadValues; ++i) { + struct reg_value * v = sinst->ReadValues[i]; + assert(v->NumReaders > 0); + v->NumReaders--; + if (!v->NumReaders) { + if (v->Next) + decrease_dependencies(s, v->Next->Writer); + } + } + + for(unsigned int i = 0; i < sinst->NumWriteValues; ++i) { + struct reg_value * v = sinst->WriteValues[i]; + if (v->NumReaders) { + for(struct reg_value_reader * r = v->Readers; r; r = r->Next) { + decrease_dependencies(s, r->Reader); + } + } else { + /* This happens in instruction sequences of the type + * OP r.x, ...; + * OP r.x, r.x, ...; + * See also the subtlety in how instructions that both + * read and write the same register are scanned. + */ + if (v->Next) + decrease_dependencies(s, v->Next->Writer); + } + } +} + +/** + * Emit all ready texture instructions in a single block. + * + * Emit as a single block to (hopefully) sample many textures in parallel, + * and to avoid hardware indirections on R300. + */ +static void emit_all_tex(struct schedule_state * s, struct rc_instruction * before) +{ + struct schedule_instruction *readytex; + + assert(s->ReadyTEX); + + /* Don't let the ready list change under us! */ + readytex = s->ReadyTEX; + s->ReadyTEX = 0; + + /* Node marker for R300 */ + struct rc_instruction * inst_begin = rc_insert_new_instruction(s->C, before->Prev); + inst_begin->U.I.Opcode = RC_OPCODE_BEGIN_TEX; + + /* Link texture instructions back in */ + while(readytex) { + struct schedule_instruction * tex = readytex; + readytex = readytex->NextReady; + + rc_insert_instruction(before->Prev, tex->Instruction); + commit_instruction(s, tex); + } +} + + +static int destructive_merge_instructions( + struct rc_pair_instruction * rgb, + struct rc_pair_instruction * alpha) +{ + assert(rgb->Alpha.Opcode == RC_OPCODE_NOP); + assert(alpha->RGB.Opcode == RC_OPCODE_NOP); + + /* Copy alpha args into rgb */ + const struct rc_opcode_info * opcode = rc_get_opcode_info(alpha->Alpha.Opcode); + + for(unsigned int arg = 0; arg < opcode->NumSrcRegs; ++arg) { + unsigned int srcrgb = 0; + unsigned int srcalpha = 0; + unsigned int oldsrc = alpha->Alpha.Arg[arg].Source; + rc_register_file file = 0; + unsigned int index = 0; + + if (alpha->Alpha.Arg[arg].Swizzle < 3) { + srcrgb = 1; + file = alpha->RGB.Src[oldsrc].File; + index = alpha->RGB.Src[oldsrc].Index; + } else if (alpha->Alpha.Arg[arg].Swizzle < 4) { + srcalpha = 1; + file = alpha->Alpha.Src[oldsrc].File; + index = alpha->Alpha.Src[oldsrc].Index; + } + + int source = rc_pair_alloc_source(rgb, srcrgb, srcalpha, file, index); + if (source < 0) + return 0; + + rgb->Alpha.Arg[arg].Source = source; + rgb->Alpha.Arg[arg].Swizzle = alpha->Alpha.Arg[arg].Swizzle; + rgb->Alpha.Arg[arg].Abs = alpha->Alpha.Arg[arg].Abs; + rgb->Alpha.Arg[arg].Negate = alpha->Alpha.Arg[arg].Negate; + } + + /* Copy alpha opcode into rgb */ + rgb->Alpha.Opcode = alpha->Alpha.Opcode; + rgb->Alpha.DestIndex = alpha->Alpha.DestIndex; + rgb->Alpha.WriteMask = alpha->Alpha.WriteMask; + rgb->Alpha.OutputWriteMask = alpha->Alpha.OutputWriteMask; + rgb->Alpha.DepthWriteMask = alpha->Alpha.DepthWriteMask; + rgb->Alpha.Saturate = alpha->Alpha.Saturate; + + /* Merge ALU result writing */ + if (alpha->WriteALUResult) { + if (rgb->WriteALUResult) + return 0; + + rgb->WriteALUResult = alpha->WriteALUResult; + rgb->ALUResultCompare = alpha->ALUResultCompare; + } + + return 1; +} + +/** + * Try to merge the given instructions into the rgb instructions. + * + * Return true on success; on failure, return false, and keep + * the instructions untouched. + */ +static int merge_instructions(struct rc_pair_instruction * rgb, struct rc_pair_instruction * alpha) +{ + struct rc_pair_instruction backup; + + memcpy(&backup, rgb, sizeof(struct rc_pair_instruction)); + + if (destructive_merge_instructions(rgb, alpha)) + return 1; + + memcpy(rgb, &backup, sizeof(struct rc_pair_instruction)); + return 0; +} + + +/** + * Find a good ALU instruction or pair of ALU instruction and emit it. + * + * Prefer emitting full ALU instructions, so that when we reach a point + * where no full ALU instruction can be emitted, we have more candidates + * for RGB/Alpha pairing. + */ +static void emit_one_alu(struct schedule_state *s, struct rc_instruction * before) +{ + struct schedule_instruction * sinst; + + if (s->ReadyFullALU || !(s->ReadyRGB && s->ReadyAlpha)) { + if (s->ReadyFullALU) { + sinst = s->ReadyFullALU; + s->ReadyFullALU = s->ReadyFullALU->NextReady; + } else if (s->ReadyRGB) { + sinst = s->ReadyRGB; + s->ReadyRGB = s->ReadyRGB->NextReady; + } else { + sinst = s->ReadyAlpha; + s->ReadyAlpha = s->ReadyAlpha->NextReady; + } + + rc_insert_instruction(before->Prev, sinst->Instruction); + commit_instruction(s, sinst); + } else { + struct schedule_instruction **prgb; + struct schedule_instruction **palpha; + + /* Some pairings might fail because they require too + * many source slots; try all possible pairings if necessary */ + for(prgb = &s->ReadyRGB; *prgb; prgb = &(*prgb)->NextReady) { + for(palpha = &s->ReadyAlpha; *palpha; palpha = &(*palpha)->NextReady) { + struct schedule_instruction * psirgb = *prgb; + struct schedule_instruction * psialpha = *palpha; + + if (!merge_instructions(&psirgb->Instruction->U.P, &psialpha->Instruction->U.P)) + continue; + + *prgb = (*prgb)->NextReady; + *palpha = (*palpha)->NextReady; + rc_insert_instruction(before->Prev, psirgb->Instruction); + commit_instruction(s, psirgb); + commit_instruction(s, psialpha); + goto success; + } + } + + /* No success in pairing; just take the first RGB instruction */ + sinst = s->ReadyRGB; + s->ReadyRGB = s->ReadyRGB->NextReady; + + rc_insert_instruction(before->Prev, sinst->Instruction); + commit_instruction(s, sinst); + success: ; + } +} + +static void scan_read(void * data, struct rc_instruction * inst, + rc_register_file file, unsigned int index, unsigned int chan) +{ + struct schedule_state * s = data; + struct reg_value * v = get_reg_value(s, file, index, chan); + + if (!v) + return; + + if (v->Writer == s->Current) { + /* The instruction reads and writes to a register component. + * In this case, we only want to increment dependencies by one. + */ + return; + } + + DBG("%i: read %i[%i] chan %i\n", s->Current->Instruction->IP, file, index, chan); + + struct reg_value_reader * reader = memory_pool_malloc(&s->C->Pool, sizeof(*reader)); + reader->Reader = s->Current; + reader->Next = v->Readers; + v->Readers = reader; + v->NumReaders++; + + s->Current->NumDependencies++; + + if (s->Current->NumReadValues >= 12) { + rc_error(s->C, "%s: NumReadValues overflow\n", __FUNCTION__); + } else { + s->Current->ReadValues[s->Current->NumReadValues++] = v; + } +} + +static void scan_write(void * data, struct rc_instruction * inst, + rc_register_file file, unsigned int index, unsigned int chan) +{ + struct schedule_state * s = data; + struct reg_value ** pv = get_reg_valuep(s, file, index, chan); + + if (!pv) + return; + + DBG("%i: write %i[%i] chan %i\n", s->Current->Instruction->IP, file, index, chan); + + struct reg_value * newv = memory_pool_malloc(&s->C->Pool, sizeof(*newv)); + memset(newv, 0, sizeof(*newv)); + + newv->Writer = s->Current; + + if (*pv) { + (*pv)->Next = newv; + s->Current->NumDependencies++; + } + + *pv = newv; + + if (s->Current->NumWriteValues >= 4) { + rc_error(s->C, "%s: NumWriteValues overflow\n", __FUNCTION__); + } else { + s->Current->WriteValues[s->Current->NumWriteValues++] = newv; + } +} + +static void schedule_block(struct r300_fragment_program_compiler * c, + struct rc_instruction * begin, struct rc_instruction * end) +{ + struct schedule_state s; + + memset(&s, 0, sizeof(s)); + s.C = &c->Base; + + /* Scan instructions for data dependencies */ + unsigned int ip = 0; + for(struct rc_instruction * inst = begin; inst != end; inst = inst->Next) { + s.Current = memory_pool_malloc(&c->Base.Pool, sizeof(*s.Current)); + memset(s.Current, 0, sizeof(struct schedule_instruction)); + + s.Current->Instruction = inst; + inst->IP = ip++; + + DBG("%i: Scanning\n", inst->IP); + + /* The order of things here is subtle and maybe slightly + * counter-intuitive, to account for the case where an + * instruction writes to the same register as it reads + * from. */ + rc_for_all_writes(inst, &scan_write, &s); + rc_for_all_reads(inst, &scan_read, &s); + + DBG("%i: Has %i dependencies\n", inst->IP, s.Current->NumDependencies); + + if (!s.Current->NumDependencies) + instruction_ready(&s, s.Current); + } + + /* Temporarily unlink all instructions */ + begin->Prev->Next = end; + end->Prev = begin->Prev; + + /* Schedule instructions back */ + while(!s.C->Error && + (s.ReadyTEX || s.ReadyRGB || s.ReadyAlpha || s.ReadyFullALU)) { + if (s.ReadyTEX) + emit_all_tex(&s, end); + + while(!s.C->Error && (s.ReadyFullALU || s.ReadyRGB || s.ReadyAlpha)) + emit_one_alu(&s, end); + } +} + +static int is_controlflow(struct rc_instruction * inst) +{ + if (inst->Type == RC_INSTRUCTION_NORMAL) { + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->U.I.Opcode); + return opcode->IsFlowControl; + } + return 0; +} + +void rc_pair_schedule(struct r300_fragment_program_compiler *c) +{ + struct rc_instruction * inst = c->Base.Program.Instructions.Next; + while(inst != &c->Base.Program.Instructions) { + if (is_controlflow(inst)) { + inst = inst->Next; + continue; + } + + struct rc_instruction * first = inst; + + while(inst != &c->Base.Program.Instructions && !is_controlflow(inst)) + inst = inst->Next; + + DBG("Schedule one block\n"); + schedule_block(c, first, inst); + } +} diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_pair_translate.c b/src/mesa/drivers/dri/r300/compiler/radeon_pair_translate.c new file mode 100644 index 00000000000..72117682725 --- /dev/null +++ b/src/mesa/drivers/dri/r300/compiler/radeon_pair_translate.c @@ -0,0 +1,253 @@ +/* + * Copyright (C) 2009 Nicolai Haehnle. + * + * 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, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "radeon_program_pair.h" + +#include "radeon_compiler.h" + + +/** + * Finally rewrite ADD, MOV, MUL as the appropriate native instruction + * and reverse the order of arguments for CMP. + */ +static void final_rewrite(struct rc_sub_instruction *inst) +{ + struct rc_src_register tmp; + + switch(inst->Opcode) { + case RC_OPCODE_ADD: + inst->SrcReg[2] = inst->SrcReg[1]; + inst->SrcReg[1].File = RC_FILE_NONE; + inst->SrcReg[1].Swizzle = RC_SWIZZLE_1111; + inst->SrcReg[1].Negate = RC_MASK_NONE; + inst->Opcode = RC_OPCODE_MAD; + break; + case RC_OPCODE_CMP: + tmp = inst->SrcReg[2]; + inst->SrcReg[2] = inst->SrcReg[0]; + inst->SrcReg[0] = tmp; + break; + case RC_OPCODE_MOV: + /* AMD say we should use CMP. + * However, when we transform + * KIL -r0; + * into + * CMP tmp, -r0, -r0, 0; + * KIL tmp; + * we get incorrect behaviour on R500 when r0 == 0.0. + * It appears that the R500 KIL hardware treats -0.0 as less + * than zero. + */ + inst->SrcReg[1].File = RC_FILE_NONE; + inst->SrcReg[1].Swizzle = RC_SWIZZLE_1111; + inst->SrcReg[2].File = RC_FILE_NONE; + inst->SrcReg[2].Swizzle = RC_SWIZZLE_0000; + inst->Opcode = RC_OPCODE_MAD; + break; + case RC_OPCODE_MUL: + inst->SrcReg[2].File = RC_FILE_NONE; + inst->SrcReg[2].Swizzle = RC_SWIZZLE_0000; + inst->Opcode = RC_OPCODE_MAD; + break; + default: + /* nothing to do */ + break; + } +} + + +/** + * Classify an instruction according to which ALUs etc. it needs + */ +static void classify_instruction(struct rc_sub_instruction * inst, + int * needrgb, int * needalpha, int * istranscendent) +{ + *needrgb = (inst->DstReg.WriteMask & RC_MASK_XYZ) ? 1 : 0; + *needalpha = (inst->DstReg.WriteMask & RC_MASK_W) ? 1 : 0; + *istranscendent = 0; + + if (inst->WriteALUResult == RC_ALURESULT_X) + *needrgb = 1; + else if (inst->WriteALUResult == RC_ALURESULT_W) + *needalpha = 1; + + switch(inst->Opcode) { + case RC_OPCODE_ADD: + case RC_OPCODE_CMP: + case RC_OPCODE_DDX: + case RC_OPCODE_DDY: + case RC_OPCODE_FRC: + case RC_OPCODE_MAD: + case RC_OPCODE_MAX: + case RC_OPCODE_MIN: + case RC_OPCODE_MOV: + case RC_OPCODE_MUL: + break; + case RC_OPCODE_COS: + case RC_OPCODE_EX2: + case RC_OPCODE_LG2: + case RC_OPCODE_RCP: + case RC_OPCODE_RSQ: + case RC_OPCODE_SIN: + *istranscendent = 1; + *needalpha = 1; + break; + case RC_OPCODE_DP4: + *needalpha = 1; + /* fall through */ + case RC_OPCODE_DP3: + *needrgb = 1; + break; + default: + break; + } +} + + +/** + * Fill the given ALU instruction's opcodes and source operands into the given pair, + * if possible. + */ +static void set_pair_instruction(struct r300_fragment_program_compiler *c, + struct rc_pair_instruction * pair, + struct rc_sub_instruction * inst) +{ + memset(pair, 0, sizeof(struct rc_pair_instruction)); + + int needrgb, needalpha, istranscendent; + classify_instruction(inst, &needrgb, &needalpha, &istranscendent); + + if (needrgb) { + if (istranscendent) + pair->RGB.Opcode = RC_OPCODE_REPL_ALPHA; + else + pair->RGB.Opcode = inst->Opcode; + if (inst->SaturateMode == RC_SATURATE_ZERO_ONE) + pair->RGB.Saturate = 1; + } + if (needalpha) { + pair->Alpha.Opcode = inst->Opcode; + if (inst->SaturateMode == RC_SATURATE_ZERO_ONE) + pair->Alpha.Saturate = 1; + } + + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->Opcode); + int nargs = opcode->NumSrcRegs; + int i; + + /* Special case for DDX/DDY (MDH/MDV). */ + if (inst->Opcode == RC_OPCODE_DDX || inst->Opcode == RC_OPCODE_DDY) { + nargs++; + } + + for(i = 0; i < opcode->NumSrcRegs; ++i) { + int source; + if (needrgb && !istranscendent) { + unsigned int srcrgb = 0; + unsigned int srcalpha = 0; + int j; + for(j = 0; j < 3; ++j) { + unsigned int swz = GET_SWZ(inst->SrcReg[i].Swizzle, j); + if (swz < 3) + srcrgb = 1; + else if (swz < 4) + srcalpha = 1; + } + source = rc_pair_alloc_source(pair, srcrgb, srcalpha, + inst->SrcReg[i].File, inst->SrcReg[i].Index); + pair->RGB.Arg[i].Source = source; + pair->RGB.Arg[i].Swizzle = inst->SrcReg[i].Swizzle & 0x1ff; + pair->RGB.Arg[i].Abs = inst->SrcReg[i].Abs; + pair->RGB.Arg[i].Negate = !!(inst->SrcReg[i].Negate & (RC_MASK_X | RC_MASK_Y | RC_MASK_Z)); + } + if (needalpha) { + unsigned int srcrgb = 0; + unsigned int srcalpha = 0; + unsigned int swz = GET_SWZ(inst->SrcReg[i].Swizzle, istranscendent ? 0 : 3); + if (swz < 3) + srcrgb = 1; + else if (swz < 4) + srcalpha = 1; + source = rc_pair_alloc_source(pair, srcrgb, srcalpha, + inst->SrcReg[i].File, inst->SrcReg[i].Index); + pair->Alpha.Arg[i].Source = source; + pair->Alpha.Arg[i].Swizzle = swz; + pair->Alpha.Arg[i].Abs = inst->SrcReg[i].Abs; + pair->Alpha.Arg[i].Negate = !!(inst->SrcReg[i].Negate & RC_MASK_W); + } + } + + /* Destination handling */ + if (inst->DstReg.File == RC_FILE_OUTPUT) { + if (inst->DstReg.Index == c->OutputColor) { + pair->RGB.OutputWriteMask |= inst->DstReg.WriteMask & RC_MASK_XYZ; + pair->Alpha.OutputWriteMask |= GET_BIT(inst->DstReg.WriteMask, 3); + } else if (inst->DstReg.Index == c->OutputDepth) { + pair->Alpha.DepthWriteMask |= GET_BIT(inst->DstReg.WriteMask, 3); + } + } else { + if (needrgb) { + pair->RGB.DestIndex = inst->DstReg.Index; + pair->RGB.WriteMask |= inst->DstReg.WriteMask & RC_MASK_XYZ; + } + if (needalpha) { + pair->Alpha.DestIndex = inst->DstReg.Index; + pair->Alpha.WriteMask |= GET_BIT(inst->DstReg.WriteMask, 3); + } + } + + if (inst->WriteALUResult) { + pair->WriteALUResult = inst->WriteALUResult; + pair->ALUResultCompare = inst->ALUResultCompare; + } +} + + +/** + * Translate all ALU instructions into corresponding pair instructions, + * performing no other changes. + */ +void rc_pair_translate(struct r300_fragment_program_compiler *c) +{ + for(struct rc_instruction * inst = c->Base.Program.Instructions.Next; + inst != &c->Base.Program.Instructions; + inst = inst->Next) { + if (inst->Type != RC_INSTRUCTION_NORMAL) + continue; + + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->U.I.Opcode); + + if (opcode->HasTexture || opcode->IsFlowControl || opcode->Opcode == RC_OPCODE_KIL) + continue; + + struct rc_sub_instruction copy = inst->U.I; + + final_rewrite(©); + inst->Type = RC_INSTRUCTION_PAIR; + set_pair_instruction(c, &inst->U.P, ©); + } +} diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_program.c b/src/mesa/drivers/dri/r300/compiler/radeon_program.c index bbbf0dd7768..0dbc5380bb5 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_program.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_program.c @@ -27,9 +27,9 @@ #include "radeon_program.h" +#include <stdio.h> + #include "radeon_compiler.h" -#include "shader/prog_parameter.h" -#include "shader/prog_print.h" /** @@ -69,37 +69,57 @@ void radeonLocalTransform( } } +/** + * Left multiplication of a register with a swizzle + */ +struct rc_src_register lmul_swizzle(unsigned int swizzle, struct rc_src_register srcreg) +{ + struct rc_src_register tmp = srcreg; + int i; + tmp.Swizzle = 0; + tmp.Negate = 0; + for(i = 0; i < 4; ++i) { + rc_swizzle swz = GET_SWZ(swizzle, i); + if (swz < 4) { + tmp.Swizzle |= GET_SWZ(srcreg.Swizzle, swz) << (i*3); + tmp.Negate |= GET_BIT(srcreg.Negate, swz) << i; + } else { + tmp.Swizzle |= swz << (i*3); + } + } + return tmp; +} -GLint rc_find_free_temporary(struct radeon_compiler * c) +unsigned int rc_find_free_temporary(struct radeon_compiler * c) { - GLboolean used[MAX_PROGRAM_TEMPS]; - GLuint i; + char used[RC_REGISTER_MAX_INDEX]; + unsigned int i; memset(used, 0, sizeof(used)); for (struct rc_instruction * rcinst = c->Program.Instructions.Next; rcinst != &c->Program.Instructions; rcinst = rcinst->Next) { - const struct prog_instruction *inst = &rcinst->I; - const GLuint nsrc = _mesa_num_inst_src_regs(inst->Opcode); - const GLuint ndst = _mesa_num_inst_dst_regs(inst->Opcode); - GLuint k; - - for (k = 0; k < nsrc; k++) { - if (inst->SrcReg[k].File == PROGRAM_TEMPORARY) - used[inst->SrcReg[k].Index] = GL_TRUE; + const struct rc_sub_instruction *inst = &rcinst->U.I; + const struct rc_opcode_info *opcode = rc_get_opcode_info(inst->Opcode); + unsigned int k; + + for (k = 0; k < opcode->NumSrcRegs; k++) { + if (inst->SrcReg[k].File == RC_FILE_TEMPORARY) + used[inst->SrcReg[k].Index] = 1; } - if (ndst) { - if (inst->DstReg.File == PROGRAM_TEMPORARY) - used[inst->DstReg.Index] = GL_TRUE; + if (opcode->HasDstReg) { + if (inst->DstReg.File == RC_FILE_TEMPORARY) + used[inst->DstReg.Index] = 1; } } - for (i = 0; i < MAX_PROGRAM_TEMPS; i++) { + for (i = 0; i < RC_REGISTER_MAX_INDEX; i++) { if (!used[i]) return i; } - return -1; + rc_error(c, "Ran out of temporary registers\n"); + return 0; } @@ -107,24 +127,31 @@ struct rc_instruction *rc_alloc_instruction(struct radeon_compiler * c) { struct rc_instruction * inst = memory_pool_malloc(&c->Pool, sizeof(struct rc_instruction)); - inst->Prev = 0; - inst->Next = 0; + memset(inst, 0, sizeof(struct rc_instruction)); - _mesa_init_instructions(&inst->I, 1); + inst->U.I.Opcode = RC_OPCODE_ILLEGAL_OPCODE; + inst->U.I.DstReg.WriteMask = RC_MASK_XYZW; + inst->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_XYZW; + inst->U.I.SrcReg[1].Swizzle = RC_SWIZZLE_XYZW; + inst->U.I.SrcReg[2].Swizzle = RC_SWIZZLE_XYZW; return inst; } - -struct rc_instruction *rc_insert_new_instruction(struct radeon_compiler * c, struct rc_instruction * after) +void rc_insert_instruction(struct rc_instruction * after, struct rc_instruction * inst) { - struct rc_instruction * inst = rc_alloc_instruction(c); - inst->Prev = after; inst->Next = after->Next; inst->Prev->Next = inst; inst->Next->Prev = inst; +} + +struct rc_instruction *rc_insert_new_instruction(struct radeon_compiler * c, struct rc_instruction * after) +{ + struct rc_instruction * inst = rc_alloc_instruction(c); + + rc_insert_instruction(after, inst); return inst; } @@ -135,53 +162,20 @@ void rc_remove_instruction(struct rc_instruction * inst) inst->Next->Prev = inst->Prev; } - -void rc_mesa_to_rc_program(struct radeon_compiler * c, struct gl_program * program) -{ - struct prog_instruction *source; - unsigned int i; - - for(source = program->Instructions; source->Opcode != OPCODE_END; ++source) { - struct rc_instruction * dest = rc_insert_new_instruction(c, c->Program.Instructions.Prev); - dest->I = *source; - } - - c->Program.ShadowSamplers = program->ShadowSamplers; - c->Program.InputsRead = program->InputsRead; - c->Program.OutputsWritten = program->OutputsWritten; - - for(i = 0; i < program->Parameters->NumParameters; ++i) { - struct rc_constant constant; - - constant.Type = RC_CONSTANT_EXTERNAL; - constant.Size = 4; - constant.u.External = i; - - rc_constants_add(&c->Program.Constants, &constant); - } -} - - /** - * Print program to stderr, default options. + * Return the number of instructions in the program. */ -void rc_print_program(const struct rc_program *prog) +unsigned int rc_recompute_ips(struct radeon_compiler * c) { - GLuint indent = 0; - GLuint linenum = 1; - struct rc_instruction *inst; - - fprintf(stderr, "# Radeon Compiler Program\n"); + unsigned int ip = 0; - for(inst = prog->Instructions.Next; inst != &prog->Instructions; inst = inst->Next) { - fprintf(stderr, "%3d: ", linenum); + for(struct rc_instruction * inst = c->Program.Instructions.Next; + inst != &c->Program.Instructions; + inst = inst->Next) { + inst->IP = ip++; + } - /* Massive hack: We rely on the fact that the printers do not actually - * use the gl_program argument (last argument) in debug mode */ - indent = _mesa_fprint_instruction_opt( - stderr, &inst->I, - indent, PROG_PRINT_DEBUG, 0); + c->Program.Instructions.IP = 0xcafedead; - linenum++; - } + return ip; } diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_program.h b/src/mesa/drivers/dri/r300/compiler/radeon_program.h index 561958608ca..03592884eb2 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_program.h +++ b/src/mesa/drivers/dri/r300/compiler/radeon_program.h @@ -28,37 +28,144 @@ #ifndef __RADEON_PROGRAM_H_ #define __RADEON_PROGRAM_H_ -#include "main/glheader.h" -#include "main/macros.h" -#include "main/enums.h" -#include "shader/program.h" -#include "shader/prog_instruction.h" +#include <stdint.h> +#include <string.h> + +#include "radeon_opcodes.h" +#include "radeon_code.h" +#include "radeon_program_constants.h" +#include "radeon_program_pair.h" struct radeon_compiler; -struct rc_instruction; -struct rc_program; -enum { - PROGRAM_BUILTIN = PROGRAM_FILE_MAX /**< not a real register, but a special swizzle constant */ +struct rc_src_register { + rc_register_file File:3; + + /** Negative values may be used for relative addressing. */ + signed int Index:(RC_REGISTER_INDEX_BITS+1); + unsigned int RelAddr:1; + + unsigned int Swizzle:12; + + /** Take the component-wise absolute value */ + unsigned int Abs:1; + + /** Post-Abs negation. */ + unsigned int Negate:4; +}; + +struct rc_dst_register { + rc_register_file File:3; + + /** Negative values may be used for relative addressing. */ + signed int Index:(RC_REGISTER_INDEX_BITS+1); + unsigned int RelAddr:1; + + unsigned int WriteMask:4; +}; + +/** + * Instructions are maintained by the compiler in a doubly linked list + * of these structures. + * + * This instruction format is intended to be expanded for hardware-specific + * trickery. At different stages of compilation, a different set of + * instruction types may be valid. + */ +struct rc_sub_instruction { + struct rc_src_register SrcReg[3]; + struct rc_dst_register DstReg; + + /** + * Opcode of this instruction, according to \ref rc_opcode enums. + */ + rc_opcode Opcode:8; + + /** + * Saturate each value of the result to the range [0,1] or [-1,1], + * according to \ref rc_saturate_mode enums. + */ + rc_saturate_mode SaturateMode:2; + + /** + * Writing to the special register RC_SPECIAL_ALU_RESULT + */ + /*@{*/ + rc_write_aluresult WriteALUResult:2; + rc_compare_func ALUResultCompare:3; + /*@}*/ + + /** + * \name Extra fields for TEX, TXB, TXD, TXL, TXP instructions. + */ + /*@{*/ + /** Source texture unit. */ + unsigned int TexSrcUnit:5; + + /** Source texture target, one of the \ref rc_texture_target enums */ + rc_texture_target TexSrcTarget:3; + + /** True if tex instruction should do shadow comparison */ + unsigned int TexShadow:1; + /*@}*/ +}; + +typedef enum { + RC_INSTRUCTION_NORMAL = 0, + RC_INSTRUCTION_PAIR +} rc_instruction_type; + +struct rc_instruction { + struct rc_instruction * Prev; + struct rc_instruction * Next; + + rc_instruction_type Type; + union { + struct rc_sub_instruction I; + struct rc_pair_instruction P; + } U; + + /** + * Warning: IPs are not stable. If you want to use them, + * you need to recompute them at the beginning of each pass + * using \ref rc_recompute_ips + */ + unsigned int IP; +}; + +struct rc_program { + /** + * Instructions.Next points to the first instruction, + * Instructions.Prev points to the last instruction. + */ + struct rc_instruction Instructions; + + /* Long term, we should probably remove InputsRead & OutputsWritten, + * since updating dependent state can be fragile, and they aren't + * actually used very often. */ + uint32_t InputsRead; + uint32_t OutputsWritten; + uint32_t ShadowSamplers; /**< Texture units used for shadow sampling. */ + + struct rc_constant_list Constants; }; enum { - OPCODE_REPL_ALPHA = MAX_OPCODE /**< used in paired instructions */ + OPCODE_REPL_ALPHA = MAX_RC_OPCODE /**< used in paired instructions */ }; -#define SWIZZLE_0000 MAKE_SWIZZLE4(SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ZERO) -#define SWIZZLE_1111 MAKE_SWIZZLE4(SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE) -static inline GLuint get_swz(GLuint swz, GLuint idx) +static inline rc_swizzle get_swz(unsigned int swz, rc_swizzle idx) { if (idx & 0x4) return idx; return GET_SWZ(swz, idx); } -static inline GLuint combine_swizzles4(GLuint src, GLuint swz_x, GLuint swz_y, GLuint swz_z, GLuint swz_w) +static inline unsigned int combine_swizzles4(unsigned int src, + rc_swizzle swz_x, rc_swizzle swz_y, rc_swizzle swz_z, rc_swizzle swz_w) { - GLuint ret = 0; + unsigned int ret = 0; ret |= get_swz(src, swz_x); ret |= get_swz(src, swz_y) << 3; @@ -68,22 +175,24 @@ static inline GLuint combine_swizzles4(GLuint src, GLuint swz_x, GLuint swz_y, G return ret; } -static inline GLuint combine_swizzles(GLuint src, GLuint swz) +static inline unsigned int combine_swizzles(unsigned int src, unsigned int swz) { - GLuint ret = 0; + unsigned int ret = 0; - ret |= get_swz(src, GET_SWZ(swz, SWIZZLE_X)); - ret |= get_swz(src, GET_SWZ(swz, SWIZZLE_Y)) << 3; - ret |= get_swz(src, GET_SWZ(swz, SWIZZLE_Z)) << 6; - ret |= get_swz(src, GET_SWZ(swz, SWIZZLE_W)) << 9; + ret |= get_swz(src, GET_SWZ(swz, RC_SWIZZLE_X)); + ret |= get_swz(src, GET_SWZ(swz, RC_SWIZZLE_Y)) << 3; + ret |= get_swz(src, GET_SWZ(swz, RC_SWIZZLE_Z)) << 6; + ret |= get_swz(src, GET_SWZ(swz, RC_SWIZZLE_W)) << 9; return ret; } -static INLINE void reset_srcreg(struct prog_src_register* reg) +struct rc_src_register lmul_swizzle(unsigned int swizzle, struct rc_src_register srcreg); + +static inline void reset_srcreg(struct rc_src_register* reg) { - _mesa_bzero(reg, sizeof(*reg)); - reg->Swizzle = SWIZZLE_NOOP; + memset(reg, 0, sizeof(struct rc_src_register)); + reg->Swizzle = RC_SWIZZLE_XYZW; } @@ -92,13 +201,13 @@ static INLINE void reset_srcreg(struct prog_src_register* reg) * * The function will be called once for each instruction. * It has to either emit the appropriate transformed code for the instruction - * and return GL_TRUE, or return GL_FALSE if it doesn't understand the + * and return true, or return false if it doesn't understand the * instruction. * * The function gets passed the userData as last parameter. */ struct radeon_program_transformation { - GLboolean (*function)( + int (*function)( struct radeon_compiler*, struct rc_instruction*, void*); @@ -110,12 +219,15 @@ void radeonLocalTransform( int num_transformations, struct radeon_program_transformation* transformations); -GLint rc_find_free_temporary(struct radeon_compiler * c); +unsigned int rc_find_free_temporary(struct radeon_compiler * c); struct rc_instruction *rc_alloc_instruction(struct radeon_compiler * c); struct rc_instruction *rc_insert_new_instruction(struct radeon_compiler * c, struct rc_instruction * after); +void rc_insert_instruction(struct rc_instruction * after, struct rc_instruction * inst); void rc_remove_instruction(struct rc_instruction * inst); +unsigned int rc_recompute_ips(struct radeon_compiler * c); + void rc_print_program(const struct rc_program *prog); #endif diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c b/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c index 8071899eaa4..ced66af1eb5 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c @@ -40,175 +40,164 @@ static struct rc_instruction *emit1( struct radeon_compiler * c, struct rc_instruction * after, - gl_inst_opcode Opcode, GLuint Saturate, struct prog_dst_register DstReg, - struct prog_src_register SrcReg) + rc_opcode Opcode, rc_saturate_mode Saturate, struct rc_dst_register DstReg, + struct rc_src_register SrcReg) { struct rc_instruction *fpi = rc_insert_new_instruction(c, after); - fpi->I.Opcode = Opcode; - fpi->I.SaturateMode = Saturate; - fpi->I.DstReg = DstReg; - fpi->I.SrcReg[0] = SrcReg; + fpi->U.I.Opcode = Opcode; + fpi->U.I.SaturateMode = Saturate; + fpi->U.I.DstReg = DstReg; + fpi->U.I.SrcReg[0] = SrcReg; return fpi; } static struct rc_instruction *emit2( struct radeon_compiler * c, struct rc_instruction * after, - gl_inst_opcode Opcode, GLuint Saturate, struct prog_dst_register DstReg, - struct prog_src_register SrcReg0, struct prog_src_register SrcReg1) + rc_opcode Opcode, rc_saturate_mode Saturate, struct rc_dst_register DstReg, + struct rc_src_register SrcReg0, struct rc_src_register SrcReg1) { struct rc_instruction *fpi = rc_insert_new_instruction(c, after); - fpi->I.Opcode = Opcode; - fpi->I.SaturateMode = Saturate; - fpi->I.DstReg = DstReg; - fpi->I.SrcReg[0] = SrcReg0; - fpi->I.SrcReg[1] = SrcReg1; + fpi->U.I.Opcode = Opcode; + fpi->U.I.SaturateMode = Saturate; + fpi->U.I.DstReg = DstReg; + fpi->U.I.SrcReg[0] = SrcReg0; + fpi->U.I.SrcReg[1] = SrcReg1; return fpi; } static struct rc_instruction *emit3( struct radeon_compiler * c, struct rc_instruction * after, - gl_inst_opcode Opcode, GLuint Saturate, struct prog_dst_register DstReg, - struct prog_src_register SrcReg0, struct prog_src_register SrcReg1, - struct prog_src_register SrcReg2) + rc_opcode Opcode, rc_saturate_mode Saturate, struct rc_dst_register DstReg, + struct rc_src_register SrcReg0, struct rc_src_register SrcReg1, + struct rc_src_register SrcReg2) { struct rc_instruction *fpi = rc_insert_new_instruction(c, after); - fpi->I.Opcode = Opcode; - fpi->I.SaturateMode = Saturate; - fpi->I.DstReg = DstReg; - fpi->I.SrcReg[0] = SrcReg0; - fpi->I.SrcReg[1] = SrcReg1; - fpi->I.SrcReg[2] = SrcReg2; + fpi->U.I.Opcode = Opcode; + fpi->U.I.SaturateMode = Saturate; + fpi->U.I.DstReg = DstReg; + fpi->U.I.SrcReg[0] = SrcReg0; + fpi->U.I.SrcReg[1] = SrcReg1; + fpi->U.I.SrcReg[2] = SrcReg2; return fpi; } -static struct prog_dst_register dstreg(int file, int index) +static struct rc_dst_register dstreg(int file, int index) { - struct prog_dst_register dst; + struct rc_dst_register dst; dst.File = file; dst.Index = index; - dst.WriteMask = WRITEMASK_XYZW; - dst.CondMask = COND_TR; + dst.WriteMask = RC_MASK_XYZW; dst.RelAddr = 0; - dst.CondSwizzle = SWIZZLE_NOOP; - dst.CondSrc = 0; - dst.pad = 0; return dst; } -static struct prog_dst_register dstregtmpmask(int index, int mask) +static struct rc_dst_register dstregtmpmask(int index, int mask) { - struct prog_dst_register dst = {0}; - dst.File = PROGRAM_TEMPORARY; + struct rc_dst_register dst = {0}; + dst.File = RC_FILE_TEMPORARY; dst.Index = index; dst.WriteMask = mask; dst.RelAddr = 0; - dst.CondMask = COND_TR; - dst.CondSwizzle = SWIZZLE_NOOP; - dst.CondSrc = 0; - dst.pad = 0; return dst; } -static const struct prog_src_register builtin_zero = { - .File = PROGRAM_BUILTIN, +static const struct rc_src_register builtin_zero = { + .File = RC_FILE_NONE, .Index = 0, - .Swizzle = SWIZZLE_0000 + .Swizzle = RC_SWIZZLE_0000 }; -static const struct prog_src_register builtin_one = { - .File = PROGRAM_BUILTIN, +static const struct rc_src_register builtin_one = { + .File = RC_FILE_NONE, .Index = 0, - .Swizzle = SWIZZLE_1111 + .Swizzle = RC_SWIZZLE_1111 }; -static const struct prog_src_register srcreg_undefined = { - .File = PROGRAM_UNDEFINED, +static const struct rc_src_register srcreg_undefined = { + .File = RC_FILE_NONE, .Index = 0, - .Swizzle = SWIZZLE_NOOP + .Swizzle = RC_SWIZZLE_XYZW }; -static struct prog_src_register srcreg(int file, int index) +static struct rc_src_register srcreg(int file, int index) { - struct prog_src_register src = srcreg_undefined; + struct rc_src_register src = srcreg_undefined; src.File = file; src.Index = index; return src; } -static struct prog_src_register srcregswz(int file, int index, int swz) +static struct rc_src_register srcregswz(int file, int index, int swz) { - struct prog_src_register src = srcreg_undefined; + struct rc_src_register src = srcreg_undefined; src.File = file; src.Index = index; src.Swizzle = swz; return src; } -static struct prog_src_register absolute(struct prog_src_register reg) +static struct rc_src_register absolute(struct rc_src_register reg) { - struct prog_src_register newreg = reg; + struct rc_src_register newreg = reg; newreg.Abs = 1; - newreg.Negate = NEGATE_NONE; + newreg.Negate = RC_MASK_NONE; return newreg; } -static struct prog_src_register negate(struct prog_src_register reg) +static struct rc_src_register negate(struct rc_src_register reg) { - struct prog_src_register newreg = reg; - newreg.Negate = newreg.Negate ^ NEGATE_XYZW; + struct rc_src_register newreg = reg; + newreg.Negate = newreg.Negate ^ RC_MASK_XYZW; return newreg; } -static struct prog_src_register swizzle(struct prog_src_register reg, GLuint x, GLuint y, GLuint z, GLuint w) +static struct rc_src_register swizzle(struct rc_src_register reg, + rc_swizzle x, rc_swizzle y, rc_swizzle z, rc_swizzle w) { - struct prog_src_register swizzled = reg; - swizzled.Swizzle = MAKE_SWIZZLE4( - x >= 4 ? x : GET_SWZ(reg.Swizzle, x), - y >= 4 ? y : GET_SWZ(reg.Swizzle, y), - z >= 4 ? z : GET_SWZ(reg.Swizzle, z), - w >= 4 ? w : GET_SWZ(reg.Swizzle, w)); + struct rc_src_register swizzled = reg; + swizzled.Swizzle = combine_swizzles4(reg.Swizzle, x, y, z, w); return swizzled; } -static struct prog_src_register scalar(struct prog_src_register reg) +static struct rc_src_register scalar(struct rc_src_register reg) { - return swizzle(reg, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X); + return swizzle(reg, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X); } static void transform_ABS(struct radeon_compiler* c, struct rc_instruction* inst) { - struct prog_src_register src = inst->I.SrcReg[0]; + struct rc_src_register src = inst->U.I.SrcReg[0]; src.Abs = 1; - src.Negate = NEGATE_NONE; - emit1(c, inst->Prev, OPCODE_MOV, inst->I.SaturateMode, inst->I.DstReg, src); + src.Negate = RC_MASK_NONE; + emit1(c, inst->Prev, RC_OPCODE_MOV, inst->U.I.SaturateMode, inst->U.I.DstReg, src); rc_remove_instruction(inst); } static void transform_DP3(struct radeon_compiler* c, struct rc_instruction* inst) { - struct prog_src_register src0 = inst->I.SrcReg[0]; - struct prog_src_register src1 = inst->I.SrcReg[1]; - src0.Negate &= ~NEGATE_W; + struct rc_src_register src0 = inst->U.I.SrcReg[0]; + struct rc_src_register src1 = inst->U.I.SrcReg[1]; + src0.Negate &= ~RC_MASK_W; src0.Swizzle &= ~(7 << (3 * 3)); - src0.Swizzle |= SWIZZLE_ZERO << (3 * 3); - src1.Negate &= ~NEGATE_W; + src0.Swizzle |= RC_SWIZZLE_ZERO << (3 * 3); + src1.Negate &= ~RC_MASK_W; src1.Swizzle &= ~(7 << (3 * 3)); - src1.Swizzle |= SWIZZLE_ZERO << (3 * 3); - emit2(c, inst->Prev, OPCODE_DP4, inst->I.SaturateMode, inst->I.DstReg, src0, src1); + src1.Swizzle |= RC_SWIZZLE_ZERO << (3 * 3); + emit2(c, inst->Prev, RC_OPCODE_DP4, inst->U.I.SaturateMode, inst->U.I.DstReg, src0, src1); rc_remove_instruction(inst); } static void transform_DPH(struct radeon_compiler* c, struct rc_instruction* inst) { - struct prog_src_register src0 = inst->I.SrcReg[0]; - src0.Negate &= ~NEGATE_W; + struct rc_src_register src0 = inst->U.I.SrcReg[0]; + src0.Negate &= ~RC_MASK_W; src0.Swizzle &= ~(7 << (3 * 3)); - src0.Swizzle |= SWIZZLE_ONE << (3 * 3); - emit2(c, inst->Prev, OPCODE_DP4, inst->I.SaturateMode, inst->I.DstReg, src0, inst->I.SrcReg[1]); + src0.Swizzle |= RC_SWIZZLE_ONE << (3 * 3); + emit2(c, inst->Prev, RC_OPCODE_DP4, inst->U.I.SaturateMode, inst->U.I.DstReg, src0, inst->U.I.SrcReg[1]); rc_remove_instruction(inst); } @@ -219,9 +208,9 @@ static void transform_DPH(struct radeon_compiler* c, static void transform_DST(struct radeon_compiler* c, struct rc_instruction* inst) { - emit2(c, inst->Prev, OPCODE_MUL, inst->I.SaturateMode, inst->I.DstReg, - swizzle(inst->I.SrcReg[0], SWIZZLE_ONE, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ONE), - swizzle(inst->I.SrcReg[1], SWIZZLE_ONE, SWIZZLE_Y, SWIZZLE_ONE, SWIZZLE_W)); + emit2(c, inst->Prev, RC_OPCODE_MUL, inst->U.I.SaturateMode, inst->U.I.DstReg, + swizzle(inst->U.I.SrcReg[0], RC_SWIZZLE_ONE, RC_SWIZZLE_Y, RC_SWIZZLE_Z, RC_SWIZZLE_ONE), + swizzle(inst->U.I.SrcReg[1], RC_SWIZZLE_ONE, RC_SWIZZLE_Y, RC_SWIZZLE_ONE, RC_SWIZZLE_W)); rc_remove_instruction(inst); } @@ -229,9 +218,9 @@ static void transform_FLR(struct radeon_compiler* c, struct rc_instruction* inst) { int tempreg = rc_find_free_temporary(c); - emit1(c, inst->Prev, OPCODE_FRC, 0, dstreg(PROGRAM_TEMPORARY, tempreg), inst->I.SrcReg[0]); - emit2(c, inst->Prev, OPCODE_ADD, inst->I.SaturateMode, inst->I.DstReg, - inst->I.SrcReg[0], negate(srcreg(PROGRAM_TEMPORARY, tempreg))); + emit1(c, inst->Prev, RC_OPCODE_FRC, 0, dstreg(RC_FILE_TEMPORARY, tempreg), inst->U.I.SrcReg[0]); + emit2(c, inst->Prev, RC_OPCODE_ADD, inst->U.I.SaturateMode, inst->U.I.DstReg, + inst->U.I.SrcReg[0], negate(srcreg(RC_FILE_TEMPORARY, tempreg))); rc_remove_instruction(inst); } @@ -256,64 +245,64 @@ static void transform_FLR(struct radeon_compiler* c, static void transform_LIT(struct radeon_compiler* c, struct rc_instruction* inst) { - GLuint constant; - GLuint constant_swizzle; - GLuint temp; - struct prog_src_register srctemp; + unsigned int constant; + unsigned int constant_swizzle; + unsigned int temp; + struct rc_src_register srctemp; constant = rc_constants_add_immediate_scalar(&c->Program.Constants, -127.999999, &constant_swizzle); - if (inst->I.DstReg.WriteMask != WRITEMASK_XYZW || inst->I.DstReg.File != PROGRAM_TEMPORARY) { + if (inst->U.I.DstReg.WriteMask != RC_MASK_XYZW || inst->U.I.DstReg.File != RC_FILE_TEMPORARY) { struct rc_instruction * inst_mov; inst_mov = emit1(c, inst, - OPCODE_MOV, 0, inst->I.DstReg, - srcreg(PROGRAM_TEMPORARY, rc_find_free_temporary(c))); + RC_OPCODE_MOV, 0, inst->U.I.DstReg, + srcreg(RC_FILE_TEMPORARY, rc_find_free_temporary(c))); - inst->I.DstReg.File = PROGRAM_TEMPORARY; - inst->I.DstReg.Index = inst_mov->I.SrcReg[0].Index; - inst->I.DstReg.WriteMask = WRITEMASK_XYZW; + inst->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst->U.I.DstReg.Index = inst_mov->U.I.SrcReg[0].Index; + inst->U.I.DstReg.WriteMask = RC_MASK_XYZW; } - temp = inst->I.DstReg.Index; - srctemp = srcreg(PROGRAM_TEMPORARY, temp); + temp = inst->U.I.DstReg.Index; + srctemp = srcreg(RC_FILE_TEMPORARY, temp); // tmp.x = max(0.0, Src.x); // tmp.y = max(0.0, Src.y); // tmp.w = clamp(Src.z, -128+eps, 128-eps); - emit2(c, inst->Prev, OPCODE_MAX, 0, - dstregtmpmask(temp, WRITEMASK_XYW), - inst->I.SrcReg[0], - swizzle(srcreg(PROGRAM_CONSTANT, constant), - SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ZERO, constant_swizzle&3)); - emit2(c, inst->Prev, OPCODE_MIN, 0, - dstregtmpmask(temp, WRITEMASK_Z), - swizzle(srctemp, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W), - negate(srcregswz(PROGRAM_CONSTANT, constant, constant_swizzle))); + emit2(c, inst->Prev, RC_OPCODE_MAX, 0, + dstregtmpmask(temp, RC_MASK_XYW), + inst->U.I.SrcReg[0], + swizzle(srcreg(RC_FILE_CONSTANT, constant), + RC_SWIZZLE_ZERO, RC_SWIZZLE_ZERO, RC_SWIZZLE_ZERO, constant_swizzle&3)); + emit2(c, inst->Prev, RC_OPCODE_MIN, 0, + dstregtmpmask(temp, RC_MASK_Z), + swizzle(srctemp, RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W), + negate(srcregswz(RC_FILE_CONSTANT, constant, constant_swizzle))); // tmp.w = Pow(tmp.y, tmp.w) - emit1(c, inst->Prev, OPCODE_LG2, 0, - dstregtmpmask(temp, WRITEMASK_W), - swizzle(srctemp, SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y)); - emit2(c, inst->Prev, OPCODE_MUL, 0, - dstregtmpmask(temp, WRITEMASK_W), - swizzle(srctemp, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W), - swizzle(srctemp, SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z)); - emit1(c, inst->Prev, OPCODE_EX2, 0, - dstregtmpmask(temp, WRITEMASK_W), - swizzle(srctemp, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W)); + emit1(c, inst->Prev, RC_OPCODE_LG2, 0, + dstregtmpmask(temp, RC_MASK_W), + swizzle(srctemp, RC_SWIZZLE_Y, RC_SWIZZLE_Y, RC_SWIZZLE_Y, RC_SWIZZLE_Y)); + emit2(c, inst->Prev, RC_OPCODE_MUL, 0, + dstregtmpmask(temp, RC_MASK_W), + swizzle(srctemp, RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W), + swizzle(srctemp, RC_SWIZZLE_Z, RC_SWIZZLE_Z, RC_SWIZZLE_Z, RC_SWIZZLE_Z)); + emit1(c, inst->Prev, RC_OPCODE_EX2, 0, + dstregtmpmask(temp, RC_MASK_W), + swizzle(srctemp, RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W)); // tmp.z = (tmp.x > 0) ? tmp.w : 0.0 - emit3(c, inst->Prev, OPCODE_CMP, inst->I.SaturateMode, - dstregtmpmask(temp, WRITEMASK_Z), - negate(swizzle(srctemp, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X)), - swizzle(srctemp, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W), + emit3(c, inst->Prev, RC_OPCODE_CMP, inst->U.I.SaturateMode, + dstregtmpmask(temp, RC_MASK_Z), + negate(swizzle(srctemp, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X)), + swizzle(srctemp, RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W), builtin_zero); // tmp.x, tmp.y, tmp.w = 1.0, tmp.x, 1.0 - emit1(c, inst->Prev, OPCODE_MOV, inst->I.SaturateMode, - dstregtmpmask(temp, WRITEMASK_XYW), - swizzle(srctemp, SWIZZLE_ONE, SWIZZLE_X, SWIZZLE_ONE, SWIZZLE_ONE)); + emit1(c, inst->Prev, RC_OPCODE_MOV, inst->U.I.SaturateMode, + dstregtmpmask(temp, RC_MASK_XYW), + swizzle(srctemp, RC_SWIZZLE_ONE, RC_SWIZZLE_X, RC_SWIZZLE_ONE, RC_SWIZZLE_ONE)); rc_remove_instruction(inst); } @@ -323,12 +312,12 @@ static void transform_LRP(struct radeon_compiler* c, { int tempreg = rc_find_free_temporary(c); - emit2(c, inst->Prev, OPCODE_ADD, 0, - dstreg(PROGRAM_TEMPORARY, tempreg), - inst->I.SrcReg[1], negate(inst->I.SrcReg[2])); - emit3(c, inst->Prev, OPCODE_MAD, inst->I.SaturateMode, - inst->I.DstReg, - inst->I.SrcReg[0], srcreg(PROGRAM_TEMPORARY, tempreg), inst->I.SrcReg[2]); + emit2(c, inst->Prev, RC_OPCODE_ADD, 0, + dstreg(RC_FILE_TEMPORARY, tempreg), + inst->U.I.SrcReg[1], negate(inst->U.I.SrcReg[2])); + emit3(c, inst->Prev, RC_OPCODE_MAD, inst->U.I.SaturateMode, + inst->U.I.DstReg, + inst->U.I.SrcReg[0], srcreg(RC_FILE_TEMPORARY, tempreg), inst->U.I.SrcReg[2]); rc_remove_instruction(inst); } @@ -337,14 +326,14 @@ static void transform_POW(struct radeon_compiler* c, struct rc_instruction* inst) { int tempreg = rc_find_free_temporary(c); - struct prog_dst_register tempdst = dstreg(PROGRAM_TEMPORARY, tempreg); - struct prog_src_register tempsrc = srcreg(PROGRAM_TEMPORARY, tempreg); - tempdst.WriteMask = WRITEMASK_W; - tempsrc.Swizzle = SWIZZLE_WWWW; + struct rc_dst_register tempdst = dstreg(RC_FILE_TEMPORARY, tempreg); + struct rc_src_register tempsrc = srcreg(RC_FILE_TEMPORARY, tempreg); + tempdst.WriteMask = RC_MASK_W; + tempsrc.Swizzle = RC_SWIZZLE_WWWW; - emit1(c, inst->Prev, OPCODE_LG2, 0, tempdst, scalar(inst->I.SrcReg[0])); - emit2(c, inst->Prev, OPCODE_MUL, 0, tempdst, tempsrc, scalar(inst->I.SrcReg[1])); - emit1(c, inst->Prev, OPCODE_EX2, inst->I.SaturateMode, inst->I.DstReg, tempsrc); + emit1(c, inst->Prev, RC_OPCODE_LG2, 0, tempdst, scalar(inst->U.I.SrcReg[0])); + emit2(c, inst->Prev, RC_OPCODE_MUL, 0, tempdst, tempsrc, scalar(inst->U.I.SrcReg[1])); + emit1(c, inst->Prev, RC_OPCODE_EX2, inst->U.I.SaturateMode, inst->U.I.DstReg, tempsrc); rc_remove_instruction(inst); } @@ -352,7 +341,26 @@ static void transform_POW(struct radeon_compiler* c, static void transform_RSQ(struct radeon_compiler* c, struct rc_instruction* inst) { - inst->I.SrcReg[0] = absolute(inst->I.SrcReg[0]); + inst->U.I.SrcReg[0] = absolute(inst->U.I.SrcReg[0]); +} + +static void transform_SEQ(struct radeon_compiler* c, + struct rc_instruction* inst) +{ + int tempreg = rc_find_free_temporary(c); + + emit2(c, inst->Prev, RC_OPCODE_ADD, 0, dstreg(RC_FILE_TEMPORARY, tempreg), inst->U.I.SrcReg[0], negate(inst->U.I.SrcReg[1])); + emit3(c, inst->Prev, RC_OPCODE_CMP, inst->U.I.SaturateMode, inst->U.I.DstReg, + negate(absolute(srcreg(RC_FILE_TEMPORARY, tempreg))), builtin_zero, builtin_one); + + rc_remove_instruction(inst); +} + +static void transform_SFL(struct radeon_compiler* c, + struct rc_instruction* inst) +{ + emit1(c, inst->Prev, RC_OPCODE_MOV, inst->U.I.SaturateMode, inst->U.I.DstReg, builtin_zero); + rc_remove_instruction(inst); } static void transform_SGE(struct radeon_compiler* c, @@ -360,9 +368,33 @@ static void transform_SGE(struct radeon_compiler* c, { int tempreg = rc_find_free_temporary(c); - emit2(c, inst->Prev, OPCODE_ADD, 0, dstreg(PROGRAM_TEMPORARY, tempreg), inst->I.SrcReg[0], negate(inst->I.SrcReg[1])); - emit3(c, inst->Prev, OPCODE_CMP, inst->I.SaturateMode, inst->I.DstReg, - srcreg(PROGRAM_TEMPORARY, tempreg), builtin_zero, builtin_one); + emit2(c, inst->Prev, RC_OPCODE_ADD, 0, dstreg(RC_FILE_TEMPORARY, tempreg), inst->U.I.SrcReg[0], negate(inst->U.I.SrcReg[1])); + emit3(c, inst->Prev, RC_OPCODE_CMP, inst->U.I.SaturateMode, inst->U.I.DstReg, + srcreg(RC_FILE_TEMPORARY, tempreg), builtin_zero, builtin_one); + + rc_remove_instruction(inst); +} + +static void transform_SGT(struct radeon_compiler* c, + struct rc_instruction* inst) +{ + int tempreg = rc_find_free_temporary(c); + + emit2(c, inst->Prev, RC_OPCODE_ADD, 0, dstreg(RC_FILE_TEMPORARY, tempreg), negate(inst->U.I.SrcReg[0]), inst->U.I.SrcReg[1]); + emit3(c, inst->Prev, RC_OPCODE_CMP, inst->U.I.SaturateMode, inst->U.I.DstReg, + srcreg(RC_FILE_TEMPORARY, tempreg), builtin_one, builtin_zero); + + rc_remove_instruction(inst); +} + +static void transform_SLE(struct radeon_compiler* c, + struct rc_instruction* inst) +{ + int tempreg = rc_find_free_temporary(c); + + emit2(c, inst->Prev, RC_OPCODE_ADD, 0, dstreg(RC_FILE_TEMPORARY, tempreg), negate(inst->U.I.SrcReg[0]), inst->U.I.SrcReg[1]); + emit3(c, inst->Prev, RC_OPCODE_CMP, inst->U.I.SaturateMode, inst->U.I.DstReg, + srcreg(RC_FILE_TEMPORARY, tempreg), builtin_zero, builtin_one); rc_remove_instruction(inst); } @@ -372,9 +404,21 @@ static void transform_SLT(struct radeon_compiler* c, { int tempreg = rc_find_free_temporary(c); - emit2(c, inst->Prev, OPCODE_ADD, 0, dstreg(PROGRAM_TEMPORARY, tempreg), inst->I.SrcReg[0], negate(inst->I.SrcReg[1])); - emit3(c, inst->Prev, OPCODE_CMP, inst->I.SaturateMode, inst->I.DstReg, - srcreg(PROGRAM_TEMPORARY, tempreg), builtin_one, builtin_zero); + emit2(c, inst->Prev, RC_OPCODE_ADD, 0, dstreg(RC_FILE_TEMPORARY, tempreg), inst->U.I.SrcReg[0], negate(inst->U.I.SrcReg[1])); + emit3(c, inst->Prev, RC_OPCODE_CMP, inst->U.I.SaturateMode, inst->U.I.DstReg, + srcreg(RC_FILE_TEMPORARY, tempreg), builtin_one, builtin_zero); + + rc_remove_instruction(inst); +} + +static void transform_SNE(struct radeon_compiler* c, + struct rc_instruction* inst) +{ + int tempreg = rc_find_free_temporary(c); + + emit2(c, inst->Prev, RC_OPCODE_ADD, 0, dstreg(RC_FILE_TEMPORARY, tempreg), inst->U.I.SrcReg[0], negate(inst->U.I.SrcReg[1])); + emit3(c, inst->Prev, RC_OPCODE_CMP, inst->U.I.SaturateMode, inst->U.I.DstReg, + negate(absolute(srcreg(RC_FILE_TEMPORARY, tempreg))), builtin_one, builtin_zero); rc_remove_instruction(inst); } @@ -382,14 +426,14 @@ static void transform_SLT(struct radeon_compiler* c, static void transform_SUB(struct radeon_compiler* c, struct rc_instruction* inst) { - inst->I.Opcode = OPCODE_ADD; - inst->I.SrcReg[1] = negate(inst->I.SrcReg[1]); + inst->U.I.Opcode = RC_OPCODE_ADD; + inst->U.I.SrcReg[1] = negate(inst->U.I.SrcReg[1]); } static void transform_SWZ(struct radeon_compiler* c, struct rc_instruction* inst) { - inst->I.Opcode = OPCODE_MOV; + inst->U.I.Opcode = RC_OPCODE_MOV; } static void transform_XPD(struct radeon_compiler* c, @@ -397,13 +441,13 @@ static void transform_XPD(struct radeon_compiler* c, { int tempreg = rc_find_free_temporary(c); - emit2(c, inst->Prev, OPCODE_MUL, 0, dstreg(PROGRAM_TEMPORARY, tempreg), - swizzle(inst->I.SrcReg[0], SWIZZLE_Z, SWIZZLE_X, SWIZZLE_Y, SWIZZLE_W), - swizzle(inst->I.SrcReg[1], SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_X, SWIZZLE_W)); - emit3(c, inst->Prev, OPCODE_MAD, inst->I.SaturateMode, inst->I.DstReg, - swizzle(inst->I.SrcReg[0], SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_X, SWIZZLE_W), - swizzle(inst->I.SrcReg[1], SWIZZLE_Z, SWIZZLE_X, SWIZZLE_Y, SWIZZLE_W), - negate(srcreg(PROGRAM_TEMPORARY, tempreg))); + emit2(c, inst->Prev, RC_OPCODE_MUL, 0, dstreg(RC_FILE_TEMPORARY, tempreg), + swizzle(inst->U.I.SrcReg[0], RC_SWIZZLE_Z, RC_SWIZZLE_X, RC_SWIZZLE_Y, RC_SWIZZLE_W), + swizzle(inst->U.I.SrcReg[1], RC_SWIZZLE_Y, RC_SWIZZLE_Z, RC_SWIZZLE_X, RC_SWIZZLE_W)); + emit3(c, inst->Prev, RC_OPCODE_MAD, inst->U.I.SaturateMode, inst->U.I.DstReg, + swizzle(inst->U.I.SrcReg[0], RC_SWIZZLE_Y, RC_SWIZZLE_Z, RC_SWIZZLE_X, RC_SWIZZLE_W), + swizzle(inst->U.I.SrcReg[1], RC_SWIZZLE_Z, RC_SWIZZLE_X, RC_SWIZZLE_Y, RC_SWIZZLE_W), + negate(srcreg(RC_FILE_TEMPORARY, tempreg))); rc_remove_instruction(inst); } @@ -414,7 +458,7 @@ static void transform_XPD(struct radeon_compiler* c, * no userData necessary. * * Eliminates the following ALU instructions: - * ABS, DPH, DST, FLR, LIT, LRP, POW, SGE, SLT, SUB, SWZ, XPD + * ABS, DPH, DST, FLR, LIT, LRP, POW, SEQ, SFL, SGE, SGT, SLE, SLT, SNE, SUB, SWZ, XPD * using: * MOV, ADD, MUL, MAD, FRC, DP3, LG2, EX2, CMP * @@ -423,27 +467,32 @@ static void transform_XPD(struct radeon_compiler* c, * * @note should be applicable to R300 and R500 fragment programs. */ -GLboolean radeonTransformALU( +int radeonTransformALU( struct radeon_compiler * c, struct rc_instruction* inst, void* unused) { - switch(inst->I.Opcode) { - case OPCODE_ABS: transform_ABS(c, inst); return GL_TRUE; - case OPCODE_DPH: transform_DPH(c, inst); return GL_TRUE; - case OPCODE_DST: transform_DST(c, inst); return GL_TRUE; - case OPCODE_FLR: transform_FLR(c, inst); return GL_TRUE; - case OPCODE_LIT: transform_LIT(c, inst); return GL_TRUE; - case OPCODE_LRP: transform_LRP(c, inst); return GL_TRUE; - case OPCODE_POW: transform_POW(c, inst); return GL_TRUE; - case OPCODE_RSQ: transform_RSQ(c, inst); return GL_TRUE; - case OPCODE_SGE: transform_SGE(c, inst); return GL_TRUE; - case OPCODE_SLT: transform_SLT(c, inst); return GL_TRUE; - case OPCODE_SUB: transform_SUB(c, inst); return GL_TRUE; - case OPCODE_SWZ: transform_SWZ(c, inst); return GL_TRUE; - case OPCODE_XPD: transform_XPD(c, inst); return GL_TRUE; + switch(inst->U.I.Opcode) { + case RC_OPCODE_ABS: transform_ABS(c, inst); return 1; + case RC_OPCODE_DPH: transform_DPH(c, inst); return 1; + case RC_OPCODE_DST: transform_DST(c, inst); return 1; + case RC_OPCODE_FLR: transform_FLR(c, inst); return 1; + case RC_OPCODE_LIT: transform_LIT(c, inst); return 1; + case RC_OPCODE_LRP: transform_LRP(c, inst); return 1; + case RC_OPCODE_POW: transform_POW(c, inst); return 1; + case RC_OPCODE_RSQ: transform_RSQ(c, inst); return 1; + case RC_OPCODE_SEQ: transform_SEQ(c, inst); return 1; + case RC_OPCODE_SFL: transform_SFL(c, inst); return 1; + case RC_OPCODE_SGE: transform_SGE(c, inst); return 1; + case RC_OPCODE_SGT: transform_SGT(c, inst); return 1; + case RC_OPCODE_SLE: transform_SLE(c, inst); return 1; + case RC_OPCODE_SLT: transform_SLT(c, inst); return 1; + case RC_OPCODE_SNE: transform_SNE(c, inst); return 1; + case RC_OPCODE_SUB: transform_SUB(c, inst); return 1; + case RC_OPCODE_SWZ: transform_SWZ(c, inst); return 1; + case RC_OPCODE_XPD: transform_XPD(c, inst); return 1; default: - return GL_FALSE; + return 0; } } @@ -452,37 +501,37 @@ static void transform_r300_vertex_ABS(struct radeon_compiler* c, struct rc_instruction* inst) { /* Note: r500 can take absolute values, but r300 cannot. */ - inst->I.Opcode = OPCODE_MAX; - inst->I.SrcReg[1] = inst->I.SrcReg[0]; - inst->I.SrcReg[1].Negate ^= NEGATE_XYZW; + inst->U.I.Opcode = RC_OPCODE_MAX; + inst->U.I.SrcReg[1] = inst->U.I.SrcReg[0]; + inst->U.I.SrcReg[1].Negate ^= RC_MASK_XYZW; } /** * For use with radeonLocalTransform, this transforms non-native ALU * instructions of the r300 up to r500 vertex engine. */ -GLboolean r300_transform_vertex_alu( +int r300_transform_vertex_alu( struct radeon_compiler * c, struct rc_instruction* inst, void* unused) { - switch(inst->I.Opcode) { - case OPCODE_ABS: transform_r300_vertex_ABS(c, inst); return GL_TRUE; - case OPCODE_DP3: transform_DP3(c, inst); return GL_TRUE; - case OPCODE_DPH: transform_DPH(c, inst); return GL_TRUE; - case OPCODE_FLR: transform_FLR(c, inst); return GL_TRUE; - case OPCODE_LRP: transform_LRP(c, inst); return GL_TRUE; - case OPCODE_SUB: transform_SUB(c, inst); return GL_TRUE; - case OPCODE_SWZ: transform_SWZ(c, inst); return GL_TRUE; - case OPCODE_XPD: transform_XPD(c, inst); return GL_TRUE; + switch(inst->U.I.Opcode) { + case RC_OPCODE_ABS: transform_r300_vertex_ABS(c, inst); return 1; + case RC_OPCODE_DP3: transform_DP3(c, inst); return 1; + case RC_OPCODE_DPH: transform_DPH(c, inst); return 1; + case RC_OPCODE_FLR: transform_FLR(c, inst); return 1; + case RC_OPCODE_LRP: transform_LRP(c, inst); return 1; + case RC_OPCODE_SUB: transform_SUB(c, inst); return 1; + case RC_OPCODE_SWZ: transform_SWZ(c, inst); return 1; + case RC_OPCODE_XPD: transform_XPD(c, inst); return 1; default: - return GL_FALSE; + return 0; } } -static void sincos_constants(struct radeon_compiler* c, GLuint *constants) +static void sincos_constants(struct radeon_compiler* c, unsigned int *constants) { - static const GLfloat SinCosConsts[2][4] = { + static const float SinCosConsts[2][4] = { { 1.273239545, // 4/PI -0.405284735, // -4/(PI*PI) @@ -511,26 +560,26 @@ static void sincos_constants(struct radeon_compiler* c, GLuint *constants) * MAD dest, tmp.y, weight, tmp.x */ static void sin_approx( - struct radeon_compiler* c, struct rc_instruction * after, - struct prog_dst_register dst, struct prog_src_register src, const GLuint* constants) -{ - GLuint tempreg = rc_find_free_temporary(c); - - emit2(c, after->Prev, OPCODE_MUL, 0, dstregtmpmask(tempreg, WRITEMASK_XY), - swizzle(src, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X), - srcreg(PROGRAM_CONSTANT, constants[0])); - emit3(c, after->Prev, OPCODE_MAD, 0, dstregtmpmask(tempreg, WRITEMASK_X), - swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y), - absolute(swizzle(src, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X)), - swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X)); - emit3(c, after->Prev, OPCODE_MAD, 0, dstregtmpmask(tempreg, WRITEMASK_Y), - swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X), - absolute(swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X)), - negate(swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X))); - emit3(c, after->Prev, OPCODE_MAD, 0, dst, - swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y), - swizzle(srcreg(PROGRAM_CONSTANT, constants[0]), SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W), - swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X)); + struct radeon_compiler* c, struct rc_instruction * inst, + struct rc_dst_register dst, struct rc_src_register src, const unsigned int* constants) +{ + unsigned int tempreg = rc_find_free_temporary(c); + + emit2(c, inst->Prev, RC_OPCODE_MUL, 0, dstregtmpmask(tempreg, RC_MASK_XY), + swizzle(src, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X), + srcreg(RC_FILE_CONSTANT, constants[0])); + emit3(c, inst->Prev, RC_OPCODE_MAD, 0, dstregtmpmask(tempreg, RC_MASK_X), + swizzle(srcreg(RC_FILE_TEMPORARY, tempreg), RC_SWIZZLE_Y, RC_SWIZZLE_Y, RC_SWIZZLE_Y, RC_SWIZZLE_Y), + absolute(swizzle(src, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X)), + swizzle(srcreg(RC_FILE_TEMPORARY, tempreg), RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X)); + emit3(c, inst->Prev, RC_OPCODE_MAD, 0, dstregtmpmask(tempreg, RC_MASK_Y), + swizzle(srcreg(RC_FILE_TEMPORARY, tempreg), RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X), + absolute(swizzle(srcreg(RC_FILE_TEMPORARY, tempreg), RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X)), + negate(swizzle(srcreg(RC_FILE_TEMPORARY, tempreg), RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X))); + emit3(c, inst->Prev, RC_OPCODE_MAD, 0, dst, + swizzle(srcreg(RC_FILE_TEMPORARY, tempreg), RC_SWIZZLE_Y, RC_SWIZZLE_Y, RC_SWIZZLE_Y, RC_SWIZZLE_Y), + swizzle(srcreg(RC_FILE_CONSTANT, constants[0]), RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W), + swizzle(srcreg(RC_FILE_TEMPORARY, tempreg), RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X)); } /** @@ -538,81 +587,81 @@ static void sin_approx( * using only the basic instructions * MOV, ADD, MUL, MAD, FRC */ -GLboolean radeonTransformTrigSimple(struct radeon_compiler* c, +int radeonTransformTrigSimple(struct radeon_compiler* c, struct rc_instruction* inst, void* unused) { - if (inst->I.Opcode != OPCODE_COS && - inst->I.Opcode != OPCODE_SIN && - inst->I.Opcode != OPCODE_SCS) - return GL_FALSE; + if (inst->U.I.Opcode != RC_OPCODE_COS && + inst->U.I.Opcode != RC_OPCODE_SIN && + inst->U.I.Opcode != RC_OPCODE_SCS) + return 0; - GLuint constants[2]; - GLuint tempreg = rc_find_free_temporary(c); + unsigned int constants[2]; + unsigned int tempreg = rc_find_free_temporary(c); sincos_constants(c, constants); - if (inst->I.Opcode == OPCODE_COS) { + if (inst->U.I.Opcode == RC_OPCODE_COS) { // MAD tmp.x, src, 1/(2*PI), 0.75 // FRC tmp.x, tmp.x // MAD tmp.z, tmp.x, 2*PI, -PI - emit3(c, inst->Prev, OPCODE_MAD, 0, dstregtmpmask(tempreg, WRITEMASK_W), - swizzle(inst->I.SrcReg[0], SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X), - swizzle(srcreg(PROGRAM_CONSTANT, constants[1]), SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z), - swizzle(srcreg(PROGRAM_CONSTANT, constants[1]), SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X)); - emit1(c, inst->Prev, OPCODE_FRC, 0, dstregtmpmask(tempreg, WRITEMASK_W), - swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W)); - emit3(c, inst->Prev, OPCODE_MAD, 0, dstregtmpmask(tempreg, WRITEMASK_W), - swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W), - swizzle(srcreg(PROGRAM_CONSTANT, constants[1]), SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W), - negate(swizzle(srcreg(PROGRAM_CONSTANT, constants[0]), SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z))); - - sin_approx(c, inst->Prev, inst->I.DstReg, - swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W), + emit3(c, inst->Prev, RC_OPCODE_MAD, 0, dstregtmpmask(tempreg, RC_MASK_W), + swizzle(inst->U.I.SrcReg[0], RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X), + swizzle(srcreg(RC_FILE_CONSTANT, constants[1]), RC_SWIZZLE_Z, RC_SWIZZLE_Z, RC_SWIZZLE_Z, RC_SWIZZLE_Z), + swizzle(srcreg(RC_FILE_CONSTANT, constants[1]), RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X)); + emit1(c, inst->Prev, RC_OPCODE_FRC, 0, dstregtmpmask(tempreg, RC_MASK_W), + swizzle(srcreg(RC_FILE_TEMPORARY, tempreg), RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W)); + emit3(c, inst->Prev, RC_OPCODE_MAD, 0, dstregtmpmask(tempreg, RC_MASK_W), + swizzle(srcreg(RC_FILE_TEMPORARY, tempreg), RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W), + swizzle(srcreg(RC_FILE_CONSTANT, constants[1]), RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W), + negate(swizzle(srcreg(RC_FILE_CONSTANT, constants[0]), RC_SWIZZLE_Z, RC_SWIZZLE_Z, RC_SWIZZLE_Z, RC_SWIZZLE_Z))); + + sin_approx(c, inst, inst->U.I.DstReg, + swizzle(srcreg(RC_FILE_TEMPORARY, tempreg), RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W), constants); - } else if (inst->I.Opcode == OPCODE_SIN) { - emit3(c, inst->Prev, OPCODE_MAD, 0, dstregtmpmask(tempreg, WRITEMASK_W), - swizzle(inst->I.SrcReg[0], SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X), - swizzle(srcreg(PROGRAM_CONSTANT, constants[1]), SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z), - swizzle(srcreg(PROGRAM_CONSTANT, constants[1]), SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y)); - emit1(c, inst->Prev, OPCODE_FRC, 0, dstregtmpmask(tempreg, WRITEMASK_W), - swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W)); - emit3(c, inst->Prev, OPCODE_MAD, 0, dstregtmpmask(tempreg, WRITEMASK_W), - swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W), - swizzle(srcreg(PROGRAM_CONSTANT, constants[1]), SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W), - negate(swizzle(srcreg(PROGRAM_CONSTANT, constants[0]), SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z))); - - sin_approx(c, inst->Prev, inst->I.DstReg, - swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W), + } else if (inst->U.I.Opcode == RC_OPCODE_SIN) { + emit3(c, inst->Prev, RC_OPCODE_MAD, 0, dstregtmpmask(tempreg, RC_MASK_W), + swizzle(inst->U.I.SrcReg[0], RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X), + swizzle(srcreg(RC_FILE_CONSTANT, constants[1]), RC_SWIZZLE_Z, RC_SWIZZLE_Z, RC_SWIZZLE_Z, RC_SWIZZLE_Z), + swizzle(srcreg(RC_FILE_CONSTANT, constants[1]), RC_SWIZZLE_Y, RC_SWIZZLE_Y, RC_SWIZZLE_Y, RC_SWIZZLE_Y)); + emit1(c, inst->Prev, RC_OPCODE_FRC, 0, dstregtmpmask(tempreg, RC_MASK_W), + swizzle(srcreg(RC_FILE_TEMPORARY, tempreg), RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W)); + emit3(c, inst->Prev, RC_OPCODE_MAD, 0, dstregtmpmask(tempreg, RC_MASK_W), + swizzle(srcreg(RC_FILE_TEMPORARY, tempreg), RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W), + swizzle(srcreg(RC_FILE_CONSTANT, constants[1]), RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W), + negate(swizzle(srcreg(RC_FILE_CONSTANT, constants[0]), RC_SWIZZLE_Z, RC_SWIZZLE_Z, RC_SWIZZLE_Z, RC_SWIZZLE_Z))); + + sin_approx(c, inst, inst->U.I.DstReg, + swizzle(srcreg(RC_FILE_TEMPORARY, tempreg), RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W), constants); } else { - emit3(c, inst->Prev, OPCODE_MAD, 0, dstregtmpmask(tempreg, WRITEMASK_XY), - swizzle(inst->I.SrcReg[0], SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X), - swizzle(srcreg(PROGRAM_CONSTANT, constants[1]), SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z), - swizzle(srcreg(PROGRAM_CONSTANT, constants[1]), SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_W)); - emit1(c, inst->Prev, OPCODE_FRC, 0, dstregtmpmask(tempreg, WRITEMASK_XY), - srcreg(PROGRAM_TEMPORARY, tempreg)); - emit3(c, inst->Prev, OPCODE_MAD, 0, dstregtmpmask(tempreg, WRITEMASK_XY), - srcreg(PROGRAM_TEMPORARY, tempreg), - swizzle(srcreg(PROGRAM_CONSTANT, constants[1]), SWIZZLE_W, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W), - negate(swizzle(srcreg(PROGRAM_CONSTANT, constants[0]), SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z, SWIZZLE_Z))); - - struct prog_dst_register dst = inst->I.DstReg; - - dst.WriteMask = inst->I.DstReg.WriteMask & WRITEMASK_X; - sin_approx(c, inst->Prev, dst, - swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X), + emit3(c, inst->Prev, RC_OPCODE_MAD, 0, dstregtmpmask(tempreg, RC_MASK_XY), + swizzle(inst->U.I.SrcReg[0], RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X), + swizzle(srcreg(RC_FILE_CONSTANT, constants[1]), RC_SWIZZLE_Z, RC_SWIZZLE_Z, RC_SWIZZLE_Z, RC_SWIZZLE_Z), + swizzle(srcreg(RC_FILE_CONSTANT, constants[1]), RC_SWIZZLE_X, RC_SWIZZLE_Y, RC_SWIZZLE_Z, RC_SWIZZLE_W)); + emit1(c, inst->Prev, RC_OPCODE_FRC, 0, dstregtmpmask(tempreg, RC_MASK_XY), + srcreg(RC_FILE_TEMPORARY, tempreg)); + emit3(c, inst->Prev, RC_OPCODE_MAD, 0, dstregtmpmask(tempreg, RC_MASK_XY), + srcreg(RC_FILE_TEMPORARY, tempreg), + swizzle(srcreg(RC_FILE_CONSTANT, constants[1]), RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W, RC_SWIZZLE_W), + negate(swizzle(srcreg(RC_FILE_CONSTANT, constants[0]), RC_SWIZZLE_Z, RC_SWIZZLE_Z, RC_SWIZZLE_Z, RC_SWIZZLE_Z))); + + struct rc_dst_register dst = inst->U.I.DstReg; + + dst.WriteMask = inst->U.I.DstReg.WriteMask & RC_MASK_X; + sin_approx(c, inst, dst, + swizzle(srcreg(RC_FILE_TEMPORARY, tempreg), RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X), constants); - dst.WriteMask = inst->I.DstReg.WriteMask & WRITEMASK_Y; - sin_approx(c, inst->Prev, dst, - swizzle(srcreg(PROGRAM_TEMPORARY, tempreg), SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y), + dst.WriteMask = inst->U.I.DstReg.WriteMask & RC_MASK_Y; + sin_approx(c, inst, dst, + swizzle(srcreg(RC_FILE_TEMPORARY, tempreg), RC_SWIZZLE_Y, RC_SWIZZLE_Y, RC_SWIZZLE_Y, RC_SWIZZLE_Y), constants); } rc_remove_instruction(inst); - return GL_TRUE; + return 1; } @@ -624,53 +673,53 @@ GLboolean radeonTransformTrigSimple(struct radeon_compiler* c, * * @warning This transformation implicitly changes the semantics of SIN and COS! */ -GLboolean radeonTransformTrigScale(struct radeon_compiler* c, +int radeonTransformTrigScale(struct radeon_compiler* c, struct rc_instruction* inst, void* unused) { - if (inst->I.Opcode != OPCODE_COS && - inst->I.Opcode != OPCODE_SIN && - inst->I.Opcode != OPCODE_SCS) - return GL_FALSE; + if (inst->U.I.Opcode != RC_OPCODE_COS && + inst->U.I.Opcode != RC_OPCODE_SIN && + inst->U.I.Opcode != RC_OPCODE_SCS) + return 0; - static const GLfloat RCP_2PI = 0.15915494309189535; - GLuint temp; - GLuint constant; - GLuint constant_swizzle; + static const float RCP_2PI = 0.15915494309189535; + unsigned int temp; + unsigned int constant; + unsigned int constant_swizzle; temp = rc_find_free_temporary(c); constant = rc_constants_add_immediate_scalar(&c->Program.Constants, RCP_2PI, &constant_swizzle); - emit2(c, inst->Prev, OPCODE_MUL, 0, dstregtmpmask(temp, WRITEMASK_W), - swizzle(inst->I.SrcReg[0], SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X), - srcregswz(PROGRAM_CONSTANT, constant, constant_swizzle)); - emit1(c, inst->Prev, OPCODE_FRC, 0, dstregtmpmask(temp, WRITEMASK_W), - srcreg(PROGRAM_TEMPORARY, temp)); - - if (inst->I.Opcode == OPCODE_COS) { - emit1(c, inst->Prev, OPCODE_COS, inst->I.SaturateMode, inst->I.DstReg, - srcregswz(PROGRAM_TEMPORARY, temp, SWIZZLE_WWWW)); - } else if (inst->I.Opcode == OPCODE_SIN) { - emit1(c, inst->Prev, OPCODE_SIN, inst->I.SaturateMode, - inst->I.DstReg, srcregswz(PROGRAM_TEMPORARY, temp, SWIZZLE_WWWW)); - } else if (inst->I.Opcode == OPCODE_SCS) { - struct prog_dst_register moddst = inst->I.DstReg; - - if (inst->I.DstReg.WriteMask & WRITEMASK_X) { - moddst.WriteMask = WRITEMASK_X; - emit1(c, inst->Prev, OPCODE_COS, inst->I.SaturateMode, moddst, - srcregswz(PROGRAM_TEMPORARY, temp, SWIZZLE_WWWW)); + emit2(c, inst->Prev, RC_OPCODE_MUL, 0, dstregtmpmask(temp, RC_MASK_W), + swizzle(inst->U.I.SrcReg[0], RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X, RC_SWIZZLE_X), + srcregswz(RC_FILE_CONSTANT, constant, constant_swizzle)); + emit1(c, inst->Prev, RC_OPCODE_FRC, 0, dstregtmpmask(temp, RC_MASK_W), + srcreg(RC_FILE_TEMPORARY, temp)); + + if (inst->U.I.Opcode == RC_OPCODE_COS) { + emit1(c, inst->Prev, RC_OPCODE_COS, inst->U.I.SaturateMode, inst->U.I.DstReg, + srcregswz(RC_FILE_TEMPORARY, temp, RC_SWIZZLE_WWWW)); + } else if (inst->U.I.Opcode == RC_OPCODE_SIN) { + emit1(c, inst->Prev, RC_OPCODE_SIN, inst->U.I.SaturateMode, + inst->U.I.DstReg, srcregswz(RC_FILE_TEMPORARY, temp, RC_SWIZZLE_WWWW)); + } else if (inst->U.I.Opcode == RC_OPCODE_SCS) { + struct rc_dst_register moddst = inst->U.I.DstReg; + + if (inst->U.I.DstReg.WriteMask & RC_MASK_X) { + moddst.WriteMask = RC_MASK_X; + emit1(c, inst->Prev, RC_OPCODE_COS, inst->U.I.SaturateMode, moddst, + srcregswz(RC_FILE_TEMPORARY, temp, RC_SWIZZLE_WWWW)); } - if (inst->I.DstReg.WriteMask & WRITEMASK_Y) { - moddst.WriteMask = WRITEMASK_Y; - emit1(c, inst->Prev, OPCODE_SIN, inst->I.SaturateMode, moddst, - srcregswz(PROGRAM_TEMPORARY, temp, SWIZZLE_WWWW)); + if (inst->U.I.DstReg.WriteMask & RC_MASK_Y) { + moddst.WriteMask = RC_MASK_Y; + emit1(c, inst->Prev, RC_OPCODE_SIN, inst->U.I.SaturateMode, moddst, + srcregswz(RC_FILE_TEMPORARY, temp, RC_SWIZZLE_WWWW)); } } rc_remove_instruction(inst); - return GL_TRUE; + return 1; } /** @@ -681,15 +730,15 @@ GLboolean radeonTransformTrigScale(struct radeon_compiler* c, * @warning This explicitly changes the form of DDX and DDY! */ -GLboolean radeonTransformDeriv(struct radeon_compiler* c, +int radeonTransformDeriv(struct radeon_compiler* c, struct rc_instruction* inst, void* unused) { - if (inst->I.Opcode != OPCODE_DDX && inst->I.Opcode != OPCODE_DDY) - return GL_FALSE; + if (inst->U.I.Opcode != RC_OPCODE_DDX && inst->U.I.Opcode != RC_OPCODE_DDY) + return 0; - inst->I.SrcReg[1].Swizzle = MAKE_SWIZZLE4(SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE); - inst->I.SrcReg[1].Negate = NEGATE_XYZW; + inst->U.I.SrcReg[1].Swizzle = RC_MAKE_SWIZZLE(RC_SWIZZLE_ONE, RC_SWIZZLE_ONE, RC_SWIZZLE_ONE, RC_SWIZZLE_ONE); + inst->U.I.SrcReg[1].Negate = RC_MASK_XYZW; - return GL_TRUE; + return 1; } diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.h b/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.h index 147efec6fcb..7cb5f84b7f0 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.h +++ b/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.h @@ -30,27 +30,27 @@ #include "radeon_program.h" -GLboolean radeonTransformALU( +int radeonTransformALU( struct radeon_compiler * c, struct rc_instruction * inst, void*); -GLboolean r300_transform_vertex_alu( +int r300_transform_vertex_alu( struct radeon_compiler * c, struct rc_instruction * inst, void*); -GLboolean radeonTransformTrigSimple( +int radeonTransformTrigSimple( struct radeon_compiler * c, struct rc_instruction * inst, void*); -GLboolean radeonTransformTrigScale( +int radeonTransformTrigScale( struct radeon_compiler * c, struct rc_instruction * inst, void*); -GLboolean radeonTransformDeriv( +int radeonTransformDeriv( struct radeon_compiler * c, struct rc_instruction * inst, void*); diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_program_constants.h b/src/mesa/drivers/dri/r300/compiler/radeon_program_constants.h new file mode 100644 index 00000000000..7c0d6720b11 --- /dev/null +++ b/src/mesa/drivers/dri/r300/compiler/radeon_program_constants.h @@ -0,0 +1,147 @@ +/* + * Copyright (C) 2009 Nicolai Haehnle. + * + * 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, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef RADEON_PROGRAM_CONSTANTS_H +#define RADEON_PROGRAM_CONSTANTS_H + +typedef enum { + RC_SATURATE_NONE = 0, + RC_SATURATE_ZERO_ONE, + RC_SATURATE_MINUS_PLUS_ONE +} rc_saturate_mode; + +typedef enum { + RC_TEXTURE_2D_ARRAY, + RC_TEXTURE_1D_ARRAY, + RC_TEXTURE_CUBE, + RC_TEXTURE_3D, + RC_TEXTURE_RECT, + RC_TEXTURE_2D, + RC_TEXTURE_1D +} rc_texture_target; + +typedef enum { + /** + * Used to indicate unused register descriptions and + * source register that use a constant swizzle. + */ + RC_FILE_NONE = 0, + RC_FILE_TEMPORARY, + + /** + * Input register. + * + * \note The compiler attaches no implicit semantics to input registers. + * Fragment/vertex program specific semantics must be defined explicitly + * using the appropriate compiler interfaces. + */ + RC_FILE_INPUT, + + /** + * Output register. + * + * \note The compiler attaches no implicit semantics to input registers. + * Fragment/vertex program specific semantics must be defined explicitly + * using the appropriate compiler interfaces. + */ + RC_FILE_OUTPUT, + RC_FILE_ADDRESS, + + /** + * Indicates a constant from the \ref rc_constant_list . + */ + RC_FILE_CONSTANT, + + /** + * Indicates a special register, see RC_SPECIAL_xxx. + */ + RC_FILE_SPECIAL +} rc_register_file; + +enum { + /** R500 fragment program ALU result "register" */ + RC_SPECIAL_ALU_RESULT = 0, + + /** Must be last */ + RC_NUM_SPECIAL_REGISTERS +}; + +#define RC_REGISTER_INDEX_BITS 10 +#define RC_REGISTER_MAX_INDEX (1 << RC_REGISTER_INDEX_BITS) + +typedef enum { + RC_SWIZZLE_X = 0, + RC_SWIZZLE_Y, + RC_SWIZZLE_Z, + RC_SWIZZLE_W, + RC_SWIZZLE_ZERO, + RC_SWIZZLE_ONE, + RC_SWIZZLE_HALF, + RC_SWIZZLE_UNUSED +} rc_swizzle; + +#define RC_MAKE_SWIZZLE(a,b,c,d) (((a)<<0) | ((b)<<3) | ((c)<<6) | ((d)<<9)) +#define RC_MAKE_SWIZZLE_SMEAR(a) RC_MAKE_SWIZZLE((a),(a),(a),(a)) +#define GET_SWZ(swz, idx) (((swz) >> ((idx)*3)) & 0x7) +#define GET_BIT(msk, idx) (((msk) >> (idx)) & 0x1) +#define SET_SWZ(swz, idx, newv) \ + do { \ + (swz) = ((swz) & ~(7 << ((idx)*3))) | ((newv) << ((idx)*3)); \ + } while(0) + +#define RC_SWIZZLE_XYZW RC_MAKE_SWIZZLE(RC_SWIZZLE_X, RC_SWIZZLE_Y, RC_SWIZZLE_Z, RC_SWIZZLE_W) +#define RC_SWIZZLE_XXXX RC_MAKE_SWIZZLE_SMEAR(RC_SWIZZLE_X) +#define RC_SWIZZLE_YYYY RC_MAKE_SWIZZLE_SMEAR(RC_SWIZZLE_Y) +#define RC_SWIZZLE_ZZZZ RC_MAKE_SWIZZLE_SMEAR(RC_SWIZZLE_Z) +#define RC_SWIZZLE_WWWW RC_MAKE_SWIZZLE_SMEAR(RC_SWIZZLE_W) +#define RC_SWIZZLE_0000 RC_MAKE_SWIZZLE_SMEAR(RC_SWIZZLE_ZERO) +#define RC_SWIZZLE_1111 RC_MAKE_SWIZZLE_SMEAR(RC_SWIZZLE_ONE) + +/** + * \name Bitmasks for components of vectors. + * + * Used for write masks, negation masks, etc. + */ +/*@{*/ +#define RC_MASK_NONE 0 +#define RC_MASK_X 1 +#define RC_MASK_Y 2 +#define RC_MASK_Z 4 +#define RC_MASK_W 8 +#define RC_MASK_XY (RC_MASK_X|RC_MASK_Y) +#define RC_MASK_XYZ (RC_MASK_X|RC_MASK_Y|RC_MASK_Z) +#define RC_MASK_XYW (RC_MASK_X|RC_MASK_Y|RC_MASK_W) +#define RC_MASK_XYZW (RC_MASK_X|RC_MASK_Y|RC_MASK_Z|RC_MASK_W) +/*@}*/ + +typedef enum { + RC_ALURESULT_NONE = 0, + RC_ALURESULT_X, + RC_ALURESULT_W +} rc_write_aluresult; + +#endif /* RADEON_PROGRAM_CONSTANTS_H */ diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_program_pair.c b/src/mesa/drivers/dri/r300/compiler/radeon_program_pair.c index 4c26db5d244..ee839596aab 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_program_pair.c +++ b/src/mesa/drivers/dri/r300/compiler/radeon_program_pair.c @@ -1,5 +1,5 @@ /* - * Copyright (C) 2008 Nicolai Haehnle. + * Copyright (C) 2008-2009 Nicolai Haehnle. * * All Rights Reserved. * @@ -25,584 +25,29 @@ * */ -/** - * @file - * - * Perform temporary register allocation and attempt to pair off instructions - * in RGB and Alpha pairs. Also attempts to optimize the TEX instruction - * vs. ALU instruction scheduling. - */ - #include "radeon_program_pair.h" -#include "memory_pool.h" -#include "radeon_compiler.h" -#include "shader/prog_print.h" - -#define error(fmt, args...) do { \ - rc_error(&s->Compiler->Base, "%s::%s(): " fmt "\n", \ - __FILE__, __FUNCTION__, ##args); \ -} while(0) - -struct pair_state_instruction { - struct prog_instruction Instruction; - GLuint IP; /**< Position of this instruction in original program */ - - GLuint IsTex:1; /**< Is a texture instruction */ - GLuint NeedRGB:1; /**< Needs the RGB ALU */ - GLuint NeedAlpha:1; /**< Needs the Alpha ALU */ - GLuint IsTranscendent:1; /**< Is a special transcendent instruction */ - - /** - * Number of (read and write) dependencies that must be resolved before - * this instruction can be scheduled. - */ - GLuint NumDependencies:5; - - /** - * Next instruction in the linked list of ready instructions. - */ - struct pair_state_instruction *NextReady; - - /** - * Values that this instruction writes - */ - struct reg_value *Values[4]; -}; - - -/** - * Used to keep track of which instructions read a value. - */ -struct reg_value_reader { - struct pair_state_instruction *Reader; - struct reg_value_reader *Next; -}; - -/** - * Used to keep track which values are stored in each component of a - * PROGRAM_TEMPORARY. - */ -struct reg_value { - struct pair_state_instruction *Writer; - struct reg_value *Next; /**< Pointer to the next value to be written to the same PROGRAM_TEMPORARY component */ - - /** - * Unordered linked list of instructions that read from this value. - */ - struct reg_value_reader *Readers; - - /** - * Number of readers of this value. This is calculated during @ref scan_instructions - * and continually decremented during code emission. - * When this count reaches zero, the instruction that writes the @ref Next value - * can be scheduled. - */ - GLuint NumReaders; -}; - -/** - * Used to translate a PROGRAM_INPUT or PROGRAM_TEMPORARY Mesa register - * to the proper hardware temporary. - */ -struct pair_register_translation { - GLuint Allocated:1; - GLuint HwIndex:8; - GLuint RefCount:23; /**< # of times this occurs in an unscheduled instruction SrcReg or DstReg */ - - /** - * Notes the value that is currently contained in each component - * (only used for PROGRAM_TEMPORARY registers). - */ - struct reg_value *Value[4]; -}; - -struct pair_state { - struct r300_fragment_program_compiler * Compiler; - const struct radeon_pair_handler *Handler; - GLboolean Verbose; - void *UserData; - - /** - * Translate Mesa registers to hardware registers - */ - struct pair_register_translation Inputs[FRAG_ATTRIB_MAX]; - struct pair_register_translation Temps[MAX_PROGRAM_TEMPS]; - - struct { - GLuint RefCount; /**< # of times this occurs in an unscheduled SrcReg or DstReg */ - } HwTemps[128]; - - /** - * Linked list of instructions that can be scheduled right now, - * based on which ALU/TEX resources they require. - */ - struct pair_state_instruction *ReadyFullALU; - struct pair_state_instruction *ReadyRGB; - struct pair_state_instruction *ReadyAlpha; - struct pair_state_instruction *ReadyTEX; -}; - - -static struct pair_register_translation *get_register(struct pair_state *s, GLuint file, GLuint index) -{ - switch(file) { - case PROGRAM_TEMPORARY: return &s->Temps[index]; - case PROGRAM_INPUT: return &s->Inputs[index]; - default: return 0; - } -} - -static void alloc_hw_reg(struct pair_state *s, GLuint file, GLuint index, GLuint hwindex) -{ - struct pair_register_translation *t = get_register(s, file, index); - ASSERT(!s->HwTemps[hwindex].RefCount); - ASSERT(!t->Allocated); - s->HwTemps[hwindex].RefCount = t->RefCount; - t->Allocated = 1; - t->HwIndex = hwindex; -} - -static GLuint get_hw_reg(struct pair_state *s, GLuint file, GLuint index) -{ - GLuint hwindex; - - struct pair_register_translation *t = get_register(s, file, index); - if (!t) { - error("get_hw_reg: %i[%i]\n", file, index); - return 0; - } - - if (t->Allocated) - return t->HwIndex; - - for(hwindex = 0; hwindex < s->Handler->MaxHwTemps; ++hwindex) - if (!s->HwTemps[hwindex].RefCount) - break; - - if (hwindex >= s->Handler->MaxHwTemps) { - error("Ran out of hardware temporaries"); - return 0; - } - - alloc_hw_reg(s, file, index, hwindex); - return hwindex; -} - - -static void deref_hw_reg(struct pair_state *s, GLuint hwindex) -{ - if (!s->HwTemps[hwindex].RefCount) { - error("Hwindex %i refcount error", hwindex); - return; - } - - s->HwTemps[hwindex].RefCount--; -} - -static void add_pairinst_to_list(struct pair_state_instruction **list, struct pair_state_instruction *pairinst) -{ - pairinst->NextReady = *list; - *list = pairinst; -} - -/** - * The given instruction has become ready. Link it into the ready - * instructions. - */ -static void instruction_ready(struct pair_state *s, struct pair_state_instruction *pairinst) -{ - if (s->Verbose) - _mesa_printf("instruction_ready(%i)\n", pairinst->IP); - - if (pairinst->IsTex) - add_pairinst_to_list(&s->ReadyTEX, pairinst); - else if (!pairinst->NeedAlpha) - add_pairinst_to_list(&s->ReadyRGB, pairinst); - else if (!pairinst->NeedRGB) - add_pairinst_to_list(&s->ReadyAlpha, pairinst); - else - add_pairinst_to_list(&s->ReadyFullALU, pairinst); -} - - -/** - * Finally rewrite ADD, MOV, MUL as the appropriate native instruction - * and reverse the order of arguments for CMP. - */ -static void final_rewrite(struct pair_state *s, struct prog_instruction *inst) -{ - struct prog_src_register tmp; - - switch(inst->Opcode) { - case OPCODE_ADD: - inst->SrcReg[2] = inst->SrcReg[1]; - inst->SrcReg[1].File = PROGRAM_BUILTIN; - inst->SrcReg[1].Swizzle = SWIZZLE_1111; - inst->SrcReg[1].Negate = NEGATE_NONE; - inst->Opcode = OPCODE_MAD; - break; - case OPCODE_CMP: - tmp = inst->SrcReg[2]; - inst->SrcReg[2] = inst->SrcReg[0]; - inst->SrcReg[0] = tmp; - break; - case OPCODE_MOV: - /* AMD say we should use CMP. - * However, when we transform - * KIL -r0; - * into - * CMP tmp, -r0, -r0, 0; - * KIL tmp; - * we get incorrect behaviour on R500 when r0 == 0.0. - * It appears that the R500 KIL hardware treats -0.0 as less - * than zero. - */ - inst->SrcReg[1].File = PROGRAM_BUILTIN; - inst->SrcReg[1].Swizzle = SWIZZLE_1111; - inst->SrcReg[2].File = PROGRAM_BUILTIN; - inst->SrcReg[2].Swizzle = SWIZZLE_0000; - inst->Opcode = OPCODE_MAD; - break; - case OPCODE_MUL: - inst->SrcReg[2].File = PROGRAM_BUILTIN; - inst->SrcReg[2].Swizzle = SWIZZLE_0000; - inst->Opcode = OPCODE_MAD; - break; - default: - /* nothing to do */ - break; - } -} - - -/** - * Classify an instruction according to which ALUs etc. it needs - */ -static void classify_instruction(struct pair_state *s, - struct pair_state_instruction *psi) -{ - psi->NeedRGB = (psi->Instruction.DstReg.WriteMask & WRITEMASK_XYZ) ? 1 : 0; - psi->NeedAlpha = (psi->Instruction.DstReg.WriteMask & WRITEMASK_W) ? 1 : 0; - - switch(psi->Instruction.Opcode) { - case OPCODE_ADD: - case OPCODE_CMP: - case OPCODE_DDX: - case OPCODE_DDY: - case OPCODE_FRC: - case OPCODE_MAD: - case OPCODE_MAX: - case OPCODE_MIN: - case OPCODE_MOV: - case OPCODE_MUL: - break; - case OPCODE_COS: - case OPCODE_EX2: - case OPCODE_LG2: - case OPCODE_RCP: - case OPCODE_RSQ: - case OPCODE_SIN: - psi->IsTranscendent = 1; - psi->NeedAlpha = 1; - break; - case OPCODE_DP4: - psi->NeedAlpha = 1; - /* fall through */ - case OPCODE_DP3: - psi->NeedRGB = 1; - break; - case OPCODE_KIL: - case OPCODE_TEX: - case OPCODE_TXB: - case OPCODE_TXP: - case OPCODE_END: - psi->IsTex = 1; - break; - default: - error("Unknown opcode %d\n", psi->Instruction.Opcode); - break; - } -} - - -/** - * Count which (input, temporary) register is read and written how often, - * and scan the instruction stream to find dependencies. - */ -static void scan_instructions(struct pair_state *s) -{ - struct rc_instruction *source; - GLuint ip; - - for(source = s->Compiler->Base.Program.Instructions.Next, ip = 0; - source != &s->Compiler->Base.Program.Instructions; - source = source->Next, ++ip) { - struct pair_state_instruction *pairinst = memory_pool_malloc(&s->Compiler->Base.Pool, sizeof(*pairinst)); - memset(pairinst, 0, sizeof(struct pair_state_instruction)); - - pairinst->Instruction = source->I; - pairinst->IP = ip; - final_rewrite(s, &pairinst->Instruction); - classify_instruction(s, pairinst); - - int nsrc = _mesa_num_inst_src_regs(pairinst->Instruction.Opcode); - int j; - for(j = 0; j < nsrc; j++) { - struct pair_register_translation *t = - get_register(s, pairinst->Instruction.SrcReg[j].File, pairinst->Instruction.SrcReg[j].Index); - if (!t) - continue; - - t->RefCount++; - - if (pairinst->Instruction.SrcReg[j].File == PROGRAM_TEMPORARY) { - int i; - for(i = 0; i < 4; ++i) { - GLuint swz = GET_SWZ(pairinst->Instruction.SrcReg[j].Swizzle, i); - if (swz >= 4) - continue; /* constant or NIL swizzle */ - if (!t->Value[swz]) - continue; /* this is an undefined read */ - - /* Do not add a dependency if this instruction - * also rewrites the value. The code below adds - * a dependency for the DstReg, which is a superset - * of the SrcReg dependency. */ - if (pairinst->Instruction.DstReg.File == PROGRAM_TEMPORARY && - pairinst->Instruction.DstReg.Index == pairinst->Instruction.SrcReg[j].Index && - GET_BIT(pairinst->Instruction.DstReg.WriteMask, swz)) - continue; - - struct reg_value_reader* r = memory_pool_malloc(&s->Compiler->Base.Pool, sizeof(*r)); - pairinst->NumDependencies++; - t->Value[swz]->NumReaders++; - r->Reader = pairinst; - r->Next = t->Value[swz]->Readers; - t->Value[swz]->Readers = r; - } - } - } - - int ndst = _mesa_num_inst_dst_regs(pairinst->Instruction.Opcode); - if (ndst) { - struct pair_register_translation *t = - get_register(s, pairinst->Instruction.DstReg.File, pairinst->Instruction.DstReg.Index); - if (t) { - t->RefCount++; - - if (pairinst->Instruction.DstReg.File == PROGRAM_TEMPORARY) { - int j; - for(j = 0; j < 4; ++j) { - if (!GET_BIT(pairinst->Instruction.DstReg.WriteMask, j)) - continue; - - struct reg_value* v = memory_pool_malloc(&s->Compiler->Base.Pool, sizeof(*v)); - memset(v, 0, sizeof(struct reg_value)); - v->Writer = pairinst; - if (t->Value[j]) { - pairinst->NumDependencies++; - t->Value[j]->Next = v; - } - t->Value[j] = v; - pairinst->Values[j] = v; - } - } - } - } - - if (s->Verbose) - _mesa_printf("scan(%i): NumDeps = %i\n", ip, pairinst->NumDependencies); - - if (!pairinst->NumDependencies) - instruction_ready(s, pairinst); - } - - /* Clear the PROGRAM_TEMPORARY state */ - int i, j; - for(i = 0; i < MAX_PROGRAM_TEMPS; ++i) { - for(j = 0; j < 4; ++j) - s->Temps[i].Value[j] = 0; - } -} - - -static void decrement_dependencies(struct pair_state *s, struct pair_state_instruction *pairinst) -{ - ASSERT(pairinst->NumDependencies > 0); - if (!--pairinst->NumDependencies) - instruction_ready(s, pairinst); -} /** - * Update the dependency tracking state based on what the instruction - * at the given IP does. + * Return the source slot where we installed the given register access, + * or -1 if no slot was free anymore. */ -static void commit_instruction(struct pair_state *s, struct pair_state_instruction *pairinst) -{ - struct prog_instruction *inst = &pairinst->Instruction; - - if (s->Verbose) - _mesa_printf("commit_instruction(%i)\n", pairinst->IP); - - if (inst->DstReg.File == PROGRAM_TEMPORARY) { - struct pair_register_translation *t = &s->Temps[inst->DstReg.Index]; - deref_hw_reg(s, t->HwIndex); - - int i; - for(i = 0; i < 4; ++i) { - if (!GET_BIT(inst->DstReg.WriteMask, i)) - continue; - - t->Value[i] = pairinst->Values[i]; - if (t->Value[i]->NumReaders) { - struct reg_value_reader *r; - for(r = pairinst->Values[i]->Readers; r; r = r->Next) - decrement_dependencies(s, r->Reader); - } else if (t->Value[i]->Next) { - /* This happens when the only reader writes - * the register at the same time */ - decrement_dependencies(s, t->Value[i]->Next->Writer); - } - } - } - - int nsrc = _mesa_num_inst_src_regs(inst->Opcode); - int i; - for(i = 0; i < nsrc; i++) { - struct pair_register_translation *t = get_register(s, inst->SrcReg[i].File, inst->SrcReg[i].Index); - if (!t) - continue; - - deref_hw_reg(s, get_hw_reg(s, inst->SrcReg[i].File, inst->SrcReg[i].Index)); - - if (inst->SrcReg[i].File != PROGRAM_TEMPORARY) - continue; - - int j; - for(j = 0; j < 4; ++j) { - GLuint swz = GET_SWZ(inst->SrcReg[i].Swizzle, j); - if (swz >= 4) - continue; - if (!t->Value[swz]) - continue; - - /* Do not free a dependency if this instruction - * also rewrites the value. See scan_instructions. */ - if (inst->DstReg.File == PROGRAM_TEMPORARY && - inst->DstReg.Index == inst->SrcReg[i].Index && - GET_BIT(inst->DstReg.WriteMask, swz)) - continue; - - if (!--t->Value[swz]->NumReaders) { - if (t->Value[swz]->Next) - decrement_dependencies(s, t->Value[swz]->Next->Writer); - } - } - } -} - - -/** - * Emit all ready texture instructions in a single block. - * - * Emit as a single block to (hopefully) sample many textures in parallel, - * and to avoid hardware indirections on R300. - * - * In R500, we don't really know when the result of a texture instruction - * arrives. So allocate all destinations first, to make sure they do not - * arrive early and overwrite a texture coordinate we're going to use later - * in the block. - */ -static void emit_all_tex(struct pair_state *s) -{ - struct pair_state_instruction *readytex; - struct pair_state_instruction *pairinst; - - ASSERT(s->ReadyTEX); - - // Don't let the ready list change under us! - readytex = s->ReadyTEX; - s->ReadyTEX = 0; - - // Allocate destination hardware registers in one block to avoid conflicts. - for(pairinst = readytex; pairinst; pairinst = pairinst->NextReady) { - struct prog_instruction *inst = &pairinst->Instruction; - if (inst->Opcode != OPCODE_KIL) - get_hw_reg(s, inst->DstReg.File, inst->DstReg.Index); - } - - if (s->Compiler->Base.Debug) - _mesa_printf(" BEGIN_TEX\n"); - - if (s->Handler->BeginTexBlock) - s->Compiler->Base.Error = s->Compiler->Base.Error || !s->Handler->BeginTexBlock(s->UserData); - - for(pairinst = readytex; pairinst; pairinst = pairinst->NextReady) { - struct prog_instruction *inst = &pairinst->Instruction; - commit_instruction(s, pairinst); - - if (inst->Opcode != OPCODE_KIL) - inst->DstReg.Index = get_hw_reg(s, inst->DstReg.File, inst->DstReg.Index); - inst->SrcReg[0].Index = get_hw_reg(s, inst->SrcReg[0].File, inst->SrcReg[0].Index); - - if (s->Compiler->Base.Debug) { - _mesa_printf(" "); - _mesa_print_instruction(inst); - fflush(stderr); - } - - struct radeon_pair_texture_instruction rpti; - - switch(inst->Opcode) { - case OPCODE_TEX: rpti.Opcode = RADEON_OPCODE_TEX; break; - case OPCODE_TXB: rpti.Opcode = RADEON_OPCODE_TXB; break; - case OPCODE_TXP: rpti.Opcode = RADEON_OPCODE_TXP; break; - default: - case OPCODE_KIL: rpti.Opcode = RADEON_OPCODE_KIL; break; - } - - rpti.DestIndex = inst->DstReg.Index; - rpti.WriteMask = inst->DstReg.WriteMask; - rpti.TexSrcUnit = inst->TexSrcUnit; - rpti.TexSrcTarget = inst->TexSrcTarget; - rpti.SrcIndex = inst->SrcReg[0].Index; - rpti.SrcSwizzle = inst->SrcReg[0].Swizzle; - - s->Compiler->Base.Error = s->Compiler->Base.Error || !s->Handler->EmitTex(s->UserData, &rpti); - } - - if (s->Compiler->Base.Debug) - _mesa_printf(" END_TEX\n"); -} - - -static int alloc_pair_source(struct pair_state *s, struct radeon_pair_instruction *pair, - struct prog_src_register src, GLboolean rgb, GLboolean alpha) +int rc_pair_alloc_source(struct rc_pair_instruction *pair, + unsigned int rgb, unsigned int alpha, + rc_register_file file, unsigned int index) { int candidate = -1; int candidate_quality = -1; int i; - if (!rgb && !alpha) + if ((!rgb && !alpha) || file == RC_FILE_NONE) return 0; - GLuint constant; - GLuint index; - - if (src.File == PROGRAM_TEMPORARY || src.File == PROGRAM_INPUT) { - constant = 0; - index = get_hw_reg(s, src.File, src.Index); - } else { - constant = 1; - index = src.Index; - } - for(i = 0; i < 3; ++i) { int q = 0; if (rgb) { if (pair->RGB.Src[i].Used) { - if (pair->RGB.Src[i].Constant != constant || + if (pair->RGB.Src[i].File != file || pair->RGB.Src[i].Index != index) continue; q++; @@ -610,7 +55,7 @@ static int alloc_pair_source(struct pair_state *s, struct radeon_pair_instructio } if (alpha) { if (pair->Alpha.Src[i].Used) { - if (pair->Alpha.Src[i].Constant != constant || + if (pair->Alpha.Src[i].File != file || pair->Alpha.Src[i].Index != index) continue; q++; @@ -625,334 +70,15 @@ static int alloc_pair_source(struct pair_state *s, struct radeon_pair_instructio if (candidate >= 0) { if (rgb) { pair->RGB.Src[candidate].Used = 1; - pair->RGB.Src[candidate].Constant = constant; + pair->RGB.Src[candidate].File = file; pair->RGB.Src[candidate].Index = index; } if (alpha) { pair->Alpha.Src[candidate].Used = 1; - pair->Alpha.Src[candidate].Constant = constant; + pair->Alpha.Src[candidate].File = file; pair->Alpha.Src[candidate].Index = index; } } return candidate; } - -/** - * Fill the given ALU instruction's opcodes and source operands into the given pair, - * if possible. - */ -static GLboolean fill_instruction_into_pair( - struct pair_state *s, - struct radeon_pair_instruction *pair, - struct pair_state_instruction *pairinst) -{ - struct prog_instruction *inst = &pairinst->Instruction; - - ASSERT(!pairinst->NeedRGB || pair->RGB.Opcode == OPCODE_NOP); - ASSERT(!pairinst->NeedAlpha || pair->Alpha.Opcode == OPCODE_NOP); - - if (pairinst->NeedRGB) { - if (pairinst->IsTranscendent) - pair->RGB.Opcode = OPCODE_REPL_ALPHA; - else - pair->RGB.Opcode = inst->Opcode; - if (inst->SaturateMode == SATURATE_ZERO_ONE) - pair->RGB.Saturate = 1; - } - if (pairinst->NeedAlpha) { - pair->Alpha.Opcode = inst->Opcode; - if (inst->SaturateMode == SATURATE_ZERO_ONE) - pair->Alpha.Saturate = 1; - } - - int nargs = _mesa_num_inst_src_regs(inst->Opcode); - int i; - - /* Special case for DDX/DDY (MDH/MDV). */ - if (inst->Opcode == OPCODE_DDX || inst->Opcode == OPCODE_DDY) { - if (pair->RGB.Src[0].Used || pair->Alpha.Src[0].Used) - return GL_FALSE; - else - nargs++; - } - - for(i = 0; i < nargs; ++i) { - int source; - if (pairinst->NeedRGB && !pairinst->IsTranscendent) { - GLboolean srcrgb = GL_FALSE; - GLboolean srcalpha = GL_FALSE; - int j; - for(j = 0; j < 3; ++j) { - GLuint swz = GET_SWZ(inst->SrcReg[i].Swizzle, j); - if (swz < 3) - srcrgb = GL_TRUE; - else if (swz < 4) - srcalpha = GL_TRUE; - } - source = alloc_pair_source(s, pair, inst->SrcReg[i], srcrgb, srcalpha); - if (source < 0) - return GL_FALSE; - pair->RGB.Arg[i].Source = source; - pair->RGB.Arg[i].Swizzle = inst->SrcReg[i].Swizzle & 0x1ff; - pair->RGB.Arg[i].Abs = inst->SrcReg[i].Abs; - pair->RGB.Arg[i].Negate = !!(inst->SrcReg[i].Negate & (NEGATE_X | NEGATE_Y | NEGATE_Z)); - } - if (pairinst->NeedAlpha) { - GLboolean srcrgb = GL_FALSE; - GLboolean srcalpha = GL_FALSE; - GLuint swz = GET_SWZ(inst->SrcReg[i].Swizzle, pairinst->IsTranscendent ? 0 : 3); - if (swz < 3) - srcrgb = GL_TRUE; - else if (swz < 4) - srcalpha = GL_TRUE; - source = alloc_pair_source(s, pair, inst->SrcReg[i], srcrgb, srcalpha); - if (source < 0) - return GL_FALSE; - pair->Alpha.Arg[i].Source = source; - pair->Alpha.Arg[i].Swizzle = swz; - pair->Alpha.Arg[i].Abs = inst->SrcReg[i].Abs; - pair->Alpha.Arg[i].Negate = !!(inst->SrcReg[i].Negate & NEGATE_W); - } - } - - return GL_TRUE; -} - - -/** - * Fill in the destination register information. - * - * This is split from filling in source registers because we want - * to avoid allocating hardware temporaries for destinations until - * we are absolutely certain that we're going to emit a certain - * instruction pairing. - */ -static void fill_dest_into_pair( - struct pair_state *s, - struct radeon_pair_instruction *pair, - struct pair_state_instruction *pairinst) -{ - struct prog_instruction *inst = &pairinst->Instruction; - - if (inst->DstReg.File == PROGRAM_OUTPUT) { - if (inst->DstReg.Index == s->Compiler->OutputColor) { - pair->RGB.OutputWriteMask |= inst->DstReg.WriteMask & WRITEMASK_XYZ; - pair->Alpha.OutputWriteMask |= GET_BIT(inst->DstReg.WriteMask, 3); - } else if (inst->DstReg.Index == s->Compiler->OutputDepth) { - pair->Alpha.DepthWriteMask |= GET_BIT(inst->DstReg.WriteMask, 3); - } - } else { - GLuint hwindex = get_hw_reg(s, inst->DstReg.File, inst->DstReg.Index); - if (pairinst->NeedRGB) { - pair->RGB.DestIndex = hwindex; - pair->RGB.WriteMask |= inst->DstReg.WriteMask & WRITEMASK_XYZ; - } - if (pairinst->NeedAlpha) { - pair->Alpha.DestIndex = hwindex; - pair->Alpha.WriteMask |= GET_BIT(inst->DstReg.WriteMask, 3); - } - } -} - - -/** - * Find a good ALU instruction or pair of ALU instruction and emit it. - * - * Prefer emitting full ALU instructions, so that when we reach a point - * where no full ALU instruction can be emitted, we have more candidates - * for RGB/Alpha pairing. - */ -static void emit_alu(struct pair_state *s) -{ - struct radeon_pair_instruction pair; - struct pair_state_instruction *psi; - - if (s->ReadyFullALU || !(s->ReadyRGB && s->ReadyAlpha)) { - if (s->ReadyFullALU) { - psi = s->ReadyFullALU; - s->ReadyFullALU = s->ReadyFullALU->NextReady; - } else if (s->ReadyRGB) { - psi = s->ReadyRGB; - s->ReadyRGB = s->ReadyRGB->NextReady; - } else { - psi = s->ReadyAlpha; - s->ReadyAlpha = s->ReadyAlpha->NextReady; - } - - _mesa_bzero(&pair, sizeof(pair)); - fill_instruction_into_pair(s, &pair, psi); - fill_dest_into_pair(s, &pair, psi); - commit_instruction(s, psi); - } else { - struct pair_state_instruction **prgb; - struct pair_state_instruction **palpha; - - /* Some pairings might fail because they require too - * many source slots; try all possible pairings if necessary */ - for(prgb = &s->ReadyRGB; *prgb; prgb = &(*prgb)->NextReady) { - for(palpha = &s->ReadyAlpha; *palpha; palpha = &(*palpha)->NextReady) { - struct pair_state_instruction * psirgb = *prgb; - struct pair_state_instruction * psialpha = *palpha; - _mesa_bzero(&pair, sizeof(pair)); - fill_instruction_into_pair(s, &pair, psirgb); - if (!fill_instruction_into_pair(s, &pair, psialpha)) - continue; - *prgb = (*prgb)->NextReady; - *palpha = (*palpha)->NextReady; - fill_dest_into_pair(s, &pair, psirgb); - fill_dest_into_pair(s, &pair, psialpha); - commit_instruction(s, psirgb); - commit_instruction(s, psialpha); - goto success; - } - } - - /* No success in pairing; just take the first RGB instruction */ - psi = s->ReadyRGB; - s->ReadyRGB = s->ReadyRGB->NextReady; - - _mesa_bzero(&pair, sizeof(pair)); - fill_instruction_into_pair(s, &pair, psi); - fill_dest_into_pair(s, &pair, psi); - commit_instruction(s, psi); - success: ; - } - - if (s->Compiler->Base.Debug) - radeonPrintPairInstruction(&pair); - - s->Compiler->Base.Error = s->Compiler->Base.Error || !s->Handler->EmitPaired(s->UserData, &pair); -} - -/* Callback function for assigning input registers to hardware registers */ -static void alloc_helper(void * data, unsigned input, unsigned hwreg) -{ - struct pair_state * s = data; - alloc_hw_reg(s, PROGRAM_INPUT, input, hwreg); -} - -void radeonPairProgram( - struct r300_fragment_program_compiler * compiler, - const struct radeon_pair_handler* handler, void *userdata) -{ - struct pair_state s; - - _mesa_bzero(&s, sizeof(s)); - s.Compiler = compiler; - s.Handler = handler; - s.UserData = userdata; - s.Verbose = GL_FALSE && s.Compiler->Base.Debug; - - if (s.Compiler->Base.Debug) - _mesa_printf("Emit paired program\n"); - - scan_instructions(&s); - s.Compiler->AllocateHwInputs(s.Compiler, &alloc_helper, &s); - - while(!s.Compiler->Base.Error && - (s.ReadyTEX || s.ReadyRGB || s.ReadyAlpha || s.ReadyFullALU)) { - if (s.ReadyTEX) - emit_all_tex(&s); - - while(s.ReadyFullALU || s.ReadyRGB || s.ReadyAlpha) - emit_alu(&s); - } - - if (s.Compiler->Base.Debug) - _mesa_printf(" END\n"); -} - - -static void print_pair_src(int i, struct radeon_pair_instruction_source* src) -{ - _mesa_printf(" Src%i = %s[%i]", i, src->Constant ? "CNST" : "TEMP", src->Index); -} - -static const char* opcode_string(GLuint opcode) -{ - if (opcode == OPCODE_REPL_ALPHA) - return "SOP"; - else - return _mesa_opcode_string(opcode); -} - -static int num_pairinst_args(GLuint opcode) -{ - if (opcode == OPCODE_REPL_ALPHA) - return 0; - else - return _mesa_num_inst_src_regs(opcode); -} - -static char swizzle_char(GLuint swz) -{ - switch(swz) { - case SWIZZLE_X: return 'x'; - case SWIZZLE_Y: return 'y'; - case SWIZZLE_Z: return 'z'; - case SWIZZLE_W: return 'w'; - case SWIZZLE_ZERO: return '0'; - case SWIZZLE_ONE: return '1'; - case SWIZZLE_NIL: return '_'; - default: return '?'; - } -} - -void radeonPrintPairInstruction(struct radeon_pair_instruction *inst) -{ - int nargs; - int i; - - _mesa_printf(" RGB: "); - for(i = 0; i < 3; ++i) { - if (inst->RGB.Src[i].Used) - print_pair_src(i, inst->RGB.Src + i); - } - _mesa_printf("\n"); - _mesa_printf(" Alpha:"); - for(i = 0; i < 3; ++i) { - if (inst->Alpha.Src[i].Used) - print_pair_src(i, inst->Alpha.Src + i); - } - _mesa_printf("\n"); - - _mesa_printf(" %s%s", opcode_string(inst->RGB.Opcode), inst->RGB.Saturate ? "_SAT" : ""); - if (inst->RGB.WriteMask) - _mesa_printf(" TEMP[%i].%s%s%s", inst->RGB.DestIndex, - (inst->RGB.WriteMask & 1) ? "x" : "", - (inst->RGB.WriteMask & 2) ? "y" : "", - (inst->RGB.WriteMask & 4) ? "z" : ""); - if (inst->RGB.OutputWriteMask) - _mesa_printf(" COLOR.%s%s%s", - (inst->RGB.OutputWriteMask & 1) ? "x" : "", - (inst->RGB.OutputWriteMask & 2) ? "y" : "", - (inst->RGB.OutputWriteMask & 4) ? "z" : ""); - nargs = num_pairinst_args(inst->RGB.Opcode); - for(i = 0; i < nargs; ++i) { - const char* abs = inst->RGB.Arg[i].Abs ? "|" : ""; - const char* neg = inst->RGB.Arg[i].Negate ? "-" : ""; - _mesa_printf(", %s%sSrc%i.%c%c%c%s", neg, abs, inst->RGB.Arg[i].Source, - swizzle_char(GET_SWZ(inst->RGB.Arg[i].Swizzle, 0)), - swizzle_char(GET_SWZ(inst->RGB.Arg[i].Swizzle, 1)), - swizzle_char(GET_SWZ(inst->RGB.Arg[i].Swizzle, 2)), - abs); - } - _mesa_printf("\n"); - - _mesa_printf(" %s%s", opcode_string(inst->Alpha.Opcode), inst->Alpha.Saturate ? "_SAT" : ""); - if (inst->Alpha.WriteMask) - _mesa_printf(" TEMP[%i].w", inst->Alpha.DestIndex); - if (inst->Alpha.OutputWriteMask) - _mesa_printf(" COLOR.w"); - if (inst->Alpha.DepthWriteMask) - _mesa_printf(" DEPTH.w"); - nargs = num_pairinst_args(inst->Alpha.Opcode); - for(i = 0; i < nargs; ++i) { - const char* abs = inst->Alpha.Arg[i].Abs ? "|" : ""; - const char* neg = inst->Alpha.Arg[i].Negate ? "-" : ""; - _mesa_printf(", %s%sSrc%i.%c%s", neg, abs, inst->Alpha.Arg[i].Source, - swizzle_char(inst->Alpha.Arg[i].Swizzle), abs); - } - _mesa_printf("\n"); -} diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_program_pair.h b/src/mesa/drivers/dri/r300/compiler/radeon_program_pair.h index ff761785518..16005984287 100644 --- a/src/mesa/drivers/dri/r300/compiler/radeon_program_pair.h +++ b/src/mesa/drivers/dri/r300/compiler/radeon_program_pair.h @@ -28,116 +28,97 @@ #ifndef __RADEON_PROGRAM_PAIR_H_ #define __RADEON_PROGRAM_PAIR_H_ -#include "radeon_program.h" +#include "radeon_code.h" +#include "radeon_opcodes.h" +#include "radeon_program_constants.h" struct r300_fragment_program_compiler; /** - * Represents a paired instruction, as found in R300 and R500 + * \file + * Represents a paired ALU instruction, as found in R300 and R500 * fragment programs. + * + * Note that this representation is taking some liberties as far + * as register files are concerned, to allow separate register + * allocation. + * + * Also note that there are some subtleties in that the semantics + * of certain opcodes are implicitly changed in this representation; + * see \ref rc_pair_translate */ + + struct radeon_pair_instruction_source { - GLuint Index:8; - GLuint Constant:1; - GLuint Used:1; + unsigned int Used:1; + rc_register_file File:3; + unsigned int Index:RC_REGISTER_INDEX_BITS; }; struct radeon_pair_instruction_rgb { - GLuint Opcode:8; - GLuint DestIndex:8; - GLuint WriteMask:3; - GLuint OutputWriteMask:3; - GLuint Saturate:1; + rc_opcode Opcode:8; + unsigned int DestIndex:RC_REGISTER_INDEX_BITS; + unsigned int WriteMask:3; + unsigned int OutputWriteMask:3; + unsigned int Saturate:1; struct radeon_pair_instruction_source Src[3]; struct { - GLuint Source:2; - GLuint Swizzle:9; - GLuint Abs:1; - GLuint Negate:1; + unsigned int Source:2; + unsigned int Swizzle:9; + unsigned int Abs:1; + unsigned int Negate:1; } Arg[3]; }; struct radeon_pair_instruction_alpha { - GLuint Opcode:8; - GLuint DestIndex:8; - GLuint WriteMask:1; - GLuint OutputWriteMask:1; - GLuint DepthWriteMask:1; - GLuint Saturate:1; + rc_opcode Opcode:8; + unsigned int DestIndex:RC_REGISTER_INDEX_BITS; + unsigned int WriteMask:1; + unsigned int OutputWriteMask:1; + unsigned int DepthWriteMask:1; + unsigned int Saturate:1; struct radeon_pair_instruction_source Src[3]; struct { - GLuint Source:2; - GLuint Swizzle:3; - GLuint Abs:1; - GLuint Negate:1; + unsigned int Source:2; + unsigned int Swizzle:3; + unsigned int Abs:1; + unsigned int Negate:1; } Arg[3]; }; -struct radeon_pair_instruction { +struct rc_pair_instruction { struct radeon_pair_instruction_rgb RGB; struct radeon_pair_instruction_alpha Alpha; -}; - -enum { - RADEON_OPCODE_TEX = 0, - RADEON_OPCODE_TXB, - RADEON_OPCODE_TXP, - RADEON_OPCODE_KIL + rc_write_aluresult WriteALUResult:2; + rc_compare_func ALUResultCompare:3; }; -struct radeon_pair_texture_instruction { - GLuint Opcode:2; /**< one of RADEON_OPCODE_xxx */ - - GLuint DestIndex:8; - GLuint WriteMask:4; - GLuint TexSrcUnit:5; - GLuint TexSrcTarget:3; - - GLuint SrcIndex:8; - GLuint SrcSwizzle:12; -}; +/** + * General helper functions for dealing with the paired instruction format. + */ +/*@{*/ +int rc_pair_alloc_source(struct rc_pair_instruction *pair, + unsigned int rgb, unsigned int alpha, + rc_register_file file, unsigned int index); +/*@}*/ /** - * + * Compiler passes that operate with the paired format. */ -struct radeon_pair_handler { - /** - * Write a paired instruction to the hardware. - * - * @return GL_FALSE on error. - */ - GLboolean (*EmitPaired)(void*, struct radeon_pair_instruction*); - - /** - * Write a texture instruction to the hardware. - * Register indices have already been rewritten to the allocated - * hardware register numbers. - * - * @return GL_FALSE on error. - */ - GLboolean (*EmitTex)(void*, struct radeon_pair_texture_instruction*); - - /** - * Called before a block of contiguous, independent texture - * instructions is emitted. - */ - GLboolean (*BeginTexBlock)(void*); - - unsigned MaxHwTemps; -}; - -void radeonPairProgram( - struct r300_fragment_program_compiler * compiler, - const struct radeon_pair_handler*, void *userdata); +/*@{*/ +struct radeon_pair_handler; -void radeonPrintPairInstruction(struct radeon_pair_instruction *inst); +void rc_pair_translate(struct r300_fragment_program_compiler *c); +void rc_pair_schedule(struct r300_fragment_program_compiler *c); +void rc_pair_regalloc(struct r300_fragment_program_compiler *c, unsigned maxtemps); +/*@}*/ #endif /* __RADEON_PROGRAM_PAIR_H_ */ diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_program_print.c b/src/mesa/drivers/dri/r300/compiler/radeon_program_print.c new file mode 100644 index 00000000000..d863b82d53f --- /dev/null +++ b/src/mesa/drivers/dri/r300/compiler/radeon_program_print.c @@ -0,0 +1,300 @@ +/* + * Copyright 2009 Nicolai Hähnle <[email protected]> + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. */ + +#include "radeon_program.h" + +#include <stdio.h> + +static const char * textarget_to_string(rc_texture_target target) +{ + switch(target) { + case RC_TEXTURE_2D_ARRAY: return "2D_ARRAY"; + case RC_TEXTURE_1D_ARRAY: return "1D_ARRAY"; + case RC_TEXTURE_CUBE: return "CUBE"; + case RC_TEXTURE_3D: return "3D"; + case RC_TEXTURE_RECT: return "RECT"; + case RC_TEXTURE_2D: return "2D"; + case RC_TEXTURE_1D: return "1D"; + default: return "BAD_TEXTURE_TARGET"; + } +} + +static void rc_print_comparefunc(FILE * f, const char * lhs, rc_compare_func func, const char * rhs) +{ + if (func == RC_COMPARE_FUNC_NEVER) { + fprintf(f, "false"); + } else if (func == RC_COMPARE_FUNC_ALWAYS) { + fprintf(f, "true"); + } else { + const char * op; + switch(func) { + case RC_COMPARE_FUNC_LESS: op = "<"; break; + case RC_COMPARE_FUNC_EQUAL: op = "=="; break; + case RC_COMPARE_FUNC_LEQUAL: op = "<="; break; + case RC_COMPARE_FUNC_GREATER: op = ">"; break; + case RC_COMPARE_FUNC_NOTEQUAL: op = "!="; break; + case RC_COMPARE_FUNC_GEQUAL: op = ">="; break; + default: op = "???"; break; + } + fprintf(f, "%s %s %s", lhs, op, rhs); + } +} + +static void rc_print_register(FILE * f, rc_register_file file, int index, unsigned int reladdr) +{ + if (file == RC_FILE_NONE) { + fprintf(f, "none"); + } else if (file == RC_FILE_SPECIAL) { + switch(index) { + case RC_SPECIAL_ALU_RESULT: fprintf(f, "aluresult"); break; + default: fprintf(f, "special[%i]", index); break; + } + } else { + const char * filename; + switch(file) { + case RC_FILE_TEMPORARY: filename = "temp"; break; + case RC_FILE_INPUT: filename = "input"; break; + case RC_FILE_OUTPUT: filename = "output"; break; + case RC_FILE_ADDRESS: filename = "addr"; break; + case RC_FILE_CONSTANT: filename = "const"; break; + default: filename = "BAD FILE"; break; + } + fprintf(f, "%s[%i%s]", filename, index, reladdr ? " + addr[0]" : ""); + } +} + +static void rc_print_mask(FILE * f, unsigned int mask) +{ + if (mask & RC_MASK_X) fprintf(f, "x"); + if (mask & RC_MASK_Y) fprintf(f, "y"); + if (mask & RC_MASK_Z) fprintf(f, "z"); + if (mask & RC_MASK_W) fprintf(f, "w"); +} + +static void rc_print_dst_register(FILE * f, struct rc_dst_register dst) +{ + rc_print_register(f, dst.File, dst.Index, dst.RelAddr); + if (dst.WriteMask != RC_MASK_XYZW) { + fprintf(f, "."); + rc_print_mask(f, dst.WriteMask); + } +} + +static char rc_swizzle_char(unsigned int swz) +{ + switch(swz) { + case RC_SWIZZLE_X: return 'x'; + case RC_SWIZZLE_Y: return 'y'; + case RC_SWIZZLE_Z: return 'z'; + case RC_SWIZZLE_W: return 'w'; + case RC_SWIZZLE_ZERO: return '0'; + case RC_SWIZZLE_ONE: return '1'; + case RC_SWIZZLE_HALF: return 'H'; + case RC_SWIZZLE_UNUSED: return '_'; + } + return '?'; +} + +static void rc_print_swizzle(FILE * f, unsigned int swizzle, unsigned int negate) +{ + unsigned int comp; + for(comp = 0; comp < 4; ++comp) { + rc_swizzle swz = GET_SWZ(swizzle, comp); + if (GET_BIT(negate, comp)) + fprintf(f, "-"); + fprintf(f, "%c", rc_swizzle_char(swz)); + } +} + +static void rc_print_src_register(FILE * f, struct rc_src_register src) +{ + int trivial_negate = (src.Negate == RC_MASK_NONE || src.Negate == RC_MASK_XYZW); + + if (src.Negate == RC_MASK_XYZW) + fprintf(f, "-"); + if (src.Abs) + fprintf(f, "|"); + + rc_print_register(f, src.File, src.Index, src.RelAddr); + + if (src.Abs && !trivial_negate) + fprintf(f, "|"); + + if (src.Swizzle != RC_SWIZZLE_XYZW || !trivial_negate) { + fprintf(f, "."); + rc_print_swizzle(f, src.Swizzle, trivial_negate ? 0 : src.Negate); + } + + if (src.Abs && trivial_negate) + fprintf(f, "|"); +} + +static void rc_print_normal_instruction(FILE * f, struct rc_instruction * inst) +{ + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->U.I.Opcode); + unsigned int reg; + + fprintf(f, "%s", opcode->Name); + + switch(inst->U.I.SaturateMode) { + case RC_SATURATE_NONE: break; + case RC_SATURATE_ZERO_ONE: fprintf(f, "_SAT"); break; + case RC_SATURATE_MINUS_PLUS_ONE: fprintf(f, "_SAT2"); break; + default: fprintf(f, "_BAD_SAT"); break; + } + + if (opcode->HasDstReg) { + fprintf(f, " "); + rc_print_dst_register(f, inst->U.I.DstReg); + if (opcode->NumSrcRegs) + fprintf(f, ","); + } + + for(reg = 0; reg < opcode->NumSrcRegs; ++reg) { + if (reg > 0) + fprintf(f, ","); + fprintf(f, " "); + rc_print_src_register(f, inst->U.I.SrcReg[reg]); + } + + if (opcode->HasTexture) { + fprintf(f, ", %s%s[%u]", + textarget_to_string(inst->U.I.TexSrcTarget), + inst->U.I.TexShadow ? "SHADOW" : "", + inst->U.I.TexSrcUnit); + } + + fprintf(f, ";"); + + if (inst->U.I.WriteALUResult) { + fprintf(f, " [aluresult = ("); + rc_print_comparefunc(f, + (inst->U.I.WriteALUResult == RC_ALURESULT_X) ? "x" : "w", + inst->U.I.ALUResultCompare, "0"); + fprintf(f, ")]"); + } + + fprintf(f, "\n"); +} + +static void rc_print_pair_instruction(FILE * f, struct rc_instruction * fullinst) +{ + struct rc_pair_instruction * inst = &fullinst->U.P; + int printedsrc = 0; + + for(unsigned int src = 0; src < 3; ++src) { + if (inst->RGB.Src[src].Used) { + if (printedsrc) + fprintf(f, ", "); + fprintf(f, "src%i.xyz = ", src); + rc_print_register(f, inst->RGB.Src[src].File, inst->RGB.Src[src].Index, 0); + printedsrc = 1; + } + if (inst->Alpha.Src[src].Used) { + if (printedsrc) + fprintf(f, ", "); + fprintf(f, "src%i.w = ", src); + rc_print_register(f, inst->Alpha.Src[src].File, inst->Alpha.Src[src].Index, 0); + printedsrc = 1; + } + } + fprintf(f, "\n"); + + if (inst->RGB.Opcode != RC_OPCODE_NOP) { + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->RGB.Opcode); + + fprintf(f, " %s%s", opcode->Name, inst->RGB.Saturate ? "_SAT" : ""); + if (inst->RGB.WriteMask) + fprintf(f, " temp[%i].%s%s%s", inst->RGB.DestIndex, + (inst->RGB.WriteMask & 1) ? "x" : "", + (inst->RGB.WriteMask & 2) ? "y" : "", + (inst->RGB.WriteMask & 4) ? "z" : ""); + if (inst->RGB.OutputWriteMask) + fprintf(f, " color.%s%s%s", + (inst->RGB.OutputWriteMask & 1) ? "x" : "", + (inst->RGB.OutputWriteMask & 2) ? "y" : "", + (inst->RGB.OutputWriteMask & 4) ? "z" : ""); + if (inst->WriteALUResult == RC_ALURESULT_X) + fprintf(f, " aluresult"); + + for(unsigned int arg = 0; arg < opcode->NumSrcRegs; ++arg) { + const char* abs = inst->RGB.Arg[arg].Abs ? "|" : ""; + const char* neg = inst->RGB.Arg[arg].Negate ? "-" : ""; + fprintf(f, ", %s%ssrc%i.%c%c%c%s", neg, abs, inst->RGB.Arg[arg].Source, + rc_swizzle_char(GET_SWZ(inst->RGB.Arg[arg].Swizzle, 0)), + rc_swizzle_char(GET_SWZ(inst->RGB.Arg[arg].Swizzle, 1)), + rc_swizzle_char(GET_SWZ(inst->RGB.Arg[arg].Swizzle, 2)), + abs); + } + fprintf(f, "\n"); + } + + if (inst->Alpha.Opcode != RC_OPCODE_NOP) { + const struct rc_opcode_info * opcode = rc_get_opcode_info(inst->Alpha.Opcode); + + fprintf(f, " %s%s", opcode->Name, inst->Alpha.Saturate ? "_SAT" : ""); + if (inst->Alpha.WriteMask) + fprintf(f, " temp[%i].w", inst->Alpha.DestIndex); + if (inst->Alpha.OutputWriteMask) + fprintf(f, " color.w"); + if (inst->Alpha.DepthWriteMask) + fprintf(f, " depth.w"); + if (inst->WriteALUResult == RC_ALURESULT_W) + fprintf(f, " aluresult"); + + for(unsigned int arg = 0; arg < opcode->NumSrcRegs; ++arg) { + const char* abs = inst->Alpha.Arg[arg].Abs ? "|" : ""; + const char* neg = inst->Alpha.Arg[arg].Negate ? "-" : ""; + fprintf(f, ", %s%ssrc%i.%c%s", neg, abs, inst->Alpha.Arg[arg].Source, + rc_swizzle_char(inst->Alpha.Arg[arg].Swizzle), abs); + } + fprintf(f, "\n"); + } + + if (inst->WriteALUResult) { + fprintf(f, " [aluresult = ("); + rc_print_comparefunc(f, "result", inst->ALUResultCompare, "0"); + fprintf(f, ")]\n"); + } +} + +/** + * Print program to stderr, default options. + */ +void rc_print_program(const struct rc_program *prog) +{ + unsigned int linenum = 0; + struct rc_instruction *inst; + + fprintf(stderr, "# Radeon Compiler Program\n"); + + for(inst = prog->Instructions.Next; inst != &prog->Instructions; inst = inst->Next) { + fprintf(stderr, "%3d: ", linenum); + + if (inst->Type == RC_INSTRUCTION_PAIR) + rc_print_pair_instruction(stderr, inst); + else + rc_print_normal_instruction(stderr, inst); + + linenum++; + } +} diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_swizzle.h b/src/mesa/drivers/dri/r300/compiler/radeon_swizzle.h new file mode 100644 index 00000000000..c81d5f7a5e9 --- /dev/null +++ b/src/mesa/drivers/dri/r300/compiler/radeon_swizzle.h @@ -0,0 +1,57 @@ +/* + * Copyright (C) 2009 Nicolai Haehnle. + * + * 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, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef RADEON_SWIZZLE_H +#define RADEON_SWIZZLE_H + +#include "radeon_program.h" + +struct rc_swizzle_split { + unsigned char NumPhases; + unsigned char Phase[4]; +}; + +/** + * Describe the swizzling capability of target hardware. + */ +struct rc_swizzle_caps { + /** + * Check whether the given swizzle, absolute and negate combination + * can be implemented natively by the hardware for this opcode. + * + * \return 1 if the swizzle is native for the given opcode + */ + int (*IsNative)(rc_opcode opcode, struct rc_src_register reg); + + /** + * Determine how to split access to the masked channels of the + * given source register to obtain ALU-native swizzles. + */ + void (*Split)(struct rc_src_register reg, unsigned int mask, struct rc_swizzle_split * split); +}; + +#endif /* RADEON_SWIZZLE_H */ diff --git a/src/mesa/drivers/dri/r300/r300_blit.c b/src/mesa/drivers/dri/r300/r300_blit.c new file mode 100644 index 00000000000..10e1b3c9124 --- /dev/null +++ b/src/mesa/drivers/dri/r300/r300_blit.c @@ -0,0 +1,524 @@ +/* + * Copyright (C) 2009 Maciej Cencora <[email protected]> + * + * 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, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "radeon_common.h" +#include "r300_context.h" + +#include "r300_blit.h" +#include "r300_cmdbuf.h" +#include "r300_emit.h" +#include "r300_tex.h" +#include "compiler/radeon_compiler.h" +#include "compiler/radeon_opcodes.h" + +static void vp_ins_outs(struct r300_vertex_program_compiler *c) +{ + c->code->inputs[VERT_ATTRIB_POS] = 0; + c->code->inputs[VERT_ATTRIB_TEX0] = 1; + c->code->outputs[VERT_RESULT_HPOS] = 0; + c->code->outputs[VERT_RESULT_TEX0] = 1; +} + +static void fp_allocate_hw_inputs( + struct r300_fragment_program_compiler * c, + void (*allocate)(void * data, unsigned input, unsigned hwreg), + void * mydata) +{ + allocate(mydata, FRAG_ATTRIB_TEX0, 0); +} + +static void create_vertex_program(struct r300_context *r300) +{ + struct r300_vertex_program_compiler compiler; + struct rc_instruction *inst; + + rc_init(&compiler.Base); + + inst = rc_insert_new_instruction(&compiler.Base, compiler.Base.Program.Instructions.Prev); + inst->U.I.Opcode = RC_OPCODE_MOV; + inst->U.I.DstReg.File = RC_FILE_OUTPUT; + inst->U.I.DstReg.Index = VERT_RESULT_HPOS; + inst->U.I.DstReg.RelAddr = 0; + inst->U.I.DstReg.WriteMask = RC_MASK_XYZW; + inst->U.I.SrcReg[0].Abs = 0; + inst->U.I.SrcReg[0].File = RC_FILE_INPUT; + inst->U.I.SrcReg[0].Index = VERT_ATTRIB_POS; + inst->U.I.SrcReg[0].Negate = 0; + inst->U.I.SrcReg[0].RelAddr = 0; + inst->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_XYZW; + + inst = rc_insert_new_instruction(&compiler.Base, compiler.Base.Program.Instructions.Prev); + inst->U.I.Opcode = RC_OPCODE_MOV; + inst->U.I.DstReg.File = RC_FILE_OUTPUT; + inst->U.I.DstReg.Index = VERT_RESULT_TEX0; + inst->U.I.DstReg.RelAddr = 0; + inst->U.I.DstReg.WriteMask = RC_MASK_XYZW; + inst->U.I.SrcReg[0].Abs = 0; + inst->U.I.SrcReg[0].File = RC_FILE_INPUT; + inst->U.I.SrcReg[0].Index = VERT_ATTRIB_TEX0; + inst->U.I.SrcReg[0].Negate = 0; + inst->U.I.SrcReg[0].RelAddr = 0; + inst->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_XYZW; + + compiler.Base.Program.InputsRead = (1 << VERT_ATTRIB_POS) | (1 << VERT_ATTRIB_TEX0); + compiler.RequiredOutputs = compiler.Base.Program.OutputsWritten = (1 << VERT_RESULT_HPOS) | (1 << VERT_RESULT_TEX0); + compiler.SetHwInputOutput = vp_ins_outs; + compiler.code = &r300->blit.vp_code; + + r3xx_compile_vertex_program(&compiler); +} + +static void create_fragment_program(struct r300_context *r300) +{ + struct r300_fragment_program_compiler compiler; + struct rc_instruction *inst; + + rc_init(&compiler.Base); + + inst = rc_insert_new_instruction(&compiler.Base, compiler.Base.Program.Instructions.Prev); + inst->U.I.Opcode = RC_OPCODE_TEX; + inst->U.I.TexSrcTarget = RC_TEXTURE_2D; + inst->U.I.TexSrcUnit = 0; + inst->U.I.DstReg.File = RC_FILE_OUTPUT; + inst->U.I.DstReg.Index = FRAG_RESULT_COLOR; + inst->U.I.DstReg.WriteMask = RC_MASK_XYZW; + inst->U.I.SrcReg[0].Abs = 0; + inst->U.I.SrcReg[0].File = RC_FILE_INPUT; + inst->U.I.SrcReg[0].Index = FRAG_ATTRIB_TEX0; + inst->U.I.SrcReg[0].Negate = 0; + inst->U.I.SrcReg[0].RelAddr = 0; + inst->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_XYZW; + + compiler.Base.Program.InputsRead = (1 << FRAG_ATTRIB_TEX0); + compiler.OutputColor = FRAG_RESULT_COLOR; + compiler.OutputDepth = FRAG_RESULT_DEPTH; + compiler.is_r500 = (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515); + compiler.code = &r300->blit.fp_code; + compiler.AllocateHwInputs = fp_allocate_hw_inputs; + + r3xx_compile_fragment_program(&compiler); +} + +void r300_blit_init(struct r300_context *r300) +{ + create_vertex_program(r300); + create_fragment_program(r300); +} + +static void r300_emit_tx_setup(struct r300_context *r300, + gl_format mesa_format, + struct radeon_bo *bo, + intptr_t offset, + unsigned width, + unsigned height, + unsigned pitch) +{ + BATCH_LOCALS(&r300->radeon); + + assert(width <= 2048); + assert(height <= 2048); + assert(r300TranslateTexFormat(mesa_format) != 0); + assert(offset % 32 == 0); + + BEGIN_BATCH(17); + OUT_BATCH_REGVAL(R300_TX_FILTER0_0, + (R300_TX_CLAMP_TO_EDGE << R300_TX_WRAP_S_SHIFT) | + (R300_TX_CLAMP_TO_EDGE << R300_TX_WRAP_T_SHIFT) | + (R300_TX_CLAMP_TO_EDGE << R300_TX_WRAP_R_SHIFT) | + R300_TX_MIN_FILTER_MIP_NONE | + R300_TX_MIN_FILTER_LINEAR | + R300_TX_MAG_FILTER_LINEAR | + (0 << 28)); + OUT_BATCH_REGVAL(R300_TX_FILTER1_0, 0); + OUT_BATCH_REGVAL(R300_TX_SIZE_0, + ((width-1) << R300_TX_WIDTHMASK_SHIFT) | + ((height-1) << R300_TX_HEIGHTMASK_SHIFT) | + (0 << R300_TX_DEPTHMASK_SHIFT) | + (0 << R300_TX_MAX_MIP_LEVEL_SHIFT) | + R300_TX_SIZE_TXPITCH_EN); + + OUT_BATCH_REGVAL(R300_TX_FORMAT_0, r300TranslateTexFormat(mesa_format)); + OUT_BATCH_REGVAL(R300_TX_FORMAT2_0, pitch/_mesa_get_format_bytes(mesa_format) - 1); + OUT_BATCH_REGSEQ(R300_TX_OFFSET_0, 1); + OUT_BATCH_RELOC(0, bo, offset, RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0); + + OUT_BATCH_REGSEQ(R300_TX_INVALTAGS, 2); + OUT_BATCH(0); + OUT_BATCH(1); + + END_BATCH(); +} + +#define EASY_US_FORMAT(FMT, C0, C1, C2, C3, SIGN) \ + (FMT | R500_C0_SEL_##C0 | R500_C1_SEL_##C1 | \ + R500_C2_SEL_##C2 | R500_C3_SEL_##C3 | R500_OUT_SIGN(SIGN)) + +static uint32_t mesa_format_to_us_format(gl_format mesa_format) +{ + switch(mesa_format) + { + case MESA_FORMAT_S8_Z24: + case MESA_FORMAT_X8_Z24: + case MESA_FORMAT_RGBA8888: // x + return EASY_US_FORMAT(R500_OUT_FMT_C4_8, A, B, G, R, 0); + case MESA_FORMAT_RGB565: // x + case MESA_FORMAT_ARGB1555: // x + case MESA_FORMAT_RGBA8888_REV: // x + return EASY_US_FORMAT(R500_OUT_FMT_C4_8, R, G, B, A, 0); + case MESA_FORMAT_ARGB8888: // x + return EASY_US_FORMAT(R500_OUT_FMT_C4_8, B, G, R, A, 0); + case MESA_FORMAT_ARGB8888_REV: + return EASY_US_FORMAT(R500_OUT_FMT_C4_8, A, R, G, B, 0); + case MESA_FORMAT_XRGB8888: + return EASY_US_FORMAT(R500_OUT_FMT_C4_8, A, R, G, B, 0); + + case MESA_FORMAT_RGB332: + return EASY_US_FORMAT(R500_OUT_FMT_C_3_3_2, A, R, G, B, 0); + + case MESA_FORMAT_RGBA_FLOAT32: + return EASY_US_FORMAT(R500_OUT_FMT_C4_32_FP, R, G, B, A, 0); + case MESA_FORMAT_RGBA_FLOAT16: + return EASY_US_FORMAT(R500_OUT_FMT_C4_16_FP, R, G, B, A, 0); + case MESA_FORMAT_ALPHA_FLOAT32: + return EASY_US_FORMAT(R500_OUT_FMT_C_32_FP, A, A, A, A, 0); + case MESA_FORMAT_ALPHA_FLOAT16: + return EASY_US_FORMAT(R500_OUT_FMT_C_16_FP, A, A, A, A, 0); + + case MESA_FORMAT_SIGNED_RGBA8888: + return EASY_US_FORMAT(R500_OUT_FMT_C4_8, R, G, B, A, 0xf); + case MESA_FORMAT_SIGNED_RGBA8888_REV: + return EASY_US_FORMAT(R500_OUT_FMT_C4_8, A, B, G, R, 0xf); + case MESA_FORMAT_SIGNED_RGBA_16: + return EASY_US_FORMAT(R500_OUT_FMT_C4_16, R, G, B, A, 0xf); + + default: + assert(!"Invalid format for US output\n"); + return 0; + } +} +#undef EASY_US_FORMAT + +static void r500_emit_fp_setup(struct r300_context *r300, + struct r500_fragment_program_code *fp, + gl_format dst_format) +{ + r500_emit_fp(r300, (uint32_t *)fp->inst, (fp->inst_end + 1) * 6, 0, 0, 0); + BATCH_LOCALS(&r300->radeon); + + BEGIN_BATCH(10); + OUT_BATCH_REGSEQ(R500_US_CODE_ADDR, 3); + OUT_BATCH(R500_US_CODE_START_ADDR(0) | R500_US_CODE_END_ADDR(fp->inst_end)); + OUT_BATCH(R500_US_CODE_RANGE_ADDR(0) | R500_US_CODE_RANGE_SIZE(fp->inst_end)); + OUT_BATCH(0); + OUT_BATCH_REGVAL(R500_US_CONFIG, 0); + OUT_BATCH_REGVAL(R500_US_OUT_FMT_0, mesa_format_to_us_format(dst_format)); + OUT_BATCH_REGVAL(R500_US_PIXSIZE, fp->max_temp_idx); + END_BATCH(); +} + +static void r500_emit_rs_setup(struct r300_context *r300) +{ + BATCH_LOCALS(&r300->radeon); + + BEGIN_BATCH(7); + OUT_BATCH_REGSEQ(R300_RS_COUNT, 2); + OUT_BATCH((4 << R300_IT_COUNT_SHIFT) | R300_HIRES_EN); + OUT_BATCH(0); + OUT_BATCH_REGVAL(R500_RS_INST_0, + (0 << R500_RS_INST_TEX_ID_SHIFT) | + (0 << R500_RS_INST_TEX_ADDR_SHIFT) | + R500_RS_INST_TEX_CN_WRITE | + R500_RS_INST_COL_CN_NO_WRITE); + OUT_BATCH_REGVAL(R500_RS_IP_0, + (0 << R500_RS_IP_TEX_PTR_S_SHIFT) | + (1 << R500_RS_IP_TEX_PTR_T_SHIFT) | + (2 << R500_RS_IP_TEX_PTR_R_SHIFT) | + (3 << R500_RS_IP_TEX_PTR_Q_SHIFT)); + END_BATCH(); +} + +static void r300_emit_fp_setup(struct r300_context *r300, + struct r300_fragment_program_code *code, + gl_format dst_format) +{ + unsigned i; + BATCH_LOCALS(&r300->radeon); + + BEGIN_BATCH((code->alu.length + 1) * 4 + code->tex.length + 1 + 11); + + OUT_BATCH_REGSEQ(R300_US_ALU_RGB_INST_0, code->alu.length); + for (i = 0; i < code->alu.length; i++) { + OUT_BATCH(code->alu.inst[i].rgb_inst); + } + OUT_BATCH_REGSEQ(R300_US_ALU_RGB_ADDR_0, code->alu.length); + for (i = 0; i < code->alu.length; i++) { + OUT_BATCH(code->alu.inst[i].rgb_addr); + } + OUT_BATCH_REGSEQ(R300_US_ALU_ALPHA_INST_0, code->alu.length); + for (i = 0; i < code->alu.length; i++) { + OUT_BATCH(code->alu.inst[i].alpha_inst); + } + OUT_BATCH_REGSEQ(R300_US_ALU_ALPHA_ADDR_0, code->alu.length); + for (i = 0; i < code->alu.length; i++) { + OUT_BATCH(code->alu.inst[i].alpha_addr); + } + + OUT_BATCH_REGSEQ(R300_US_TEX_INST_0, code->tex.length); + OUT_BATCH_TABLE(code->tex.inst, code->tex.length); + + OUT_BATCH_REGSEQ(R300_US_CONFIG, 3); + OUT_BATCH(R300_PFS_CNTL_FIRST_NODE_HAS_TEX); + OUT_BATCH(code->pixsize); + OUT_BATCH(code->code_offset); + OUT_BATCH_REGSEQ(R300_US_CODE_ADDR_0, 4); + OUT_BATCH_TABLE(code->code_addr, 4); + OUT_BATCH_REGVAL(R500_US_OUT_FMT_0, mesa_format_to_us_format(dst_format)); + END_BATCH(); +} + +static void r300_emit_rs_setup(struct r300_context *r300) +{ + BATCH_LOCALS(&r300->radeon); + + BEGIN_BATCH(7); + OUT_BATCH_REGSEQ(R300_RS_COUNT, 2); + OUT_BATCH((4 << R300_IT_COUNT_SHIFT) | R300_HIRES_EN); + OUT_BATCH(0); + OUT_BATCH_REGVAL(R300_RS_INST_0, + R300_RS_INST_TEX_ID(0) | + R300_RS_INST_TEX_ADDR(0) | + R300_RS_INST_TEX_CN_WRITE); + OUT_BATCH_REGVAL(R300_RS_IP_0, + R300_RS_TEX_PTR(0) | + R300_RS_SEL_S(R300_RS_SEL_C0) | + R300_RS_SEL_R(R300_RS_SEL_C1) | + R300_RS_SEL_T(R300_RS_SEL_K0) | + R300_RS_SEL_Q(R300_RS_SEL_K1)); + END_BATCH(); +} + +static void emit_pvs_setup(struct r300_context *r300, + uint32_t *vp_code, + unsigned vp_len) +{ + BATCH_LOCALS(&r300->radeon); + + r300_emit_vpu(r300, vp_code, vp_len * 4, R300_PVS_CODE_START); + + BEGIN_BATCH(4); + OUT_BATCH_REGSEQ(R300_VAP_PVS_CODE_CNTL_0, 3); + OUT_BATCH((0 << R300_PVS_FIRST_INST_SHIFT) | + ((vp_len - 1) << R300_PVS_XYZW_VALID_INST_SHIFT) | + ((vp_len - 1)<< R300_PVS_LAST_INST_SHIFT)); + OUT_BATCH(0); + OUT_BATCH((vp_len - 1) << R300_PVS_LAST_VTX_SRC_INST_SHIFT); + END_BATCH(); +} + +static void emit_vap_setup(struct r300_context *r300, unsigned width, unsigned height) +{ + BATCH_LOCALS(&r300->radeon); + + BEGIN_BATCH(12); + OUT_BATCH_REGSEQ(R300_SE_VTE_CNTL, 2); + OUT_BATCH(R300_VTX_XY_FMT | R300_VTX_Z_FMT); + OUT_BATCH(4); + + OUT_BATCH_REGVAL(R300_VAP_PSC_SGN_NORM_CNTL, 0xaaaaaaaa); + OUT_BATCH_REGVAL(R300_VAP_PROG_STREAM_CNTL_0, + ((R300_DATA_TYPE_FLOAT_2 | (0 << R300_DST_VEC_LOC_SHIFT)) << 0) | + (((1 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_2 | R300_LAST_VEC) << 16)); + OUT_BATCH_REGVAL(R300_VAP_PROG_STREAM_CNTL_EXT_0, + ((((R300_SWIZZLE_SELECT_X << R300_SWIZZLE_SELECT_X_SHIFT) | + (R300_SWIZZLE_SELECT_Y << R300_SWIZZLE_SELECT_Y_SHIFT) | + (R300_SWIZZLE_SELECT_FP_ZERO << R300_SWIZZLE_SELECT_Z_SHIFT) | + (R300_SWIZZLE_SELECT_FP_ONE << R300_SWIZZLE_SELECT_W_SHIFT) | + (0xf << R300_WRITE_ENA_SHIFT) ) << 0) | + (((R300_SWIZZLE_SELECT_X << R300_SWIZZLE_SELECT_X_SHIFT) | + (R300_SWIZZLE_SELECT_Y << R300_SWIZZLE_SELECT_Y_SHIFT) | + (R300_SWIZZLE_SELECT_FP_ZERO << R300_SWIZZLE_SELECT_Z_SHIFT) | + (R300_SWIZZLE_SELECT_FP_ONE << R300_SWIZZLE_SELECT_W_SHIFT) | + (0xf << R300_WRITE_ENA_SHIFT) ) << 16) ) ); + OUT_BATCH_REGSEQ(R300_VAP_OUTPUT_VTX_FMT_0, 2); + OUT_BATCH(R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT); + OUT_BATCH(R300_VAP_OUTPUT_VTX_FMT_1__4_COMPONENTS); + END_BATCH(); +} + +static GLboolean validate_buffers(struct r300_context *r300, + struct radeon_bo *src_bo, + struct radeon_bo *dst_bo) +{ + int ret; + radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs, + src_bo, RADEON_GEM_DOMAIN_VRAM, 0); + + radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs, + dst_bo, 0, RADEON_GEM_DOMAIN_VRAM); + + ret = radeon_cs_space_check_with_bo(r300->radeon.cmdbuf.cs, + first_elem(&r300->radeon.dma.reserved)->bo, + RADEON_GEM_DOMAIN_GTT, 0); + if (ret) + return GL_FALSE; + + return GL_TRUE; +} + +static void emit_draw_packet(struct r300_context *r300, float width, float height) +{ + float verts[] = { 0.0, 0.0, 0.0, 1.0, + 0.0, height, 0.0, 0.0, + width, height, 1.0, 0.0, + width, 0.0, 1.0, 1.0 }; + + BATCH_LOCALS(&r300->radeon); + + BEGIN_BATCH(19); + OUT_BATCH_PACKET3(R300_PACKET3_3D_DRAW_IMMD_2, 16); + OUT_BATCH(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_EMBEDDED | + (4 << 16) | R300_VAP_VF_CNTL__PRIM_QUADS); + OUT_BATCH_TABLE(verts, 16); + END_BATCH(); +} + +static void other_stuff(struct r300_context *r300) +{ + BATCH_LOCALS(&r300->radeon); + + BEGIN_BATCH(15); + OUT_BATCH_REGVAL(R300_GA_POLY_MODE, + R300_GA_POLY_MODE_FRONT_PTYPE_TRI | R300_GA_POLY_MODE_BACK_PTYPE_TRI); + OUT_BATCH_REGVAL(R300_SU_CULL_MODE, R300_FRONT_FACE_CCW); + OUT_BATCH_REGVAL(R300_FG_FOG_BLEND, 0); + OUT_BATCH_REGVAL(R300_FG_ALPHA_FUNC, 0); + OUT_BATCH_REGSEQ(R300_RB3D_CBLEND, 2); + OUT_BATCH(0x0); + OUT_BATCH(0x0); + OUT_BATCH_REGVAL(R300_VAP_CLIP_CNTL, R300_CLIP_DISABLE); + OUT_BATCH_REGVAL(R300_ZB_CNTL, 0); + END_BATCH(); +} + +static void emit_cb_setup(struct r300_context *r300, + struct radeon_bo *bo, + intptr_t offset, + gl_format mesa_format, + unsigned width, + unsigned height) +{ + BATCH_LOCALS(&r300->radeon); + + unsigned x1, y1, x2, y2; + x1 = 0; + y1 = 0; + x2 = width - 1; + y2 = height - 1; + + if (r300->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV515) { + x1 += R300_SCISSORS_OFFSET; + y1 += R300_SCISSORS_OFFSET; + x2 += R300_SCISSORS_OFFSET; + y2 += R300_SCISSORS_OFFSET; + } + + r300_emit_cb_setup(r300, bo, offset, mesa_format, + _mesa_get_format_bytes(mesa_format), + _mesa_format_row_stride(mesa_format, width)); + + BEGIN_BATCH_NO_AUTOSTATE(5); + OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2); + OUT_BATCH((x1 << R300_SCISSORS_X_SHIFT)|(y1 << R300_SCISSORS_Y_SHIFT)); + OUT_BATCH((x2 << R300_SCISSORS_X_SHIFT)|(y2 << R300_SCISSORS_Y_SHIFT)); + OUT_BATCH_REGVAL(R300_RB3D_CCTL, 0); + END_BATCH(); +} + +GLboolean r300_blit(struct r300_context *r300, + struct radeon_bo *src_bo, + intptr_t src_offset, + gl_format src_mesaformat, + unsigned src_pitch, + unsigned src_width, + unsigned src_height, + struct radeon_bo *dst_bo, + intptr_t dst_offset, + gl_format dst_mesaformat, + unsigned dst_width, + unsigned dst_height) +{ + //assert(src_width == dst_width); + //assert(src_height == dst_height); + + if (src_bo == dst_bo) { + return GL_FALSE; + } + + //return GL_FALSE; + + if (0) { + fprintf(stderr, "src: width %d, height %d, pitch %d vs %d, format %s\n", + src_width, src_height, src_pitch, + _mesa_format_row_stride(src_mesaformat, src_width), + _mesa_get_format_name(src_mesaformat)); + fprintf(stderr, "dst: width %d, height %d, pitch %d, format %s\n", + dst_width, dst_height, + _mesa_format_row_stride(dst_mesaformat, dst_width), + _mesa_get_format_name(dst_mesaformat)); + } + + if (!validate_buffers(r300, src_bo, dst_bo)) + return GL_FALSE; + + rcommonEnsureCmdBufSpace(&r300->radeon, 200, __FUNCTION__); + + other_stuff(r300); + + r300_emit_tx_setup(r300, src_mesaformat, src_bo, src_offset, src_width, src_height, src_pitch); + + if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) { + r500_emit_fp_setup(r300, &r300->blit.fp_code.code.r500, dst_mesaformat); + r500_emit_rs_setup(r300); + } else { + r300_emit_fp_setup(r300, &r300->blit.fp_code.code.r300, dst_mesaformat); + r300_emit_rs_setup(r300); + } + + emit_pvs_setup(r300, r300->blit.vp_code.body.d, 2); + emit_vap_setup(r300, dst_width, dst_height); + + emit_cb_setup(r300, dst_bo, dst_offset, dst_mesaformat, dst_width, dst_height); + + emit_draw_packet(r300, dst_width, dst_height); + + r300EmitCacheFlush(r300); + + radeonFlush(r300->radeon.glCtx); + //r300ResetHwState(r300); + + return GL_TRUE; +}
\ No newline at end of file diff --git a/src/mesa/drivers/dri/r300/r300_blit.h b/src/mesa/drivers/dri/r300/r300_blit.h new file mode 100644 index 00000000000..29c5aa95149 --- /dev/null +++ b/src/mesa/drivers/dri/r300/r300_blit.h @@ -0,0 +1,46 @@ +/* + * Copyright (C) 2009 Maciej Cencora <[email protected]> + * + * 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, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef R300_BLIT_H +#define R300_BLIT_H + +void r300_blit_init(struct r300_context *r300); + +GLboolean r300_blit(struct r300_context *r300, + struct radeon_bo *src_bo, + intptr_t src_offset, + gl_format src_mesaformat, + unsigned src_pitch, + unsigned src_width, + unsigned src_height, + struct radeon_bo *dst_bo, + intptr_t dst_offset, + gl_format dst_mesaformat, + unsigned dst_width, + unsigned dst_height); + +#endif // R300_BLIT_H
\ No newline at end of file diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.c b/src/mesa/drivers/dri/r300/r300_cmdbuf.c index da5b7ba642f..e1c33bbb2cf 100644 --- a/src/mesa/drivers/dri/r300/r300_cmdbuf.c +++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.c @@ -45,15 +45,12 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "radeon_drm.h" #include "r300_context.h" -#include "r300_ioctl.h" -#include "radeon_reg.h" #include "r300_reg.h" #include "r300_cmdbuf.h" #include "r300_emit.h" #include "radeon_bocs_wrapper.h" #include "radeon_mipmap_tree.h" #include "r300_state.h" -#include "radeon_reg.h" #include "radeon_queryobj.h" /** # of dwords reserved for additional instructions that may need to be written @@ -74,7 +71,7 @@ static unsigned packet0_count(r300ContextPtr r300, uint32_t *pkt) #define vpu_count(ptr) (((drm_r300_cmd_header_t*)(ptr))->vpu.count) #define r500fp_count(ptr) (((drm_r300_cmd_header_t*)(ptr))->r500fp.count) -int check_vpu(GLcontext *ctx, struct radeon_state_atom *atom) +static int check_vpu(GLcontext *ctx, struct radeon_state_atom *atom) { r300ContextPtr r300 = R300_CONTEXT(ctx); int cnt; @@ -88,54 +85,73 @@ int check_vpu(GLcontext *ctx, struct radeon_state_atom *atom) return cnt ? (cnt * 4) + extra : 0; } +void r300_emit_vpu(struct r300_context *r300, + uint32_t *data, + unsigned len, + uint32_t addr) +{ + BATCH_LOCALS(&r300->radeon); + + BEGIN_BATCH_NO_AUTOSTATE(5 + len); + OUT_BATCH_REGVAL(R300_VAP_PVS_STATE_FLUSH_REG, 0); + OUT_BATCH_REGVAL(R300_VAP_PVS_VECTOR_INDX_REG, addr); + OUT_BATCH(CP_PACKET0(R300_VAP_PVS_UPLOAD_DATA, len-1) | RADEON_ONE_REG_WR); + OUT_BATCH_TABLE(data, len); + END_BATCH(); +} -void emit_vpu(GLcontext *ctx, struct radeon_state_atom * atom) +static void emit_vpu_state(GLcontext *ctx, struct radeon_state_atom * atom) { - r300ContextPtr r300 = R300_CONTEXT(ctx); - BATCH_LOCALS(&r300->radeon); - drm_r300_cmd_header_t cmd; - uint32_t addr, ndw; + r300ContextPtr r300 = R300_CONTEXT(ctx); + drm_r300_cmd_header_t cmd; + uint32_t addr, ndw; + + cmd.u = atom->cmd[0]; + addr = (cmd.vpu.adrhi << 8) | cmd.vpu.adrlo; + ndw = atom->check(ctx, atom); - cmd.u = atom->cmd[0]; - addr = (cmd.vpu.adrhi << 8) | cmd.vpu.adrlo; - ndw = atom->check(ctx, atom); + r300_emit_vpu(r300, &atom->cmd[1], vpu_count(atom->cmd) * 4, addr); +} + +void r500_emit_fp(struct r300_context *r300, + uint32_t *data, + unsigned len, + uint32_t addr, + unsigned type, + unsigned clamp) +{ + BATCH_LOCALS(&r300->radeon); - BEGIN_BATCH_NO_AUTOSTATE(ndw); + addr |= (type << 16); + addr |= (clamp << 17); - ndw -= 5; - OUT_BATCH_REGVAL(R300_VAP_PVS_VECTOR_INDX_REG, addr); - OUT_BATCH(CP_PACKET0(R300_VAP_PVS_UPLOAD_DATA, ndw-1) | RADEON_ONE_REG_WR); - OUT_BATCH_TABLE(&atom->cmd[1], ndw); - OUT_BATCH_REGVAL(R300_VAP_PVS_STATE_FLUSH_REG, 0); - END_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(len + 3); + OUT_BATCH(CP_PACKET0(R500_GA_US_VECTOR_INDEX, 0)); + OUT_BATCH(addr); + OUT_BATCH(CP_PACKET0(R500_GA_US_VECTOR_DATA, len-1) | RADEON_ONE_REG_WR); + OUT_BATCH_TABLE(data, len); + END_BATCH(); } -void emit_r500fp(GLcontext *ctx, struct radeon_state_atom * atom) +static void emit_r500fp_atom(GLcontext *ctx, struct radeon_state_atom * atom) { - r300ContextPtr r300 = R300_CONTEXT(ctx); - BATCH_LOCALS(&r300->radeon); - drm_r300_cmd_header_t cmd; - uint32_t addr, ndw, sz; - int type, clamp; - - ndw = atom->check(ctx, atom); - - cmd.u = atom->cmd[0]; - sz = cmd.r500fp.count; - addr = ((cmd.r500fp.adrhi_flags & 1) << 8) | cmd.r500fp.adrlo; - type = !!(cmd.r500fp.adrhi_flags & R500FP_CONSTANT_TYPE); - clamp = !!(cmd.r500fp.adrhi_flags & R500FP_CONSTANT_CLAMP); - - addr |= (type << 16); - addr |= (clamp << 17); - - BEGIN_BATCH_NO_AUTOSTATE(ndw); - OUT_BATCH(CP_PACKET0(R500_GA_US_VECTOR_INDEX, 0)); - OUT_BATCH(addr); - ndw-=3; - OUT_BATCH(CP_PACKET0(R500_GA_US_VECTOR_DATA, ndw-1) | RADEON_ONE_REG_WR); - OUT_BATCH_TABLE(&atom->cmd[1], ndw); - END_BATCH(); + r300ContextPtr r300 = R300_CONTEXT(ctx); + drm_r300_cmd_header_t cmd; + uint32_t addr, count; + int type, clamp; + + cmd.u = atom->cmd[0]; + addr = ((cmd.r500fp.adrhi_flags & 1) << 8) | cmd.r500fp.adrlo; + type = !!(cmd.r500fp.adrhi_flags & R500FP_CONSTANT_TYPE); + clamp = !!(cmd.r500fp.adrhi_flags & R500FP_CONSTANT_CLAMP); + + if (type) { + count = r500fp_count(atom->cmd) * 4; + } else { + count = r500fp_count(atom->cmd) * 6; + } + + r500_emit_fp(r300, &atom->cmd[1], count, addr, type, clamp); } static int check_tex_offsets(GLcontext *ctx, struct radeon_state_atom * atom) @@ -171,7 +187,7 @@ static void emit_tex_offsets(GLcontext *ctx, struct radeon_state_atom * atom) if (t && !t->image_override) { BEGIN_BATCH_NO_AUTOSTATE(4); OUT_BATCH_REGSEQ(R300_TX_OFFSET_0 + (i * 4), 1); - OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, 0, + OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, get_base_teximage_offset(t), RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0); END_BATCH(); } else if (!t) { @@ -258,93 +274,136 @@ static int check_cb_offset(GLcontext *ctx, struct radeon_state_atom * atom) return dw; } -static void emit_cb_offset(GLcontext *ctx, struct radeon_state_atom * atom) +static void emit_scissor(struct r300_context *r300, + unsigned width, + unsigned height) { - r300ContextPtr r300 = R300_CONTEXT(ctx); - BATCH_LOCALS(&r300->radeon); - struct radeon_renderbuffer *rrb; - uint32_t cbpitch; - uint32_t offset = r300->radeon.state.color.draw_offset; - uint32_t dw = 6; - int i; + int i; + BATCH_LOCALS(&r300->radeon); + if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) { + BEGIN_BATCH_NO_AUTOSTATE(3); + OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2); + OUT_BATCH(0); + OUT_BATCH(((width - 1) << R300_SCISSORS_X_SHIFT) | + ((height - 1) << R300_SCISSORS_Y_SHIFT)); + END_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(16); + for (i = 0; i < 4; i++) { + OUT_BATCH_REGSEQ(R300_SC_CLIPRECT_TL_0 + (i * 8), 2); + OUT_BATCH((0 << R300_CLIPRECT_X_SHIFT) | (0 << R300_CLIPRECT_Y_SHIFT)); + OUT_BATCH(((width - 1) << R300_CLIPRECT_X_SHIFT) | ((height - 1) << R300_CLIPRECT_Y_SHIFT)); + } + OUT_BATCH_REGSEQ(R300_SC_CLIP_RULE, 1); + OUT_BATCH(0xAAAA); + OUT_BATCH_REGSEQ(R300_SC_SCREENDOOR, 1); + OUT_BATCH(0xffffff); + END_BATCH(); + } else { + BEGIN_BATCH_NO_AUTOSTATE(3); + OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2); + OUT_BATCH((R300_SCISSORS_OFFSET << R300_SCISSORS_X_SHIFT) | + (R300_SCISSORS_OFFSET << R300_SCISSORS_Y_SHIFT)); + OUT_BATCH(((width + R300_SCISSORS_OFFSET - 1) << R300_SCISSORS_X_SHIFT) | + ((height + R300_SCISSORS_OFFSET - 1) << R300_SCISSORS_Y_SHIFT)); + END_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(16); + for (i = 0; i < 4; i++) { + OUT_BATCH_REGSEQ(R300_SC_CLIPRECT_TL_0 + (i * 8), 2); + OUT_BATCH((R300_SCISSORS_OFFSET << R300_CLIPRECT_X_SHIFT) | (R300_SCISSORS_OFFSET << R300_CLIPRECT_Y_SHIFT)); + OUT_BATCH(((R300_SCISSORS_OFFSET + width - 1) << R300_CLIPRECT_X_SHIFT) | + ((R300_SCISSORS_OFFSET + height - 1) << R300_CLIPRECT_Y_SHIFT)); + } + OUT_BATCH_REGSEQ(R300_SC_CLIP_RULE, 1); + OUT_BATCH(0xAAAA); + OUT_BATCH_REGSEQ(R300_SC_SCREENDOOR, 1); + OUT_BATCH(0xffffff); + END_BATCH(); + } +} - rrb = radeon_get_colorbuffer(&r300->radeon); - if (!rrb || !rrb->bo) { - fprintf(stderr, "no rrb\n"); - return; - } +void r300_emit_cb_setup(struct r300_context *r300, + struct radeon_bo *bo, + uint32_t offset, + GLuint format, + unsigned cpp, + unsigned pitch) +{ + BATCH_LOCALS(&r300->radeon); + uint32_t cbpitch = pitch / cpp; + uint32_t dw = 6; + + assert(offset % 32 == 0); + + switch (format) { + case MESA_FORMAT_RGB565: + assert(_mesa_little_endian()); + cbpitch |= R300_COLOR_FORMAT_RGB565; + break; + case MESA_FORMAT_RGB565_REV: + assert(!_mesa_little_endian()); + cbpitch |= R300_COLOR_FORMAT_RGB565; + break; + case MESA_FORMAT_ARGB4444: + assert(_mesa_little_endian()); + cbpitch |= R300_COLOR_FORMAT_ARGB4444; + break; + case MESA_FORMAT_ARGB4444_REV: + assert(!_mesa_little_endian()); + cbpitch |= R300_COLOR_FORMAT_ARGB4444; + break; + case MESA_FORMAT_ARGB1555: + assert(_mesa_little_endian()); + cbpitch |= R300_COLOR_FORMAT_ARGB1555; + break; + case MESA_FORMAT_ARGB1555_REV: + assert(!_mesa_little_endian()); + cbpitch |= R300_COLOR_FORMAT_ARGB1555; + break; + default: + if (cpp == 4) { + cbpitch |= R300_COLOR_FORMAT_ARGB8888; + } else { + _mesa_problem(r300->radeon.glCtx, "unexpected format in emit_cb_offset()");; + } + break; + } - if (RADEON_DEBUG & RADEON_STATE) - fprintf(stderr,"rrb is %p %d %dx%d\n", rrb, offset, rrb->base.Width, rrb->base.Height); - cbpitch = (rrb->pitch / rrb->cpp); - if (rrb->cpp == 4) - cbpitch |= R300_COLOR_FORMAT_ARGB8888; - else switch (rrb->base._ActualFormat) { - case GL_RGB5: - cbpitch |= R300_COLOR_FORMAT_RGB565; - break; - case GL_RGBA4: - cbpitch |= R300_COLOR_FORMAT_ARGB4444; - break; - case GL_RGB5_A1: - cbpitch |= R300_COLOR_FORMAT_ARGB1555; - break; - } + if (bo->flags & RADEON_BO_FLAGS_MACRO_TILE) + cbpitch |= R300_COLOR_TILE_ENABLE; - if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE) - cbpitch |= R300_COLOR_TILE_ENABLE; + if (r300->radeon.radeonScreen->kernel_mm) + dw += 2; + + BEGIN_BATCH_NO_AUTOSTATE(dw); + OUT_BATCH_REGSEQ(R300_RB3D_COLOROFFSET0, 1); + OUT_BATCH_RELOC(offset, bo, offset, 0, RADEON_GEM_DOMAIN_VRAM, 0); + OUT_BATCH_REGSEQ(R300_RB3D_COLORPITCH0, 1); + if (!r300->radeon.radeonScreen->kernel_mm) + OUT_BATCH(cbpitch); + else + OUT_BATCH_RELOC(cbpitch, bo, cbpitch, 0, RADEON_GEM_DOMAIN_VRAM, 0); + END_BATCH(); +} + +static void emit_cb_offset_atom(GLcontext *ctx, struct radeon_state_atom * atom) +{ + r300ContextPtr r300 = R300_CONTEXT(ctx); + struct radeon_renderbuffer *rrb; + uint32_t offset = r300->radeon.state.color.draw_offset; + + rrb = radeon_get_colorbuffer(&r300->radeon); + if (!rrb || !rrb->bo) { + fprintf(stderr, "no rrb\n"); + return; + } + + if (RADEON_DEBUG & RADEON_STATE) + fprintf(stderr,"rrb is %p %d %dx%d\n", rrb, offset, rrb->base.Width, rrb->base.Height); + + r300_emit_cb_setup(r300, rrb->bo, offset, rrb->base.Format, rrb->cpp, rrb->pitch); - if (r300->radeon.radeonScreen->kernel_mm) - dw += 2; - BEGIN_BATCH_NO_AUTOSTATE(dw); - OUT_BATCH_REGSEQ(R300_RB3D_COLOROFFSET0, 1); - OUT_BATCH_RELOC(offset, rrb->bo, offset, 0, RADEON_GEM_DOMAIN_VRAM, 0); - OUT_BATCH_REGSEQ(R300_RB3D_COLORPITCH0, 1); - if (!r300->radeon.radeonScreen->kernel_mm) - OUT_BATCH(cbpitch); - else - OUT_BATCH_RELOC(cbpitch, rrb->bo, cbpitch, 0, RADEON_GEM_DOMAIN_VRAM, 0); - END_BATCH(); if (r300->radeon.radeonScreen->driScreen->dri2.enabled) { - if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) { - BEGIN_BATCH_NO_AUTOSTATE(3); - OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2); - OUT_BATCH(0); - OUT_BATCH(((rrb->base.Width - 1) << R300_SCISSORS_X_SHIFT) | - ((rrb->base.Height - 1) << R300_SCISSORS_Y_SHIFT)); - END_BATCH(); - BEGIN_BATCH_NO_AUTOSTATE(16); - for (i = 0; i < 4; i++) { - OUT_BATCH_REGSEQ(R300_SC_CLIPRECT_TL_0 + (i * 8), 2); - OUT_BATCH((0 << R300_CLIPRECT_X_SHIFT) | (0 << R300_CLIPRECT_Y_SHIFT)); - OUT_BATCH(((rrb->base.Width - 1) << R300_CLIPRECT_X_SHIFT) | ((rrb->base.Height - 1) << R300_CLIPRECT_Y_SHIFT)); - } - OUT_BATCH_REGSEQ(R300_SC_CLIP_RULE, 1); - OUT_BATCH(0xAAAA); - OUT_BATCH_REGSEQ(R300_SC_SCREENDOOR, 1); - OUT_BATCH(0xffffff); - END_BATCH(); - } else { - BEGIN_BATCH_NO_AUTOSTATE(3); - OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2); - OUT_BATCH((R300_SCISSORS_OFFSET << R300_SCISSORS_X_SHIFT) | - (R300_SCISSORS_OFFSET << R300_SCISSORS_Y_SHIFT)); - OUT_BATCH(((rrb->base.Width + R300_SCISSORS_OFFSET - 1) << R300_SCISSORS_X_SHIFT) | - ((rrb->base.Height + R300_SCISSORS_OFFSET - 1) << R300_SCISSORS_Y_SHIFT)); - END_BATCH(); - BEGIN_BATCH_NO_AUTOSTATE(16); - for (i = 0; i < 4; i++) { - OUT_BATCH_REGSEQ(R300_SC_CLIPRECT_TL_0 + (i * 8), 2); - OUT_BATCH((R300_SCISSORS_OFFSET << R300_CLIPRECT_X_SHIFT) | (R300_SCISSORS_OFFSET << R300_CLIPRECT_Y_SHIFT)); - OUT_BATCH(((R300_SCISSORS_OFFSET + rrb->base.Width - 1) << R300_CLIPRECT_X_SHIFT) | - ((R300_SCISSORS_OFFSET + rrb->base.Height - 1) << R300_CLIPRECT_Y_SHIFT)); - } - OUT_BATCH_REGSEQ(R300_SC_CLIP_RULE, 1); - OUT_BATCH(0xAAAA); - OUT_BATCH_REGSEQ(R300_SC_SCREENDOOR, 1); - OUT_BATCH(0xffffff); - END_BATCH(); - } + emit_scissor(r300, rrb->base.Width, rrb->base.Height); } } @@ -440,7 +499,7 @@ static int check_variable(GLcontext *ctx, struct radeon_state_atom *atom) return cnt ? cnt + 1 : 0; } -int check_r500fp(GLcontext *ctx, struct radeon_state_atom *atom) +static int check_r500fp(GLcontext *ctx, struct radeon_state_atom *atom) { int cnt; r300ContextPtr r300 = R300_CONTEXT(ctx); @@ -452,7 +511,7 @@ int check_r500fp(GLcontext *ctx, struct radeon_state_atom *atom) return cnt ? (cnt * 6) + extra : 0; } -int check_r500fp_const(GLcontext *ctx, struct radeon_state_atom *atom) +static int check_r500fp_const(GLcontext *ctx, struct radeon_state_atom *atom) { int cnt; r300ContextPtr r300 = R300_CONTEXT(ctx); @@ -629,13 +688,13 @@ void r300InitCmdBuf(r300ContextPtr r300) r300->hw.r500fp.cmd[R300_FPI_CMD_0] = cmdr500fp(r300->radeon.radeonScreen, 0, 0, 0, 0); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.r500fp.emit = emit_r500fp; + r300->hw.r500fp.emit = emit_r500fp_atom; ALLOC_STATE(r500fp_const, r500fp_const, R500_FPP_CMDSIZE, 0); r300->hw.r500fp_const.cmd[R300_FPI_CMD_0] = cmdr500fp(r300->radeon.radeonScreen, 0, 0, 1, 0); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.r500fp_const.emit = emit_r500fp; + r300->hw.r500fp_const.emit = emit_r500fp_atom; } else { ALLOC_STATE(fp, always, R300_FP_CMDSIZE, 0); r300->hw.fp.cmd[R300_FP_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_US_CONFIG, 3); @@ -679,7 +738,7 @@ void r300InitCmdBuf(r300ContextPtr r300) ALLOC_STATE(rop, always, 2, 0); r300->hw.rop.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_RB3D_ROPCNTL, 1); ALLOC_STATE(cb, cb_offset, R300_CB_CMDSIZE, 0); - r300->hw.cb.emit = &emit_cb_offset; + r300->hw.cb.emit = &emit_cb_offset_atom; ALLOC_STATE(rb3d_dither_ctl, always, 10, 0); r300->hw.rb3d_dither_ctl.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_RB3D_DITHER_CTL, 9); ALLOC_STATE(rb3d_aaresolve_ctl, always, 2, 0); @@ -743,20 +802,20 @@ void r300InitCmdBuf(r300ContextPtr r300) r300->hw.vpi.cmd[0] = cmdvpu(r300->radeon.radeonScreen, R300_PVS_CODE_START, 0); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.vpi.emit = emit_vpu; + r300->hw.vpi.emit = emit_vpu_state; if (is_r500) { ALLOC_STATE(vpp, vpu, R300_VPP_CMDSIZE, 0); r300->hw.vpp.cmd[0] = cmdvpu(r300->radeon.radeonScreen, R500_PVS_CONST_START, 0); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.vpp.emit = emit_vpu; + r300->hw.vpp.emit = emit_vpu_state; ALLOC_STATE(vps, vpu, R300_VPS_CMDSIZE, 0); r300->hw.vps.cmd[0] = cmdvpu(r300->radeon.radeonScreen, R500_POINT_VPORT_SCALE_OFFSET, 1); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.vps.emit = emit_vpu; + r300->hw.vps.emit = emit_vpu_state; for (i = 0; i < 6; i++) { ALLOC_STATE(vpucp[i], vpu, R300_VPUCP_CMDSIZE, 0); @@ -764,20 +823,20 @@ void r300InitCmdBuf(r300ContextPtr r300) cmdvpu(r300->radeon.radeonScreen, R500_PVS_UCP_START + i, 1); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.vpucp[i].emit = emit_vpu; + r300->hw.vpucp[i].emit = emit_vpu_state; } } else { ALLOC_STATE(vpp, vpu, R300_VPP_CMDSIZE, 0); r300->hw.vpp.cmd[0] = cmdvpu(r300->radeon.radeonScreen, R300_PVS_CONST_START, 0); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.vpp.emit = emit_vpu; + r300->hw.vpp.emit = emit_vpu_state; ALLOC_STATE(vps, vpu, R300_VPS_CMDSIZE, 0); r300->hw.vps.cmd[0] = cmdvpu(r300->radeon.radeonScreen, R300_POINT_VPORT_SCALE_OFFSET, 1); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.vps.emit = emit_vpu; + r300->hw.vps.emit = emit_vpu_state; for (i = 0; i < 6; i++) { ALLOC_STATE(vpucp[i], vpu, R300_VPUCP_CMDSIZE, 0); @@ -785,7 +844,7 @@ void r300InitCmdBuf(r300ContextPtr r300) cmdvpu(r300->radeon.radeonScreen, R300_PVS_UCP_START + i, 1); if (r300->radeon.radeonScreen->kernel_mm) - r300->hw.vpucp[i].emit = emit_vpu; + r300->hw.vpucp[i].emit = emit_vpu_state; } } } diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.h b/src/mesa/drivers/dri/r300/r300_cmdbuf.h index 1b703e518a0..0e68da928ed 100644 --- a/src/mesa/drivers/dri/r300/r300_cmdbuf.h +++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.h @@ -44,14 +44,26 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define FIREAOS_BUFSZ (3) #define SCISSORS_BUFSZ (3) -extern void r300InitCmdBuf(r300ContextPtr r300); +void r300InitCmdBuf(r300ContextPtr r300); void r300_emit_scissor(GLcontext *ctx); -void emit_vpu(GLcontext *ctx, struct radeon_state_atom * atom); -int check_vpu(GLcontext *ctx, struct radeon_state_atom *atom); +void r300_emit_vpu(struct r300_context *ctx, + uint32_t *data, + unsigned len, + uint32_t addr); -void emit_r500fp(GLcontext *ctx, struct radeon_state_atom * atom); -int check_r500fp(GLcontext *ctx, struct radeon_state_atom *atom); -int check_r500fp_const(GLcontext *ctx, struct radeon_state_atom *atom); +void r500_emit_fp(struct r300_context *r300, + uint32_t *data, + unsigned len, + uint32_t addr, + unsigned type, + unsigned clamp); -#endif /* __R300_CMDBUF_H__ */ +void r300_emit_cb_setup(struct r300_context *r300, + struct radeon_bo *bo, + uint32_t offset, + GLuint format, + unsigned cpp, + unsigned pitch); + +#endif /* __R300_CMDBUF_H__ */ diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c index 6fcf209af66..05005f61c30 100644 --- a/src/mesa/drivers/dri/r300/r300_context.c +++ b/src/mesa/drivers/dri/r300/r300_context.c @@ -55,13 +55,14 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "tnl/t_vp_build.h" #include "drivers/common/driverfuncs.h" +#include "drivers/common/meta.h" #include "r300_context.h" #include "radeon_context.h" #include "radeon_span.h" +#include "r300_blit.h" #include "r300_cmdbuf.h" #include "r300_state.h" -#include "r300_ioctl.h" #include "r300_tex.h" #include "r300_emit.h" #include "r300_swtcl.h" @@ -90,10 +91,11 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define need_GL_ATI_separate_stencil #define need_GL_NV_vertex_program -#include "extension_helper.h" +#include "main/remap_helper.h" +void r300_init_texcopy_functions(struct dd_function_table *table); -const struct dri_extension card_extensions[] = { +static const struct dri_extension card_extensions[] = { /* *INDENT-OFF* */ {"GL_ARB_depth_texture", NULL}, {"GL_ARB_fragment_program", NULL}, @@ -145,7 +147,7 @@ const struct dri_extension card_extensions[] = { }; -const struct dri_extension mm_extensions[] = { +static const struct dri_extension mm_extensions[] = { { "GL_EXT_framebuffer_blit", GL_EXT_framebuffer_blit_functions }, { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions }, { NULL, NULL } @@ -155,7 +157,7 @@ const struct dri_extension mm_extensions[] = { * The GL 2.0 functions are needed to make display lists work with * functions added by GL_ATI_separate_stencil. */ -const struct dri_extension gl_20_extension[] = { +static const struct dri_extension gl_20_extension[] = { {"GL_VERSION_2_0", GL_VERSION_2_0_functions }, }; @@ -376,11 +378,21 @@ static void r300InitConstValues(GLcontext *ctx, radeonScreenPtr screen) if (screen->chip_family >= CHIP_FAMILY_RV515) { ctx->Const.FragmentProgram.MaxNativeTemps = R500_PFS_NUM_TEMP_REGS; ctx->Const.FragmentProgram.MaxNativeAttribs = 11; /* copy i915... */ - ctx->Const.FragmentProgram.MaxNativeParameters = R500_PFS_NUM_CONST_REGS; - ctx->Const.FragmentProgram.MaxNativeAluInstructions = R500_PFS_MAX_INST; - ctx->Const.FragmentProgram.MaxNativeTexInstructions = R500_PFS_MAX_INST; - ctx->Const.FragmentProgram.MaxNativeInstructions = R500_PFS_MAX_INST; - ctx->Const.FragmentProgram.MaxNativeTexIndirections = R500_PFS_MAX_INST; + + /* The hardware limits are higher than this, + * but the non-KMS DRM interface artificially limits us + * to this many instructions. + * + * We could of course work around it in the KMS path, + * but it would be a mess, so it seems wiser + * to leave it as is. Going forward, the Gallium driver + * will not be subject to these limitations. + */ + ctx->Const.FragmentProgram.MaxNativeParameters = 255; + ctx->Const.FragmentProgram.MaxNativeAluInstructions = 255; + ctx->Const.FragmentProgram.MaxNativeTexInstructions = 255; + ctx->Const.FragmentProgram.MaxNativeInstructions = 255; + ctx->Const.FragmentProgram.MaxNativeTexIndirections = 255; ctx->Const.FragmentProgram.MaxNativeAddressRegs = 0; } else { ctx->Const.FragmentProgram.MaxNativeTemps = R300_PFS_NUM_TEMP_REGS; @@ -429,11 +441,11 @@ static void r300InitGLExtensions(GLcontext *ctx) if (r300->options.stencil_two_side_disabled) _mesa_disable_extension(ctx, "GL_EXT_stencil_two_side"); - if (r300->options.s3tc_force_enabled) { + if (r300->options.s3tc_force_disabled) { + _mesa_disable_extension(ctx, "GL_EXT_texture_compression_s3tc"); + } else if (ctx->Mesa_DXTn || r300->options.s3tc_force_enabled) { _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); _mesa_enable_extension(ctx, "GL_S3_s3tc"); - } else if (r300->options.s3tc_force_disabled) { - _mesa_disable_extension(ctx, "GL_EXT_texture_compression_s3tc"); } if (!r300->radeon.radeonScreen->drmSupportsOcclusionQueries) { @@ -441,6 +453,13 @@ static void r300InitGLExtensions(GLcontext *ctx) } } +static void r300InitIoctlFuncs(struct dd_function_table *functions) +{ + functions->Clear = _mesa_meta_Clear; + functions->Finish = radeonFinish; + functions->Flush = radeonFlush; +} + /* Create the device specific rendering context. */ GLboolean r300CreateContext(const __GLcontextModes * glVisual, @@ -467,6 +486,7 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual, r300_init_vtbl(&r300->radeon); _mesa_init_driver_functions(&functions); + r300_init_texcopy_functions(&functions); r300InitIoctlFuncs(&functions); r300InitStateFuncs(&functions); r300InitTextureFuncs(&functions); @@ -520,6 +540,7 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual, r300InitSwtcl(ctx); } + r300_blit_init(r300); radeon_fbo_init(&r300->radeon); radeonInitSpanFuncs( ctx ); r300InitCmdBuf(r300); diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h index 518d5cdbf4f..54a92a2e447 100644 --- a/src/mesa/drivers/dri/r300/r300_context.h +++ b/src/mesa/drivers/dri/r300/r300_context.h @@ -533,6 +533,11 @@ struct r300_context { uint32_t fallback; + struct { + struct r300_vertex_program_code vp_code; + struct rX00_fragment_program_code fp_code; + } blit; + DECLARE_RENDERINPUTS(render_inputs_bitset); }; @@ -549,6 +554,8 @@ extern void r300InitShaderFunctions(r300ContextPtr r300); extern void r300InitDraw(GLcontext *ctx); +extern void r300_init_texcopy_functions(struct dd_function_table *table); + #define r300PackFloat32 radeonPackFloat32 #define r300PackFloat24 radeonPackFloat24 diff --git a/src/mesa/drivers/dri/r300/r300_draw.c b/src/mesa/drivers/dri/r300/r300_draw.c index e9968f9ffef..3dcd986e226 100644 --- a/src/mesa/drivers/dri/r300/r300_draw.c +++ b/src/mesa/drivers/dri/r300/r300_draw.c @@ -100,7 +100,7 @@ static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer GLubyte *in = (GLubyte *)src_ptr; radeonAllocDmaRegion(&r300->radeon, &r300->ind_buf.bo, &r300->ind_buf.bo_offset, size, 4); - + radeon_bo_map(r300->ind_buf.bo, 1); assert(r300->ind_buf.bo->ptr != NULL); out = (GLuint *)ADD_POINTERS(r300->ind_buf.bo->ptr, r300->ind_buf.bo_offset); @@ -111,7 +111,7 @@ static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer if (i < mesa_ind_buf->count) { *out++ = in[i]; } - + radeon_bo_unmap(r300->ind_buf.bo); #if MESA_BIG_ENDIAN } else { /* if (mesa_ind_buf->type == GL_UNSIGNED_SHORT) */ GLushort *in = (GLushort *)src_ptr; @@ -120,6 +120,7 @@ static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer radeonAllocDmaRegion(&r300->radeon, &r300->ind_buf.bo, &r300->ind_buf.bo_offset, size, 4); + radeon_bo_map(r300->ind_buf.bo, 1); assert(r300->ind_buf.bo->ptr != NULL); out = (GLuint *)ADD_POINTERS(r300->ind_buf.bo->ptr, r300->ind_buf.bo_offset); @@ -130,6 +131,7 @@ static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer if (i < mesa_ind_buf->count) { *out++ = in[i]; } + radeon_bo_unmap(r300->ind_buf.bo); #endif } @@ -173,10 +175,12 @@ static void r300SetupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer radeonAllocDmaRegion(&r300->radeon, &r300->ind_buf.bo, &r300->ind_buf.bo_offset, size, 4); + radeon_bo_map(r300->ind_buf.bo, 1); assert(r300->ind_buf.bo->ptr != NULL); dst_ptr = ADD_POINTERS(r300->ind_buf.bo->ptr, r300->ind_buf.bo_offset); _mesa_memcpy(dst_ptr, src_ptr, size); + radeon_bo_unmap(r300->ind_buf.bo); r300->ind_buf.is_32bit = (mesa_ind_buf->type == GL_UNSIGNED_INT); r300->ind_buf.count = mesa_ind_buf->count; @@ -242,6 +246,7 @@ static void r300ConvertAttrib(GLcontext *ctx, int count, const struct gl_client_ } radeonAllocDmaRegion(&r300->radeon, &attr->bo, &attr->bo_offset, sizeof(GLfloat) * input->Size * count, 32); + radeon_bo_map(attr->bo, 1); dst_ptr = (GLfloat *)ADD_POINTERS(attr->bo->ptr, attr->bo_offset); radeon_print(RADEON_FALLBACKS, RADEON_IMPORTANT, @@ -280,6 +285,7 @@ static void r300ConvertAttrib(GLcontext *ctx, int count, const struct gl_client_ break; } + radeon_bo_unmap(attr->bo); if (mapped_named_bo) { ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER, input->BufferObj); } @@ -294,6 +300,8 @@ static void r300AlignDataToDword(GLcontext *ctx, const struct gl_client_array *i radeonAllocDmaRegion(&r300->radeon, &attr->bo, &attr->bo_offset, size, 32); + radeon_bo_map(attr->bo, 1); + if (!input->BufferObj->Pointer) { ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER, GL_READ_ONLY_ARB, input->BufferObj); mapped_named_bo = GL_TRUE; @@ -317,6 +325,7 @@ static void r300AlignDataToDword(GLcontext *ctx, const struct gl_client_array *i ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER, input->BufferObj); } + radeon_bo_unmap(attr->bo); attr->stride = dst_stride; } @@ -527,6 +536,7 @@ static void r300AllocDmaRegions(GLcontext *ctx, const struct gl_client_array *in } radeonAllocDmaRegion(&r300->radeon, &vbuf->attribs[index].bo, &vbuf->attribs[index].bo_offset, size, 32); + radeon_bo_map(vbuf->attribs[index].bo, 1); assert(vbuf->attribs[index].bo->ptr != NULL); dst = (uint32_t *)ADD_POINTERS(vbuf->attribs[index].bo->ptr, vbuf->attribs[index].bo_offset); switch (vbuf->attribs[index].dwords) { @@ -536,6 +546,7 @@ static void r300AllocDmaRegions(GLcontext *ctx, const struct gl_client_array *in case 4: radeonEmitVec16(dst, input[i]->Ptr, input[i]->StrideB, local_count); break; default: assert(0); break; } + radeon_bo_unmap(vbuf->attribs[index].bo); } } diff --git a/src/mesa/drivers/dri/r300/r300_emit.c b/src/mesa/drivers/dri/r300/r300_emit.c index 07e62230874..3759ca2bea7 100644 --- a/src/mesa/drivers/dri/r300/r300_emit.c +++ b/src/mesa/drivers/dri/r300/r300_emit.c @@ -49,7 +49,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r300_context.h" #include "r300_state.h" #include "r300_emit.h" -#include "r300_ioctl.h" #include "r300_render.h" #include "r300_swtcl.h" diff --git a/src/mesa/drivers/dri/r300/r300_emit.h b/src/mesa/drivers/dri/r300/r300_emit.h index 8e57e354d1d..a456d8867c4 100644 --- a/src/mesa/drivers/dri/r300/r300_emit.h +++ b/src/mesa/drivers/dri/r300/r300_emit.h @@ -42,7 +42,6 @@ #include "main/glheader.h" #include "r300_context.h" #include "r300_cmdbuf.h" -#include "radeon_reg.h" static INLINE uint32_t cmdpacket0(struct radeon_screen *rscrn, int reg, int count) diff --git a/src/mesa/drivers/dri/r300/r300_fragprog_common.c b/src/mesa/drivers/dri/r300/r300_fragprog_common.c index 469c278b510..267ee81a7a6 100644 --- a/src/mesa/drivers/dri/r300/r300_fragprog_common.c +++ b/src/mesa/drivers/dri/r300/r300_fragprog_common.c @@ -44,6 +44,7 @@ #include "compiler/radeon_compiler.h" +#include "radeon_mesa_to_rc.h" #include "r300_state.h" @@ -99,8 +100,8 @@ static void insert_WPOS_trailer(struct r300_fragment_program_compiler *compiler, { int i; + fp->wpos_attr = FRAG_ATTRIB_MAX; if (!(compiler->Base.Program.InputsRead & FRAG_BIT_WPOS)) { - fp->wpos_attr = FRAG_ATTRIB_MAX; return; } @@ -112,6 +113,13 @@ static void insert_WPOS_trailer(struct r300_fragment_program_compiler *compiler, } } + /* No free texcoord found, fall-back to software rendering */ + if (fp->wpos_attr == FRAG_ATTRIB_MAX) + { + compiler->Base.Error = 1; + return; + } + rc_transform_fragment_wpos(&compiler->Base, FRAG_ATTRIB_WPOS, fp->wpos_attr); } @@ -124,11 +132,11 @@ static void insert_WPOS_trailer(struct r300_fragment_program_compiler *compiler, */ static void rewriteFog(struct r300_fragment_program_compiler *compiler, struct r300_fragment_program * fp) { - struct prog_src_register src; + struct rc_src_register src; int i; + fp->fog_attr = FRAG_ATTRIB_MAX; if (!(compiler->Base.Program.InputsRead & FRAG_BIT_FOGC)) { - fp->fog_attr = FRAG_ATTRIB_MAX; return; } @@ -140,8 +148,15 @@ static void rewriteFog(struct r300_fragment_program_compiler *compiler, struct r } } + /* No free texcoord found, fall-back to software rendering */ + if (fp->fog_attr == FRAG_ATTRIB_MAX) + { + compiler->Base.Error = 1; + return; + } + memset(&src, 0, sizeof(src)); - src.File = PROGRAM_INPUT; + src.File = RC_FILE_INPUT; src.Index = fp->fog_attr; src.Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ONE); rc_move_input(&compiler->Base, FRAG_ATTRIB_FOGC, src); @@ -218,13 +233,26 @@ static void translate_fragment_program(GLcontext *ctx, struct r300_fragment_prog fflush(stderr); } - rc_mesa_to_rc_program(&compiler.Base, &cont->Base.Base); + radeon_mesa_to_rc_program(&compiler.Base, &cont->Base.Base); insert_WPOS_trailer(&compiler, fp); rewriteFog(&compiler, fp); r3xx_compile_fragment_program(&compiler); + + if (compiler.is_r500) { + /* We need to support the non-KMS DRM interface, which + * artificially limits the number of instructions and + * constants which are available to us. + * + * See also the comment in r300_context.c where we + * set the MAX_NATIVE_xxx values. + */ + if (fp->code.code.r500.inst_end >= 255 || fp->code.constants.Count > 255) + rc_error(&compiler.Base, "Program is too big (upgrade to r300g to avoid this limitation).\n"); + } + fp->error = compiler.Base.Error; fp->InputsRead = compiler.Base.Program.InputsRead; diff --git a/src/mesa/drivers/dri/r300/r300_ioctl.c b/src/mesa/drivers/dri/r300/r300_ioctl.c deleted file mode 100644 index 5cb04e2bb6d..00000000000 --- a/src/mesa/drivers/dri/r300/r300_ioctl.c +++ /dev/null @@ -1,782 +0,0 @@ -/* -Copyright (C) The Weather Channel, Inc. 2002. -Copyright (C) 2004 Nicolai Haehnle. -All Rights Reserved. - -The Weather Channel (TM) funded Tungsten Graphics to develop the -initial release of the Radeon 8500 driver under the XFree86 license. -This notice must be preserved. - -Permission is hereby granted, free of charge, to any person obtaining -a copy of this software and associated documentation files (the -"Software"), to deal in the Software without restriction, including -without limitation the rights to use, copy, modify, merge, publish, -distribute, sublicense, and/or sell copies of the Software, and to -permit persons to whom the Software is furnished to do so, subject to -the following conditions: - -The above copyright notice and this permission notice (including the -next paragraph) shall be included in all copies or substantial -portions of the Software. - -THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, -EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF -MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. -IN NO EVENT SHALL THE COPYRIGHT OWNER(S) 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. - -**************************************************************************/ - -/** - * \file - * - * \author Keith Whitwell <[email protected]> - * - * \author Nicolai Haehnle <[email protected]> - */ - -#include <sched.h> -#include <errno.h> - -#include "main/glheader.h" -#include "main/imports.h" -#include "main/macros.h" -#include "main/context.h" -#include "main/simple_list.h" -#include "swrast/swrast.h" - -#include "radeon_common.h" -#include "radeon_lock.h" -#include "r300_context.h" -#include "r300_ioctl.h" -#include "r300_cmdbuf.h" -#include "r300_state.h" -#include "r300_vertprog.h" -#include "radeon_reg.h" -#include "r300_emit.h" -#include "r300_context.h" - -#include "vblank.h" - -#define R200_3D_DRAW_IMMD_2 0xC0003500 - -#define CLEARBUFFER_COLOR 0x1 -#define CLEARBUFFER_DEPTH 0x2 -#define CLEARBUFFER_STENCIL 0x4 - -#if 1 - -/** - * Fragment program helper macros - */ - -/* Produce unshifted source selectors */ -#define FP_TMP(idx) (idx) -#define FP_CONST(idx) ((idx) | (1 << 5)) - -/* Produce source/dest selector dword */ -#define FP_SELC_MASK_NO 0 -#define FP_SELC_MASK_X 1 -#define FP_SELC_MASK_Y 2 -#define FP_SELC_MASK_XY 3 -#define FP_SELC_MASK_Z 4 -#define FP_SELC_MASK_XZ 5 -#define FP_SELC_MASK_YZ 6 -#define FP_SELC_MASK_XYZ 7 - -#define FP_SELC(destidx,regmask,outmask,src0,src1,src2) \ - (((destidx) << R300_ALU_DSTC_SHIFT) | \ - (FP_SELC_MASK_##regmask << 23) | \ - (FP_SELC_MASK_##outmask << 26) | \ - ((src0) << R300_ALU_SRC0C_SHIFT) | \ - ((src1) << R300_ALU_SRC1C_SHIFT) | \ - ((src2) << R300_ALU_SRC2C_SHIFT)) - -#define FP_SELA_MASK_NO 0 -#define FP_SELA_MASK_W 1 - -#define FP_SELA(destidx,regmask,outmask,src0,src1,src2) \ - (((destidx) << R300_ALU_DSTA_SHIFT) | \ - (FP_SELA_MASK_##regmask << 23) | \ - (FP_SELA_MASK_##outmask << 24) | \ - ((src0) << R300_ALU_SRC0A_SHIFT) | \ - ((src1) << R300_ALU_SRC1A_SHIFT) | \ - ((src2) << R300_ALU_SRC2A_SHIFT)) - -/* Produce unshifted argument selectors */ -#define FP_ARGC(source) R300_ALU_ARGC_##source -#define FP_ARGA(source) R300_ALU_ARGA_##source -#define FP_ABS(arg) ((arg) | (1 << 6)) -#define FP_NEG(arg) ((arg) ^ (1 << 5)) - -/* Produce instruction dword */ -#define FP_INSTRC(opcode,arg0,arg1,arg2) \ - (R300_ALU_OUTC_##opcode | \ - ((arg0) << R300_ALU_ARG0C_SHIFT) | \ - ((arg1) << R300_ALU_ARG1C_SHIFT) | \ - ((arg2) << R300_ALU_ARG2C_SHIFT)) - -#define FP_INSTRA(opcode,arg0,arg1,arg2) \ - (R300_ALU_OUTA_##opcode | \ - ((arg0) << R300_ALU_ARG0A_SHIFT) | \ - ((arg1) << R300_ALU_ARG1A_SHIFT) | \ - ((arg2) << R300_ALU_ARG2A_SHIFT)) - -#endif - -static void r300EmitClearState(GLcontext * ctx); - -static void r300ClearBuffer(r300ContextPtr r300, int flags, - struct radeon_renderbuffer *rrb, - struct radeon_renderbuffer *rrbd) -{ - BATCH_LOCALS(&r300->radeon); - GLcontext *ctx = r300->radeon.glCtx; - __DRIdrawablePrivate *dPriv = radeon_get_drawable(&r300->radeon); - GLuint cbpitch = 0; - r300ContextPtr rmesa = r300; - - if (RADEON_DEBUG & RADEON_IOCTL) - fprintf(stderr, "%s: buffer %p (%i,%i %ix%i)\n", - __FUNCTION__, rrb, dPriv->x, dPriv->y, - dPriv->w, dPriv->h); - - if (rrb) { - cbpitch = (rrb->pitch / rrb->cpp); - if (rrb->cpp == 4) - cbpitch |= R300_COLOR_FORMAT_ARGB8888; - else - cbpitch |= R300_COLOR_FORMAT_RGB565; - - if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE){ - cbpitch |= R300_COLOR_TILE_ENABLE; - } - } - - /* TODO in bufmgr */ - cp_wait(&r300->radeon, R300_WAIT_3D | R300_WAIT_3D_CLEAN); - end_3d(&rmesa->radeon); - - if (flags & CLEARBUFFER_COLOR) { - assert(rrb != 0); - BEGIN_BATCH_NO_AUTOSTATE(6); - OUT_BATCH_REGSEQ(R300_RB3D_COLOROFFSET0, 1); - OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); - OUT_BATCH_REGVAL(R300_RB3D_COLORPITCH0, cbpitch); - END_BATCH(); - } -#if 1 - if (flags & (CLEARBUFFER_DEPTH | CLEARBUFFER_STENCIL)) { - uint32_t zbpitch = (rrbd->pitch / rrbd->cpp); - if (rrbd->bo->flags & RADEON_BO_FLAGS_MACRO_TILE){ - zbpitch |= R300_DEPTHMACROTILE_ENABLE; - } - if (rrbd->bo->flags & RADEON_BO_FLAGS_MICRO_TILE){ - zbpitch |= R300_DEPTHMICROTILE_TILED; - } - BEGIN_BATCH_NO_AUTOSTATE(6); - OUT_BATCH_REGSEQ(R300_ZB_DEPTHOFFSET, 1); - OUT_BATCH_RELOC(0, rrbd->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); - OUT_BATCH_REGSEQ(R300_ZB_DEPTHPITCH, 1); - if (!r300->radeon.radeonScreen->kernel_mm) - OUT_BATCH(zbpitch); - else - OUT_BATCH_RELOC(zbpitch, rrbd->bo, zbpitch, 0, RADEON_GEM_DOMAIN_VRAM, 0); - END_BATCH(); - } -#endif - BEGIN_BATCH_NO_AUTOSTATE(6); - OUT_BATCH_REGSEQ(RB3D_COLOR_CHANNEL_MASK, 1); - if (flags & CLEARBUFFER_COLOR) { - OUT_BATCH((ctx->Color.ColorMask[BCOMP] ? RB3D_COLOR_CHANNEL_MASK_BLUE_MASK0 : 0) | - (ctx->Color.ColorMask[GCOMP] ? RB3D_COLOR_CHANNEL_MASK_GREEN_MASK0 : 0) | - (ctx->Color.ColorMask[RCOMP] ? RB3D_COLOR_CHANNEL_MASK_RED_MASK0 : 0) | - (ctx->Color.ColorMask[ACOMP] ? RB3D_COLOR_CHANNEL_MASK_ALPHA_MASK0 : 0)); - } else { - OUT_BATCH(0); - } - - - { - uint32_t t1, t2; - - t1 = 0x0; - t2 = 0x0; - - if (flags & CLEARBUFFER_DEPTH) { - t1 |= R300_Z_ENABLE | R300_Z_WRITE_ENABLE; - t2 |= - (R300_ZS_ALWAYS << R300_Z_FUNC_SHIFT); - } - - if (flags & CLEARBUFFER_STENCIL) { - t1 |= R300_STENCIL_ENABLE; - t2 |= - (R300_ZS_ALWAYS << - R300_S_FRONT_FUNC_SHIFT) | - (R300_ZS_REPLACE << - R300_S_FRONT_SFAIL_OP_SHIFT) | - (R300_ZS_REPLACE << - R300_S_FRONT_ZPASS_OP_SHIFT) | - (R300_ZS_REPLACE << - R300_S_FRONT_ZFAIL_OP_SHIFT); - } - - OUT_BATCH_REGSEQ(R300_ZB_CNTL, 3); - OUT_BATCH(t1); - OUT_BATCH(t2); - OUT_BATCH(((ctx->Stencil.WriteMask[0] & R300_STENCILREF_MASK) << - R300_STENCILWRITEMASK_SHIFT) | - (ctx->Stencil.Clear & R300_STENCILREF_MASK)); - END_BATCH(); - } - - if (!rmesa->radeon.radeonScreen->kernel_mm) { - BEGIN_BATCH_NO_AUTOSTATE(9); - OUT_BATCH(cmdpacket3(r300->radeon.radeonScreen, R300_CMD_PACKET3_CLEAR)); - OUT_BATCH_FLOAT32(dPriv->w / 2.0); - OUT_BATCH_FLOAT32(dPriv->h / 2.0); - OUT_BATCH_FLOAT32(ctx->Depth.Clear); - OUT_BATCH_FLOAT32(1.0); - OUT_BATCH_FLOAT32(ctx->Color.ClearColor[0]); - OUT_BATCH_FLOAT32(ctx->Color.ClearColor[1]); - OUT_BATCH_FLOAT32(ctx->Color.ClearColor[2]); - OUT_BATCH_FLOAT32(ctx->Color.ClearColor[3]); - END_BATCH(); - } else { - OUT_BATCH(CP_PACKET3(R200_3D_DRAW_IMMD_2, 8)); - OUT_BATCH(R300_PRIM_TYPE_POINT | R300_PRIM_WALK_RING | - (1 << R300_PRIM_NUM_VERTICES_SHIFT)); - OUT_BATCH_FLOAT32(dPriv->w / 2.0); - OUT_BATCH_FLOAT32(dPriv->h / 2.0); - OUT_BATCH_FLOAT32(ctx->Depth.Clear); - OUT_BATCH_FLOAT32(1.0); - OUT_BATCH_FLOAT32(ctx->Color.ClearColor[0]); - OUT_BATCH_FLOAT32(ctx->Color.ClearColor[1]); - OUT_BATCH_FLOAT32(ctx->Color.ClearColor[2]); - OUT_BATCH_FLOAT32(ctx->Color.ClearColor[3]); - } - - r300EmitCacheFlush(rmesa); - cp_wait(&r300->radeon, R300_WAIT_3D | R300_WAIT_3D_CLEAN); - - R300_STATECHANGE(r300, cb); - R300_STATECHANGE(r300, cmk); - R300_STATECHANGE(r300, zs); -} - -static void r300EmitClearState(GLcontext * ctx) -{ - r300ContextPtr r300 = R300_CONTEXT(ctx); - BATCH_LOCALS(&r300->radeon); - __DRIdrawablePrivate *dPriv = radeon_get_drawable(&r300->radeon); - int i; - int has_tcl; - int is_r500 = 0; - GLuint vap_cntl; - - has_tcl = r300->options.hw_tcl_enabled; - - if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) - is_r500 = 1; - - /* State atom dirty tracking is a little subtle here. - * - * On the one hand, we need to make sure base state is emitted - * here if we start with an empty batch buffer, otherwise clear - * works incorrectly with multiple processes. Therefore, the first - * BEGIN_BATCH cannot be a BEGIN_BATCH_NO_AUTOSTATE. - * - * On the other hand, implicit state emission clears the state atom - * dirty bits, so we have to call R300_STATECHANGE later than the - * first BEGIN_BATCH. - * - * The final trickiness is that, because we change state, we need - * to ensure that any stored swtcl primitives are flushed properly - * before we start changing state. See the R300_NEWPRIM in r300Clear - * for this. - */ - BEGIN_BATCH(31); - OUT_BATCH_REGSEQ(R300_VAP_PROG_STREAM_CNTL_0, 1); - if (!has_tcl) - OUT_BATCH(((((0 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_0_SHIFT) | - ((R300_LAST_VEC | (2 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_1_SHIFT))); - else - OUT_BATCH(((((0 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_0_SHIFT) | - ((R300_LAST_VEC | (1 << R300_DST_VEC_LOC_SHIFT) | R300_DATA_TYPE_FLOAT_4) << R300_DATA_TYPE_1_SHIFT))); - - OUT_BATCH_REGVAL(R300_FG_FOG_BLEND, 0); - OUT_BATCH_REGVAL(R300_VAP_PROG_STREAM_CNTL_EXT_0, - ((((R300_SWIZZLE_SELECT_X << R300_SWIZZLE_SELECT_X_SHIFT) | - (R300_SWIZZLE_SELECT_Y << R300_SWIZZLE_SELECT_Y_SHIFT) | - (R300_SWIZZLE_SELECT_Z << R300_SWIZZLE_SELECT_Z_SHIFT) | - (R300_SWIZZLE_SELECT_W << R300_SWIZZLE_SELECT_W_SHIFT) | - ((R300_WRITE_ENA_X | R300_WRITE_ENA_Y | R300_WRITE_ENA_Z | R300_WRITE_ENA_W) << R300_WRITE_ENA_SHIFT)) - << R300_SWIZZLE0_SHIFT) | - (((R300_SWIZZLE_SELECT_X << R300_SWIZZLE_SELECT_X_SHIFT) | - (R300_SWIZZLE_SELECT_Y << R300_SWIZZLE_SELECT_Y_SHIFT) | - (R300_SWIZZLE_SELECT_Z << R300_SWIZZLE_SELECT_Z_SHIFT) | - (R300_SWIZZLE_SELECT_W << R300_SWIZZLE_SELECT_W_SHIFT) | - ((R300_WRITE_ENA_X | R300_WRITE_ENA_Y | R300_WRITE_ENA_Z | R300_WRITE_ENA_W) << R300_WRITE_ENA_SHIFT)) - << R300_SWIZZLE1_SHIFT))); - - /* R300_VAP_INPUT_CNTL_0, R300_VAP_INPUT_CNTL_1 */ - OUT_BATCH_REGSEQ(R300_VAP_VTX_STATE_CNTL, 2); - OUT_BATCH((R300_SEL_USER_COLOR_0 << R300_COLOR_0_ASSEMBLY_SHIFT)); - OUT_BATCH(R300_INPUT_CNTL_POS | R300_INPUT_CNTL_COLOR | R300_INPUT_CNTL_TC0); - - /* comes from fglrx startup of clear */ - OUT_BATCH_REGSEQ(R300_SE_VTE_CNTL, 2); - OUT_BATCH(R300_VTX_W0_FMT | R300_VPORT_X_SCALE_ENA | - R300_VPORT_X_OFFSET_ENA | R300_VPORT_Y_SCALE_ENA | - R300_VPORT_Y_OFFSET_ENA | R300_VPORT_Z_SCALE_ENA | - R300_VPORT_Z_OFFSET_ENA); - OUT_BATCH(0x8); - - OUT_BATCH_REGVAL(R300_VAP_PSC_SGN_NORM_CNTL, 0xaaaaaaaa); - - OUT_BATCH_REGSEQ(R300_VAP_OUTPUT_VTX_FMT_0, 2); - OUT_BATCH(R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT | - R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT); - OUT_BATCH(0); /* no textures */ - - OUT_BATCH_REGVAL(R300_TX_ENABLE, 0); - - OUT_BATCH_REGSEQ(R300_SE_VPORT_XSCALE, 6); - OUT_BATCH_FLOAT32(1.0); - OUT_BATCH_FLOAT32(dPriv->x); - OUT_BATCH_FLOAT32(1.0); - OUT_BATCH_FLOAT32(dPriv->y); - OUT_BATCH_FLOAT32(1.0); - OUT_BATCH_FLOAT32(0.0); - - OUT_BATCH_REGVAL(R300_FG_ALPHA_FUNC, 0); - - OUT_BATCH_REGSEQ(R300_RB3D_CBLEND, 2); - OUT_BATCH(0x0); - OUT_BATCH(0x0); - END_BATCH(); - - R300_STATECHANGE(r300, vir[0]); - R300_STATECHANGE(r300, fogs); - R300_STATECHANGE(r300, vir[1]); - R300_STATECHANGE(r300, vic); - R300_STATECHANGE(r300, vte); - R300_STATECHANGE(r300, vof); - R300_STATECHANGE(r300, txe); - R300_STATECHANGE(r300, vpt); - R300_STATECHANGE(r300, at); - R300_STATECHANGE(r300, bld); - R300_STATECHANGE(r300, ps); - - if (has_tcl) { - R300_STATECHANGE(r300, vap_clip_cntl); - - BEGIN_BATCH_NO_AUTOSTATE(2); - OUT_BATCH_REGVAL(R300_VAP_CLIP_CNTL, R300_PS_UCP_MODE_CLIP_AS_TRIFAN | R300_CLIP_DISABLE); - END_BATCH(); - } - - BEGIN_BATCH_NO_AUTOSTATE(2); - OUT_BATCH_REGVAL(R300_GA_POINT_SIZE, - ((dPriv->w * 6) << R300_POINTSIZE_X_SHIFT) | - ((dPriv->h * 6) << R300_POINTSIZE_Y_SHIFT)); - END_BATCH(); - - if (!is_r500) { - R300_STATECHANGE(r300, ri); - R300_STATECHANGE(r300, rc); - R300_STATECHANGE(r300, rr); - - BEGIN_BATCH(14); - OUT_BATCH_REGSEQ(R300_RS_IP_0, 8); - for (i = 0; i < 8; ++i) - OUT_BATCH(R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(3)); - - OUT_BATCH_REGSEQ(R300_RS_COUNT, 2); - OUT_BATCH((1 << R300_IC_COUNT_SHIFT) | R300_HIRES_EN); - OUT_BATCH(0x0); - - OUT_BATCH_REGVAL(R300_RS_INST_0, R300_RS_INST_COL_CN_WRITE); - END_BATCH(); - } else { - R300_STATECHANGE(r300, ri); - R300_STATECHANGE(r300, rc); - R300_STATECHANGE(r300, rr); - - BEGIN_BATCH(14); - OUT_BATCH_REGSEQ(R500_RS_IP_0, 8); - for (i = 0; i < 8; ++i) { - OUT_BATCH((R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_S_SHIFT) | - (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_T_SHIFT) | - (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT) | - (R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT)); - } - - OUT_BATCH_REGSEQ(R300_RS_COUNT, 2); - OUT_BATCH((1 << R300_IC_COUNT_SHIFT) | R300_HIRES_EN); - OUT_BATCH(0x0); - - OUT_BATCH_REGVAL(R500_RS_INST_0, R500_RS_INST_COL_CN_WRITE); - END_BATCH(); - } - - if (!is_r500) { - R300_STATECHANGE(r300, fp); - R300_STATECHANGE(r300, fpi[0]); - R300_STATECHANGE(r300, fpi[1]); - R300_STATECHANGE(r300, fpi[2]); - R300_STATECHANGE(r300, fpi[3]); - - BEGIN_BATCH(17); - OUT_BATCH_REGSEQ(R300_US_CONFIG, 3); - OUT_BATCH(0x0); - OUT_BATCH(0x0); - OUT_BATCH(0x0); - OUT_BATCH_REGSEQ(R300_US_CODE_ADDR_0, 4); - OUT_BATCH(0x0); - OUT_BATCH(0x0); - OUT_BATCH(0x0); - OUT_BATCH(R300_RGBA_OUT); - - OUT_BATCH_REGVAL(R300_US_ALU_RGB_INST_0, - FP_INSTRC(MAD, FP_ARGC(SRC0C_XYZ), FP_ARGC(ONE), FP_ARGC(ZERO))); - OUT_BATCH_REGVAL(R300_US_ALU_RGB_ADDR_0, - FP_SELC(0, NO, XYZ, FP_TMP(0), 0, 0)); - OUT_BATCH_REGVAL(R300_US_ALU_ALPHA_INST_0, - FP_INSTRA(MAD, FP_ARGA(SRC0A), FP_ARGA(ONE), FP_ARGA(ZERO))); - OUT_BATCH_REGVAL(R300_US_ALU_ALPHA_ADDR_0, - FP_SELA(0, NO, W, FP_TMP(0), 0, 0)); - END_BATCH(); - } else { - struct radeon_state_atom r500fp; - uint32_t _cmd[10]; - - R300_STATECHANGE(r300, fp); - R300_STATECHANGE(r300, r500fp); - - BEGIN_BATCH(7); - OUT_BATCH_REGSEQ(R500_US_CONFIG, 2); - OUT_BATCH(R500_ZERO_TIMES_ANYTHING_EQUALS_ZERO); - OUT_BATCH(0x0); - OUT_BATCH_REGSEQ(R500_US_CODE_ADDR, 3); - OUT_BATCH(R500_US_CODE_START_ADDR(0) | R500_US_CODE_END_ADDR(1)); - OUT_BATCH(R500_US_CODE_RANGE_ADDR(0) | R500_US_CODE_RANGE_SIZE(1)); - OUT_BATCH(R500_US_CODE_OFFSET_ADDR(0)); - END_BATCH(); - - r500fp.check = check_r500fp; - r500fp.cmd = _cmd; - r500fp.cmd[0] = cmdr500fp(r300->radeon.radeonScreen, 0, 1, 0, 0); - r500fp.cmd[1] = R500_INST_TYPE_OUT | - R500_INST_TEX_SEM_WAIT | - R500_INST_LAST | - R500_INST_RGB_OMASK_R | - R500_INST_RGB_OMASK_G | - R500_INST_RGB_OMASK_B | - R500_INST_ALPHA_OMASK | - R500_INST_RGB_CLAMP | - R500_INST_ALPHA_CLAMP; - r500fp.cmd[2] = R500_RGB_ADDR0(0) | - R500_RGB_ADDR1(0) | - R500_RGB_ADDR1_CONST | - R500_RGB_ADDR2(0) | - R500_RGB_ADDR2_CONST; - r500fp.cmd[3] = R500_ALPHA_ADDR0(0) | - R500_ALPHA_ADDR1(0) | - R500_ALPHA_ADDR1_CONST | - R500_ALPHA_ADDR2(0) | - R500_ALPHA_ADDR2_CONST; - r500fp.cmd[4] = R500_ALU_RGB_SEL_A_SRC0 | - R500_ALU_RGB_R_SWIZ_A_R | - R500_ALU_RGB_G_SWIZ_A_G | - R500_ALU_RGB_B_SWIZ_A_B | - R500_ALU_RGB_SEL_B_SRC0 | - R500_ALU_RGB_R_SWIZ_B_R | - R500_ALU_RGB_B_SWIZ_B_G | - R500_ALU_RGB_G_SWIZ_B_B; - r500fp.cmd[5] = R500_ALPHA_OP_CMP | - R500_ALPHA_SWIZ_A_A | - R500_ALPHA_SWIZ_B_A; - r500fp.cmd[6] = R500_ALU_RGBA_OP_CMP | - R500_ALU_RGBA_R_SWIZ_0 | - R500_ALU_RGBA_G_SWIZ_0 | - R500_ALU_RGBA_B_SWIZ_0 | - R500_ALU_RGBA_A_SWIZ_0; - - r500fp.cmd[7] = 0; - if (r300->radeon.radeonScreen->kernel_mm) { - emit_r500fp(ctx, &r500fp); - } else { - int dwords = r500fp.check(ctx,&r500fp); - BEGIN_BATCH_NO_AUTOSTATE(dwords); - OUT_BATCH_TABLE(r500fp.cmd, dwords); - END_BATCH(); - } - - } - - BEGIN_BATCH(2); - OUT_BATCH_REGVAL(R300_VAP_PVS_STATE_FLUSH_REG, 0); - END_BATCH(); - - if (has_tcl) { - vap_cntl = ((10 << R300_PVS_NUM_SLOTS_SHIFT) | - (5 << R300_PVS_NUM_CNTLRS_SHIFT) | - (12 << R300_VF_MAX_VTX_NUM_SHIFT)); - if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) - vap_cntl |= R500_TCL_STATE_OPTIMIZATION; - } else { - vap_cntl = ((10 << R300_PVS_NUM_SLOTS_SHIFT) | - (5 << R300_PVS_NUM_CNTLRS_SHIFT) | - (5 << R300_VF_MAX_VTX_NUM_SHIFT)); - } - - if (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV515) - vap_cntl |= (2 << R300_PVS_NUM_FPUS_SHIFT); - else if ((r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV530) || - (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV560) || - (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV570)) - vap_cntl |= (5 << R300_PVS_NUM_FPUS_SHIFT); - else if ((r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV410) || - (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_R420)) - vap_cntl |= (6 << R300_PVS_NUM_FPUS_SHIFT); - else if ((r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_R520) || - (r300->radeon.radeonScreen->chip_family == CHIP_FAMILY_R580)) - vap_cntl |= (8 << R300_PVS_NUM_FPUS_SHIFT); - else - vap_cntl |= (4 << R300_PVS_NUM_FPUS_SHIFT); - - R300_STATECHANGE(r300, vap_cntl); - - BEGIN_BATCH(2); - OUT_BATCH_REGVAL(R300_VAP_CNTL, vap_cntl); - END_BATCH(); - - if (has_tcl) { - struct radeon_state_atom vpu; - uint32_t _cmd[10]; - R300_STATECHANGE(r300, pvs); - R300_STATECHANGE(r300, vap_flush); - R300_STATECHANGE(r300, vpi); - - BEGIN_BATCH(4); - OUT_BATCH_REGSEQ(R300_VAP_PVS_CODE_CNTL_0, 3); - OUT_BATCH((0 << R300_PVS_FIRST_INST_SHIFT) | - (0 << R300_PVS_XYZW_VALID_INST_SHIFT) | - (1 << R300_PVS_LAST_INST_SHIFT)); - OUT_BATCH((0 << R300_PVS_CONST_BASE_OFFSET_SHIFT) | - (0 << R300_PVS_MAX_CONST_ADDR_SHIFT)); - OUT_BATCH(1 << R300_PVS_LAST_VTX_SRC_INST_SHIFT); - END_BATCH(); - - vpu.check = check_vpu; - vpu.cmd = _cmd; - vpu.cmd[0] = cmdvpu(r300->radeon.radeonScreen, 0, 2); - - vpu.cmd[1] = PVS_OP_DST_OPERAND(VE_ADD, GL_FALSE, GL_FALSE, - 0, 0xf, PVS_DST_REG_OUT); - vpu.cmd[2] = PVS_SRC_OPERAND(0, PVS_SRC_SELECT_X, PVS_SRC_SELECT_Y, - PVS_SRC_SELECT_Z, PVS_SRC_SELECT_W, - PVS_SRC_REG_INPUT, NEGATE_NONE); - vpu.cmd[3] = PVS_SRC_OPERAND(0, PVS_SRC_SELECT_FORCE_0, - PVS_SRC_SELECT_FORCE_0, - PVS_SRC_SELECT_FORCE_0, - PVS_SRC_SELECT_FORCE_0, - PVS_SRC_REG_INPUT, NEGATE_NONE); - vpu.cmd[4] = 0x0; - - vpu.cmd[5] = PVS_OP_DST_OPERAND(VE_ADD, GL_FALSE, GL_FALSE, 1, 0xf, - PVS_DST_REG_OUT); - vpu.cmd[6] = PVS_SRC_OPERAND(1, PVS_SRC_SELECT_X, - PVS_SRC_SELECT_Y, PVS_SRC_SELECT_Z, - PVS_SRC_SELECT_W, PVS_SRC_REG_INPUT, - NEGATE_NONE); - vpu.cmd[7] = PVS_SRC_OPERAND(1, PVS_SRC_SELECT_FORCE_0, - PVS_SRC_SELECT_FORCE_0, - PVS_SRC_SELECT_FORCE_0, - PVS_SRC_SELECT_FORCE_0, - PVS_SRC_REG_INPUT, NEGATE_NONE); - vpu.cmd[8] = 0x0; - - if (r300->radeon.radeonScreen->kernel_mm) { - int dwords = r300->hw.vap_flush.check(ctx,&r300->hw.vap_flush); - BEGIN_BATCH_NO_AUTOSTATE(dwords); - OUT_BATCH_TABLE(r300->hw.vap_flush.cmd, dwords); - END_BATCH(); - emit_vpu(ctx, &vpu); - } else { - int dwords = vpu.check(ctx,&vpu); - BEGIN_BATCH_NO_AUTOSTATE(dwords); - OUT_BATCH_TABLE(vpu.cmd, dwords); - END_BATCH(); - } - - } -} - -static int r300KernelClear(GLcontext *ctx, GLuint flags) -{ - r300ContextPtr r300 = R300_CONTEXT(ctx); - __DRIdrawablePrivate *dPriv = radeon_get_drawable(&r300->radeon); - struct radeon_framebuffer *rfb = dPriv->driverPrivate; - struct radeon_renderbuffer *rrb; - struct radeon_renderbuffer *rrbd; - int bits = 0, ret; - - /* Make sure it fits there. */ - radeon_cs_space_reset_bos(r300->radeon.cmdbuf.cs); - - if (flags & BUFFER_BIT_COLOR0) { - rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_COLOR0); - radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs, - rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM); - } - - if (flags & BUFFER_BIT_FRONT_LEFT) { - rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_FRONT_LEFT); - radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs, - rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM); - } - - if (flags & BUFFER_BIT_BACK_LEFT) { - rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_BACK_LEFT); - radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs, - rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM); - } - - rrbd = radeon_get_renderbuffer(&rfb->base, BUFFER_DEPTH); - if (rrbd) { - radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs, - rrbd->bo, 0, RADEON_GEM_DOMAIN_VRAM); - } - - ret = radeon_cs_space_check(r300->radeon.cmdbuf.cs); - if (ret) - return -1; - - rcommonEnsureCmdBufSpace(&r300->radeon, 421 * 3, __FUNCTION__); - if (flags || bits) - r300EmitClearState(ctx); - - rrbd = radeon_get_renderbuffer(&rfb->base, BUFFER_DEPTH); - if (rrbd && (flags & BUFFER_BIT_DEPTH)) - bits |= CLEARBUFFER_DEPTH; - - if (rrbd && (flags & BUFFER_BIT_STENCIL)) - bits |= CLEARBUFFER_STENCIL; - - if (flags & BUFFER_BIT_COLOR0) { - rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_COLOR0); - r300ClearBuffer(r300, CLEARBUFFER_COLOR, rrb, NULL); - bits = 0; - } - - if (flags & BUFFER_BIT_FRONT_LEFT) { - rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_FRONT_LEFT); - r300ClearBuffer(r300, bits | CLEARBUFFER_COLOR, rrb, rrbd); - bits = 0; - } - - if (flags & BUFFER_BIT_BACK_LEFT) { - rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_BACK_LEFT); - r300ClearBuffer(r300, bits | CLEARBUFFER_COLOR, rrb, rrbd); - bits = 0; - } - - if (bits) - r300ClearBuffer(r300, bits, NULL, rrbd); - - COMMIT_BATCH(); - return 0; -} - -/** - * Buffer clear - */ -static void r300Clear(GLcontext * ctx, GLbitfield mask) -{ - r300ContextPtr r300 = R300_CONTEXT(ctx); - __DRIdrawablePrivate *dPriv = radeon_get_drawable(&r300->radeon); - const GLuint colorMask = *((GLuint *) & ctx->Color.ColorMask); - GLbitfield swrast_mask = 0, tri_mask = 0; - int i, ret; - struct gl_framebuffer *fb = ctx->DrawBuffer; - - if (RADEON_DEBUG & RADEON_IOCTL) - fprintf(stderr, "r300Clear\n"); - - if (!r300->radeon.radeonScreen->driScreen->dri2.enabled) { - LOCK_HARDWARE(&r300->radeon); - UNLOCK_HARDWARE(&r300->radeon); - if (dPriv->numClipRects == 0) - return; - } - - /* Flush swtcl vertices if necessary, because we will change hardware - * state during clear. See also the state-related comment in - * r300EmitClearState. - */ - R300_NEWPRIM(r300); - - if (colorMask == ~0) - tri_mask |= (mask & BUFFER_BITS_COLOR); - else - tri_mask |= (mask & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_BACK_LEFT)); - - - /* HW stencil */ - if (mask & BUFFER_BIT_STENCIL) { - tri_mask |= BUFFER_BIT_STENCIL; - } - - /* HW depth */ - if (mask & BUFFER_BIT_DEPTH) { - tri_mask |= BUFFER_BIT_DEPTH; - } - - /* If we're doing a tri pass for depth/stencil, include a likely color - * buffer with it. - */ - - for (i = 0; i < BUFFER_COUNT; i++) { - GLuint bufBit = 1 << i; - if ((tri_mask) & bufBit) { - if (!fb->Attachment[i].Renderbuffer->ClassID) { - tri_mask &= ~bufBit; - swrast_mask |= bufBit; - } - } - } - - /* SW fallback clearing */ - swrast_mask = mask & ~tri_mask; - - ret = 0; - if (tri_mask) { - if (r300->radeon.radeonScreen->kernel_mm) - radeonUserClear(ctx, tri_mask); - else { - /* if kernel clear fails due to size restraints fallback */ - ret = r300KernelClear(ctx, tri_mask); - if (ret < 0) - swrast_mask |= tri_mask; - } - } - - if (swrast_mask) { - if (RADEON_DEBUG & RADEON_FALLBACKS) - fprintf(stderr, "%s: swrast clear, mask: %x\n", - __FUNCTION__, swrast_mask); - _swrast_Clear(ctx, swrast_mask); - } -} - -void r300InitIoctlFuncs(struct dd_function_table *functions) -{ - functions->Clear = r300Clear; - functions->Finish = radeonFinish; - functions->Flush = radeonFlush; -} diff --git a/src/mesa/drivers/dri/r300/r300_ioctl.h b/src/mesa/drivers/dri/r300/r300_ioctl.h deleted file mode 100644 index 3abfa71a6e8..00000000000 --- a/src/mesa/drivers/dri/r300/r300_ioctl.h +++ /dev/null @@ -1,44 +0,0 @@ -/* -Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. - -The Weather Channel (TM) funded Tungsten Graphics to develop the -initial release of the Radeon 8500 driver under the XFree86 license. -This notice must be preserved. - -Permission is hereby granted, free of charge, to any person obtaining -a copy of this software and associated documentation files (the -"Software"), to deal in the Software without restriction, including -without limitation the rights to use, copy, modify, merge, publish, -distribute, sublicense, and/or sell copies of the Software, and to -permit persons to whom the Software is furnished to do so, subject to -the following conditions: - -The above copyright notice and this permission notice (including the -next paragraph) shall be included in all copies or substantial -portions of the Software. - -THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, -EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF -MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. -IN NO EVENT SHALL THE COPYRIGHT OWNER(S) 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]> - * Nicolai Haehnle <[email protected]> - */ - -#ifndef __R300_IOCTL_H__ -#define __R300_IOCTL_H__ - -#include "r300_context.h" -#include "radeon_drm.h" - -extern void r300InitIoctlFuncs(struct dd_function_table *functions); - -#endif /* __R300_IOCTL_H__ */ diff --git a/src/mesa/drivers/dri/r300/r300_reg.h b/src/mesa/drivers/dri/r300/r300_reg.h index b9ccd098dc8..ea684e7df18 100644 --- a/src/mesa/drivers/dri/r300/r300_reg.h +++ b/src/mesa/drivers/dri/r300/r300_reg.h @@ -1789,6 +1789,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_ALU_DSTC_OUTPUT_X (1 << 26) # define R300_ALU_DSTC_OUTPUT_Y (1 << 27) # define R300_ALU_DSTC_OUTPUT_Z (1 << 28) +# define R300_RGB_TARGET(x) ((x) << 29) #define R300_US_ALU_ALPHA_ADDR_0 0x47C0 # define R300_ALU_SRC0A_SHIFT 0 @@ -1806,6 +1807,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_ALU_DSTA_REG (1 << 23) # define R300_ALU_DSTA_OUTPUT (1 << 24) # define R300_ALU_DSTA_DEPTH (1 << 27) +# define R300_ALPHA_TARGET(x) ((x) << 25) #define R300_US_ALU_RGB_INST_0 0x48C0 # define R300_ALU_ARGC_SRC0C_XYZ 0 @@ -3002,6 +3004,8 @@ enum { # define R500_INST_RGB_CLAMP (1 << 19) # define R500_INST_ALPHA_CLAMP (1 << 20) # define R500_INST_ALU_RESULT_SEL (1 << 21) +# define R500_INST_ALU_RESULT_SEL_RED (0 << 21) +# define R500_INST_ALU_RESULT_SEL_ALPHA (1 << 21) # define R500_INST_ALPHA_PRED_INV (1 << 22) # define R500_INST_ALU_RESULT_OP_EQ (0 << 23) # define R500_INST_ALU_RESULT_OP_LT (1 << 23) diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c index b5ddfdc9f82..02c94250a8f 100644 --- a/src/mesa/drivers/dri/r300/r300_render.c +++ b/src/mesa/drivers/dri/r300/r300_render.c @@ -67,10 +67,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #include "vbo/vbo_split.h" #include "tnl/tnl.h" #include "tnl/t_vp_build.h" -#include "radeon_reg.h" -#include "radeon_macros.h" #include "r300_context.h" -#include "r300_ioctl.h" #include "r300_state.h" #include "r300_reg.h" #include "r300_tex.h" @@ -475,7 +472,7 @@ void r300SwitchFallback(GLcontext *ctx, uint32_t bit, GLboolean mode) /* update only if we have disabled all tcl fallbacks */ if (rmesa->options.hw_tcl_enabled) { - if ((old_fallback & R300_RASTER_FALLBACK_MASK) == bit) { + if ((old_fallback & R300_TCL_FALLBACK_MASK) == bit) { R300_STATECHANGE(rmesa, vap_cntl_status); rmesa->hw.vap_cntl_status.cmd[1] &= ~R300_VAP_TCL_BYPASS; } diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c index 3060f49aaf8..da0a9dfb4cf 100644 --- a/src/mesa/drivers/dri/r300/r300_state.c +++ b/src/mesa/drivers/dri/r300/r300_state.c @@ -45,7 +45,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/framebuffer.h" #include "main/simple_list.h" #include "main/api_arrayelt.h" -#include "main/texformat.h" #include "swrast/swrast.h" #include "swrast_setup/swrast_setup.h" @@ -56,7 +55,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "tnl/t_vp_build.h" #include "r300_context.h" -#include "r300_ioctl.h" #include "r300_state.h" #include "r300_reg.h" #include "r300_emit.h" diff --git a/src/mesa/drivers/dri/r300/r300_swtcl.c b/src/mesa/drivers/dri/r300/r300_swtcl.c index ee2c71e1a7f..383c8a274b5 100644 --- a/src/mesa/drivers/dri/r300/r300_swtcl.c +++ b/src/mesa/drivers/dri/r300/r300_swtcl.c @@ -124,7 +124,7 @@ void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead, GLuint *_ } if (ctx->Light.Enabled && ctx->Light.Model.TwoSide) { - VB->AttribPtr[VERT_ATTRIB_GENERIC0] = VB->ColorPtr[1]; + VB->AttribPtr[VERT_ATTRIB_GENERIC0] = VB->BackfaceColorPtr; OutputsWritten |= 1 << VERT_RESULT_BFC0; #if MESA_LITTLE_ENDIAN EMIT_ATTR( _TNL_ATTRIB_GENERIC0, EMIT_4UB_4F_RGBA ); @@ -134,7 +134,7 @@ void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead, GLuint *_ ADD_ATTR(VERT_ATTRIB_GENERIC0, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR2, SWIZZLE_XYZW, MASK_XYZW, 1); #endif if (fp_reads & FRAG_BIT_COL1) { - VB->AttribPtr[VERT_ATTRIB_GENERIC1] = VB->SecondaryColorPtr[1]; + VB->AttribPtr[VERT_ATTRIB_GENERIC1] = VB->BackfaceSecondaryColorPtr; GLuint swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ONE); OutputsWritten |= 1 << VERT_RESULT_BFC1; #if MESA_LITTLE_ENDIAN @@ -159,7 +159,7 @@ void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead, GLuint *_ int tex_id = rmesa->selected_fp->wpos_attr - FRAG_ATTRIB_TEX0; VB->AttribPtr[VERT_ATTRIB_TEX0 + tex_id] = VB->AttribPtr[VERT_ATTRIB_POS]; - VB->TexCoordPtr[tex_id] = VB->AttribPtr[VERT_ATTRIB_POS]; + VB->AttribPtr[_TNL_ATTRIB_TEX0 + tex_id] = VB->AttribPtr[VERT_ATTRIB_POS]; RENDERINPUTS_SET(tnl->render_inputs_bitset, _TNL_ATTRIB_TEX0 + tex_id); } @@ -167,7 +167,7 @@ void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead, GLuint *_ int tex_id = rmesa->selected_fp->fog_attr - FRAG_ATTRIB_TEX0; VB->AttribPtr[VERT_ATTRIB_TEX0 + tex_id] = VB->AttribPtr[VERT_ATTRIB_FOG]; - VB->TexCoordPtr[tex_id] = VB->AttribPtr[VERT_ATTRIB_FOG]; + VB->AttribPtr[_TNL_ATTRIB_TEX0 + tex_id] = VB->AttribPtr[VERT_ATTRIB_FOG]; RENDERINPUTS_SET(tnl->render_inputs_bitset, _TNL_ATTRIB_TEX0 + tex_id); } @@ -180,7 +180,7 @@ void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead, GLuint *_ GLuint swiz, format, hw_format; for (i = 0; i < ctx->Const.MaxTextureUnits; i++) { if (fp_reads & FRAG_BIT_TEX(i)) { - switch (VB->TexCoordPtr[i]->size) { + switch (VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size) { case 1: format = EMIT_1F; hw_format = R300_DATA_TYPE_FLOAT_1; @@ -665,11 +665,11 @@ void r300_swtcl_flush(GLcontext *ctx, uint32_t current_offset) r300EmitCacheFlush(rmesa); radeonEmitState(&rmesa->radeon); - r300_emit_scissor(ctx); + r300_emit_scissor(ctx); r300EmitVertexAOS(rmesa, - rmesa->radeon.swtcl.vertex_size, - first_elem(&rmesa->radeon.dma.reserved)->bo, - current_offset); + rmesa->radeon.swtcl.vertex_size, + rmesa->radeon.swtcl.bo, + current_offset); r300EmitVbufPrim(rmesa, rmesa->radeon.swtcl.hw_primitive, diff --git a/src/mesa/drivers/dri/r300/r300_tex.c b/src/mesa/drivers/dri/r300/r300_tex.c index 433e5a87d48..ac3d5b1bec3 100644 --- a/src/mesa/drivers/dri/r300/r300_tex.c +++ b/src/mesa/drivers/dri/r300/r300_tex.c @@ -40,7 +40,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/image.h" #include "main/mipmap.h" #include "main/simple_list.h" -#include "main/texformat.h" #include "main/texstore.h" #include "main/teximage.h" #include "main/texobj.h" @@ -49,7 +48,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r300_context.h" #include "r300_state.h" -#include "r300_ioctl.h" #include "radeon_mipmap_tree.h" #include "r300_tex.h" @@ -196,6 +194,7 @@ static void r300TexParameter(GLcontext * ctx, GLenum target, GLenum pname, const GLfloat * params) { radeonTexObj* t = radeon_tex_obj(texObj); + GLenum texBaseFormat; if (RADEON_DEBUG & (RADEON_STATE | RADEON_TEXTURE)) { fprintf(stderr, "%s( %s )\n", __FUNCTION__, @@ -223,23 +222,16 @@ static void r300TexParameter(GLcontext * ctx, GLenum target, case GL_TEXTURE_MAX_LEVEL: case GL_TEXTURE_MIN_LOD: case GL_TEXTURE_MAX_LOD: - /* This isn't the most efficient solution but there doesn't appear to - * be a nice alternative. Since there's no LOD clamping, - * we just have to rely on loading the right subset of mipmap levels - * to simulate a clamped LOD. - */ - if (t->mt) { - radeon_miptree_unreference(t->mt); - t->mt = 0; - t->validated = GL_FALSE; - } + t->validated = GL_FALSE; break; case GL_DEPTH_TEXTURE_MODE: if (!texObj->Image[0][texObj->BaseLevel]) return; - if (texObj->Image[0][texObj->BaseLevel]->TexFormat->BaseFormat - == GL_DEPTH_COMPONENT) { + texBaseFormat = texObj->Image[0][texObj->BaseLevel]->_BaseFormat; + + if (texBaseFormat == GL_DEPTH_COMPONENT || + texBaseFormat == GL_DEPTH_STENCIL) { r300SetDepthTexMode(texObj); break; } else { @@ -268,7 +260,11 @@ static void r300DeleteTexture(GLcontext * ctx, struct gl_texture_object *texObj) if (rmesa) { int i; - radeon_firevertices(&rmesa->radeon); + struct radeon_bo *bo; + bo = !t->mt ? t->bo : t->mt->bo; + if (bo && radeon_bo_is_referenced_by_cs(bo, rmesa->radeon.cmdbuf.cs)) { + radeon_firevertices(&rmesa->radeon); + } for(i = 0; i < R300_MAX_TEXTURE_UNITS; ++i) if (rmesa->hw.textures[i] == t) @@ -280,10 +276,8 @@ static void r300DeleteTexture(GLcontext * ctx, struct gl_texture_object *texObj) t->bo = NULL; } - if (t->mt) { - radeon_miptree_unreference(t->mt); - t->mt = 0; - } + radeon_miptree_unreference(&t->mt); + _mesa_delete_texture_object(ctx, texObj); } diff --git a/src/mesa/drivers/dri/r300/r300_tex.h b/src/mesa/drivers/dri/r300/r300_tex.h index 8a653ea2d11..beb10072e9c 100644 --- a/src/mesa/drivers/dri/r300/r300_tex.h +++ b/src/mesa/drivers/dri/r300/r300_tex.h @@ -51,4 +51,6 @@ extern GLboolean r300ValidateBuffers(GLcontext * ctx); extern void r300InitTextureFuncs(struct dd_function_table *functions); +uint32_t r300TranslateTexFormat(gl_format mesaFormat); + #endif /* __r300_TEX_H__ */ diff --git a/src/mesa/drivers/dri/r300/r300_texcopy.c b/src/mesa/drivers/dri/r300/r300_texcopy.c new file mode 100644 index 00000000000..5e3a724d4e6 --- /dev/null +++ b/src/mesa/drivers/dri/r300/r300_texcopy.c @@ -0,0 +1,167 @@ +/* + * Copyright (C) 2009 Maciej Cencora <[email protected]> + * + * 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, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "radeon_common.h" +#include "r300_context.h" + +#include "main/image.h" +#include "main/teximage.h" +#include "main/texstate.h" +#include "drivers/common/meta.h" + +#include "radeon_mipmap_tree.h" +#include "r300_blit.h" +#include <main/debug.h> + +static GLboolean +do_copy_texsubimage(GLcontext *ctx, + GLenum target, GLint level, + struct radeon_tex_obj *tobj, + radeon_texture_image *timg, + GLint dstx, GLint dsty, + GLint x, GLint y, + GLsizei width, GLsizei height) +{ + struct r300_context *r300 = R300_CONTEXT(ctx); + struct radeon_renderbuffer *rrb; + + if (_mesa_get_format_bits(timg->base.TexFormat, GL_DEPTH_BITS) > 0) { + rrb = radeon_get_depthbuffer(&r300->radeon); + } else { + rrb = radeon_get_colorbuffer(&r300->radeon); + } + + if (!timg->mt) { + radeon_validate_texture_miptree(ctx, &tobj->base); + } + + assert(rrb && rrb->bo); + assert(timg->mt->bo); + assert(timg->base.Width >= dstx + width); + assert(timg->base.Height >= dsty + height); + //assert(tobj->mt == timg->mt); + + intptr_t src_offset = rrb->draw_offset + x * rrb->cpp + y * rrb->pitch; + intptr_t dst_offset = radeon_miptree_image_offset(timg->mt, _mesa_tex_target_to_face(target), level); + dst_offset += dstx * _mesa_get_format_bytes(timg->base.TexFormat) + + dsty * _mesa_format_row_stride(timg->base.TexFormat, timg->base.Width); + + if (src_offset % 32 || dst_offset % 32) { + return GL_FALSE; + } + + if (0) { + fprintf(stderr, "%s: copying to face %d, level %d\n", + __FUNCTION__, _mesa_tex_target_to_face(target), level); + fprintf(stderr, "to: x %d, y %d, offset %d\n", dstx, dsty, (uint32_t) dst_offset); + fprintf(stderr, "from (%dx%d) width %d, height %d, offset %d, pitch %d, width %d\n", + x, y, width, height, (uint32_t) src_offset, rrb->pitch, rrb->pitch/rrb->cpp); + fprintf(stderr, "src size %d, dst size %d\n", rrb->bo->size, timg->mt->bo->size); + + } + + /* blit from src buffer to texture */ + return r300_blit(r300, rrb->bo, src_offset, rrb->base.Format, rrb->pitch, + rrb->base.Width, rrb->base.Height, timg->mt->bo ? timg->mt->bo : timg->bo, dst_offset, + timg->base.TexFormat, width, height); +} + +static void +r300CopyTexImage2D(GLcontext *ctx, GLenum target, GLint level, + GLenum internalFormat, + GLint x, GLint y, GLsizei width, GLsizei height, + GLint border) +{ + struct gl_texture_unit *texUnit = _mesa_get_current_tex_unit(ctx); + struct gl_texture_object *texObj = + _mesa_select_tex_object(ctx, texUnit, target); + struct gl_texture_image *texImage = + _mesa_select_tex_image(ctx, texObj, target, level); + int srcx, srcy, dstx, dsty; + + if (border) + goto fail; + + /* Setup or redefine the texture object, mipmap tree and texture + * image. Don't populate yet. + */ + ctx->Driver.TexImage2D(ctx, target, level, internalFormat, + width, height, border, + GL_RGBA, GL_UNSIGNED_BYTE, NULL, + &ctx->DefaultPacking, texObj, texImage); + + srcx = x; + srcy = y; + dstx = 0; + dsty = 0; + if (!_mesa_clip_copytexsubimage(ctx, + &dstx, &dsty, + &srcx, &srcy, + &width, &height)) { + return; + } + + if (!do_copy_texsubimage(ctx, target, level, + radeon_tex_obj(texObj), (radeon_texture_image *)texImage, + 0, 0, x, y, width, height)) { + goto fail; + } + + return; + +fail: + _mesa_meta_CopyTexImage2D(ctx, target, level, internalFormat, x, y, + width, height, border); +} + +static void +r300CopyTexSubImage2D(GLcontext *ctx, GLenum target, GLint level, + GLint xoffset, GLint yoffset, + GLint x, GLint y, + GLsizei width, GLsizei height) +{ + struct gl_texture_unit *texUnit = _mesa_get_current_tex_unit(ctx); + struct gl_texture_object *texObj = _mesa_select_tex_object(ctx, texUnit, target); + struct gl_texture_image *texImage = _mesa_select_tex_image(ctx, texObj, target, level); + + if (!do_copy_texsubimage(ctx, target, level, + radeon_tex_obj(texObj), (radeon_texture_image *)texImage, + xoffset, yoffset, x, y, width, height)) { + + //DEBUG_FALLBACKS + + _mesa_meta_CopyTexSubImage2D(ctx, target, level, + xoffset, yoffset, x, y, width, height); + } +} + + +void r300_init_texcopy_functions(struct dd_function_table *table) +{ + table->CopyTexImage2D = r300CopyTexImage2D; + table->CopyTexSubImage2D = r300CopyTexSubImage2D; +}
\ No newline at end of file diff --git a/src/mesa/drivers/dri/r300/r300_texstate.c b/src/mesa/drivers/dri/r300/r300_texstate.c index f030451b282..6db56ba618f 100644 --- a/src/mesa/drivers/dri/r300/r300_texstate.c +++ b/src/mesa/drivers/dri/r300/r300_texstate.c @@ -39,7 +39,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/imports.h" #include "main/context.h" #include "main/macros.h" -#include "main/texformat.h" #include "main/teximage.h" #include "main/texobj.h" #include "main/enums.h" @@ -47,19 +46,10 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r300_context.h" #include "r300_state.h" -#include "r300_ioctl.h" #include "radeon_mipmap_tree.h" #include "r300_tex.h" #include "r300_reg.h" -#define VALID_FORMAT(f) ( ((f) <= MESA_FORMAT_RGBA_DXT5 \ - || ((f) >= MESA_FORMAT_RGBA_FLOAT32 && \ - (f) <= MESA_FORMAT_INTENSITY_FLOAT16)) \ - && tx_table[f].flag ) - -#define _ASSIGN(entry, format) \ - [ MESA_FORMAT_ ## entry ] = { format, 0, 1} - /* * Note that the _REV formats are the same as the non-REV formats. This is * because the REV and non-REV formats are identical as a byte string, but @@ -69,66 +59,119 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. * identically. -- paulus */ -static const struct tx_table { - GLuint format, filter, flag; -} tx_table[] = { - /* *INDENT-OFF* */ +uint32_t r300TranslateTexFormat(gl_format mesaFormat) +{ + switch (mesaFormat) + { #ifdef MESA_LITTLE_ENDIAN - _ASSIGN(RGBA8888, R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8)), - _ASSIGN(RGBA8888_REV, R300_EASY_TX_FORMAT(Z, Y, X, W, W8Z8Y8X8)), - _ASSIGN(ARGB8888, R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8)), - _ASSIGN(ARGB8888_REV, R300_EASY_TX_FORMAT(W, Z, Y, X, W8Z8Y8X8)), + case MESA_FORMAT_RGBA8888: + return R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8); + case MESA_FORMAT_RGBA8888_REV: + return R300_EASY_TX_FORMAT(Z, Y, X, W, W8Z8Y8X8); + case MESA_FORMAT_ARGB8888: + return R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8); + case MESA_FORMAT_ARGB8888_REV: + return R300_EASY_TX_FORMAT(W, Z, Y, X, W8Z8Y8X8); #else - _ASSIGN(RGBA8888, R300_EASY_TX_FORMAT(Z, Y, X, W, W8Z8Y8X8)), - _ASSIGN(RGBA8888_REV, R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8)), - _ASSIGN(ARGB8888, R300_EASY_TX_FORMAT(W, Z, Y, X, W8Z8Y8X8)), - _ASSIGN(ARGB8888_REV, R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8)), + case MESA_FORMAT_RGBA8888: + return R300_EASY_TX_FORMAT(Z, Y, X, W, W8Z8Y8X8); + case MESA_FORMAT_RGBA8888_REV: + return R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8); + case MESA_FORMAT_ARGB8888: + return R300_EASY_TX_FORMAT(W, Z, Y, X, W8Z8Y8X8); + case MESA_FORMAT_ARGB8888_REV: + return R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8); #endif - _ASSIGN(RGB888, R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8)), - _ASSIGN(RGB565, R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z5Y6X5)), - _ASSIGN(RGB565_REV, R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z5Y6X5)), - _ASSIGN(ARGB4444, R300_EASY_TX_FORMAT(X, Y, Z, W, W4Z4Y4X4)), - _ASSIGN(ARGB4444_REV, R300_EASY_TX_FORMAT(X, Y, Z, W, W4Z4Y4X4)), - _ASSIGN(ARGB1555, R300_EASY_TX_FORMAT(X, Y, Z, W, W1Z5Y5X5)), - _ASSIGN(ARGB1555_REV, R300_EASY_TX_FORMAT(X, Y, Z, W, W1Z5Y5X5)), - _ASSIGN(AL88, R300_EASY_TX_FORMAT(X, X, X, Y, Y8X8)), - _ASSIGN(AL88_REV, R300_EASY_TX_FORMAT(X, X, X, Y, Y8X8)), - _ASSIGN(RGB332, R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z3Y3X2)), - _ASSIGN(A8, R300_EASY_TX_FORMAT(ZERO, ZERO, ZERO, X, X8)), - _ASSIGN(L8, R300_EASY_TX_FORMAT(X, X, X, ONE, X8)), - _ASSIGN(I8, R300_EASY_TX_FORMAT(X, X, X, X, X8)), - _ASSIGN(CI8, R300_EASY_TX_FORMAT(X, X, X, X, X8)), - _ASSIGN(YCBCR, R300_EASY_TX_FORMAT(X, Y, Z, ONE, G8R8_G8B8) | R300_TX_FORMAT_YUV_MODE), - _ASSIGN(YCBCR_REV, R300_EASY_TX_FORMAT(X, Y, Z, ONE, G8R8_G8B8) | R300_TX_FORMAT_YUV_MODE), - _ASSIGN(RGB_DXT1, R300_EASY_TX_FORMAT(X, Y, Z, ONE, DXT1)), - _ASSIGN(RGBA_DXT1, R300_EASY_TX_FORMAT(X, Y, Z, W, DXT1)), - _ASSIGN(RGBA_DXT3, R300_EASY_TX_FORMAT(X, Y, Z, W, DXT3)), - _ASSIGN(RGBA_DXT5, R300_EASY_TX_FORMAT(Y, Z, W, X, DXT5)), - _ASSIGN(RGBA_FLOAT32, R300_EASY_TX_FORMAT(Z, Y, X, W, FL_R32G32B32A32)), - _ASSIGN(RGBA_FLOAT16, R300_EASY_TX_FORMAT(Z, Y, X, W, FL_R16G16B16A16)), - _ASSIGN(RGB_FLOAT32, 0xffffffff), - _ASSIGN(RGB_FLOAT16, 0xffffffff), - _ASSIGN(ALPHA_FLOAT32, R300_EASY_TX_FORMAT(ZERO, ZERO, ZERO, X, FL_I32)), - _ASSIGN(ALPHA_FLOAT16, R300_EASY_TX_FORMAT(ZERO, ZERO, ZERO, X, FL_I16)), - _ASSIGN(LUMINANCE_FLOAT32, R300_EASY_TX_FORMAT(X, X, X, ONE, FL_I32)), - _ASSIGN(LUMINANCE_FLOAT16, R300_EASY_TX_FORMAT(X, X, X, ONE, FL_I16)), - _ASSIGN(LUMINANCE_ALPHA_FLOAT32, R300_EASY_TX_FORMAT(X, X, X, Y, FL_I32A32)), - _ASSIGN(LUMINANCE_ALPHA_FLOAT16, R300_EASY_TX_FORMAT(X, X, X, Y, FL_I16A16)), - _ASSIGN(INTENSITY_FLOAT32, R300_EASY_TX_FORMAT(X, X, X, X, FL_I32)), - _ASSIGN(INTENSITY_FLOAT16, R300_EASY_TX_FORMAT(X, X, X, X, FL_I16)), - _ASSIGN(Z16, R300_EASY_TX_FORMAT(X, X, X, X, X16)), - _ASSIGN(Z24_S8, R300_EASY_TX_FORMAT(X, X, X, X, X24_Y8)), - _ASSIGN(S8_Z24, R300_EASY_TX_FORMAT(Y, Y, Y, Y, X24_Y8)), - _ASSIGN(Z32, R300_EASY_TX_FORMAT(X, X, X, X, X32)), - /* EXT_texture_sRGB */ - _ASSIGN(SRGBA8, R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8) | R300_TX_FORMAT_GAMMA), - _ASSIGN(SLA8, R300_EASY_TX_FORMAT(X, X, X, Y, Y8X8) | R300_TX_FORMAT_GAMMA), - _ASSIGN(SL8, R300_EASY_TX_FORMAT(X, X, X, ONE, X8) | R300_TX_FORMAT_GAMMA), - /* *INDENT-ON* */ + case MESA_FORMAT_XRGB8888: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8); + case MESA_FORMAT_RGB888: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8); + case MESA_FORMAT_RGB565: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z5Y6X5); + case MESA_FORMAT_RGB565_REV: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z5Y6X5); + case MESA_FORMAT_ARGB4444: + return R300_EASY_TX_FORMAT(X, Y, Z, W, W4Z4Y4X4); + case MESA_FORMAT_ARGB4444_REV: + return R300_EASY_TX_FORMAT(X, Y, Z, W, W4Z4Y4X4); + case MESA_FORMAT_ARGB1555: + return R300_EASY_TX_FORMAT(X, Y, Z, W, W1Z5Y5X5); + case MESA_FORMAT_ARGB1555_REV: + return R300_EASY_TX_FORMAT(X, Y, Z, W, W1Z5Y5X5); + case MESA_FORMAT_AL88: + return R300_EASY_TX_FORMAT(X, X, X, Y, Y8X8); + case MESA_FORMAT_AL88_REV: + return R300_EASY_TX_FORMAT(X, X, X, Y, Y8X8); + case MESA_FORMAT_RGB332: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z3Y3X2); + case MESA_FORMAT_A8: + return R300_EASY_TX_FORMAT(ZERO, ZERO, ZERO, X, X8); + case MESA_FORMAT_L8: + return R300_EASY_TX_FORMAT(X, X, X, ONE, X8); + case MESA_FORMAT_I8: + return R300_EASY_TX_FORMAT(X, X, X, X, X8); + case MESA_FORMAT_CI8: + return R300_EASY_TX_FORMAT(X, X, X, X, X8); + case MESA_FORMAT_YCBCR: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, G8R8_G8B8) | R300_TX_FORMAT_YUV_MODE; + case MESA_FORMAT_YCBCR_REV: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, G8R8_G8B8) | R300_TX_FORMAT_YUV_MODE; + case MESA_FORMAT_RGB_DXT1: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, DXT1); + case MESA_FORMAT_RGBA_DXT1: + return R300_EASY_TX_FORMAT(X, Y, Z, W, DXT1); + case MESA_FORMAT_RGBA_DXT3: + return R300_EASY_TX_FORMAT(X, Y, Z, W, DXT3); + case MESA_FORMAT_RGBA_DXT5: + return R300_EASY_TX_FORMAT(Y, Z, W, X, DXT5); + case MESA_FORMAT_RGBA_FLOAT32: + return R300_EASY_TX_FORMAT(Z, Y, X, W, FL_R32G32B32A32); + case MESA_FORMAT_RGBA_FLOAT16: + return R300_EASY_TX_FORMAT(Z, Y, X, W, FL_R16G16B16A16); + case MESA_FORMAT_ALPHA_FLOAT32: + return R300_EASY_TX_FORMAT(ZERO, ZERO, ZERO, X, FL_I32); + case MESA_FORMAT_ALPHA_FLOAT16: + return R300_EASY_TX_FORMAT(ZERO, ZERO, ZERO, X, FL_I16); + case MESA_FORMAT_LUMINANCE_FLOAT32: + return R300_EASY_TX_FORMAT(X, X, X, ONE, FL_I32); + case MESA_FORMAT_LUMINANCE_FLOAT16: + return R300_EASY_TX_FORMAT(X, X, X, ONE, FL_I16); + case MESA_FORMAT_LUMINANCE_ALPHA_FLOAT32: + return R300_EASY_TX_FORMAT(X, X, X, Y, FL_I32A32); + case MESA_FORMAT_LUMINANCE_ALPHA_FLOAT16: + return R300_EASY_TX_FORMAT(X, X, X, Y, FL_I16A16); + case MESA_FORMAT_INTENSITY_FLOAT32: + return R300_EASY_TX_FORMAT(X, X, X, X, FL_I32); + case MESA_FORMAT_INTENSITY_FLOAT16: + return R300_EASY_TX_FORMAT(X, X, X, X, FL_I16); + case MESA_FORMAT_Z16: + return R300_EASY_TX_FORMAT(X, X, X, X, X16); + case MESA_FORMAT_Z24_S8: + return R300_EASY_TX_FORMAT(X, X, X, X, X24_Y8); + case MESA_FORMAT_S8_Z24: + return R300_EASY_TX_FORMAT(Y, Y, Y, Y, X24_Y8); + case MESA_FORMAT_Z32: + return R300_EASY_TX_FORMAT(X, X, X, X, X32); + /* EXT_texture_sRGB */ + case MESA_FORMAT_SRGBA8: + return R300_EASY_TX_FORMAT(Y, Z, W, X, W8Z8Y8X8) | R300_TX_FORMAT_GAMMA; + case MESA_FORMAT_SLA8: + return R300_EASY_TX_FORMAT(X, X, X, Y, Y8X8) | R300_TX_FORMAT_GAMMA; + case MESA_FORMAT_SL8: + return R300_EASY_TX_FORMAT(X, X, X, ONE, X8) | R300_TX_FORMAT_GAMMA; + case MESA_FORMAT_SRGB_DXT1: + return R300_EASY_TX_FORMAT(X, Y, Z, ONE, DXT1) | R300_TX_FORMAT_GAMMA; + case MESA_FORMAT_SRGBA_DXT1: + return R300_EASY_TX_FORMAT(X, Y, Z, W, DXT1) | R300_TX_FORMAT_GAMMA; + case MESA_FORMAT_SRGBA_DXT3: + return R300_EASY_TX_FORMAT(X, Y, Z, W, DXT3) | R300_TX_FORMAT_GAMMA; + case MESA_FORMAT_SRGBA_DXT5: + return R300_EASY_TX_FORMAT(Y, Z, W, X, DXT5) | R300_TX_FORMAT_GAMMA; + default: + return 0; + } }; -#undef _ASSIGN - void r300SetDepthTexMode(struct gl_texture_object *tObj) { static const GLuint formats[3][3] = { @@ -156,11 +199,11 @@ void r300SetDepthTexMode(struct gl_texture_object *tObj) t = radeon_tex_obj(tObj); - switch (tObj->Image[0][tObj->BaseLevel]->TexFormat->MesaFormat) { + switch (tObj->Image[0][tObj->BaseLevel]->TexFormat) { case MESA_FORMAT_Z16: format = formats[0]; break; - case MESA_FORMAT_Z24_S8: + case MESA_FORMAT_S8_Z24: format = formats[1]; break; case MESA_FORMAT_Z32: @@ -203,32 +246,28 @@ void r300SetDepthTexMode(struct gl_texture_object *tObj) static void setup_hardware_state(r300ContextPtr rmesa, radeonTexObj *t) { const struct gl_texture_image *firstImage; - int firstlevel = t->mt ? t->mt->firstLevel : 0; - - firstImage = t->base.Image[0][firstlevel]; + firstImage = t->base.Image[0][t->minLod]; - if (!t->image_override - && VALID_FORMAT(firstImage->TexFormat->MesaFormat)) { - if (firstImage->TexFormat->BaseFormat == GL_DEPTH_COMPONENT) { + if (!t->image_override) { + if (firstImage->_BaseFormat == GL_DEPTH_COMPONENT) { r300SetDepthTexMode(&t->base); } else { - t->pp_txformat = tx_table[firstImage->TexFormat->MesaFormat].format; + t->pp_txformat = r300TranslateTexFormat(firstImage->TexFormat); + if (t->pp_txformat == 0) { + _mesa_problem(rmesa->radeon.glCtx, "%s: Invalid format %s", + __FUNCTION__, _mesa_get_format_name(firstImage->TexFormat)); + _mesa_exit(1); + } } - - t->pp_txfilter |= tx_table[firstImage->TexFormat->MesaFormat].filter; - } else if (!t->image_override) { - _mesa_problem(NULL, "unexpected texture format in %s", - __FUNCTION__); - return; } if (t->image_override && t->bo) return; - t->pp_txsize = (((firstImage->Width - 1) << R300_TX_WIDTHMASK_SHIFT) - | ((firstImage->Height - 1) << R300_TX_HEIGHTMASK_SHIFT) - | ((firstImage->DepthLog2) << R300_TX_DEPTHMASK_SHIFT) - | ((t->mt->lastLevel - t->mt->firstLevel) << R300_TX_MAX_MIP_LEVEL_SHIFT)); + t->pp_txsize = (((R300_TX_WIDTHMASK_MASK & ((firstImage->Width - 1) << R300_TX_WIDTHMASK_SHIFT))) + | ((R300_TX_HEIGHTMASK_MASK & ((firstImage->Height - 1) << R300_TX_HEIGHTMASK_SHIFT))) + | ((R300_TX_DEPTHMASK_MASK & ((firstImage->DepthLog2) << R300_TX_DEPTHMASK_SHIFT))) + | ((R300_TX_MAX_MIP_LEVEL_MASK & ((t->maxLod - t->minLod) << R300_TX_MAX_MIP_LEVEL_SHIFT)))); t->tile_bits = 0; @@ -239,7 +278,7 @@ static void setup_hardware_state(r300ContextPtr rmesa, radeonTexObj *t) if (t->base.Target == GL_TEXTURE_RECTANGLE_NV) { - unsigned int align = (64 / t->mt->bpp) - 1; + unsigned int align = (64 / _mesa_get_format_bytes(firstImage->TexFormat)) - 1; t->pp_txsize |= R300_TX_SIZE_TXPITCH_EN; if (!t->image_override) t->pp_txpitch = ((firstImage->Width + align) & ~align) - 1; @@ -248,8 +287,12 @@ static void setup_hardware_state(r300ContextPtr rmesa, radeonTexObj *t) if (rmesa->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) { if (firstImage->Width > 2048) t->pp_txpitch |= R500_TXWIDTH_BIT11; + else + t->pp_txpitch &= ~R500_TXWIDTH_BIT11; if (firstImage->Height > 2048) t->pp_txpitch |= R500_TXHEIGHT_BIT11; + else + t->pp_txpitch &= ~R500_TXHEIGHT_BIT11; } } @@ -355,18 +398,15 @@ void r300SetTexOffset(__DRIcontext * pDRICtx, GLint texname, switch (depth) { case 32: t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8); - t->pp_txfilter |= tx_table[2].filter; pitch_val /= 4; break; case 24: default: t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8); - t->pp_txfilter |= tx_table[4].filter; pitch_val /= 4; break; case 16: t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z5Y6X5); - t->pp_txfilter |= tx_table[5].filter; pitch_val /= 2; break; } @@ -407,18 +447,7 @@ void r300SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_fo return; } - radeon_update_renderbuffers(pDRICtx, dPriv); - /* back & depth buffer are useless free them right away */ - rb = (void*)rfb->base.Attachment[BUFFER_DEPTH].Renderbuffer; - if (rb && rb->bo) { - radeon_bo_unref(rb->bo); - rb->bo = NULL; - } - rb = (void*)rfb->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer; - if (rb && rb->bo) { - radeon_bo_unref(rb->bo); - rb->bo = NULL; - } + radeon_update_renderbuffers(pDRICtx, dPriv, GL_TRUE); rb = rfb->color_rb[0]; if (rb->bo == NULL) { /* Failed to BO for the buffer */ @@ -434,20 +463,13 @@ void r300SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_fo radeon_bo_unref(rImage->bo); rImage->bo = NULL; } - if (t->mt) { - radeon_miptree_unreference(t->mt); - t->mt = NULL; - } - if (rImage->mt) { - radeon_miptree_unreference(rImage->mt); - rImage->mt = NULL; - } + + radeon_miptree_unreference(&t->mt); + radeon_miptree_unreference(&rImage->mt); + _mesa_init_teximage_fields(radeon->glCtx, target, texImage, rb->base.Width, rb->base.Height, 1, 0, rb->cpp); texImage->RowStride = rb->pitch / rb->cpp; - texImage->TexFormat = radeonChooseTextureFormat(radeon->glCtx, - internalFormat, - type, format, 0); rImage->bo = rb->bo; radeon_bo_ref(rImage->bo); t->bo = rb->bo; @@ -463,32 +485,33 @@ void r300SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_fo t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8); else t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8); - t->pp_txfilter |= tx_table[2].filter; pitch_val /= 4; break; case 3: default: t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8); - t->pp_txfilter |= tx_table[4].filter; pitch_val /= 4; break; case 2: t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, ONE, Z5Y6X5); - t->pp_txfilter |= tx_table[5].filter; pitch_val /= 2; break; } pitch_val--; - t->pp_txsize = ((rb->base.Width - 1) << R300_TX_WIDTHMASK_SHIFT) | - ((rb->base.Height - 1) << R300_TX_HEIGHTMASK_SHIFT); + t->pp_txsize = (((R300_TX_WIDTHMASK_MASK & ((rb->base.Width - 1) << R300_TX_WIDTHMASK_SHIFT))) + | ((R300_TX_HEIGHTMASK_MASK & ((rb->base.Height - 1) << R300_TX_HEIGHTMASK_SHIFT)))); t->pp_txsize |= R300_TX_SIZE_TXPITCH_EN; t->pp_txpitch |= pitch_val; if (rmesa->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) { if (rb->base.Width > 2048) t->pp_txpitch |= R500_TXWIDTH_BIT11; + else + t->pp_txpitch &= ~R500_TXWIDTH_BIT11; if (rb->base.Height > 2048) t->pp_txpitch |= R500_TXHEIGHT_BIT11; + else + t->pp_txpitch &= ~R500_TXHEIGHT_BIT11; } t->validated = GL_TRUE; _mesa_unlock_texture(radeon->glCtx, texObj); diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.c b/src/mesa/drivers/dri/r300/r300_vertprog.c index dd0f27f9cba..c2f96af2c12 100644 --- a/src/mesa/drivers/dri/r300/r300_vertprog.c +++ b/src/mesa/drivers/dri/r300/r300_vertprog.c @@ -41,8 +41,9 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #include "tnl/tnl.h" #include "compiler/radeon_compiler.h" -#include "compiler/radeon_nqssadce.h" +#include "radeon_mesa_to_rc.h" #include "r300_context.h" +#include "r300_fragprog_common.h" #include "r300_state.h" /** @@ -61,12 +62,6 @@ static int r300VertexProgUpdateParams(GLcontext * ctx, struct r300_vertex_progra } } - if (vp->code.constants.Count * 4 > VSF_MAX_FRAGMENT_LENGTH) { - /* Should have checked this earlier... */ - fprintf(stderr, "%s:Params exhausted\n", __FUNCTION__); - _mesa_exit(-1); - } - for(i = 0; i < vp->code.constants.Count; ++i) { const float * src = 0; const struct rc_constant * constant = &vp->code.constants.Constants[i]; @@ -203,6 +198,34 @@ static void t_inputs_outputs(struct r300_vertex_program_compiler * c) } } +/** + * The NV_vertex_program spec mandates that all registers be + * initialized to zero. We do this here unconditionally. + * + * \note We rely on dead-code elimination in the compiler. + */ +static void initialize_NV_registers(struct radeon_compiler * compiler) +{ + unsigned int reg; + struct rc_instruction * inst; + + for(reg = 0; reg < 12; ++reg) { + inst = rc_insert_new_instruction(compiler, &compiler->Program.Instructions); + inst->U.I.Opcode = RC_OPCODE_MOV; + inst->U.I.DstReg.File = RC_FILE_TEMPORARY; + inst->U.I.DstReg.Index = reg; + inst->U.I.SrcReg[0].File = RC_FILE_NONE; + inst->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_0000; + } + + inst = rc_insert_new_instruction(compiler, &compiler->Program.Instructions); + inst->U.I.Opcode = RC_OPCODE_ARL; + inst->U.I.DstReg.File = RC_FILE_ADDRESS; + inst->U.I.DstReg.Index = 0; + inst->U.I.DstReg.WriteMask = WRITEMASK_X; + inst->U.I.SrcReg[0].File = RC_FILE_NONE; + inst->U.I.SrcReg[0].Swizzle = RC_SWIZZLE_0000; +} static struct r300_vertex_program *build_program(GLcontext *ctx, struct r300_vertex_program_key *wanted_key, @@ -232,7 +255,10 @@ static struct r300_vertex_program *build_program(GLcontext *ctx, _mesa_insert_mvp_code(ctx, vp->Base); } - rc_mesa_to_rc_program(&compiler.Base, &vp->Base->Base); + radeon_mesa_to_rc_program(&compiler.Base, &vp->Base->Base); + + if (mesa_vp->IsNVProgram) + initialize_NV_registers(&compiler.Base); rc_move_output(&compiler.Base, VERT_RESULT_PSIZ, VERT_RESULT_PSIZ, WRITEMASK_X); @@ -249,6 +275,11 @@ static struct r300_vertex_program *build_program(GLcontext *ctx, } r3xx_compile_vertex_program(&compiler); + + if (vp->code.constants.Count > ctx->Const.VertexProgram.MaxParameters) { + rc_error(&compiler.Base, "Program exceeds constant buffer size limit\n"); + } + vp->error = compiler.Base.Error; vp->Base->Base.InputsRead = vp->code.InputsRead; @@ -267,6 +298,20 @@ struct r300_vertex_program * r300SelectAndTranslateVertexShader(GLcontext *ctx) struct r300_vertex_program *vp; vpc = (struct r300_vertex_program_cont *)ctx->VertexProgram._Current; + + if (!r300->selected_fp) { + /* This can happen when GetProgramiv is called to check + * whether the program runs natively. + * + * To be honest, this is not a very good solution, + * but solving the problem of reporting good values + * for those queries is tough anyway considering that + * we recompile vertex programs based on the precise + * fragment program that is in use. + */ + r300SelectAndTranslateFragmentShader(ctx); + } + wanted_key.FpReads = r300->selected_fp->InputsRead; wanted_key.FogAttr = r300->selected_fp->fog_attr; wanted_key.WPosAttr = r300->selected_fp->wpos_attr; @@ -288,7 +333,6 @@ struct r300_vertex_program * r300SelectAndTranslateVertexShader(GLcontext *ctx) #define bump_vpu_count(ptr, new_count) do { \ drm_r300_cmd_header_t* _p=((drm_r300_cmd_header_t*)(ptr)); \ int _nc=(new_count)/4; \ - assert(_nc < 256); \ if(_nc>_p->vpu.count)_p->vpu.count=_nc; \ } while(0) diff --git a/src/mesa/drivers/dri/r300/radeon_mesa_to_rc.c b/src/mesa/drivers/dri/r300/radeon_mesa_to_rc.c new file mode 100644 index 00000000000..9f9dec840b4 --- /dev/null +++ b/src/mesa/drivers/dri/r300/radeon_mesa_to_rc.c @@ -0,0 +1,223 @@ +/* + * Copyright (C) 2009 Nicolai Haehnle. + * + * 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, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "radeon_mesa_to_rc.h" + +#include "main/mtypes.h" +#include "shader/prog_instruction.h" +#include "shader/prog_parameter.h" + +#include "compiler/radeon_compiler.h" +#include "compiler/radeon_program.h" + + +static rc_opcode translate_opcode(gl_inst_opcode opcode) +{ + switch(opcode) { + case OPCODE_NOP: return RC_OPCODE_NOP; + case OPCODE_ABS: return RC_OPCODE_ABS; + case OPCODE_ADD: return RC_OPCODE_ADD; + case OPCODE_ARL: return RC_OPCODE_ARL; + case OPCODE_CMP: return RC_OPCODE_CMP; + case OPCODE_COS: return RC_OPCODE_COS; + case OPCODE_DDX: return RC_OPCODE_DDX; + case OPCODE_DDY: return RC_OPCODE_DDY; + case OPCODE_DP3: return RC_OPCODE_DP3; + case OPCODE_DP4: return RC_OPCODE_DP4; + case OPCODE_DPH: return RC_OPCODE_DPH; + case OPCODE_DST: return RC_OPCODE_DST; + case OPCODE_EX2: return RC_OPCODE_EX2; + case OPCODE_EXP: return RC_OPCODE_EXP; + case OPCODE_FLR: return RC_OPCODE_FLR; + case OPCODE_FRC: return RC_OPCODE_FRC; + case OPCODE_KIL: return RC_OPCODE_KIL; + case OPCODE_LG2: return RC_OPCODE_LG2; + case OPCODE_LIT: return RC_OPCODE_LIT; + case OPCODE_LOG: return RC_OPCODE_LOG; + case OPCODE_LRP: return RC_OPCODE_LRP; + case OPCODE_MAD: return RC_OPCODE_MAD; + case OPCODE_MAX: return RC_OPCODE_MAX; + case OPCODE_MIN: return RC_OPCODE_MIN; + case OPCODE_MOV: return RC_OPCODE_MOV; + case OPCODE_MUL: return RC_OPCODE_MUL; + case OPCODE_POW: return RC_OPCODE_POW; + case OPCODE_RCP: return RC_OPCODE_RCP; + case OPCODE_RSQ: return RC_OPCODE_RSQ; + case OPCODE_SCS: return RC_OPCODE_SCS; + case OPCODE_SEQ: return RC_OPCODE_SEQ; + case OPCODE_SFL: return RC_OPCODE_SFL; + case OPCODE_SGE: return RC_OPCODE_SGE; + case OPCODE_SGT: return RC_OPCODE_SGT; + case OPCODE_SIN: return RC_OPCODE_SIN; + case OPCODE_SLE: return RC_OPCODE_SLE; + case OPCODE_SLT: return RC_OPCODE_SLT; + case OPCODE_SNE: return RC_OPCODE_SNE; + case OPCODE_SUB: return RC_OPCODE_SUB; + case OPCODE_SWZ: return RC_OPCODE_SWZ; + case OPCODE_TEX: return RC_OPCODE_TEX; + case OPCODE_TXB: return RC_OPCODE_TXB; + case OPCODE_TXD: return RC_OPCODE_TXD; + case OPCODE_TXL: return RC_OPCODE_TXL; + case OPCODE_TXP: return RC_OPCODE_TXP; + case OPCODE_XPD: return RC_OPCODE_XPD; + default: return RC_OPCODE_ILLEGAL_OPCODE; + } +} + +static rc_saturate_mode translate_saturate(unsigned int saturate) +{ + switch(saturate) { + default: + case SATURATE_OFF: return RC_SATURATE_NONE; + case SATURATE_ZERO_ONE: return RC_SATURATE_ZERO_ONE; + } +} + +static rc_register_file translate_register_file(unsigned int file) +{ + switch(file) { + case PROGRAM_TEMPORARY: return RC_FILE_TEMPORARY; + case PROGRAM_INPUT: return RC_FILE_INPUT; + case PROGRAM_OUTPUT: return RC_FILE_OUTPUT; + case PROGRAM_LOCAL_PARAM: + case PROGRAM_ENV_PARAM: + case PROGRAM_STATE_VAR: + case PROGRAM_NAMED_PARAM: + case PROGRAM_CONSTANT: + case PROGRAM_UNIFORM: return RC_FILE_CONSTANT; + case PROGRAM_ADDRESS: return RC_FILE_ADDRESS; + default: return RC_FILE_NONE; + } +} + +static void translate_srcreg(struct rc_src_register * dest, struct prog_src_register * src) +{ + dest->File = translate_register_file(src->File); + dest->Index = src->Index; + dest->RelAddr = src->RelAddr; + dest->Swizzle = src->Swizzle; + dest->Abs = src->Abs; + dest->Negate = src->Negate; +} + +static void translate_dstreg(struct rc_dst_register * dest, struct prog_dst_register * src) +{ + dest->File = translate_register_file(src->File); + dest->Index = src->Index; + dest->RelAddr = src->RelAddr; + dest->WriteMask = src->WriteMask; +} + +static rc_texture_target translate_tex_target(gl_texture_index target) +{ + switch(target) { + case TEXTURE_2D_ARRAY_INDEX: return RC_TEXTURE_2D_ARRAY; + case TEXTURE_1D_ARRAY_INDEX: return RC_TEXTURE_1D_ARRAY; + case TEXTURE_CUBE_INDEX: return RC_TEXTURE_CUBE; + case TEXTURE_3D_INDEX: return RC_TEXTURE_3D; + case TEXTURE_RECT_INDEX: return RC_TEXTURE_RECT; + default: + case TEXTURE_2D_INDEX: return RC_TEXTURE_2D; + case TEXTURE_1D_INDEX: return RC_TEXTURE_1D; + } +} + +static void translate_instruction(struct radeon_compiler * c, + struct rc_instruction * dest, struct prog_instruction * src) +{ + const struct rc_opcode_info * opcode; + unsigned int i; + + dest->U.I.Opcode = translate_opcode(src->Opcode); + if (dest->U.I.Opcode == RC_OPCODE_ILLEGAL_OPCODE) { + rc_error(c, "Unsupported opcode %i\n", src->Opcode); + return; + } + dest->U.I.SaturateMode = translate_saturate(src->SaturateMode); + + opcode = rc_get_opcode_info(dest->U.I.Opcode); + + for(i = 0; i < opcode->NumSrcRegs; ++i) + translate_srcreg(&dest->U.I.SrcReg[i], &src->SrcReg[i]); + + if (opcode->HasDstReg) + translate_dstreg(&dest->U.I.DstReg, &src->DstReg); + + if (opcode->HasTexture) { + dest->U.I.TexSrcUnit = src->TexSrcUnit; + dest->U.I.TexSrcTarget = translate_tex_target(src->TexSrcTarget); + dest->U.I.TexShadow = src->TexShadow; + } +} + +void radeon_mesa_to_rc_program(struct radeon_compiler * c, struct gl_program * program) +{ + struct prog_instruction *source; + unsigned int i; + + for(source = program->Instructions; source->Opcode != OPCODE_END; ++source) { + struct rc_instruction * dest = rc_insert_new_instruction(c, c->Program.Instructions.Prev); + translate_instruction(c, dest, source); + } + + c->Program.ShadowSamplers = program->ShadowSamplers; + c->Program.InputsRead = program->InputsRead; + c->Program.OutputsWritten = program->OutputsWritten; + + int isNVProgram = 0; + + if (program->Target == GL_VERTEX_PROGRAM_ARB) { + struct gl_vertex_program * vp = (struct gl_vertex_program *) program; + isNVProgram = vp->IsNVProgram; + } + + if (isNVProgram) { + /* NV_vertex_program has a fixed-sized constant environment. + * This could be handled more efficiently for programs that + * do not use relative addressing. + */ + for(i = 0; i < 96; ++i) { + struct rc_constant constant; + + constant.Type = RC_CONSTANT_EXTERNAL; + constant.Size = 4; + constant.u.External = i; + + rc_constants_add(&c->Program.Constants, &constant); + } + } else { + for(i = 0; i < program->Parameters->NumParameters; ++i) { + struct rc_constant constant; + + constant.Type = RC_CONSTANT_EXTERNAL; + constant.Size = 4; + constant.u.External = i; + + rc_constants_add(&c->Program.Constants, &constant); + } + } +} diff --git a/src/mesa/drivers/dri/r300/radeon_mesa_to_rc.h b/src/mesa/drivers/dri/r300/radeon_mesa_to_rc.h new file mode 100644 index 00000000000..9511a04f36f --- /dev/null +++ b/src/mesa/drivers/dri/r300/radeon_mesa_to_rc.h @@ -0,0 +1,36 @@ +/* + * Copyright (C) 2009 Nicolai Haehnle. + * + * 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, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef RADEON_MESA_TO_RC_H +#define RADEON_MESA_TO_RC_H + +struct gl_program; +struct radeon_compiler; + +void radeon_mesa_to_rc_program(struct radeon_compiler * c, struct gl_program * program); + +#endif /* RADEON_MESA_TO_RC_H */ diff --git a/src/mesa/drivers/dri/r600/Makefile b/src/mesa/drivers/dri/r600/Makefile index 36bf773c054..9b7c42042e6 100644 --- a/src/mesa/drivers/dri/r600/Makefile +++ b/src/mesa/drivers/dri/r600/Makefile @@ -29,6 +29,7 @@ COMMON_SOURCES = \ RADEON_COMMON_SOURCES = \ radeon_bo_legacy.c \ radeon_common_context.c \ + radeon_buffer_objects.c \ radeon_common.c \ radeon_cs_legacy.c \ radeon_dma.c \ @@ -75,4 +76,3 @@ DRI_LIB_DEPS += $(RADEON_LDFLAGS) include ../Makefile.template -symlinks: diff --git a/src/mesa/drivers/dri/r600/r600_cmdbuf.c b/src/mesa/drivers/dri/r600/r600_cmdbuf.c index 3cfe03a45f7..d27a3245a39 100644 --- a/src/mesa/drivers/dri/r600/r600_cmdbuf.c +++ b/src/mesa/drivers/dri/r600/r600_cmdbuf.c @@ -254,7 +254,7 @@ static int r600_cs_process_relocs(struct radeon_cs *cs, relocs = (struct r600_cs_reloc_legacy *)cs->relocs; restart: for (i = 0; i < cs->crelocs; i++) { - uint32_t soffset, eoffset, asicoffset; + uint32_t soffset, eoffset; r = radeon_bo_legacy_validate(relocs[i].base.bo, &soffset, &eoffset); @@ -262,24 +262,12 @@ restart: goto restart; } if (r) { - fprintf(stderr, "validated %p [0x%08X, 0x%08X]\n", + fprintf(stderr, "invalid bo(%p) [0x%08X, 0x%08X]\n", relocs[i].base.bo, soffset, eoffset); return r; } - asicoffset = soffset; for (j = 0; j < relocs[i].cindices; j++) { - if (asicoffset >= eoffset) { - /* radeon_bo_debug(relocs[i].base.bo, 12); */ - fprintf(stderr, "validated %p [0x%08X, 0x%08X]\n", - relocs[i].base.bo, soffset, eoffset); - fprintf(stderr, "above end: %p 0x%08X 0x%08X\n", - relocs[i].base.bo, - cs->packets[relocs[i].indices[j]], - eoffset); - exit(0); - return -EINVAL; - } /* pkt3 nop header in ib chunk */ cs->packets[relocs[i].reloc_indices[j]] = 0xC0001000; /* reloc index in ib chunk */ @@ -287,7 +275,7 @@ restart: } /* asic offset in reloc chunk */ /* see alex drm r600_nomm_relocate */ - reloc_chunk[offset_dw] = asicoffset; + reloc_chunk[offset_dw] = soffset; reloc_chunk[offset_dw + 3] = 0; offset_dw += 4; diff --git a/src/mesa/drivers/dri/r600/r600_context.c b/src/mesa/drivers/dri/r600/r600_context.c index f8fd9c13d77..25314eff563 100644 --- a/src/mesa/drivers/dri/r600/r600_context.c +++ b/src/mesa/drivers/dri/r600/r600_context.c @@ -59,10 +59,12 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "radeon_debug.h" #include "r600_context.h" #include "radeon_common_context.h" +#include "radeon_buffer_objects.h" #include "radeon_span.h" #include "r600_cmdbuf.h" #include "r600_emit.h" #include "radeon_bocs_wrapper.h" +#include "radeon_queryobj.h" #include "r700_state.h" #include "r700_ioctl.h" @@ -72,11 +74,10 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "utils.h" #include "xmlpool.h" /* for symbolic values of enum-type options */ -/* hw_tcl_on derives from future_hw_tcl_on when its safe to change it. */ -int future_hw_tcl_on = 1; -int hw_tcl_on = 1; +//#define R600_ENABLE_GLSL_TEST 1 #define need_GL_VERSION_2_0 +#define need_GL_ARB_occlusion_query #define need_GL_ARB_point_parameters #define need_GL_ARB_vertex_program #define need_GL_EXT_blend_equation_separate @@ -91,14 +92,14 @@ int hw_tcl_on = 1; #define need_GL_ATI_separate_stencil #define need_GL_NV_vertex_program -#include "extension_helper.h" +#include "main/remap_helper.h" -extern const struct tnl_pipeline_stage *r700_pipeline[]; - -const struct dri_extension card_extensions[] = { +static const struct dri_extension card_extensions[] = { /* *INDENT-OFF* */ + {"GL_ARB_depth_clamp", NULL}, {"GL_ARB_depth_texture", NULL}, {"GL_ARB_fragment_program", NULL}, + {"GL_ARB_occlusion_query", GL_ARB_occlusion_query_functions}, {"GL_ARB_multitexture", NULL}, {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions}, {"GL_ARB_shadow", NULL}, @@ -110,6 +111,7 @@ const struct dri_extension card_extensions[] = { {"GL_ARB_texture_env_crossbar", NULL}, {"GL_ARB_texture_env_dot3", NULL}, {"GL_ARB_texture_mirrored_repeat", NULL}, + {"GL_ARB_texture_non_power_of_two", NULL}, {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions}, {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions}, {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions}, @@ -130,6 +132,7 @@ const struct dri_extension card_extensions[] = { {"GL_EXT_texture_lod_bias", NULL}, {"GL_EXT_texture_mirror_clamp", NULL}, {"GL_EXT_texture_rectangle", NULL}, + {"GL_EXT_vertex_array_bgra", NULL}, {"GL_EXT_texture_sRGB", NULL}, {"GL_ATI_separate_stencil", GL_ATI_separate_stencil_functions}, {"GL_ATI_texture_env_combine3", NULL}, @@ -145,7 +148,7 @@ const struct dri_extension card_extensions[] = { }; -const struct dri_extension mm_extensions[] = { +static const struct dri_extension mm_extensions[] = { { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions }, { NULL, NULL } }; @@ -154,21 +157,28 @@ const struct dri_extension mm_extensions[] = { * The GL 2.0 functions are needed to make display lists work with * functions added by GL_ATI_separate_stencil. */ -const struct dri_extension gl_20_extension[] = { +static const struct dri_extension gl_20_extension[] = { +#ifdef R600_ENABLE_GLSL_TEST + {"GL_ARB_shading_language_100", GL_VERSION_2_0_functions }, +#else {"GL_VERSION_2_0", GL_VERSION_2_0_functions }, +#endif /* R600_ENABLE_GLSL_TEST */ }; - -static void r600RunPipeline(GLcontext * ctx) -{ - _mesa_lock_context_textures(ctx); - - if (ctx->NewState) - _mesa_update_state_locked(ctx); - - _tnl_run_pipeline(ctx); - _mesa_unlock_context_textures(ctx); -} +static const struct tnl_pipeline_stage *r600_pipeline[] = { + /* Catch any t&l fallbacks + */ + &_tnl_vertex_transform_stage, + &_tnl_normal_transform_stage, + &_tnl_lighting_stage, + &_tnl_fog_coordinate_stage, + &_tnl_texgen_stage, + &_tnl_texture_transform_stage, + &_tnl_point_attenuation_stage, + &_tnl_vertex_program_stage, + &_tnl_render_stage, + 0, +}; static void r600_get_lock(radeonContextPtr rmesa) { @@ -179,7 +189,7 @@ static void r600_get_lock(radeonContextPtr rmesa) if (!rmesa->radeonScreen->kernel_mm) radeon_bo_legacy_texture_age(rmesa->radeonScreen->bom); } -} +} static void r600_vtbl_emit_cs_header(struct radeon_cs *cs, radeonContextPtr rmesa) { @@ -201,6 +211,24 @@ static void r600_fallback(GLcontext *ctx, GLuint bit, GLboolean mode) context->radeon.Fallback &= ~bit; } +static void r600_emit_query_finish(radeonContextPtr radeon) +{ + context_t *context = (context_t*) radeon; + BATCH_LOCALS(&context->radeon); + + struct radeon_query_object *query = radeon->query.current; + + BEGIN_BATCH_NO_AUTOSTATE(4 + 2); + R600_OUT_BATCH(CP_PACKET3(R600_IT_EVENT_WRITE, 2)); + R600_OUT_BATCH(ZPASS_DONE); + R600_OUT_BATCH(query->curr_offset + 8); /* hw writes qwords */ + R600_OUT_BATCH(0x00000000); + R600_OUT_BATCH_RELOC(VGT_EVENT_INITIATOR, query->bo, 0, 0, RADEON_GEM_DOMAIN_GTT, 0); + END_BATCH(); + assert(query->curr_offset < RADEON_QUERY_PAGE_SIZE); + query->emitted_begin = GL_FALSE; +} + static void r600_init_vtbl(radeonContextPtr radeon) { radeon->vtbl.get_lock = r600_get_lock; @@ -209,6 +237,121 @@ static void r600_init_vtbl(radeonContextPtr radeon) radeon->vtbl.swtcl_flush = NULL; radeon->vtbl.pre_emit_atoms = r600_vtbl_pre_emit_atoms; radeon->vtbl.fallback = r600_fallback; + radeon->vtbl.emit_query_finish = r600_emit_query_finish; +} + +static void r600InitConstValues(GLcontext *ctx, radeonScreenPtr screen) +{ + context_t *r600 = R700_CONTEXT(ctx); + + ctx->Const.MaxTextureImageUnits = + driQueryOptioni(&r600->radeon.optionCache, "texture_image_units"); + ctx->Const.MaxTextureCoordUnits = + driQueryOptioni(&r600->radeon.optionCache, "texture_coord_units"); + ctx->Const.MaxTextureUnits = + MIN2(ctx->Const.MaxTextureImageUnits, + ctx->Const.MaxTextureCoordUnits); + ctx->Const.MaxTextureMaxAnisotropy = 16.0; + ctx->Const.MaxTextureLodBias = 16.0; + + ctx->Const.MaxTextureLevels = 13; /* hw support 14 */ + ctx->Const.MaxTextureRectSize = 4096; /* hw support 8192 */ + + ctx->Const.MinPointSize = 0x0001 / 8.0; + ctx->Const.MinPointSizeAA = 0x0001 / 8.0; + ctx->Const.MaxPointSize = 0xffff / 8.0; + ctx->Const.MaxPointSizeAA = 0xffff / 8.0; + + ctx->Const.MinLineWidth = 0x0001 / 8.0; + ctx->Const.MinLineWidthAA = 0x0001 / 8.0; + ctx->Const.MaxLineWidth = 0xffff / 8.0; + ctx->Const.MaxLineWidthAA = 0xffff / 8.0; + + ctx->Const.MaxDrawBuffers = 1; /* hw supports 8 */ + + /* 256 for reg-based consts, inline consts also supported */ + ctx->Const.VertexProgram.MaxInstructions = 8192; /* in theory no limit */ + ctx->Const.VertexProgram.MaxNativeInstructions = 8192; + ctx->Const.VertexProgram.MaxNativeAttribs = 160; + ctx->Const.VertexProgram.MaxTemps = 128; + ctx->Const.VertexProgram.MaxNativeTemps = 128; + ctx->Const.VertexProgram.MaxNativeParameters = 256; + ctx->Const.VertexProgram.MaxNativeAddressRegs = 1; /* ??? */ + + ctx->Const.FragmentProgram.MaxNativeTemps = 128; + ctx->Const.FragmentProgram.MaxNativeAttribs = 32; + ctx->Const.FragmentProgram.MaxNativeParameters = 256; + ctx->Const.FragmentProgram.MaxNativeAluInstructions = 8192; + /* 8 per clause on r6xx, 16 on rv670/r7xx */ + if ((screen->chip_family == CHIP_FAMILY_RV670) || + (screen->chip_family >= CHIP_FAMILY_RV770)) + ctx->Const.FragmentProgram.MaxNativeTexInstructions = 16; + else + ctx->Const.FragmentProgram.MaxNativeTexInstructions = 8; + ctx->Const.FragmentProgram.MaxNativeInstructions = 8192; + ctx->Const.FragmentProgram.MaxNativeTexIndirections = 8; /* ??? */ + ctx->Const.FragmentProgram.MaxNativeAddressRegs = 0; /* and these are?? */ +} + +static void r600ParseOptions(context_t *r600, radeonScreenPtr screen) +{ + /* Parse configuration files. + * Do this here so that initialMaxAnisotropy is set before we create + * the default textures. + */ + driParseConfigFiles(&r600->radeon.optionCache, &screen->optionCache, + screen->driScreen->myNum, "r600"); + + r600->radeon.initialMaxAnisotropy = driQueryOptionf(&r600->radeon.optionCache, + "def_max_anisotropy"); + +} + +static void r600InitGLExtensions(GLcontext *ctx) +{ + context_t *r600 = R700_CONTEXT(ctx); + + driInitExtensions(ctx, card_extensions, GL_TRUE); + if (r600->radeon.radeonScreen->kernel_mm) + driInitExtensions(ctx, mm_extensions, GL_FALSE); + +#ifdef R600_ENABLE_GLSL_TEST + driInitExtensions(ctx, gl_20_extension, GL_TRUE); + //_mesa_enable_2_0_extensions(ctx); + //1.5 + ctx->Extensions.ARB_occlusion_query = GL_TRUE; + ctx->Extensions.ARB_vertex_buffer_object = GL_TRUE; + ctx->Extensions.EXT_shadow_funcs = GL_TRUE; + //2.0 + ctx->Extensions.ARB_draw_buffers = GL_TRUE; + ctx->Extensions.ARB_point_sprite = GL_TRUE; + ctx->Extensions.ARB_shader_objects = GL_TRUE; + ctx->Extensions.ARB_vertex_shader = GL_TRUE; + ctx->Extensions.ARB_fragment_shader = GL_TRUE; + ctx->Extensions.EXT_blend_equation_separate = GL_TRUE; + ctx->Extensions.ATI_separate_stencil = GL_TRUE; + + /* glsl compiler has problem if this is not GL_TRUE */ + ctx->Shader.EmitCondCodes = GL_TRUE; +#endif /* R600_ENABLE_GLSL_TEST */ + + if (driQueryOptionb + (&r600->radeon.optionCache, "disable_stencil_two_side")) + _mesa_disable_extension(ctx, "GL_EXT_stencil_two_side"); + + if (r600->radeon.glCtx->Mesa_DXTn + && !driQueryOptionb(&r600->radeon.optionCache, "disable_s3tc")) { + _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); + _mesa_enable_extension(ctx, "GL_S3_s3tc"); + } else + if (driQueryOptionb(&r600->radeon.optionCache, "force_s3tc_enable")) + { + _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); + } + + /* XXX: RV740 only seems to report results from half of its DBs */ + if (r600->radeon.radeonScreen->chip_family == CHIP_FAMILY_RV740) + _mesa_disable_extension(ctx, "GL_ARB_occlusion_query"); } /* Create the device specific rendering context. @@ -234,19 +377,10 @@ GLboolean r600CreateContext(const __GLcontextModes * glVisual, return GL_FALSE; } - if (!(screen->chip_flags & RADEON_CHIPSET_TCL)) - hw_tcl_on = future_hw_tcl_on = 0; + r600ParseOptions(r600, screen); + r600->radeon.radeonScreen = screen; r600_init_vtbl(&r600->radeon); - /* Parse configuration files. - * Do this here so that initialMaxAnisotropy is set before we create - * the default textures. - */ - driParseConfigFiles(&r600->radeon.optionCache, &screen->optionCache, - screen->driScreen->myNum, "r600"); - - r600->radeon.initialMaxAnisotropy = driQueryOptionf(&r600->radeon.optionCache, - "def_max_anisotropy"); /* Init default driver functions then plug in our R600-specific functions * (the texture functions are especially important) @@ -256,7 +390,9 @@ GLboolean r600CreateContext(const __GLcontextModes * glVisual, r700InitStateFuncs(&functions); r600InitTextureFuncs(&functions); r700InitShaderFuncs(&functions); + radeonInitQueryObjFunctions(&functions); r700InitIoctlFuncs(&functions); + radeonInitBufferObjectFuncs(&functions); if (!radeonInitContext(&r600->radeon, &functions, glVisual, driContextPriv, @@ -266,44 +402,14 @@ GLboolean r600CreateContext(const __GLcontextModes * glVisual, return GL_FALSE; } - /* Init r600 context data */ - /* Set the maximum texture size small enough that we can guarentee that - * all texture units can bind a maximal texture and have them both in - * texturable memory at once. - */ - ctx = r600->radeon.glCtx; - ctx->Const.MaxTextureImageUnits = - driQueryOptioni(&r600->radeon.optionCache, "texture_image_units"); - ctx->Const.MaxTextureCoordUnits = - driQueryOptioni(&r600->radeon.optionCache, "texture_coord_units"); - ctx->Const.MaxTextureUnits = - MIN2(ctx->Const.MaxTextureImageUnits, - ctx->Const.MaxTextureCoordUnits); - ctx->Const.MaxTextureMaxAnisotropy = 16.0; - ctx->Const.MaxTextureLodBias = 16.0; - - ctx->Const.MaxTextureLevels = 13; - ctx->Const.MaxTextureRectSize = 4096; - - ctx->Const.MinPointSize = 0x0001 / 8.0; - ctx->Const.MinPointSizeAA = 0x0001 / 8.0; - ctx->Const.MaxPointSize = 0xffff / 8.0; - ctx->Const.MaxPointSizeAA = 0xffff / 8.0; - - ctx->Const.MinLineWidth = 0x0001 / 8.0; - ctx->Const.MinLineWidthAA = 0x0001 / 8.0; - ctx->Const.MaxLineWidth = 0xffff / 8.0; - ctx->Const.MaxLineWidthAA = 0xffff / 8.0; + ctx->VertexProgram._MaintainTnlProgram = GL_TRUE; + ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE; - /* Needs further modifications */ -#if 0 - ctx->Const.MaxArrayLockSize = - ( /*512 */ RADEON_BUFFER_SIZE * 16 * 1024) / (4 * 4); -#endif + r600InitConstValues(ctx, screen); - ctx->Const.MaxDrawBuffers = 1; + _mesa_set_mvp_with_dp4( ctx, GL_TRUE ); /* Initialize the software rasterizer and helper modules. */ @@ -312,16 +418,12 @@ GLboolean r600CreateContext(const __GLcontextModes * glVisual, _tnl_CreateContext(ctx); _swsetup_CreateContext(ctx); _swsetup_Wakeup(ctx); - _ae_create_context(ctx); /* Install the customized pipeline: */ _tnl_destroy_pipeline(ctx); - _tnl_install_pipeline(ctx, r700_pipeline); - - /* Try and keep materials and vertices separate: - */ -/* _tnl_isolate_materials(ctx, GL_TRUE); */ + _tnl_install_pipeline(ctx, r600_pipeline); + TNL_CONTEXT(ctx)->Driver.RunPipeline = _tnl_run_pipeline; /* Configure swrast and TNL to match hardware characteristics: */ @@ -330,62 +432,16 @@ GLboolean r600CreateContext(const __GLcontextModes * glVisual, _tnl_allow_pixel_fog(ctx, GL_FALSE); _tnl_allow_vertex_fog(ctx, GL_TRUE); - /* currently bogus data */ - ctx->Const.VertexProgram.MaxInstructions = VSF_MAX_FRAGMENT_LENGTH / 4; - ctx->Const.VertexProgram.MaxNativeInstructions = - VSF_MAX_FRAGMENT_LENGTH / 4; - ctx->Const.VertexProgram.MaxNativeAttribs = 16; /* r420 */ - ctx->Const.VertexProgram.MaxTemps = 32; - ctx->Const.VertexProgram.MaxNativeTemps = - /*VSF_MAX_FRAGMENT_TEMPS */ 32; - ctx->Const.VertexProgram.MaxNativeParameters = 256; /* r420 */ - ctx->Const.VertexProgram.MaxNativeAddressRegs = 1; - - ctx->Const.FragmentProgram.MaxNativeTemps = PFS_NUM_TEMP_REGS; - ctx->Const.FragmentProgram.MaxNativeAttribs = 11; /* copy i915... */ - ctx->Const.FragmentProgram.MaxNativeParameters = PFS_NUM_CONST_REGS; - ctx->Const.FragmentProgram.MaxNativeAluInstructions = PFS_MAX_ALU_INST; - ctx->Const.FragmentProgram.MaxNativeTexInstructions = PFS_MAX_TEX_INST; - ctx->Const.FragmentProgram.MaxNativeInstructions = - PFS_MAX_ALU_INST + PFS_MAX_TEX_INST; - ctx->Const.FragmentProgram.MaxNativeTexIndirections = - PFS_MAX_TEX_INDIRECT; - ctx->Const.FragmentProgram.MaxNativeAddressRegs = 0; /* and these are?? */ - ctx->VertexProgram._MaintainTnlProgram = GL_TRUE; - ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE; - radeon_init_debug(); - driInitExtensions(ctx, card_extensions, GL_TRUE); - if (r600->radeon.radeonScreen->kernel_mm) - driInitExtensions(ctx, mm_extensions, GL_FALSE); - - if (driQueryOptionb - (&r600->radeon.optionCache, "disable_stencil_two_side")) - _mesa_disable_extension(ctx, "GL_EXT_stencil_two_side"); - - if (r600->radeon.glCtx->Mesa_DXTn - && !driQueryOptionb(&r600->radeon.optionCache, "disable_s3tc")) { - _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); - _mesa_enable_extension(ctx, "GL_S3_s3tc"); - } else - if (driQueryOptionb(&r600->radeon.optionCache, "force_s3tc_enable")) - { - _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); - } + r700InitDraw(ctx); radeon_fbo_init(&r600->radeon); radeonInitSpanFuncs( ctx ); - r600InitCmdBuf(r600); - r700InitState(r600->radeon.glCtx); - TNL_CONTEXT(ctx)->Driver.RunPipeline = r600RunPipeline; - - if (driQueryOptionb(&r600->radeon.optionCache, "no_rast")) { - radeon_warning("disabling 3D acceleration\n"); - } + r600InitGLExtensions(ctx); return GL_TRUE; } diff --git a/src/mesa/drivers/dri/r600/r600_context.h b/src/mesa/drivers/dri/r600/r600_context.h index c59df7505af..394fd757d4b 100644 --- a/src/mesa/drivers/dri/r600/r600_context.h +++ b/src/mesa/drivers/dri/r600/r600_context.h @@ -58,57 +58,15 @@ typedef struct r600_context context_t; #include "main/mm.h" -/************ DMA BUFFERS **************/ - -/* The blit width for texture uploads - */ -#define R600_BLIT_WIDTH_BYTES 1024 -#define R600_MAX_TEXTURE_UNITS 8 - -struct r600_texture_state { - int tc_count; /* number of incoming texture coordinates from VAP */ -}; - -/* Perhaps more if we store programs in vmem? */ -/* drm_r600_cmd_header_t->vpu->count is unsigned char */ -#define VSF_MAX_FRAGMENT_LENGTH (255*4) - -/* Can be tested with colormat currently. */ -#define VSF_MAX_FRAGMENT_TEMPS (14) - -#define STATE_R600_WINDOW_DIMENSION (STATE_INTERNAL_DRIVER+0) -#define STATE_R600_TEXRECT_FACTOR (STATE_INTERNAL_DRIVER+1) - -extern int hw_tcl_on; - #define COLOR_IS_RGBA #define TAG(x) r600##x #include "tnl_dd/t_dd_vertex.h" #undef TAG -#define PFS_MAX_ALU_INST 64 -#define PFS_MAX_TEX_INST 64 -#define PFS_MAX_TEX_INDIRECT 4 -#define PFS_NUM_TEMP_REGS 32 -#define PFS_NUM_CONST_REGS 16 - -#define R600_MAX_AOS_ARRAYS 16 - -#define REG_COORDS 0 -#define REG_COLOR0 1 -#define REG_TEX0 2 - #define R600_FALLBACK_NONE 0 #define R600_FALLBACK_TCL 1 #define R600_FALLBACK_RAST 2 -enum -{ - NO_SHIFT = 0, - LEFT_SHIFT = 1, - RIGHT_SHIFT = 2, -}; - struct r600_hw_state { struct radeon_state_atom sq; struct radeon_state_atom db; @@ -145,6 +103,32 @@ struct r600_hw_state { struct radeon_state_atom tx_brdr_clr; }; +typedef struct StreamDesc +{ + GLint size; //number of data element + GLenum type; //data element type + GLsizei stride; + + struct radeon_bo *bo; + GLint bo_offset; + + GLuint dwords; + GLuint dst_loc; + GLuint _signed; + GLboolean normalize; + GLboolean is_named_bo; + GLubyte element; +} StreamDesc; + +typedef struct r700_index_buffer +{ + struct radeon_bo *bo; + int bo_offset; + + GLboolean is_32bit; + GLuint count; +} r700_index_buffer; + /** * \brief R600 context structure. */ @@ -160,9 +144,9 @@ struct r600_context { /* Vertex buffers */ - GLvector4f dummy_attrib[_TNL_ATTRIB_MAX]; - GLvector4f *temp_attrib[_TNL_ATTRIB_MAX]; - + GLint nNumActiveAos; + StreamDesc stream_desc[VERT_ATTRIB_MAX]; + struct r700_index_buffer ind_buf; }; #define R700_CONTEXT(ctx) ((context_t *)(ctx->DriverCtx)) @@ -193,16 +177,13 @@ extern GLboolean r700SyncSurf(context_t *context, uint32_t write_domain, uint32_t sync_type); -extern void r700SetupStreams(GLcontext * ctx); extern void r700Start3D(context_t *context); extern void r600InitAtoms(context_t *context); +extern void r700InitDraw(GLcontext *ctx); #define RADEON_D_CAPTURE 0 #define RADEON_D_PLAYBACK 1 #define RADEON_D_PLAYBACK_RAW 2 #define RADEON_D_T 3 -#define r600PackFloat32 radeonPackFloat32 -#define r600PackFloat24 radeonPackFloat24 - #endif /* __R600_CONTEXT_H__ */ diff --git a/src/mesa/drivers/dri/r600/r600_reg_r6xx.h b/src/mesa/drivers/dri/r600/r600_reg_r6xx.h index f7702c46de5..74af7b4fed7 100644 --- a/src/mesa/drivers/dri/r600/r600_reg_r6xx.h +++ b/src/mesa/drivers/dri/r600/r600_reg_r6xx.h @@ -415,11 +415,11 @@ enum { ALPHA_TO_MASK_ENABLE = 1 << 0, ALPHA_TO_MASK_OFFSET0_mask = 0x03 << 8, ALPHA_TO_MASK_OFFSET0_shift = 8, - ALPHA_TO_MASK_OFFSET1_mask = 0x03 << 8, + ALPHA_TO_MASK_OFFSET1_mask = 0x03 << 10, ALPHA_TO_MASK_OFFSET1_shift = 10, - ALPHA_TO_MASK_OFFSET2_mask = 0x03 << 8, + ALPHA_TO_MASK_OFFSET2_mask = 0x03 << 12, ALPHA_TO_MASK_OFFSET2_shift = 12, - ALPHA_TO_MASK_OFFSET3_mask = 0x03 << 8, + ALPHA_TO_MASK_OFFSET3_mask = 0x03 << 14, ALPHA_TO_MASK_OFFSET3_shift = 14, // SQ_VTX_CONSTANT_WORD2_0 = 0x00038008, diff --git a/src/mesa/drivers/dri/r600/r600_reg_r7xx.h b/src/mesa/drivers/dri/r600/r600_reg_r7xx.h index e5c01c861aa..eb169bd885d 100644 --- a/src/mesa/drivers/dri/r600/r600_reg_r7xx.h +++ b/src/mesa/drivers/dri/r600/r600_reg_r7xx.h @@ -143,6 +143,8 @@ enum { // SQ_TEX_SAMPLER_MISC_0 = 0x0003d03c, R7xx_TRUNCATE_COORD_bit = 1 << 9, R7xx_DISABLE_CUBE_WRAP_bit = 1 << 10, +// DB_RENDER_CONTROL = 0x00028d0c, + PERFECT_ZPASS_COUNTS_bit = 1 << 15, } ; diff --git a/src/mesa/drivers/dri/r600/r600_tex.c b/src/mesa/drivers/dri/r600/r600_tex.c index d105b90cd18..9d83a64e220 100644 --- a/src/mesa/drivers/dri/r600/r600_tex.c +++ b/src/mesa/drivers/dri/r600/r600_tex.c @@ -40,7 +40,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/image.h" #include "main/mipmap.h" #include "main/simple_list.h" -#include "main/texformat.h" #include "main/texstore.h" #include "main/teximage.h" #include "main/texobj.h" @@ -286,6 +285,7 @@ static void r600TexParameter(GLcontext * ctx, GLenum target, GLenum pname, const GLfloat * params) { radeonTexObj* t = radeon_tex_obj(texObj); + GLenum baseFormat; radeon_print(RADEON_STATE | RADEON_TEXTURE, RADEON_VERBOSE, "%s( %s )\n", __FUNCTION__, @@ -312,23 +312,15 @@ static void r600TexParameter(GLcontext * ctx, GLenum target, case GL_TEXTURE_MAX_LEVEL: case GL_TEXTURE_MIN_LOD: case GL_TEXTURE_MAX_LOD: - /* This isn't the most efficient solution but there doesn't appear to - * be a nice alternative. Since there's no LOD clamping, - * we just have to rely on loading the right subset of mipmap levels - * to simulate a clamped LOD. - */ - if (t->mt) { - radeon_miptree_unreference(t->mt); - t->mt = 0; - t->validated = GL_FALSE; - } + t->validated = GL_FALSE; break; case GL_DEPTH_TEXTURE_MODE: if (!texObj->Image[0][texObj->BaseLevel]) return; - if (texObj->Image[0][texObj->BaseLevel]->TexFormat->BaseFormat - == GL_DEPTH_COMPONENT) { + baseFormat = texObj->Image[0][texObj->BaseLevel]->_BaseFormat; + if (baseFormat == GL_DEPTH_COMPONENT || + baseFormat == GL_DEPTH_STENCIL) { r600SetDepthTexMode(texObj); break; } else { @@ -368,10 +360,8 @@ static void r600DeleteTexture(GLcontext * ctx, struct gl_texture_object *texObj) t->bo = NULL; } - if (t->mt) { - radeon_miptree_unreference(t->mt); - t->mt = 0; - } + radeon_miptree_unreference(&t->mt); + _mesa_delete_texture_object(ctx, texObj); } diff --git a/src/mesa/drivers/dri/r600/r600_texstate.c b/src/mesa/drivers/dri/r600/r600_texstate.c index 7d7e77d355b..2a4a6e6ee14 100644 --- a/src/mesa/drivers/dri/r600/r600_texstate.c +++ b/src/mesa/drivers/dri/r600/r600_texstate.c @@ -39,7 +39,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/imports.h" #include "main/context.h" #include "main/macros.h" -#include "main/texformat.h" #include "main/teximage.h" #include "main/texobj.h" #include "main/enums.h" @@ -78,7 +77,7 @@ void r600UpdateTextureState(GLcontext * ctx) } } -static GLboolean r600GetTexFormat(struct gl_texture_object *tObj, GLuint mesa_format) +static GLboolean r600GetTexFormat(struct gl_texture_object *tObj, gl_format mesa_format) { radeonTexObj *t = radeon_tex_obj(tObj); @@ -87,9 +86,19 @@ static GLboolean r600GetTexFormat(struct gl_texture_object *tObj, GLuint mesa_fo CLEARfield(t->SQ_TEX_RESOURCE4, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask); CLEARfield(t->SQ_TEX_RESOURCE4, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask); + SETfield(t->SQ_TEX_RESOURCE4, SQ_FORMAT_COMP_UNSIGNED, + FORMAT_COMP_X_shift, FORMAT_COMP_X_mask); + SETfield(t->SQ_TEX_RESOURCE4, SQ_FORMAT_COMP_UNSIGNED, + FORMAT_COMP_Y_shift, FORMAT_COMP_Y_mask); + SETfield(t->SQ_TEX_RESOURCE4, SQ_FORMAT_COMP_UNSIGNED, + FORMAT_COMP_X_shift, FORMAT_COMP_Z_mask); + SETfield(t->SQ_TEX_RESOURCE4, SQ_FORMAT_COMP_UNSIGNED, + FORMAT_COMP_W_shift, FORMAT_COMP_W_mask); + switch (mesa_format) /* This is mesa format. */ { case MESA_FORMAT_RGBA8888: + case MESA_FORMAT_SIGNED_RGBA8888: SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask); @@ -101,8 +110,19 @@ static GLboolean r600GetTexFormat(struct gl_texture_object *tObj, GLuint mesa_fo SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask); SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask); + if (mesa_format == MESA_FORMAT_SIGNED_RGBA8888) { + SETfield(t->SQ_TEX_RESOURCE4, SQ_FORMAT_COMP_SIGNED, + FORMAT_COMP_X_shift, FORMAT_COMP_X_mask); + SETfield(t->SQ_TEX_RESOURCE4, SQ_FORMAT_COMP_SIGNED, + FORMAT_COMP_Y_shift, FORMAT_COMP_Y_mask); + SETfield(t->SQ_TEX_RESOURCE4, SQ_FORMAT_COMP_SIGNED, + FORMAT_COMP_Z_shift, FORMAT_COMP_Z_mask); + SETfield(t->SQ_TEX_RESOURCE4, SQ_FORMAT_COMP_SIGNED, + FORMAT_COMP_W_shift, FORMAT_COMP_W_mask); + } break; case MESA_FORMAT_RGBA8888_REV: + case MESA_FORMAT_SIGNED_RGBA8888_REV: SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask); @@ -114,6 +134,16 @@ static GLboolean r600GetTexFormat(struct gl_texture_object *tObj, GLuint mesa_fo SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask); SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_W, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask); + if (mesa_format == MESA_FORMAT_SIGNED_RGBA8888_REV) { + SETfield(t->SQ_TEX_RESOURCE4, SQ_FORMAT_COMP_SIGNED, + FORMAT_COMP_X_shift, FORMAT_COMP_X_mask); + SETfield(t->SQ_TEX_RESOURCE4, SQ_FORMAT_COMP_SIGNED, + FORMAT_COMP_Y_shift, FORMAT_COMP_Y_mask); + SETfield(t->SQ_TEX_RESOURCE4, SQ_FORMAT_COMP_SIGNED, + FORMAT_COMP_Z_shift, FORMAT_COMP_Z_mask); + SETfield(t->SQ_TEX_RESOURCE4, SQ_FORMAT_COMP_SIGNED, + FORMAT_COMP_W_shift, FORMAT_COMP_W_mask); + } break; case MESA_FORMAT_ARGB8888: SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8, @@ -480,13 +510,21 @@ static GLboolean r600GetTexFormat(struct gl_texture_object *tObj, GLuint mesa_fo SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask); break; case MESA_FORMAT_Z16: + case MESA_FORMAT_X8_Z24: + case MESA_FORMAT_S8_Z24: case MESA_FORMAT_Z24_S8: case MESA_FORMAT_Z32: + case MESA_FORMAT_S8: switch (mesa_format) { case MESA_FORMAT_Z16: SETfield(t->SQ_TEX_RESOURCE1, FMT_16, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask); break; + case MESA_FORMAT_X8_Z24: + case MESA_FORMAT_S8_Z24: + SETfield(t->SQ_TEX_RESOURCE1, FMT_8_24, + SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask); + break; case MESA_FORMAT_Z24_S8: SETfield(t->SQ_TEX_RESOURCE1, FMT_24_8, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask); @@ -495,6 +533,12 @@ static GLboolean r600GetTexFormat(struct gl_texture_object *tObj, GLuint mesa_fo SETfield(t->SQ_TEX_RESOURCE1, FMT_32, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask); break; + case MESA_FORMAT_S8: + SETfield(t->SQ_TEX_RESOURCE1, FMT_8, + SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask); + break; + default: + break; }; switch (tObj->DepthMode) { case GL_LUMINANCE: /* X, X, X, ONE */ @@ -591,7 +635,7 @@ void r600SetDepthTexMode(struct gl_texture_object *tObj) t = radeon_tex_obj(tObj); - r600GetTexFormat(tObj, tObj->Image[0][tObj->BaseLevel]->TexFormat->MesaFormat); + r600GetTexFormat(tObj, tObj->Image[0][tObj->BaseLevel]->TexFormat); } @@ -605,7 +649,6 @@ static void setup_hardware_state(context_t *rmesa, struct gl_texture_object *tex { radeonTexObj *t = radeon_tex_obj(texObj); const struct gl_texture_image *firstImage; - int firstlevel = t->mt ? t->mt->firstLevel : 0; GLuint uTexelPitch, row_align; if (rmesa->radeon.radeonScreen->driScreen->dri2.enabled && @@ -613,10 +656,10 @@ static void setup_hardware_state(context_t *rmesa, struct gl_texture_object *tex t->bo) return; - firstImage = t->base.Image[0][firstlevel]; + firstImage = t->base.Image[0][t->minLod]; if (!t->image_override) { - if (!r600GetTexFormat(texObj, firstImage->TexFormat->MesaFormat)) { + if (!r600GetTexFormat(texObj, firstImage->TexFormat)) { radeon_error("unexpected texture format in %s\n", __FUNCTION__); return; @@ -648,7 +691,8 @@ static void setup_hardware_state(context_t *rmesa, struct gl_texture_object *tex } row_align = rmesa->radeon.texture_row_align - 1; - uTexelPitch = ((firstImage->Width * t->mt->bpp + row_align) & ~row_align) / t->mt->bpp; + uTexelPitch = (_mesa_format_row_stride(firstImage->TexFormat, firstImage->Width) + row_align) & ~row_align; + uTexelPitch = uTexelPitch / _mesa_get_format_bytes(firstImage->TexFormat); uTexelPitch = (uTexelPitch + R700_TEXEL_PITCH_ALIGNMENT_MASK) & ~R700_TEXEL_PITCH_ALIGNMENT_MASK; @@ -662,10 +706,10 @@ static void setup_hardware_state(context_t *rmesa, struct gl_texture_object *tex SETfield(t->SQ_TEX_RESOURCE1, firstImage->Height - 1, TEX_HEIGHT_shift, TEX_HEIGHT_mask); - if ((t->mt->lastLevel - t->mt->firstLevel) > 0) { - t->SQ_TEX_RESOURCE3 = t->mt->levels[0].size / 256; - SETfield(t->SQ_TEX_RESOURCE4, t->mt->firstLevel, BASE_LEVEL_shift, BASE_LEVEL_mask); - SETfield(t->SQ_TEX_RESOURCE5, t->mt->lastLevel, LAST_LEVEL_shift, LAST_LEVEL_mask); + if ((t->maxLod - t->minLod) > 0) { + t->SQ_TEX_RESOURCE3 = t->mt->levels[t->minLod].size / 256; + SETfield(t->SQ_TEX_RESOURCE4, 0, BASE_LEVEL_shift, BASE_LEVEL_mask); + SETfield(t->SQ_TEX_RESOURCE5, t->maxLod - t->minLod, LAST_LEVEL_shift, LAST_LEVEL_mask); } } @@ -764,7 +808,8 @@ void r600SetTexOffset(__DRIcontext * pDRICtx, GLint texname, struct gl_texture_object *tObj = _mesa_lookup_texture(rmesa->radeon.glCtx, texname); radeonTexObjPtr t = radeon_tex_obj(tObj); - uint32_t pitch_val, size; + const struct gl_texture_image *firstImage; + uint32_t pitch_val, size, row_align; if (!tObj) return; @@ -774,7 +819,9 @@ void r600SetTexOffset(__DRIcontext * pDRICtx, GLint texname, if (!offset) return; - size = pitch;//h * w * (depth / 8); + firstImage = t->base.Image[0][t->minLod]; + row_align = rmesa->radeon.texture_row_align - 1; + size = ((_mesa_format_row_stride(firstImage->TexFormat, firstImage->Width) + row_align) & ~row_align) * firstImage->Height; if (t->bo) { radeon_bo_unref(t->bo); t->bo = NULL; @@ -870,18 +917,7 @@ void r600SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_fo return; } - radeon_update_renderbuffers(pDRICtx, dPriv); - /* back & depth buffer are useless free them right away */ - rb = (void*)rfb->base.Attachment[BUFFER_DEPTH].Renderbuffer; - if (rb && rb->bo) { - radeon_bo_unref(rb->bo); - rb->bo = NULL; - } - rb = (void*)rfb->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer; - if (rb && rb->bo) { - radeon_bo_unref(rb->bo); - rb->bo = NULL; - } + radeon_update_renderbuffers(pDRICtx, dPriv, GL_TRUE); rb = rfb->color_rb[0]; if (rb->bo == NULL) { /* Failed to BO for the buffer */ @@ -897,20 +933,14 @@ void r600SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_fo radeon_bo_unref(rImage->bo); rImage->bo = NULL; } - if (t->mt) { - radeon_miptree_unreference(t->mt); - t->mt = NULL; - } - if (rImage->mt) { - radeon_miptree_unreference(rImage->mt); - rImage->mt = NULL; - } + + radeon_miptree_unreference(&t->mt); + radeon_miptree_unreference(&rImage->mt); + _mesa_init_teximage_fields(radeon->glCtx, target, texImage, rb->base.Width, rb->base.Height, 1, 0, rb->cpp); texImage->RowStride = rb->pitch / rb->cpp; - texImage->TexFormat = radeonChooseTextureFormat(radeon->glCtx, - internalFormat, - type, format, 0); + rImage->bo = rb->bo; radeon_bo_ref(rImage->bo); t->bo = rb->bo; diff --git a/src/mesa/drivers/dri/r600/r700_assembler.c b/src/mesa/drivers/dri/r600/r700_assembler.c index efeccb25f1e..e84f5245253 100644 --- a/src/mesa/drivers/dri/r600/r700_assembler.c +++ b/src/mesa/drivers/dri/r600/r700_assembler.c @@ -32,12 +32,49 @@ #include "main/mtypes.h" #include "main/imports.h" +#include "shader/prog_parameter.h" #include "radeon_debug.h" #include "r600_context.h" #include "r700_assembler.h" +#define USE_CF_FOR_CONTINUE_BREAK 1 +#define USE_CF_FOR_POP_AFTER 1 + +struct prog_instruction noise1_insts[12] = { + {OPCODE_BGNSUB , {{13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {13, 0, 15, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_MOV , {{0, 0, 0, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {0, 0, 2, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_MOV , {{8, 0, 0, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {0, 0, 4, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_MOV , {{8, 0, 585, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {0, 0, 8, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_SGT , {{0, 0, 585, 0, 0, 0}, {8, 0, 1170, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {0, 1, 1, 0, 8, 1672, 0}, 1, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_IF , {{13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {13, 0, 15, 0, 7, 0, 0}, 0, 0, 0, 1, 0, 0, 0, 15, 0, 0, 0}, + {OPCODE_MOV , {{0, 0, 1755, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {0, 0, 1, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_RET , {{13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {13, 0, 15, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_ENDIF , {{13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {13, 0, 15, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_MOV , {{0, 0, 1170, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {0, 0, 1, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_RET , {{13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {13, 0, 15, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0}, + {OPCODE_ENDSUB , {{13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}, {13, 0, 1672, 0, 0, 0}}, {13, 0, 15, 0, 8, 1672, 0}, 0, 0, 0, 1, 0, 0, 0, -1, 0, 0, 0} +}; +float noise1_const[2][4] = { + {0.300000f, 0.900000f, 0.500000f, 0.300000f} +}; + +COMPILED_SUB noise1_presub = { + &(noise1_insts[0]), + 12, + 2, + 1, + 0, + &(noise1_const[0]), + SWIZZLE_X, + SWIZZLE_X, + SWIZZLE_X, + SWIZZLE_X, + {0,0,0}, + 0 +}; + BITS addrmode_PVSDST(PVSDST * pPVSDST) { return pPVSDST->addrmode0 | ((BITS)pPVSDST->addrmode1 << 1); @@ -213,7 +250,7 @@ GLboolean is_reduction_opcode(PVSDWORD* dest) { if (dest->dst.op3 == 0) { - if ( (dest->dst.opcode == SQ_OP2_INST_DOT4 || dest->dst.opcode == SQ_OP2_INST_DOT4_IEEE) ) + if ( (dest->dst.opcode == SQ_OP2_INST_DOT4 || dest->dst.opcode == SQ_OP2_INST_DOT4_IEEE || dest->dst.opcode == SQ_OP2_INST_CUBE) ) { return GL_TRUE; } @@ -327,21 +364,27 @@ GLuint GetSurfaceFormat(GLenum eType, GLuint nChannels, GLuint * pClient_size) return(format); } -unsigned int r700GetNumOperands(r700_AssemblerBase* pAsm) +unsigned int r700GetNumOperands(GLuint opcode, GLuint nIsOp3) { - if(pAsm->D.dst.op3) + if(nIsOp3 > 0) { return 3; } - switch (pAsm->D.dst.opcode) + switch (opcode) { - case SQ_OP2_INST_ADD: + case SQ_OP2_INST_ADD: + case SQ_OP2_INST_KILLE: + case SQ_OP2_INST_KILLGT: + case SQ_OP2_INST_KILLGE: + case SQ_OP2_INST_KILLNE: case SQ_OP2_INST_MUL: case SQ_OP2_INST_MAX: case SQ_OP2_INST_MIN: //case SQ_OP2_INST_MAX_DX10: //case SQ_OP2_INST_MIN_DX10: + case SQ_OP2_INST_SETE: + case SQ_OP2_INST_SETNE: case SQ_OP2_INST_SETGT: case SQ_OP2_INST_SETGE: case SQ_OP2_INST_PRED_SETE: @@ -350,12 +393,14 @@ unsigned int r700GetNumOperands(r700_AssemblerBase* pAsm) case SQ_OP2_INST_PRED_SETNE: case SQ_OP2_INST_DOT4: case SQ_OP2_INST_DOT4_IEEE: + case SQ_OP2_INST_CUBE: return 2; case SQ_OP2_INST_MOV: + case SQ_OP2_INST_MOVA_FLOOR: case SQ_OP2_INST_FRACT: case SQ_OP2_INST_FLOOR: - case SQ_OP2_INST_KILLGT: + case SQ_OP2_INST_TRUNC: case SQ_OP2_INST_EXP_IEEE: case SQ_OP2_INST_LOG_CLAMPED: case SQ_OP2_INST_LOG_IEEE: @@ -367,7 +412,7 @@ unsigned int r700GetNumOperands(r700_AssemblerBase* pAsm) return 1; default: radeon_error( - "Need instruction operand number for %x.\n", pAsm->D.dst.opcode); + "Need instruction operand number for %x.\n", opcode); }; return 3; @@ -381,95 +426,120 @@ int Init_r700_AssemblerBase(SHADER_PIPE_TYPE spt, r700_AssemblerBase* pAsm, R700 pAsm->pR700Shader = pShader; pAsm->currentShaderType = spt; - pAsm->cf_last_export_ptr = NULL; + pAsm->cf_last_export_ptr = NULL; - pAsm->cf_current_export_clause_ptr = NULL; - pAsm->cf_current_alu_clause_ptr = NULL; - pAsm->cf_current_tex_clause_ptr = NULL; - pAsm->cf_current_vtx_clause_ptr = NULL; - pAsm->cf_current_cf_clause_ptr = NULL; + pAsm->cf_current_export_clause_ptr = NULL; + pAsm->cf_current_alu_clause_ptr = NULL; + pAsm->cf_current_tex_clause_ptr = NULL; + pAsm->cf_current_vtx_clause_ptr = NULL; + pAsm->cf_current_cf_clause_ptr = NULL; - // No clause has been created yet - pAsm->cf_current_clause_type = CF_EMPTY_CLAUSE; + // No clause has been created yet + pAsm->cf_current_clause_type = CF_EMPTY_CLAUSE; - pAsm->number_of_colorandz_exports = 0; - pAsm->number_of_exports = 0; - pAsm->number_of_export_opcodes = 0; + pAsm->number_of_colorandz_exports = 0; + pAsm->number_of_exports = 0; + pAsm->number_of_export_opcodes = 0; + pAsm->alu_x_opcode = 0; - pAsm->D.bits = 0; - pAsm->S[0].bits = 0; - pAsm->S[1].bits = 0; - pAsm->S[2].bits = 0; + pAsm->D2.bits = 0; + + pAsm->D.bits = 0; + pAsm->S[0].bits = 0; + pAsm->S[1].bits = 0; + pAsm->S[2].bits = 0; - pAsm->uLastPosUpdate = 0; + pAsm->uLastPosUpdate = 0; - *(BITS *) &pAsm->fp_stOutFmt0 = 0; + *(BITS *) &pAsm->fp_stOutFmt0 = 0; - pAsm->uIIns = 0; - pAsm->uOIns = 0; - pAsm->number_used_registers = 0; - pAsm->uUsedConsts = 256; + pAsm->uIIns = 0; + pAsm->uOIns = 0; + pAsm->number_used_registers = 0; + pAsm->uUsedConsts = 256; - // Fragment programs - pAsm->uBoolConsts = 0; - pAsm->uIntConsts = 0; - pAsm->uInsts = 0; - pAsm->uConsts = 0; + // Fragment programs + pAsm->uBoolConsts = 0; + pAsm->uIntConsts = 0; + pAsm->uInsts = 0; + pAsm->uConsts = 0; - pAsm->FCSP = 0; - pAsm->fc_stack[0].type = FC_NONE; + pAsm->FCSP = 0; + pAsm->fc_stack[0].type = FC_NONE; - pAsm->branch_depth = 0; - pAsm->max_branch_depth = 0; + pAsm->aArgSubst[0] = + pAsm->aArgSubst[1] = + pAsm->aArgSubst[2] = + pAsm->aArgSubst[3] = (-1); - pAsm->aArgSubst[0] = - pAsm->aArgSubst[1] = - pAsm->aArgSubst[2] = - pAsm->aArgSubst[3] = (-1); + pAsm->uOutputs = 0; - pAsm->uOutputs = 0; + for (i=0; i<NUMBER_OF_OUTPUT_COLORS; i++) + { + pAsm->color_export_register_number[i] = (-1); + } - for (i=0; i<NUMBER_OF_OUTPUT_COLORS; i++) - { - pAsm->color_export_register_number[i] = (-1); - } + pAsm->depth_export_register_number = (-1); + pAsm->stencil_export_register_number = (-1); + pAsm->coverage_to_mask_export_register_number = (-1); + pAsm->mask_export_register_number = (-1); - pAsm->depth_export_register_number = (-1); - pAsm->stencil_export_register_number = (-1); - pAsm->coverage_to_mask_export_register_number = (-1); - pAsm->mask_export_register_number = (-1); + pAsm->starting_export_register_number = 0; + pAsm->starting_vfetch_register_number = 0; + pAsm->starting_temp_register_number = 0; + pAsm->uFirstHelpReg = 0; - pAsm->starting_export_register_number = 0; - pAsm->starting_vfetch_register_number = 0; - pAsm->starting_temp_register_number = 0; - pAsm->uFirstHelpReg = 0; + pAsm->input_position_is_used = GL_FALSE; + pAsm->input_normal_is_used = GL_FALSE; + for (i=0; i<NUMBER_OF_INPUT_COLORS; i++) + { + pAsm->input_color_is_used[ i ] = GL_FALSE; + } - pAsm->input_position_is_used = GL_FALSE; - pAsm->input_normal_is_used = GL_FALSE; + for (i=0; i<NUMBER_OF_TEXTURE_UNITS; i++) + { + pAsm->input_texture_unit_is_used[ i ] = GL_FALSE; + } + for (i=0; i<VERT_ATTRIB_MAX; i++) + { + pAsm->vfetch_instruction_ptr_array[ i ] = NULL; + } - for (i=0; i<NUMBER_OF_INPUT_COLORS; i++) - { - pAsm->input_color_is_used[ i ] = GL_FALSE; - } + pAsm->number_of_inputs = 0; - for (i=0; i<NUMBER_OF_TEXTURE_UNITS; i++) - { - pAsm->input_texture_unit_is_used[ i ] = GL_FALSE; - } + pAsm->is_tex = GL_FALSE; + pAsm->need_tex_barrier = GL_FALSE; - for (i=0; i<VERT_ATTRIB_MAX; i++) - { - pAsm->vfetch_instruction_ptr_array[ i ] = NULL; - } + pAsm->subs = NULL; + pAsm->unSubArraySize = 0; + pAsm->unSubArrayPointer = 0; + pAsm->callers = NULL; + pAsm->unCallerArraySize = 0; + pAsm->unCallerArrayPointer = 0; + + pAsm->CALLSP = 0; + pAsm->CALLSTACK[0].FCSP_BeforeEntry = 0; + pAsm->CALLSTACK[0].plstCFInstructions_local + = &(pAsm->pR700Shader->lstCFInstructions); + + pAsm->CALLSTACK[0].max = 0; + pAsm->CALLSTACK[0].current = 0; + + SetActiveCFlist(pAsm->pR700Shader, pAsm->CALLSTACK[0].plstCFInstructions_local); - pAsm->number_of_inputs = 0; + pAsm->unCFflags = 0; - return 0; + pAsm->presubs = NULL; + pAsm->unPresubArraySize = 0; + pAsm->unNumPresub = 0; + pAsm->unCurNumILInsts = 0; + + return 0; } GLboolean IsTex(gl_inst_opcode Opcode) @@ -587,6 +657,31 @@ int check_current_clause(r700_AssemblerBase* pAsm, return GL_TRUE; } +GLboolean add_cf_instruction(r700_AssemblerBase* pAsm) +{ + if(GL_FALSE == check_current_clause(pAsm, CF_OTHER_CLAUSE)) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr = + (R700ControlFlowGenericClause*) CALLOC_STRUCT(R700ControlFlowGenericClause); + + if (pAsm->cf_current_cf_clause_ptr != NULL) + { + Init_R700ControlFlowGenericClause(pAsm->cf_current_cf_clause_ptr); + AddCFInstruction( pAsm->pR700Shader, + (R700ControlFlowInstruction *)pAsm->cf_current_cf_clause_ptr ); + } + else + { + radeon_error("Could not allocate a new VFetch CF instruction.\n"); + return GL_FALSE; + } + + return GL_TRUE; +} + GLboolean add_vfetch_instruction(r700_AssemblerBase* pAsm, R700VertexInstruction* vertex_instruction_ptr) { @@ -682,7 +777,7 @@ GLboolean add_tex_instruction(r700_AssemblerBase* pAsm, // If this clause constains any TEX instruction that is dependent on a previous instruction, // set the barrier bit - if( pAsm->pInstDeps[pAsm->uiCurInst].nDstDep > (-1) ) + if( pAsm->pInstDeps[pAsm->uiCurInst].nDstDep > (-1) || pAsm->need_tex_barrier == GL_TRUE ) { pAsm->cf_current_tex_clause_ptr->m_Word1.f.barrier = 0x1; } @@ -786,6 +881,133 @@ GLboolean assemble_vfetch_instruction(r700_AssemblerBase* pAsm, return GL_TRUE; } +GLboolean assemble_vfetch_instruction2(r700_AssemblerBase* pAsm, + GLuint destination_register, + GLenum type, + GLint size, + GLubyte element, + GLuint _signed, + GLboolean normalize, + VTX_FETCH_METHOD * pFetchMethod) +{ + GLuint client_size_inbyte; + GLuint data_format; + GLuint mega_fetch_count; + GLuint is_mega_fetch_flag; + + R700VertexGenericFetch* vfetch_instruction_ptr; + R700VertexGenericFetch* assembled_vfetch_instruction_ptr + = pAsm->vfetch_instruction_ptr_array[element]; + + if (assembled_vfetch_instruction_ptr == NULL) + { + vfetch_instruction_ptr = (R700VertexGenericFetch*) CALLOC_STRUCT(R700VertexGenericFetch); + if (vfetch_instruction_ptr == NULL) + { + return GL_FALSE; + } + Init_R700VertexGenericFetch(vfetch_instruction_ptr); + } + else + { + vfetch_instruction_ptr = assembled_vfetch_instruction_ptr; + } + + data_format = GetSurfaceFormat(type, size, &client_size_inbyte); + + if(GL_TRUE == pFetchMethod->bEnableMini) //More conditions here + { + //TODO : mini fetch + } + else + { + mega_fetch_count = MEGA_FETCH_BYTES - 1; + is_mega_fetch_flag = 0x1; + pFetchMethod->mega_fetch_remainder = MEGA_FETCH_BYTES - client_size_inbyte; + } + + vfetch_instruction_ptr->m_Word0.f.vtx_inst = SQ_VTX_INST_FETCH; + vfetch_instruction_ptr->m_Word0.f.fetch_type = SQ_VTX_FETCH_VERTEX_DATA; + vfetch_instruction_ptr->m_Word0.f.fetch_whole_quad = 0x0; + + vfetch_instruction_ptr->m_Word0.f.buffer_id = element; + vfetch_instruction_ptr->m_Word0.f.src_gpr = 0x0; + vfetch_instruction_ptr->m_Word0.f.src_rel = SQ_ABSOLUTE; + vfetch_instruction_ptr->m_Word0.f.src_sel_x = SQ_SEL_X; + vfetch_instruction_ptr->m_Word0.f.mega_fetch_count = mega_fetch_count; + + vfetch_instruction_ptr->m_Word1.f.dst_sel_x = (size < 1) ? SQ_SEL_0 : SQ_SEL_X; + vfetch_instruction_ptr->m_Word1.f.dst_sel_y = (size < 2) ? SQ_SEL_0 : SQ_SEL_Y; + vfetch_instruction_ptr->m_Word1.f.dst_sel_z = (size < 3) ? SQ_SEL_0 : SQ_SEL_Z; + vfetch_instruction_ptr->m_Word1.f.dst_sel_w = (size < 4) ? SQ_SEL_1 : SQ_SEL_W; + + vfetch_instruction_ptr->m_Word1.f.use_const_fields = 1; + vfetch_instruction_ptr->m_Word1.f.data_format = data_format; + vfetch_instruction_ptr->m_Word2.f.endian_swap = SQ_ENDIAN_NONE; + + if(1 == _signed) + { + vfetch_instruction_ptr->m_Word1.f.format_comp_all = SQ_FORMAT_COMP_SIGNED; + } + else + { + vfetch_instruction_ptr->m_Word1.f.format_comp_all = SQ_FORMAT_COMP_UNSIGNED; + } + + if(GL_TRUE == normalize) + { + vfetch_instruction_ptr->m_Word1.f.num_format_all = SQ_NUM_FORMAT_NORM; + } + else + { + vfetch_instruction_ptr->m_Word1.f.num_format_all = SQ_NUM_FORMAT_INT; + } + + // Destination register + vfetch_instruction_ptr->m_Word1_GPR.f.dst_gpr = destination_register; + vfetch_instruction_ptr->m_Word1_GPR.f.dst_rel = SQ_ABSOLUTE; + + vfetch_instruction_ptr->m_Word2.f.offset = 0; + vfetch_instruction_ptr->m_Word2.f.const_buf_no_stride = 0x0; + + vfetch_instruction_ptr->m_Word2.f.mega_fetch = is_mega_fetch_flag; + + if (assembled_vfetch_instruction_ptr == NULL) + { + if ( GL_FALSE == add_vfetch_instruction(pAsm, (R700VertexInstruction *)vfetch_instruction_ptr) ) + { + return GL_FALSE; + } + + if (pAsm->vfetch_instruction_ptr_array[element] != NULL) + { + return GL_FALSE; + } + else + { + pAsm->vfetch_instruction_ptr_array[element] = vfetch_instruction_ptr; + } + } + + return GL_TRUE; +} + +GLboolean cleanup_vfetch_instructions(r700_AssemblerBase* pAsm) +{ + GLint i; + pAsm->cf_current_clause_type = CF_EMPTY_CLAUSE; + pAsm->cf_current_vtx_clause_ptr = NULL; + + for (i=0; i<VERT_ATTRIB_MAX; i++) + { + pAsm->vfetch_instruction_ptr_array[ i ] = NULL; + } + + cleanup_vfetch_shaderinst(pAsm->pR700Shader); + + return GL_TRUE; +} + GLuint gethelpr(r700_AssemblerBase* pAsm) { GLuint r = pAsm->uHelpReg; @@ -855,7 +1077,8 @@ GLboolean checkop2(r700_AssemblerBase* pAsm) checkop_init(pAsm); - if( (pILInst->SrcReg[0].File == PROGRAM_CONSTANT) || + if( (pILInst->SrcReg[0].File == PROGRAM_UNIFORM) || + (pILInst->SrcReg[0].File == PROGRAM_CONSTANT) || (pILInst->SrcReg[0].File == PROGRAM_LOCAL_PARAM) || (pILInst->SrcReg[0].File == PROGRAM_ENV_PARAM) || (pILInst->SrcReg[0].File == PROGRAM_STATE_VAR) ) @@ -866,7 +1089,8 @@ GLboolean checkop2(r700_AssemblerBase* pAsm) { bSrcConst[0] = GL_FALSE; } - if( (pILInst->SrcReg[1].File == PROGRAM_CONSTANT) || + if( (pILInst->SrcReg[1].File == PROGRAM_UNIFORM) || + (pILInst->SrcReg[1].File == PROGRAM_CONSTANT) || (pILInst->SrcReg[1].File == PROGRAM_LOCAL_PARAM) || (pILInst->SrcReg[1].File == PROGRAM_ENV_PARAM) || (pILInst->SrcReg[1].File == PROGRAM_STATE_VAR) ) @@ -899,7 +1123,8 @@ GLboolean checkop3(r700_AssemblerBase* pAsm) checkop_init(pAsm); - if( (pILInst->SrcReg[0].File == PROGRAM_CONSTANT) || + if( (pILInst->SrcReg[0].File == PROGRAM_UNIFORM) || + (pILInst->SrcReg[0].File == PROGRAM_CONSTANT) || (pILInst->SrcReg[0].File == PROGRAM_LOCAL_PARAM) || (pILInst->SrcReg[0].File == PROGRAM_ENV_PARAM) || (pILInst->SrcReg[0].File == PROGRAM_STATE_VAR) ) @@ -910,7 +1135,8 @@ GLboolean checkop3(r700_AssemblerBase* pAsm) { bSrcConst[0] = GL_FALSE; } - if( (pILInst->SrcReg[1].File == PROGRAM_CONSTANT) || + if( (pILInst->SrcReg[1].File == PROGRAM_UNIFORM) || + (pILInst->SrcReg[1].File == PROGRAM_CONSTANT) || (pILInst->SrcReg[1].File == PROGRAM_LOCAL_PARAM) || (pILInst->SrcReg[1].File == PROGRAM_ENV_PARAM) || (pILInst->SrcReg[1].File == PROGRAM_STATE_VAR) ) @@ -921,7 +1147,8 @@ GLboolean checkop3(r700_AssemblerBase* pAsm) { bSrcConst[1] = GL_FALSE; } - if( (pILInst->SrcReg[2].File == PROGRAM_CONSTANT) || + if( (pILInst->SrcReg[2].File == PROGRAM_UNIFORM) || + (pILInst->SrcReg[2].File == PROGRAM_CONSTANT) || (pILInst->SrcReg[2].File == PROGRAM_LOCAL_PARAM) || (pILInst->SrcReg[2].File == PROGRAM_ENV_PARAM) || (pILInst->SrcReg[2].File == PROGRAM_STATE_VAR) ) @@ -1021,6 +1248,7 @@ GLboolean assemble_src(r700_AssemblerBase *pAsm, case PROGRAM_LOCAL_PARAM: case PROGRAM_ENV_PARAM: case PROGRAM_STATE_VAR: + case PROGRAM_UNIFORM: if (1 == pILInst->SrcReg[src].RelAddr) { setaddrmode_PVSSRC(&(pAsm->S[fld].src), ADDR_RELATIVE_A0); @@ -1034,7 +1262,7 @@ GLboolean assemble_src(r700_AssemblerBase *pAsm, pAsm->S[fld].src.reg = pILInst->SrcReg[src].Index; break; case PROGRAM_INPUT: - setaddrmode_PVSSRC(&(pAsm->S[fld].src), ADDR_ABSOLUTE); + setaddrmode_PVSSRC(&(pAsm->S[fld].src), ADDR_ABSOLUTE); pAsm->S[fld].src.rtype = SRC_REG_INPUT; switch (pAsm->currentShaderType) { @@ -1047,7 +1275,7 @@ GLboolean assemble_src(r700_AssemblerBase *pAsm, } break; default: - radeon_error("Invalid source argument type\n"); + radeon_error("Invalid source argument type : %d \n", pILInst->SrcReg[src].File); return GL_FALSE; } } @@ -1103,6 +1331,15 @@ GLboolean assemble_dst(r700_AssemblerBase *pAsm) pAsm->D.dst.writez = (pILInst->DstReg.WriteMask >> 2) & 0x1; pAsm->D.dst.writew = (pILInst->DstReg.WriteMask >> 3) & 0x1; + if(pILInst->SaturateMode == SATURATE_ZERO_ONE) + { + pAsm->D2.dst2.SaturateMode = 1; + } + else + { + pAsm->D2.dst2.SaturateMode = 0; + } + return GL_TRUE; } @@ -1152,42 +1389,67 @@ GLboolean tex_src(r700_AssemblerBase *pAsm) GLboolean bValidTexCoord = GL_FALSE; + if(pAsm->aArgSubst[1] >= 0) + { + bValidTexCoord = GL_TRUE; + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[0].src.reg = pAsm->aArgSubst[1]; + } + else + { switch (pILInst->SrcReg[0].File) { - case PROGRAM_CONSTANT: - case PROGRAM_LOCAL_PARAM: - case PROGRAM_ENV_PARAM: - case PROGRAM_STATE_VAR: - bValidTexCoord = GL_TRUE; - setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); - pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; - pAsm->S[0].src.reg = pAsm->aArgSubst[1]; - break; - case PROGRAM_TEMPORARY: - bValidTexCoord = GL_TRUE; - pAsm->S[0].src.reg = pILInst->SrcReg[0].Index + - pAsm->starting_temp_register_number; - pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; - break; - case PROGRAM_INPUT: - switch (pILInst->SrcReg[0].Index) - { - case FRAG_ATTRIB_COL0: - case FRAG_ATTRIB_COL1: - case FRAG_ATTRIB_TEX0: - case FRAG_ATTRIB_TEX1: - case FRAG_ATTRIB_TEX2: - case FRAG_ATTRIB_TEX3: - case FRAG_ATTRIB_TEX4: - case FRAG_ATTRIB_TEX5: - case FRAG_ATTRIB_TEX6: - case FRAG_ATTRIB_TEX7: - bValidTexCoord = GL_TRUE; - pAsm->S[0].src.reg = - pAsm->uiFP_AttributeMap[pILInst->SrcReg[0].Index]; - pAsm->S[0].src.rtype = SRC_REG_INPUT; - break; - } - break; + case PROGRAM_UNIFORM: + case PROGRAM_CONSTANT: + case PROGRAM_LOCAL_PARAM: + case PROGRAM_ENV_PARAM: + case PROGRAM_STATE_VAR: + break; + case PROGRAM_TEMPORARY: + bValidTexCoord = GL_TRUE; + pAsm->S[0].src.reg = pILInst->SrcReg[0].Index + + pAsm->starting_temp_register_number; + pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; + break; + case PROGRAM_INPUT: + switch (pILInst->SrcReg[0].Index) + { + case FRAG_ATTRIB_WPOS: + case FRAG_ATTRIB_COL0: + case FRAG_ATTRIB_COL1: + case FRAG_ATTRIB_FOGC: + case FRAG_ATTRIB_TEX0: + case FRAG_ATTRIB_TEX1: + case FRAG_ATTRIB_TEX2: + case FRAG_ATTRIB_TEX3: + case FRAG_ATTRIB_TEX4: + case FRAG_ATTRIB_TEX5: + case FRAG_ATTRIB_TEX6: + case FRAG_ATTRIB_TEX7: + bValidTexCoord = GL_TRUE; + pAsm->S[0].src.reg = + pAsm->uiFP_AttributeMap[pILInst->SrcReg[0].Index]; + pAsm->S[0].src.rtype = SRC_REG_INPUT; + break; + case FRAG_ATTRIB_FACE: + fprintf(stderr, "FRAG_ATTRIB_FACE unsupported\n"); + break; + case FRAG_ATTRIB_PNTC: + fprintf(stderr, "FRAG_ATTRIB_PNTC unsupported\n"); + break; + } + + if( (pILInst->SrcReg[0].Index >= FRAG_ATTRIB_VAR0) || + (pILInst->SrcReg[0].Index < FRAG_ATTRIB_MAX) ) + { + bValidTexCoord = GL_TRUE; + pAsm->S[0].src.reg = + pAsm->uiFP_AttributeMap[pILInst->SrcReg[0].Index]; + pAsm->S[0].src.rtype = SRC_REG_INPUT; + } + + break; + } } if(GL_TRUE == bValidTexCoord) @@ -1368,6 +1630,10 @@ GLboolean assemble_alu_src(R700ALUInstruction* alu_instruction_ptr, { src_sel = pSource->reg + CFILE_REGISTER_OFFSET; } + else if (pSource->rtype == SRC_REC_LITERAL) + { + src_sel = SQ_ALU_SRC_LITERAL; + } else { radeon_error("Source (%d) register type (%d) not one of TEMP, INPUT, or CONSTANT.\n", @@ -1457,7 +1723,8 @@ GLboolean add_alu_instruction(r700_AssemblerBase* pAsm, return GL_FALSE; } - if ( pAsm->cf_current_alu_clause_ptr == NULL || + if ( pAsm->alu_x_opcode != 0 || + pAsm->cf_current_alu_clause_ptr == NULL || ( (pAsm->cf_current_alu_clause_ptr != NULL) && (pAsm->cf_current_alu_clause_ptr->m_Word1.f.count >= (GetCFMaxInstructions(pAsm->cf_current_alu_clause_ptr->m_ShaderInstType)-contiguous_slots_needed-1) ) ) ) @@ -1487,9 +1754,17 @@ GLboolean add_alu_instruction(r700_AssemblerBase* pAsm, pAsm->cf_current_alu_clause_ptr->m_Word1.f.kcache_addr0 = 0x0; pAsm->cf_current_alu_clause_ptr->m_Word1.f.kcache_addr1 = 0x0; - //cf_current_alu_clause_ptr->m_Word1.f.count = number_of_scalar_operations - 1; pAsm->cf_current_alu_clause_ptr->m_Word1.f.count = 0x0; - pAsm->cf_current_alu_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_ALU; + + if(pAsm->alu_x_opcode != 0) + { + pAsm->cf_current_alu_clause_ptr->m_Word1.f.cf_inst = pAsm->alu_x_opcode; + pAsm->alu_x_opcode = 0; + } + else + { + pAsm->cf_current_alu_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_ALU; + } pAsm->cf_current_alu_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; @@ -1497,7 +1772,7 @@ GLboolean add_alu_instruction(r700_AssemblerBase* pAsm, } else { - pAsm->cf_current_alu_clause_ptr->m_Word1.f.count++; + pAsm->cf_current_alu_clause_ptr->m_Word1.f.count += (GetInstructionSize(alu_instruction_ptr->m_ShaderInstType) / 2); } // If this clause constains any instruction that is forward dependent on a TEX instruction, @@ -1774,7 +2049,7 @@ GLboolean check_scalar(r700_AssemblerBase* pAsm, GLuint swizzle_key; - GLuint number_of_operands = r700GetNumOperands(pAsm); + GLuint number_of_operands = r700GetNumOperands(pAsm->D.dst.opcode, pAsm->D.dst.op3); for (src=0; src<number_of_operands; src++) { @@ -1863,7 +2138,7 @@ GLboolean check_vector(r700_AssemblerBase* pAsm, GLuint swizzle_key; - GLuint number_of_operands = r700GetNumOperands(pAsm); + GLuint number_of_operands = r700GetNumOperands(pAsm->D.dst.opcode, pAsm->D.dst.op3); for (src=0; src<number_of_operands; src++) { @@ -1896,7 +2171,7 @@ GLboolean check_vector(r700_AssemblerBase* pAsm, if( is_gpr(sel) ) { if( GL_FALSE == cycle_for_vector_bank_swizzle(bank_swizzle, src, &cycle) ) - { + { return GL_FALSE; } @@ -1908,7 +2183,7 @@ GLboolean check_vector(r700_AssemblerBase* pAsm, else { if( GL_FALSE == reserve_gpr(pAsm, sel, chan, cycle) ) - { + { return GL_FALSE; } } @@ -1920,7 +2195,7 @@ GLboolean check_vector(r700_AssemblerBase* pAsm, if( is_cfile(sel) ) { if( GL_FALSE == reserve_cfile(pAsm, sel, chan) ) - { + { return GL_FALSE; } } @@ -1932,6 +2207,10 @@ GLboolean check_vector(r700_AssemblerBase* pAsm, GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) { + R700ALUInstruction * alu_instruction_ptr; + R700ALUInstructionHalfLiteral * alu_instruction_ptr_hl; + R700ALUInstructionFullLiteral * alu_instruction_ptr_fl; + GLuint number_of_scalar_operations; GLboolean is_single_scalar_operation; GLuint scalar_channel_index; @@ -1940,10 +2219,10 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) int current_source_index; GLuint contiguous_slots_needed; - GLuint uNumSrc = r700GetNumOperands(pAsm); - GLuint channel_swizzle, j; - GLuint chan_counter[4] = {0, 0, 0, 0}; - PVSSRC * pSource[3]; + GLuint uNumSrc = r700GetNumOperands(pAsm->D.dst.opcode, pAsm->D.dst.op3); + //GLuint channel_swizzle, j; + //GLuint chan_counter[4] = {0, 0, 0, 0}; + //PVSSRC * pSource[3]; GLboolean bSplitInst = GL_FALSE; if (1 == pAsm->D.dst.math) @@ -1955,7 +2234,9 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) { is_single_scalar_operation = GL_FALSE; number_of_scalar_operations = 4; - + +/* current assembler doesn't do more than 1 register per source */ +#if 0 /* check read port, only very preliminary algorithm, not count in src0/1 same comp case and prev slot repeat case; also not count relative addressing. TODO: improve performance. */ @@ -1990,27 +2271,49 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) { bSplitInst = GL_TRUE; } +#endif } contiguous_slots_needed = 0; - if(GL_TRUE == is_reduction_opcode(&(pAsm->D)) ) + if(!is_single_scalar_operation) { contiguous_slots_needed = 4; } + contiguous_slots_needed += pAsm->D2.dst2.literal_slots; + initialize(pAsm); for (scalar_channel_index=0; scalar_channel_index < number_of_scalar_operations; scalar_channel_index++) { - R700ALUInstruction* alu_instruction_ptr = (R700ALUInstruction*) CALLOC_STRUCT(R700ALUInstruction); - if (alu_instruction_ptr == NULL) - { - return GL_FALSE; - } - Init_R700ALUInstruction(alu_instruction_ptr); + if(scalar_channel_index == (number_of_scalar_operations-1)) + { + switch(pAsm->D2.dst2.literal_slots) + { + case 0: + alu_instruction_ptr = (R700ALUInstruction*) CALLOC_STRUCT(R700ALUInstruction); + Init_R700ALUInstruction(alu_instruction_ptr); + break; + case 1: + alu_instruction_ptr_hl = (R700ALUInstructionHalfLiteral*) CALLOC_STRUCT(R700ALUInstructionHalfLiteral); + Init_R700ALUInstructionHalfLiteral(alu_instruction_ptr_hl, pAsm->C[0].f, pAsm->C[1].f); + alu_instruction_ptr = (R700ALUInstruction*)alu_instruction_ptr_hl; + break; + case 2: + alu_instruction_ptr_fl = (R700ALUInstructionFullLiteral*) CALLOC_STRUCT(R700ALUInstructionFullLiteral); + Init_R700ALUInstructionFullLiteral(alu_instruction_ptr_fl,pAsm->C[0].f, pAsm->C[1].f, pAsm->C[2].f, pAsm->C[3].f); + alu_instruction_ptr = (R700ALUInstruction*)alu_instruction_ptr_fl; + break; + }; + } + else + { + alu_instruction_ptr = (R700ALUInstruction*) CALLOC_STRUCT(R700ALUInstruction); + Init_R700ALUInstruction(alu_instruction_ptr); + } //src 0 current_source_index = 0; @@ -2020,11 +2323,11 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) current_source_index, pcurrent_source, scalar_channel_index) ) - { + { return GL_FALSE; } - if (pAsm->D.dst.math == 0) + if (uNumSrc > 1) { // Process source 1 current_source_index = 1; @@ -2034,13 +2337,13 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) current_source_index, pcurrent_source, scalar_channel_index) ) - { + { return GL_FALSE; } } //other bits - alu_instruction_ptr->m_Word0.f.index_mode = SQ_INDEX_LOOP; + alu_instruction_ptr->m_Word0.f.index_mode = pAsm->D2.dst2.index_mode; if( (is_single_scalar_operation == GL_TRUE) || (GL_TRUE == bSplitInst) ) @@ -2052,9 +2355,17 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) alu_instruction_ptr->m_Word0.f.last = (scalar_channel_index == 3) ? 1 : 0; } - alu_instruction_ptr->m_Word0.f.pred_sel = 0x0; - alu_instruction_ptr->m_Word1_OP2.f.update_pred = 0x0; - alu_instruction_ptr->m_Word1_OP2.f.update_execute_mask = 0x0; + alu_instruction_ptr->m_Word0.f.pred_sel = (pAsm->D.dst.pred_inv > 0) ? 1 : 0; + if(1 == pAsm->D.dst.predicated) + { + alu_instruction_ptr->m_Word1_OP2.f.update_pred = 0x1; + alu_instruction_ptr->m_Word1_OP2.f.update_execute_mask = 0x1; + } + else + { + alu_instruction_ptr->m_Word1_OP2.f.update_pred = 0x0; + alu_instruction_ptr->m_Word1_OP2.f.update_execute_mask = 0x0; + } // dst if( (pAsm->D.dst.rtype == DST_REG_TEMPORARY) || @@ -2063,7 +2374,7 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) alu_instruction_ptr->m_Word1.f.dst_gpr = pAsm->D.dst.reg; } else - { + { radeon_error("Only temp destination registers supported for ALU dest regs.\n"); return GL_FALSE; } @@ -2093,7 +2404,7 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) alu_instruction_ptr->m_Word1.f.dst_chan = scalar_channel_index; - alu_instruction_ptr->m_Word1.f.clamp = pAsm->pILInst[pAsm->uiCurInst].SaturateMode; + alu_instruction_ptr->m_Word1.f.clamp = pAsm->D2.dst2.SaturateMode; if (pAsm->D.dst.op3) { @@ -2120,8 +2431,8 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) { alu_instruction_ptr->m_Word1_OP2.f6.alu_inst = pAsm->D.dst.opcode; - alu_instruction_ptr->m_Word1_OP2.f6.src0_abs = 0x0; - alu_instruction_ptr->m_Word1_OP2.f6.src1_abs = 0x0; + alu_instruction_ptr->m_Word1_OP2.f6.src0_abs = pAsm->S[0].src.abs; + alu_instruction_ptr->m_Word1_OP2.f6.src1_abs = pAsm->S[1].src.abs; //alu_instruction_ptr->m_Word1_OP2.f6.update_execute_mask = 0x0; //alu_instruction_ptr->m_Word1_OP2.f6.update_pred = 0x0; @@ -2149,8 +2460,8 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) { alu_instruction_ptr->m_Word1_OP2.f.alu_inst = pAsm->D.dst.opcode; - alu_instruction_ptr->m_Word1_OP2.f.src0_abs = 0x0; - alu_instruction_ptr->m_Word1_OP2.f.src1_abs = 0x0; + alu_instruction_ptr->m_Word1_OP2.f.src0_abs = pAsm->S[0].src.abs; + alu_instruction_ptr->m_Word1_OP2.f.src1_abs = pAsm->S[1].src.abs; //alu_instruction_ptr->m_Word1_OP2.f.update_execute_mask = 0x0; //alu_instruction_ptr->m_Word1_OP2.f.update_pred = 0x0; @@ -2177,7 +2488,7 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) } if(GL_FALSE == add_alu_instruction(pAsm, alu_instruction_ptr, contiguous_slots_needed) ) - { + { return GL_FALSE; } @@ -2188,19 +2499,19 @@ GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm) if (is_single_scalar_operation) { if(GL_FALSE == check_scalar(pAsm, alu_instruction_ptr) ) - { + { return GL_FALSE; } } else { if(GL_FALSE == check_vector(pAsm, alu_instruction_ptr) ) - { - return 1; + { + return GL_FALSE; } } - contiguous_slots_needed = 0; + contiguous_slots_needed -= 1; } return GL_TRUE; @@ -2210,9 +2521,7 @@ GLboolean next_ins(r700_AssemblerBase *pAsm) { struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]); - if( GL_TRUE == IsTex(pILInst->Opcode) && - /* handle const moves to temp register */ - !(pAsm->D.dst.opcode == SQ_OP2_INST_MOV) ) + if( GL_TRUE == pAsm->is_tex ) { if (pILInst->TexSrcTarget == TEXTURE_RECT_INDEX) { if( GL_FALSE == assemble_tex_instruction(pAsm, GL_FALSE) ) @@ -2253,10 +2562,14 @@ GLboolean next_ins(r700_AssemblerBase *pAsm) //reset for next inst. pAsm->D.bits = 0; + pAsm->D2.bits = 0; pAsm->S[0].bits = 0; pAsm->S[1].bits = 0; pAsm->S[2].bits = 0; - + pAsm->is_tex = GL_FALSE; + pAsm->need_tex_barrier = GL_FALSE; + pAsm->D2.bits = 0; + pAsm->C[0].bits = pAsm->C[1].bits = pAsm->C[2].bits = pAsm->C[3].bits = 0; return GL_TRUE; } @@ -2375,6 +2688,35 @@ GLboolean assemble_ADD(r700_AssemblerBase *pAsm) return GL_TRUE; } +GLboolean assemble_ARL(r700_AssemblerBase *pAsm) +{ /* TODO: ar values dont' persist between clauses */ + if( GL_FALSE == checkop1(pAsm) ) + { + return GL_FALSE; + } + + pAsm->D.dst.opcode = SQ_OP2_INST_MOVA_FLOOR; + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = 0; + pAsm->D.dst.writex = 0; + pAsm->D.dst.writey = 0; + pAsm->D.dst.writez = 0; + pAsm->D.dst.writew = 0; + + if( GL_FALSE == assemble_src(pAsm, 0, -1) ) + { + return GL_FALSE; + } + + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } + + return GL_TRUE; +} + GLboolean assemble_BAD(char *opcode_str) { radeon_error("Not yet implemented instruction (%s)\n", opcode_str); @@ -2460,9 +2802,44 @@ GLboolean assemble_CMP(r700_AssemblerBase *pAsm) return GL_TRUE; } -GLboolean assemble_COS(r700_AssemblerBase *pAsm) +GLboolean assemble_TRIG(r700_AssemblerBase *pAsm, BITS opcode) { - return assemble_math_function(pAsm, SQ_OP2_INST_COS); + int tmp; + checkop1(pAsm); + + tmp = gethelpr(pAsm); + + pAsm->D.dst.opcode = SQ_OP2_INST_MUL; + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = tmp; + pAsm->D.dst.writex = 1; + + assemble_src(pAsm, 0, -1); + + pAsm->S[1].src.rtype = SRC_REC_LITERAL; + setswizzle_PVSSRC(&(pAsm->S[1].src), SQ_SEL_X); + pAsm->D2.dst2.literal_slots = 1; + pAsm->C[0].f = 1/(3.1415926535 * 2); + pAsm->C[1].f = 0.0F; + next_ins(pAsm); + + pAsm->D.dst.opcode = opcode; + pAsm->D.dst.math = 1; + + assemble_dst(pAsm); + + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[0].src.reg = tmp; + setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X); + noneg_PVSSRC(&(pAsm->S[0].src)); + + next_ins(pAsm); + + //TODO - replicate if more channels set in WriteMask + return GL_TRUE; + } GLboolean assemble_DOT(r700_AssemblerBase *pAsm) @@ -2496,7 +2873,7 @@ GLboolean assemble_DOT(r700_AssemblerBase *pAsm) } else if(pAsm->pILInst[pAsm->uiCurInst].Opcode == OPCODE_DPH) { - onecomp_PVSSRC(&(pAsm->S[1].src), 3); + onecomp_PVSSRC(&(pAsm->S[0].src), 3); } if ( GL_FALSE == next_ins(pAsm) ) @@ -2549,6 +2926,133 @@ GLboolean assemble_EX2(r700_AssemblerBase *pAsm) { return assemble_math_function(pAsm, SQ_OP2_INST_EXP_IEEE); } + +GLboolean assemble_EXP(r700_AssemblerBase *pAsm) +{ + BITS tmp; + + checkop1(pAsm); + + tmp = gethelpr(pAsm); + + // FLOOR tmp.x, a.x + // EX2 dst.x tmp.x + + if (pAsm->pILInst->DstReg.WriteMask & 0x1) { + pAsm->D.dst.opcode = SQ_OP2_INST_FLOOR; + + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = tmp; + pAsm->D.dst.writex = 1; + + if( GL_FALSE == assemble_src(pAsm, 0, -1) ) + { + return GL_FALSE; + } + + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } + + pAsm->D.dst.opcode = SQ_OP2_INST_EXP_IEEE; + pAsm->D.dst.math = 1; + + if( GL_FALSE == assemble_dst(pAsm) ) + { + return GL_FALSE; + } + + pAsm->D.dst.writey = pAsm->D.dst.writez = pAsm->D.dst.writew = 0; + + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = DST_REG_TEMPORARY; + pAsm->S[0].src.reg = tmp; + + setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X); + noneg_PVSSRC(&(pAsm->S[0].src)); + + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } + } + + // FRACT dst.y a.x + + if ((pAsm->pILInst->DstReg.WriteMask >> 1) & 0x1) { + pAsm->D.dst.opcode = SQ_OP2_INST_FRACT; + + if( GL_FALSE == assemble_dst(pAsm) ) + { + return GL_FALSE; + } + + if( GL_FALSE == assemble_src(pAsm, 0, -1) ) + { + return GL_FALSE; + } + + pAsm->D.dst.writex = pAsm->D.dst.writez = pAsm->D.dst.writew = 0; + + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } + } + + // EX2 dst.z, a.x + + if ((pAsm->pILInst->DstReg.WriteMask >> 2) & 0x1) { + pAsm->D.dst.opcode = SQ_OP2_INST_EXP_IEEE; + pAsm->D.dst.math = 1; + + if( GL_FALSE == assemble_dst(pAsm) ) + { + return GL_FALSE; + } + + if( GL_FALSE == assemble_src(pAsm, 0, -1) ) + { + return GL_FALSE; + } + + pAsm->D.dst.writex = pAsm->D.dst.writey = pAsm->D.dst.writew = 0; + + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } + } + + // MOV dst.w 1.0 + + if ((pAsm->pILInst->DstReg.WriteMask >> 3) & 0x1) { + pAsm->D.dst.opcode = SQ_OP2_INST_MOV; + + if( GL_FALSE == assemble_dst(pAsm) ) + { + return GL_FALSE; + } + + pAsm->D.dst.writex = pAsm->D.dst.writey = pAsm->D.dst.writez = 0; + + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[0].src.reg = tmp; + + setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_1); + noneg_PVSSRC(&(pAsm->S[0].src)); + + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } + } + + return GL_TRUE; +} GLboolean assemble_FLR(r700_AssemblerBase *pAsm) { @@ -2603,17 +3107,19 @@ GLboolean assemble_FRC(r700_AssemblerBase *pAsm) return GL_TRUE; } -GLboolean assemble_KIL(r700_AssemblerBase *pAsm) -{ - checkop1(pAsm); +GLboolean assemble_KIL(r700_AssemblerBase *pAsm, GLuint opcode) +{ + struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]); - pAsm->D.dst.opcode = SQ_OP2_INST_KILLGT; - - if ( GL_FALSE == assemble_dst(pAsm) ) - { - return GL_FALSE; - } + if(pILInst->Opcode == OPCODE_KIL) + checkop1(pAsm); + + pAsm->D.dst.opcode = opcode; + //pAsm->D.dst.math = 1; + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = 0; pAsm->D.dst.writex = 0; pAsm->D.dst.writey = 0; pAsm->D.dst.writez = 0; @@ -2622,30 +3128,34 @@ GLboolean assemble_KIL(r700_AssemblerBase *pAsm) setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; pAsm->S[0].src.reg = 0; - setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_0); noneg_PVSSRC(&(pAsm->S[0].src)); - pAsm->S[1].src.rtype = SRC_REG_TEMPORARY; - - if(PROGRAM_TEMPORARY == pAsm->pILInst[pAsm->uiCurInst].DstReg.File) + if(pILInst->Opcode == OPCODE_KIL_NV) { - pAsm->S[1].src.reg = pAsm->pILInst[pAsm->uiCurInst].DstReg.Index + pAsm->starting_temp_register_number; + setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE); + pAsm->S[1].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[1].src.reg = 0; + setswizzle_PVSSRC(&(pAsm->S[1].src), SQ_SEL_1); + neg_PVSSRC(&(pAsm->S[1].src)); } else - { //PROGRAM_OUTPUT - pAsm->S[1].src.reg = pAsm->uiFP_OutputMap[pAsm->pILInst[pAsm->uiCurInst].DstReg.Index]; + { + if( GL_FALSE == assemble_src(pAsm, 0, 1) ) + { + return GL_FALSE; + } + } - - setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE); - noswizzle_PVSSRC(&(pAsm->S[1].src)); - + if ( GL_FALSE == next_ins(pAsm) ) { return GL_FALSE; } + /* Doc says KILL has to be last(end) ALU clause */ pAsm->pR700Shader->killIsUsed = GL_TRUE; + pAsm->alu_x_opcode = SQ_CF_INST_ALU; return GL_TRUE; } @@ -2709,6 +3219,7 @@ GLboolean assemble_LRP(r700_AssemblerBase *pAsm) { return GL_FALSE; } + if( GL_FALSE == assemble_src(pAsm, 2, -1) ) { return GL_FALSE; @@ -2739,6 +3250,217 @@ GLboolean assemble_LRP(r700_AssemblerBase *pAsm) return GL_TRUE; } +GLboolean assemble_LOG(r700_AssemblerBase *pAsm) +{ + BITS tmp1, tmp2, tmp3; + + checkop1(pAsm); + + tmp1 = gethelpr(pAsm); + tmp2 = gethelpr(pAsm); + tmp3 = gethelpr(pAsm); + + // FIXME: The hardware can do fabs() directly on input + // elements, but the compiler doesn't have the + // capability to use that. + + // MAX tmp1.x, a.x, -a.x (fabs(a.x)) + + pAsm->D.dst.opcode = SQ_OP2_INST_MAX; + + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = tmp1; + pAsm->D.dst.writex = 1; + + if( GL_FALSE == assemble_src(pAsm, 0, -1) ) + { + return GL_FALSE; + } + + pAsm->S[1].bits = pAsm->S[0].bits; + flipneg_PVSSRC(&(pAsm->S[1].src)); + + if ( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } + + // Entire algo: + // + // LG2 tmp2.x, tmp1.x + // FLOOR tmp3.x, tmp2.x + // MOV dst.x, tmp3.x + // ADD tmp3.x, tmp2.x, -tmp3.x + // EX2 dst.y, tmp3.x + // MOV dst.z, tmp2.x + // MOV dst.w, 1.0 + + // LG2 tmp2.x, tmp1.x + // FLOOR tmp3.x, tmp2.x + + pAsm->D.dst.opcode = SQ_OP2_INST_LOG_IEEE; + pAsm->D.dst.math = 1; + + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = tmp2; + pAsm->D.dst.writex = 1; + + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = DST_REG_TEMPORARY; + pAsm->S[0].src.reg = tmp1; + + setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X); + noneg_PVSSRC(&(pAsm->S[0].src)); + + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } + + pAsm->D.dst.opcode = SQ_OP2_INST_FLOOR; + + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = tmp3; + pAsm->D.dst.writex = 1; + + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = DST_REG_TEMPORARY; + pAsm->S[0].src.reg = tmp2; + + setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X); + noneg_PVSSRC(&(pAsm->S[0].src)); + + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } + + // MOV dst.x, tmp3.x + + pAsm->D.dst.opcode = SQ_OP2_INST_MOV; + + if( GL_FALSE == assemble_dst(pAsm) ) + { + return GL_FALSE; + } + + pAsm->D.dst.writey = pAsm->D.dst.writez = pAsm->D.dst.writew = 0; + + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = DST_REG_TEMPORARY; + pAsm->S[0].src.reg = tmp3; + + setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X); + noneg_PVSSRC(&(pAsm->S[0].src)); + + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } + + // ADD tmp3.x, tmp2.x, -tmp3.x + // EX2 dst.y, tmp3.x + + pAsm->D.dst.opcode = SQ_OP2_INST_ADD; + + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = tmp3; + pAsm->D.dst.writex = 1; + + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = DST_REG_TEMPORARY; + pAsm->S[0].src.reg = tmp2; + + setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X); + noneg_PVSSRC(&(pAsm->S[0].src)); + + setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE); + pAsm->S[1].src.rtype = DST_REG_TEMPORARY; + pAsm->S[1].src.reg = tmp3; + + setswizzle_PVSSRC(&(pAsm->S[1].src), SQ_SEL_X); + neg_PVSSRC(&(pAsm->S[1].src)); + + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } + + pAsm->D.dst.opcode = SQ_OP2_INST_EXP_IEEE; + pAsm->D.dst.math = 1; + + if( GL_FALSE == assemble_dst(pAsm) ) + { + return GL_FALSE; + } + + pAsm->D.dst.writex = pAsm->D.dst.writez = pAsm->D.dst.writew = 0; + + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = DST_REG_TEMPORARY; + pAsm->S[0].src.reg = tmp3; + + setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X); + noneg_PVSSRC(&(pAsm->S[0].src)); + + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } + + // MOV dst.z, tmp2.x + + pAsm->D.dst.opcode = SQ_OP2_INST_MOV; + + if( GL_FALSE == assemble_dst(pAsm) ) + { + return GL_FALSE; + } + + pAsm->D.dst.writex = pAsm->D.dst.writey = pAsm->D.dst.writew = 0; + + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = DST_REG_TEMPORARY; + pAsm->S[0].src.reg = tmp2; + + setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X); + noneg_PVSSRC(&(pAsm->S[0].src)); + + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } + + // MOV dst.w 1.0 + + pAsm->D.dst.opcode = SQ_OP2_INST_MOV; + + if( GL_FALSE == assemble_dst(pAsm) ) + { + return GL_FALSE; + } + + pAsm->D.dst.writex = pAsm->D.dst.writey = pAsm->D.dst.writez = 0; + + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[0].src.reg = tmp1; + + setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_1); + noneg_PVSSRC(&(pAsm->S[0].src)); + + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } + + return GL_TRUE; +} + GLboolean assemble_MAD(struct r700_AssemblerBase *pAsm) { int tmp, ii; @@ -2880,6 +3602,11 @@ GLboolean assemble_LIT(r700_AssemblerBase *pAsm) return GL_FALSE; } + if( GL_FALSE == assemble_src(pAsm, 0, -1) ) + { + return GL_FALSE; + } + /* dst.y = max(src.x, 0.0) */ pAsm->D.dst.opcode = SQ_OP2_INST_MAX; pAsm->D.dst.rtype = dstType; @@ -2891,11 +3618,7 @@ GLboolean assemble_LIT(r700_AssemblerBase *pAsm) pAsm->S[0].src.rtype = srcType; pAsm->S[0].src.reg = srcReg; setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); - noneg_PVSSRC(&(pAsm->S[0].src)); - pAsm->S[0].src.swizzlex = SQ_SEL_X; - pAsm->S[0].src.swizzley = SQ_SEL_X; - pAsm->S[0].src.swizzlez = SQ_SEL_X; - pAsm->S[0].src.swizzlew = SQ_SEL_X; + swizzleagain_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X, SQ_SEL_X, SQ_SEL_X, SQ_SEL_X); pAsm->S[1].src.rtype = SRC_REG_TEMPORARY; pAsm->S[1].src.reg = tmp; setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE); @@ -2909,34 +3632,47 @@ GLboolean assemble_LIT(r700_AssemblerBase *pAsm) return GL_FALSE; } - /* before: dst.w = log(src.y) - * after : dst.x = log(src.y) - * why change dest register is that dst.w has been initialized as 1 before - */ + if( GL_FALSE == assemble_src(pAsm, 0, -1) ) + { + return GL_FALSE; + } + + swizzleagain_PVSSRC(&(pAsm->S[0].src), SQ_SEL_Y, SQ_SEL_Y, SQ_SEL_Y, SQ_SEL_Y); + + /* dst.z = log(src.y) */ pAsm->D.dst.opcode = SQ_OP2_INST_LOG_CLAMPED; pAsm->D.dst.math = 1; pAsm->D.dst.rtype = dstType; pAsm->D.dst.reg = dstReg; - pAsm->D.dst.writex = 1; + pAsm->D.dst.writex = 0; pAsm->D.dst.writey = 0; - pAsm->D.dst.writez = 0; + pAsm->D.dst.writez = 1; pAsm->D.dst.writew = 0; pAsm->S[0].src.rtype = srcType; pAsm->S[0].src.reg = srcReg; setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); - noneg_PVSSRC(&(pAsm->S[0].src)); - pAsm->S[0].src.swizzlex = SQ_SEL_Y; - pAsm->S[0].src.swizzley = SQ_SEL_Y; - pAsm->S[0].src.swizzlez = SQ_SEL_Y; - pAsm->S[0].src.swizzlew = SQ_SEL_Y; if( GL_FALSE == next_ins(pAsm) ) { return GL_FALSE; } - /* before: tmp.x = amd MUL_LIT(src.w, dst.w, src.x ) */ - /* after : tmp.x = amd MUL_LIT(src.w, dst.x, src.x ) */ + if( GL_FALSE == assemble_src(pAsm, 0, -1) ) + { + return GL_FALSE; + } + + if( GL_FALSE == assemble_src(pAsm, 0, 2) ) + { + return GL_FALSE; + } + + swizzleagain_PVSSRC(&(pAsm->S[0].src), SQ_SEL_W, SQ_SEL_W, SQ_SEL_W, SQ_SEL_W); + + swizzleagain_PVSSRC(&(pAsm->S[2].src), SQ_SEL_X, SQ_SEL_X, SQ_SEL_X, SQ_SEL_X); + + /* tmp.x = amd MUL_LIT(src.w, dst.z, src.x ) */ pAsm->D.dst.opcode = SQ_OP3_INST_MUL_LIT; + pAsm->D.dst.math = 1; pAsm->D.dst.op3 = 1; pAsm->D.dst.rtype = DST_REG_TEMPORARY; pAsm->D.dst.reg = tmp; @@ -2948,29 +3684,19 @@ GLboolean assemble_LIT(r700_AssemblerBase *pAsm) pAsm->S[0].src.rtype = srcType; pAsm->S[0].src.reg = srcReg; setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); - noneg_PVSSRC(&(pAsm->S[0].src)); - pAsm->S[0].src.swizzlex = SQ_SEL_W; - pAsm->S[0].src.swizzley = SQ_SEL_W; - pAsm->S[0].src.swizzlez = SQ_SEL_W; - pAsm->S[0].src.swizzlew = SQ_SEL_W; pAsm->S[1].src.rtype = SRC_REG_TEMPORARY; pAsm->S[1].src.reg = dstReg; setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE); noneg_PVSSRC(&(pAsm->S[1].src)); - pAsm->S[1].src.swizzlex = SQ_SEL_X; - pAsm->S[1].src.swizzley = SQ_SEL_X; - pAsm->S[1].src.swizzlez = SQ_SEL_X; - pAsm->S[1].src.swizzlew = SQ_SEL_X; + pAsm->S[1].src.swizzlex = SQ_SEL_Z; + pAsm->S[1].src.swizzley = SQ_SEL_Z; + pAsm->S[1].src.swizzlez = SQ_SEL_Z; + pAsm->S[1].src.swizzlew = SQ_SEL_Z; pAsm->S[2].src.rtype = srcType; pAsm->S[2].src.reg = srcReg; setaddrmode_PVSSRC(&(pAsm->S[2].src), ADDR_ABSOLUTE); - noneg_PVSSRC(&(pAsm->S[2].src)); - pAsm->S[2].src.swizzlex = SQ_SEL_X; - pAsm->S[2].src.swizzley = SQ_SEL_X; - pAsm->S[2].src.swizzlez = SQ_SEL_X; - pAsm->S[2].src.swizzlew = SQ_SEL_X; if( GL_FALSE == next_ins(pAsm) ) { @@ -3229,77 +3955,137 @@ GLboolean assemble_RSQ(r700_AssemblerBase *pAsm) return assemble_math_function(pAsm, SQ_OP2_INST_RECIPSQRT_IEEE); } -GLboolean assemble_SIN(r700_AssemblerBase *pAsm) -{ - return assemble_math_function(pAsm, SQ_OP2_INST_SIN); -} - GLboolean assemble_SCS(r700_AssemblerBase *pAsm) { BITS tmp; - checkop1(pAsm); + checkop1(pAsm); - tmp = gethelpr(pAsm); + tmp = gethelpr(pAsm); + /* tmp.x = src /2*PI */ + pAsm->D.dst.opcode = SQ_OP2_INST_MUL; + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = tmp; + pAsm->D.dst.writex = 1; - // COS tmp.x, a.x - pAsm->D.dst.opcode = SQ_OP2_INST_COS; - pAsm->D.dst.math = 1; + assemble_src(pAsm, 0, -1); - setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); - pAsm->D.dst.rtype = DST_REG_TEMPORARY; - pAsm->D.dst.reg = tmp; - pAsm->D.dst.writex = 1; + pAsm->S[1].src.rtype = SRC_REC_LITERAL; + setswizzle_PVSSRC(&(pAsm->S[1].src), SQ_SEL_X); + pAsm->D2.dst2.literal_slots = 1; + pAsm->C[0].f = 1/(3.1415926535 * 2); + pAsm->C[1].f = 0.0F; - if( GL_FALSE == assemble_src(pAsm, 0, -1) ) - { - return GL_FALSE; - } + next_ins(pAsm); - if ( GL_FALSE == next_ins(pAsm) ) - { - return GL_FALSE; - } + // COS dst.x, a.x + pAsm->D.dst.opcode = SQ_OP2_INST_COS; + pAsm->D.dst.math = 1; - // SIN tmp.y, a.x - pAsm->D.dst.opcode = SQ_OP2_INST_SIN; - pAsm->D.dst.math = 1; + assemble_dst(pAsm); + /* mask y */ + pAsm->D.dst.writey = 0; - setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); - pAsm->D.dst.rtype = DST_REG_TEMPORARY; - pAsm->D.dst.reg = tmp; - pAsm->D.dst.writey = 1; + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[0].src.reg = tmp; + setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X); + noneg_PVSSRC(&(pAsm->S[0].src)); - if( GL_FALSE == assemble_src(pAsm, 0, -1) ) - { - return GL_FALSE; - } + if ( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } - if( GL_FALSE == next_ins(pAsm) ) - { - return GL_FALSE; - } + // SIN dst.y, a.x + pAsm->D.dst.opcode = SQ_OP2_INST_SIN; + pAsm->D.dst.math = 1; - // MOV dst.mask, tmp - pAsm->D.dst.opcode = SQ_OP2_INST_MOV; + assemble_dst(pAsm); + /* mask x */ + pAsm->D.dst.writex = 0; - if( GL_FALSE == assemble_dst(pAsm) ) - { - return GL_FALSE; - } + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[0].src.reg = tmp; + setswizzle_PVSSRC(&(pAsm->S[0].src), SQ_SEL_X); + noneg_PVSSRC(&(pAsm->S[0].src)); - setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); - pAsm->S[0].src.rtype = DST_REG_TEMPORARY; - pAsm->S[0].src.reg = tmp; + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } - noswizzle_PVSSRC(&(pAsm->S[0].src)); - pAsm->S[0].src.swizzlez = SQ_SEL_0; - pAsm->S[0].src.swizzlew = SQ_SEL_0; + return GL_TRUE; +} - if ( GL_FALSE == next_ins(pAsm) ) - { - return GL_FALSE; - } +GLboolean assemble_LOGIC(r700_AssemblerBase *pAsm, BITS opcode) +{ + if( GL_FALSE == checkop2(pAsm) ) + { + return GL_FALSE; + } + + pAsm->D.dst.opcode = opcode; + //pAsm->D.dst.math = 1; + + if( GL_FALSE == assemble_dst(pAsm) ) + { + return GL_FALSE; + } + + if( GL_FALSE == assemble_src(pAsm, 0, -1) ) + { + return GL_FALSE; + } + + if( GL_FALSE == assemble_src(pAsm, 1, -1) ) + { + return GL_FALSE; + } + + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } + + return GL_TRUE; +} + +GLboolean assemble_LOGIC_PRED(r700_AssemblerBase *pAsm, BITS opcode) +{ + struct prog_instruction *pILInst = &(pAsm->pILInst[pAsm->uiCurInst]); + + pAsm->D.dst.opcode = opcode; + pAsm->D.dst.math = 1; + pAsm->D.dst.predicated = 1; + + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = pAsm->uHelpReg; + pAsm->D.dst.writex = 1; + pAsm->D.dst.writey = pAsm->D.dst.writez = pAsm->D.dst.writew = 0; + + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[0].src.reg = pAsm->last_cond_register + pAsm->starting_temp_register_number; + pAsm->S[0].src.swizzlex = pILInst->DstReg.CondSwizzle & 0x7; + noneg_PVSSRC(&(pAsm->S[0].src)); + + pAsm->S[1].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[1].src.reg = pAsm->uHelpReg; + setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE); + noneg_PVSSRC(&(pAsm->S[1].src)); + pAsm->S[1].src.swizzlex = SQ_SEL_0; + pAsm->S[1].src.swizzley = SQ_SEL_0; + pAsm->S[1].src.swizzlez = SQ_SEL_0; + pAsm->S[1].src.swizzlew = SQ_SEL_0; + + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } return GL_TRUE; } @@ -3376,9 +4162,13 @@ GLboolean assemble_STP(r700_AssemblerBase *pAsm) GLboolean assemble_TEX(r700_AssemblerBase *pAsm) { GLboolean src_const; + GLboolean need_barrier = GL_FALSE; + checkop1(pAsm); + switch (pAsm->pILInst[pAsm->uiCurInst].SrcReg[0].File) { + case PROGRAM_UNIFORM: case PROGRAM_CONSTANT: case PROGRAM_LOCAL_PARAM: case PROGRAM_ENV_PARAM: @@ -3396,29 +4186,170 @@ GLboolean assemble_TEX(r700_AssemblerBase *pAsm) { if ( GL_FALSE == mov_temp(pAsm, 0) ) return GL_FALSE; + need_barrier = GL_TRUE; } - switch (pAsm->pILInst[pAsm->uiCurInst].Opcode) + if (pAsm->pILInst[pAsm->uiCurInst].Opcode == OPCODE_TXP) { - case OPCODE_TEX: - pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE; - break; - case OPCODE_TXB: - radeon_error("do not support TXB yet\n"); + GLuint tmp = gethelpr(pAsm); + pAsm->D.dst.opcode = SQ_OP2_INST_RECIP_IEEE; + pAsm->D.dst.math = 1; + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = tmp; + pAsm->D.dst.writew = 1; + + if( GL_FALSE == assemble_src(pAsm, 0, -1) ) + { return GL_FALSE; - break; - case OPCODE_TXP: - /* TODO : tex proj version : divid first 3 components by 4th */ - pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE; - break; - default: - radeon_error("Internal error: bad texture op (not TEX)\n"); + } + swizzleagain_PVSSRC(&(pAsm->S[0].src), SQ_SEL_W, SQ_SEL_W, SQ_SEL_W, SQ_SEL_W); + if( GL_FALSE == next_ins(pAsm) ) + { return GL_FALSE; - break; + } + + pAsm->D.dst.opcode = SQ_OP2_INST_MUL; + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = tmp; + pAsm->D.dst.writex = 1; + pAsm->D.dst.writey = 1; + pAsm->D.dst.writez = 1; + pAsm->D.dst.writew = 0; + + if( GL_FALSE == assemble_src(pAsm, 0, -1) ) + { + return GL_FALSE; + } + setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE); + pAsm->S[1].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[1].src.reg = tmp; + setswizzle_PVSSRC(&(pAsm->S[1].src), SQ_SEL_W); + + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } + + pAsm->aArgSubst[1] = tmp; + need_barrier = GL_TRUE; } + if (pAsm->pILInst[pAsm->uiCurInst].TexSrcTarget == TEXTURE_CUBE_INDEX ) + { + GLuint tmp1 = gethelpr(pAsm); + GLuint tmp2 = gethelpr(pAsm); + + /* tmp1.xyzw = CUBE(R0.zzxy, R0.yxzz) */ + pAsm->D.dst.opcode = SQ_OP2_INST_CUBE; + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = tmp1; + nomask_PVSDST(&(pAsm->D.dst)); + + if( GL_FALSE == assemble_src(pAsm, 0, -1) ) + { + return GL_FALSE; + } + + if( GL_FALSE == assemble_src(pAsm, 0, 1) ) + { + return GL_FALSE; + } + + swizzleagain_PVSSRC(&(pAsm->S[0].src), SQ_SEL_Z, SQ_SEL_Z, SQ_SEL_X, SQ_SEL_Y); + swizzleagain_PVSSRC(&(pAsm->S[1].src), SQ_SEL_Y, SQ_SEL_X, SQ_SEL_Z, SQ_SEL_Z); + + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } + + /* tmp1.z = RCP_e(|tmp1.z|) */ + pAsm->D.dst.opcode = SQ_OP2_INST_RECIP_IEEE; + pAsm->D.dst.math = 1; + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = tmp1; + pAsm->D.dst.writez = 1; + + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[0].src.reg = tmp1; + pAsm->S[0].src.swizzlex = SQ_SEL_Z; + pAsm->S[0].src.abs = 1; + + next_ins(pAsm); + + /* MULADD R0.x, R0.x, PS1, (0x3FC00000, 1.5f).x + * MULADD R0.y, R0.y, PS1, (0x3FC00000, 1.5f).x + * muladd has no writemask, have to use another temp + */ + pAsm->D.dst.opcode = SQ_OP3_INST_MULADD; + pAsm->D.dst.op3 = 1; + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = tmp2; + + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[0].src.reg = tmp1; + noswizzle_PVSSRC(&(pAsm->S[0].src)); + setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE); + pAsm->S[1].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[1].src.reg = tmp1; + setswizzle_PVSSRC(&(pAsm->S[1].src), SQ_SEL_Z); + setaddrmode_PVSSRC(&(pAsm->S[2].src), ADDR_ABSOLUTE); + /* immediate c 1.5 */ + pAsm->D2.dst2.literal_slots = 1; + pAsm->C[0].f = 1.5F; + pAsm->S[2].src.rtype = SRC_REC_LITERAL; + pAsm->S[2].src.reg = tmp1; + setswizzle_PVSSRC(&(pAsm->S[2].src), SQ_SEL_X); + + next_ins(pAsm); + + /* tmp1.xy = temp2.xy */ + pAsm->D.dst.opcode = SQ_OP2_INST_MOV; + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = tmp1; + pAsm->D.dst.writex = 1; + pAsm->D.dst.writey = 1; + pAsm->D.dst.writez = 0; + pAsm->D.dst.writew = 0; + + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[0].src.reg = tmp2; + noswizzle_PVSSRC(&(pAsm->S[0].src)); + + next_ins(pAsm); + pAsm->aArgSubst[1] = tmp1; + need_barrier = GL_TRUE; + + } + + if(pAsm->pILInst[pAsm->uiCurInst].Opcode == OPCODE_TXB) + { + pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE_L; + } + else + { + pAsm->D.dst.opcode = SQ_TEX_INST_SAMPLE; + } + + pAsm->is_tex = GL_TRUE; + if ( GL_TRUE == need_barrier ) + + pAsm->is_tex = GL_TRUE; + if ( GL_TRUE == need_barrier ) + { + pAsm->need_tex_barrier = GL_TRUE; + } // Set src1 to tex unit id - pAsm->S[1].src.reg = pAsm->pILInst[pAsm->uiCurInst].TexSrcUnit; + pAsm->S[1].src.reg = pAsm->SamplerUnits[pAsm->pILInst[pAsm->uiCurInst].TexSrcUnit]; pAsm->S[1].src.rtype = SRC_REG_TEMPORARY; //No sw info from mesa compiler, so hard code here. @@ -3437,10 +4368,25 @@ GLboolean assemble_TEX(r700_AssemblerBase *pAsm) return GL_FALSE; } - if ( GL_FALSE == next_ins(pAsm) ) + if(pAsm->pILInst[pAsm->uiCurInst].Opcode == OPCODE_TXP) { - return GL_FALSE; + /* hopefully did swizzles before */ + noswizzle_PVSSRC(&(pAsm->S[0].src)); + } + + if(pAsm->pILInst[pAsm->uiCurInst].TexSrcTarget == TEXTURE_CUBE_INDEX) + { + /* SAMPLE dst, tmp.yxwy, CUBE */ + pAsm->S[0].src.swizzlex = SQ_SEL_Y; + pAsm->S[0].src.swizzley = SQ_SEL_X; + pAsm->S[0].src.swizzlez = SQ_SEL_W; + pAsm->S[0].src.swizzlew = SQ_SEL_Y; } + + if ( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } return GL_TRUE; } @@ -3560,27 +4506,909 @@ GLboolean assemble_EXPORT(r700_AssemblerBase *pAsm) return GL_TRUE; } -GLboolean assemble_IF(r700_AssemblerBase *pAsm) +static inline void decreaseCurrent(r700_AssemblerBase *pAsm, GLuint uReason) +{ + switch (uReason) + { + case FC_PUSH_VPM: + pAsm->CALLSTACK[pAsm->CALLSP].current--; + break; + case FC_PUSH_WQM: + pAsm->CALLSTACK[pAsm->CALLSP].current -= 4; + break; + case FC_LOOP: + pAsm->CALLSTACK[pAsm->CALLSP].current -= 4; + break; + case FC_REP: + /* TODO : for 16 vp asic, should -= 2; */ + pAsm->CALLSTACK[pAsm->CALLSP].current -= 1; + break; + }; +} + +static inline void checkStackDepth(r700_AssemblerBase *pAsm, GLuint uReason, GLboolean bCheckMaxOnly) +{ + if(GL_TRUE == bCheckMaxOnly) + { + switch (uReason) + { + case FC_PUSH_VPM: + if((pAsm->CALLSTACK[pAsm->CALLSP].current + 1) + > pAsm->CALLSTACK[pAsm->CALLSP].max) + { + pAsm->CALLSTACK[pAsm->CALLSP].max = + pAsm->CALLSTACK[pAsm->CALLSP].current + 1; + } + break; + case FC_PUSH_WQM: + if((pAsm->CALLSTACK[pAsm->CALLSP].current + 4) + > pAsm->CALLSTACK[pAsm->CALLSP].max) + { + pAsm->CALLSTACK[pAsm->CALLSP].max = + pAsm->CALLSTACK[pAsm->CALLSP].current + 4; + } + break; + } + return; + } + + switch (uReason) + { + case FC_PUSH_VPM: + pAsm->CALLSTACK[pAsm->CALLSP].current++; + break; + case FC_PUSH_WQM: + pAsm->CALLSTACK[pAsm->CALLSP].current += 4; + break; + case FC_LOOP: + pAsm->CALLSTACK[pAsm->CALLSP].current += 4; + break; + case FC_REP: + /* TODO : for 16 vp asic, should += 2; */ + pAsm->CALLSTACK[pAsm->CALLSP].current += 1; + break; + }; + + if(pAsm->CALLSTACK[pAsm->CALLSP].current + > pAsm->CALLSTACK[pAsm->CALLSP].max) + { + pAsm->CALLSTACK[pAsm->CALLSP].max = + pAsm->CALLSTACK[pAsm->CALLSP].current; + } +} + +GLboolean jumpToOffest(r700_AssemblerBase *pAsm, GLuint pops, GLint offset) +{ + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = pops; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_JUMP; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->cf_current_cf_clause_ptr->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex + offset; + + return GL_TRUE; +} + +GLboolean pops(r700_AssemblerBase *pAsm, GLuint pops) +{ + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = pops; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_POP; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + pAsm->cf_current_cf_clause_ptr->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex + 1; + + return GL_TRUE; +} + +GLboolean assemble_IF(r700_AssemblerBase *pAsm, GLboolean bHasElse) { + pAsm->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; + + assemble_LOGIC_PRED(pAsm, SQ_OP2_INST_PRED_SETNE); + + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + if(GL_TRUE != bHasElse) + { + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + } + else + { + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 0; + } + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_JUMP; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->FCSP++; + pAsm->fc_stack[pAsm->FCSP].type = FC_IF; + pAsm->fc_stack[pAsm->FCSP].mid = NULL; + pAsm->fc_stack[pAsm->FCSP].midLen= 0; + pAsm->fc_stack[pAsm->FCSP].first = pAsm->cf_current_cf_clause_ptr; + +#ifndef USE_CF_FOR_POP_AFTER + if(GL_TRUE != bHasElse) + { + pAsm->alu_x_opcode = SQ_CF_INST_ALU_POP_AFTER; + } +#endif /* USE_CF_FOR_POP_AFTER */ + + checkStackDepth(pAsm, FC_PUSH_VPM, GL_FALSE); + + return GL_TRUE; +} + +GLboolean assemble_ELSE(r700_AssemblerBase *pAsm) +{ + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; /// + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_ELSE; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->fc_stack[pAsm->FCSP].mid = (R700ControlFlowGenericClause **)_mesa_realloc( (void *)pAsm->fc_stack[pAsm->FCSP].mid, + 0, + sizeof(R700ControlFlowGenericClause *) ); + pAsm->fc_stack[pAsm->FCSP].mid[0] = pAsm->cf_current_cf_clause_ptr; + //pAsm->fc_stack[pAsm->FCSP].unNumMid = 1; + +#ifndef USE_CF_FOR_POP_AFTER + pAsm->alu_x_opcode = SQ_CF_INST_ALU_POP_AFTER; +#endif /* USE_CF_FOR_POP_AFTER */ + + pAsm->fc_stack[pAsm->FCSP].first->m_Word0.f.addr = pAsm->pR700Shader->plstCFInstructions_active->uNumOfNode - 1; + return GL_TRUE; } GLboolean assemble_ENDIF(r700_AssemblerBase *pAsm) { +#ifdef USE_CF_FOR_POP_AFTER + pops(pAsm, 1); +#endif /* USE_CF_FOR_POP_AFTER */ + + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + + if(NULL == pAsm->fc_stack[pAsm->FCSP].mid) + { + /* no else in between */ + pAsm->fc_stack[pAsm->FCSP].first->m_Word0.f.addr = pAsm->pR700Shader->plstCFInstructions_active->uNumOfNode; + } + else + { + pAsm->fc_stack[pAsm->FCSP].mid[0]->m_Word0.f.addr = pAsm->pR700Shader->plstCFInstructions_active->uNumOfNode; + } + + if(NULL != pAsm->fc_stack[pAsm->FCSP].mid) + { + FREE(pAsm->fc_stack[pAsm->FCSP].mid); + } + + if(pAsm->fc_stack[pAsm->FCSP].type != FC_IF) + { + radeon_error("if/endif in shader code are not paired. \n"); + return GL_FALSE; + } + + pAsm->FCSP--; + + decreaseCurrent(pAsm, FC_PUSH_VPM); + return GL_TRUE; } -GLboolean AssembleInstr(GLuint uiNumberInsts, +GLboolean assemble_BGNLOOP(r700_AssemblerBase *pAsm) +{ + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_LOOP_START_NO_AL; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->FCSP++; + pAsm->fc_stack[pAsm->FCSP].type = FC_LOOP; + pAsm->fc_stack[pAsm->FCSP].mid = NULL; + pAsm->fc_stack[pAsm->FCSP].unNumMid = 0; + pAsm->fc_stack[pAsm->FCSP].midLen = 0; + pAsm->fc_stack[pAsm->FCSP].first = pAsm->cf_current_cf_clause_ptr; + + checkStackDepth(pAsm, FC_LOOP, GL_FALSE); + + return GL_TRUE; +} + +GLboolean assemble_BRK(r700_AssemblerBase *pAsm) +{ +#ifdef USE_CF_FOR_CONTINUE_BREAK + + pAsm->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; + + assemble_LOGIC_PRED(pAsm, SQ_OP2_INST_PRED_SETNE); + + unsigned int unFCSP; + for(unFCSP=pAsm->FCSP; unFCSP>0; unFCSP--) + { + if(FC_LOOP == pAsm->fc_stack[unFCSP].type) + { + break; + } + } + if(0 == FC_LOOP) + { + radeon_error("Break is not inside loop/endloop pair.\n"); + return GL_FALSE; + } + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_LOOP_BREAK; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->fc_stack[unFCSP].mid = (R700ControlFlowGenericClause **)_mesa_realloc( + (void *)pAsm->fc_stack[unFCSP].mid, + sizeof(R700ControlFlowGenericClause *) * pAsm->fc_stack[unFCSP].unNumMid, + sizeof(R700ControlFlowGenericClause *) * (pAsm->fc_stack[unFCSP].unNumMid + 1) ); + pAsm->fc_stack[unFCSP].mid[pAsm->fc_stack[unFCSP].unNumMid] = pAsm->cf_current_cf_clause_ptr; + pAsm->fc_stack[unFCSP].unNumMid++; + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_POP; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + pAsm->cf_current_cf_clause_ptr->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex + 1; + + checkStackDepth(pAsm, FC_PUSH_VPM, GL_TRUE); + +#endif //USE_CF_FOR_CONTINUE_BREAK + return GL_TRUE; +} + +GLboolean assemble_CONT(r700_AssemblerBase *pAsm) +{ +#ifdef USE_CF_FOR_CONTINUE_BREAK + pAsm->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; + + assemble_LOGIC_PRED(pAsm, SQ_OP2_INST_PRED_SETNE); + + unsigned int unFCSP; + for(unFCSP=pAsm->FCSP; unFCSP>0; unFCSP--) + { + if(FC_LOOP == pAsm->fc_stack[unFCSP].type) + { + break; + } + } + if(0 == FC_LOOP) + { + radeon_error("Continue is not inside loop/endloop pair.\n"); + return GL_FALSE; + } + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_LOOP_CONTINUE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->fc_stack[unFCSP].mid = (R700ControlFlowGenericClause **)_mesa_realloc( + (void *)pAsm->fc_stack[unFCSP].mid, + sizeof(R700ControlFlowGenericClause *) * pAsm->fc_stack[unFCSP].unNumMid, + sizeof(R700ControlFlowGenericClause *) * (pAsm->fc_stack[unFCSP].unNumMid + 1) ); + pAsm->fc_stack[unFCSP].mid[pAsm->fc_stack[unFCSP].unNumMid] = pAsm->cf_current_cf_clause_ptr; + pAsm->fc_stack[unFCSP].unNumMid++; + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_POP; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + pAsm->cf_current_cf_clause_ptr->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex + 1; + + checkStackDepth(pAsm, FC_PUSH_VPM, GL_TRUE); + +#endif /* USE_CF_FOR_CONTINUE_BREAK */ + + return GL_TRUE; +} + +GLboolean assemble_ENDLOOP(r700_AssemblerBase *pAsm) +{ + GLuint i; + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_LOOP_END; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->cf_current_cf_clause_ptr->m_Word0.f.addr = pAsm->fc_stack[pAsm->FCSP].first->m_uIndex + 1; + pAsm->fc_stack[pAsm->FCSP].first->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex + 1; + +#ifdef USE_CF_FOR_CONTINUE_BREAK + for(i=0; i<pAsm->fc_stack[pAsm->FCSP].unNumMid; i++) + { + pAsm->fc_stack[pAsm->FCSP].mid[i]->m_Word0.f.addr = pAsm->cf_current_cf_clause_ptr->m_uIndex; + } + if(NULL != pAsm->fc_stack[pAsm->FCSP].mid) + { + FREE(pAsm->fc_stack[pAsm->FCSP].mid); + } +#endif + + if(pAsm->fc_stack[pAsm->FCSP].type != FC_LOOP) + { + radeon_error("loop/endloop in shader code are not paired. \n"); + return GL_FALSE; + } + + GLuint unFCSP; + GLuint unIF = 0; + if((pAsm->unCFflags & HAS_CURRENT_LOOPRET) > 0) + { + for(unFCSP=(pAsm->FCSP-1); unFCSP>pAsm->CALLSTACK[pAsm->CALLSP].FCSP_BeforeEntry; unFCSP--) + { + if(FC_LOOP == pAsm->fc_stack[unFCSP].type) + { + breakLoopOnFlag(pAsm, unFCSP); + break; + } + else if(FC_IF == pAsm->fc_stack[unFCSP].type) + { + unIF++; + } + } + if(unFCSP <= pAsm->CALLSTACK[pAsm->CALLSP].FCSP_BeforeEntry) + { +#ifdef USE_CF_FOR_POP_AFTER + returnOnFlag(pAsm, unIF); +#else + returnOnFlag(pAsm, 0); +#endif /* USE_CF_FOR_POP_AFTER */ + pAsm->unCFflags &= ~HAS_CURRENT_LOOPRET; + } + } + + pAsm->FCSP--; + + decreaseCurrent(pAsm, FC_LOOP); + + return GL_TRUE; +} + +void add_return_inst(r700_AssemblerBase *pAsm) +{ + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + //pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_RETURN; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; +} + +GLboolean assemble_BGNSUB(r700_AssemblerBase *pAsm, GLint nILindex, GLuint uiIL_Shift) +{ + /* Put in sub */ + if( (pAsm->unSubArrayPointer + 1) > pAsm->unSubArraySize ) + { + pAsm->subs = (SUB_OFFSET*)_mesa_realloc( (void *)pAsm->subs, + sizeof(SUB_OFFSET) * pAsm->unSubArraySize, + sizeof(SUB_OFFSET) * (pAsm->unSubArraySize + 10) ); + if(NULL == pAsm->subs) + { + return GL_FALSE; + } + pAsm->unSubArraySize += 10; + } + + pAsm->subs[pAsm->unSubArrayPointer].subIL_Offset = nILindex + uiIL_Shift; + pAsm->subs[pAsm->unSubArrayPointer].lstCFInstructions_local.pHead=NULL; + pAsm->subs[pAsm->unSubArrayPointer].lstCFInstructions_local.pTail=NULL; + pAsm->subs[pAsm->unSubArrayPointer].lstCFInstructions_local.uNumOfNode=0; + + pAsm->CALLSP++; + pAsm->CALLSTACK[pAsm->CALLSP].subDescIndex = pAsm->unSubArrayPointer; + pAsm->CALLSTACK[pAsm->CALLSP].FCSP_BeforeEntry = pAsm->FCSP; + pAsm->CALLSTACK[pAsm->CALLSP].plstCFInstructions_local + = &(pAsm->subs[pAsm->unSubArrayPointer].lstCFInstructions_local); + pAsm->CALLSTACK[pAsm->CALLSP].max = 0; + pAsm->CALLSTACK[pAsm->CALLSP].current = 0; + SetActiveCFlist(pAsm->pR700Shader, + pAsm->CALLSTACK[pAsm->CALLSP].plstCFInstructions_local); + + pAsm->unSubArrayPointer++; + + /* start sub */ + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + + pAsm->FCSP++; + pAsm->fc_stack[pAsm->FCSP].type = FC_REP; + + checkStackDepth(pAsm, FC_REP, GL_FALSE); + + return GL_TRUE; +} + +GLboolean assemble_ENDSUB(r700_AssemblerBase *pAsm) +{ + if(pAsm->fc_stack[pAsm->FCSP].type != FC_REP) + { + radeon_error("BGNSUB/ENDSUB in shader code are not paired. \n"); + return GL_FALSE; + } + + /* copy max to sub structure */ + pAsm->subs[pAsm->CALLSTACK[pAsm->CALLSP].subDescIndex].unStackDepthMax + = pAsm->CALLSTACK[pAsm->CALLSP].max; + + decreaseCurrent(pAsm, FC_REP); + + pAsm->CALLSP--; + SetActiveCFlist(pAsm->pR700Shader, + pAsm->CALLSTACK[pAsm->CALLSP].plstCFInstructions_local); + + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + + pAsm->FCSP--; + + return GL_TRUE; +} + +GLboolean assemble_RET(r700_AssemblerBase *pAsm) +{ + GLuint unIF = 0; + + if(pAsm->CALLSP > 0) + { /* in sub */ + GLuint unFCSP; + for(unFCSP=pAsm->FCSP; unFCSP>pAsm->CALLSTACK[pAsm->CALLSP].FCSP_BeforeEntry; unFCSP--) + { + if(FC_LOOP == pAsm->fc_stack[unFCSP].type) + { + setRetInLoopFlag(pAsm, SQ_SEL_1); + breakLoopOnFlag(pAsm, unFCSP); + pAsm->unCFflags |= LOOPRET_FLAGS; + + return GL_TRUE; + } + else if(FC_IF == pAsm->fc_stack[unFCSP].type) + { + unIF++; + } + } + } + +#ifdef USE_CF_FOR_POP_AFTER + if(unIF > 0) + { + pops(pAsm, unIF); + } +#endif /* USE_CF_FOR_POP_AFTER */ + + add_return_inst(pAsm); + + return GL_TRUE; +} + +GLboolean assemble_CAL(r700_AssemblerBase *pAsm, + GLint nILindex, + GLuint uiIL_Shift, + GLuint uiNumberInsts, + struct prog_instruction *pILInst, + PRESUB_DESC * pPresubDesc) +{ + GLint uiIL_Offset; + + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.call_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_CALL; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + /* Put in caller */ + if( (pAsm->unCallerArrayPointer + 1) > pAsm->unCallerArraySize ) + { + pAsm->callers = (CALLER_POINTER*)_mesa_realloc( (void *)pAsm->callers, + sizeof(CALLER_POINTER) * pAsm->unCallerArraySize, + sizeof(CALLER_POINTER) * (pAsm->unCallerArraySize + 10) ); + if(NULL == pAsm->callers) + { + return GL_FALSE; + } + pAsm->unCallerArraySize += 10; + } + + uiIL_Offset = nILindex + uiIL_Shift; + pAsm->callers[pAsm->unCallerArrayPointer].subIL_Offset = uiIL_Offset; + pAsm->callers[pAsm->unCallerArrayPointer].cf_ptr = pAsm->cf_current_cf_clause_ptr; + + pAsm->callers[pAsm->unCallerArrayPointer].finale_cf_ptr = NULL; + pAsm->callers[pAsm->unCallerArrayPointer].prelude_cf_ptr = NULL; + + pAsm->unCallerArrayPointer++; + + int j; + GLuint max; + GLuint unSubID; + GLboolean bRet; + for(j=0; j<pAsm->unSubArrayPointer; j++) + { + if(uiIL_Offset == pAsm->subs[j].subIL_Offset) + { /* compiled before */ + + max = pAsm->subs[j].unStackDepthMax + + pAsm->CALLSTACK[pAsm->CALLSP].current; + if(max > pAsm->CALLSTACK[pAsm->CALLSP].max) + { + pAsm->CALLSTACK[pAsm->CALLSP].max = max; + } + + pAsm->callers[pAsm->unCallerArrayPointer - 1].subDescIndex = j; + return GL_TRUE; + } + } + + pAsm->callers[pAsm->unCallerArrayPointer - 1].subDescIndex = pAsm->unSubArrayPointer; + unSubID = pAsm->unSubArrayPointer; + + bRet = AssembleInstr(nILindex, uiIL_Shift, uiNumberInsts, pILInst, pAsm); + + if(GL_TRUE == bRet) + { + max = pAsm->subs[unSubID].unStackDepthMax + + pAsm->CALLSTACK[pAsm->CALLSP].current; + if(max > pAsm->CALLSTACK[pAsm->CALLSP].max) + { + pAsm->CALLSTACK[pAsm->CALLSP].max = max; + } + + pAsm->subs[unSubID].pPresubDesc = pPresubDesc; + } + + return bRet; +} + +GLboolean setRetInLoopFlag(r700_AssemblerBase *pAsm, GLuint flagValue) +{ + GLfloat fLiteral[2] = {0.1, 0.0}; + + pAsm->D.dst.opcode = SQ_OP2_INST_MOV; + pAsm->D.dst.op3 = 0; + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = pAsm->flag_reg_index; + pAsm->D.dst.writex = 1; + pAsm->D.dst.writey = 0; + pAsm->D.dst.writez = 0; + pAsm->D.dst.writew = 0; + pAsm->D2.dst2.literal_slots = 1; + pAsm->D2.dst2.SaturateMode = SATURATE_OFF; + pAsm->D.dst.predicated = 0; + /* in reloc where dislink flag init inst, only one slot alu inst is handled. */ + pAsm->D.dst.math = 1; /* TODO : not math really, but one channel op, more generic alu assembler needed */ + pAsm->D2.dst2.index_mode = SQ_INDEX_LOOP; /* Check this ! */ +#if 0 + pAsm->S[0].src.rtype = SRC_REC_LITERAL; + //pAsm->S[0].src.reg = 0; + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + noneg_PVSSRC(&(pAsm->S[0].src)); + pAsm->S[0].src.swizzlex = SQ_SEL_X; + pAsm->S[0].src.swizzley = SQ_SEL_Y; + pAsm->S[0].src.swizzlez = SQ_SEL_Z; + pAsm->S[0].src.swizzlew = SQ_SEL_W; + + if( GL_FALSE == next_ins_literal(pAsm, &(fLiteral[0])) ) + { + return GL_FALSE; + } +#else + pAsm->S[0].src.rtype = DST_REG_TEMPORARY; + pAsm->S[0].src.reg = 0; + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + noneg_PVSSRC(&(pAsm->S[0].src)); + pAsm->S[0].src.swizzlex = flagValue; + pAsm->S[0].src.swizzley = flagValue; + pAsm->S[0].src.swizzlez = flagValue; + pAsm->S[0].src.swizzlew = flagValue; + + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } +#endif + + return GL_TRUE; +} + +GLboolean testFlag(r700_AssemblerBase *pAsm) +{ + GLfloat fLiteral[2] = {0.1, 0.0}; + + //Test flag + GLuint tmp = gethelpr(pAsm); + pAsm->alu_x_opcode = SQ_CF_INST_ALU_PUSH_BEFORE; + + pAsm->D.dst.opcode = SQ_OP2_INST_PRED_SETE; + pAsm->D.dst.math = 1; + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = tmp; + pAsm->D.dst.writex = 1; + pAsm->D.dst.writey = 0; + pAsm->D.dst.writez = 0; + pAsm->D.dst.writew = 0; + pAsm->D2.dst2.literal_slots = 1; + pAsm->D2.dst2.SaturateMode = SATURATE_OFF; + pAsm->D.dst.predicated = 1; + pAsm->D2.dst2.index_mode = SQ_INDEX_LOOP; /* Check this ! */ + + pAsm->S[0].src.rtype = DST_REG_TEMPORARY; + pAsm->S[0].src.reg = pAsm->flag_reg_index; + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + noneg_PVSSRC(&(pAsm->S[0].src)); + pAsm->S[0].src.swizzlex = SQ_SEL_X; + pAsm->S[0].src.swizzley = SQ_SEL_Y; + pAsm->S[0].src.swizzlez = SQ_SEL_Z; + pAsm->S[0].src.swizzlew = SQ_SEL_W; +#if 0 + pAsm->S[1].src.rtype = SRC_REC_LITERAL; + //pAsm->S[1].src.reg = 0; + setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE); + noneg_PVSSRC(&(pAsm->S[1].src)); + pAsm->S[1].src.swizzlex = SQ_SEL_X; + pAsm->S[1].src.swizzley = SQ_SEL_Y; + pAsm->S[1].src.swizzlez = SQ_SEL_Z; + pAsm->S[1].src.swizzlew = SQ_SEL_W; + + if( GL_FALSE == next_ins_literal(pAsm, &(fLiteral[0])) ) + { + return GL_FALSE; + } +#else + pAsm->S[1].src.rtype = DST_REG_TEMPORARY; + pAsm->S[1].src.reg = 0; + setaddrmode_PVSSRC(&(pAsm->S[1].src), ADDR_ABSOLUTE); + noneg_PVSSRC(&(pAsm->S[1].src)); + pAsm->S[1].src.swizzlex = SQ_SEL_1; + pAsm->S[1].src.swizzley = SQ_SEL_1; + pAsm->S[1].src.swizzlez = SQ_SEL_1; + pAsm->S[1].src.swizzlew = SQ_SEL_1; + + if( GL_FALSE == next_ins(pAsm) ) + { + return GL_FALSE; + } +#endif + + checkStackDepth(pAsm, FC_PUSH_VPM, GL_TRUE); + + return GL_TRUE; +} + +GLboolean returnOnFlag(r700_AssemblerBase *pAsm, GLuint unIF) +{ + testFlag(pAsm); + jumpToOffest(pAsm, 1, 4); + setRetInLoopFlag(pAsm, SQ_SEL_0); + pops(pAsm, unIF + 1); + add_return_inst(pAsm); + + return GL_TRUE; +} + +GLboolean breakLoopOnFlag(r700_AssemblerBase *pAsm, GLuint unFCSP) +{ + testFlag(pAsm); + + //break + if(GL_FALSE == add_cf_instruction(pAsm) ) + { + return GL_FALSE; + } + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.pop_count = 1; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_const = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cond = SQ_CF_COND_ACTIVE; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.end_of_program = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.valid_pixel_mode = 0x0; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.cf_inst = SQ_CF_INST_LOOP_BREAK; + pAsm->cf_current_cf_clause_ptr->m_Word1.f.whole_quad_mode = 0x0; + + pAsm->cf_current_cf_clause_ptr->m_Word1.f.barrier = 0x1; + + pAsm->fc_stack[unFCSP].mid = (R700ControlFlowGenericClause **)_mesa_realloc( + (void *)pAsm->fc_stack[unFCSP].mid, + sizeof(R700ControlFlowGenericClause *) * pAsm->fc_stack[unFCSP].unNumMid, + sizeof(R700ControlFlowGenericClause *) * (pAsm->fc_stack[unFCSP].unNumMid + 1) ); + pAsm->fc_stack[unFCSP].mid[pAsm->fc_stack[unFCSP].unNumMid] = pAsm->cf_current_cf_clause_ptr; + pAsm->fc_stack[unFCSP].unNumMid++; + + pops(pAsm, 1); + + return GL_TRUE; +} + +GLboolean AssembleInstr(GLuint uiFirstInst, + GLuint uiIL_Shift, + GLuint uiNumberInsts, struct prog_instruction *pILInst, r700_AssemblerBase *pR700AsmCode) { GLuint i; pR700AsmCode->pILInst = pILInst; - for(i=0; i<uiNumberInsts; i++) + for(i=uiFirstInst; i<uiNumberInsts; i++) { pR700AsmCode->uiCurInst = i; +#ifndef USE_CF_FOR_CONTINUE_BREAK + if(OPCODE_BRK == pILInst[i+1].Opcode) + { + switch(pILInst[i].Opcode) + { + case OPCODE_SLE: + pILInst[i].Opcode = OPCODE_SGT; + break; + case OPCODE_SLT: + pILInst[i].Opcode = OPCODE_SGE; + break; + case OPCODE_SGE: + pILInst[i].Opcode = OPCODE_SLT; + break; + case OPCODE_SGT: + pILInst[i].Opcode = OPCODE_SLE; + break; + case OPCODE_SEQ: + pILInst[i].Opcode = OPCODE_SNE; + break; + case OPCODE_SNE: + pILInst[i].Opcode = OPCODE_SEQ; + break; + default: + break; + } + } +#endif + if(pILInst[i].CondUpdate == 1) + { + /* remember dest register used for cond evaluation */ + /* XXX also handle PROGRAM_OUTPUT registers here? */ + pR700AsmCode->last_cond_register = pILInst[i].DstReg.Index; + } + switch (pILInst[i].Opcode) { case OPCODE_ABS: @@ -3594,8 +5422,7 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, break; case OPCODE_ARL: - radeon_error("Not yet implemented instruction OPCODE_ARL \n"); - //if ( GL_FALSE == assemble_BAD("ARL") ) + if ( GL_FALSE == assemble_ARL(pR700AsmCode) ) return GL_FALSE; break; case OPCODE_ARR: @@ -3609,7 +5436,7 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, return GL_FALSE; break; case OPCODE_COS: - if ( GL_FALSE == assemble_COS(pR700AsmCode) ) + if ( GL_FALSE == assemble_TRIG(pR700AsmCode, SQ_OP2_INST_COS) ) return GL_FALSE; break; @@ -3630,16 +5457,16 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, return GL_FALSE; break; case OPCODE_EXP: - radeon_error("Not yet implemented instruction OPCODE_EXP \n"); - //if ( GL_FALSE == assemble_BAD("EXP") ) + if ( GL_FALSE == assemble_EXP(pR700AsmCode) ) return GL_FALSE; - break; // approx of EX2 + break; case OPCODE_FLR: if ( GL_FALSE == assemble_FLR(pR700AsmCode) ) return GL_FALSE; break; - //case OP_FLR_INT: + //case OP_FLR_INT: ; + // if ( GL_FALSE == assemble_FLR_INT() ) // return GL_FALSE; // break; @@ -3650,7 +5477,8 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, break; case OPCODE_KIL: - if ( GL_FALSE == assemble_KIL(pR700AsmCode) ) + case OPCODE_KIL_NV: + if ( GL_FALSE == assemble_KIL(pR700AsmCode, SQ_OP2_INST_KILLGT) ) return GL_FALSE; break; case OPCODE_LG2: @@ -3666,10 +5494,9 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, return GL_FALSE; break; case OPCODE_LOG: - radeon_error("Not yet implemented instruction OPCODE_LOG \n"); - //if ( GL_FALSE == assemble_BAD("LOG") ) + if ( GL_FALSE == assemble_LOG(pR700AsmCode) ) return GL_FALSE; - break; // approx of LG2 + break; case OPCODE_MAD: if ( GL_FALSE == assemble_MAD(pR700AsmCode) ) @@ -3691,6 +5518,26 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, case OPCODE_MUL: if ( GL_FALSE == assemble_MUL(pR700AsmCode) ) return GL_FALSE; + break; + + case OPCODE_NOISE1: + { + callPreSub(pR700AsmCode, + GLSL_NOISE1, + &noise1_presub, + pILInst->DstReg.Index + pR700AsmCode->starting_temp_register_number, + 1); + radeon_error("noise1: not yet supported shader instruction\n"); + }; + break; + case OPCODE_NOISE2: + radeon_error("noise2: not yet supported shader instruction\n"); + break; + case OPCODE_NOISE3: + radeon_error("noise3: not yet supported shader instruction\n"); + break; + case OPCODE_NOISE4: + radeon_error("noise4: not yet supported shader instruction\n"); break; case OPCODE_POW: @@ -3706,22 +5553,78 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, return GL_FALSE; break; case OPCODE_SIN: - if ( GL_FALSE == assemble_SIN(pR700AsmCode) ) + if ( GL_FALSE == assemble_TRIG(pR700AsmCode, SQ_OP2_INST_SIN) ) return GL_FALSE; break; case OPCODE_SCS: if ( GL_FALSE == assemble_SCS(pR700AsmCode) ) return GL_FALSE; - break; + break; + + case OPCODE_SEQ: + if ( GL_FALSE == assemble_LOGIC(pR700AsmCode, SQ_OP2_INST_SETE) ) + { + return GL_FALSE; + } + break; + + case OPCODE_SGT: + if ( GL_FALSE == assemble_LOGIC(pR700AsmCode, SQ_OP2_INST_SETGT) ) + { + return GL_FALSE; + } + break; case OPCODE_SGE: if ( GL_FALSE == assemble_SGE(pR700AsmCode) ) + { return GL_FALSE; - break; + } + break; + + /* NO LT, LE, TODO : use GE => LE, GT => LT : reverse 2 src order would be simpliest. Or use SQ_CF_COND_FALSE for SQ_CF_COND_ACTIVE.*/ case OPCODE_SLT: - if ( GL_FALSE == assemble_SLT(pR700AsmCode) ) + { + struct prog_src_register SrcRegSave[2]; + SrcRegSave[0] = pILInst[i].SrcReg[0]; + SrcRegSave[1] = pILInst[i].SrcReg[1]; + pILInst[i].SrcReg[0] = SrcRegSave[1]; + pILInst[i].SrcReg[1] = SrcRegSave[0]; + if ( GL_FALSE == assemble_LOGIC(pR700AsmCode, SQ_OP2_INST_SETGT) ) + { + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + return GL_FALSE; + } + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + } + break; + + case OPCODE_SLE: + { + struct prog_src_register SrcRegSave[2]; + SrcRegSave[0] = pILInst[i].SrcReg[0]; + SrcRegSave[1] = pILInst[i].SrcReg[1]; + pILInst[i].SrcReg[0] = SrcRegSave[1]; + pILInst[i].SrcReg[1] = SrcRegSave[0]; + if ( GL_FALSE == assemble_LOGIC(pR700AsmCode, SQ_OP2_INST_SETGE) ) + { + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + return GL_FALSE; + } + pILInst[i].SrcReg[0] = SrcRegSave[0]; + pILInst[i].SrcReg[1] = SrcRegSave[1]; + } + break; + + case OPCODE_SNE: + if ( GL_FALSE == assemble_LOGIC(pR700AsmCode, SQ_OP2_INST_SETNE) ) + { return GL_FALSE; - break; + } + break; //case OP_STP: // if ( GL_FALSE == assemble_STP(pR700AsmCode) ) @@ -3755,30 +5658,104 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, return GL_FALSE; break; + case OPCODE_TRUNC: + if ( GL_FALSE == assemble_math_function(pR700AsmCode, SQ_OP2_INST_TRUNC) ) + return GL_FALSE; + break; + case OPCODE_XPD: if ( GL_FALSE == assemble_XPD(pR700AsmCode) ) return GL_FALSE; break; case OPCODE_IF : - if ( GL_FALSE == assemble_IF(pR700AsmCode) ) - return GL_FALSE; + { + GLboolean bHasElse = GL_FALSE; + + if(pILInst[pILInst[i].BranchTarget - 1].Opcode == OPCODE_ELSE) + { + bHasElse = GL_TRUE; + } + + if ( GL_FALSE == assemble_IF(pR700AsmCode, bHasElse) ) + { + return GL_FALSE; + } + } break; + case OPCODE_ELSE : - radeon_error("Not yet implemented instruction OPCODE_ELSE \n"); - //if ( GL_FALSE == assemble_BAD("ELSE") ) + if ( GL_FALSE == assemble_ELSE(pR700AsmCode) ) return GL_FALSE; break; + case OPCODE_ENDIF: if ( GL_FALSE == assemble_ENDIF(pR700AsmCode) ) return GL_FALSE; break; + case OPCODE_BGNLOOP: + if( GL_FALSE == assemble_BGNLOOP(pR700AsmCode) ) + { + return GL_FALSE; + } + break; + + case OPCODE_BRK: + if( GL_FALSE == assemble_BRK(pR700AsmCode) ) + { + return GL_FALSE; + } + break; + + case OPCODE_CONT: + if( GL_FALSE == assemble_CONT(pR700AsmCode) ) + { + return GL_FALSE; + } + break; + + case OPCODE_ENDLOOP: + if( GL_FALSE == assemble_ENDLOOP(pR700AsmCode) ) + { + return GL_FALSE; + } + break; + + case OPCODE_BGNSUB: + if( GL_FALSE == assemble_BGNSUB(pR700AsmCode, i, uiIL_Shift) ) + { + return GL_FALSE; + } + break; + + case OPCODE_RET: + if( GL_FALSE == assemble_RET(pR700AsmCode) ) + { + return GL_FALSE; + } + break; + + case OPCODE_CAL: + if( GL_FALSE == assemble_CAL(pR700AsmCode, + pILInst[i].BranchTarget, + uiIL_Shift, + uiNumberInsts, + pILInst, + NULL) ) + { + return GL_FALSE; + } + break; + //case OPCODE_EXPORT: // if ( GL_FALSE == assemble_EXPORT() ) // return GL_FALSE; // break; + case OPCODE_ENDSUB: + return assemble_ENDSUB(pR700AsmCode); + case OPCODE_END: //pR700AsmCode->uiCurInst = i; //This is to remaind that if in later exoort there is depth/stencil @@ -3795,6 +5772,417 @@ GLboolean AssembleInstr(GLuint uiNumberInsts, return GL_TRUE; } +GLboolean InitShaderProgram(r700_AssemblerBase * pAsm) +{ + setRetInLoopFlag(pAsm, SQ_SEL_0); + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + return GL_TRUE; +} + +GLboolean RelocProgram(r700_AssemblerBase * pAsm, struct gl_program * pILProg) +{ + GLuint i; + GLuint unCFoffset; + TypedShaderList * plstCFmain; + TypedShaderList * plstCFsub; + + R700ShaderInstruction * pInst; + R700ControlFlowGenericClause * pCFInst; + + R700ControlFlowALUClause * pCF_ALU; + R700ALUInstruction * pALU; + GLuint unConstOffset = 0; + GLuint unRegOffset; + GLuint unMinRegIndex; + + plstCFmain = pAsm->CALLSTACK[0].plstCFInstructions_local; + + /* remove flags init if they are not used */ + if((pAsm->unCFflags & HAS_LOOPRET) == 0) + { + R700ControlFlowALUClause * pCF_ALU; + pInst = plstCFmain->pHead; + while(pInst) + { + if(SIT_CF_ALU == pInst->m_ShaderInstType) + { + pCF_ALU = (R700ControlFlowALUClause *)pInst; + if(0 == pCF_ALU->m_Word1.f.count) + { + pCF_ALU->m_Word1.f.cf_inst = SQ_CF_INST_NOP; + } + else + { + R700ALUInstruction * pALU = pCF_ALU->m_pLinkedALUInstruction; + + pALU->m_pLinkedALUClause = NULL; + pALU = (R700ALUInstruction *)(pALU->pNextInst); + pALU->m_pLinkedALUClause = pCF_ALU; + pCF_ALU->m_pLinkedALUInstruction = pALU; + + pCF_ALU->m_Word1.f.count--; + } + break; + } + pInst = pInst->pNextInst; + }; + } + + if(pAsm->CALLSTACK[0].max > 0) + { + pAsm->pR700Shader->uStackSize = ((pAsm->CALLSTACK[0].max + 3)>>2) + 2; + } + + if(0 == pAsm->unSubArrayPointer) + { + return GL_TRUE; + } + + unCFoffset = plstCFmain->uNumOfNode; + + if(NULL != pILProg->Parameters) + { + unConstOffset = pILProg->Parameters->NumParameters; + } + + /* Reloc subs */ + for(i=0; i<pAsm->unSubArrayPointer; i++) + { + pAsm->subs[i].unCFoffset = unCFoffset; + plstCFsub = &(pAsm->subs[i].lstCFInstructions_local); + + pInst = plstCFsub->pHead; + + /* reloc instructions */ + while(pInst) + { + if(SIT_CF_GENERIC == pInst->m_ShaderInstType) + { + pCFInst = (R700ControlFlowGenericClause *)pInst; + + switch (pCFInst->m_Word1.f.cf_inst) + { + case SQ_CF_INST_POP: + case SQ_CF_INST_JUMP: + case SQ_CF_INST_ELSE: + case SQ_CF_INST_LOOP_END: + case SQ_CF_INST_LOOP_START: + case SQ_CF_INST_LOOP_START_NO_AL: + case SQ_CF_INST_LOOP_CONTINUE: + case SQ_CF_INST_LOOP_BREAK: + pCFInst->m_Word0.f.addr += unCFoffset; + break; + default: + break; + } + } + + pInst->m_uIndex += unCFoffset; + + pInst = pInst->pNextInst; + }; + + if(NULL != pAsm->subs[i].pPresubDesc) + { + GLuint uNumSrc; + + unMinRegIndex = pAsm->subs[i].pPresubDesc->pCompiledSub->MinRegIndex; + unRegOffset = pAsm->subs[i].pPresubDesc->maxStartReg; + unConstOffset += pAsm->subs[i].pPresubDesc->unConstantsStart; + + pInst = plstCFsub->pHead; + while(pInst) + { + if(SIT_CF_ALU == pInst->m_ShaderInstType) + { + pCF_ALU = (R700ControlFlowALUClause *)pInst; + + pALU = pCF_ALU->m_pLinkedALUInstruction; + for(int j=0; j<=pCF_ALU->m_Word1.f.count; j++) + { + pALU->m_Word1.f.dst_gpr = pALU->m_Word1.f.dst_gpr + unRegOffset - unMinRegIndex; + + if(pALU->m_Word0.f.src0_sel < SQ_ALU_SRC_GPR_SIZE) + { + pALU->m_Word0.f.src0_sel = pALU->m_Word0.f.src0_sel + unRegOffset - unMinRegIndex; + } + else if(pALU->m_Word0.f.src0_sel >= SQ_ALU_SRC_CFILE_BASE) + { + pALU->m_Word0.f.src0_sel += unConstOffset; + } + + if( ((pALU->m_Word1.val >> SQ_ALU_WORD1_OP3_ALU_INST_SHIFT) & 0x0000001F) + >= SQ_OP3_INST_MUL_LIT ) + { /* op3 : 3 srcs */ + if(pALU->m_Word1_OP3.f.src2_sel < SQ_ALU_SRC_GPR_SIZE) + { + pALU->m_Word1_OP3.f.src2_sel = pALU->m_Word1_OP3.f.src2_sel + unRegOffset - unMinRegIndex; + } + else if(pALU->m_Word1_OP3.f.src2_sel >= SQ_ALU_SRC_CFILE_BASE) + { + pALU->m_Word1_OP3.f.src2_sel += unConstOffset; + } + if(pALU->m_Word0.f.src1_sel < SQ_ALU_SRC_GPR_SIZE) + { + pALU->m_Word0.f.src1_sel = pALU->m_Word0.f.src1_sel + unRegOffset - unMinRegIndex; + } + else if(pALU->m_Word0.f.src1_sel >= SQ_ALU_SRC_CFILE_BASE) + { + pALU->m_Word0.f.src1_sel += unConstOffset; + } + } + else + { + if(pAsm->bR6xx) + { + uNumSrc = r700GetNumOperands(pALU->m_Word1_OP2.f6.alu_inst, 0); + } + else + { + uNumSrc = r700GetNumOperands(pALU->m_Word1_OP2.f.alu_inst, 0); + } + if(2 == uNumSrc) + { /* 2 srcs */ + if(pALU->m_Word0.f.src1_sel < SQ_ALU_SRC_GPR_SIZE) + { + pALU->m_Word0.f.src1_sel = pALU->m_Word0.f.src1_sel + unRegOffset - unMinRegIndex; + } + else if(pALU->m_Word0.f.src1_sel >= SQ_ALU_SRC_CFILE_BASE) + { + pALU->m_Word0.f.src1_sel += unConstOffset; + } + } + } + pALU = (R700ALUInstruction*)(pALU->pNextInst); + } + } + pInst = pInst->pNextInst; + }; + } + + /* Put sub into main */ + plstCFmain->pTail->pNextInst = plstCFsub->pHead; + plstCFmain->pTail = plstCFsub->pTail; + plstCFmain->uNumOfNode += plstCFsub->uNumOfNode; + + unCFoffset += plstCFsub->uNumOfNode; + } + + /* reloc callers */ + for(i=0; i<pAsm->unCallerArrayPointer; i++) + { + pAsm->callers[i].cf_ptr->m_Word0.f.addr + = pAsm->subs[pAsm->callers[i].subDescIndex].unCFoffset; + + if(NULL != pAsm->subs[pAsm->callers[i].subDescIndex].pPresubDesc) + { + unMinRegIndex = pAsm->subs[pAsm->callers[i].subDescIndex].pPresubDesc->pCompiledSub->MinRegIndex; + unRegOffset = pAsm->subs[pAsm->callers[i].subDescIndex].pPresubDesc->maxStartReg; + + if(NULL != pAsm->callers[i].prelude_cf_ptr) + { + pCF_ALU = (R700ControlFlowALUClause * )(pAsm->callers[i].prelude_cf_ptr); + pALU = pCF_ALU->m_pLinkedALUInstruction; + for(int j=0; j<=pCF_ALU->m_Word1.f.count; j++) + { + pALU->m_Word1.f.dst_gpr = pALU->m_Word1.f.dst_gpr + unRegOffset - unMinRegIndex; + pALU = (R700ALUInstruction*)(pALU->pNextInst); + } + } + if(NULL != pAsm->callers[i].finale_cf_ptr) + { + pCF_ALU = (R700ControlFlowALUClause * )(pAsm->callers[i].finale_cf_ptr); + pALU = pCF_ALU->m_pLinkedALUInstruction; + for(int j=0; j<=pCF_ALU->m_Word1.f.count; j++) + { + pALU->m_Word0.f.src0_sel = pALU->m_Word0.f.src0_sel + unRegOffset - unMinRegIndex; + pALU = (R700ALUInstruction*)(pALU->pNextInst); + } + } + } + } + + return GL_TRUE; +} + +GLboolean callPreSub(r700_AssemblerBase* pAsm, + LOADABLE_SCRIPT_SIGNITURE scriptSigniture, + COMPILED_SUB * pCompiledSub, + GLshort uOutReg, + GLshort uNumValidSrc) +{ + /* save assemble context */ + GLuint starting_temp_register_number_save; + GLuint number_used_registers_save; + GLuint uFirstHelpReg_save; + GLuint uHelpReg_save; + GLuint uiCurInst_save; + struct prog_instruction *pILInst_save; + PRESUB_DESC * pPresubDesc; + GLboolean bRet; + int i; + + R700ControlFlowGenericClause* prelude_cf_ptr = NULL; + + /* copy srcs to presub inputs */ + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + for(i=0; i<uNumValidSrc; i++) + { + pAsm->D.dst.opcode = SQ_OP2_INST_MOV; + setaddrmode_PVSDST(&(pAsm->D.dst), ADDR_ABSOLUTE); + pAsm->D.dst.rtype = DST_REG_TEMPORARY; + pAsm->D.dst.reg = pCompiledSub->srcRegIndex[i]; + pAsm->D.dst.writex = 1; + pAsm->D.dst.writey = 1; + pAsm->D.dst.writez = 1; + pAsm->D.dst.writew = 1; + + if( GL_FALSE == assemble_src(pAsm, i, 0) ) + { + return GL_FALSE; + } + + next_ins(pAsm); + } + if(uNumValidSrc > 0) + { + prelude_cf_ptr = pAsm->cf_current_alu_clause_ptr; + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + } + + /* browse thro existing presubs. */ + for(i=0; i<pAsm->unNumPresub; i++) + { + if(pAsm->presubs[i].sptSigniture == scriptSigniture) + { + break; + } + } + + if(i == pAsm->unNumPresub) + { /* not loaded yet */ + /* save assemble context */ + number_used_registers_save = pAsm->number_used_registers; + uFirstHelpReg_save = pAsm->uFirstHelpReg; + uHelpReg_save = pAsm->uHelpReg; + starting_temp_register_number_save = pAsm->starting_temp_register_number; + pILInst_save = pAsm->pILInst; + uiCurInst_save = pAsm->uiCurInst; + + /* alloc in presub */ + if( (pAsm->unNumPresub + 1) > pAsm->unPresubArraySize ) + { + pAsm->presubs = (PRESUB_DESC*)_mesa_realloc( (void *)pAsm->presubs, + sizeof(PRESUB_DESC) * pAsm->unPresubArraySize, + sizeof(PRESUB_DESC) * (pAsm->unPresubArraySize + 4) ); + if(NULL == pAsm->presubs) + { + radeon_error("No memeory to allocate built in shader function description structures. \n"); + return GL_FALSE; + } + pAsm->unPresubArraySize += 4; + } + + pPresubDesc = &(pAsm->presubs[i]); + pPresubDesc->sptSigniture = scriptSigniture; + + /* constants offsets need to be final resolved at reloc. */ + if(0 == pAsm->unNumPresub) + { + pPresubDesc->unConstantsStart = 0; + } + else + { + pPresubDesc->unConstantsStart = pAsm->presubs[i-1].unConstantsStart + + pAsm->presubs[i-1].pCompiledSub->NumParameters; + } + + pPresubDesc->pCompiledSub = pCompiledSub; + + pPresubDesc->subIL_Shift = pAsm->unCurNumILInsts; + pPresubDesc->maxStartReg = uFirstHelpReg_save; + pAsm->unCurNumILInsts += pCompiledSub->NumInstructions; + + pAsm->unNumPresub++; + + /* setup new assemble context */ + pAsm->starting_temp_register_number = 0; + pAsm->number_used_registers = pCompiledSub->NumTemporaries; + pAsm->uFirstHelpReg = pAsm->number_used_registers; + pAsm->uHelpReg = pAsm->uFirstHelpReg; + + bRet = assemble_CAL(pAsm, + 0, + pPresubDesc->subIL_Shift, + pCompiledSub->NumInstructions, + pCompiledSub->Instructions, + pPresubDesc); + + + pPresubDesc->number_used_registers = pAsm->number_used_registers; + + /* restore assemble context */ + pAsm->number_used_registers = number_used_registers_save; + pAsm->uFirstHelpReg = uFirstHelpReg_save; + pAsm->uHelpReg = uHelpReg_save; + pAsm->starting_temp_register_number = starting_temp_register_number_save; + pAsm->pILInst = pILInst_save; + pAsm->uiCurInst = uiCurInst_save; + } + else + { /* was loaded */ + pPresubDesc = &(pAsm->presubs[i]); + + bRet = assemble_CAL(pAsm, + 0, + pPresubDesc->subIL_Shift, + pCompiledSub->NumInstructions, + pCompiledSub->Instructions, + pPresubDesc); + } + + if(GL_FALSE == bRet) + { + radeon_error("Shader presub assemble failed. \n"); + } + else + { + /* copy presub output to real dst */ + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + pAsm->D.dst.opcode = SQ_OP2_INST_MOV; + + if( GL_FALSE == assemble_dst(pAsm) ) + { + return GL_FALSE; + } + + setaddrmode_PVSSRC(&(pAsm->S[0].src), ADDR_ABSOLUTE); + pAsm->S[0].src.rtype = SRC_REG_TEMPORARY; + pAsm->S[0].src.reg = pCompiledSub->dstRegIndex; + pAsm->S[0].src.swizzlex = pCompiledSub->outputSwizzleX; + pAsm->S[0].src.swizzley = pCompiledSub->outputSwizzleY; + pAsm->S[0].src.swizzlez = pCompiledSub->outputSwizzleZ; + pAsm->S[0].src.swizzlew = pCompiledSub->outputSwizzleW; + + next_ins(pAsm); + + pAsm->callers[pAsm->unCallerArrayPointer - 1].finale_cf_ptr = pAsm->cf_current_alu_clause_ptr; + pAsm->callers[pAsm->unCallerArrayPointer - 1].prelude_cf_ptr = prelude_cf_ptr; + pAsm->alu_x_opcode = SQ_CF_INST_ALU; + } + + if( (pPresubDesc->number_used_registers + pAsm->uFirstHelpReg) > pAsm->number_used_registers ) + { + pAsm->number_used_registers = pPresubDesc->number_used_registers + pAsm->uFirstHelpReg; + } + if(pAsm->uFirstHelpReg > pPresubDesc->maxStartReg) + { + pPresubDesc->maxStartReg = pAsm->uFirstHelpReg; + } + + return bRet; +} + GLboolean Process_Export(r700_AssemblerBase* pAsm, GLuint type, GLuint export_starting_index, @@ -3940,6 +6328,7 @@ GLboolean Process_Fragment_Exports(r700_AssemblerBase *pR700AsmCode, GLbitfield OutputsWritten) { unsigned int unBit; + GLuint export_count = 0; if(pR700AsmCode->depth_export_register_number >= 0) { @@ -3961,6 +6350,7 @@ GLboolean Process_Fragment_Exports(r700_AssemblerBase *pR700AsmCode, { return GL_FALSE; } + export_count++; } unBit = 1 << FRAG_RESULT_DEPTH; if(OutputsWritten & unBit) @@ -3974,8 +6364,15 @@ GLboolean Process_Fragment_Exports(r700_AssemblerBase *pR700AsmCode, { return GL_FALSE; } + export_count++; } - + /* Need to export something, otherwise we'll hang + * results are undefined anyway */ + if(export_count == 0) + { + Process_Export(pR700AsmCode, SQ_EXPORT_PIXEL, 0, 1, 0, GL_FALSE); + } + if(pR700AsmCode->cf_last_export_ptr != NULL) { pR700AsmCode->cf_last_export_ptr->m_Word1.f.cf_inst = SQ_CF_INST_EXPORT_DONE; @@ -4080,6 +6477,25 @@ GLboolean Process_Vertex_Exports(r700_AssemblerBase *pR700AsmCode, export_starting_index++; } } + + for(i=VERT_RESULT_VAR0; i<VERT_RESULT_MAX; i++) + { + unBit = 1 << i; + if(OutputsWritten & unBit) + { + if( GL_FALSE == Process_Export(pR700AsmCode, + SQ_EXPORT_PARAM, + export_starting_index, + 1, + pR700AsmCode->ucVP_OutputMap[i], + GL_FALSE) ) + { + return GL_FALSE; + } + + export_starting_index++; + } + } // At least one param should be exported if (export_count) @@ -4114,6 +6530,21 @@ GLboolean Clean_Up_Assembler(r700_AssemblerBase *pR700AsmCode) { FREE(pR700AsmCode->pucOutMask); FREE(pR700AsmCode->pInstDeps); + + if(NULL != pR700AsmCode->subs) + { + FREE(pR700AsmCode->subs); + } + if(NULL != pR700AsmCode->callers) + { + FREE(pR700AsmCode->callers); + } + + if(NULL != pR700AsmCode->presubs) + { + FREE(pR700AsmCode->presubs); + } + return GL_TRUE; } diff --git a/src/mesa/drivers/dri/r600/r700_assembler.h b/src/mesa/drivers/dri/r600/r700_assembler.h index f9c4d849c65..6ef945dfda3 100644 --- a/src/mesa/drivers/dri/r600/r700_assembler.h +++ b/src/mesa/drivers/dri/r600/r700_assembler.h @@ -34,6 +34,45 @@ #include "r700_shaderinst.h" #include "r700_shader.h" +typedef enum LOADABLE_SCRIPT_SIGNITURE +{ + GLSL_NOISE1 = 0x10000001, + GLSL_NOISE2 = 0x10000002, + GLSL_NOISE3 = 0x10000003, + GLSL_NOISE4 = 0x10000004 +}LOADABLE_SCRIPT_SIGNITURE; + +typedef struct COMPILED_SUB +{ + struct prog_instruction *Instructions; + GLuint NumInstructions; + GLuint NumTemporaries; + GLuint NumParameters; + GLuint MinRegIndex; + GLfloat (*ParameterValues)[4]; + GLbyte outputSwizzleX; + GLbyte outputSwizzleY; + GLbyte outputSwizzleZ; + GLbyte outputSwizzleW; + GLshort srcRegIndex[3]; + GLushort dstRegIndex; +}COMPILED_SUB; + +typedef struct PRESUB_DESCtag +{ + LOADABLE_SCRIPT_SIGNITURE sptSigniture; + GLint subIL_Shift; + struct prog_src_register InReg[3]; + struct prog_dst_register OutReg; + + GLushort maxStartReg; + GLushort number_used_registers; + + GLuint unConstantsStart; + + COMPILED_SUB * pCompiledSub; +} PRESUB_DESC; + typedef enum SHADER_PIPE_TYPE { SPT_VP = 0, @@ -72,7 +111,8 @@ typedef enum SrcRegisterType SRC_REG_INPUT = 1, SRC_REG_CONSTANT = 2, SRC_REG_ALT_TEMPORARY = 3, - NUMBER_OF_SRC_REG_TYPE = 4 + SRC_REC_LITERAL = 4, + NUMBER_OF_SRC_REG_TYPE = 5 } SrcRegisterType; typedef enum DstRegisterType @@ -111,16 +151,24 @@ typedef struct PVSDSTtag BITS addrmode1:1; //32 } PVSDST; +typedef struct PVSINSTtag +{ + BITS literal_slots :2; + BITS SaturateMode :2; + BITS index_mode :3; +} PVSINST; + typedef struct PVSSRCtag { - BITS rtype:4; + BITS rtype:3; BITS addrmode0:1; - BITS reg:10; //15 (8) + BITS reg:10; //14 (8) BITS swizzlex:3; BITS swizzley:3; BITS swizzlez:3; - BITS swizzlew:3; //27 + BITS swizzlew:3; //26 + BITS abs:1; BITS negx:1; BITS negy:1; BITS negz:1; @@ -148,6 +196,7 @@ typedef union PVSDWORDtag { BITS bits; PVSDST dst; + PVSINST dst2; PVSSRC src; PVSMATH math; float f; @@ -251,6 +300,8 @@ enum FC_IF = 1, FC_LOOP = 2, FC_REP = 3, + FC_PUSH_VPM = 4, + FC_PUSH_WQM = 5, COND_NONE = 0, COND_BOOL = 1, @@ -263,22 +314,56 @@ enum typedef struct FC_LEVEL { - unsigned int first; ///< first fc instruction on level (if, rep, loop) - unsigned int* mid; ///< middle instructions - else or all breaks on this level - unsigned int midLen; - unsigned int type; - unsigned int cond; - unsigned int inv; - unsigned int bpush; ///< 1 if first instruction does branch stack push - int id; ///< id of bool or int variable + R700ControlFlowGenericClause * first; + R700ControlFlowGenericClause ** mid; + unsigned int unNumMid; + unsigned int midLen; + unsigned int type; + unsigned int cond; + unsigned int inv; + int id; ///< id of bool or int variable } FC_LEVEL; typedef struct VTX_FETCH_METHOD { - GLboolean bEnableMini; - GLuint mega_fetch_remainder; + GLboolean bEnableMini; + GLuint mega_fetch_remainder; } VTX_FETCH_METHOD; +typedef struct SUB_OFFSET +{ + GLint subIL_Offset; + GLuint unCFoffset; + GLuint unStackDepthMax; + PRESUB_DESC * pPresubDesc; + TypedShaderList lstCFInstructions_local; +} SUB_OFFSET; + +typedef struct CALLER_POINTER +{ + GLint subIL_Offset; + GLint subDescIndex; + R700ControlFlowGenericClause* cf_ptr; + + R700ControlFlowGenericClause* prelude_cf_ptr; + R700ControlFlowGenericClause* finale_cf_ptr; +} CALLER_POINTER; + +#define SQ_MAX_CALL_DEPTH 0x00000020 + +typedef struct CALL_LEVEL +{ + unsigned int FCSP_BeforeEntry; + GLint subDescIndex; + GLushort current; + GLushort max; + TypedShaderList * plstCFInstructions_local; +} CALL_LEVEL; + +#define HAS_CURRENT_LOOPRET 0x1L +#define HAS_LOOPRET 0x2L +#define LOOPRET_FLAGS HAS_LOOPRET | HAS_CURRENT_LOOPRET + typedef struct r700_AssemblerBase { R700ControlFlowSXClause* cf_last_export_ptr; @@ -294,14 +379,19 @@ typedef struct r700_AssemblerBase // No clause has been created yet CF_CLAUSE_TYPE cf_current_clause_type; + BITS alu_x_opcode; + GLuint number_of_exports; GLuint number_of_colorandz_exports; GLuint number_of_export_opcodes; PVSDWORD D; + PVSDWORD D2; PVSDWORD S[3]; + PVSDWORD C[4]; unsigned int uLastPosUpdate; + unsigned int last_cond_register; OUT_FRAGMENT_FMT_0 fp_stOutFmt0; @@ -310,6 +400,8 @@ typedef struct r700_AssemblerBase unsigned int number_used_registers; unsigned int uUsedConsts; + unsigned int flag_reg_index; + // Fragment programs unsigned int uiFP_AttributeMap[FRAG_ATTRIB_MAX]; unsigned int uiFP_OutputMap[FRAG_RESULT_MAX]; @@ -330,9 +422,6 @@ typedef struct r700_AssemblerBase unsigned int FCSP; FC_LEVEL fc_stack[32]; - unsigned int branch_depth; - unsigned int max_branch_depth; - //----------------------------------------------------------------------------------- // ArgSubst used in Assemble_Source() function //----------------------------------------------------------------------------------- @@ -373,7 +462,29 @@ typedef struct r700_AssemblerBase SHADER_PIPE_TYPE currentShaderType; struct prog_instruction * pILInst; GLuint uiCurInst; + GLubyte SamplerUnits[MAX_SAMPLERS]; GLboolean bR6xx; + /* helper to decide which type of instruction to assemble */ + GLboolean is_tex; + /* we inserted helper intructions and need barrier on next TEX ins */ + GLboolean need_tex_barrier; + + SUB_OFFSET * subs; + GLuint unSubArraySize; + GLuint unSubArrayPointer; + CALLER_POINTER * callers; + GLuint unCallerArraySize; + GLuint unCallerArrayPointer; + unsigned int CALLSP; + CALL_LEVEL CALLSTACK[SQ_MAX_CALL_DEPTH]; + + GLuint unCFflags; + + PRESUB_DESC * presubs; + GLuint unPresubArraySize; + GLuint unNumPresub; + GLuint unCurNumILInsts; + } r700_AssemblerBase; //Internal use @@ -395,7 +506,7 @@ BITS is_depth_component_exported(OUT_FRAGMENT_FMT_0* pFPOutFmt) ; GLboolean is_reduction_opcode(PVSDWORD * dest); GLuint GetSurfaceFormat(GLenum eType, GLuint nChannels, GLuint * pClient_size); -unsigned int r700GetNumOperands(r700_AssemblerBase* pAsm); +unsigned int r700GetNumOperands(GLuint opcode, GLuint nIsOp3); GLboolean IsTex(gl_inst_opcode Opcode); GLboolean IsAlu(gl_inst_opcode Opcode); @@ -411,6 +522,15 @@ GLboolean assemble_vfetch_instruction(r700_AssemblerBase* pAsm, GLuint number_of_elements, GLenum dataElementType, VTX_FETCH_METHOD* pFetchMethod); +GLboolean assemble_vfetch_instruction2(r700_AssemblerBase* pAsm, + GLuint destination_register, + GLenum type, + GLint size, + GLubyte element, + GLuint _signed, + GLboolean normalize, + VTX_FETCH_METHOD * pFetchMethod); +GLboolean cleanup_vfetch_instructions(r700_AssemblerBase* pAsm); GLuint gethelpr(r700_AssemblerBase* pAsm); void resethelpr(r700_AssemblerBase* pAsm); void checkop_init(r700_AssemblerBase* pAsm); @@ -433,6 +553,10 @@ GLboolean assemble_alu_src(R700ALUInstruction* alu_instruction_ptr, GLboolean add_alu_instruction(r700_AssemblerBase* pAsm, R700ALUInstruction* alu_instruction_ptr, GLuint contiguous_slots_needed); + +GLboolean add_cf_instruction(r700_AssemblerBase* pAsm); +void add_return_inst(r700_AssemblerBase *pAsm); + void get_src_properties(R700ALUInstruction* alu_instruction_ptr, int source_index, BITS* psrc_sel, @@ -454,21 +578,31 @@ GLboolean check_vector(r700_AssemblerBase* pAsm, R700ALUInstruction* alu_instruction_ptr); GLboolean assemble_alu_instruction(r700_AssemblerBase *pAsm); GLboolean next_ins(r700_AssemblerBase *pAsm); + +GLboolean pops(r700_AssemblerBase *pAsm, GLuint pops); +GLboolean jumpToOffest(r700_AssemblerBase *pAsm, GLuint pops, GLint offset); +GLboolean setRetInLoopFlag(r700_AssemblerBase *pAsm, GLuint flagValue); +GLboolean testFlag(r700_AssemblerBase *pAsm); +GLboolean breakLoopOnFlag(r700_AssemblerBase *pAsm, GLuint unFCSP); +GLboolean returnOnFlag(r700_AssemblerBase *pAsm, GLuint unIF); + GLboolean assemble_math_function(r700_AssemblerBase* pAsm, BITS opcode); GLboolean assemble_ABS(r700_AssemblerBase *pAsm); GLboolean assemble_ADD(r700_AssemblerBase *pAsm); +GLboolean assemble_ARL(r700_AssemblerBase *pAsm); GLboolean assemble_BAD(char *opcode_str); GLboolean assemble_CMP(r700_AssemblerBase *pAsm); -GLboolean assemble_COS(r700_AssemblerBase *pAsm); GLboolean assemble_DOT(r700_AssemblerBase *pAsm); GLboolean assemble_DST(r700_AssemblerBase *pAsm); GLboolean assemble_EX2(r700_AssemblerBase *pAsm); +GLboolean assemble_EXP(r700_AssemblerBase *pAsm); GLboolean assemble_FLR(r700_AssemblerBase *pAsm); GLboolean assemble_FLR_INT(r700_AssemblerBase *pAsm); GLboolean assemble_FRC(r700_AssemblerBase *pAsm); -GLboolean assemble_KIL(r700_AssemblerBase *pAsm); +GLboolean assemble_KIL(r700_AssemblerBase *pAsm, GLuint opcode); GLboolean assemble_LG2(r700_AssemblerBase *pAsm); GLboolean assemble_LRP(r700_AssemblerBase *pAsm); +GLboolean assemble_LOG(r700_AssemblerBase *pAsm); GLboolean assemble_MAD(r700_AssemblerBase *pAsm); GLboolean assemble_LIT(r700_AssemblerBase *pAsm); GLboolean assemble_MAX(r700_AssemblerBase *pAsm); @@ -478,17 +612,37 @@ GLboolean assemble_MUL(r700_AssemblerBase *pAsm); GLboolean assemble_POW(r700_AssemblerBase *pAsm); GLboolean assemble_RCP(r700_AssemblerBase *pAsm); GLboolean assemble_RSQ(r700_AssemblerBase *pAsm); -GLboolean assemble_SIN(r700_AssemblerBase *pAsm); GLboolean assemble_SCS(r700_AssemblerBase *pAsm); GLboolean assemble_SGE(r700_AssemblerBase *pAsm); + +GLboolean assemble_LOGIC(r700_AssemblerBase *pAsm, BITS opcode); +GLboolean assemble_LOGIC_PRED(r700_AssemblerBase *pAsm, BITS opcode); +GLboolean assemble_TRIG(r700_AssemblerBase *pAsm, BITS opcode); + GLboolean assemble_SLT(r700_AssemblerBase *pAsm); GLboolean assemble_STP(r700_AssemblerBase *pAsm); GLboolean assemble_TEX(r700_AssemblerBase *pAsm); GLboolean assemble_XPD(r700_AssemblerBase *pAsm); GLboolean assemble_EXPORT(r700_AssemblerBase *pAsm); -GLboolean assemble_IF(r700_AssemblerBase *pAsm); +GLboolean assemble_IF(r700_AssemblerBase *pAsm, GLboolean bHasElse); +GLboolean assemble_ELSE(r700_AssemblerBase *pAsm); GLboolean assemble_ENDIF(r700_AssemblerBase *pAsm); +GLboolean assemble_BGNLOOP(r700_AssemblerBase *pAsm); +GLboolean assemble_BRK(r700_AssemblerBase *pAsm); +GLboolean assemble_COND(r700_AssemblerBase *pAsm); +GLboolean assemble_ENDLOOP(r700_AssemblerBase *pAsm); + +GLboolean assemble_BGNSUB(r700_AssemblerBase *pAsm, GLint nILindex, GLuint uiIL_Shift); +GLboolean assemble_ENDSUB(r700_AssemblerBase *pAsm); +GLboolean assemble_RET(r700_AssemblerBase *pAsm); +GLboolean assemble_CAL(r700_AssemblerBase *pAsm, + GLint nILindex, + GLuint uiIL_Offest, + GLuint uiNumberInsts, + struct prog_instruction *pILInst, + PRESUB_DESC * pPresubDesc); + GLboolean Process_Export(r700_AssemblerBase* pAsm, GLuint type, GLuint export_starting_index, @@ -498,14 +652,25 @@ GLboolean Process_Export(r700_AssemblerBase* pAsm, GLboolean Move_Depth_Exports_To_Correct_Channels(r700_AssemblerBase *pAsm, BITS depth_channel_select); +GLboolean callPreSub(r700_AssemblerBase* pAsm, + LOADABLE_SCRIPT_SIGNITURE scriptSigniture, + /* struct prog_instruction ** pILInstParent, */ + COMPILED_SUB * pCompiledSub, + GLshort uOutReg, + GLshort uNumValidSrc); //Interface -GLboolean AssembleInstr(GLuint uiNumberInsts, +GLboolean AssembleInstr(GLuint uiFirstInst, + GLuint uiIL_Shift, + GLuint uiNumberInsts, struct prog_instruction *pILInst, r700_AssemblerBase *pR700AsmCode); GLboolean Process_Fragment_Exports(r700_AssemblerBase *pR700AsmCode, GLbitfield OutputsWritten); GLboolean Process_Vertex_Exports(r700_AssemblerBase *pR700AsmCode, GLbitfield OutputsWritten); +GLboolean RelocProgram(r700_AssemblerBase * pAsm, struct gl_program * pILProg); +GLboolean InitShaderProgram(r700_AssemblerBase * pAsm); + int Init_r700_AssemblerBase(SHADER_PIPE_TYPE spt, r700_AssemblerBase* pAsm, R700_Shader* pShader); GLboolean Clean_Up_Assembler(r700_AssemblerBase *pR700AsmCode); diff --git a/src/mesa/drivers/dri/r600/r700_chip.c b/src/mesa/drivers/dri/r600/r700_chip.c index 06d7e9c9ab1..ee2a0a4c8a6 100644 --- a/src/mesa/drivers/dri/r600/r700_chip.c +++ b/src/mesa/drivers/dri/r600/r700_chip.c @@ -54,11 +54,15 @@ static void r700SendTexState(GLcontext *ctx, struct radeon_state_atom *atom) for (i = 0; i < R700_TEXTURE_NUMBERUNITS; i++) { if (ctx->Texture.Unit[i]._ReallyEnabled) { radeonTexObj *t = r700->textures[i]; + uint32_t offset; if (t) { - if (!t->image_override) + if (!t->image_override) { bo = t->mt->bo; - else + offset = get_base_teximage_offset(t); + } else { bo = t->bo; + offset = 0; + } if (bo) { r700SyncSurf(context, bo, @@ -77,7 +81,7 @@ static void r700SendTexState(GLcontext *ctx, struct radeon_state_atom *atom) R600_OUT_BATCH(r700->textures[i]->SQ_TEX_RESOURCE6); R600_OUT_BATCH_RELOC(r700->textures[i]->SQ_TEX_RESOURCE2, bo, - 0, + offset, RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0); R600_OUT_BATCH_RELOC(r700->textures[i]->SQ_TEX_RESOURCE3, bo, @@ -141,17 +145,15 @@ static void r700SendTexBorderColorState(GLcontext *ctx, struct radeon_state_atom } } +extern int getTypeSize(GLenum type); static void r700SetupVTXConstants(GLcontext * ctx, - unsigned int nStreamID, void * pAos, - unsigned int size, /* number of elements in vector */ - unsigned int stride, - unsigned int count) /* number of vectors in stream */ + StreamDesc * pStreamDesc) { context_t *context = R700_CONTEXT(ctx); struct radeon_aos * paos = (struct radeon_aos *)pAos; + unsigned int nVBsize; BATCH_LOCALS(&context->radeon); - radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); unsigned int uSQ_VTX_CONSTANT_WORD0_0; unsigned int uSQ_VTX_CONSTANT_WORD1_0; @@ -171,18 +173,40 @@ static void r700SetupVTXConstants(GLcontext * ctx, else r700SyncSurf(context, paos->bo, RADEON_GEM_DOMAIN_GTT, 0, VC_ACTION_ENA_bit); + if(0 == pStreamDesc->stride) + { + nVBsize = paos->count * pStreamDesc->size * getTypeSize(pStreamDesc->type); + } + else + { + nVBsize = paos->count * pStreamDesc->stride; + } + uSQ_VTX_CONSTANT_WORD0_0 = paos->offset; - uSQ_VTX_CONSTANT_WORD1_0 = count * (size * 4) - 1; + uSQ_VTX_CONSTANT_WORD1_0 = nVBsize - 1; SETfield(uSQ_VTX_CONSTANT_WORD2_0, 0, BASE_ADDRESS_HI_shift, BASE_ADDRESS_HI_mask); /* TODO */ - SETfield(uSQ_VTX_CONSTANT_WORD2_0, stride, SQ_VTX_CONSTANT_WORD2_0__STRIDE_shift, + SETfield(uSQ_VTX_CONSTANT_WORD2_0, pStreamDesc->stride, SQ_VTX_CONSTANT_WORD2_0__STRIDE_shift, SQ_VTX_CONSTANT_WORD2_0__STRIDE_mask); - SETfield(uSQ_VTX_CONSTANT_WORD2_0, GetSurfaceFormat(GL_FLOAT, size, NULL), + SETfield(uSQ_VTX_CONSTANT_WORD2_0, GetSurfaceFormat(pStreamDesc->type, pStreamDesc->size, NULL), SQ_VTX_CONSTANT_WORD2_0__DATA_FORMAT_shift, SQ_VTX_CONSTANT_WORD2_0__DATA_FORMAT_mask); /* TODO : trace back api for initial data type, not only GL_FLOAT */ - SETfield(uSQ_VTX_CONSTANT_WORD2_0, SQ_NUM_FORMAT_SCALED, - SQ_VTX_CONSTANT_WORD2_0__NUM_FORMAT_ALL_shift, SQ_VTX_CONSTANT_WORD2_0__NUM_FORMAT_ALL_mask); - SETbit(uSQ_VTX_CONSTANT_WORD2_0, SQ_VTX_CONSTANT_WORD2_0__FORMAT_COMP_ALL_bit); + + if(GL_TRUE == pStreamDesc->normalize) + { + SETfield(uSQ_VTX_CONSTANT_WORD2_0, SQ_NUM_FORMAT_NORM, + SQ_VTX_CONSTANT_WORD2_0__NUM_FORMAT_ALL_shift, SQ_VTX_CONSTANT_WORD2_0__NUM_FORMAT_ALL_mask); + } + //else + //{ + // SETfield(uSQ_VTX_CONSTANT_WORD2_0, SQ_NUM_FORMAT_INT, + // SQ_VTX_CONSTANT_WORD2_0__NUM_FORMAT_ALL_shift, SQ_VTX_CONSTANT_WORD2_0__NUM_FORMAT_ALL_mask); + //} + + if(1 == pStreamDesc->_signed) + { + SETbit(uSQ_VTX_CONSTANT_WORD2_0, SQ_VTX_CONSTANT_WORD2_0__FORMAT_COMP_ALL_bit); + } SETfield(uSQ_VTX_CONSTANT_WORD3_0, 1, MEM_REQUEST_SIZE_shift, MEM_REQUEST_SIZE_mask); SETfield(uSQ_VTX_CONSTANT_WORD6_0, SQ_TEX_VTX_VALID_BUFFER, @@ -191,7 +215,7 @@ static void r700SetupVTXConstants(GLcontext * ctx, BEGIN_BATCH_NO_AUTOSTATE(9 + 2); R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_RESOURCE, 7)); - R600_OUT_BATCH((nStreamID + SQ_FETCH_RESOURCE_VS_OFFSET) * FETCH_RESOURCE_STRIDE); + R600_OUT_BATCH((pStreamDesc->element + SQ_FETCH_RESOURCE_VS_OFFSET) * FETCH_RESOURCE_STRIDE); R600_OUT_BATCH(uSQ_VTX_CONSTANT_WORD0_0); R600_OUT_BATCH(uSQ_VTX_CONSTANT_WORD1_0); R600_OUT_BATCH(uSQ_VTX_CONSTANT_WORD2_0); @@ -208,31 +232,6 @@ static void r700SetupVTXConstants(GLcontext * ctx, } -void r700SetupStreams(GLcontext *ctx) -{ - context_t *context = R700_CONTEXT(ctx); - struct r700_vertex_program *vp = context->selected_vp; - TNLcontext *tnl = TNL_CONTEXT(ctx); - struct vertex_buffer *vb = &tnl->vb; - unsigned int i, j = 0; - radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); - - R600_STATECHANGE(context, vtx); - - for(i=0; i<VERT_ATTRIB_MAX; i++) { - if(vp->mesa_program->Base.InputsRead & (1 << i)) { - rcommon_emit_vector(ctx, - &context->radeon.tcl.aos[j], - vb->AttribPtr[i]->data, - vb->AttribPtr[i]->size, - vb->AttribPtr[i]->stride, - vb->Count); - j++; - } - } - context->radeon.tcl.aos_count = j; -} - static void r700SendVTXState(GLcontext *ctx, struct radeon_state_atom *atom) { context_t *context = R700_CONTEXT(ctx); @@ -256,15 +255,12 @@ static void r700SendVTXState(GLcontext *ctx, struct radeon_state_atom *atom) COMMIT_BATCH(); for(i=0; i<VERT_ATTRIB_MAX; i++) { - if(vp->mesa_program->Base.InputsRead & (1 << i)) { - /* currently aos are packed */ - r700SetupVTXConstants(ctx, - i, - (void*)(&context->radeon.tcl.aos[j]), - (unsigned int)context->radeon.tcl.aos[j].components, - (unsigned int)context->radeon.tcl.aos[j].stride * 4, - (unsigned int)context->radeon.tcl.aos[j].count); - j++; + if(vp->mesa_program->Base.InputsRead & (1 << i)) + { + r700SetupVTXConstants(ctx, + (void*)(&context->radeon.tcl.aos[j]), + &(context->stream_desc[j])); + j++; } } } @@ -366,7 +362,6 @@ static void r700SendDepthTargetState(GLcontext *ctx, struct radeon_state_atom *a rrb = radeon_get_depthbuffer(&context->radeon); if (!rrb || !rrb->bo) { - fprintf(stderr, "no rrb\n"); return; } @@ -408,7 +403,6 @@ static void r700SendRenderTargetState(GLcontext *ctx, struct radeon_state_atom * rrb = radeon_get_colorbuffer(&context->radeon); if (!rrb || !rrb->bo) { - fprintf(stderr, "no rrb\n"); return; } @@ -452,68 +446,77 @@ static void r700SendRenderTargetState(GLcontext *ctx, struct radeon_state_atom * static void r700SendPSState(GLcontext *ctx, struct radeon_state_atom *atom) { - context_t *context = R700_CONTEXT(ctx); - R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); - struct radeon_bo * pbo; - BATCH_LOCALS(&context->radeon); - radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); + context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); + struct radeon_bo * pbo; + BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); - pbo = (struct radeon_bo *)r700GetActiveFpShaderBo(GL_CONTEXT(context)); + pbo = (struct radeon_bo *)r700GetActiveFpShaderBo(GL_CONTEXT(context)); - if (!pbo) - return; + if (!pbo) + return; - r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit); + r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit); - BEGIN_BATCH_NO_AUTOSTATE(3 + 2); - R600_OUT_BATCH_REGSEQ(SQ_PGM_START_PS, 1); - R600_OUT_BATCH(r700->ps.SQ_PGM_START_PS.u32All); - R600_OUT_BATCH_RELOC(r700->ps.SQ_PGM_START_PS.u32All, - pbo, - r700->ps.SQ_PGM_START_PS.u32All, - RADEON_GEM_DOMAIN_GTT, 0, 0); - END_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(3 + 2); + R600_OUT_BATCH_REGSEQ(SQ_PGM_START_PS, 1); + R600_OUT_BATCH(r700->ps.SQ_PGM_START_PS.u32All); + R600_OUT_BATCH_RELOC(r700->ps.SQ_PGM_START_PS.u32All, + pbo, + r700->ps.SQ_PGM_START_PS.u32All, + RADEON_GEM_DOMAIN_GTT, 0, 0); + END_BATCH(); - BEGIN_BATCH_NO_AUTOSTATE(9); - R600_OUT_BATCH_REGVAL(SQ_PGM_RESOURCES_PS, r700->ps.SQ_PGM_RESOURCES_PS.u32All); - R600_OUT_BATCH_REGVAL(SQ_PGM_EXPORTS_PS, r700->ps.SQ_PGM_EXPORTS_PS.u32All); - R600_OUT_BATCH_REGVAL(SQ_PGM_CF_OFFSET_PS, r700->ps.SQ_PGM_CF_OFFSET_PS.u32All); - END_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(9); + R600_OUT_BATCH_REGVAL(SQ_PGM_RESOURCES_PS, r700->ps.SQ_PGM_RESOURCES_PS.u32All); + R600_OUT_BATCH_REGVAL(SQ_PGM_EXPORTS_PS, r700->ps.SQ_PGM_EXPORTS_PS.u32All); + R600_OUT_BATCH_REGVAL(SQ_PGM_CF_OFFSET_PS, r700->ps.SQ_PGM_CF_OFFSET_PS.u32All); + END_BATCH(); - COMMIT_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(3); + R600_OUT_BATCH_REGVAL(SQ_LOOP_CONST_0, 0x01000FFF); + END_BATCH(); + + COMMIT_BATCH(); } static void r700SendVSState(GLcontext *ctx, struct radeon_state_atom *atom) { - context_t *context = R700_CONTEXT(ctx); - R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); - struct radeon_bo * pbo; - BATCH_LOCALS(&context->radeon); - radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); + context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context); + struct radeon_bo * pbo; + BATCH_LOCALS(&context->radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); - pbo = (struct radeon_bo *)r700GetActiveVpShaderBo(GL_CONTEXT(context)); + pbo = (struct radeon_bo *)r700GetActiveVpShaderBo(GL_CONTEXT(context)); - if (!pbo) - return; + if (!pbo) + return; - r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit); + r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit); - BEGIN_BATCH_NO_AUTOSTATE(3 + 2); - R600_OUT_BATCH_REGSEQ(SQ_PGM_START_VS, 1); - R600_OUT_BATCH(r700->vs.SQ_PGM_START_VS.u32All); - R600_OUT_BATCH_RELOC(r700->vs.SQ_PGM_START_VS.u32All, - pbo, - r700->vs.SQ_PGM_START_VS.u32All, - RADEON_GEM_DOMAIN_GTT, 0, 0); - END_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(3 + 2); + R600_OUT_BATCH_REGSEQ(SQ_PGM_START_VS, 1); + R600_OUT_BATCH(r700->vs.SQ_PGM_START_VS.u32All); + R600_OUT_BATCH_RELOC(r700->vs.SQ_PGM_START_VS.u32All, + pbo, + r700->vs.SQ_PGM_START_VS.u32All, + RADEON_GEM_DOMAIN_GTT, 0, 0); + END_BATCH(); - BEGIN_BATCH_NO_AUTOSTATE(6); - R600_OUT_BATCH_REGVAL(SQ_PGM_RESOURCES_VS, r700->vs.SQ_PGM_RESOURCES_VS.u32All); - R600_OUT_BATCH_REGVAL(SQ_PGM_CF_OFFSET_VS, r700->vs.SQ_PGM_CF_OFFSET_VS.u32All); - END_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(6); + R600_OUT_BATCH_REGVAL(SQ_PGM_RESOURCES_VS, r700->vs.SQ_PGM_RESOURCES_VS.u32All); + R600_OUT_BATCH_REGVAL(SQ_PGM_CF_OFFSET_VS, r700->vs.SQ_PGM_CF_OFFSET_VS.u32All); + END_BATCH(); - COMMIT_BATCH(); + BEGIN_BATCH_NO_AUTOSTATE(3); + R600_OUT_BATCH_REGVAL((SQ_LOOP_CONST_0 + 32*4), 0x0100000F); + //R600_OUT_BATCH_REGVAL((SQ_LOOP_CONST_0 + (SQ_LOOP_CONST_vs<2)), 0x0100000F); + END_BATCH(); + + COMMIT_BATCH(); } static void r700SendFSState(GLcontext *ctx, struct radeon_state_atom *atom) @@ -794,8 +797,7 @@ static void r700SendDBState(GLcontext *ctx, struct radeon_state_atom *atom) BATCH_LOCALS(&context->radeon); radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); - BEGIN_BATCH_NO_AUTOSTATE(23); - R600_OUT_BATCH_REGVAL(DB_HTILE_DATA_BASE, r700->DB_HTILE_DATA_BASE.u32All); + BEGIN_BATCH_NO_AUTOSTATE(17); R600_OUT_BATCH_REGSEQ(DB_STENCIL_CLEAR, 2); R600_OUT_BATCH(r700->DB_STENCIL_CLEAR.u32All); @@ -808,7 +810,6 @@ static void r700SendDBState(GLcontext *ctx, struct radeon_state_atom *atom) R600_OUT_BATCH(r700->DB_RENDER_CONTROL.u32All); R600_OUT_BATCH(r700->DB_RENDER_OVERRIDE.u32All); - R600_OUT_BATCH_REGVAL(DB_HTILE_SURFACE, r700->DB_HTILE_SURFACE.u32All); R600_OUT_BATCH_REGVAL(DB_ALPHA_TO_MASK, r700->DB_ALPHA_TO_MASK.u32All); END_BATCH(); @@ -1108,6 +1109,32 @@ static void r700SendVSConsts(GLcontext *ctx, struct radeon_state_atom *atom) COMMIT_BATCH(); } +static void r700SendQueryBegin(GLcontext *ctx, struct radeon_state_atom *atom) +{ + radeonContextPtr radeon = RADEON_CONTEXT(ctx); + struct radeon_query_object *query = radeon->query.current; + BATCH_LOCALS(radeon); + radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s\n", __func__); + + /* clear the buffer */ + radeon_bo_map(query->bo, GL_FALSE); + memset(query->bo->ptr, 0, 4 * 2 * sizeof(uint64_t)); /* 4 DBs, 2 qwords each */ + radeon_bo_unmap(query->bo); + + radeon_cs_space_check_with_bo(radeon->cmdbuf.cs, + query->bo, + 0, RADEON_GEM_DOMAIN_GTT); + + BEGIN_BATCH_NO_AUTOSTATE(4 + 2); + R600_OUT_BATCH(CP_PACKET3(R600_IT_EVENT_WRITE, 2)); + R600_OUT_BATCH(ZPASS_DONE); + R600_OUT_BATCH(query->curr_offset); /* hw writes qwords */ + R600_OUT_BATCH(0x00000000); + R600_OUT_BATCH_RELOC(VGT_EVENT_INITIATOR, query->bo, 0, 0, RADEON_GEM_DOMAIN_GTT, 0); + END_BATCH(); + query->emitted_begin = GL_TRUE; +} + static int check_always(GLcontext *ctx, struct radeon_state_atom *atom) { return atom->cmd_size; @@ -1136,7 +1163,11 @@ static int check_blnd(GLcontext *ctx, struct radeon_state_atom *atom) count += 3; if (context->radeon.radeonScreen->chip_family > CHIP_FAMILY_R600) { - for (ui = 0; ui < R700_MAX_RENDER_TARGETS; ui++) { + /* targets are enabled in r700SetRenderTarget but state + size is calculated before that. Until MRT's are done + hardcode target0 as enabled. */ + count += 3; + for (ui = 1; ui < R700_MAX_RENDER_TARGETS; ui++) { if (r700->render_target[ui].enabled) count += 3; } @@ -1216,6 +1247,20 @@ static int check_vs_consts(GLcontext *ctx, struct radeon_state_atom *atom) return count; } +static int check_queryobj(GLcontext *ctx, struct radeon_state_atom *atom) +{ + radeonContextPtr radeon = RADEON_CONTEXT(ctx); + struct radeon_query_object *query = radeon->query.current; + int count; + + if (!query || query->emitted_begin) + count = 0; + else + count = atom->cmd_size; + radeon_print(RADEON_STATE, RADEON_TRACE, "%s %d\n", __func__, count); + return count; +} + #define ALLOC_STATE( ATOM, CHK, SZ, EMIT ) \ do { \ context->atoms.ATOM.cmd_size = (SZ); \ @@ -1229,6 +1274,19 @@ do { \ insert_at_tail(&context->radeon.hw.atomlist, &context->atoms.ATOM); \ } while (0) +static void r600_init_query_stateobj(radeonContextPtr radeon, int SZ) +{ + radeon->query.queryobj.cmd_size = (SZ); + radeon->query.queryobj.cmd = NULL; + radeon->query.queryobj.name = "queryobj"; + radeon->query.queryobj.idx = 0; + radeon->query.queryobj.check = check_queryobj; + radeon->query.queryobj.dirty = GL_FALSE; + radeon->query.queryobj.emit = r700SendQueryBegin; + radeon->hw.max_state_size += (SZ); + insert_at_tail(&radeon->hw.atomlist, &radeon->query.queryobj); +} + void r600InitAtoms(context_t *context) { radeon_print(RADEON_STATE, RADEON_NORMAL, "%s %p\n", __func__, context); @@ -1239,7 +1297,7 @@ void r600InitAtoms(context_t *context) context->radeon.hw.atomlist.name = "atom-list"; ALLOC_STATE(sq, always, 34, r700SendSQConfig); - ALLOC_STATE(db, always, 23, r700SendDBState); + ALLOC_STATE(db, always, 17, r700SendDBState); ALLOC_STATE(stencil, always, 4, r700SendStencilState); ALLOC_STATE(db_target, always, 12, r700SendDepthTargetState); ALLOC_STATE(sc, always, 15, r700SendSCState); @@ -1252,22 +1310,23 @@ void r600InitAtoms(context_t *context) ALLOC_STATE(poly, always, 10, r700SendPolyState); ALLOC_STATE(cb, cb, 18, r700SendCBState); ALLOC_STATE(clrcmp, always, 6, r700SendCBCLRCMPState); + ALLOC_STATE(cb_target, always, 25, r700SendRenderTargetState); ALLOC_STATE(blnd, blnd, (6 + (R700_MAX_RENDER_TARGETS * 3)), r700SendCBBlendState); ALLOC_STATE(blnd_clr, always, 6, r700SendCBBlendColorState); - ALLOC_STATE(cb_target, always, 25, r700SendRenderTargetState); ALLOC_STATE(sx, always, 9, r700SendSXState); ALLOC_STATE(vgt, always, 41, r700SendVGTState); ALLOC_STATE(spi, always, (59 + R700_MAX_SHADER_EXPORTS), r700SendSPIState); ALLOC_STATE(vpt, always, 16, r700SendViewportState); ALLOC_STATE(fs, always, 18, r700SendFSState); - ALLOC_STATE(vs, always, 18, r700SendVSState); - ALLOC_STATE(ps, always, 21, r700SendPSState); + ALLOC_STATE(vs, always, 21, r700SendVSState); + ALLOC_STATE(ps, always, 24, r700SendPSState); ALLOC_STATE(vs_consts, vs_consts, (2 + (R700_MAX_DX9_CONSTS * 4)), r700SendVSConsts); ALLOC_STATE(ps_consts, ps_consts, (2 + (R700_MAX_DX9_CONSTS * 4)), r700SendPSConsts); ALLOC_STATE(vtx, vtx, (6 + (VERT_ATTRIB_MAX * 18)), r700SendVTXState); ALLOC_STATE(tx, tx, (R700_TEXTURE_NUMBERUNITS * 20), r700SendTexState); ALLOC_STATE(tx_smplr, tx, (R700_TEXTURE_NUMBERUNITS * 5), r700SendTexSamplerState); ALLOC_STATE(tx_brdr_clr, tx, (R700_TEXTURE_NUMBERUNITS * 6), r700SendTexBorderColorState); + r600_init_query_stateobj(&context->radeon, 6 * 2); context->radeon.hw.is_dirty = GL_TRUE; context->radeon.hw.all_dirty = GL_TRUE; diff --git a/src/mesa/drivers/dri/r600/r700_fragprog.c b/src/mesa/drivers/dri/r600/r700_fragprog.c index 78ce3ae4366..d15f0137107 100644 --- a/src/mesa/drivers/dri/r600/r700_fragprog.c +++ b/src/mesa/drivers/dri/r600/r700_fragprog.c @@ -44,12 +44,18 @@ //TODO : Validate FP input with VP output. void Map_Fragment_Program(r700_AssemblerBase *pAsm, - struct gl_fragment_program *mesa_fp) + struct gl_fragment_program *mesa_fp, + GLcontext *ctx) { unsigned int unBit; unsigned int i; GLuint ui; + /* match fp inputs with vp exports. */ + struct r700_vertex_program_cont *vpc = + (struct r700_vertex_program_cont *)ctx->VertexProgram._Current; + GLbitfield OutputsWritten = vpc->mesa_program.Base.OutputsWritten; + pAsm->number_used_registers = 0; //Input mapping : mesa_fp->Base.InputsRead set the flag, set in @@ -61,32 +67,93 @@ void Map_Fragment_Program(r700_AssemblerBase *pAsm, pAsm->uiFP_AttributeMap[FRAG_ATTRIB_WPOS] = pAsm->number_used_registers++; } - unBit = 1 << FRAG_ATTRIB_COL0; - if(mesa_fp->Base.InputsRead & unBit) + unBit = 1 << VERT_RESULT_COL0; + if(OutputsWritten & unBit) { pAsm->uiFP_AttributeMap[FRAG_ATTRIB_COL0] = pAsm->number_used_registers++; } - unBit = 1 << FRAG_ATTRIB_COL1; - if(mesa_fp->Base.InputsRead & unBit) + unBit = 1 << VERT_RESULT_COL1; + if(OutputsWritten & unBit) { pAsm->uiFP_AttributeMap[FRAG_ATTRIB_COL1] = pAsm->number_used_registers++; } - unBit = 1 << FRAG_ATTRIB_FOGC; - if(mesa_fp->Base.InputsRead & unBit) - { - pAsm->uiFP_AttributeMap[FRAG_ATTRIB_FOGC] = pAsm->number_used_registers++; - } + unBit = 1 << VERT_RESULT_FOGC; + if(OutputsWritten & unBit) + { + pAsm->uiFP_AttributeMap[FRAG_ATTRIB_FOGC] = pAsm->number_used_registers++; + } for(i=0; i<8; i++) { - unBit = 1 << (FRAG_ATTRIB_TEX0 + i); - if(mesa_fp->Base.InputsRead & unBit) + unBit = 1 << (VERT_RESULT_TEX0 + i); + if(OutputsWritten & unBit) { pAsm->uiFP_AttributeMap[FRAG_ATTRIB_TEX0 + i] = pAsm->number_used_registers++; } } + +/* order has been taken care of */ +#if 1 + for(i=VERT_RESULT_VAR0; i<VERT_RESULT_MAX; i++) + { + unBit = 1 << i; + if(OutputsWritten & unBit) + { + pAsm->uiFP_AttributeMap[i-VERT_RESULT_VAR0+FRAG_ATTRIB_VAR0] = pAsm->number_used_registers++; + } + } +#else + if( (mesa_fp->Base.InputsRead >> FRAG_ATTRIB_VAR0) > 0 ) + { + struct r700_vertex_program_cont *vpc = + (struct r700_vertex_program_cont *)ctx->VertexProgram._Current; + struct gl_program_parameter_list * VsVarying = vpc->mesa_program.Base.Varying; + struct gl_program_parameter_list * PsVarying = mesa_fp->Base.Varying; + struct gl_program_parameter * pVsParam; + struct gl_program_parameter * pPsParam; + GLuint j, k; + GLuint unMaxVarying = 0; + + for(i=0; i<VsVarying->NumParameters; i++) + { + pAsm->uiFP_AttributeMap[i + FRAG_ATTRIB_VAR0] = 0; + } + + for(i=FRAG_ATTRIB_VAR0; i<FRAG_ATTRIB_MAX; i++) + { + unBit = 1 << i; + if(mesa_fp->Base.InputsRead & unBit) + { + j = i - FRAG_ATTRIB_VAR0; + pPsParam = PsVarying->Parameters + j; + + for(k=0; k<VsVarying->NumParameters; k++) + { + pVsParam = VsVarying->Parameters + k; + + if( strcmp(pPsParam->Name, pVsParam->Name) == 0) + { + pAsm->uiFP_AttributeMap[i] = pAsm->number_used_registers + k; + if(k > unMaxVarying) + { + unMaxVarying = k; + } + break; + } + } + } + } + + pAsm->number_used_registers += unMaxVarying + 1; + } +#endif + unBit = 1 << FRAG_ATTRIB_FACE; + if(mesa_fp->Base.InputsRead & unBit) + { + pAsm->uiFP_AttributeMap[FRAG_ATTRIB_FACE] = pAsm->number_used_registers++; + } /* Map temporary registers (GPRs) */ pAsm->starting_temp_register_number = pAsm->number_used_registers; @@ -127,6 +194,8 @@ void Map_Fragment_Program(r700_AssemblerBase *pAsm, pAsm->pucOutMask[ui] = 0x0; } + pAsm->flag_reg_index = pAsm->number_used_registers++; + pAsm->uFirstHelpReg = pAsm->number_used_registers; } @@ -135,15 +204,19 @@ GLboolean Find_Instruction_Dependencies_fp(struct r700_fragment_program *fp, { GLuint i, j; GLint * puiTEMPwrites; + GLint * puiTEMPreads; struct prog_instruction * pILInst; InstDeps *pInstDeps; struct prog_instruction * texcoord_DepInst; GLint nDepInstID; puiTEMPwrites = (GLint*) MALLOC(sizeof(GLuint)*mesa_fp->Base.NumTemporaries); + puiTEMPreads = (GLint*) MALLOC(sizeof(GLuint)*mesa_fp->Base.NumTemporaries); + for(i=0; i<mesa_fp->Base.NumTemporaries; i++) { puiTEMPwrites[i] = -1; + puiTEMPreads[i] = -1; } pInstDeps = (InstDeps*)MALLOC(sizeof(InstDeps)*mesa_fp->Base.NumInstructions); @@ -167,6 +240,11 @@ GLboolean Find_Instruction_Dependencies_fp(struct r700_fragment_program *fp, { //Set dep. pInstDeps[i].nSrcDeps[j] = puiTEMPwrites[pILInst->SrcReg[j].Index]; + //Set first read + if(puiTEMPreads[pILInst->SrcReg[j].Index] < 0 ) + { + puiTEMPreads[pILInst->SrcReg[j].Index] = i; + } } else { @@ -177,8 +255,6 @@ GLboolean Find_Instruction_Dependencies_fp(struct r700_fragment_program *fp, fp->r700AsmCode.pInstDeps = pInstDeps; - FREE(puiTEMPwrites); - //Find dep for tex inst for(i=0; i<mesa_fp->Base.NumInstructions; i++) { @@ -203,29 +279,58 @@ GLboolean Find_Instruction_Dependencies_fp(struct r700_fragment_program *fp, { //... other deps? } } + // make sure that we dont overwrite src used earlier + nDepInstID = puiTEMPreads[pILInst->DstReg.Index]; + if(nDepInstID < i) + { + pInstDeps[i].nDstDep = puiTEMPreads[pILInst->DstReg.Index]; + texcoord_DepInst = &(mesa_fp->Base.Instructions[nDepInstID]); + if(GL_TRUE == IsAlu(texcoord_DepInst->Opcode) ) + { + pInstDeps[nDepInstID].nDstDep = i; + } + + } + } } + FREE(puiTEMPwrites); + FREE(puiTEMPreads); + return GL_TRUE; } GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp, - struct gl_fragment_program *mesa_fp) + struct gl_fragment_program *mesa_fp, + GLcontext *ctx) { GLuint number_of_colors_exported; GLboolean z_enabled = GL_FALSE; GLuint unBit; + int i; //Init_Program Init_r700_AssemblerBase( SPT_FP, &(fp->r700AsmCode), &(fp->r700Shader) ); - Map_Fragment_Program(&(fp->r700AsmCode), mesa_fp); + Map_Fragment_Program(&(fp->r700AsmCode), mesa_fp, ctx); if( GL_FALSE == Find_Instruction_Dependencies_fp(fp, mesa_fp) ) { return GL_FALSE; } + + InitShaderProgram(&(fp->r700AsmCode)); - if( GL_FALSE == AssembleInstr(mesa_fp->Base.NumInstructions, + for(i=0; i < MAX_SAMPLERS; i++) + { + fp->r700AsmCode.SamplerUnits[i] = fp->mesa_program.Base.SamplerUnits[i]; + } + + fp->r700AsmCode.unCurNumILInsts = mesa_fp->Base.NumInstructions; + + if( GL_FALSE == AssembleInstr(0, + 0, + mesa_fp->Base.NumInstructions, &(mesa_fp->Base.Instructions[0]), &(fp->r700AsmCode)) ) { @@ -237,6 +342,11 @@ GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp, return GL_FALSE; } + if( GL_FALSE == RelocProgram(&(fp->r700AsmCode), &(mesa_fp->Base)) ) + { + return GL_FALSE; + } + fp->r700Shader.nRegs = (fp->r700AsmCode.number_used_registers == 0) ? 0 : (fp->r700AsmCode.number_used_registers - 1); @@ -251,7 +361,15 @@ GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp, number_of_colors_exported--; } - fp->r700Shader.exportMode = number_of_colors_exported << 1 | z_enabled; + /* illegal to set this to 0 */ + if(number_of_colors_exported || z_enabled) + { + fp->r700Shader.exportMode = number_of_colors_exported << 1 | z_enabled; + } + else + { + fp->r700Shader.exportMode = (1 << 1); + } fp->translated = GL_TRUE; @@ -269,7 +387,7 @@ void r700SelectFragmentShader(GLcontext *ctx) } if (GL_FALSE == fp->translated) - r700TranslateFragmentShader(fp, &(fp->mesa_program)); + r700TranslateFragmentShader(fp, &(fp->mesa_program), ctx); } void * r700GetActiveFpShaderBo(GLcontext * ctx) @@ -341,6 +459,25 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) SETbit(r700->SPI_PS_IN_CONTROL_0.u32All, POSITION_ENA_bit); SETbit(r700->SPI_INPUT_Z.u32All, PROVIDE_Z_TO_SPI_bit); } + else + { + CLEARbit(r700->SPI_PS_IN_CONTROL_0.u32All, POSITION_ENA_bit); + CLEARbit(r700->SPI_INPUT_Z.u32All, PROVIDE_Z_TO_SPI_bit); + } + + if (mesa_fp->Base.InputsRead & (1 << FRAG_ATTRIB_FACE)) + { + ui += 1; + SETfield(r700->SPI_PS_IN_CONTROL_0.u32All, ui, NUM_INTERP_shift, NUM_INTERP_mask); + SETbit(r700->SPI_PS_IN_CONTROL_1.u32All, FRONT_FACE_ENA_bit); + SETbit(r700->SPI_PS_IN_CONTROL_1.u32All, FRONT_FACE_ALL_BITS_bit); + SETfield(r700->SPI_PS_IN_CONTROL_1.u32All, pAsm->uiFP_AttributeMap[FRAG_ATTRIB_FACE], FRONT_FACE_ADDR_shift, FRONT_FACE_ADDR_mask); + } + else + { + CLEARbit(r700->SPI_PS_IN_CONTROL_1.u32All, FRONT_FACE_ENA_bit); + } + ui = (unNumOfReg < ui) ? ui : unNumOfReg; @@ -357,27 +494,10 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) SETfield(r700->ps.SQ_PGM_EXPORTS_PS.u32All, fp->r700Shader.exportMode, EXPORT_MODE_shift, EXPORT_MODE_mask); - R600_STATECHANGE(context, db); - - if(fp->r700Shader.killIsUsed) - { - SETbit(r700->DB_SHADER_CONTROL.u32All, KILL_ENABLE_bit); - } - else - { - CLEARbit(r700->DB_SHADER_CONTROL.u32All, KILL_ENABLE_bit); - } - - if(fp->r700Shader.depthIsExported) - { - SETbit(r700->DB_SHADER_CONTROL.u32All, Z_EXPORT_ENABLE_bit); - } - else - { - CLEARbit(r700->DB_SHADER_CONTROL.u32All, Z_EXPORT_ENABLE_bit); - } - // emit ps input map + struct r700_vertex_program_cont *vpc = + (struct r700_vertex_program_cont *)ctx->VertexProgram._Current; + GLbitfield OutputsWritten = vpc->mesa_program.Base.OutputsWritten; unBit = 1 << FRAG_ATTRIB_WPOS; if(mesa_fp->Base.InputsRead & unBit) { @@ -391,8 +511,8 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); } - unBit = 1 << FRAG_ATTRIB_COL0; - if(mesa_fp->Base.InputsRead & unBit) + unBit = 1 << VERT_RESULT_COL0; + if(OutputsWritten & unBit) { ui = pAsm->uiFP_AttributeMap[FRAG_ATTRIB_COL0]; SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit); @@ -404,8 +524,8 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); } - unBit = 1 << FRAG_ATTRIB_COL1; - if(mesa_fp->Base.InputsRead & unBit) + unBit = 1 << VERT_RESULT_COL1; + if(OutputsWritten & unBit) { ui = pAsm->uiFP_AttributeMap[FRAG_ATTRIB_COL1]; SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit); @@ -417,8 +537,8 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); } - unBit = 1 << FRAG_ATTRIB_FOGC; - if(mesa_fp->Base.InputsRead & unBit) + unBit = 1 << VERT_RESULT_FOGC; + if(OutputsWritten & unBit) { ui = pAsm->uiFP_AttributeMap[FRAG_ATTRIB_FOGC]; SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit); @@ -432,8 +552,8 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) for(i=0; i<8; i++) { - unBit = 1 << (FRAG_ATTRIB_TEX0 + i); - if(mesa_fp->Base.InputsRead & unBit) + unBit = 1 << (VERT_RESULT_TEX0 + i); + if(OutputsWritten & unBit) { ui = pAsm->uiFP_AttributeMap[FRAG_ATTRIB_TEX0 + i]; SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit); @@ -443,14 +563,47 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) } } - R600_STATECHANGE(context, cb); + unBit = 1 << FRAG_ATTRIB_FACE; + if(mesa_fp->Base.InputsRead & unBit) + { + ui = pAsm->uiFP_AttributeMap[FRAG_ATTRIB_FACE]; + SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit); + SETfield(r700->SPI_PS_INPUT_CNTL[ui].u32All, ui, + SEMANTIC_shift, SEMANTIC_mask); + if (r700->SPI_INTERP_CONTROL_0.u32All & FLAT_SHADE_ENA_bit) + SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); + else + CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); + } + + for(i=VERT_RESULT_VAR0; i<VERT_RESULT_MAX; i++) + { + unBit = 1 << i; + if(OutputsWritten & unBit) + { + ui = pAsm->uiFP_AttributeMap[i-VERT_RESULT_VAR0+FRAG_ATTRIB_VAR0]; + SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, SEL_CENTROID_bit); + SETfield(r700->SPI_PS_INPUT_CNTL[ui].u32All, ui, + SEMANTIC_shift, SEMANTIC_mask); + if (r700->SPI_INTERP_CONTROL_0.u32All & FLAT_SHADE_ENA_bit) + SETbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); + else + CLEARbit(r700->SPI_PS_INPUT_CNTL[ui].u32All, FLAT_SHADE_bit); + } + } + exportCount = (r700->ps.SQ_PGM_EXPORTS_PS.u32All & EXPORT_MODE_mask) / (1 << EXPORT_MODE_shift); - r700->CB_SHADER_CONTROL.u32All = (1 << exportCount) - 1; + if (r700->CB_SHADER_CONTROL.u32All != ((1 << exportCount) - 1)) + { + R600_STATECHANGE(context, cb); + r700->CB_SHADER_CONTROL.u32All = (1 << exportCount) - 1; + } /* sent out shader constants. */ paramList = fp->mesa_program.Base.Parameters; - if(NULL != paramList) { + if(NULL != paramList) + { _mesa_load_state_parameters(ctx, paramList); if (paramList->NumParameters > R700_MAX_DX9_CONSTS) @@ -463,14 +616,33 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx) unNumParamData = paramList->NumParameters; for(ui=0; ui<unNumParamData; ui++) { - r700->ps.consts[ui][0].f32All = paramList->ParameterValues[ui][0]; - r700->ps.consts[ui][1].f32All = paramList->ParameterValues[ui][1]; - r700->ps.consts[ui][2].f32All = paramList->ParameterValues[ui][2]; - r700->ps.consts[ui][3].f32All = paramList->ParameterValues[ui][3]; + r700->ps.consts[ui][0].f32All = paramList->ParameterValues[ui][0]; + r700->ps.consts[ui][1].f32All = paramList->ParameterValues[ui][1]; + r700->ps.consts[ui][2].f32All = paramList->ParameterValues[ui][2]; + r700->ps.consts[ui][3].f32All = paramList->ParameterValues[ui][3]; } } else r700->ps.num_consts = 0; + COMPILED_SUB * pCompiledSub; + GLuint uj; + GLuint unConstOffset = r700->ps.num_consts; + for(ui=0; ui<pAsm->unNumPresub; ui++) + { + pCompiledSub = pAsm->presubs[ui].pCompiledSub; + + r700->ps.num_consts += pCompiledSub->NumParameters; + + for(uj=0; uj<pCompiledSub->NumParameters; uj++) + { + r700->ps.consts[uj + unConstOffset][0].f32All = pCompiledSub->ParameterValues[uj][0]; + r700->ps.consts[uj + unConstOffset][1].f32All = pCompiledSub->ParameterValues[uj][1]; + r700->ps.consts[uj + unConstOffset][2].f32All = pCompiledSub->ParameterValues[uj][2]; + r700->ps.consts[uj + unConstOffset][3].f32All = pCompiledSub->ParameterValues[uj][3]; + } + unConstOffset += pCompiledSub->NumParameters; + } + return GL_TRUE; } diff --git a/src/mesa/drivers/dri/r600/r700_fragprog.h b/src/mesa/drivers/dri/r600/r700_fragprog.h index cbb108d2128..e562bfa4789 100644 --- a/src/mesa/drivers/dri/r600/r700_fragprog.h +++ b/src/mesa/drivers/dri/r600/r700_fragprog.h @@ -49,12 +49,14 @@ struct r700_fragment_program /* Internal */ void Map_Fragment_Program(r700_AssemblerBase *pAsm, - struct gl_fragment_program *mesa_fp); + struct gl_fragment_program *mesa_fp, + GLcontext *ctx); GLboolean Find_Instruction_Dependencies_fp(struct r700_fragment_program *fp, struct gl_fragment_program *mesa_fp); GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp, - struct gl_fragment_program *mesa_vp); + struct gl_fragment_program *mesa_vp, + GLcontext *ctx); /* Interface */ extern void r700SelectFragmentShader(GLcontext *ctx); diff --git a/src/mesa/drivers/dri/r600/r700_oglprog.c b/src/mesa/drivers/dri/r600/r700_oglprog.c index 5290ef31be3..0d476fcd863 100644 --- a/src/mesa/drivers/dri/r600/r700_oglprog.c +++ b/src/mesa/drivers/dri/r600/r700_oglprog.c @@ -40,6 +40,24 @@ #include "r700_vertprog.h" +static void freeVertProgCache(GLcontext *ctx, struct r700_vertex_program_cont *cache) +{ + struct r700_vertex_program *tmp, *vp = cache->progs; + + while (vp) { + tmp = vp->next; + /* Release DMA region */ + r600DeleteShader(ctx, vp->shaderbo); + /* Clean up */ + Clean_Up_Assembler(&(vp->r700AsmCode)); + Clean_Up_Shader(&(vp->r700Shader)); + + _mesa_reference_vertprog(ctx, &vp->mesa_program, NULL); + _mesa_free(vp); + vp = tmp; + } +} + static struct gl_program *r700NewProgram(GLcontext * ctx, GLenum target, GLuint id) @@ -84,8 +102,7 @@ static struct gl_program *r700NewProgram(GLcontext * ctx, static void r700DeleteProgram(GLcontext * ctx, struct gl_program *prog) { - struct r700_vertex_program_cont * vpc; - struct r700_vertex_program *vp, *tmp; + struct r700_vertex_program_cont *vpc = (struct r700_vertex_program_cont *)prog; struct r700_fragment_program * fp; radeon_print(RADEON_SHADER, RADEON_VERBOSE, @@ -95,20 +112,7 @@ static void r700DeleteProgram(GLcontext * ctx, struct gl_program *prog) { case GL_VERTEX_STATE_PROGRAM_NV: case GL_VERTEX_PROGRAM_ARB: - vpc = (struct r700_vertex_program_cont*)prog; - vp = vpc->progs; - while (vp) { - tmp = vp->next; - /* Release DMA region */ - - r600DeleteShader(ctx, vp->shaderbo); - - /* Clean up */ - Clean_Up_Assembler(&(vp->r700AsmCode)); - Clean_Up_Shader(&(vp->r700Shader)); - _mesa_free(vp); - vp = tmp; - } + freeVertProgCache(ctx, vpc); break; case GL_FRAGMENT_PROGRAM_NV: case GL_FRAGMENT_PROGRAM_ARB: @@ -131,7 +135,24 @@ static void r700DeleteProgram(GLcontext * ctx, struct gl_program *prog) static void r700ProgramStringNotify(GLcontext * ctx, GLenum target, struct gl_program *prog) { - + struct r700_vertex_program_cont *vpc = (struct r700_vertex_program_cont *)prog; + struct r700_fragment_program * fp = (struct r700_fragment_program*)prog; + + switch (target) { + case GL_VERTEX_PROGRAM_ARB: + freeVertProgCache(ctx, vpc); + vpc->progs = NULL; + break; + case GL_FRAGMENT_PROGRAM_ARB: + r600DeleteShader(ctx, fp->shaderbo); + Clean_Up_Assembler(&(fp->r700AsmCode)); + Clean_Up_Shader(&(fp->r700Shader)); + fp->translated = GL_FALSE; + fp->loaded = GL_FALSE; + fp->shaderbo = NULL; + break; + } + } static GLboolean r700IsProgramNative(GLcontext * ctx, GLenum target, struct gl_program *prog) diff --git a/src/mesa/drivers/dri/r600/r700_render.c b/src/mesa/drivers/dri/r600/r700_render.c index b1c3648ca56..eab27cbd842 100644 --- a/src/mesa/drivers/dri/r600/r700_render.c +++ b/src/mesa/drivers/dri/r600/r700_render.c @@ -43,6 +43,7 @@ #include "tnl/t_context.h" #include "tnl/t_vertex.h" #include "tnl/t_pipeline.h" +#include "vbo/vbo_context.h" #include "r600_context.h" #include "r600_cmdbuf.h" @@ -53,13 +54,12 @@ #include "r700_fragprog.h" #include "r700_state.h" +#include "radeon_buffer_objects.h" #include "radeon_common_context.h" void r700WaitForIdle(context_t *context); void r700WaitForIdleClean(context_t *context); -GLboolean r700SendTextureState(context_t *context); static unsigned int r700PrimitiveType(int prim); -void r600UpdateTextureState(GLcontext * ctx); GLboolean r700SyncSurf(context_t *context, struct radeon_bo *pbo, uint32_t read_domain, @@ -249,113 +249,654 @@ static int r700NumVerts(int num_verts, int prim) static void r700RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim) { - context_t *context = R700_CONTEXT(ctx); - BATCH_LOCALS(&context->radeon); - int type, i, total_emit; - int num_indices; - uint32_t vgt_draw_initiator = 0; - uint32_t vgt_index_type = 0; - uint32_t vgt_primitive_type = 0; - uint32_t vgt_num_indices = 0; - TNLcontext *tnl = TNL_CONTEXT(ctx); - struct vertex_buffer *vb = &tnl->vb; - - type = r700PrimitiveType(prim); - num_indices = r700NumVerts(end - start, prim); - - radeon_print(RADEON_RENDER, RADEON_TRACE, - "%s type %x num_indices %d\n", - __func__, type, num_indices); - - if (type < 0 || num_indices <= 0) - return; + context_t *context = R700_CONTEXT(ctx); + BATCH_LOCALS(&context->radeon); + int type, total_emit; + int num_indices; + uint32_t vgt_draw_initiator = 0; + uint32_t vgt_index_type = 0; + uint32_t vgt_primitive_type = 0; + uint32_t vgt_num_indices = 0; + + type = r700PrimitiveType(prim); + num_indices = r700NumVerts(end - start, prim); + + radeon_print(RADEON_RENDER, RADEON_TRACE, + "%s type %x num_indices %d\n", + __func__, type, num_indices); + + if (type < 0 || num_indices <= 0) + return; + + SETfield(vgt_primitive_type, type, + VGT_PRIMITIVE_TYPE__PRIM_TYPE_shift, VGT_PRIMITIVE_TYPE__PRIM_TYPE_mask); + + SETfield(vgt_index_type, DI_INDEX_SIZE_32_BIT, INDEX_TYPE_shift, INDEX_TYPE_mask); + + if(GL_TRUE != context->ind_buf.is_32bit) + { + SETfield(vgt_index_type, DI_INDEX_SIZE_16_BIT, INDEX_TYPE_shift, INDEX_TYPE_mask); + } + + vgt_num_indices = num_indices; + SETfield(vgt_draw_initiator, DI_SRC_SEL_DMA, SOURCE_SELECT_shift, SOURCE_SELECT_mask); + SETfield(vgt_draw_initiator, DI_MAJOR_MODE_0, MAJOR_MODE_shift, MAJOR_MODE_mask); + + total_emit = 3 /* VGT_PRIMITIVE_TYPE */ + + 2 /* VGT_INDEX_TYPE */ + + 2 /* NUM_INSTANCES */ + + 5 + 2; /* DRAW_INDEX */ + + BEGIN_BATCH_NO_AUTOSTATE(total_emit); + // prim + R600_OUT_BATCH_REGSEQ(VGT_PRIMITIVE_TYPE, 1); + R600_OUT_BATCH(vgt_primitive_type); + // index type + R600_OUT_BATCH(CP_PACKET3(R600_IT_INDEX_TYPE, 0)); + R600_OUT_BATCH(vgt_index_type); + // num instances + R600_OUT_BATCH(CP_PACKET3(R600_IT_NUM_INSTANCES, 0)); + R600_OUT_BATCH(1); + // draw packet + R600_OUT_BATCH(CP_PACKET3(R600_IT_DRAW_INDEX, 3)); + R600_OUT_BATCH(context->ind_buf.bo_offset); + R600_OUT_BATCH(0); + R600_OUT_BATCH(vgt_num_indices); + R600_OUT_BATCH(vgt_draw_initiator); + R600_OUT_BATCH_RELOC(context->ind_buf.bo_offset, + context->ind_buf.bo, + context->ind_buf.bo_offset, + RADEON_GEM_DOMAIN_GTT, 0, 0); + END_BATCH(); + COMMIT_BATCH(); +} + +static void r700RunRenderPrimitiveImmediate(GLcontext * ctx, int start, int end, int prim) +{ + context_t *context = R700_CONTEXT(ctx); + BATCH_LOCALS(&context->radeon); + int type, i; + uint32_t num_indices, total_emit = 0; + uint32_t vgt_draw_initiator = 0; + uint32_t vgt_index_type = 0; + uint32_t vgt_primitive_type = 0; + uint32_t vgt_num_indices = 0; + + type = r700PrimitiveType(prim); + num_indices = r700NumVerts(end - start, prim); + + radeon_print(RADEON_RENDER, RADEON_TRACE, + "%s type %x num_indices %d\n", + __func__, type, num_indices); + + if (type < 0 || num_indices <= 0) + return; + + SETfield(vgt_primitive_type, type, + VGT_PRIMITIVE_TYPE__PRIM_TYPE_shift, VGT_PRIMITIVE_TYPE__PRIM_TYPE_mask); - total_emit = 3 /* VGT_PRIMITIVE_TYPE */ - + 2 /* VGT_INDEX_TYPE */ - + 2 /* NUM_INSTANCES */ - + num_indices + 3; /* DRAW_INDEX_IMMD */ - - BEGIN_BATCH_NO_AUTOSTATE(total_emit); - // prim - SETfield(vgt_primitive_type, type, - VGT_PRIMITIVE_TYPE__PRIM_TYPE_shift, VGT_PRIMITIVE_TYPE__PRIM_TYPE_mask); - R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CONFIG_REG, 1)); - R600_OUT_BATCH(mmVGT_PRIMITIVE_TYPE - ASIC_CONFIG_BASE_INDEX); - R600_OUT_BATCH(vgt_primitive_type); - - // index type - SETfield(vgt_index_type, DI_INDEX_SIZE_32_BIT, INDEX_TYPE_shift, INDEX_TYPE_mask); - R600_OUT_BATCH(CP_PACKET3(R600_IT_INDEX_TYPE, 0)); - R600_OUT_BATCH(vgt_index_type); - - // num instances - R600_OUT_BATCH(CP_PACKET3(R600_IT_NUM_INSTANCES, 0)); - R600_OUT_BATCH(1); - - // draw packet - vgt_num_indices = num_indices; - SETfield(vgt_draw_initiator, DI_SRC_SEL_IMMEDIATE, SOURCE_SELECT_shift, SOURCE_SELECT_mask); - SETfield(vgt_draw_initiator, DI_MAJOR_MODE_0, MAJOR_MODE_shift, MAJOR_MODE_mask); - - R600_OUT_BATCH(CP_PACKET3(R600_IT_DRAW_INDEX_IMMD, (num_indices + 1))); + if (num_indices > 0xffff) + { + SETfield(vgt_index_type, DI_INDEX_SIZE_32_BIT, INDEX_TYPE_shift, INDEX_TYPE_mask); + } + else + { + SETfield(vgt_index_type, DI_INDEX_SIZE_16_BIT, INDEX_TYPE_shift, INDEX_TYPE_mask); + } + + vgt_num_indices = num_indices; + SETfield(vgt_draw_initiator, DI_MAJOR_MODE_0, MAJOR_MODE_shift, MAJOR_MODE_mask); + + if (start == 0) + { + SETfield(vgt_draw_initiator, DI_SRC_SEL_AUTO_INDEX, SOURCE_SELECT_shift, SOURCE_SELECT_mask); + } + else + { + if (num_indices > 0xffff) + { + total_emit += num_indices; + } + else + { + total_emit += (num_indices + 1) / 2; + } + SETfield(vgt_draw_initiator, DI_SRC_SEL_IMMEDIATE, SOURCE_SELECT_shift, SOURCE_SELECT_mask); + } + + total_emit += 3 /* VGT_PRIMITIVE_TYPE */ + + 2 /* VGT_INDEX_TYPE */ + + 2 /* NUM_INSTANCES */ + + 3; /* DRAW */ + + BEGIN_BATCH_NO_AUTOSTATE(total_emit); + // prim + R600_OUT_BATCH_REGSEQ(VGT_PRIMITIVE_TYPE, 1); + R600_OUT_BATCH(vgt_primitive_type); + // index type + R600_OUT_BATCH(CP_PACKET3(R600_IT_INDEX_TYPE, 0)); + R600_OUT_BATCH(vgt_index_type); + // num instances + R600_OUT_BATCH(CP_PACKET3(R600_IT_NUM_INSTANCES, 0)); + R600_OUT_BATCH(1); + // draw packet + if(start == 0) + { + R600_OUT_BATCH(CP_PACKET3(R600_IT_DRAW_INDEX_AUTO, 1)); R600_OUT_BATCH(vgt_num_indices); R600_OUT_BATCH(vgt_draw_initiator); - - for (i = start; i < (start + num_indices); i++) { - if(vb->Elts) - R600_OUT_BATCH(vb->Elts[i]); + } + else + { + if (num_indices > 0xffff) + { + R600_OUT_BATCH(CP_PACKET3(R600_IT_DRAW_INDEX_IMMD, (num_indices + 1))); + R600_OUT_BATCH(vgt_num_indices); + R600_OUT_BATCH(vgt_draw_initiator); + for (i = start; i < (start + num_indices); i++) + { + R600_OUT_BATCH(i); + } + } + else + { + R600_OUT_BATCH(CP_PACKET3(R600_IT_DRAW_INDEX_IMMD, (((num_indices + 1) / 2) + 1))); + R600_OUT_BATCH(vgt_num_indices); + R600_OUT_BATCH(vgt_draw_initiator); + for (i = start; i < (start + num_indices); i += 2) + { + if ((i + 1) == (start + num_indices)) + { + R600_OUT_BATCH(i); + } else - R600_OUT_BATCH(i); - } - END_BATCH(); - COMMIT_BATCH(); + { + R600_OUT_BATCH(((i + 1) << 16) | (i)); + } + } + } + } + END_BATCH(); + COMMIT_BATCH(); } /* start 3d, idle, cb/db flush */ #define PRE_EMIT_STATE_BUFSZ 10 + 5 + 14 -static GLuint r700PredictRenderSize(GLcontext* ctx) +static GLuint r700PredictRenderSize(GLcontext* ctx, + const struct _mesa_prim *prim, + const struct _mesa_index_buffer *ib, + GLuint nr_prims) { context_t *context = R700_CONTEXT(ctx); - TNLcontext *tnl = TNL_CONTEXT(ctx); - struct r700_vertex_program *vp = context->selected_vp; - struct vertex_buffer *vb = &tnl->vb; GLboolean flushed; GLuint dwords, i; GLuint state_size; - /* pre calculate aos count so state prediction works */ - context->radeon.tcl.aos_count = _mesa_bitcount(vp->mesa_program->Base.InputsRead); dwords = PRE_EMIT_STATE_BUFSZ; - for (i = 0; i < vb->PrimitiveCount; i++) - dwords += vb->Primitive[i].count + 10; + if (ib) + dwords += nr_prims * 14; + else { + for (i = 0; i < nr_prims; ++i) + { + if (prim[i].start == 0) + dwords += 10; + else if (prim[i].count > 0xffff) + dwords += prim[i].count + 10; + else + dwords += ((prim[i].count + 1) / 2) + 10; + } + } + state_size = radeonCountStateEmitSize(&context->radeon); flushed = rcommonEnsureCmdBufSpace(&context->radeon, - dwords + state_size, __FUNCTION__); - + dwords + state_size, + __FUNCTION__); if (flushed) - dwords += radeonCountStateEmitSize(&context->radeon); + dwords += radeonCountStateEmitSize(&context->radeon); else - dwords += state_size; + dwords += state_size; - radeon_print(RADEON_RENDER, RADEON_VERBOSE, - "%s: total prediction size is %d.\n", __FUNCTION__, dwords); + radeon_print(RADEON_RENDER, RADEON_VERBOSE, "%s: total prediction size is %d.\n", __FUNCTION__, dwords); return dwords; + +} + +#define CONVERT( TYPE, MACRO ) do { \ + GLuint i, j, sz; \ + sz = input->Size; \ + if (input->Normalized) { \ + for (i = 0; i < count; i++) { \ + const TYPE *in = (TYPE *)src_ptr; \ + for (j = 0; j < sz; j++) { \ + *dst_ptr++ = MACRO(*in); \ + in++; \ + } \ + src_ptr += stride; \ + } \ + } else { \ + for (i = 0; i < count; i++) { \ + const TYPE *in = (TYPE *)src_ptr; \ + for (j = 0; j < sz; j++) { \ + *dst_ptr++ = (GLfloat)(*in); \ + in++; \ + } \ + src_ptr += stride; \ + } \ + } \ +} while (0) + +/** + * Convert attribute data type to float + * If the attribute uses named buffer object replace the bo with newly allocated bo + */ +static void r700ConvertAttrib(GLcontext *ctx, int count, + const struct gl_client_array *input, + struct StreamDesc *attr) +{ + context_t *context = R700_CONTEXT(ctx); + const GLvoid *src_ptr; + GLboolean mapped_named_bo = GL_FALSE; + GLfloat *dst_ptr; + GLuint stride; + + stride = (input->StrideB == 0) ? getTypeSize(input->Type) * input->Size : input->StrideB; + + /* Convert value for first element only */ + if (input->StrideB == 0) + { + count = 1; + } + + if (input->BufferObj->Name) + { + if (!input->BufferObj->Pointer) + { + ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER, GL_READ_ONLY_ARB, input->BufferObj); + mapped_named_bo = GL_TRUE; + } + + src_ptr = ADD_POINTERS(input->BufferObj->Pointer, input->Ptr); + } + else + { + src_ptr = input->Ptr; + } + + radeonAllocDmaRegion(&context->radeon, &attr->bo, &attr->bo_offset, + sizeof(GLfloat) * input->Size * count, 32); + + radeon_bo_map(attr->bo, 1); + + dst_ptr = (GLfloat *)ADD_POINTERS(attr->bo->ptr, attr->bo_offset); + + assert(src_ptr != NULL); + + switch (input->Type) + { + case GL_DOUBLE: + CONVERT(GLdouble, (GLfloat)); + break; + case GL_UNSIGNED_INT: + CONVERT(GLuint, UINT_TO_FLOAT); + break; + case GL_INT: + CONVERT(GLint, INT_TO_FLOAT); + break; + case GL_UNSIGNED_SHORT: + CONVERT(GLushort, USHORT_TO_FLOAT); + break; + case GL_SHORT: + CONVERT(GLshort, SHORT_TO_FLOAT); + break; + case GL_UNSIGNED_BYTE: + assert(input->Format != GL_BGRA); + CONVERT(GLubyte, UBYTE_TO_FLOAT); + break; + case GL_BYTE: + CONVERT(GLbyte, BYTE_TO_FLOAT); + break; + default: + assert(0); + break; + } + + radeon_bo_unmap(attr->bo); + + if (mapped_named_bo) + { + ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER, input->BufferObj); + } +} + +static void r700AlignDataToDword(GLcontext *ctx, + const struct gl_client_array *input, + int count, + struct StreamDesc *attr) +{ + context_t *context = R700_CONTEXT(ctx); + const int dst_stride = (input->StrideB + 3) & ~3; + const int size = getTypeSize(input->Type) * input->Size * count; + GLboolean mapped_named_bo = GL_FALSE; + + radeonAllocDmaRegion(&context->radeon, &attr->bo, &attr->bo_offset, size, 32); + + radeon_bo_map(attr->bo, 1); + + if (!input->BufferObj->Pointer) + { + ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER, GL_READ_ONLY_ARB, input->BufferObj); + mapped_named_bo = GL_TRUE; + } + + { + GLvoid *src_ptr = ADD_POINTERS(input->BufferObj->Pointer, input->Ptr); + GLvoid *dst_ptr = ADD_POINTERS(attr->bo->ptr, attr->bo_offset); + int i; + + for (i = 0; i < count; ++i) + { + _mesa_memcpy(dst_ptr, src_ptr, input->StrideB); + src_ptr += input->StrideB; + dst_ptr += dst_stride; + } + } + + radeon_bo_unmap(attr->bo); + if (mapped_named_bo) + { + ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER, input->BufferObj); + } + + attr->stride = dst_stride; +} + +static void r700SetupStreams(GLcontext *ctx, const struct gl_client_array *input[], int count) +{ + context_t *context = R700_CONTEXT(ctx); + GLuint stride; + int ret; + int i, index; + + R600_STATECHANGE(context, vtx); + + for(index = 0; index < context->nNumActiveAos; index++) + { + struct radeon_aos *aos = &context->radeon.tcl.aos[index]; + i = context->stream_desc[index].element; + + stride = (input[i]->StrideB == 0) ? getTypeSize(input[i]->Type) * input[i]->Size : input[i]->StrideB; + + if (input[i]->Type == GL_DOUBLE || input[i]->Type == GL_UNSIGNED_INT || input[i]->Type == GL_INT || +#if MESA_BIG_ENDIAN + getTypeSize(input[i]->Type) != 4 || +#endif + stride < 4) + { + r700ConvertAttrib(ctx, count, input[i], &context->stream_desc[index]); + } + else + { + if (input[i]->BufferObj->Name) + { + if (stride % 4 != 0) + { + assert(((intptr_t) input[i]->Ptr) % input[i]->StrideB == 0); + r700AlignDataToDword(ctx, input[i], count, &context->stream_desc[index]); + context->stream_desc[index].is_named_bo = GL_FALSE; + } + else + { + context->stream_desc[index].stride = input[i]->StrideB; + context->stream_desc[index].bo_offset = (intptr_t) input[i]->Ptr; + context->stream_desc[index].bo = get_radeon_buffer_object(input[i]->BufferObj)->bo; + context->stream_desc[index].is_named_bo = GL_TRUE; + } + } + else + { + int size; + int local_count = count; + uint32_t *dst; + + if (input[i]->StrideB == 0) + { + size = getTypeSize(input[i]->Type) * input[i]->Size; + local_count = 1; + } + else + { + size = getTypeSize(input[i]->Type) * input[i]->Size * local_count; + } + + radeonAllocDmaRegion(&context->radeon, &context->stream_desc[index].bo, + &context->stream_desc[index].bo_offset, size, 32); + + radeon_bo_map(context->stream_desc[index].bo, 1); + assert(context->stream_desc[index].bo->ptr != NULL); + + + dst = (uint32_t *)ADD_POINTERS(context->stream_desc[index].bo->ptr, + context->stream_desc[index].bo_offset); + + switch (context->stream_desc[index].dwords) + { + case 1: + radeonEmitVec4(dst, input[i]->Ptr, input[i]->StrideB, local_count); + break; + case 2: + radeonEmitVec8(dst, input[i]->Ptr, input[i]->StrideB, local_count); + break; + case 3: + radeonEmitVec12(dst, input[i]->Ptr, input[i]->StrideB, local_count); + break; + case 4: + radeonEmitVec16(dst, input[i]->Ptr, input[i]->StrideB, local_count); + break; + default: + assert(0); + break; + } + radeon_bo_unmap(context->stream_desc[index].bo); + } + } + + aos->count = context->stream_desc[index].stride == 0 ? 1 : count; + aos->stride = context->stream_desc[index].stride / sizeof(float); + aos->components = context->stream_desc[index].dwords; + aos->bo = context->stream_desc[index].bo; + aos->offset = context->stream_desc[index].bo_offset; + + if(context->stream_desc[index].is_named_bo) + { + radeon_cs_space_add_persistent_bo(context->radeon.cmdbuf.cs, + context->stream_desc[index].bo, + RADEON_GEM_DOMAIN_GTT, 0); + } + } + + ret = radeon_cs_space_check_with_bo(context->radeon.cmdbuf.cs, + first_elem(&context->radeon.dma.reserved)->bo, + RADEON_GEM_DOMAIN_GTT, 0); +} + +static void r700FreeData(GLcontext *ctx) +{ + /* Need to zero tcl.aos[n].bo and tcl.elt_dma_bo + * to prevent double unref in radeonReleaseArrays + * called during context destroy + */ + context_t *context = R700_CONTEXT(ctx); + + int i; + + for (i = 0; i < context->nNumActiveAos; i++) + { + if (!context->stream_desc[i].is_named_bo) + { + radeon_bo_unref(context->stream_desc[i].bo); + } + context->radeon.tcl.aos[i].bo = NULL; + } + + if (context->ind_buf.bo != NULL) + { + radeon_bo_unref(context->ind_buf.bo); + } +} + +static void r700FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer *mesa_ind_buf) +{ + context_t *context = R700_CONTEXT(ctx); + GLvoid *src_ptr; + GLuint *out; + int i; + GLboolean mapped_named_bo = GL_FALSE; + + if (mesa_ind_buf->obj->Name && !mesa_ind_buf->obj->Pointer) + { + ctx->Driver.MapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER, GL_READ_ONLY_ARB, mesa_ind_buf->obj); + mapped_named_bo = GL_TRUE; + assert(mesa_ind_buf->obj->Pointer != NULL); + } + src_ptr = ADD_POINTERS(mesa_ind_buf->obj->Pointer, mesa_ind_buf->ptr); + + if (mesa_ind_buf->type == GL_UNSIGNED_BYTE) + { + GLuint size = sizeof(GLushort) * ((mesa_ind_buf->count + 1) & ~1); + GLubyte *in = (GLubyte *)src_ptr; + + radeonAllocDmaRegion(&context->radeon, &context->ind_buf.bo, + &context->ind_buf.bo_offset, size, 4); + + radeon_bo_map(context->ind_buf.bo, 1); + assert(context->ind_buf.bo->ptr != NULL); + out = (GLuint *)ADD_POINTERS(context->ind_buf.bo->ptr, context->ind_buf.bo_offset); + + for (i = 0; i + 1 < mesa_ind_buf->count; i += 2) + { + *out++ = in[i] | in[i + 1] << 16; + } + + if (i < mesa_ind_buf->count) + { + *out++ = in[i]; + } + + radeon_bo_unmap(context->ind_buf.bo); +#if MESA_BIG_ENDIAN + } + else + { /* if (mesa_ind_buf->type == GL_UNSIGNED_SHORT) */ + GLushort *in = (GLushort *)src_ptr; + GLuint size = sizeof(GLushort) * ((mesa_ind_buf->count + 1) & ~1); + + radeonAllocDmaRegion(&context->radeon, &context->ind_buf.bo, + &context->ind_buf.bo_offset, size, 4); + + radeon_bo_map(context->ind_buf.bo, 1); + assert(context->ind_buf.bo->ptr != NULL); + out = (GLuint *)ADD_POINTERS(context->ind_buf.bo->ptr, context->ind_buf.bo_offset); + + for (i = 0; i + 1 < mesa_ind_buf->count; i += 2) + { + *out++ = in[i] | in[i + 1] << 16; + } + + if (i < mesa_ind_buf->count) + { + *out++ = in[i]; + } + radeon_bo_unmap(context->ind_buf.bo); +#endif + } + + context->ind_buf.is_32bit = GL_FALSE; + context->ind_buf.count = mesa_ind_buf->count; + + if (mapped_named_bo) + { + ctx->Driver.UnmapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER, mesa_ind_buf->obj); + } } -static GLboolean r700RunRender(GLcontext * ctx, - struct tnl_pipeline_stage *stage) +static void r700SetupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer *mesa_ind_buf) +{ + context_t *context = R700_CONTEXT(ctx); + + if (!mesa_ind_buf) { + context->ind_buf.bo = NULL; + return; + } + +#if MESA_BIG_ENDIAN + if (mesa_ind_buf->type == GL_UNSIGNED_INT) + { +#else + if (mesa_ind_buf->type != GL_UNSIGNED_BYTE) + { +#endif + const GLvoid *src_ptr; + GLvoid *dst_ptr; + GLboolean mapped_named_bo = GL_FALSE; + + if (mesa_ind_buf->obj->Name && !mesa_ind_buf->obj->Pointer) + { + ctx->Driver.MapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER, GL_READ_ONLY_ARB, mesa_ind_buf->obj); + assert(mesa_ind_buf->obj->Pointer != NULL); + mapped_named_bo = GL_TRUE; + } + + src_ptr = ADD_POINTERS(mesa_ind_buf->obj->Pointer, mesa_ind_buf->ptr); + + const GLuint size = mesa_ind_buf->count * getTypeSize(mesa_ind_buf->type); + + radeonAllocDmaRegion(&context->radeon, &context->ind_buf.bo, + &context->ind_buf.bo_offset, size, 4); + radeon_bo_map(context->ind_buf.bo, 1); + assert(context->ind_buf.bo->ptr != NULL); + dst_ptr = ADD_POINTERS(context->ind_buf.bo->ptr, context->ind_buf.bo_offset); + + _mesa_memcpy(dst_ptr, src_ptr, size); + + radeon_bo_unmap(context->ind_buf.bo); + context->ind_buf.is_32bit = (mesa_ind_buf->type == GL_UNSIGNED_INT); + context->ind_buf.count = mesa_ind_buf->count; + + if (mapped_named_bo) + { + ctx->Driver.UnmapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER, mesa_ind_buf->obj); + } + } + else + { + r700FixupIndexBuffer(ctx, mesa_ind_buf); + } +} + +static GLboolean r700TryDrawPrims(GLcontext *ctx, + const struct gl_client_array *arrays[], + const struct _mesa_prim *prim, + GLuint nr_prims, + const struct _mesa_index_buffer *ib, + GLuint min_index, + GLuint max_index ) { context_t *context = R700_CONTEXT(ctx); radeonContextPtr radeon = &context->radeon; - unsigned int i, id = 0; - TNLcontext *tnl = TNL_CONTEXT(ctx); - struct vertex_buffer *vb = &tnl->vb; + GLuint i, id = 0; struct radeon_renderbuffer *rrb; - radeon_print(RADEON_RENDER, RADEON_NORMAL, "%s: cs begin at %d\n", - __func__, context->radeon.cmdbuf.cs->cdw); + if (ctx->NewState) + _mesa_update_state( ctx ); + + _tnl_UpdateFixedFunctionProgram(ctx); + r700SetVertexFormat(ctx, arrays, max_index + 1); + /* shaders need to be updated before buffers are validated */ + r700UpdateShaders(ctx); + if (!r600ValidateBuffers(ctx)) + return GL_FALSE; /* always emit CB base to prevent * lock ups on some chips. @@ -367,21 +908,29 @@ static GLboolean r700RunRender(GLcontext * ctx, r700SetScissor(context); r700SetupVertexProgram(ctx); r700SetupFragmentProgram(ctx); - r600UpdateTextureState(ctx); + r700UpdateShaderStates(ctx); - GLuint emit_end = r700PredictRenderSize(ctx) - + context->radeon.cmdbuf.cs->cdw; - r700SetupStreams(ctx); + GLuint emit_end = r700PredictRenderSize(ctx, prim, ib, nr_prims) + + context->radeon.cmdbuf.cs->cdw; + + r700SetupIndexBuffer(ctx, ib); + r700SetupStreams(ctx, arrays, max_index + 1); radeonEmitState(radeon); radeon_debug_add_indent(); - /* richard test code */ - for (i = 0; i < vb->PrimitiveCount; i++) { - GLuint prim = _tnl_translate_prim(&vb->Primitive[i]); - GLuint start = vb->Primitive[i].start; - GLuint end = vb->Primitive[i].start + vb->Primitive[i].count; - r700RunRenderPrimitive(ctx, start, end, prim); + for (i = 0; i < nr_prims; ++i) + { + if (context->ind_buf.bo) + r700RunRenderPrimitive(ctx, + prim[i].start, + prim[i].start + prim[i].count, + prim[i].mode); + else + r700RunRenderPrimitiveImmediate(ctx, + prim[i].start, + prim[i].start + prim[i].count, + prim[i].mode); } radeon_debug_remove_indent(); @@ -398,83 +947,54 @@ static GLboolean r700RunRender(GLcontext * ctx, r700SyncSurf(context, rrb->bo, 0, RADEON_GEM_DOMAIN_VRAM, DB_ACTION_ENA_bit | DB_DEST_BASE_ENA_bit); - radeonReleaseArrays(ctx, ~0); - - radeon_print(RADEON_RENDER, RADEON_TRACE, "%s: cs end at %d\n", - __func__, context->radeon.cmdbuf.cs->cdw); + r700FreeData(ctx); - if ( emit_end < context->radeon.cmdbuf.cs->cdw ) - WARN_ONCE("Rendering was %d commands larger than predicted size." - " We might overflow command buffer.\n", context->radeon.cmdbuf.cs->cdw - emit_end); - - return GL_FALSE; -} + if (emit_end < context->radeon.cmdbuf.cs->cdw) + { + WARN_ONCE("Rendering was %d commands larger than predicted size." + " We might overflow command buffer.\n", context->radeon.cmdbuf.cs->cdw - emit_end); + } -static GLboolean r700RunNonTCLRender(GLcontext * ctx, - struct tnl_pipeline_stage *stage) /* -------------------- */ -{ - GLboolean bRet = GL_TRUE; - - return bRet; + return GL_TRUE; } -static GLboolean r700RunTCLRender(GLcontext * ctx, /*----------------------*/ - struct tnl_pipeline_stage *stage) +static void r700DrawPrims(GLcontext *ctx, + const struct gl_client_array *arrays[], + const struct _mesa_prim *prim, + GLuint nr_prims, + const struct _mesa_index_buffer *ib, + GLboolean index_bounds_valid, + GLuint min_index, + GLuint max_index) { - GLboolean bRet = GL_FALSE; + GLboolean retval = GL_FALSE; - /* TODO : sw fallback */ - - /* Need shader bo's setup before bo check */ - r700UpdateShaders(ctx); - /** + /* This check should get folded into just the places that + * min/max index are really needed. + */ + if (!index_bounds_valid) { + vbo_get_minmax_index(ctx, prim, ib, &min_index, &max_index); + } - * Ensure all enabled and complete textures are uploaded along with any buffers being used. - */ - if(!r600ValidateBuffers(ctx)) - { - return GL_TRUE; - } + if (min_index) { + vbo_rebase_prims( ctx, arrays, prim, nr_prims, ib, min_index, max_index, r700DrawPrims ); + return; + } - bRet = r700RunRender(ctx, stage); + /* Make an attempt at drawing */ + retval = r700TryDrawPrims(ctx, arrays, prim, nr_prims, ib, min_index, max_index); - return bRet; - //GL_FALSE will stop to do other pipe stage in _tnl_run_pipeline - //The render here DOES finish the whole pipe, so GL_FALSE should be returned for success. + /* If failed run tnl pipeline - it should take care of fallbacks */ + if (!retval) + _tnl_draw_prims(ctx, arrays, prim, nr_prims, ib, min_index, max_index); } -const struct tnl_pipeline_stage _r700_render_stage = { - "r700 Hardware Rasterization", - NULL, - NULL, - NULL, - NULL, - r700RunNonTCLRender -}; - -const struct tnl_pipeline_stage _r700_tcl_stage = { - "r700 Hardware Transform, Clipping and Lighting", - NULL, - NULL, - NULL, - NULL, - r700RunTCLRender -}; - -const struct tnl_pipeline_stage *r700_pipeline[] = +void r700InitDraw(GLcontext *ctx) { - &_r700_tcl_stage, - &_tnl_vertex_transform_stage, - &_tnl_normal_transform_stage, - &_tnl_lighting_stage, - &_tnl_fog_coordinate_stage, - &_tnl_texgen_stage, - &_tnl_texture_transform_stage, - &_tnl_vertex_program_stage, - - &_r700_render_stage, - &_tnl_render_stage, - 0, -}; + struct vbo_context *vbo = vbo_context(ctx); + + /* to be enabled */ + vbo->draw_prims = r700DrawPrims; +} diff --git a/src/mesa/drivers/dri/r600/r700_shader.c b/src/mesa/drivers/dri/r600/r700_shader.c index b4fd51c1370..2eed1acc2f5 100644 --- a/src/mesa/drivers/dri/r600/r700_shader.c +++ b/src/mesa/drivers/dri/r600/r700_shader.c @@ -60,6 +60,55 @@ void AddInstToList(TypedShaderList * plstCFInstructions, R700ShaderInstruction * plstCFInstructions->uNumOfNode++; } +void TakeInstOutFromList(TypedShaderList * plstCFInstructions, R700ShaderInstruction * pInst) +{ + GLuint ulIndex = 0; + GLboolean bFound = GL_FALSE; + R700ShaderInstruction * pPrevInst = NULL; + R700ShaderInstruction * pCurInst = plstCFInstructions->pHead; + + /* Need go thro list to make sure pInst is there. */ + while(NULL != pCurInst) + { + if(pCurInst == pInst) + { + bFound = GL_TRUE; + break; + } + + pPrevInst = pCurInst; + pCurInst = pCurInst->pNextInst; + } + if(GL_TRUE == bFound) + { + plstCFInstructions->uNumOfNode--; + + pCurInst = pInst->pNextInst; + ulIndex = pInst->m_uIndex; + while(NULL != pCurInst) + { + pCurInst->m_uIndex = ulIndex; + ulIndex++; + pCurInst = pCurInst->pNextInst; + } + + if(plstCFInstructions->pHead == pInst) + { + plstCFInstructions->pHead = pInst->pNextInst; + } + if(plstCFInstructions->pTail == pInst) + { + plstCFInstructions->pTail = pPrevInst; + } + if(NULL != pPrevInst) + { + pPrevInst->pNextInst = pInst->pNextInst; + } + + FREE(pInst); + } +} + void Init_R700_Shader(R700_Shader * pShader) { pShader->Type = R700_SHADER_INVALID; @@ -110,13 +159,18 @@ void Init_R700_Shader(R700_Shader * pShader) pShader->lstVTXInstructions.uNumOfNode=0; } +void SetActiveCFlist(R700_Shader *pShader, TypedShaderList * plstCF) +{ + pShader->plstCFInstructions_active = plstCF; +} + void AddCFInstruction(R700_Shader *pShader, R700ControlFlowInstruction *pCFInst) { R700ControlFlowSXClause* pSXClause; R700ControlFlowSMXClause* pSMXClause; - pCFInst->m_uIndex = pShader->lstCFInstructions.uNumOfNode; - AddInstToList(&(pShader->lstCFInstructions), + pCFInst->m_uIndex = pShader->plstCFInstructions_active->uNumOfNode; + AddInstToList(pShader->plstCFInstructions_active, (R700ShaderInstruction*)pCFInst); pShader->uShaderBinaryDWORDSize += GetInstructionSize(pCFInst->m_ShaderInstType); @@ -488,6 +542,47 @@ void DebugPrint(void) { } +void cleanup_vfetch_shaderinst(R700_Shader *pShader) +{ + R700ShaderInstruction *pInst; + R700ShaderInstruction *pInstToFree; + R700VertexInstruction *pVTXInst; + R700ControlFlowInstruction *pCFInst; + + pInst = pShader->lstVTXInstructions.pHead; + while(NULL != pInst) + { + pVTXInst = (R700VertexInstruction *)pInst; + pShader->uShaderBinaryDWORDSize -= GetInstructionSize(pVTXInst->m_ShaderInstType); + + if(NULL != pVTXInst->m_pLinkedGenericClause) + { + pCFInst = (R700ControlFlowInstruction*)(pVTXInst->m_pLinkedGenericClause); + + TakeInstOutFromList(&(pShader->lstCFInstructions), + (R700ShaderInstruction*)pCFInst); + + pShader->uShaderBinaryDWORDSize -= GetInstructionSize(pCFInst->m_ShaderInstType); + } + + pInst = pInst->pNextInst; + }; + + //destroy each item in pShader->lstVTXInstructions; + pInst = pShader->lstVTXInstructions.pHead; + while(NULL != pInst) + { + pInstToFree = pInst; + pInst = pInst->pNextInst; + FREE(pInstToFree); + }; + + //set NULL pShader->lstVTXInstructions + pShader->lstVTXInstructions.pHead=NULL; + pShader->lstVTXInstructions.pTail=NULL; + pShader->lstVTXInstructions.uNumOfNode=0; +} + void Clean_Up_Shader(R700_Shader *pShader) { FREE(pShader->pProgram); diff --git a/src/mesa/drivers/dri/r600/r700_shader.h b/src/mesa/drivers/dri/r600/r700_shader.h index bfd01e1a93a..0599ffd901f 100644 --- a/src/mesa/drivers/dri/r600/r700_shader.h +++ b/src/mesa/drivers/dri/r600/r700_shader.h @@ -109,6 +109,7 @@ typedef struct R700_Shader GLuint uStackSize; GLuint uMaxCallDepth; + TypedShaderList * plstCFInstructions_active; TypedShaderList lstCFInstructions; TypedShaderList lstALUInstructions; TypedShaderList lstTEXInstructions; @@ -128,21 +129,23 @@ typedef struct R700_Shader //Internal void AddInstToList(TypedShaderList * plstCFInstructions, R700ShaderInstruction * pInst); +void TakeInstOutFromList(TypedShaderList * plstCFInstructions, R700ShaderInstruction * pInst); void ResolveLinks(R700_Shader *pShader); void Assemble(R700_Shader *pShader); - //Interface void Init_R700_Shader(R700_Shader * pShader); void AddCFInstruction(R700_Shader *pShader, R700ControlFlowInstruction *pCFInst); void AddVTXInstruction(R700_Shader *pShader, R700VertexInstruction *pVTXInst); void AddTEXInstruction(R700_Shader *pShader, R700TextureInstruction *pTEXInst); void AddALUInstruction(R700_Shader *pShader, R700ALUInstruction *pALUInst); +void SetActiveCFlist(R700_Shader *pShader, TypedShaderList * plstCF); void LoadProgram(R700_Shader *pShader); void UpdateShaderRegisters(R700_Shader *pShader); void DeleteInstructions(R700_Shader *pShader); void DebugPrint(void); +void cleanup_vfetch_shaderinst(R700_Shader *pShader); void Clean_Up_Shader(R700_Shader *pShader); diff --git a/src/mesa/drivers/dri/r600/r700_state.c b/src/mesa/drivers/dri/r600/r700_state.c index e91aa43118f..d7420678ff0 100644 --- a/src/mesa/drivers/dri/r600/r700_state.c +++ b/src/mesa/drivers/dri/r600/r700_state.c @@ -46,7 +46,6 @@ #include "shader/prog_parameter.h" #include "shader/prog_statevars.h" #include "vbo/vbo.h" -#include "main/texformat.h" #include "r600_context.h" @@ -55,18 +54,15 @@ #include "r700_fragprog.h" #include "r700_vertprog.h" - +void r600UpdateTextureState(GLcontext * ctx); static void r700SetClipPlaneState(GLcontext * ctx, GLenum cap, GLboolean state); static void r700UpdatePolygonMode(GLcontext * ctx); static void r700SetPolygonOffsetState(GLcontext * ctx, GLboolean state); static void r700SetStencilState(GLcontext * ctx, GLboolean state); -void r700UpdateShaders (GLcontext * ctx) //---------------------------------- +void r700UpdateShaders(GLcontext * ctx) { context_t *context = R700_CONTEXT(ctx); - GLvector4f dummy_attrib[_TNL_ATTRIB_MAX]; - GLvector4f *temp_attrib[_TNL_ATTRIB_MAX]; - int i; /* should only happenen once, just after context is created */ /* TODO: shouldn't we fallback to sw here? */ @@ -77,21 +73,6 @@ void r700UpdateShaders (GLcontext * ctx) //---------------------------------- r700SelectFragmentShader(ctx); - if (context->radeon.NewGLState) { - for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) { - /* mat states from state var not array for sw */ - dummy_attrib[i].stride = 0; - temp_attrib[i] = TNL_CONTEXT(ctx)->vb.AttribPtr[i]; - TNL_CONTEXT(ctx)->vb.AttribPtr[i] = &(dummy_attrib[i]); - } - - _tnl_UpdateFixedFunctionProgram(ctx); - - for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) { - TNL_CONTEXT(ctx)->vb.AttribPtr[i] = temp_attrib[i]; - } - } - r700SelectVertexShader(ctx); r700UpdateStateParameters(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS); context->radeon.NewGLState = 0; @@ -210,6 +191,67 @@ static void r700InvalidateState(GLcontext * ctx, GLuint new_state) //----------- context->radeon.NewGLState |= new_state; } +static void r700SetDBRenderState(GLcontext * ctx) +{ + context_t *context = R700_CONTEXT(ctx); + R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); + struct r700_fragment_program *fp = (struct r700_fragment_program *) + (ctx->FragmentProgram._Current); + + R600_STATECHANGE(context, db); + + SETbit(r700->DB_SHADER_CONTROL.u32All, DUAL_EXPORT_ENABLE_bit); + SETfield(r700->DB_SHADER_CONTROL.u32All, EARLY_Z_THEN_LATE_Z, Z_ORDER_shift, Z_ORDER_mask); + /* XXX need to enable htile for hiz/s */ + SETfield(r700->DB_RENDER_OVERRIDE.u32All, FORCE_DISABLE, FORCE_HIZ_ENABLE_shift, FORCE_HIZ_ENABLE_mask); + SETfield(r700->DB_RENDER_OVERRIDE.u32All, FORCE_DISABLE, FORCE_HIS_ENABLE0_shift, FORCE_HIS_ENABLE0_mask); + SETfield(r700->DB_RENDER_OVERRIDE.u32All, FORCE_DISABLE, FORCE_HIS_ENABLE1_shift, FORCE_HIS_ENABLE1_mask); + + if (context->radeon.query.current) + { + SETbit(r700->DB_RENDER_OVERRIDE.u32All, NOOP_CULL_DISABLE_bit); + if (context->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV770) + { + SETbit(r700->DB_RENDER_CONTROL.u32All, PERFECT_ZPASS_COUNTS_bit); + } + } + else + { + CLEARbit(r700->DB_RENDER_OVERRIDE.u32All, NOOP_CULL_DISABLE_bit); + if (context->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV770) + { + CLEARbit(r700->DB_RENDER_CONTROL.u32All, PERFECT_ZPASS_COUNTS_bit); + } + } + + if (fp) + { + if (fp->r700Shader.killIsUsed) + { + SETbit(r700->DB_SHADER_CONTROL.u32All, KILL_ENABLE_bit); + } + else + { + CLEARbit(r700->DB_SHADER_CONTROL.u32All, KILL_ENABLE_bit); + } + + if (fp->r700Shader.depthIsExported) + { + SETbit(r700->DB_SHADER_CONTROL.u32All, Z_EXPORT_ENABLE_bit); + } + else + { + CLEARbit(r700->DB_SHADER_CONTROL.u32All, Z_EXPORT_ENABLE_bit); + } + } +} + +void r700UpdateShaderStates(GLcontext * ctx) +{ + r700SetDBRenderState(ctx); + r600UpdateTextureState(ctx); +} + static void r700SetDepthState(GLcontext * ctx) { context_t *context = R700_CONTEXT(ctx); @@ -475,10 +517,10 @@ static void r700SetBlendState(GLcontext * ctx) eqn, COLOR_COMB_FCN_shift, COLOR_COMB_FCN_mask); SETfield(blend_reg, - blend_factor(ctx->Color.BlendSrcRGB, GL_TRUE), + blend_factor(ctx->Color.BlendSrcA, GL_TRUE), ALPHA_SRCBLEND_shift, ALPHA_SRCBLEND_mask); SETfield(blend_reg, - blend_factor(ctx->Color.BlendDstRGB, GL_FALSE), + blend_factor(ctx->Color.BlendDstA, GL_FALSE), ALPHA_DESTBLEND_shift, ALPHA_DESTBLEND_mask); switch (ctx->Color.BlendEquationA) { @@ -753,9 +795,9 @@ static void r700ColorMask(GLcontext * ctx, (b ? 4 : 0) | (a ? 8 : 0)); - if (mask != r700->CB_SHADER_MASK.u32All) { + if (mask != r700->CB_TARGET_MASK.u32All) { R600_STATECHANGE(context, cb); - SETfield(r700->CB_SHADER_MASK.u32All, mask, OUTPUT0_ENABLE_shift, OUTPUT0_ENABLE_mask); + SETfield(r700->CB_TARGET_MASK.u32All, mask, TARGET0_ENABLE_shift, TARGET0_ENABLE_mask); } } @@ -845,9 +887,9 @@ static void r700PointSize(GLcontext * ctx, GLfloat size) size = CLAMP(size, ctx->Const.MinPointSize, ctx->Const.MaxPointSize); /* format is 12.4 fixed point */ - SETfield(r700->PA_SU_POINT_SIZE.u32All, (int)(size * 16), + SETfield(r700->PA_SU_POINT_SIZE.u32All, (int)(size * 8.0), PA_SU_POINT_SIZE__HEIGHT_shift, PA_SU_POINT_SIZE__HEIGHT_mask); - SETfield(r700->PA_SU_POINT_SIZE.u32All, (int)(size * 16), + SETfield(r700->PA_SU_POINT_SIZE.u32All, (int)(size * 8.0), PA_SU_POINT_SIZE__WIDTH_shift, PA_SU_POINT_SIZE__WIDTH_mask); } @@ -862,11 +904,11 @@ static void r700PointParameter(GLcontext * ctx, GLenum pname, const GLfloat * pa /* format is 12.4 fixed point */ switch (pname) { case GL_POINT_SIZE_MIN: - SETfield(r700->PA_SU_POINT_MINMAX.u32All, (int)(ctx->Point.MinSize * 16.0), + SETfield(r700->PA_SU_POINT_MINMAX.u32All, (int)(ctx->Point.MinSize * 8.0), MIN_SIZE_shift, MIN_SIZE_mask); break; case GL_POINT_SIZE_MAX: - SETfield(r700->PA_SU_POINT_MINMAX.u32All, (int)(ctx->Point.MaxSize * 16.0), + SETfield(r700->PA_SU_POINT_MINMAX.u32All, (int)(ctx->Point.MaxSize * 8.0), MAX_SIZE_shift, MAX_SIZE_mask); break; case GL_POINT_DISTANCE_ATTENUATION: @@ -1049,6 +1091,7 @@ static void r700UpdateWindow(GLcontext * ctx, int id) //-------------------- GLfloat tz = v[MAT_TZ] * depthScale; R600_STATECHANGE(context, vpt); + R600_STATECHANGE(context, cl); r700->viewport[id].PA_CL_VPORT_XSCALE.f32All = sx; r700->viewport[id].PA_CL_VPORT_XOFFSET.f32All = tx; @@ -1059,6 +1102,18 @@ static void r700UpdateWindow(GLcontext * ctx, int id) //-------------------- r700->viewport[id].PA_CL_VPORT_ZSCALE.f32All = sz; r700->viewport[id].PA_CL_VPORT_ZOFFSET.f32All = tz; + if (ctx->Transform.DepthClamp) { + r700->viewport[id].PA_SC_VPORT_ZMIN_0.f32All = MIN2(ctx->Viewport.Near, ctx->Viewport.Far); + r700->viewport[id].PA_SC_VPORT_ZMAX_0.f32All = MAX2(ctx->Viewport.Near, ctx->Viewport.Far); + SETbit(r700->PA_CL_CLIP_CNTL.u32All, ZCLIP_NEAR_DISABLE_bit); + SETbit(r700->PA_CL_CLIP_CNTL.u32All, ZCLIP_FAR_DISABLE_bit); + } else { + r700->viewport[id].PA_SC_VPORT_ZMIN_0.f32All = 0.0; + r700->viewport[id].PA_SC_VPORT_ZMAX_0.f32All = 1.0; + CLEARbit(r700->PA_CL_CLIP_CNTL.u32All, ZCLIP_NEAR_DISABLE_bit); + CLEARbit(r700->PA_CL_CLIP_CNTL.u32All, ZCLIP_FAR_DISABLE_bit); + } + r700->viewport[id].enabled = GL_TRUE; r700SetScissor(context); @@ -1130,20 +1185,25 @@ static void r700PolygonOffset(GLcontext * ctx, GLfloat factor, GLfloat units) // context_t *context = R700_CONTEXT(ctx); R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw); GLfloat constant = units; + GLchar depth = 0; + + R600_STATECHANGE(context, poly); switch (ctx->Visual.depthBits) { case 16: constant *= 4.0; + depth = -16; break; case 24: constant *= 2.0; + depth = -24; break; } factor *= 12.0; - - R600_STATECHANGE(context, poly); - + SETfield(r700->PA_SU_POLY_OFFSET_DB_FMT_CNTL.u32All, depth, + POLY_OFFSET_NEG_NUM_DB_BITS_shift, POLY_OFFSET_NEG_NUM_DB_BITS_mask); + //r700->PA_SU_POLY_OFFSET_CLAMP.f32All = constant; //??? r700->PA_SU_POLY_OFFSET_FRONT_SCALE.f32All = factor; r700->PA_SU_POLY_OFFSET_FRONT_OFFSET.f32All = constant; r700->PA_SU_POLY_OFFSET_BACK_SCALE.f32All = factor; @@ -1276,6 +1336,11 @@ void r700SetScissor(context_t *context) //--------------- y1 = context->radeon.state.scissor.rect.y1; x2 = context->radeon.state.scissor.rect.x2; y2 = context->radeon.state.scissor.rect.y2; + /* r600 has exclusive BR scissors */ + if (context->radeon.radeonScreen->kernel_mm) { + x2++; + y2++; + } } else { if (context->radeon.radeonScreen->driScreen->dri2.enabled) { x1 = 0; @@ -1354,8 +1419,6 @@ void r700SetScissor(context_t *context) //--------------- SETfield(r700->viewport[id].PA_SC_VPORT_SCISSOR_0_BR.u32All, y2, PA_SC_VPORT_SCISSOR_0_BR__BR_Y_shift, PA_SC_VPORT_SCISSOR_0_BR__BR_Y_mask); - r700->viewport[id].PA_SC_VPORT_ZMIN_0.u32All = 0; - r700->viewport[id].PA_SC_VPORT_ZMAX_0.u32All = 0x3F800000; r700->viewport[id].enabled = GL_TRUE; } @@ -1670,19 +1733,10 @@ void r700InitState(GLcontext * ctx) //------------------- r700Enable(ctx, GL_DEPTH_TEST, ctx->Depth.Test); r700DepthMask(ctx, ctx->Depth.Mask); r700DepthFunc(ctx, ctx->Depth.Func); - SETbit(r700->DB_SHADER_CONTROL.u32All, DUAL_EXPORT_ENABLE_bit); - r700->DB_DEPTH_CLEAR.u32All = 0x3F800000; - - r700->DB_RENDER_CONTROL.u32All = 0; SETbit(r700->DB_RENDER_CONTROL.u32All, STENCIL_COMPRESS_DISABLE_bit); SETbit(r700->DB_RENDER_CONTROL.u32All, DEPTH_COMPRESS_DISABLE_bit); - r700->DB_RENDER_OVERRIDE.u32All = 0; - if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770) - SETbit(r700->DB_RENDER_OVERRIDE.u32All, FORCE_SHADER_Z_ORDER_bit); - SETfield(r700->DB_RENDER_OVERRIDE.u32All, FORCE_DISABLE, FORCE_HIZ_ENABLE_shift, FORCE_HIZ_ENABLE_mask); - SETfield(r700->DB_RENDER_OVERRIDE.u32All, FORCE_DISABLE, FORCE_HIS_ENABLE0_shift, FORCE_HIS_ENABLE0_mask); - SETfield(r700->DB_RENDER_OVERRIDE.u32All, FORCE_DISABLE, FORCE_HIS_ENABLE1_shift, FORCE_HIS_ENABLE1_mask); + r700SetDBRenderState(ctx); r700->DB_ALPHA_TO_MASK.u32All = 0; SETfield(r700->DB_ALPHA_TO_MASK.u32All, 2, ALPHA_TO_MASK_OFFSET0_shift, ALPHA_TO_MASK_OFFSET0_mask); @@ -1756,7 +1810,7 @@ void r700InitState(GLcontext * ctx) //------------------- r700->CB_CLRCMP_MSK.u32All = 0xFFFFFFFF; /* screen/window/view */ - SETfield(r700->CB_TARGET_MASK.u32All, 0xF, (4 * id), TARGET0_ENABLE_mask); + SETfield(r700->CB_SHADER_MASK.u32All, 0xF, (4 * id), OUTPUT0_ENABLE_mask); context->radeon.hw.all_dirty = GL_TRUE; diff --git a/src/mesa/drivers/dri/r600/r700_state.h b/src/mesa/drivers/dri/r600/r700_state.h index 0f53d5b4c59..60c6a7f23ca 100644 --- a/src/mesa/drivers/dri/r600/r700_state.h +++ b/src/mesa/drivers/dri/r600/r700_state.h @@ -35,6 +35,7 @@ extern void r700UpdateStateParameters(GLcontext * ctx, GLuint new_state); extern void r700UpdateShaders (GLcontext * ctx); +extern void r700UpdateShaderStates(GLcontext * ctx); extern void r700UpdateViewportOffset(GLcontext * ctx); diff --git a/src/mesa/drivers/dri/r600/r700_vertprog.c b/src/mesa/drivers/dri/r600/r700_vertprog.c index 9ee26286d9b..90fac078ff0 100644 --- a/src/mesa/drivers/dri/r600/r700_vertprog.c +++ b/src/mesa/drivers/dri/r600/r700_vertprog.c @@ -111,6 +111,15 @@ unsigned int Map_Vertex_Output(r700_AssemblerBase *pAsm, } } + for(i=VERT_RESULT_VAR0; i<VERT_RESULT_MAX; i++) + { + unBit = 1 << i; + if(mesa_vp->Base.OutputsWritten & unBit) + { + pAsm->ucVP_OutputMap[i] = unTotal++; + } + } + return (unTotal - unStart); } @@ -159,7 +168,35 @@ GLboolean Process_Vertex_Program_Vfetch_Instructions( return GL_TRUE; } -void Map_Vertex_Program(struct r700_vertex_program *vp, +GLboolean Process_Vertex_Program_Vfetch_Instructions2( + GLcontext *ctx, + struct r700_vertex_program *vp, + struct gl_vertex_program *mesa_vp) +{ + int i; + context_t *context = R700_CONTEXT(ctx); + + VTX_FETCH_METHOD vtxFetchMethod; + vtxFetchMethod.bEnableMini = GL_FALSE; + vtxFetchMethod.mega_fetch_remainder = 0; + + for(i=0; i<context->nNumActiveAos; i++) + { + assemble_vfetch_instruction2(&vp->r700AsmCode, + vp->r700AsmCode.ucVP_AttributeMap[context->stream_desc[i].element], + context->stream_desc[i].type, + context->stream_desc[i].size, + context->stream_desc[i].element, + context->stream_desc[i]._signed, + context->stream_desc[i].normalize, + &vtxFetchMethod); + } + + return GL_TRUE; +} + +void Map_Vertex_Program(GLcontext *ctx, + struct r700_vertex_program *vp, struct gl_vertex_program *mesa_vp) { GLuint ui; @@ -175,10 +212,10 @@ void Map_Vertex_Program(struct r700_vertex_program *vp, pAsm->number_used_registers += num_inputs; // Create VFETCH instructions for inputs - if (GL_TRUE != Process_Vertex_Program_Vfetch_Instructions(vp, mesa_vp) ) + if (GL_TRUE != Process_Vertex_Program_Vfetch_Instructions2(ctx, vp, mesa_vp) ) { - radeon_error("Calling Process_Vertex_Program_Vfetch_Instructions return error. \n"); - return; //error + radeon_error("Calling Process_Vertex_Program_Vfetch_Instructions2 return error. \n"); + return; } // Map Outputs @@ -189,7 +226,7 @@ void Map_Vertex_Program(struct r700_vertex_program *vp, pAsm->number_used_registers += pAsm->number_of_exports; pAsm->pucOutMask = (unsigned char*) MALLOC(pAsm->number_of_exports); - + for(ui=0; ui<pAsm->number_of_exports; ui++) { pAsm->pucOutMask[ui] = 0x0; @@ -206,7 +243,9 @@ void Map_Vertex_Program(struct r700_vertex_program *vp, { /* fix func t_vp uses NumTemporaries */ pAsm->number_used_registers += mesa_vp->Base.NumTemporaries; } - + + pAsm->flag_reg_index = pAsm->number_used_registers++; + pAsm->uFirstHelpReg = pAsm->number_used_registers; } @@ -261,13 +300,10 @@ GLboolean Find_Instruction_Dependencies_vp(struct r700_vertex_program *vp, } struct r700_vertex_program* r700TranslateVertexShader(GLcontext *ctx, - struct gl_vertex_program *mesa_vp) + struct gl_vertex_program *mesa_vp) { context_t *context = R700_CONTEXT(ctx); struct r700_vertex_program *vp; - TNLcontext *tnl = TNL_CONTEXT(ctx); - struct vertex_buffer *vb = &tnl->vb; - unsigned int unBit; unsigned int i; vp = _mesa_calloc(sizeof(*vp)); @@ -278,17 +314,13 @@ struct r700_vertex_program* r700TranslateVertexShader(GLcontext *ctx, _mesa_insert_mvp_code(ctx, vp->mesa_program); } - for(i=0; i<VERT_ATTRIB_MAX; i++) + for(i=0; i<context->nNumActiveAos; i++) { - unBit = 1 << i; - if(vp->mesa_program->Base.InputsRead & unBit) /* ctx->Array.ArrayObj->xxxxxxx */ - { - vp->aos_desc[i].size = vb->AttribPtr[i]->size; - vp->aos_desc[i].stride = vb->AttribPtr[i]->size * sizeof(GL_FLOAT);/* when emit array, data is packed. vb->AttribPtr[i]->stride;*/ - vp->aos_desc[i].type = GL_FLOAT; - } + vp->aos_desc[i].size = context->stream_desc[i].size; + vp->aos_desc[i].stride = context->stream_desc[i].stride; + vp->aos_desc[i].type = context->stream_desc[i].type; } - + if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770) { vp->r700AsmCode.bR6xx = 1; @@ -296,25 +328,41 @@ struct r700_vertex_program* r700TranslateVertexShader(GLcontext *ctx, //Init_Program Init_r700_AssemblerBase(SPT_VP, &(vp->r700AsmCode), &(vp->r700Shader) ); - Map_Vertex_Program( vp, vp->mesa_program ); + Map_Vertex_Program(ctx, vp, vp->mesa_program ); if(GL_FALSE == Find_Instruction_Dependencies_vp(vp, vp->mesa_program)) { return NULL; + } + + InitShaderProgram(&(vp->r700AsmCode)); + + for(i=0; i < MAX_SAMPLERS; i++) + { + vp->r700AsmCode.SamplerUnits[i] = vp->mesa_program->Base.SamplerUnits[i]; } - if(GL_FALSE == AssembleInstr(vp->mesa_program->Base.NumInstructions, - &(vp->mesa_program->Base.Instructions[0]), + vp->r700AsmCode.unCurNumILInsts = vp->mesa_program->Base.NumInstructions; + + if(GL_FALSE == AssembleInstr(0, + 0, + vp->mesa_program->Base.NumInstructions, + &(vp->mesa_program->Base.Instructions[0]), &(vp->r700AsmCode)) ) { return NULL; - } + } if(GL_FALSE == Process_Vertex_Exports(&(vp->r700AsmCode), vp->mesa_program->Base.OutputsWritten) ) { return NULL; } + if( GL_FALSE == RelocProgram(&(vp->r700AsmCode), &(vp->mesa_program->Base)) ) + { + return GL_FALSE; + } + vp->r700Shader.nRegs = (vp->r700AsmCode.number_used_registers == 0) ? 0 : (vp->r700AsmCode.number_used_registers - 1); @@ -330,9 +378,6 @@ void r700SelectVertexShader(GLcontext *ctx) context_t *context = R700_CONTEXT(ctx); struct r700_vertex_program_cont *vpc; struct r700_vertex_program *vp; - TNLcontext *tnl = TNL_CONTEXT(ctx); - struct vertex_buffer *vb = &tnl->vb; - unsigned int unBit; unsigned int i; GLboolean match; GLbitfield InputsRead; @@ -343,29 +388,27 @@ void r700SelectVertexShader(GLcontext *ctx) if (vpc->mesa_program.IsPositionInvariant) { InputsRead |= VERT_BIT_POS; - } - + } + for (vp = vpc->progs; vp; vp = vp->next) { - match = GL_TRUE; - for(i=0; i<VERT_ATTRIB_MAX; i++) + match = GL_TRUE; + for(i=0; i<context->nNumActiveAos; i++) { - unBit = 1 << i; - if(InputsRead & unBit) + if (vp->aos_desc[i].size != context->stream_desc[i].size) { - if (vp->aos_desc[i].size != vb->AttribPtr[i]->size) - match = GL_FALSE; - break; + match = GL_FALSE; + break; } } - if (match) + if (match) { context->selected_vp = vp; return; } } - vp = r700TranslateVertexShader(ctx, &(vpc->mesa_program) ); + vp = r700TranslateVertexShader(ctx, &(vpc->mesa_program)); if(!vp) { radeon_error("Failed to translate vertex shader. \n"); @@ -377,6 +420,146 @@ void r700SelectVertexShader(GLcontext *ctx) return; } +int getTypeSize(GLenum type) +{ + switch (type) + { + case GL_DOUBLE: + return sizeof(GLdouble); + case GL_FLOAT: + return sizeof(GLfloat); + case GL_INT: + return sizeof(GLint); + case GL_UNSIGNED_INT: + return sizeof(GLuint); + case GL_SHORT: + return sizeof(GLshort); + case GL_UNSIGNED_SHORT: + return sizeof(GLushort); + case GL_BYTE: + return sizeof(GLbyte); + case GL_UNSIGNED_BYTE: + return sizeof(GLubyte); + default: + assert(0); + return 0; + } +} + +static void r700TranslateAttrib(GLcontext *ctx, GLuint unLoc, int count, const struct gl_client_array *input) +{ + context_t *context = R700_CONTEXT(ctx); + + StreamDesc * pStreamDesc = &(context->stream_desc[context->nNumActiveAos]); + + GLuint stride; + + stride = (input->StrideB == 0) ? getTypeSize(input->Type) * input->Size + : input->StrideB; + + if (input->Type == GL_DOUBLE || input->Type == GL_UNSIGNED_INT || input->Type == GL_INT || +#if MESA_BIG_ENDIAN + getTypeSize(input->Type) != 4 || +#endif + stride < 4) + { + pStreamDesc->type = GL_FLOAT; + + if (input->StrideB == 0) + { + pStreamDesc->stride = 0; + } + else + { + pStreamDesc->stride = sizeof(GLfloat) * input->Size; + } + pStreamDesc->dwords = input->Size; + pStreamDesc->is_named_bo = GL_FALSE; + } + else + { + pStreamDesc->type = input->Type; + pStreamDesc->dwords = (getTypeSize(input->Type) * input->Size + 3)/ 4; + if (!input->BufferObj->Name) + { + if (input->StrideB == 0) + { + pStreamDesc->stride = 0; + } + else + { + pStreamDesc->stride = (getTypeSize(pStreamDesc->type) * input->Size + 3) & ~3; + } + + pStreamDesc->is_named_bo = GL_FALSE; + } + } + + pStreamDesc->size = input->Size; + pStreamDesc->dst_loc = context->nNumActiveAos; + pStreamDesc->element = unLoc; + + switch (pStreamDesc->type) + { //GetSurfaceFormat + case GL_FLOAT: + pStreamDesc->_signed = 0; + pStreamDesc->normalize = GL_FALSE; + break; + case GL_SHORT: + pStreamDesc->_signed = 1; + pStreamDesc->normalize = input->Normalized; + break; + case GL_BYTE: + pStreamDesc->_signed = 1; + pStreamDesc->normalize = input->Normalized; + break; + case GL_UNSIGNED_SHORT: + pStreamDesc->_signed = 0; + pStreamDesc->normalize = input->Normalized; + break; + case GL_UNSIGNED_BYTE: + pStreamDesc->_signed = 0; + pStreamDesc->normalize = input->Normalized; + break; + default: + case GL_INT: + case GL_UNSIGNED_INT: + case GL_DOUBLE: + assert(0); + break; + } + context->nNumActiveAos++; +} + +void r700SetVertexFormat(GLcontext *ctx, const struct gl_client_array *arrays[], int count) +{ + context_t *context = R700_CONTEXT(ctx); + struct r700_vertex_program *vpc + = (struct r700_vertex_program *)ctx->VertexProgram._Current; + + struct gl_vertex_program * mesa_vp = (struct gl_vertex_program *)&(vpc->mesa_program); + unsigned int unLoc = 0; + unsigned int unBit = mesa_vp->Base.InputsRead; + context->nNumActiveAos = 0; + + if (mesa_vp->IsPositionInvariant) + { + unBit |= VERT_BIT_POS; + } + + while(unBit) + { + if(unBit & 1) + { + r700TranslateAttrib(ctx, unLoc, count, arrays[unLoc]); + } + + unBit >>= 1; + ++unLoc; + } + context->radeon.tcl.aos_count = context->nNumActiveAos; +} + void * r700GetActiveVpShaderBo(GLcontext * ctx) { context_t *context = R700_CONTEXT(ctx); @@ -456,6 +639,12 @@ GLboolean r700SetupVertexProgram(GLcontext * ctx) paramList = vp->mesa_program->Base.Parameters; if(NULL != paramList) { + /* vp->mesa_program was cloned, not updated by glsl shader api. */ + /* _mesa_reference_program has already checked glsl shProg is ok and set ctx->VertexProgem._Current */ + /* so, use ctx->VertexProgem._Current */ + struct gl_program_parameter_list *paramListOrginal = + paramListOrginal = ctx->VertexProgram._Current->Base.Parameters; + _mesa_load_state_parameters(ctx, paramList); if (paramList->NumParameters > R700_MAX_DX9_CONSTS) @@ -468,13 +657,42 @@ GLboolean r700SetupVertexProgram(GLcontext * ctx) unNumParamData = paramList->NumParameters; for(ui=0; ui<unNumParamData; ui++) { - r700->vs.consts[ui][0].f32All = paramList->ParameterValues[ui][0]; - r700->vs.consts[ui][1].f32All = paramList->ParameterValues[ui][1]; - r700->vs.consts[ui][2].f32All = paramList->ParameterValues[ui][2]; - r700->vs.consts[ui][3].f32All = paramList->ParameterValues[ui][3]; + if(paramList->Parameters[ui].Type == PROGRAM_UNIFORM) + { + r700->vs.consts[ui][0].f32All = paramListOrginal->ParameterValues[ui][0]; + r700->vs.consts[ui][1].f32All = paramListOrginal->ParameterValues[ui][1]; + r700->vs.consts[ui][2].f32All = paramListOrginal->ParameterValues[ui][2]; + r700->vs.consts[ui][3].f32All = paramListOrginal->ParameterValues[ui][3]; + } + else + { + r700->vs.consts[ui][0].f32All = paramList->ParameterValues[ui][0]; + r700->vs.consts[ui][1].f32All = paramList->ParameterValues[ui][1]; + r700->vs.consts[ui][2].f32All = paramList->ParameterValues[ui][2]; + r700->vs.consts[ui][3].f32All = paramList->ParameterValues[ui][3]; + } } } else r700->vs.num_consts = 0; + COMPILED_SUB * pCompiledSub; + GLuint uj; + GLuint unConstOffset = r700->vs.num_consts; + for(ui=0; ui<vp->r700AsmCode.unNumPresub; ui++) + { + pCompiledSub = vp->r700AsmCode.presubs[ui].pCompiledSub; + + r700->vs.num_consts += pCompiledSub->NumParameters; + + for(uj=0; uj<pCompiledSub->NumParameters; uj++) + { + r700->vs.consts[uj + unConstOffset][0].f32All = pCompiledSub->ParameterValues[uj][0]; + r700->vs.consts[uj + unConstOffset][1].f32All = pCompiledSub->ParameterValues[uj][1]; + r700->vs.consts[uj + unConstOffset][2].f32All = pCompiledSub->ParameterValues[uj][2]; + r700->vs.consts[uj + unConstOffset][3].f32All = pCompiledSub->ParameterValues[uj][3]; + } + unConstOffset += pCompiledSub->NumParameters; + } + return GL_TRUE; } diff --git a/src/mesa/drivers/dri/r600/r700_vertprog.h b/src/mesa/drivers/dri/r600/r700_vertprog.h index c48764c43ba..00824c29d32 100644 --- a/src/mesa/drivers/dri/r600/r700_vertprog.h +++ b/src/mesa/drivers/dri/r600/r700_vertprog.h @@ -52,8 +52,7 @@ struct r700_vertex_program GLboolean translated; GLboolean loaded; - GLboolean needUpdateVF; - + void * shaderbo; ArrayDesc aos_desc[VERT_ATTRIB_MAX]; @@ -76,19 +75,27 @@ unsigned int Map_Vertex_Input(r700_AssemblerBase *pAsm, GLboolean Process_Vertex_Program_Vfetch_Instructions( struct r700_vertex_program *vp, struct gl_vertex_program *mesa_vp); -void Map_Vertex_Program(struct r700_vertex_program *vp, +GLboolean Process_Vertex_Program_Vfetch_Instructions2( + GLcontext *ctx, + struct r700_vertex_program *vp, + struct gl_vertex_program *mesa_vp); +void Map_Vertex_Program(GLcontext *ctx, + struct r700_vertex_program *vp, struct gl_vertex_program *mesa_vp); GLboolean Find_Instruction_Dependencies_vp(struct r700_vertex_program *vp, struct gl_vertex_program *mesa_vp); struct r700_vertex_program* r700TranslateVertexShader(GLcontext *ctx, - struct gl_vertex_program *mesa_vp); + struct gl_vertex_program *mesa_vp); /* Interface */ extern void r700SelectVertexShader(GLcontext *ctx); +extern void r700SetVertexFormat(GLcontext *ctx, const struct gl_client_array *arrays[], int count); extern GLboolean r700SetupVertexProgram(GLcontext * ctx); extern void * r700GetActiveVpShaderBo(GLcontext * ctx); +extern int getTypeSize(GLenum type); + #endif /* _R700_VERTPROG_H_ */ diff --git a/src/mesa/drivers/dri/r600/radeon_buffer_objects.c b/src/mesa/drivers/dri/r600/radeon_buffer_objects.c new file mode 120000 index 00000000000..f6a5f664701 --- /dev/null +++ b/src/mesa/drivers/dri/r600/radeon_buffer_objects.c @@ -0,0 +1 @@ +../radeon/radeon_buffer_objects.c
\ No newline at end of file diff --git a/src/mesa/drivers/dri/r600/radeon_buffer_objects.h b/src/mesa/drivers/dri/r600/radeon_buffer_objects.h new file mode 120000 index 00000000000..2f134fd17b8 --- /dev/null +++ b/src/mesa/drivers/dri/r600/radeon_buffer_objects.h @@ -0,0 +1 @@ +../radeon/radeon_buffer_objects.h
\ No newline at end of file diff --git a/src/mesa/drivers/dri/radeon/Makefile b/src/mesa/drivers/dri/radeon/Makefile index b1efc72872f..ae2e695bfcc 100644 --- a/src/mesa/drivers/dri/radeon/Makefile +++ b/src/mesa/drivers/dri/radeon/Makefile @@ -55,4 +55,3 @@ X86_SOURCES = include ../Makefile.template -symlinks: diff --git a/src/mesa/drivers/dri/radeon/radeon_bo_drm.h b/src/mesa/drivers/dri/radeon/radeon_bo_drm.h index 71413716333..46e30b905a0 100644 --- a/src/mesa/drivers/dri/radeon/radeon_bo_drm.h +++ b/src/mesa/drivers/dri/radeon/radeon_bo_drm.h @@ -39,6 +39,7 @@ #define RADEON_BO_FLAGS_MICRO_TILE 2 struct radeon_bo_manager; +struct radeon_cs; struct radeon_bo { uint32_t alignment; @@ -74,6 +75,7 @@ struct radeon_bo_funcs { int (*bo_get_tiling)(struct radeon_bo *bo, uint32_t *tiling_flags, uint32_t *pitch); int (*bo_is_busy)(struct radeon_bo *bo, uint32_t *domain); + int (*bo_is_referenced_by_cs)(struct radeon_bo *bo, struct radeon_cs *cs); }; struct radeon_bo_manager { @@ -199,6 +201,15 @@ static inline int radeon_bo_is_static(struct radeon_bo *bo) return 0; } +static inline int _radeon_bo_is_referenced_by_cs(struct radeon_bo *bo, + struct radeon_cs *cs, + const char *file, + const char *func, + unsigned line) +{ + return bo->cref > 1; +} + #define radeon_bo_open(bom, h, s, a, d, f)\ _radeon_bo_open(bom, h, s, a, d, f, __FILE__, __FUNCTION__, __LINE__) #define radeon_bo_ref(bo)\ @@ -215,5 +226,7 @@ static inline int radeon_bo_is_static(struct radeon_bo *bo) _radeon_bo_wait(bo, __FILE__, __func__, __LINE__) #define radeon_bo_is_busy(bo, domain) \ _radeon_bo_is_busy(bo, domain, __FILE__, __func__, __LINE__) +#define radeon_bo_is_referenced_by_cs(bo, cs) \ + _radeon_bo_is_referenced_by_cs(bo, cs, __FILE__, __FUNCTION__, __LINE__) #endif diff --git a/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c b/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c index 3e7547d2f9d..ce60a2f7eae 100644 --- a/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c +++ b/src/mesa/drivers/dri/radeon/radeon_bo_legacy.c @@ -710,6 +710,10 @@ int radeon_bo_legacy_validate(struct radeon_bo *bo, bo, bo->size, bo_legacy->map_count); return -EINVAL; } + if(bo->size == 0) { + fprintf(stderr, "bo(%p) has size 0.\n", bo); + return -EINVAL; + } if (bo_legacy->static_bo || bo_legacy->validated) { *soffset = bo_legacy->offset; *eoffset = bo_legacy->offset + bo->size; diff --git a/src/mesa/drivers/dri/radeon/radeon_buffer_objects.c b/src/mesa/drivers/dri/radeon/radeon_buffer_objects.c index a24b6dac265..99d3ec7005e 100644 --- a/src/mesa/drivers/dri/radeon/radeon_buffer_objects.c +++ b/src/mesa/drivers/dri/radeon/radeon_buffer_objects.c @@ -136,8 +136,13 @@ radeonBufferSubData(GLcontext * ctx, const GLvoid * data, struct gl_buffer_object *obj) { + radeonContextPtr radeon = RADEON_CONTEXT(ctx); struct radeon_buffer_object *radeon_obj = get_radeon_buffer_object(obj); + if (radeon_bo_is_referenced_by_cs(radeon_obj->bo, radeon->cmdbuf.cs)) { + radeon_firevertices(radeon); + } + radeon_bo_map(radeon_obj->bo, GL_TRUE); _mesa_memcpy(radeon_obj->bo->ptr + offset, data, size); @@ -187,7 +192,11 @@ radeonMapBuffer(GLcontext * ctx, radeon_bo_map(radeon_obj->bo, access == GL_WRITE_ONLY_ARB); - return obj->Pointer = radeon_obj->bo->ptr; + obj->Pointer = radeon_obj->bo->ptr; + obj->Length = obj->Size; + obj->Offset = 0; + + return obj->Pointer; } @@ -203,9 +212,12 @@ radeonUnmapBuffer(GLcontext * ctx, if (radeon_obj->bo != NULL) { radeon_bo_unmap(radeon_obj->bo); - obj->Pointer = NULL; } + obj->Pointer = NULL; + obj->Offset = 0; + obj->Length = 0; + return GL_TRUE; } diff --git a/src/mesa/drivers/dri/radeon/radeon_common.c b/src/mesa/drivers/dri/radeon/radeon_common.c index a4c7b40798a..51fa6189377 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common.c +++ b/src/mesa/drivers/dri/radeon/radeon_common.c @@ -229,7 +229,6 @@ void radeonUpdateScissor( GLcontext *ctx ) } if (!rmesa->radeonScreen->kernel_mm) { /* Fix scissors for dri 1 */ - __DRIdrawablePrivate *dPriv = radeon_get_drawable(rmesa); x1 += dPriv->x; x2 += dPriv->x + 1; @@ -263,29 +262,6 @@ void radeonScissor(GLcontext* ctx, GLint x, GLint y, GLsizei w, GLsizei h) } } -void radeonPolygonStipplePreKMS( GLcontext *ctx, const GLubyte *mask ) -{ - radeonContextPtr radeon = RADEON_CONTEXT(ctx); - GLuint i; - drm_radeon_stipple_t stipple; - - /* Must flip pattern upside down. - */ - for ( i = 0 ; i < 32 ; i++ ) { - stipple.mask[31 - i] = ((GLuint *) mask)[i]; - } - - /* TODO: push this into cmd mechanism - */ - radeon_firevertices(radeon); - LOCK_HARDWARE( radeon ); - - drmCommandWrite( radeon->dri.fd, DRM_RADEON_STIPPLE, - &stipple, sizeof(stipple) ); - UNLOCK_HARDWARE( radeon ); -} - - /* ================================================================ * SwapBuffers with client-side throttling */ @@ -841,7 +817,7 @@ void radeonDrawBuffer( GLcontext *ctx, GLenum mode ) */ if (!was_front_buffer_rendering && radeon->is_front_buffer_rendering) { radeon_update_renderbuffers(radeon->dri.context, - radeon->dri.context->driDrawablePriv); + radeon->dri.context->driDrawablePriv, GL_FALSE); } } @@ -858,7 +834,7 @@ void radeonReadBuffer( GLcontext *ctx, GLenum mode ) if (!was_front_buffer_reading && rmesa->is_front_buffer_reading) { radeon_update_renderbuffers(rmesa->dri.context, - rmesa->dri.context->driReadablePriv); + rmesa->dri.context->driReadablePriv, GL_FALSE); } } /* nothing, until we implement h/w glRead/CopyPixels or CopyTexImage */ @@ -909,9 +885,9 @@ void radeon_viewport(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei he if (radeon->is_front_buffer_rendering) { ctx->Driver.Flush(ctx); } - radeon_update_renderbuffers(driContext, driContext->driDrawablePriv); + radeon_update_renderbuffers(driContext, driContext->driDrawablePriv, GL_FALSE); if (driContext->driDrawablePriv != driContext->driReadablePriv) - radeon_update_renderbuffers(driContext, driContext->driReadablePriv); + radeon_update_renderbuffers(driContext, driContext->driReadablePriv, GL_FALSE); } old_viewport = ctx->Driver.Viewport; @@ -1124,8 +1100,6 @@ void radeonFlush(GLcontext *ctx) if (radeon->dma.flush) radeon->dma.flush( ctx ); - radeonEmitState(radeon); - if (radeon->cmdbuf.cs->cdw) rcommonFlushCmdBuf(radeon, __FUNCTION__); @@ -1148,9 +1122,6 @@ void radeonFlush(GLcontext *ctx) } } } - - make_empty_list(&radeon->query.not_flushed_head); - } /* Make sure all commands have been sent to the hardware and have @@ -1345,5 +1316,5 @@ void rcommonBeginBatch(radeonContextPtr rmesa, int n, void radeonUserClear(GLcontext *ctx, GLuint mask) { - _mesa_meta_clear(ctx, mask); + _mesa_meta_Clear(ctx, mask); } diff --git a/src/mesa/drivers/dri/radeon/radeon_common.h b/src/mesa/drivers/dri/radeon/radeon_common.h index f3201911ac6..0608fe2418c 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common.h +++ b/src/mesa/drivers/dri/radeon/radeon_common.h @@ -10,7 +10,6 @@ void radeonRecalcScissorRects(radeonContextPtr radeon); void radeonSetCliprects(radeonContextPtr radeon); void radeonUpdateScissor( GLcontext *ctx ); void radeonScissor(GLcontext* ctx, GLint x, GLint y, GLsizei w, GLsizei h); -void radeonPolygonStipplePreKMS( GLcontext *ctx, const GLubyte *mask ); void radeonWaitForIdleLocked(radeonContextPtr radeon); extern uint32_t radeonGetAge(radeonContextPtr radeon); @@ -43,7 +42,7 @@ void radeon_renderbuffer_set_bo(struct radeon_renderbuffer *rb, struct radeon_bo *bo); struct radeon_renderbuffer * -radeon_create_renderbuffer(GLenum format, __DRIdrawablePrivate *driDrawPriv); +radeon_create_renderbuffer(gl_format format, __DRIdrawablePrivate *driDrawPriv); static inline struct radeon_renderbuffer *radeon_renderbuffer(struct gl_renderbuffer *rb) { struct radeon_renderbuffer *rrb = (struct radeon_renderbuffer *)rb; diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.c b/src/mesa/drivers/dri/radeon/radeon_common_context.c index 1c53c04da77..5c68bf5df6c 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common_context.c +++ b/src/mesa/drivers/dri/radeon/radeon_common_context.c @@ -227,11 +227,8 @@ GLboolean radeonInitContext(radeonContextPtr radeon, fthrottle_mode = driQueryOptioni(&radeon->optionCache, "fthrottle_mode"); radeon->iw.irq_seq = -1; radeon->irqsEmitted = 0; - if (IS_R600_CLASS(radeon->radeonScreen)) - radeon->do_irqs = 0; - else - radeon->do_irqs = (fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS && - radeon->radeonScreen->irq); + radeon->do_irqs = (fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS && + radeon->radeonScreen->irq); radeon->do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS); @@ -265,10 +262,9 @@ GLboolean radeonInitContext(radeonContextPtr radeon, else radeon->texture_row_align = 32; radeon->texture_rect_row_align = 64; - radeon->texture_compressed_row_align = 64; + radeon->texture_compressed_row_align = 32; } - make_empty_list(&radeon->query.not_flushed_head); radeon_init_dma(radeon); return GL_TRUE; @@ -499,23 +495,12 @@ radeon_make_renderbuffer_current(radeonContextPtr radeon, static unsigned radeon_bits_per_pixel(const struct radeon_renderbuffer *rb) { - switch (rb->base._ActualFormat) { - case GL_RGB5: - case GL_DEPTH_COMPONENT16: - return 16; - case GL_RGB8: - case GL_RGBA8: - case GL_DEPTH_COMPONENT24: - case GL_DEPTH24_STENCIL8_EXT: - case GL_STENCIL_INDEX8_EXT: - return 32; - default: - return 0; - } + return _mesa_get_format_bytes(rb->base.Format) * 8; } void -radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable) +radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable, + GLboolean front_only) { unsigned int attachments[10]; __DRIbuffer *buffers = NULL; @@ -541,7 +526,7 @@ radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable) struct radeon_renderbuffer *stencil_rb; i = 0; - if ((radeon->is_front_buffer_rendering || + if ((front_only || radeon->is_front_buffer_rendering || radeon->is_front_buffer_reading || !draw->color_rb[1]) && draw->color_rb[0]) { @@ -549,23 +534,25 @@ radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable) attachments[i++] = radeon_bits_per_pixel(draw->color_rb[0]); } - if (draw->color_rb[1]) { - attachments[i++] = __DRI_BUFFER_BACK_LEFT; - attachments[i++] = radeon_bits_per_pixel(draw->color_rb[1]); - } + if (!front_only) { + if (draw->color_rb[1]) { + attachments[i++] = __DRI_BUFFER_BACK_LEFT; + attachments[i++] = radeon_bits_per_pixel(draw->color_rb[1]); + } - depth_rb = radeon_get_renderbuffer(&draw->base, BUFFER_DEPTH); - stencil_rb = radeon_get_renderbuffer(&draw->base, BUFFER_STENCIL); - - if ((depth_rb != NULL) && (stencil_rb != NULL)) { - attachments[i++] = __DRI_BUFFER_DEPTH_STENCIL; - attachments[i++] = radeon_bits_per_pixel(depth_rb); - } else if (depth_rb != NULL) { - attachments[i++] = __DRI_BUFFER_DEPTH; - attachments[i++] = radeon_bits_per_pixel(depth_rb); - } else if (stencil_rb != NULL) { - attachments[i++] = __DRI_BUFFER_STENCIL; - attachments[i++] = radeon_bits_per_pixel(stencil_rb); + depth_rb = radeon_get_renderbuffer(&draw->base, BUFFER_DEPTH); + stencil_rb = radeon_get_renderbuffer(&draw->base, BUFFER_STENCIL); + + if ((depth_rb != NULL) && (stencil_rb != NULL)) { + attachments[i++] = __DRI_BUFFER_DEPTH_STENCIL; + attachments[i++] = radeon_bits_per_pixel(depth_rb); + } else if (depth_rb != NULL) { + attachments[i++] = __DRI_BUFFER_DEPTH; + attachments[i++] = radeon_bits_per_pixel(depth_rb); + } else if (stencil_rb != NULL) { + attachments[i++] = __DRI_BUFFER_STENCIL; + attachments[i++] = radeon_bits_per_pixel(stencil_rb); + } } buffers = (*screen->dri2.loader->getBuffersWithFormat)(drawable, @@ -578,12 +565,14 @@ radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable) i = 0; if (draw->color_rb[0]) attachments[i++] = __DRI_BUFFER_FRONT_LEFT; - if (draw->color_rb[1]) - attachments[i++] = __DRI_BUFFER_BACK_LEFT; - if (radeon_get_renderbuffer(&draw->base, BUFFER_DEPTH)) - attachments[i++] = __DRI_BUFFER_DEPTH; - if (radeon_get_renderbuffer(&draw->base, BUFFER_STENCIL)) - attachments[i++] = __DRI_BUFFER_STENCIL; + if (!front_only) { + if (draw->color_rb[1]) + attachments[i++] = __DRI_BUFFER_BACK_LEFT; + if (radeon_get_renderbuffer(&draw->base, BUFFER_DEPTH)) + attachments[i++] = __DRI_BUFFER_DEPTH; + if (radeon_get_renderbuffer(&draw->base, BUFFER_STENCIL)) + attachments[i++] = __DRI_BUFFER_STENCIL; + } buffers = (*screen->dri2.loader->getBuffers)(drawable, &drawable->w, @@ -751,9 +740,9 @@ GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv, readfb = driReadPriv->driverPrivate; if (driContextPriv->driScreenPriv->dri2.enabled) { - radeon_update_renderbuffers(driContextPriv, driDrawPriv); + radeon_update_renderbuffers(driContextPriv, driDrawPriv, GL_FALSE); if (driDrawPriv != driReadPriv) - radeon_update_renderbuffers(driContextPriv, driReadPriv); + radeon_update_renderbuffers(driContextPriv, driReadPriv, GL_FALSE); _mesa_reference_renderbuffer(&radeon->state.color.rb, &(radeon_get_renderbuffer(&drfb->base, BUFFER_BACK_LEFT)->base)); _mesa_reference_renderbuffer(&radeon->state.depth.rb, diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.h b/src/mesa/drivers/dri/radeon/radeon_common_context.h index 0309345393d..49a9ec56106 100644 --- a/src/mesa/drivers/dri/radeon/radeon_common_context.h +++ b/src/mesa/drivers/dri/radeon/radeon_common_context.h @@ -208,6 +208,10 @@ struct radeon_tex_obj { * and so on. */ GLboolean validated; + /* Minimum LOD to be used during rendering */ + unsigned minLod; + /* Miximum LOD to be used during rendering */ + unsigned maxLod; GLuint override_offset; GLboolean image_override; /* Image overridden by GLX_EXT_tfp */ @@ -324,6 +328,7 @@ struct radeon_swtcl_info { GLuint vertex_attr_count; GLuint emit_prediction; + struct radeon_bo *bo; }; #define RADEON_MAX_AOS_ARRAYS 16 @@ -502,7 +507,6 @@ struct radeon_context { struct { struct radeon_query_object *current; - struct radeon_query_object not_flushed_head; struct radeon_state_atom queryobj; } query; @@ -585,7 +589,8 @@ GLboolean radeonInitContext(radeonContextPtr radeon, void radeonCleanupContext(radeonContextPtr radeon); GLboolean radeonUnbindContext(__DRIcontextPrivate * driContextPriv); -void radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable); +void radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable, + GLboolean front_only); GLboolean radeonMakeCurrent(__DRIcontextPrivate * driContextPriv, __DRIdrawablePrivate * driDrawPriv, __DRIdrawablePrivate * driReadPriv); diff --git a/src/mesa/drivers/dri/radeon/radeon_context.c b/src/mesa/drivers/dri/radeon/radeon_context.c index 8f4485aee7c..5e700be4a59 100644 --- a/src/mesa/drivers/dri/radeon/radeon_context.c +++ b/src/mesa/drivers/dri/radeon/radeon_context.c @@ -69,7 +69,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #define need_GL_EXT_fog_coord #define need_GL_EXT_secondary_color #define need_GL_EXT_framebuffer_object -#include "extension_helper.h" +#include "main/remap_helper.h" #define DRIVER_DATE "20061018" @@ -79,7 +79,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. /* Extension strings exported by the R100 driver. */ -const struct dri_extension card_extensions[] = +static const struct dri_extension card_extensions[] = { { "GL_ARB_multitexture", NULL }, { "GL_ARB_occlusion_query", GL_ARB_occlusion_query_functions}, @@ -109,7 +109,7 @@ const struct dri_extension card_extensions[] = { NULL, NULL } }; -const struct dri_extension mm_extensions[] = { +static const struct dri_extension mm_extensions[] = { { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions }, { NULL, NULL } }; diff --git a/src/mesa/drivers/dri/radeon/radeon_context.h b/src/mesa/drivers/dri/radeon/radeon_context.h index 4e2c52c835c..12ab33a009c 100644 --- a/src/mesa/drivers/dri/radeon/radeon_context.h +++ b/src/mesa/drivers/dri/radeon/radeon_context.h @@ -331,8 +331,12 @@ struct r100_hw_state { struct radeon_state_atom stp; }; +struct radeon_stipple_state { + GLuint mask[32]; +}; struct r100_state { + struct radeon_stipple_state stipple; struct radeon_texture_state texture; }; diff --git a/src/mesa/drivers/dri/radeon/radeon_debug.c b/src/mesa/drivers/dri/radeon/radeon_debug.c index 3b6f0038037..413000b6c06 100644 --- a/src/mesa/drivers/dri/radeon/radeon_debug.c +++ b/src/mesa/drivers/dri/radeon/radeon_debug.c @@ -39,7 +39,7 @@ static const struct dri_debug_control debug_control[] = { {"fall", RADEON_FALLBACKS}, {"tex", RADEON_TEXTURE}, {"ioctl", RADEON_IOCTL}, - {"verts", RADEON_RENDER}, + {"verts", RADEON_VERTS}, {"render", RADEON_RENDER}, {"swrender", RADEON_SWRENDER}, {"state", RADEON_STATE}, diff --git a/src/mesa/drivers/dri/radeon/radeon_debug.h b/src/mesa/drivers/dri/radeon/radeon_debug.h index 2a8302293b2..26da31c1c48 100644 --- a/src/mesa/drivers/dri/radeon/radeon_debug.h +++ b/src/mesa/drivers/dri/radeon/radeon_debug.h @@ -149,21 +149,22 @@ static inline void radeon_debug_remove_indent(void) } } + /* From http://gcc. gnu.org/onlinedocs/gcc-3.2.3/gcc/Variadic-Macros.html . I suppose we could inline this and use macro to fetch out __LINE__ and stuff in case we run into trouble with other compilers ... GLUE! */ -#define WARN_ONCE(a, ...) { \ - static int warn##__LINE__=1; \ - if(warn##__LINE__){ \ +#define WARN_ONCE(a, ...) do { \ + static int __warn_once=1; \ + if(__warn_once){ \ radeon_warning("*********************************WARN_ONCE*********************************\n"); \ radeon_warning("File %s function %s line %d\n", \ __FILE__, __FUNCTION__, __LINE__); \ radeon_warning( (a), ## __VA_ARGS__);\ radeon_warning("***************************************************************************\n"); \ - warn##__LINE__=0;\ + __warn_once=0;\ } \ - } + } while(0) #endif diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.c b/src/mesa/drivers/dri/radeon/radeon_dma.c index c9a32c808be..b8c65f4ce62 100644 --- a/src/mesa/drivers/dri/radeon/radeon_dma.c +++ b/src/mesa/drivers/dri/radeon/radeon_dma.c @@ -151,6 +151,7 @@ void rcommon_emit_vector(GLcontext * ctx, struct radeon_aos *aos, aos->components = size; aos->count = count; + radeon_bo_map(aos->bo, 1); out = (uint32_t*)((char*)aos->bo->ptr + aos->offset); switch (size) { case 1: radeonEmitVec4(out, data, stride, count); break; @@ -161,6 +162,7 @@ void rcommon_emit_vector(GLcontext * ctx, struct radeon_aos *aos, assert(0); break; } + radeon_bo_unmap(aos->bo); } void radeon_init_dma(radeonContextPtr rmesa) @@ -183,10 +185,6 @@ void radeonRefillCurrentDmaRegion(radeonContextPtr rmesa, int size) __FUNCTION__, size, rmesa->dma.minimum_size); - /* unmap old reserved bo */ - if (!is_empty_list(&rmesa->dma.reserved)) - radeon_bo_unmap(first_elem(&rmesa->dma.reserved)->bo); - if (is_empty_list(&rmesa->dma.free) || last_elem(&rmesa->dma.free)->bo->size < size) { dma_bo = CALLOC_STRUCT(radeon_dma_bo); @@ -224,8 +222,6 @@ again_alloc: /* Cmd buff have been flushed in radeon_revalidate_bos */ goto again_alloc; } - - radeon_bo_map(first_elem(&rmesa->dma.reserved)->bo, 1); } /* Allocates a region from rmesa->dma.current. If there isn't enough @@ -282,7 +278,6 @@ void radeonFreeDmaRegions(radeonContextPtr rmesa) foreach_s(dma_bo, temp, &rmesa->dma.reserved) { remove_from_list(dma_bo); - radeon_bo_unmap(dma_bo->bo); radeon_bo_unref(dma_bo->bo); FREE(dma_bo); } @@ -362,9 +357,6 @@ void radeonReleaseDmaRegions(radeonContextPtr rmesa) insert_at_tail(&rmesa->dma.free, dma_bo); } - /* unmap the last dma region */ - if (!is_empty_list(&rmesa->dma.reserved)) - radeon_bo_unmap(first_elem(&rmesa->dma.reserved)->bo); /* move reserved to wait list */ foreach_s(dma_bo, temp, &rmesa->dma.reserved) { /* free objects that are too small to be used because of large request */ @@ -398,11 +390,12 @@ void rcommon_flush_last_swtcl_prim( GLcontext *ctx ) radeonContextPtr rmesa = RADEON_CONTEXT(ctx); struct radeon_dma *dma = &rmesa->dma; - if (RADEON_DEBUG & RADEON_IOCTL) fprintf(stderr, "%s\n", __FUNCTION__); dma->flush = NULL; + radeon_bo_unmap(rmesa->swtcl.bo); + if (!is_empty_list(&dma->reserved)) { GLuint current_offset = dma->current_used; @@ -417,6 +410,8 @@ void rcommon_flush_last_swtcl_prim( GLcontext *ctx ) } rmesa->swtcl.numverts = 0; } + radeon_bo_unref(rmesa->swtcl.bo); + rmesa->swtcl.bo = NULL; } /* Alloc space in the current dma region. */ @@ -427,6 +422,7 @@ rcommonAllocDmaLowVerts( radeonContextPtr rmesa, int nverts, int vsize ) void *head; if (RADEON_DEBUG & RADEON_IOCTL) fprintf(stderr, "%s\n", __FUNCTION__); + if(is_empty_list(&rmesa->dma.reserved) ||rmesa->dma.current_vertexptr + bytes > first_elem(&rmesa->dma.reserved)->bo->size) { if (rmesa->dma.flush) { @@ -450,7 +446,13 @@ rcommonAllocDmaLowVerts( radeonContextPtr rmesa, int nverts, int vsize ) rmesa->swtcl.numverts * rmesa->swtcl.vertex_size * 4 == rmesa->dma.current_vertexptr ); - head = (first_elem(&rmesa->dma.reserved)->bo->ptr + rmesa->dma.current_vertexptr); + if (!rmesa->swtcl.bo) { + rmesa->swtcl.bo = first_elem(&rmesa->dma.reserved)->bo; + radeon_bo_ref(rmesa->swtcl.bo); + radeon_bo_map(rmesa->swtcl.bo, 1); + } + + head = (rmesa->swtcl.bo->ptr + rmesa->dma.current_vertexptr); rmesa->dma.current_vertexptr += bytes; rmesa->swtcl.numverts += nverts; return head; diff --git a/src/mesa/drivers/dri/radeon/radeon_fbo.c b/src/mesa/drivers/dri/radeon/radeon_fbo.c index d83b166742c..fc21069a92c 100644 --- a/src/mesa/drivers/dri/radeon/radeon_fbo.c +++ b/src/mesa/drivers/dri/radeon/radeon_fbo.c @@ -33,7 +33,6 @@ #include "main/framebuffer.h" #include "main/renderbuffer.h" #include "main/context.h" -#include "main/texformat.h" #include "main/texrender.h" #include "drivers/common/meta.h" @@ -91,11 +90,8 @@ radeon_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, case GL_R3_G3_B2: case GL_RGB4: case GL_RGB5: - rb->_ActualFormat = GL_RGB5; + rb->Format = _dri_texformat_rgb565; rb->DataType = GL_UNSIGNED_BYTE; - rb->RedBits = 5; - rb->GreenBits = 6; - rb->BlueBits = 5; cpp = 2; break; case GL_RGB: @@ -103,12 +99,8 @@ radeon_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, case GL_RGB10: case GL_RGB12: case GL_RGB16: - rb->_ActualFormat = GL_RGB8; + rb->Format = _dri_texformat_argb8888; rb->DataType = GL_UNSIGNED_BYTE; - rb->RedBits = 8; - rb->GreenBits = 8; - rb->BlueBits = 8; - rb->AlphaBits = 0; cpp = 4; break; case GL_RGBA: @@ -119,12 +111,8 @@ radeon_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, case GL_RGB10_A2: case GL_RGBA12: case GL_RGBA16: - rb->_ActualFormat = GL_RGBA8; + rb->Format = _dri_texformat_argb8888; rb->DataType = GL_UNSIGNED_BYTE; - rb->RedBits = 8; - rb->GreenBits = 8; - rb->BlueBits = 8; - rb->AlphaBits = 8; cpp = 4; break; case GL_STENCIL_INDEX: @@ -133,39 +121,36 @@ radeon_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, case GL_STENCIL_INDEX8_EXT: case GL_STENCIL_INDEX16_EXT: /* alloc a depth+stencil buffer */ - rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT; + rb->Format = MESA_FORMAT_S8_Z24; rb->DataType = GL_UNSIGNED_INT_24_8_EXT; - rb->StencilBits = 8; cpp = 4; break; case GL_DEPTH_COMPONENT16: - rb->_ActualFormat = GL_DEPTH_COMPONENT16; + rb->Format = MESA_FORMAT_Z16; rb->DataType = GL_UNSIGNED_SHORT; - rb->DepthBits = 16; cpp = 2; break; case GL_DEPTH_COMPONENT: case GL_DEPTH_COMPONENT24: case GL_DEPTH_COMPONENT32: - rb->_ActualFormat = GL_DEPTH_COMPONENT24; + rb->Format = MESA_FORMAT_X8_Z24; rb->DataType = GL_UNSIGNED_INT; - rb->DepthBits = 24; cpp = 4; break; case GL_DEPTH_STENCIL_EXT: case GL_DEPTH24_STENCIL8_EXT: - rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT; + rb->Format = MESA_FORMAT_S8_Z24; rb->DataType = GL_UNSIGNED_INT_24_8_EXT; - rb->DepthBits = 24; - rb->StencilBits = 8; cpp = 4; break; default: _mesa_problem(ctx, - "Unexpected format in intel_alloc_renderbuffer_storage"); + "Unexpected format in radeon_alloc_renderbuffer_storage"); return GL_FALSE; } + rb->_BaseFormat = _mesa_base_fbo_format(ctx, internalFormat); + if (ctx->Driver.Flush) ctx->Driver.Flush(ctx); /* +r6/r7 */ @@ -213,7 +198,7 @@ radeon_alloc_window_storage(GLcontext * ctx, struct gl_renderbuffer *rb, ASSERT(rb->Name == 0); rb->Width = width; rb->Height = height; - rb->_ActualFormat = internalFormat; + rb->InternalFormat = internalFormat; return GL_TRUE; } @@ -255,8 +240,13 @@ radeon_nop_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, return GL_FALSE; } + +/** + * Create a renderbuffer for a window's color, depth and/or stencil buffer. + * Not used for user-created renderbuffers. + */ struct radeon_renderbuffer * -radeon_create_renderbuffer(GLenum format, __DRIdrawablePrivate *driDrawPriv) +radeon_create_renderbuffer(gl_format format, __DRIdrawablePrivate *driDrawPriv) { struct radeon_renderbuffer *rrb; @@ -267,67 +257,64 @@ radeon_create_renderbuffer(GLenum format, __DRIdrawablePrivate *driDrawPriv) _mesa_init_renderbuffer(&rrb->base, 0); rrb->base.ClassID = RADEON_RB_CLASS; - /* XXX format junk */ + rrb->base.Format = format; + switch (format) { - case GL_RGB5: - rrb->base._ActualFormat = GL_RGB5; - rrb->base._BaseFormat = GL_RGBA; - rrb->base.RedBits = 5; - rrb->base.GreenBits = 6; - rrb->base.BlueBits = 5; + case MESA_FORMAT_RGB565: + assert(_mesa_little_endian()); + rrb->base.DataType = GL_UNSIGNED_BYTE; + rrb->base._BaseFormat = GL_RGB; + break; + case MESA_FORMAT_RGB565_REV: + assert(!_mesa_little_endian()); + rrb->base.DataType = GL_UNSIGNED_BYTE; + rrb->base._BaseFormat = GL_RGB; + break; + case MESA_FORMAT_XRGB8888: + assert(_mesa_little_endian()); rrb->base.DataType = GL_UNSIGNED_BYTE; + rrb->base._BaseFormat = GL_RGB; break; - case GL_RGB8: - rrb->base._ActualFormat = GL_RGB8; - rrb->base._BaseFormat = GL_RGB; - rrb->base.RedBits = 8; - rrb->base.GreenBits = 8; - rrb->base.BlueBits = 8; - rrb->base.AlphaBits = 0; + case MESA_FORMAT_XRGB8888_REV: + assert(!_mesa_little_endian()); rrb->base.DataType = GL_UNSIGNED_BYTE; + rrb->base._BaseFormat = GL_RGB; break; - case GL_RGBA8: - rrb->base._ActualFormat = GL_RGBA8; - rrb->base._BaseFormat = GL_RGBA; - rrb->base.RedBits = 8; - rrb->base.GreenBits = 8; - rrb->base.BlueBits = 8; - rrb->base.AlphaBits = 8; + case MESA_FORMAT_ARGB8888: + assert(_mesa_little_endian()); rrb->base.DataType = GL_UNSIGNED_BYTE; + rrb->base._BaseFormat = GL_RGBA; break; - case GL_STENCIL_INDEX8_EXT: - rrb->base._ActualFormat = GL_STENCIL_INDEX8_EXT; - rrb->base._BaseFormat = GL_STENCIL_INDEX; - rrb->base.StencilBits = 8; + case MESA_FORMAT_ARGB8888_REV: + assert(!_mesa_little_endian()); rrb->base.DataType = GL_UNSIGNED_BYTE; + rrb->base._BaseFormat = GL_RGBA; break; - case GL_DEPTH_COMPONENT16: - rrb->base._ActualFormat = GL_DEPTH_COMPONENT16; - rrb->base._BaseFormat = GL_DEPTH_COMPONENT; - rrb->base.DepthBits = 16; + case MESA_FORMAT_S8: + rrb->base.DataType = GL_UNSIGNED_BYTE; + rrb->base._BaseFormat = GL_STENCIL_INDEX; + break; + case MESA_FORMAT_Z16: rrb->base.DataType = GL_UNSIGNED_SHORT; + rrb->base._BaseFormat = GL_DEPTH_COMPONENT; break; - case GL_DEPTH_COMPONENT24: - rrb->base._ActualFormat = GL_DEPTH_COMPONENT24; - rrb->base._BaseFormat = GL_DEPTH_COMPONENT; - rrb->base.DepthBits = 24; + case MESA_FORMAT_X8_Z24: rrb->base.DataType = GL_UNSIGNED_INT; + rrb->base._BaseFormat = GL_DEPTH_COMPONENT; break; - case GL_DEPTH24_STENCIL8_EXT: - rrb->base._ActualFormat = GL_DEPTH24_STENCIL8_EXT; - rrb->base._BaseFormat = GL_DEPTH_STENCIL_EXT; - rrb->base.DepthBits = 24; - rrb->base.StencilBits = 8; + case MESA_FORMAT_S8_Z24: rrb->base.DataType = GL_UNSIGNED_INT_24_8_EXT; + rrb->base._BaseFormat = GL_DEPTH_STENCIL; break; default: - fprintf(stderr, "%s: Unknown format 0x%04x\n", __FUNCTION__, format); + fprintf(stderr, "%s: Unknown format %s\n", + __FUNCTION__, _mesa_get_format_name(format)); _mesa_delete_renderbuffer(&rrb->base); return NULL; } rrb->dPriv = driDrawPriv; - rrb->base.InternalFormat = format; + rrb->base.InternalFormat = _mesa_get_format_base_format(format); rrb->base.Delete = radeon_delete_renderbuffer; rrb->base.AllocStorage = radeon_alloc_window_storage; @@ -382,51 +369,41 @@ radeon_framebuffer_renderbuffer(GLcontext * ctx, } +/* TODO: According to EXT_fbo spec internal format of texture image + * once set during glTexImage call, should be preserved when + * attaching image to renderbuffer. When HW doesn't support + * rendering to format of attached image, set framebuffer + * completeness accordingly in radeon_validate_framebuffer (issue #79). + */ static GLboolean radeon_update_wrapper(GLcontext *ctx, struct radeon_renderbuffer *rrb, struct gl_texture_image *texImage) { int retry = 0; + gl_format texFormat; + restart: - if (texImage->TexFormat == &_mesa_texformat_argb8888) { - rrb->cpp = 4; - rrb->base._ActualFormat = GL_RGBA8; - rrb->base._BaseFormat = GL_RGBA; + if (texImage->TexFormat == _dri_texformat_argb8888) { rrb->base.DataType = GL_UNSIGNED_BYTE; DBG("Render to RGBA8 texture OK\n"); } - else if (texImage->TexFormat == &_mesa_texformat_rgb565) { - rrb->cpp = 2; - rrb->base._ActualFormat = GL_RGB5; - rrb->base._BaseFormat = GL_RGB; + else if (texImage->TexFormat == _dri_texformat_rgb565) { rrb->base.DataType = GL_UNSIGNED_BYTE; DBG("Render to RGB5 texture OK\n"); } - else if (texImage->TexFormat == &_mesa_texformat_argb1555) { - rrb->cpp = 2; - rrb->base._ActualFormat = GL_RGB5_A1; - rrb->base._BaseFormat = GL_RGBA; + else if (texImage->TexFormat == _dri_texformat_argb1555) { rrb->base.DataType = GL_UNSIGNED_BYTE; DBG("Render to ARGB1555 texture OK\n"); } - else if (texImage->TexFormat == &_mesa_texformat_argb4444) { - rrb->cpp = 2; - rrb->base._ActualFormat = GL_RGBA4; - rrb->base._BaseFormat = GL_RGBA; + else if (texImage->TexFormat == _dri_texformat_argb4444) { rrb->base.DataType = GL_UNSIGNED_BYTE; - DBG("Render to ARGB1555 texture OK\n"); + DBG("Render to ARGB4444 texture OK\n"); } - else if (texImage->TexFormat == &_mesa_texformat_z16) { - rrb->cpp = 2; - rrb->base._ActualFormat = GL_DEPTH_COMPONENT16; - rrb->base._BaseFormat = GL_DEPTH_COMPONENT; + else if (texImage->TexFormat == MESA_FORMAT_Z16) { rrb->base.DataType = GL_UNSIGNED_SHORT; DBG("Render to DEPTH16 texture OK\n"); } - else if (texImage->TexFormat == &_mesa_texformat_s8_z24) { - rrb->cpp = 4; - rrb->base._ActualFormat = GL_DEPTH24_STENCIL8_EXT; - rrb->base._BaseFormat = GL_DEPTH_STENCIL_EXT; + else if (texImage->TexFormat == MESA_FORMAT_S8_Z24) { rrb->base.DataType = GL_UNSIGNED_INT_24_8_EXT; DBG("Render to DEPTH_STENCIL texture OK\n"); } @@ -434,27 +411,31 @@ restart: /* try redoing the FBO */ if (retry == 1) { DBG("Render to texture BAD FORMAT %d\n", - texImage->TexFormat->MesaFormat); + texImage->TexFormat); return GL_FALSE; } + /* XXX why is the tex format being set here? + * I think this can be removed. + */ texImage->TexFormat = radeonChooseTextureFormat(ctx, texImage->InternalFormat, 0, - texImage->TexFormat->DataType, + _mesa_get_format_datatype(texImage->TexFormat), 1); retry++; goto restart; } + texFormat = texImage->TexFormat; + + rrb->base.Format = texFormat; + + rrb->cpp = _mesa_get_format_bytes(texFormat); rrb->pitch = texImage->Width * rrb->cpp; - rrb->base.InternalFormat = rrb->base._ActualFormat; + rrb->base.InternalFormat = texImage->InternalFormat; + rrb->base._BaseFormat = _mesa_base_fbo_format(ctx, rrb->base.InternalFormat); + rrb->base.Width = texImage->Width; rrb->base.Height = texImage->Height; - rrb->base.RedBits = texImage->TexFormat->RedBits; - rrb->base.GreenBits = texImage->TexFormat->GreenBits; - rrb->base.BlueBits = texImage->TexFormat->BlueBits; - rrb->base.AlphaBits = texImage->TexFormat->AlphaBits; - rrb->base.DepthBits = texImage->TexFormat->DepthBits; - rrb->base.StencilBits = texImage->TexFormat->StencilBits; rrb->base.Delete = radeon_delete_renderbuffer; rrb->base.AllocStorage = radeon_nop_alloc_storage; @@ -555,8 +536,10 @@ radeon_render_texture(GLcontext * ctx, imageOffset += offsets[att->Zoffset]; } - /* store that offset in the region */ + /* store that offset in the region, along with the correct pitch for + * the image we are rendering to */ rrb->draw_offset = imageOffset; + rrb->pitch = radeon_image->mt->levels[att->TextureLevel].rowstride; /* update drawing region, etc */ radeon_draw_buffer(ctx, fb); @@ -583,7 +566,7 @@ void radeon_fbo_init(struct radeon_context *radeon) radeon->glCtx->Driver.FinishRenderTexture = radeon_finish_render_texture; radeon->glCtx->Driver.ResizeBuffers = radeon_resize_buffers; radeon->glCtx->Driver.ValidateFramebuffer = radeon_validate_framebuffer; - radeon->glCtx->Driver.BlitFramebuffer = _mesa_meta_blit_framebuffer; + radeon->glCtx->Driver.BlitFramebuffer = _mesa_meta_BlitFramebuffer; } diff --git a/src/mesa/drivers/dri/radeon/radeon_lock.c b/src/mesa/drivers/dri/radeon/radeon_lock.c index 02de8e5fd1d..7ad781ba619 100644 --- a/src/mesa/drivers/dri/radeon/radeon_lock.c +++ b/src/mesa/drivers/dri/radeon/radeon_lock.c @@ -62,8 +62,6 @@ void radeonGetLock(radeonContextPtr rmesa, GLuint flags) __DRIdrawablePrivate *const readable = radeon_get_readable(rmesa); __DRIscreenPrivate *sPriv = rmesa->dri.screen; - assert(drawable != NULL); - drmGetLock(rmesa->dri.fd, rmesa->dri.hwContext, flags); /* The window might have moved, so we might need to get new clip @@ -74,12 +72,13 @@ void radeonGetLock(radeonContextPtr rmesa, GLuint flags) * Since the hardware state depends on having the latest drawable * clip rects, all state checking must be done _after_ this call. */ - DRI_VALIDATE_DRAWABLE_INFO(sPriv, drawable); - if (drawable != readable) { + if (drawable) + DRI_VALIDATE_DRAWABLE_INFO(sPriv, drawable); + if (readable && drawable != readable) { DRI_VALIDATE_DRAWABLE_INFO(sPriv, readable); } - if (rmesa->lastStamp != drawable->lastStamp) { + if (drawable && (rmesa->lastStamp != drawable->lastStamp)) { radeon_window_moved(rmesa); rmesa->lastStamp = drawable->lastStamp; } diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c b/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c index 08e1c5d00df..d810e6080eb 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c +++ b/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c @@ -76,12 +76,14 @@ static void emit_vecfog(GLcontext *ctx, struct radeon_aos *aos, /* Emit the data */ + radeon_bo_map(aos->bo, 1); out = (uint32_t*)((char*)aos->bo->ptr + aos->offset); for (i = 0; i < count; i++) { out[0] = radeonComputeFogBlendFactor( ctx, *(GLfloat *)data ); out++; data += stride; } + radeon_bo_unmap(aos->bo); } static void emit_s0_vec(uint32_t *out, GLvoid *data, int stride, int count) @@ -151,6 +153,7 @@ static void emit_tex_vector(GLcontext *ctx, struct radeon_aos *aos, /* Emit the data */ + radeon_bo_map(aos->bo, 1); out = (uint32_t*)((char*)aos->bo->ptr + aos->offset); switch (size) { case 1: @@ -170,6 +173,7 @@ static void emit_tex_vector(GLcontext *ctx, struct radeon_aos *aos, exit(1); break; } + radeon_bo_unmap(aos->bo); } @@ -196,12 +200,12 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (!rmesa->tcl.obj.buf) rcommon_emit_vector( ctx, &(rmesa->tcl.aos[nr]), - (char *)VB->ObjPtr->data, - VB->ObjPtr->size, - VB->ObjPtr->stride, + (char *)VB->AttribPtr[_TNL_ATTRIB_POS]->data, + VB->AttribPtr[_TNL_ATTRIB_POS]->size, + VB->AttribPtr[_TNL_ATTRIB_POS]->stride, count); - switch( VB->ObjPtr->size ) { + switch( VB->AttribPtr[_TNL_ATTRIB_POS]->size ) { case 4: vfmt |= RADEON_CP_VC_FRMT_W0; case 3: vfmt |= RADEON_CP_VC_FRMT_Z; case 2: vfmt |= RADEON_CP_VC_FRMT_XY; @@ -216,9 +220,9 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (!rmesa->tcl.norm.buf) rcommon_emit_vector( ctx, &(rmesa->tcl.aos[nr]), - (char *)VB->NormalPtr->data, + (char *)VB->AttribPtr[_TNL_ATTRIB_NORMAL]->data, 3, - VB->NormalPtr->stride, + VB->AttribPtr[_TNL_ATTRIB_NORMAL]->stride, count); vfmt |= RADEON_CP_VC_FRMT_N0; @@ -227,9 +231,9 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (inputs & VERT_BIT_COLOR0) { int emitsize; - if (VB->ColorPtr[0]->size == 4 && - (VB->ColorPtr[0]->stride != 0 || - VB->ColorPtr[0]->data[0][3] != 1.0)) { + if (VB->AttribPtr[_TNL_ATTRIB_COLOR0]->size == 4 && + (VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride != 0 || + VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data[0][3] != 1.0)) { vfmt |= RADEON_CP_VC_FRMT_FPCOLOR | RADEON_CP_VC_FRMT_FPALPHA; emitsize = 4; } @@ -242,9 +246,9 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (!rmesa->tcl.rgba.buf) rcommon_emit_vector( ctx, &(rmesa->tcl.aos[nr]), - (char *)VB->ColorPtr[0]->data, + (char *)VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data, emitsize, - VB->ColorPtr[0]->stride, + VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride, count); nr++; @@ -256,9 +260,9 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) rcommon_emit_vector( ctx, &(rmesa->tcl.aos[nr]), - (char *)VB->SecondaryColorPtr[0]->data, + (char *)VB->AttribPtr[_TNL_ATTRIB_COLOR1]->data, 3, - VB->SecondaryColorPtr[0]->stride, + VB->AttribPtr[_TNL_ATTRIB_COLOR1]->stride, count); } @@ -273,8 +277,8 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (!rmesa->tcl.fog.buf) emit_vecfog( ctx, &(rmesa->tcl.aos[nr]), - (char *)VB->FogCoordPtr->data, - VB->FogCoordPtr->stride, + (char *)VB->AttribPtr[_TNL_ATTRIB_FOG]->data, + VB->AttribPtr[_TNL_ATTRIB_FOG]->stride, count); vfmt |= RADEON_CP_VC_FRMT_FPFOG; @@ -290,24 +294,24 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (!rmesa->tcl.tex[unit].buf) emit_tex_vector( ctx, &(rmesa->tcl.aos[nr]), - (char *)VB->TexCoordPtr[unit]->data, - VB->TexCoordPtr[unit]->size, - VB->TexCoordPtr[unit]->stride, + (char *)VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->data, + VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size, + VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->stride, count ); nr++; vfmt |= RADEON_ST_BIT(unit); /* assume we need the 3rd coord if texgen is active for r/q OR at least 3 coords are submitted. This may not be 100% correct */ - if (VB->TexCoordPtr[unit]->size >= 3) { + if (VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size >= 3) { vtx |= RADEON_Q_BIT(unit); vfmt |= RADEON_Q_BIT(unit); } if ( (ctx->Texture.Unit[unit].TexGenEnabled & (R_BIT | Q_BIT)) ) vtx |= RADEON_Q_BIT(unit); - else if ((VB->TexCoordPtr[unit]->size >= 3) && + else if ((VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size >= 3) && ((ctx->Texture.Unit[unit]._ReallyEnabled & (TEXTURE_CUBE_BIT)) == 0)) { - GLuint swaptexmatcol = (VB->TexCoordPtr[unit]->size - 3); + GLuint swaptexmatcol = (VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size - 3); if (((rmesa->NeedTexMatrix >> unit) & 1) && (swaptexmatcol != ((rmesa->TexMatColSwap >> unit) & 1))) radeonUploadTexMatrix( rmesa, unit, swaptexmatcol ) ; diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h b/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h index 515783135d6..d764ccb9826 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h +++ b/src/mesa/drivers/dri/radeon/radeon_maos_vbtmp.h @@ -56,18 +56,18 @@ static void TAG(emit)( GLcontext *ctx, radeon_print(RADEON_SWRENDER, RADEON_VERBOSE, "%s\n", __FUNCTION__); - coord = (GLuint (*)[4])VB->ObjPtr->data; - coord_stride = VB->ObjPtr->stride; + coord = (GLuint (*)[4])VB->AttribPtr[_TNL_ATTRIB_POS]->data; + coord_stride = VB->AttribPtr[_TNL_ATTRIB_POS]->stride; if (DO_TEX2) { - if (VB->TexCoordPtr[2]) { + if (VB->AttribPtr[_TNL_ATTRIB_TEX2]) { const GLuint t2 = GET_TEXSOURCE(2); - tc2 = (GLuint (*)[4])VB->TexCoordPtr[t2]->data; - tc2_stride = VB->TexCoordPtr[t2]->stride; - if (DO_PTEX && VB->TexCoordPtr[t2]->size < 3) { + tc2 = (GLuint (*)[4])VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->data; + tc2_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->stride; + if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->size < 3) { fill_tex |= (1<<2); } - else if (DO_PTEX && VB->TexCoordPtr[t2]->size < 4) { + else if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t2]->size < 4) { rqcoordsnoswap |= (1<<2); } } else { @@ -77,14 +77,14 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_TEX1) { - if (VB->TexCoordPtr[1]) { + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]) { const GLuint t1 = GET_TEXSOURCE(1); - tc1 = (GLuint (*)[4])VB->TexCoordPtr[t1]->data; - tc1_stride = VB->TexCoordPtr[t1]->stride; - if (DO_PTEX && VB->TexCoordPtr[t1]->size < 3) { + tc1 = (GLuint (*)[4])VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->data; + tc1_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->stride; + if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->size < 3) { fill_tex |= (1<<1); } - else if (DO_PTEX && VB->TexCoordPtr[t1]->size < 4) { + else if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t1]->size < 4) { rqcoordsnoswap |= (1<<1); } } else { @@ -94,14 +94,14 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_TEX0) { - if (VB->TexCoordPtr[0]) { + if (VB->AttribPtr[_TNL_ATTRIB_TEX0]) { const GLuint t0 = GET_TEXSOURCE(0); - tc0_stride = VB->TexCoordPtr[t0]->stride; - tc0 = (GLuint (*)[4])VB->TexCoordPtr[t0]->data; - if (DO_PTEX && VB->TexCoordPtr[t0]->size < 3) { + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->stride; + tc0 = (GLuint (*)[4])VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->data; + if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->size < 3) { fill_tex |= (1<<0); } - else if (DO_PTEX && VB->TexCoordPtr[t0]->size < 4) { + else if (DO_PTEX && VB->AttribPtr[_TNL_ATTRIB_TEX0 + t0]->size < 4) { rqcoordsnoswap |= (1<<0); } } else { @@ -112,9 +112,9 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_NORM) { - if (VB->NormalPtr) { - norm_stride = VB->NormalPtr->stride; - norm = (GLuint (*)[4])VB->NormalPtr->data; + if (VB->AttribPtr[_TNL_ATTRIB_NORMAL]) { + norm_stride = VB->AttribPtr[_TNL_ATTRIB_NORMAL]->stride; + norm = (GLuint (*)[4])VB->AttribPtr[_TNL_ATTRIB_NORMAL]->data; } else { norm_stride = 0; norm = (GLuint (*)[4])&ctx->Current.Attrib[VERT_ATTRIB_NORMAL]; @@ -122,9 +122,9 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_RGBA) { - if (VB->ColorPtr[0]) { - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; + if (VB->AttribPtr[_TNL_ATTRIB_COLOR0]) { + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; } else { col = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_COLOR0]; col_stride = 0; @@ -132,9 +132,9 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_SPEC_OR_FOG) { - if (VB->SecondaryColorPtr[0]) { - spec = VB->SecondaryColorPtr[0]->data; - spec_stride = VB->SecondaryColorPtr[0]->stride; + if (VB->AttribPtr[_TNL_ATTRIB_COLOR1]) { + spec = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->data; + spec_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->stride; } else { spec = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_COLOR1]; spec_stride = 0; @@ -142,9 +142,9 @@ static void TAG(emit)( GLcontext *ctx, } if (DO_SPEC_OR_FOG) { - if (VB->FogCoordPtr) { - fog = VB->FogCoordPtr->data; - fog_stride = VB->FogCoordPtr->stride; + if (VB->AttribPtr[_TNL_ATTRIB_FOG]) { + fog = VB->AttribPtr[_TNL_ATTRIB_FOG]->data; + fog_stride = VB->AttribPtr[_TNL_ATTRIB_FOG]->stride; } else { fog = (GLfloat (*)[4])ctx->Current.Attrib[VERT_ATTRIB_FOG]; fog_stride = 0; diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c index 78ec1193026..98f96ff2a76 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c +++ b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c @@ -326,7 +326,7 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) if (1) { req |= RADEON_CP_VC_FRMT_Z; - if (VB->ObjPtr->size == 4) { + if (VB->AttribPtr[_TNL_ATTRIB_POS]->size == 4) { req |= RADEON_CP_VC_FRMT_W0; } } @@ -348,15 +348,15 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) req |= RADEON_ST_BIT(unit); /* assume we need the 3rd coord if texgen is active for r/q OR at least 3 coords are submitted. This may not be 100% correct */ - if (VB->TexCoordPtr[unit]->size >= 3) { + if (VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size >= 3) { req |= RADEON_Q_BIT(unit); vtx |= RADEON_Q_BIT(unit); } if ( (ctx->Texture.Unit[unit].TexGenEnabled & (R_BIT | Q_BIT)) ) vtx |= RADEON_Q_BIT(unit); - else if ((VB->TexCoordPtr[unit]->size >= 3) && + else if ((VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size >= 3) && ((ctx->Texture.Unit[unit]._ReallyEnabled & (TEXTURE_CUBE_BIT)) == 0)) { - GLuint swaptexmatcol = (VB->TexCoordPtr[unit]->size - 3); + GLuint swaptexmatcol = (VB->AttribPtr[_TNL_ATTRIB_TEX0 + unit]->size - 3); if (((rmesa->NeedTexMatrix >> unit) & 1) && (swaptexmatcol != ((rmesa->TexMatColSwap >> unit) & 1))) radeonUploadTexMatrix( rmesa, unit, swaptexmatcol ) ; @@ -390,19 +390,19 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) * this, add more vertex code (for obj-2, obj-3) or preferably move * to maos. */ - if (VB->ObjPtr->size < 3 || - (VB->ObjPtr->size == 3 && + if (VB->AttribPtr[_TNL_ATTRIB_POS]->size < 3 || + (VB->AttribPtr[_TNL_ATTRIB_POS]->size == 3 && (setup_tab[i].vertex_format & RADEON_CP_VC_FRMT_W0))) { _math_trans_4f( rmesa->tcl.ObjClean.data, - VB->ObjPtr->data, - VB->ObjPtr->stride, + VB->AttribPtr[_TNL_ATTRIB_POS]->data, + VB->AttribPtr[_TNL_ATTRIB_POS]->stride, GL_FLOAT, - VB->ObjPtr->size, + VB->AttribPtr[_TNL_ATTRIB_POS]->size, 0, VB->Count ); - switch (VB->ObjPtr->size) { + switch (VB->AttribPtr[_TNL_ATTRIB_POS]->size) { case 1: _mesa_vector4f_clean_elem(&rmesa->tcl.ObjClean, VB->Count, 1); case 2: @@ -416,14 +416,14 @@ void radeonEmitArrays( GLcontext *ctx, GLuint inputs ) break; } - VB->ObjPtr = &rmesa->tcl.ObjClean; + VB->AttribPtr[_TNL_ATTRIB_POS] = &rmesa->tcl.ObjClean; } - + radeon_bo_map(rmesa->radeon.tcl.aos[0].bo, 1); setup_tab[i].emit( ctx, 0, VB->Count, rmesa->radeon.tcl.aos[0].bo->ptr + rmesa->radeon.tcl.aos[0].offset); - + radeon_bo_unmap(rmesa->radeon.tcl.aos[0].bo); // rmesa->radeon.tcl.aos[0].size = setup_tab[i].vertex_size; rmesa->radeon.tcl.aos[0].stride = setup_tab[i].vertex_size; rmesa->tcl.vertex_format = setup_tab[i].vertex_format; diff --git a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c index 38db305e2a7..a7f347202a1 100644 --- a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c +++ b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c @@ -1,4 +1,5 @@ /* + * Copyright (C) 2009 Maciej Cencora. * Copyright (C) 2008 Nicolai Haehnle. * * All Rights Reserved. @@ -32,51 +33,52 @@ #include "main/simple_list.h" #include "main/texcompress.h" -#include "main/texformat.h" - -static GLuint radeon_compressed_texture_size(GLcontext *ctx, - GLsizei width, GLsizei height, GLsizei depth, - GLuint mesaFormat) +#include "main/teximage.h" +#include "main/texobj.h" +#include "radeon_texture.h" + +static unsigned get_aligned_compressed_row_stride( + gl_format format, + unsigned width, + unsigned minStride) { - GLuint size = _mesa_compressed_texture_size(ctx, width, height, depth, mesaFormat); - - if (mesaFormat == MESA_FORMAT_RGB_DXT1 || - mesaFormat == MESA_FORMAT_RGBA_DXT1) { - if (width + 3 < 8) /* width one block */ - size = size * 4; - else if (width + 3 < 16) - size = size * 2; - } else { - /* DXT3/5, 16 bytes per block */ - // WARN_ONCE("DXT 3/5 suffers from multitexturing problems!\n"); - if (width + 3 < 8) - size = size * 2; + const unsigned blockSize = _mesa_get_format_bytes(format); + unsigned blockWidth, blockHeight, numXBlocks; + + _mesa_get_format_block_size(format, &blockWidth, &blockHeight); + numXBlocks = (width + blockWidth - 1) / blockWidth; + + while (numXBlocks * blockSize < minStride) + { + ++numXBlocks; } - return size; + return numXBlocks * blockSize; } +static unsigned get_compressed_image_size( + gl_format format, + unsigned rowStride, + unsigned height) +{ + unsigned blockWidth, blockHeight; + + _mesa_get_format_block_size(format, &blockWidth, &blockHeight); -static int radeon_compressed_num_bytes(GLuint mesaFormat) + return rowStride * ((height + blockHeight - 1) / blockHeight); +} + +static int find_next_power_of_two(GLuint value) { - int bytes = 0; - switch(mesaFormat) { - - case MESA_FORMAT_RGB_FXT1: - case MESA_FORMAT_RGBA_FXT1: - case MESA_FORMAT_RGB_DXT1: - case MESA_FORMAT_RGBA_DXT1: - bytes = 2; - break; - - case MESA_FORMAT_RGBA_DXT3: - case MESA_FORMAT_RGBA_DXT5: - bytes = 4; - default: - break; - } - - return bytes; + int i, tmp; + + i = 0; + tmp = value - 1; + while (tmp) { + tmp >>= 1; + i++; + } + return (1 << i); } /** @@ -91,28 +93,28 @@ static void compute_tex_image_offset(radeonContextPtr rmesa, radeon_mipmap_tree { radeon_mipmap_level *lvl = &mt->levels[level]; uint32_t row_align; + GLuint height; + + height = find_next_power_of_two(lvl->height); /* Find image size in bytes */ - if (mt->compressed) { - /* TODO: Is this correct? Need test cases for compressed textures! */ - row_align = rmesa->texture_compressed_row_align - 1; - lvl->rowstride = (lvl->width * mt->bpp + row_align) & ~row_align; - lvl->size = radeon_compressed_texture_size(mt->radeon->glCtx, - lvl->width, lvl->height, lvl->depth, mt->compressed); + if (_mesa_is_format_compressed(mt->mesaFormat)) { + lvl->rowstride = get_aligned_compressed_row_stride(mt->mesaFormat, lvl->width, rmesa->texture_compressed_row_align); + lvl->size = get_compressed_image_size(mt->mesaFormat, lvl->rowstride, height); } else if (mt->target == GL_TEXTURE_RECTANGLE_NV) { row_align = rmesa->texture_rect_row_align - 1; - lvl->rowstride = (lvl->width * mt->bpp + row_align) & ~row_align; - lvl->size = lvl->rowstride * lvl->height; + lvl->rowstride = (_mesa_format_row_stride(mt->mesaFormat, lvl->width) + row_align) & ~row_align; + lvl->size = lvl->rowstride * height; } else if (mt->tilebits & RADEON_TXO_MICRO_TILE) { /* tile pattern is 16 bytes x2. mipmaps stay 32 byte aligned, * though the actual offset may be different (if texture is less than * 32 bytes width) to the untiled case */ - lvl->rowstride = (lvl->width * mt->bpp * 2 + 31) & ~31; - lvl->size = lvl->rowstride * ((lvl->height + 1) / 2) * lvl->depth; + lvl->rowstride = (_mesa_format_row_stride(mt->mesaFormat, lvl->width) * 2 + 31) & ~31; + lvl->size = lvl->rowstride * ((height + 1) / 2) * lvl->depth; } else { row_align = rmesa->texture_row_align - 1; - lvl->rowstride = (lvl->width * mt->bpp + row_align) & ~row_align; - lvl->size = lvl->rowstride * lvl->height * lvl->depth; + lvl->rowstride = (_mesa_format_row_stride(mt->mesaFormat, lvl->width) + row_align) & ~row_align; + lvl->size = lvl->rowstride * height * lvl->depth; } assert(lvl->size > 0); @@ -124,7 +126,7 @@ static void compute_tex_image_offset(radeonContextPtr rmesa, radeon_mipmap_tree if (RADEON_DEBUG & RADEON_TEXTURE) fprintf(stderr, "level %d, face %d: rs:%d %dx%d at %d\n", - level, face, lvl->rowstride, lvl->width, lvl->height, lvl->faces[face].offset); + level, face, lvl->rowstride, lvl->width, height, lvl->faces[face].offset); } static GLuint minify(GLuint size, GLuint levels) @@ -138,22 +140,19 @@ static GLuint minify(GLuint size, GLuint levels) static void calculate_miptree_layout_r100(radeonContextPtr rmesa, radeon_mipmap_tree *mt) { - GLuint curOffset; - GLuint numLevels; - GLuint i; - GLuint face; + GLuint curOffset, i, face, level; - numLevels = mt->lastLevel - mt->firstLevel + 1; - assert(numLevels <= rmesa->glCtx->Const.MaxTextureLevels); + assert(mt->numLevels <= rmesa->glCtx->Const.MaxTextureLevels); curOffset = 0; for(face = 0; face < mt->faces; face++) { - for(i = 0; i < numLevels; i++) { - mt->levels[i].width = minify(mt->width0, i); - mt->levels[i].height = minify(mt->height0, i); - mt->levels[i].depth = minify(mt->depth0, i); - compute_tex_image_offset(rmesa, mt, face, i, &curOffset); + for(i = 0, level = mt->baseLevel; i < mt->numLevels; i++, level++) { + mt->levels[level].valid = 1; + mt->levels[level].width = minify(mt->width0, i); + mt->levels[level].height = minify(mt->height0, i); + mt->levels[level].depth = minify(mt->depth0, i); + compute_tex_image_offset(rmesa, mt, face, level, &curOffset); } } @@ -163,23 +162,21 @@ static void calculate_miptree_layout_r100(radeonContextPtr rmesa, radeon_mipmap_ static void calculate_miptree_layout_r300(radeonContextPtr rmesa, radeon_mipmap_tree *mt) { - GLuint curOffset; - GLuint numLevels; - GLuint i; + GLuint curOffset, i, level; - numLevels = mt->lastLevel - mt->firstLevel + 1; - assert(numLevels <= rmesa->glCtx->Const.MaxTextureLevels); + assert(mt->numLevels <= rmesa->glCtx->Const.MaxTextureLevels); curOffset = 0; - for(i = 0; i < numLevels; i++) { + for(i = 0, level = mt->baseLevel; i < mt->numLevels; i++, level++) { GLuint face; - mt->levels[i].width = minify(mt->width0, i); - mt->levels[i].height = minify(mt->height0, i); - mt->levels[i].depth = minify(mt->depth0, i); + mt->levels[level].valid = 1; + mt->levels[level].width = minify(mt->width0, i); + mt->levels[level].height = minify(mt->height0, i); + mt->levels[level].depth = minify(mt->depth0, i); for(face = 0; face < mt->faces; face++) - compute_tex_image_offset(rmesa, mt, face, i, &curOffset); + compute_tex_image_offset(rmesa, mt, face, level, &curOffset); } /* Note the required size in memory */ @@ -189,27 +186,22 @@ static void calculate_miptree_layout_r300(radeonContextPtr rmesa, radeon_mipmap_ /** * Create a new mipmap tree, calculate its layout and allocate memory. */ -radeon_mipmap_tree* radeon_miptree_create(radeonContextPtr rmesa, radeonTexObj *t, - GLenum target, GLenum internal_format, GLuint firstLevel, GLuint lastLevel, - GLuint width0, GLuint height0, GLuint depth0, - GLuint bpp, GLuint tilebits, GLuint compressed) +static radeon_mipmap_tree* radeon_miptree_create(radeonContextPtr rmesa, + GLenum target, gl_format mesaFormat, GLuint baseLevel, GLuint numLevels, + GLuint width0, GLuint height0, GLuint depth0, GLuint tilebits) { radeon_mipmap_tree *mt = CALLOC_STRUCT(_radeon_mipmap_tree); - mt->radeon = rmesa; - mt->internal_format = internal_format; + mt->mesaFormat = mesaFormat; mt->refcount = 1; - mt->t = t; mt->target = target; mt->faces = (target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; - mt->firstLevel = firstLevel; - mt->lastLevel = lastLevel; + mt->baseLevel = baseLevel; + mt->numLevels = numLevels; mt->width0 = width0; mt->height0 = height0; mt->depth0 = depth0; - mt->bpp = compressed ? radeon_compressed_num_bytes(compressed) : bpp; mt->tilebits = tilebits; - mt->compressed = compressed; if (rmesa->radeonScreen->chip_family >= CHIP_FAMILY_R300) calculate_miptree_layout_r300(rmesa, mt); @@ -224,53 +216,43 @@ radeon_mipmap_tree* radeon_miptree_create(radeonContextPtr rmesa, radeonTexObj * return mt; } -void radeon_miptree_reference(radeon_mipmap_tree *mt) +void radeon_miptree_reference(radeon_mipmap_tree *mt, radeon_mipmap_tree **ptr) { + assert(!*ptr); + mt->refcount++; assert(mt->refcount > 0); + + *ptr = mt; } -void radeon_miptree_unreference(radeon_mipmap_tree *mt) +void radeon_miptree_unreference(radeon_mipmap_tree **ptr) { + radeon_mipmap_tree *mt = *ptr; if (!mt) return; assert(mt->refcount > 0); + mt->refcount--; if (!mt->refcount) { radeon_bo_unref(mt->bo); free(mt); } -} + *ptr = 0; +} /** - * Calculate first and last mip levels for the given texture object, - * where the dimensions are taken from the given texture image at - * the given level. - * - * Note: level is the OpenGL level number, which is not necessarily the same - * as the first level that is actually present. - * - * The base level image of the given texture face must be non-null, - * or this will fail. + * Calculate min and max LOD for the given texture object. + * @param[in] tObj texture object whose LOD values to calculate + * @param[out] pminLod minimal LOD + * @param[out] pmaxLod maximal LOD */ -static void calculate_first_last_level(struct gl_texture_object *tObj, - GLuint *pfirstLevel, GLuint *plastLevel, - GLuint face, GLuint level) +static void calculate_min_max_lod(struct gl_texture_object *tObj, + unsigned *pminLod, unsigned *pmaxLod) { - const struct gl_texture_image * const baseImage = - tObj->Image[face][level]; - - assert(baseImage); - - /* These must be signed values. MinLod and MaxLod can be negative numbers, - * and having firstLevel and lastLevel as signed prevents the need for - * extra sign checks. - */ - int firstLevel; - int lastLevel; - + int minLod, maxLod; /* Yes, this looks overly complicated, but it's all needed. */ switch (tObj->Target) { @@ -281,32 +263,30 @@ static void calculate_first_last_level(struct gl_texture_object *tObj, if (tObj->MinFilter == GL_NEAREST || tObj->MinFilter == GL_LINEAR) { /* GL_NEAREST and GL_LINEAR only care about GL_TEXTURE_BASE_LEVEL. */ - firstLevel = lastLevel = tObj->BaseLevel; + minLod = maxLod = tObj->BaseLevel; } else { - firstLevel = tObj->BaseLevel + (GLint)(tObj->MinLod + 0.5); - firstLevel = MAX2(firstLevel, tObj->BaseLevel); - firstLevel = MIN2(firstLevel, level + baseImage->MaxLog2); - lastLevel = tObj->BaseLevel + (GLint)(tObj->MaxLod + 0.5); - lastLevel = MAX2(lastLevel, tObj->BaseLevel); - lastLevel = MIN2(lastLevel, level + baseImage->MaxLog2); - lastLevel = MIN2(lastLevel, tObj->MaxLevel); - lastLevel = MAX2(firstLevel, lastLevel); /* need at least one level */ + minLod = tObj->BaseLevel + (GLint)(tObj->MinLod); + minLod = MAX2(minLod, tObj->BaseLevel); + minLod = MIN2(minLod, tObj->MaxLevel); + maxLod = tObj->BaseLevel + (GLint)(tObj->MaxLod + 0.5); + maxLod = MIN2(maxLod, tObj->MaxLevel); + maxLod = MIN2(maxLod, tObj->Image[0][minLod]->MaxLog2 + minLod); + maxLod = MAX2(maxLod, minLod); /* need at least one level */ } break; case GL_TEXTURE_RECTANGLE_NV: case GL_TEXTURE_4D_SGIS: - firstLevel = lastLevel = 0; + minLod = maxLod = 0; break; default: return; } /* save these values */ - *pfirstLevel = firstLevel; - *plastLevel = lastLevel; + *pminLod = minLod; + *pmaxLod = maxLod; } - /** * Checks whether the given miptree can hold the given texture image at the * given face and level. @@ -316,20 +296,15 @@ GLboolean radeon_miptree_matches_image(radeon_mipmap_tree *mt, { radeon_mipmap_level *lvl; - if (face >= mt->faces || level < mt->firstLevel || level > mt->lastLevel) + if (face >= mt->faces) return GL_FALSE; - if (texImage->InternalFormat != mt->internal_format || - texImage->IsCompressed != mt->compressed) + if (texImage->TexFormat != mt->mesaFormat) return GL_FALSE; - if (!texImage->IsCompressed && - !mt->compressed && - texImage->TexFormat->TexelBytes != mt->bpp) - return GL_FALSE; - - lvl = &mt->levels[level - mt->firstLevel]; - if (lvl->width != texImage->Width || + lvl = &mt->levels[level]; + if (!lvl->valid || + lvl->width != texImage->Width || lvl->height != texImage->Height || lvl->depth != texImage->Depth) return GL_FALSE; @@ -337,59 +312,72 @@ GLboolean radeon_miptree_matches_image(radeon_mipmap_tree *mt, return GL_TRUE; } - /** * Checks whether the given miptree has the right format to store the given texture object. */ -GLboolean radeon_miptree_matches_texture(radeon_mipmap_tree *mt, struct gl_texture_object *texObj) +static GLboolean radeon_miptree_matches_texture(radeon_mipmap_tree *mt, struct gl_texture_object *texObj) { struct gl_texture_image *firstImage; - GLuint compressed; - GLuint numfaces = 1; - GLuint firstLevel, lastLevel; - - calculate_first_last_level(texObj, &firstLevel, &lastLevel, 0, texObj->BaseLevel); - if (texObj->Target == GL_TEXTURE_CUBE_MAP) - numfaces = 6; - - firstImage = texObj->Image[0][firstLevel]; - compressed = firstImage->IsCompressed ? firstImage->TexFormat->MesaFormat : 0; - - return (mt->firstLevel == firstLevel && - mt->lastLevel == lastLevel && - mt->width0 == firstImage->Width && - mt->height0 == firstImage->Height && - mt->depth0 == firstImage->Depth && - mt->compressed == compressed && - (!mt->compressed ? (mt->bpp == firstImage->TexFormat->TexelBytes) : 1)); -} + unsigned numLevels; + radeon_mipmap_level *mtBaseLevel; + + if (texObj->BaseLevel < mt->baseLevel) + return GL_FALSE; + + mtBaseLevel = &mt->levels[texObj->BaseLevel - mt->baseLevel]; + firstImage = texObj->Image[0][texObj->BaseLevel]; + numLevels = MIN2(texObj->MaxLevel - texObj->BaseLevel + 1, firstImage->MaxLog2 + 1); + + if (RADEON_DEBUG & RADEON_TEXTURE) { + fprintf(stderr, "Checking if miptree %p matches texObj %p\n", mt, texObj); + fprintf(stderr, "target %d vs %d\n", mt->target, texObj->Target); + fprintf(stderr, "format %d vs %d\n", mt->mesaFormat, firstImage->TexFormat); + fprintf(stderr, "numLevels %d vs %d\n", mt->numLevels, numLevels); + fprintf(stderr, "width0 %d vs %d\n", mtBaseLevel->width, firstImage->Width); + fprintf(stderr, "height0 %d vs %d\n", mtBaseLevel->height, firstImage->Height); + fprintf(stderr, "depth0 %d vs %d\n", mtBaseLevel->depth, firstImage->Depth); + if (mt->target == texObj->Target && + mt->mesaFormat == firstImage->TexFormat && + mt->numLevels >= numLevels && + mtBaseLevel->width == firstImage->Width && + mtBaseLevel->height == firstImage->Height && + mtBaseLevel->depth == firstImage->Depth) { + fprintf(stderr, "MATCHED\n"); + } else { + fprintf(stderr, "NOT MATCHED\n"); + } + } + return (mt->target == texObj->Target && + mt->mesaFormat == firstImage->TexFormat && + mt->numLevels >= numLevels && + mtBaseLevel->width == firstImage->Width && + mtBaseLevel->height == firstImage->Height && + mtBaseLevel->depth == firstImage->Depth); +} /** - * Try to allocate a mipmap tree for the given texture that will fit the - * given image in the given position. + * Try to allocate a mipmap tree for the given texture object. + * @param[in] rmesa radeon context + * @param[in] t radeon texture object */ -void radeon_try_alloc_miptree(radeonContextPtr rmesa, radeonTexObj *t, - radeon_texture_image *image, GLuint face, GLuint level) +void radeon_try_alloc_miptree(radeonContextPtr rmesa, radeonTexObj *t) { - GLuint compressed = image->base.IsCompressed ? image->base.TexFormat->MesaFormat : 0; - GLuint numfaces = 1; - GLuint firstLevel, lastLevel; + struct gl_texture_object *texObj = &t->base; + struct gl_texture_image *texImg = texObj->Image[0][texObj->BaseLevel]; + GLuint numLevels; assert(!t->mt); - calculate_first_last_level(&t->base, &firstLevel, &lastLevel, face, level); - if (t->base.Target == GL_TEXTURE_CUBE_MAP) - numfaces = 6; - - if (level != firstLevel || face >= numfaces) + if (!texImg) return; - t->mt = radeon_miptree_create(rmesa, t, t->base.Target, - image->base.InternalFormat, - firstLevel, lastLevel, - image->base.Width, image->base.Height, image->base.Depth, - image->base.TexFormat->TexelBytes, t->tile_bits, compressed); + numLevels = MIN2(texObj->MaxLevel - texObj->BaseLevel + 1, texImg->MaxLog2 + 1); + + t->mt = radeon_miptree_create(rmesa, t->base.Target, + texImg->TexFormat, texObj->BaseLevel, + numLevels, texImg->Width, texImg->Height, + texImg->Depth, t->tile_bits); } /* Although we use the image_offset[] array to store relative offsets @@ -401,21 +389,231 @@ void radeon_try_alloc_miptree(radeonContextPtr rmesa, radeonTexObj *t, void radeon_miptree_depth_offsets(radeon_mipmap_tree *mt, GLuint level, GLuint *offsets) { - if (mt->target != GL_TEXTURE_3D || mt->faces == 1) - offsets[0] = 0; - else { - int i; - for (i = 0; i < 6; i++) - offsets[i] = mt->levels[level].faces[i].offset; - } + if (mt->target != GL_TEXTURE_3D || mt->faces == 1) { + offsets[0] = 0; + } else { + int i; + for (i = 0; i < 6; i++) { + offsets[i] = mt->levels[level].faces[i].offset; + } + } } GLuint radeon_miptree_image_offset(radeon_mipmap_tree *mt, GLuint face, GLuint level) { - if (mt->target == GL_TEXTURE_CUBE_MAP_ARB) - return (mt->levels[level].faces[face].offset); - else - return mt->levels[level].faces[0].offset; + if (mt->target == GL_TEXTURE_CUBE_MAP_ARB) + return (mt->levels[level].faces[face].offset); + else + return mt->levels[level].faces[0].offset; +} + +/** + * Ensure that the given image is stored in the given miptree from now on. + */ +static void migrate_image_to_miptree(radeon_mipmap_tree *mt, + radeon_texture_image *image, + int face, int level) +{ + radeon_mipmap_level *dstlvl = &mt->levels[level]; + unsigned char *dest; + + assert(image->mt != mt); + assert(dstlvl->valid); + assert(dstlvl->width == image->base.Width); + assert(dstlvl->height == image->base.Height); + assert(dstlvl->depth == image->base.Depth); + + radeon_bo_map(mt->bo, GL_TRUE); + dest = mt->bo->ptr + dstlvl->faces[face].offset; + + if (image->mt) { + /* Format etc. should match, so we really just need a memcpy(). + * In fact, that memcpy() could be done by the hardware in many + * cases, provided that we have a proper memory manager. + */ + assert(mt->mesaFormat == image->base.TexFormat); + + radeon_mipmap_level *srclvl = &image->mt->levels[image->mtlevel]; + + /* TODO: bring back these assertions once the FBOs are fixed */ +#if 0 + assert(image->mtlevel == level); + assert(srclvl->size == dstlvl->size); + assert(srclvl->rowstride == dstlvl->rowstride); +#endif + + radeon_bo_map(image->mt->bo, GL_FALSE); + + memcpy(dest, + image->mt->bo->ptr + srclvl->faces[face].offset, + dstlvl->size); + radeon_bo_unmap(image->mt->bo); + + radeon_miptree_unreference(&image->mt); + } else { + const uint32_t srcrowstride = _mesa_format_row_stride(image->base.TexFormat, image->base.Width); + uint32_t rows = image->base.Height * image->base.Depth; + + if (_mesa_is_format_compressed(image->base.TexFormat)) { + uint32_t blockWidth, blockHeight; + _mesa_get_format_block_size(image->base.TexFormat, &blockWidth, &blockHeight); + rows = (rows + blockHeight - 1) / blockHeight; + } + + copy_rows(dest, dstlvl->rowstride, image->base.Data, srcrowstride, + rows, srcrowstride); + + _mesa_free_texmemory(image->base.Data); + image->base.Data = 0; + } + + radeon_bo_unmap(mt->bo); + + radeon_miptree_reference(mt, &image->mt); + image->mtface = face; + image->mtlevel = level; +} + +/** + * Filter matching miptrees, and select one with the most of data. + * @param[in] texObj radeon texture object + * @param[in] firstLevel first texture level to check + * @param[in] lastLevel last texture level to check + */ +static radeon_mipmap_tree * get_biggest_matching_miptree(radeonTexObj *texObj, + unsigned firstLevel, + unsigned lastLevel) +{ + const unsigned numLevels = lastLevel - firstLevel + 1; + unsigned *mtSizes = calloc(numLevels, sizeof(unsigned)); + radeon_mipmap_tree **mts = calloc(numLevels, sizeof(radeon_mipmap_tree *)); + unsigned mtCount = 0; + unsigned maxMtIndex = 0; + radeon_mipmap_tree *tmp; + + for (unsigned level = firstLevel; level <= lastLevel; ++level) { + radeon_texture_image *img = get_radeon_texture_image(texObj->base.Image[0][level]); + unsigned found = 0; + // TODO: why this hack?? + if (!img) + break; + + if (!img->mt) + continue; + + for (int i = 0; i < mtCount; ++i) { + if (mts[i] == img->mt) { + found = 1; + mtSizes[i] += img->mt->levels[img->mtlevel].size; + break; + } + } + + if (!found && radeon_miptree_matches_texture(img->mt, &texObj->base)) { + mtSizes[mtCount] = img->mt->levels[img->mtlevel].size; + mts[mtCount] = img->mt; + mtCount++; + } + } + + if (mtCount == 0) { + return NULL; + } + + for (int i = 1; i < mtCount; ++i) { + if (mtSizes[i] > mtSizes[maxMtIndex]) { + maxMtIndex = i; + } + } + + tmp = mts[maxMtIndex]; + free(mtSizes); + free(mts); + + return tmp; } + +/** + * Validate texture mipmap tree. + * If individual images are stored in different mipmap trees + * use the mipmap tree that has the most of the correct data. + */ +int radeon_validate_texture_miptree(GLcontext * ctx, struct gl_texture_object *texObj) +{ + radeonContextPtr rmesa = RADEON_CONTEXT(ctx); + radeonTexObj *t = radeon_tex_obj(texObj); + + if (t->validated || t->image_override) { + return GL_TRUE; + } + + if (texObj->Image[0][texObj->BaseLevel]->Border > 0) + return GL_FALSE; + + _mesa_test_texobj_completeness(rmesa->glCtx, texObj); + if (!texObj->_Complete) { + return GL_FALSE; + } + + calculate_min_max_lod(&t->base, &t->minLod, &t->maxLod); + + if (RADEON_DEBUG & RADEON_TEXTURE) + fprintf(stderr, "%s: Validating texture %p now, minLod = %d, maxLod = %d\n", + __FUNCTION__, texObj ,t->minLod, t->maxLod); + + radeon_mipmap_tree *dst_miptree; + dst_miptree = get_biggest_matching_miptree(t, t->minLod, t->maxLod); + + if (!dst_miptree) { + radeon_miptree_unreference(&t->mt); + radeon_try_alloc_miptree(rmesa, t); + dst_miptree = t->mt; + if (RADEON_DEBUG & RADEON_TEXTURE) { + fprintf(stderr, "%s: No matching miptree found, allocated new one %p\n", __FUNCTION__, t->mt); + } + } else if (RADEON_DEBUG & RADEON_TEXTURE) { + fprintf(stderr, "%s: Using miptree %p\n", __FUNCTION__, t->mt); + } + + const unsigned faces = texObj->Target == GL_TEXTURE_CUBE_MAP ? 6 : 1; + unsigned face, level; + radeon_texture_image *img; + /* Validate only the levels that will actually be used during rendering */ + for (face = 0; face < faces; ++face) { + for (level = t->minLod; level <= t->maxLod; ++level) { + img = get_radeon_texture_image(texObj->Image[face][level]); + + if (RADEON_DEBUG & RADEON_TEXTURE) { + fprintf(stderr, "Checking image level %d, face %d, mt %p ... ", level, face, img->mt); + } + + if (img->mt != dst_miptree) { + if (RADEON_DEBUG & RADEON_TEXTURE) { + fprintf(stderr, "MIGRATING\n"); + } + struct radeon_bo *src_bo = (img->mt) ? img->mt->bo : img->bo; + if (src_bo && radeon_bo_is_referenced_by_cs(src_bo, rmesa->cmdbuf.cs)) { + radeon_firevertices(rmesa); + } + migrate_image_to_miptree(dst_miptree, img, face, level); + } else if (RADEON_DEBUG & RADEON_TEXTURE) { + fprintf(stderr, "OK\n"); + } + } + } + + t->validated = GL_TRUE; + + return GL_TRUE; +} + +uint32_t get_base_teximage_offset(radeonTexObj *texObj) +{ + if (!texObj->mt) { + return 0; + } else { + return radeon_miptree_image_offset(texObj->mt, 0, texObj->minLod); + } +}
\ No newline at end of file diff --git a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h index db28252da37..a10649b5aea 100644 --- a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h +++ b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h @@ -44,6 +44,7 @@ struct _radeon_mipmap_level { GLuint depth; GLuint size; /** Size of each image, in bytes */ GLuint rowstride; /** in bytes */ + GLuint valid; radeon_mipmap_image faces[6]; }; @@ -59,43 +60,35 @@ struct _radeon_mipmap_level { * changed. */ struct _radeon_mipmap_tree { - radeonContextPtr radeon; - radeonTexObj *t; struct radeon_bo *bo; GLuint refcount; GLuint totalsize; /** total size of the miptree, in bytes */ GLenum target; /** GL_TEXTURE_xxx */ - GLenum internal_format; + GLenum mesaFormat; /** MESA_FORMAT_xxx */ GLuint faces; /** # of faces: 6 for cubemaps, 1 otherwise */ - GLuint firstLevel; /** First mip level stored in this mipmap tree */ - GLuint lastLevel; /** Last mip level stored in this mipmap tree */ + GLuint baseLevel; /** gl_texture_object->baseLevel it was created for */ + GLuint numLevels; /** Number of mip levels stored in this mipmap tree */ - GLuint width0; /** Width of firstLevel image */ - GLuint height0; /** Height of firstLevel image */ - GLuint depth0; /** Depth of firstLevel image */ + GLuint width0; /** Width of baseLevel image */ + GLuint height0; /** Height of baseLevel image */ + GLuint depth0; /** Depth of baseLevel image */ - GLuint bpp; /** Bytes per texel */ GLuint tilebits; /** RADEON_TXO_xxx_TILE */ - GLuint compressed; /** MESA_FORMAT_xxx indicating a compressed format, or 0 if uncompressed */ radeon_mipmap_level levels[RADEON_MIPTREE_MAX_TEXTURE_LEVELS]; }; -radeon_mipmap_tree* radeon_miptree_create(radeonContextPtr rmesa, radeonTexObj *t, - GLenum target, GLenum internal_format, GLuint firstLevel, GLuint lastLevel, - GLuint width0, GLuint height0, GLuint depth0, - GLuint bpp, GLuint tilebits, GLuint compressed); -void radeon_miptree_reference(radeon_mipmap_tree *mt); -void radeon_miptree_unreference(radeon_mipmap_tree *mt); +void radeon_miptree_reference(radeon_mipmap_tree *mt, radeon_mipmap_tree **ptr); +void radeon_miptree_unreference(radeon_mipmap_tree **ptr); GLboolean radeon_miptree_matches_image(radeon_mipmap_tree *mt, struct gl_texture_image *texImage, GLuint face, GLuint level); -GLboolean radeon_miptree_matches_texture(radeon_mipmap_tree *mt, struct gl_texture_object *texObj); -void radeon_try_alloc_miptree(radeonContextPtr rmesa, radeonTexObj *t, - radeon_texture_image *texImage, GLuint face, GLuint level); +void radeon_try_alloc_miptree(radeonContextPtr rmesa, radeonTexObj *t); GLuint radeon_miptree_image_offset(radeon_mipmap_tree *mt, GLuint face, GLuint level); void radeon_miptree_depth_offsets(radeon_mipmap_tree *mt, GLuint level, GLuint *offsets); + +uint32_t get_base_teximage_offset(radeonTexObj *texObj); #endif /* __RADEON_MIPMAP_TREE_H_ */ diff --git a/src/mesa/drivers/dri/radeon/radeon_queryobj.c b/src/mesa/drivers/dri/radeon/radeon_queryobj.c index b79d864ba29..98117cdfc14 100644 --- a/src/mesa/drivers/dri/radeon/radeon_queryobj.c +++ b/src/mesa/drivers/dri/radeon/radeon_queryobj.c @@ -31,24 +31,11 @@ #include "main/imports.h" #include "main/simple_list.h" -static int radeonQueryIsFlushed(GLcontext *ctx, struct gl_query_object *q) -{ - radeonContextPtr radeon = RADEON_CONTEXT(ctx); - struct radeon_query_object *tmp, *query = (struct radeon_query_object *)q; - - foreach(tmp, &radeon->query.not_flushed_head) { - if (tmp == query) { - return 0; - } - } - - return 1; -} - static void radeonQueryGetResult(GLcontext *ctx, struct gl_query_object *q) { + radeonContextPtr radeon = RADEON_CONTEXT(ctx); struct radeon_query_object *query = (struct radeon_query_object *)q; - uint32_t *result; + uint32_t *result; int i; radeon_print(RADEON_STATE, RADEON_VERBOSE, @@ -56,13 +43,35 @@ static void radeonQueryGetResult(GLcontext *ctx, struct gl_query_object *q) __FUNCTION__, query->Base.Id, (int) query->Base.Result); radeon_bo_map(query->bo, GL_FALSE); - - result = query->bo->ptr; + result = query->bo->ptr; query->Base.Result = 0; - for (i = 0; i < query->curr_offset/sizeof(uint32_t); ++i) { - query->Base.Result += result[i]; - radeon_print(RADEON_STATE, RADEON_TRACE, "result[%d] = %d\n", i, result[i]); + if (IS_R600_CLASS(radeon->radeonScreen)) { + /* ZPASS EVENT writes alternating qwords + * At query start we set the start offset to 0 and + * hw writes zpass start counts to qwords 0, 2, 4, 6. + * At query end we set the start offset to 8 and + * hw writes zpass end counts to qwords 1, 3, 5, 7. + * then we substract. MSB is the valid bit. + */ + for (i = 0; i < 16; i += 4) { + uint64_t start = (uint64_t)LE32_TO_CPU(result[i]) | + (uint64_t)LE32_TO_CPU(result[i + 1]) << 32; + uint64_t end = (uint64_t)LE32_TO_CPU(result[i + 2]) | + (uint64_t)LE32_TO_CPU(result[i + 3]) << 32; + if ((start & 0x8000000000000000) && (end & 0x8000000000000000)) { + uint64_t query_count = end - start; + query->Base.Result += query_count; + + } + radeon_print(RADEON_STATE, RADEON_TRACE, + "%d start: %lx, end: %lx %ld\n", i, start, end, end - start); + } + } else { + for (i = 0; i < query->curr_offset/sizeof(uint32_t); ++i) { + query->Base.Result += LE32_TO_CPU(result[i]); + radeon_print(RADEON_STATE, RADEON_TRACE, "result[%d] = %d\n", i, LE32_TO_CPU(result[i])); + } } radeon_bo_unmap(query->bo); @@ -99,10 +108,11 @@ static void radeonDeleteQuery(GLcontext *ctx, struct gl_query_object *q) static void radeonWaitQuery(GLcontext *ctx, struct gl_query_object *q) { + radeonContextPtr radeon = RADEON_CONTEXT(ctx); struct radeon_query_object *query = (struct radeon_query_object *)q; /* If the cmdbuf with packets for this query hasn't been flushed yet, do it now */ - if (!radeonQueryIsFlushed(ctx, q)) + if (radeon_bo_is_referenced_by_cs(query->bo, radeon->cmdbuf.cs)) ctx->Driver.Flush(ctx); radeon_print(RADEON_STATE, RADEON_VERBOSE, "%s: query id %d, bo %p, offset %d\n", __FUNCTION__, q->Id, query->bo, query->curr_offset); @@ -134,8 +144,6 @@ static void radeonBeginQuery(GLcontext *ctx, struct gl_query_object *q) radeon->query.queryobj.dirty = GL_TRUE; radeon->hw.is_dirty = GL_TRUE; - insert_at_tail(&radeon->query.not_flushed_head, query); - } void radeonEmitQueryEnd(GLcontext *ctx) @@ -183,7 +191,7 @@ static void radeonCheckQuery(GLcontext *ctx, struct gl_query_object *q) uint32_t domain; /* Need to perform a flush, as per ARB_occlusion_query spec */ - if (!radeonQueryIsFlushed(ctx, q)) { + if (radeon_bo_is_referenced_by_cs(query->bo, radeon->cmdbuf.cs)) { ctx->Driver.Flush(ctx); } diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.c b/src/mesa/drivers/dri/radeon/radeon_screen.c index 573eb6c9c18..be2d8365ef5 100644 --- a/src/mesa/drivers/dri/radeon/radeon_screen.c +++ b/src/mesa/drivers/dri/radeon/radeon_screen.c @@ -141,12 +141,6 @@ DRI_CONF_BEGIN DRI_CONF_END; static const GLuint __driNConfigOptions = 17; -extern const struct dri_extension blend_extensions[]; -extern const struct dri_extension ARB_vp_extension[]; -extern const struct dri_extension NV_vp_extension[]; -extern const struct dri_extension ATI_fs_extension[]; -extern const struct dri_extension point_extensions[]; - #elif defined(RADEON_R300) || defined(RADEON_R600) #define DRI_CONF_FP_OPTIMIZATION_SPEED 0 @@ -218,13 +212,8 @@ DRI_CONF_BEGIN DRI_CONF_END; static const GLuint __driNConfigOptions = 17; -extern const struct dri_extension gl_20_extension[]; - #endif -extern const struct dri_extension card_extensions[]; -extern const struct dri_extension mm_extensions[]; - static int getSwapInfo( __DRIdrawablePrivate *dPriv, __DRIswapInfo * sInfo ); static int @@ -401,12 +390,14 @@ static int radeon_set_screen_flags(radeonScreenPtr screen, int device_id) screen->device_id = device_id; screen->chip_flags = 0; switch ( device_id ) { + case PCI_CHIP_RN50_515E: + case PCI_CHIP_RN50_5969: + return -1; + case PCI_CHIP_RADEON_LY: case PCI_CHIP_RADEON_LZ: case PCI_CHIP_RADEON_QY: case PCI_CHIP_RADEON_QZ: - case PCI_CHIP_RN50_515E: - case PCI_CHIP_RN50_5969: screen->chip_family = CHIP_FAMILY_RV100; break; @@ -1480,7 +1471,7 @@ radeonCreateBuffer( __DRIscreenPrivate *driScrnPriv, const GLboolean swAccum = mesaVis->accumRedBits > 0; const GLboolean swStencil = mesaVis->stencilBits > 0 && mesaVis->depthBits != 24; - GLenum rgbFormat; + gl_format rgbFormat; struct radeon_framebuffer *rfb; if (isPixmap) @@ -1493,11 +1484,11 @@ radeonCreateBuffer( __DRIscreenPrivate *driScrnPriv, _mesa_initialize_framebuffer(&rfb->base, mesaVis); if (mesaVis->redBits == 5) - rgbFormat = GL_RGB5; + rgbFormat = _mesa_little_endian() ? MESA_FORMAT_RGB565 : MESA_FORMAT_RGB565_REV; else if (mesaVis->alphaBits == 0) - rgbFormat = GL_RGB8; + rgbFormat = _mesa_little_endian() ? MESA_FORMAT_XRGB8888 : MESA_FORMAT_XRGB8888_REV; else - rgbFormat = GL_RGBA8; + rgbFormat = _mesa_little_endian() ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB8888_REV; /* front color renderbuffer */ rfb->color_rb[0] = radeon_create_renderbuffer(rgbFormat, driDrawPriv); @@ -1513,19 +1504,22 @@ radeonCreateBuffer( __DRIscreenPrivate *driScrnPriv, if (mesaVis->depthBits == 24) { if (mesaVis->stencilBits == 8) { - struct radeon_renderbuffer *depthStencilRb = radeon_create_renderbuffer(GL_DEPTH24_STENCIL8_EXT, driDrawPriv); + struct radeon_renderbuffer *depthStencilRb = + radeon_create_renderbuffer(MESA_FORMAT_S8_Z24, driDrawPriv); _mesa_add_renderbuffer(&rfb->base, BUFFER_DEPTH, &depthStencilRb->base); _mesa_add_renderbuffer(&rfb->base, BUFFER_STENCIL, &depthStencilRb->base); depthStencilRb->has_surface = screen->depthHasSurface; } else { /* depth renderbuffer */ - struct radeon_renderbuffer *depth = radeon_create_renderbuffer(GL_DEPTH_COMPONENT24, driDrawPriv); + struct radeon_renderbuffer *depth = + radeon_create_renderbuffer(MESA_FORMAT_X8_Z24, driDrawPriv); _mesa_add_renderbuffer(&rfb->base, BUFFER_DEPTH, &depth->base); depth->has_surface = screen->depthHasSurface; } } else if (mesaVis->depthBits == 16) { - /* just 16-bit depth buffer, no hw stencil */ - struct radeon_renderbuffer *depth = radeon_create_renderbuffer(GL_DEPTH_COMPONENT16, driDrawPriv); + /* just 16-bit depth buffer, no hw stencil */ + struct radeon_renderbuffer *depth = + radeon_create_renderbuffer(MESA_FORMAT_Z16, driDrawPriv); _mesa_add_renderbuffer(&rfb->base, BUFFER_DEPTH, &depth->base); depth->has_surface = screen->depthHasSurface; } @@ -1619,27 +1613,6 @@ radeonInitScreen(__DRIscreenPrivate *psp) return NULL; } - /* Calling driInitExtensions here, with a NULL context pointer, - * does not actually enable the extensions. It just makes sure - * that all the dispatch offsets for all the extensions that - * *might* be enables are known. This is needed because the - * dispatch offsets need to be known when _mesa_context_create - * is called, but we can't enable the extensions until we have a - * context pointer. - * - * Hello chicken. Hello egg. How are you two today? - */ - driInitExtensions( NULL, card_extensions, GL_FALSE ); -#if defined(RADEON_R200) - driInitExtensions( NULL, blend_extensions, GL_FALSE ); - driInitSingleExtension( NULL, ARB_vp_extension ); - driInitSingleExtension( NULL, NV_vp_extension ); - driInitSingleExtension( NULL, ATI_fs_extension ); - driInitExtensions( NULL, point_extensions, GL_FALSE ); -#elif (defined(RADEON_R300) || defined(RADEON_R600)) - driInitSingleExtension( NULL, gl_20_extension ); -#endif - if (!radeonInitDriver(psp)) return NULL; @@ -1672,28 +1645,6 @@ __DRIconfig **radeonInitScreen2(__DRIscreenPrivate *psp) int color; __DRIconfig **configs = NULL; - /* Calling driInitExtensions here, with a NULL context pointer, - * does not actually enable the extensions. It just makes sure - * that all the dispatch offsets for all the extensions that - * *might* be enables are known. This is needed because the - * dispatch offsets need to be known when _mesa_context_create - * is called, but we can't enable the extensions until we have a - * context pointer. - * - * Hello chicken. Hello egg. How are you two today? - */ - driInitExtensions( NULL, card_extensions, GL_FALSE ); - driInitExtensions( NULL, mm_extensions, GL_FALSE ); -#if defined(RADEON_R200) - driInitExtensions( NULL, blend_extensions, GL_FALSE ); - driInitSingleExtension( NULL, ARB_vp_extension ); - driInitSingleExtension( NULL, NV_vp_extension ); - driInitSingleExtension( NULL, ATI_fs_extension ); - driInitExtensions( NULL, point_extensions, GL_FALSE ); -#elif (defined(RADEON_R300) || defined(RADEON_R600)) - driInitSingleExtension( NULL, gl_20_extension ); -#endif - if (!radeonInitDriver(psp)) { return NULL; } diff --git a/src/mesa/drivers/dri/radeon/radeon_span.c b/src/mesa/drivers/dri/radeon/radeon_span.c index 0c49c3713af..37904dc8dc9 100644 --- a/src/mesa/drivers/dri/radeon/radeon_span.c +++ b/src/mesa/drivers/dri/radeon/radeon_span.c @@ -41,6 +41,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ #include "main/glheader.h" +#include "main/texformat.h" #include "swrast/swrast.h" #include "radeon_common.h" @@ -334,22 +335,6 @@ static GLubyte *radeon_ptr_2byte_8x2(const struct radeon_renderbuffer * rrb, #endif -#ifndef RADEON_R300 -#ifndef RADEON_R600 -static uint32_t -z24s8_to_s8z24(uint32_t val) -{ - return (val << 24) | (val >> 8); -} - -static uint32_t -s8z24_to_z24s8(uint32_t val) -{ - return (val >> 24) | (val << 8); -} -#endif -#endif - /* * Note that all information needed to access pixels in a renderbuffer * should be obtained through the gl_renderbuffer parameter, not per-context @@ -416,6 +401,18 @@ s8z24_to_z24s8(uint32_t val) #endif #include "spantmp2.h" +#define SPANTMP_PIXEL_FMT GL_RGB +#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_5_6_5_REV + +#define TAG(x) radeon##x##_RGB565_REV +#define TAG2(x,y) radeon##x##_RGB565_REV##y +#if defined(RADEON_R600) +#define GET_PTR(X,Y) r600_ptr_color(rrb, (X) + x_off, (Y) + y_off) +#else +#define GET_PTR(X,Y) radeon_ptr_2byte_8x2(rrb, (X) + x_off, (Y) + y_off) +#endif +#include "spantmp2.h" + /* 16 bit, ARGB1555 color spanline and pixel functions */ #define SPANTMP_PIXEL_FMT GL_BGRA @@ -430,6 +427,18 @@ s8z24_to_z24s8(uint32_t val) #endif #include "spantmp2.h" +#define SPANTMP_PIXEL_FMT GL_BGRA +#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_1_5_5_5 + +#define TAG(x) radeon##x##_ARGB1555_REV +#define TAG2(x,y) radeon##x##_ARGB1555_REV##y +#if defined(RADEON_R600) +#define GET_PTR(X,Y) r600_ptr_color(rrb, (X) + x_off, (Y) + y_off) +#else +#define GET_PTR(X,Y) radeon_ptr_2byte_8x2(rrb, (X) + x_off, (Y) + y_off) +#endif +#include "spantmp2.h" + /* 16 bit, RGBA4 color spanline and pixel functions */ #define SPANTMP_PIXEL_FMT GL_BGRA @@ -444,6 +453,18 @@ s8z24_to_z24s8(uint32_t val) #endif #include "spantmp2.h" +#define SPANTMP_PIXEL_FMT GL_BGRA +#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_4_4_4_4 + +#define TAG(x) radeon##x##_ARGB4444_REV +#define TAG2(x,y) radeon##x##_ARGB4444_REV##y +#if defined(RADEON_R600) +#define GET_PTR(X,Y) r600_ptr_color(rrb, (X) + x_off, (Y) + y_off) +#else +#define GET_PTR(X,Y) radeon_ptr_2byte_8x2(rrb, (X) + x_off, (Y) + y_off) +#endif +#include "spantmp2.h" + /* 32 bit, xRGB8888 color spanline and pixel functions */ #define SPANTMP_PIXEL_FMT GL_BGRA @@ -488,6 +509,42 @@ s8z24_to_z24s8(uint32_t val) #endif #include "spantmp2.h" +/* 32 bit, BGRx8888 color spanline and pixel functions + */ +#define SPANTMP_PIXEL_FMT GL_BGRA +#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8 + +#define TAG(x) radeon##x##_BGRx8888 +#define TAG2(x,y) radeon##x##_BGRx8888##y +#if defined(RADEON_R600) +#define GET_VALUE(_x, _y) ((*(GLuint*)(r600_ptr_color(rrb, _x + x_off, _y + y_off)) | 0x000000ff)) +#define PUT_VALUE(_x, _y, d) { \ + GLuint *_ptr = (GLuint*)r600_ptr_color( rrb, _x + x_off, _y + y_off ); \ + *_ptr = d; \ +} while (0) +#else +#define GET_VALUE(_x, _y) ((*(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off, _y + y_off)) | 0x000000ff)) +#define PUT_VALUE(_x, _y, d) { \ + GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off ); \ + *_ptr = d; \ +} while (0) +#endif +#include "spantmp2.h" + +/* 32 bit, BGRA8888 color spanline and pixel functions + */ +#define SPANTMP_PIXEL_FMT GL_BGRA +#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8 + +#define TAG(x) radeon##x##_BGRA8888 +#define TAG2(x,y) radeon##x##_BGRA8888##y +#if defined(RADEON_R600) +#define GET_PTR(X,Y) r600_ptr_color(rrb, (X) + x_off, (Y) + y_off) +#else +#define GET_PTR(X,Y) radeon_ptr_4byte(rrb, (X) + x_off, (Y) + y_off) +#endif +#include "spantmp2.h" + /* ================================================================ * Depth buffer */ @@ -542,10 +599,10 @@ s8z24_to_z24s8(uint32_t val) #define WRITE_DEPTH( _x, _y, d ) \ do { \ GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off ); \ - GLuint tmp = *_ptr; \ + GLuint tmp = LE32_TO_CPU(*_ptr); \ tmp &= 0x000000ff; \ tmp |= ((d << 8) & 0xffffff00); \ - *_ptr = tmp; \ + *_ptr = CPU_TO_LE32(tmp); \ } while (0) #elif defined(RADEON_R600) #define WRITE_DEPTH( _x, _y, d ) \ @@ -560,26 +617,26 @@ do { \ #define WRITE_DEPTH( _x, _y, d ) \ do { \ GLuint *_ptr = (GLuint*)r200_depth_4byte( rrb, _x + x_off, _y + y_off ); \ - GLuint tmp = *_ptr; \ + GLuint tmp = LE32_TO_CPU(*_ptr); \ tmp &= 0xff000000; \ tmp |= ((d) & 0x00ffffff); \ - *_ptr = tmp; \ + *_ptr = CPU_TO_LE32(tmp); \ } while (0) #else #define WRITE_DEPTH( _x, _y, d ) \ do { \ GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off ); \ - GLuint tmp = *_ptr; \ + GLuint tmp = LE32_TO_CPU(*_ptr); \ tmp &= 0xff000000; \ tmp |= ((d) & 0x00ffffff); \ - *_ptr = tmp; \ + *_ptr = CPU_TO_LE32(tmp); \ } while (0) #endif #if defined(RADEON_R300) #define READ_DEPTH( d, _x, _y ) \ do { \ - d = (*(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off, _y + y_off)) & 0xffffff00) >> 8; \ + d = (LE32_TO_CPU(*(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off, _y + y_off))) & 0xffffff00) >> 8; \ }while(0) #elif defined(RADEON_R600) #define READ_DEPTH( d, _x, _y ) \ @@ -589,11 +646,11 @@ do { \ #elif defined(RADEON_R200) #define READ_DEPTH( d, _x, _y ) \ do { \ - d = *(GLuint*)(r200_depth_4byte(rrb, _x + x_off, _y + y_off)) & 0x00ffffff; \ + d = LE32_TO_CPU(*(GLuint*)(r200_depth_4byte(rrb, _x + x_off, _y + y_off))) & 0x00ffffff; \ }while(0) #else #define READ_DEPTH( d, _x, _y ) \ - d = *(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off, _y + y_off)) & 0x00ffffff; + d = LE32_TO_CPU(*(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off, _y + y_off))) & 0x00ffffff; #endif #define TAG(x) radeon##x##_z24 @@ -611,7 +668,7 @@ do { \ #define WRITE_DEPTH( _x, _y, d ) \ do { \ GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off ); \ - *_ptr = d; \ + *_ptr = CPU_TO_LE32((((d) & 0xff000000) >> 24) | (((d) & 0x00ffffff) << 8)); \ } while (0) #elif defined(RADEON_R600) #define WRITE_DEPTH( _x, _y, d ) \ @@ -619,53 +676,52 @@ do { \ GLuint *_ptr = (GLuint*)r600_ptr_depth( rrb, _x + x_off, _y + y_off ); \ GLuint tmp = *_ptr; \ tmp &= 0xff000000; \ - tmp |= (((d) >> 8) & 0x00ffffff); \ + tmp |= ((d) & 0x00ffffff); \ *_ptr = tmp; \ _ptr = (GLuint*)r600_ptr_stencil(rrb, _x + x_off, _y + y_off); \ tmp = *_ptr; \ tmp &= 0xffffff00; \ - tmp |= (d) & 0xff; \ + tmp |= ((d) >> 24) & 0xff; \ *_ptr = tmp; \ } while (0) #elif defined(RADEON_R200) #define WRITE_DEPTH( _x, _y, d ) \ do { \ GLuint *_ptr = (GLuint*)r200_depth_4byte( rrb, _x + x_off, _y + y_off ); \ - GLuint tmp = z24s8_to_s8z24(d); \ - *_ptr = tmp; \ + *_ptr = CPU_TO_LE32(d); \ } while (0) #else #define WRITE_DEPTH( _x, _y, d ) \ do { \ GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off ); \ - GLuint tmp = z24s8_to_s8z24(d); \ - *_ptr = tmp; \ + *_ptr = CPU_TO_LE32(d); \ } while (0) #endif #if defined(RADEON_R300) #define READ_DEPTH( d, _x, _y ) \ do { \ - d = (*(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off, _y + y_off))); \ + GLuint tmp = (*(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off, _y + y_off))); \ + d = LE32_TO_CPU(((tmp & 0x000000ff) << 24) | ((tmp & 0xffffff00) >> 8)); \ }while(0) #elif defined(RADEON_R600) #define READ_DEPTH( d, _x, _y ) \ do { \ - d = ((*(GLuint*)(r600_ptr_depth(rrb, _x + x_off, _y + y_off))) << 8) & 0xffffff00; \ - d |= (*(GLuint*)(r600_ptr_stencil(rrb, _x + x_off, _y + y_off))) & 0x000000ff; \ + d = (*(GLuint*)(r600_ptr_depth(rrb, _x + x_off, _y + y_off))) & 0x00ffffff; \ + d |= ((*(GLuint*)(r600_ptr_stencil(rrb, _x + x_off, _y + y_off))) << 24) & 0xff000000; \ }while(0) #elif defined(RADEON_R200) #define READ_DEPTH( d, _x, _y ) \ do { \ - d = s8z24_to_z24s8(*(GLuint*)(r200_depth_4byte(rrb, _x + x_off, _y + y_off))); \ + d = LE32_TO_CPU(*(GLuint*)(r200_depth_4byte(rrb, _x + x_off, _y + y_off))); \ }while(0) #else #define READ_DEPTH( d, _x, _y ) do { \ - d = s8z24_to_z24s8(*(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off, _y + y_off ))); \ + d = LE32_TO_CPU(*(GLuint*)(radeon_ptr_4byte(rrb, _x + x_off, _y + y_off))); \ } while (0) #endif -#define TAG(x) radeon##x##_z24_s8 +#define TAG(x) radeon##x##_s8_z24 #include "depthtmp.h" /* ================================================================ @@ -678,10 +734,10 @@ do { \ #define WRITE_STENCIL( _x, _y, d ) \ do { \ GLuint *_ptr = (GLuint*)radeon_ptr_4byte(rrb, _x + x_off, _y + y_off); \ - GLuint tmp = *_ptr; \ + GLuint tmp = LE32_TO_CPU(*_ptr); \ tmp &= 0xffffff00; \ tmp |= (d) & 0xff; \ - *_ptr = tmp; \ + *_ptr = CPU_TO_LE32(tmp); \ } while (0) #elif defined(RADEON_R600) #define WRITE_STENCIL( _x, _y, d ) \ @@ -696,19 +752,19 @@ do { \ #define WRITE_STENCIL( _x, _y, d ) \ do { \ GLuint *_ptr = (GLuint*)r200_depth_4byte(rrb, _x + x_off, _y + y_off); \ - GLuint tmp = *_ptr; \ + GLuint tmp = LE32_TO_CPU(*_ptr); \ tmp &= 0x00ffffff; \ tmp |= (((d) & 0xff) << 24); \ - *_ptr = tmp; \ + *_ptr = CPU_TO_LE32(tmp); \ } while (0) #else #define WRITE_STENCIL( _x, _y, d ) \ do { \ GLuint *_ptr = (GLuint*)radeon_ptr_4byte(rrb, _x + x_off, _y + y_off); \ - GLuint tmp = *_ptr; \ + GLuint tmp = LE32_TO_CPU(*_ptr); \ tmp &= 0x00ffffff; \ tmp |= (((d) & 0xff) << 24); \ - *_ptr = tmp; \ + *_ptr = CPU_TO_LE32(tmp); \ } while (0) #endif @@ -716,7 +772,7 @@ do { \ #define READ_STENCIL( d, _x, _y ) \ do { \ GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off ); \ - GLuint tmp = *_ptr; \ + GLuint tmp = LE32_TO_CPU(*_ptr); \ d = tmp & 0x000000ff; \ } while (0) #elif defined(RADEON_R600) @@ -730,19 +786,19 @@ do { \ #define READ_STENCIL( d, _x, _y ) \ do { \ GLuint *_ptr = (GLuint*)r200_depth_4byte( rrb, _x + x_off, _y + y_off ); \ - GLuint tmp = *_ptr; \ + GLuint tmp = LE32_TO_CPU(*_ptr); \ d = (tmp & 0xff000000) >> 24; \ } while (0) #else #define READ_STENCIL( d, _x, _y ) \ do { \ GLuint *_ptr = (GLuint*)radeon_ptr_4byte( rrb, _x + x_off, _y + y_off ); \ - GLuint tmp = *_ptr; \ + GLuint tmp = LE32_TO_CPU(*_ptr); \ d = (tmp & 0xff000000) >> 24; \ } while (0) #endif -#define TAG(x) radeon##x##_z24_s8 +#define TAG(x) radeon##x##_s8_z24 #include "stenciltmp.h" @@ -864,25 +920,35 @@ void radeonInitSpanFuncs(GLcontext * ctx) */ static void radeonSetSpanFunctions(struct radeon_renderbuffer *rrb) { - if (rrb->base._ActualFormat == GL_RGB5) { + if (rrb->base.Format == MESA_FORMAT_RGB565) { radeonInitPointers_RGB565(&rrb->base); - } else if (rrb->base._ActualFormat == GL_RGB8) { + } else if (rrb->base.Format == MESA_FORMAT_RGB565_REV) { + radeonInitPointers_RGB565_REV(&rrb->base); + } else if (rrb->base.Format == MESA_FORMAT_XRGB8888) { radeonInitPointers_xRGB8888(&rrb->base); - } else if (rrb->base._ActualFormat == GL_RGBA8) { + } else if (rrb->base.Format == MESA_FORMAT_XRGB8888_REV) { + radeonInitPointers_BGRx8888(&rrb->base); + } else if (rrb->base.Format == MESA_FORMAT_ARGB8888) { radeonInitPointers_ARGB8888(&rrb->base); - } else if (rrb->base._ActualFormat == GL_RGBA4) { + } else if (rrb->base.Format == MESA_FORMAT_ARGB8888_REV) { + radeonInitPointers_BGRA8888(&rrb->base); + } else if (rrb->base.Format == MESA_FORMAT_ARGB4444) { radeonInitPointers_ARGB4444(&rrb->base); - } else if (rrb->base._ActualFormat == GL_RGB5_A1) { + } else if (rrb->base.Format == MESA_FORMAT_ARGB4444_REV) { + radeonInitPointers_ARGB4444_REV(&rrb->base); + } else if (rrb->base.Format == MESA_FORMAT_ARGB1555) { radeonInitPointers_ARGB1555(&rrb->base); - } else if (rrb->base._ActualFormat == GL_DEPTH_COMPONENT16) { + } else if (rrb->base.Format == MESA_FORMAT_ARGB1555_REV) { + radeonInitPointers_ARGB1555_REV(&rrb->base); + } else if (rrb->base.Format == MESA_FORMAT_Z16) { radeonInitDepthPointers_z16(&rrb->base); - } else if (rrb->base._ActualFormat == GL_DEPTH_COMPONENT24) { + } else if (rrb->base.Format == MESA_FORMAT_X8_Z24) { radeonInitDepthPointers_z24(&rrb->base); - } else if (rrb->base._ActualFormat == GL_DEPTH24_STENCIL8_EXT) { - radeonInitDepthPointers_z24_s8(&rrb->base); - } else if (rrb->base._ActualFormat == GL_STENCIL_INDEX8_EXT) { - radeonInitStencilPointers_z24_s8(&rrb->base); + } else if (rrb->base.Format == MESA_FORMAT_S8_Z24) { + radeonInitDepthPointers_s8_z24(&rrb->base); + } else if (rrb->base.Format == MESA_FORMAT_S8) { + radeonInitStencilPointers_s8_z24(&rrb->base); } else { - fprintf(stderr, "radeonSetSpanFunctions: bad actual format: 0x%04X\n", rrb->base._ActualFormat); + fprintf(stderr, "radeonSetSpanFunctions: bad format: 0x%04X\n", rrb->base.Format); } } diff --git a/src/mesa/drivers/dri/radeon/radeon_state.c b/src/mesa/drivers/dri/radeon/radeon_state.c index 4d0d35ee0cd..f6c733ab209 100644 --- a/src/mesa/drivers/dri/radeon/radeon_state.c +++ b/src/mesa/drivers/dri/radeon/radeon_state.c @@ -550,6 +550,31 @@ static void radeonPolygonOffset( GLcontext *ctx, rmesa->hw.zbs.cmd[ZBS_SE_ZBIAS_CONSTANT] = constant.ui32; } +static void radeonPolygonStipplePreKMS( GLcontext *ctx, const GLubyte *mask ) +{ + r100ContextPtr rmesa = R100_CONTEXT(ctx); + GLuint i; + drm_radeon_stipple_t stipple; + + /* Must flip pattern upside down. + */ + for ( i = 0 ; i < 32 ; i++ ) { + rmesa->state.stipple.mask[31 - i] = ((GLuint *) mask)[i]; + } + + /* TODO: push this into cmd mechanism + */ + radeon_firevertices(&rmesa->radeon); + LOCK_HARDWARE( &rmesa->radeon ); + + /* FIXME: Use window x,y offsets into stipple RAM. + */ + stipple.mask = rmesa->state.stipple.mask; + drmCommandWrite( rmesa->radeon.dri.fd, DRM_RADEON_STIPPLE, + &stipple, sizeof(drm_radeon_stipple_t) ); + UNLOCK_HARDWARE( &rmesa->radeon ); +} + static void radeonPolygonMode( GLcontext *ctx, GLenum face, GLenum mode ) { r100ContextPtr rmesa = R100_CONTEXT(ctx); diff --git a/src/mesa/drivers/dri/radeon/radeon_state_init.c b/src/mesa/drivers/dri/radeon/radeon_state_init.c index f3ad0dd17af..dd828882543 100644 --- a/src/mesa/drivers/dri/radeon/radeon_state_init.c +++ b/src/mesa/drivers/dri/radeon/radeon_state_init.c @@ -440,16 +440,18 @@ static void ctx_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom) atom->cmd[CTX_RB3D_CNTL] &= ~(0xf << 10); if (rrb->cpp == 4) atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB8888; - else switch (rrb->base._ActualFormat) { - case GL_RGB5: + else switch (rrb->base.Format) { + case MESA_FORMAT_RGB565: atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_RGB565; break; - case GL_RGBA4: + case MESA_FORMAT_ARGB4444: atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB4444; break; - case GL_RGB5_A1: + case MESA_FORMAT_ARGB1555: atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB1555; break; + default: + _mesa_problem(ctx, "unexpected format in ctx_emit_cs()"); } cbpitch = (rrb->pitch / rrb->cpp); @@ -643,11 +645,11 @@ static void tex_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom) OUT_BATCH(CP_PACKET0(RADEON_PP_TXOFFSET_0 + (24 * i), 0)); if (t->mt && !t->image_override) { if ((ctx->Texture.Unit[i]._ReallyEnabled & TEXTURE_CUBE_BIT)) { - lvl = &t->mt->levels[0]; + lvl = &t->mt->levels[t->minLod]; OUT_BATCH_RELOC(lvl->faces[5].offset, t->mt->bo, lvl->faces[5].offset, RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0); } else { - OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, 0, + OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, get_base_teximage_offset(t), RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0); } } else { diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c index e61f59eaeaf..8bf1bfbc575 100644 --- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c @@ -179,7 +179,7 @@ static void radeonSetVertexFormat( GLcontext *ctx ) for (i = 0; i < ctx->Const.MaxTextureUnits; i++) { if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX(i) )) { - GLuint sz = VB->TexCoordPtr[i]->size; + GLuint sz = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size; switch (sz) { case 1: @@ -309,7 +309,7 @@ void r100_swtcl_flush(GLcontext *ctx, uint32_t current_offset) radeonEmitState(&rmesa->radeon); radeonEmitVertexAOS( rmesa, rmesa->radeon.swtcl.vertex_size, - first_elem(&rmesa->radeon.dma.reserved)->bo, + rmesa->radeon.swtcl.bo, current_offset); diff --git a/src/mesa/drivers/dri/radeon/radeon_tex.c b/src/mesa/drivers/dri/radeon/radeon_tex.c index 99865fff27b..749ab75f201 100644 --- a/src/mesa/drivers/dri/radeon/radeon_tex.c +++ b/src/mesa/drivers/dri/radeon/radeon_tex.c @@ -38,7 +38,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/enums.h" #include "main/image.h" #include "main/simple_list.h" -#include "main/texformat.h" #include "main/texstore.h" #include "main/teximage.h" #include "main/texobj.h" @@ -349,17 +348,7 @@ static void radeonTexParameter( GLcontext *ctx, GLenum target, case GL_TEXTURE_MAX_LEVEL: case GL_TEXTURE_MIN_LOD: case GL_TEXTURE_MAX_LOD: - - /* This isn't the most efficient solution but there doesn't appear to - * be a nice alternative. Since there's no LOD clamping, - * we just have to rely on loading the right subset of mipmap levels - * to simulate a clamped LOD. - */ - if (t->mt) { - radeon_miptree_unreference(t->mt); - t->mt = 0; - t->validated = GL_FALSE; - } + t->validated = GL_FALSE; break; default: @@ -389,10 +378,8 @@ static void radeonDeleteTexture( GLcontext *ctx, } } - if (t->mt) { - radeon_miptree_unreference(t->mt); - t->mt = 0; - } + radeon_miptree_unreference(&t->mt); + /* Free mipmap images and the texture object itself */ _mesa_delete_texture_object(ctx, texObj); } diff --git a/src/mesa/drivers/dri/radeon/radeon_texstate.c b/src/mesa/drivers/dri/radeon/radeon_texstate.c index ae41b90efe5..84ddcfd4fd3 100644 --- a/src/mesa/drivers/dri/radeon/radeon_texstate.c +++ b/src/mesa/drivers/dri/radeon/radeon_texstate.c @@ -38,8 +38,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/colormac.h" #include "main/context.h" #include "main/macros.h" -#include "main/texformat.h" #include "main/teximage.h" +#include "main/texstate.h" #include "main/texobj.h" #include "main/enums.h" @@ -81,8 +81,10 @@ struct tx_table { GLuint format, filter; }; +/* XXX verify this table against MESA_FORMAT_x values */ static const struct tx_table tx_table[] = { + _INVALID(NONE), /* MESA_FORMAT_NONE */ _ALPHA(RGBA8888), _ALPHA_REV(RGBA8888), _ALPHA(ARGB8888), @@ -660,7 +662,7 @@ void radeonSetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_ rmesa = pDRICtx->driverPrivate; rfb = dPriv->driverPrivate; - texUnit = &radeon->glCtx->Texture.Unit[radeon->glCtx->Texture.CurrentUnit]; + texUnit = _mesa_get_current_tex_unit(radeon->glCtx); texObj = _mesa_select_tex_object(radeon->glCtx, texUnit, target); texImage = _mesa_get_tex_image(radeon->glCtx, texObj, target, 0); @@ -670,24 +672,13 @@ void radeonSetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_ return; } - radeon_update_renderbuffers(pDRICtx, dPriv); - /* back & depth buffer are useless free them right away */ - rb = (void*)rfb->base.Attachment[BUFFER_DEPTH].Renderbuffer; - if (rb && rb->bo) { - radeon_bo_unref(rb->bo); - rb->bo = NULL; - } - rb = (void*)rfb->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer; - if (rb && rb->bo) { - radeon_bo_unref(rb->bo); - rb->bo = NULL; - } + radeon_update_renderbuffers(pDRICtx, dPriv, GL_TRUE); rb = rfb->color_rb[0]; if (rb->bo == NULL) { /* Failed to BO for the buffer */ return; } - + _mesa_lock_texture(radeon->glCtx, texObj); if (t->bo) { radeon_bo_unref(t->bo); @@ -697,20 +688,14 @@ void radeonSetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_ radeon_bo_unref(rImage->bo); rImage->bo = NULL; } - if (t->mt) { - radeon_miptree_unreference(t->mt); - t->mt = NULL; - } - if (rImage->mt) { - radeon_miptree_unreference(rImage->mt); - rImage->mt = NULL; - } + + radeon_miptree_unreference(&t->mt); + radeon_miptree_unreference(&rImage->mt); + _mesa_init_teximage_fields(radeon->glCtx, target, texImage, rb->base.Width, rb->base.Height, 1, 0, rb->cpp); texImage->RowStride = rb->pitch / rb->cpp; - texImage->TexFormat = radeonChooseTextureFormat(radeon->glCtx, - internalFormat, - type, format, 0); + rImage->bo = rb->bo; radeon_bo_ref(rImage->bo); t->bo = rb->bo; @@ -718,8 +703,6 @@ void radeonSetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_ t->tile_bits = 0; t->image_override = GL_TRUE; t->override_offset = 0; - t->pp_txpitch &= (1 << 13) -1; - pitch_val = rb->pitch; switch (rb->cpp) { case 4: if (glx_texture_format == GLX_TEXTURE_FORMAT_RGB_EXT) @@ -738,12 +721,17 @@ void radeonSetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_ t->pp_txfilter |= tx_table[MESA_FORMAT_RGB565].filter; break; } - t->pp_txsize = ((rb->base.Width - 1) << RADEON_TEX_USIZE_SHIFT) - | ((rb->base.Height - 1) << RADEON_TEX_VSIZE_SHIFT); - t->pp_txformat |= RADEON_TXFORMAT_NON_POWER2; - t->pp_txpitch = pitch_val; - t->pp_txpitch -= 32; + t->pp_txpitch &= (1 << 13) -1; + pitch_val = rb->pitch; + + t->pp_txsize = ((rb->base.Width - 1) << RADEON_TEX_USIZE_SHIFT) + | ((rb->base.Height - 1) << RADEON_TEX_VSIZE_SHIFT); + if (target == GL_TEXTURE_RECTANGLE_NV) { + t->pp_txformat |= RADEON_TXFORMAT_NON_POWER2; + t->pp_txpitch = pitch_val; + t->pp_txpitch -= 32; + } t->validated = GL_TRUE; _mesa_unlock_texture(radeon->glCtx, texObj); return; @@ -1021,7 +1009,7 @@ static GLboolean setup_hardware_state(r100ContextPtr rmesa, radeonTexObj *t, int return GL_TRUE; } - firstImage = t->base.Image[0][t->mt->firstLevel]; + firstImage = t->base.Image[0][t->minLod]; if (firstImage->Border > 0) { fprintf(stderr, "%s: border\n", __FUNCTION__); @@ -1031,27 +1019,27 @@ static GLboolean setup_hardware_state(r100ContextPtr rmesa, radeonTexObj *t, int log2Width = firstImage->WidthLog2; log2Height = firstImage->HeightLog2; log2Depth = firstImage->DepthLog2; - texelBytes = firstImage->TexFormat->TexelBytes; + texelBytes = _mesa_get_format_bytes(firstImage->TexFormat); if (!t->image_override) { - if (VALID_FORMAT(firstImage->TexFormat->MesaFormat)) { + if (VALID_FORMAT(firstImage->TexFormat)) { const struct tx_table *table = tx_table; t->pp_txformat &= ~(RADEON_TXFORMAT_FORMAT_MASK | RADEON_TXFORMAT_ALPHA_IN_MAP); t->pp_txfilter &= ~RADEON_YUV_TO_RGB; - t->pp_txformat |= table[ firstImage->TexFormat->MesaFormat ].format; - t->pp_txfilter |= table[ firstImage->TexFormat->MesaFormat ].filter; + t->pp_txformat |= table[ firstImage->TexFormat ].format; + t->pp_txfilter |= table[ firstImage->TexFormat ].filter; } else { _mesa_problem(NULL, "unexpected texture format in %s", __FUNCTION__); return GL_FALSE; } } - + t->pp_txfilter &= ~RADEON_MAX_MIP_LEVEL_MASK; - t->pp_txfilter |= (t->mt->lastLevel - t->mt->firstLevel) << RADEON_MAX_MIP_LEVEL_SHIFT; + t->pp_txfilter |= (t->maxLod - t->minLod) << RADEON_MAX_MIP_LEVEL_SHIFT; t->pp_txformat &= ~(RADEON_TXFORMAT_WIDTH_MASK | RADEON_TXFORMAT_HEIGHT_MASK | @@ -1060,9 +1048,9 @@ static GLboolean setup_hardware_state(r100ContextPtr rmesa, radeonTexObj *t, int RADEON_TXFORMAT_F5_HEIGHT_MASK); t->pp_txformat |= ((log2Width << RADEON_TXFORMAT_WIDTH_SHIFT) | (log2Height << RADEON_TXFORMAT_HEIGHT_SHIFT)); - + t->tile_bits = 0; - + if (t->base.Target == GL_TEXTURE_CUBE_MAP) { ASSERT(log2Width == log2Height); t->pp_txformat |= ((log2Width << RADEON_TXFORMAT_F5_WIDTH_SHIFT) | @@ -1083,7 +1071,7 @@ static GLboolean setup_hardware_state(r100ContextPtr rmesa, radeonTexObj *t, int | ((firstImage->Height - 1) << RADEON_TEX_VSIZE_SHIFT)); if ( !t->image_override ) { - if (firstImage->IsCompressed) + if (_mesa_is_format_compressed(firstImage->TexFormat)) t->pp_txpitch = (firstImage->Width + 63) & ~(63); else t->pp_txpitch = ((firstImage->Width * texelBytes) + 63) & ~(63); diff --git a/src/mesa/drivers/dri/radeon/radeon_texture.c b/src/mesa/drivers/dri/radeon/radeon_texture.c index 049284ef8c5..28690325d12 100644 --- a/src/mesa/drivers/dri/radeon/radeon_texture.c +++ b/src/mesa/drivers/dri/radeon/radeon_texture.c @@ -1,4 +1,5 @@ /* + * Copyright (C) 2009 Maciej Cencora. * Copyright (C) 2008 Nicolai Haehnle. * Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved. * @@ -34,7 +35,6 @@ #include "main/convolve.h" #include "main/mipmap.h" #include "main/texcompress.h" -#include "main/texformat.h" #include "main/texstore.h" #include "main/teximage.h" #include "main/texobj.h" @@ -47,7 +47,7 @@ #include "radeon_mipmap_tree.h" -static void copy_rows(void* dst, GLuint dststride, const void* src, GLuint srcstride, +void copy_rows(void* dst, GLuint dststride, const void* src, GLuint srcstride, GLuint numrows, GLuint rowsize) { assert(rowsize <= dststride); @@ -82,8 +82,7 @@ void radeonFreeTexImageData(GLcontext *ctx, struct gl_texture_image *timage) radeon_texture_image* image = get_radeon_texture_image(timage); if (image->mt) { - radeon_miptree_unreference(image->mt); - image->mt = 0; + radeon_miptree_unreference(&image->mt); assert(!image->base.Data); } else { _mesa_free_texture_image_data(ctx, timage); @@ -101,10 +100,15 @@ void radeonFreeTexImageData(GLcontext *ctx, struct gl_texture_image *timage) /* Set Data pointer and additional data for mapped texture image */ static void teximage_set_map_data(radeon_texture_image *image) { - radeon_mipmap_level *lvl = &image->mt->levels[image->mtlevel]; + radeon_mipmap_level *lvl; + + if (!image->mt) + return; + + lvl = &image->mt->levels[image->mtlevel]; image->base.Data = image->mt->bo->ptr + lvl->faces[image->mtface].offset; - image->base.RowStride = lvl->rowstride / image->mt->bpp; + image->base.RowStride = lvl->rowstride / _mesa_get_format_bytes(image->base.TexFormat); } @@ -139,7 +143,6 @@ static void map_override(GLcontext *ctx, radeonTexObj *t) radeon_bo_map(t->bo, GL_FALSE); img->base.Data = t->bo->ptr; - _mesa_set_fetch_functions(&img->base, 2); } static void unmap_override(GLcontext *ctx, radeonTexObj *t) @@ -171,7 +174,7 @@ void radeonMapTexture(GLcontext *ctx, struct gl_texture_object *texObj) radeon_bo_map(t->mt->bo, GL_FALSE); for(face = 0; face < t->mt->faces; ++face) { - for(level = t->mt->firstLevel; level <= t->mt->lastLevel; ++level) + for(level = t->minLod; level <= t->maxLod; ++level) teximage_set_map_data(get_radeon_texture_image(texObj->Image[face][level])); } } @@ -188,7 +191,7 @@ void radeonUnmapTexture(GLcontext *ctx, struct gl_texture_object *texObj) return; for(face = 0; face < t->mt->faces; ++face) { - for(level = t->mt->firstLevel; level <= t->mt->lastLevel; ++level) + for(level = t->minLod; level <= t->maxLod; ++level) texObj->Image[face][level]->Data = 0; } radeon_bo_unmap(t->mt->bo); @@ -237,8 +240,7 @@ static void radeon_generate_mipmap(GLcontext *ctx, GLenum target, image->mtlevel = i; image->mtface = face; - radeon_miptree_unreference(image->mt); - image->mt = NULL; + radeon_miptree_unreference(&image->mt); } } @@ -256,9 +258,9 @@ void radeonGenerateMipmap(GLcontext* ctx, GLenum target, struct gl_texture_objec /* try to find a format which will only need a memcopy */ -static const struct gl_texture_format *radeonChoose8888TexFormat(radeonContextPtr rmesa, - GLenum srcFormat, - GLenum srcType, GLboolean fbo) +static gl_format radeonChoose8888TexFormat(radeonContextPtr rmesa, + GLenum srcFormat, + GLenum srcType, GLboolean fbo) { const GLuint ui = 1; const GLubyte littleEndian = *((const GLubyte *)&ui); @@ -271,37 +273,37 @@ static const struct gl_texture_format *radeonChoose8888TexFormat(radeonContextPt (srcFormat == GL_RGBA && srcType == GL_UNSIGNED_BYTE && !littleEndian) || (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_INT_8_8_8_8_REV) || (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_BYTE && littleEndian)) { - return &_mesa_texformat_rgba8888; + return MESA_FORMAT_RGBA8888; } else if ((srcFormat == GL_RGBA && srcType == GL_UNSIGNED_INT_8_8_8_8_REV) || (srcFormat == GL_RGBA && srcType == GL_UNSIGNED_BYTE && littleEndian) || (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_INT_8_8_8_8) || (srcFormat == GL_ABGR_EXT && srcType == GL_UNSIGNED_BYTE && !littleEndian)) { - return &_mesa_texformat_rgba8888_rev; + return MESA_FORMAT_RGBA8888_REV; } else if (IS_R200_CLASS(rmesa->radeonScreen)) { return _dri_texformat_argb8888; } else if (srcFormat == GL_BGRA && ((srcType == GL_UNSIGNED_BYTE && !littleEndian) || srcType == GL_UNSIGNED_INT_8_8_8_8)) { - return &_mesa_texformat_argb8888_rev; + return MESA_FORMAT_ARGB8888_REV; } else if (srcFormat == GL_BGRA && ((srcType == GL_UNSIGNED_BYTE && littleEndian) || srcType == GL_UNSIGNED_INT_8_8_8_8_REV)) { - return &_mesa_texformat_argb8888; + return MESA_FORMAT_ARGB8888; } else return _dri_texformat_argb8888; } -const struct gl_texture_format *radeonChooseTextureFormat_mesa(GLcontext * ctx, - GLint internalFormat, - GLenum format, - GLenum type) +gl_format radeonChooseTextureFormat_mesa(GLcontext * ctx, + GLint internalFormat, + GLenum format, + GLenum type) { return radeonChooseTextureFormat(ctx, internalFormat, format, type, 0); } -const struct gl_texture_format *radeonChooseTextureFormat(GLcontext * ctx, - GLint internalFormat, - GLenum format, - GLenum type, GLboolean fbo) +gl_format radeonChooseTextureFormat(GLcontext * ctx, + GLint internalFormat, + GLenum format, + GLenum type, GLboolean fbo) { radeonContextPtr rmesa = RADEON_CONTEXT(ctx); const GLboolean do32bpt = @@ -425,50 +427,50 @@ const struct gl_texture_format *radeonChooseTextureFormat(GLcontext * ctx, case GL_YCBCR_MESA: if (type == GL_UNSIGNED_SHORT_8_8_APPLE || type == GL_UNSIGNED_BYTE) - return &_mesa_texformat_ycbcr; + return MESA_FORMAT_YCBCR; else - return &_mesa_texformat_ycbcr_rev; + return MESA_FORMAT_YCBCR_REV; case GL_RGB_S3TC: case GL_RGB4_S3TC: case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: - return &_mesa_texformat_rgb_dxt1; + return MESA_FORMAT_RGB_DXT1; case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: - return &_mesa_texformat_rgba_dxt1; + return MESA_FORMAT_RGBA_DXT1; case GL_RGBA_S3TC: case GL_RGBA4_S3TC: case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: - return &_mesa_texformat_rgba_dxt3; + return MESA_FORMAT_RGBA_DXT3; case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: - return &_mesa_texformat_rgba_dxt5; + return MESA_FORMAT_RGBA_DXT5; case GL_ALPHA16F_ARB: - return &_mesa_texformat_alpha_float16; + return MESA_FORMAT_ALPHA_FLOAT16; case GL_ALPHA32F_ARB: - return &_mesa_texformat_alpha_float32; + return MESA_FORMAT_ALPHA_FLOAT32; case GL_LUMINANCE16F_ARB: - return &_mesa_texformat_luminance_float16; + return MESA_FORMAT_LUMINANCE_FLOAT16; case GL_LUMINANCE32F_ARB: - return &_mesa_texformat_luminance_float32; + return MESA_FORMAT_LUMINANCE_FLOAT32; case GL_LUMINANCE_ALPHA16F_ARB: - return &_mesa_texformat_luminance_alpha_float16; + return MESA_FORMAT_LUMINANCE_ALPHA_FLOAT16; case GL_LUMINANCE_ALPHA32F_ARB: - return &_mesa_texformat_luminance_alpha_float32; + return MESA_FORMAT_LUMINANCE_ALPHA_FLOAT32; case GL_INTENSITY16F_ARB: - return &_mesa_texformat_intensity_float16; + return MESA_FORMAT_INTENSITY_FLOAT16; case GL_INTENSITY32F_ARB: - return &_mesa_texformat_intensity_float32; + return MESA_FORMAT_INTENSITY_FLOAT32; case GL_RGB16F_ARB: - return &_mesa_texformat_rgba_float16; + return MESA_FORMAT_RGBA_FLOAT16; case GL_RGB32F_ARB: - return &_mesa_texformat_rgba_float32; + return MESA_FORMAT_RGBA_FLOAT32; case GL_RGBA16F_ARB: - return &_mesa_texformat_rgba_float16; + return MESA_FORMAT_RGBA_FLOAT16; case GL_RGBA32F_ARB: - return &_mesa_texformat_rgba_float32; + return MESA_FORMAT_RGBA_FLOAT32; case GL_DEPTH_COMPONENT: case GL_DEPTH_COMPONENT16: @@ -476,7 +478,7 @@ const struct gl_texture_format *radeonChooseTextureFormat(GLcontext * ctx, case GL_DEPTH_COMPONENT32: case GL_DEPTH_STENCIL_EXT: case GL_DEPTH24_STENCIL8_EXT: - return &_mesa_texformat_s8_z24; + return MESA_FORMAT_S8_Z24; /* EXT_texture_sRGB */ case GL_SRGB: @@ -485,26 +487,193 @@ const struct gl_texture_format *radeonChooseTextureFormat(GLcontext * ctx, case GL_SRGB8_ALPHA8: case GL_COMPRESSED_SRGB: case GL_COMPRESSED_SRGB_ALPHA: - return &_mesa_texformat_srgba8; + return MESA_FORMAT_SRGBA8; case GL_SLUMINANCE: case GL_SLUMINANCE8: case GL_COMPRESSED_SLUMINANCE: - return &_mesa_texformat_sl8; + return MESA_FORMAT_SL8; case GL_SLUMINANCE_ALPHA: case GL_SLUMINANCE8_ALPHA8: case GL_COMPRESSED_SLUMINANCE_ALPHA: - return &_mesa_texformat_sla8; + return MESA_FORMAT_SLA8; default: _mesa_problem(ctx, "unexpected internalFormat 0x%x in %s", (int)internalFormat, __func__); + return MESA_FORMAT_NONE; + } + + return MESA_FORMAT_NONE; /* never get here */ +} + +/** Check if given image is valid within current texture object. + */ +static int image_matches_texture_obj(struct gl_texture_object *texObj, + struct gl_texture_image *texImage, + unsigned level) +{ + const struct gl_texture_image *baseImage = texObj->Image[0][texObj->BaseLevel]; + + if (!baseImage) + return 0; + + if (level < texObj->BaseLevel || level > texObj->MaxLevel) + return 0; + + const unsigned levelDiff = level - texObj->BaseLevel; + const unsigned refWidth = MAX2(baseImage->Width >> levelDiff, 1); + const unsigned refHeight = MAX2(baseImage->Height >> levelDiff, 1); + const unsigned refDepth = MAX2(baseImage->Depth >> levelDiff, 1); + + return (texImage->Width == refWidth && + texImage->Height == refHeight && + texImage->Depth == refDepth); +} + +static void teximage_assign_miptree(radeonContextPtr rmesa, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage, + unsigned face, + unsigned level) +{ + radeonTexObj *t = radeon_tex_obj(texObj); + radeon_texture_image* image = get_radeon_texture_image(texImage); + + /* Since miptree holds only images for levels <BaseLevel..MaxLevel> + * don't allocate the miptree if the teximage won't fit. + */ + if (!image_matches_texture_obj(texObj, texImage, level)) + return; + + /* Try using current miptree, or create new if there isn't any */ + if (!t->mt || !radeon_miptree_matches_image(t->mt, texImage, face, level)) { + radeon_miptree_unreference(&t->mt); + radeon_try_alloc_miptree(rmesa, t); + if (RADEON_DEBUG & RADEON_TEXTURE) { + fprintf(stderr, "%s: texObj %p, texImage %p, face %d, level %d, " + "texObj miptree doesn't match, allocated new miptree %p\n", + __FUNCTION__, texObj, texImage, face, level, t->mt); + } + } + + /* Miptree alocation may have failed, + * when there was no image for baselevel specified */ + if (t->mt) { + image->mtface = face; + image->mtlevel = level; + radeon_miptree_reference(t->mt, &image->mt); + } +} + +static GLuint * allocate_image_offsets(GLcontext *ctx, + unsigned alignedWidth, + unsigned height, + unsigned depth) +{ + int i; + GLuint *offsets; + + offsets = _mesa_malloc(depth * sizeof(GLuint)) ; + if (!offsets) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTex[Sub]Image"); return NULL; } - return NULL; /* never get here */ + for (i = 0; i < depth; ++i) { + offsets[i] = alignedWidth * height * i; + } + + return offsets; +} + +/** + * Update a subregion of the given texture image. + */ +static void radeon_store_teximage(GLcontext* ctx, int dims, + GLint xoffset, GLint yoffset, GLint zoffset, + GLsizei width, GLsizei height, GLsizei depth, + GLsizei imageSize, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage, + int compressed) +{ + radeonTexObj *t = radeon_tex_obj(texObj); + radeon_texture_image* image = get_radeon_texture_image(texImage); + + GLuint dstRowStride; + GLuint *dstImageOffsets; + + if (image->mt) { + dstRowStride = image->mt->levels[image->mtlevel].rowstride; + } else if (t->bo) { + /* TFP case */ + /* TODO */ + assert(0); + } else { + dstRowStride = _mesa_format_row_stride(texImage->TexFormat, texImage->Width); + } + + assert(dstRowStride); + + if (dims == 3) { + unsigned alignedWidth = dstRowStride/_mesa_get_format_bytes(texImage->TexFormat); + dstImageOffsets = allocate_image_offsets(ctx, alignedWidth, texImage->Height, texImage->Depth); + if (!dstImageOffsets) { + return; + } + } else { + dstImageOffsets = texImage->ImageOffsets; + } + + radeon_teximage_map(image, GL_TRUE); + + if (compressed) { + uint32_t srcRowStride, bytesPerRow, rows, block_width, block_height; + GLubyte *img_start; + + _mesa_get_format_block_size(texImage->TexFormat, &block_width, &block_height); + + if (!image->mt) { + dstRowStride = _mesa_format_row_stride(texImage->TexFormat, texImage->Width); + img_start = _mesa_compressed_image_address(xoffset, yoffset, 0, + texImage->TexFormat, + texImage->Width, texImage->Data); + } + else { + uint32_t offset; + offset = dstRowStride / _mesa_get_format_bytes(texImage->TexFormat) * yoffset / block_height + xoffset / block_width; + offset *= _mesa_get_format_bytes(texImage->TexFormat); + img_start = texImage->Data + offset; + } + srcRowStride = _mesa_format_row_stride(texImage->TexFormat, width); + bytesPerRow = srcRowStride; + rows = (height + block_height - 1) / block_height; + + copy_rows(img_start, dstRowStride, pixels, srcRowStride, rows, bytesPerRow); + } + else { + if (!_mesa_texstore(ctx, dims, texImage->_BaseFormat, + texImage->TexFormat, texImage->Data, + xoffset, yoffset, zoffset, + dstRowStride, + dstImageOffsets, + width, height, depth, + format, type, pixels, packing)) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage"); + } + } + + if (dims == 3) { + _mesa_free(dstImageOffsets); + } + + radeon_teximage_unmap(image); } /** @@ -525,13 +694,22 @@ static void radeon_teximage( radeonContextPtr rmesa = RADEON_CONTEXT(ctx); radeonTexObj* t = radeon_tex_obj(texObj); radeon_texture_image* image = get_radeon_texture_image(texImage); - GLuint dstRowStride; GLint postConvWidth = width; GLint postConvHeight = height; - GLuint texelBytes; GLuint face = radeon_face_for_target(target); - radeon_firevertices(rmesa); + { + struct radeon_bo *bo; + bo = !image->mt ? image->bo : image->mt->bo; + if (bo && radeon_bo_is_referenced_by_cs(bo, rmesa->cmdbuf.cs)) { + radeon_firevertices(rmesa); + } + } + + if (RADEON_DEBUG & RADEON_TEXTURE) { + fprintf(stderr, "radeon_teximage%dd: texObj %p, texImage %p, face %d, level %d\n", + dims, texObj, texImage, face, level); + } t->validated = GL_FALSE; @@ -540,62 +718,35 @@ static void radeon_teximage( &postConvHeight); } - /* Choose and fill in the texture format for this image */ - texImage->TexFormat = radeonChooseTextureFormat(ctx, internalFormat, format, type, 0); - _mesa_set_fetch_functions(texImage, dims); - - if (texImage->TexFormat->TexelBytes == 0) { - texelBytes = 0; - texImage->IsCompressed = GL_TRUE; - texImage->CompressedSize = - ctx->Driver.CompressedTextureSize(ctx, texImage->Width, - texImage->Height, texImage->Depth, - texImage->TexFormat->MesaFormat); - } else { - texImage->IsCompressed = GL_FALSE; - texImage->CompressedSize = 0; - - texelBytes = texImage->TexFormat->TexelBytes; + if (!_mesa_is_format_compressed(texImage->TexFormat)) { + GLuint texelBytes = _mesa_get_format_bytes(texImage->TexFormat); /* Minimum pitch of 32 bytes */ if (postConvWidth * texelBytes < 32) { - postConvWidth = 32 / texelBytes; - texImage->RowStride = postConvWidth; + postConvWidth = 32 / texelBytes; + texImage->RowStride = postConvWidth; } - if (!image->mt) { + if (!image->mt) { assert(texImage->RowStride == postConvWidth); } } - /* Allocate memory for image */ - radeonFreeTexImageData(ctx, texImage); /* Mesa core only clears texImage->Data but not image->mt */ - - if (t->mt && - t->mt->firstLevel == level && - t->mt->lastLevel == level && - t->mt->target != GL_TEXTURE_CUBE_MAP_ARB && - !radeon_miptree_matches_image(t->mt, texImage, face, level)) { - radeon_miptree_unreference(t->mt); - t->mt = NULL; - } - - if (!t->mt) - radeon_try_alloc_miptree(rmesa, t, image, face, level); - if (t->mt && radeon_miptree_matches_image(t->mt, texImage, face, level)) { - radeon_mipmap_level *lvl; - image->mt = t->mt; - image->mtlevel = level - t->mt->firstLevel; - image->mtface = face; - radeon_miptree_reference(t->mt); - lvl = &image->mt->levels[image->mtlevel]; - dstRowStride = lvl->rowstride; - } else { - int size; - if (texImage->IsCompressed) { - size = texImage->CompressedSize; - } else { - size = texImage->Width * texImage->Height * texImage->Depth * texImage->TexFormat->TexelBytes; + /* Mesa core only clears texImage->Data but not image->mt */ + radeonFreeTexImageData(ctx, texImage); + + if (!t->bo) { + teximage_assign_miptree(rmesa, texObj, texImage, face, level); + if (!image->mt) { + int size = _mesa_format_image_size(texImage->TexFormat, + texImage->Width, + texImage->Height, + texImage->Depth); + texImage->Data = _mesa_alloc_texmemory(size); + if (RADEON_DEBUG & RADEON_TEXTURE) { + fprintf(stderr, "radeon_teximage%dd: texObj %p, texImage %p, " + " no miptree assigned, using local memory %p\n", + dims, texObj, texImage, texImage->Data); + } } - texImage->Data = _mesa_alloc_texmemory(size); } /* Upload texture image; note that the spec allows pixels to be NULL */ @@ -609,64 +760,16 @@ static void radeon_teximage( } if (pixels) { - radeon_teximage_map(image, GL_TRUE); - if (compressed) { - if (image->mt) { - uint32_t srcRowStride, bytesPerRow, rows; - srcRowStride = _mesa_compressed_row_stride(texImage->TexFormat->MesaFormat, width); - bytesPerRow = srcRowStride; - rows = (height + 3) / 4; - copy_rows(texImage->Data, image->mt->levels[level].rowstride, - pixels, srcRowStride, rows, bytesPerRow); - } else { - memcpy(texImage->Data, pixels, imageSize); - } - } else { - GLuint dstRowStride; - GLuint *dstImageOffsets; - - if (image->mt) { - radeon_mipmap_level *lvl = &image->mt->levels[image->mtlevel]; - dstRowStride = lvl->rowstride; - } else { - dstRowStride = texImage->Width * texImage->TexFormat->TexelBytes; - } - - if (dims == 3) { - int i; - - dstImageOffsets = _mesa_malloc(depth * sizeof(GLuint)) ; - if (!dstImageOffsets) - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); - - for (i = 0; i < depth; ++i) { - dstImageOffsets[i] = dstRowStride/texImage->TexFormat->TexelBytes * height * i; - } - } else { - dstImageOffsets = texImage->ImageOffsets; - } - - if (!texImage->TexFormat->StoreImage(ctx, dims, - texImage->_BaseFormat, - texImage->TexFormat, - texImage->Data, 0, 0, 0, /* dstX/Y/Zoffset */ - dstRowStride, - dstImageOffsets, - width, height, depth, - format, type, pixels, packing)) - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); - - if (dims == 3) - _mesa_free(dstImageOffsets); - } + radeon_store_teximage(ctx, dims, + 0, 0, 0, + width, height, depth, + imageSize, format, type, + pixels, packing, + texObj, texImage, + compressed); } _mesa_unmap_teximage_pbo(ctx, packing); - - if (pixels) - radeon_teximage_unmap(image); - - } void radeonTexImage1D(GLcontext * ctx, GLenum target, GLint level, @@ -719,7 +822,7 @@ void radeonTexImage3D(GLcontext * ctx, GLenum target, GLint level, } /** - * Update a subregion of the given texture image. + * All glTexSubImage calls go through this function. */ static void radeon_texsubimage(GLcontext* ctx, int dims, GLenum target, int level, GLint xoffset, GLint yoffset, GLint zoffset, @@ -736,64 +839,39 @@ static void radeon_texsubimage(GLcontext* ctx, int dims, GLenum target, int leve radeonTexObj* t = radeon_tex_obj(texObj); radeon_texture_image* image = get_radeon_texture_image(texImage); - radeon_firevertices(rmesa); + { + struct radeon_bo *bo; + bo = !image->mt ? image->bo : image->mt->bo; + if (bo && radeon_bo_is_referenced_by_cs(bo, rmesa->cmdbuf.cs)) { + radeon_firevertices(rmesa); + } + } + + if (RADEON_DEBUG & RADEON_TEXTURE) { + fprintf(stderr, "radeon_texsubimage%dd: texObj %p, texImage %p, face %d, level %d\n", + dims, texObj, texImage, radeon_face_for_target(target), level); + } t->validated = GL_FALSE; if (compressed) { pixels = _mesa_validate_pbo_compressed_teximage( - ctx, imageSize, pixels, packing, "glCompressedTexImage"); + ctx, imageSize, pixels, packing, "glCompressedTexSubImage"); } else { pixels = _mesa_validate_pbo_teximage(ctx, dims, - width, height, depth, format, type, pixels, packing, "glTexSubImage1D"); + width, height, depth, format, type, pixels, packing, "glTexSubImage"); } if (pixels) { - GLint dstRowStride; - radeon_teximage_map(image, GL_TRUE); - - if (image->mt) { - radeon_mipmap_level *lvl = &image->mt->levels[image->mtlevel]; - dstRowStride = lvl->rowstride; - } else { - dstRowStride = texImage->RowStride * texImage->TexFormat->TexelBytes; - } - - if (compressed) { - uint32_t srcRowStride, bytesPerRow, rows; - GLubyte *img_start; - if (!image->mt) { - dstRowStride = _mesa_compressed_row_stride(texImage->TexFormat->MesaFormat, texImage->Width); - img_start = _mesa_compressed_image_address(xoffset, yoffset, 0, - texImage->TexFormat->MesaFormat, - texImage->Width, texImage->Data); - } - else { - uint32_t blocks_x = dstRowStride / (image->mt->bpp * 4); - img_start = texImage->Data + image->mt->bpp * 4 * (blocks_x * (yoffset / 4) + xoffset / 4); - } - srcRowStride = _mesa_compressed_row_stride(texImage->TexFormat->MesaFormat, width); - bytesPerRow = srcRowStride; - rows = (height + 3) / 4; - - copy_rows(img_start, dstRowStride, pixels, srcRowStride, rows, bytesPerRow); - - } else { - if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat, - texImage->TexFormat, texImage->Data, - xoffset, yoffset, zoffset, - dstRowStride, - texImage->ImageOffsets, - width, height, depth, - format, type, pixels, packing)) - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage"); - } + radeon_store_teximage(ctx, dims, + xoffset, yoffset, zoffset, + width, height, depth, + imageSize, format, type, + pixels, packing, + texObj, texImage, + compressed); } - radeon_teximage_unmap(image); - _mesa_unmap_teximage_pbo(ctx, packing); - - } void radeonTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, @@ -849,143 +927,6 @@ void radeonTexSubImage3D(GLcontext * ctx, GLenum target, GLint level, format, type, pixels, packing, texObj, texImage, 0); } - - -/** - * Ensure that the given image is stored in the given miptree from now on. - */ -static void migrate_image_to_miptree(radeon_mipmap_tree *mt, radeon_texture_image *image, int face, int level) -{ - radeon_mipmap_level *dstlvl = &mt->levels[level - mt->firstLevel]; - unsigned char *dest; - - assert(image->mt != mt); - assert(dstlvl->width == image->base.Width); - assert(dstlvl->height == image->base.Height); - assert(dstlvl->depth == image->base.Depth); - - - radeon_bo_map(mt->bo, GL_TRUE); - dest = mt->bo->ptr + dstlvl->faces[face].offset; - - if (image->mt) { - /* Format etc. should match, so we really just need a memcpy(). - * In fact, that memcpy() could be done by the hardware in many - * cases, provided that we have a proper memory manager. - */ - radeon_mipmap_level *srclvl = &image->mt->levels[image->mtlevel-image->mt->firstLevel]; - - assert(srclvl->size == dstlvl->size); - assert(srclvl->rowstride == dstlvl->rowstride); - - radeon_bo_map(image->mt->bo, GL_FALSE); - - memcpy(dest, - image->mt->bo->ptr + srclvl->faces[face].offset, - dstlvl->size); - radeon_bo_unmap(image->mt->bo); - - radeon_miptree_unreference(image->mt); - } else { - uint32_t srcrowstride; - uint32_t height; - /* need to confirm this value is correct */ - if (mt->compressed) { - height = (image->base.Height + 3) / 4; - srcrowstride = _mesa_compressed_row_stride(image->base.TexFormat->MesaFormat, image->base.Width); - } else { - height = image->base.Height * image->base.Depth; - srcrowstride = image->base.Width * image->base.TexFormat->TexelBytes; - } - -// if (mt->tilebits) -// WARN_ONCE("%s: tiling not supported yet", __FUNCTION__); - - copy_rows(dest, dstlvl->rowstride, image->base.Data, srcrowstride, - height, srcrowstride); - - _mesa_free_texmemory(image->base.Data); - image->base.Data = 0; - } - - radeon_bo_unmap(mt->bo); - - image->mt = mt; - image->mtface = face; - image->mtlevel = level; - radeon_miptree_reference(image->mt); -} - -int radeon_validate_texture_miptree(GLcontext * ctx, struct gl_texture_object *texObj) -{ - radeonContextPtr rmesa = RADEON_CONTEXT(ctx); - radeonTexObj *t = radeon_tex_obj(texObj); - radeon_texture_image *baseimage = get_radeon_texture_image(texObj->Image[0][texObj->BaseLevel]); - int face, level; - - if (t->validated || t->image_override) - return GL_TRUE; - - if (RADEON_DEBUG & RADEON_TEXTURE) - fprintf(stderr, "%s: Validating texture %p now\n", __FUNCTION__, texObj); - - if (baseimage->base.Border > 0) - return GL_FALSE; - - /* Ensure a matching miptree exists. - * - * Differing mipmap trees can result when the app uses TexImage to - * change texture dimensions. - * - * Prefer to use base image's miptree if it - * exists, since that most likely contains more valid data (remember - * that the base level is usually significantly larger than the rest - * of the miptree, so cubemaps are the only possible exception). - */ - if (baseimage->mt && - baseimage->mt != t->mt && - radeon_miptree_matches_texture(baseimage->mt, &t->base)) { - radeon_miptree_unreference(t->mt); - t->mt = baseimage->mt; - radeon_miptree_reference(t->mt); - } else if (t->mt && !radeon_miptree_matches_texture(t->mt, &t->base)) { - radeon_miptree_unreference(t->mt); - t->mt = 0; - } - - if (!t->mt) { - if (RADEON_DEBUG & RADEON_TEXTURE) - fprintf(stderr, " Allocate new miptree\n"); - radeon_try_alloc_miptree(rmesa, t, baseimage, 0, texObj->BaseLevel); - if (!t->mt) { - _mesa_problem(ctx, "radeon_validate_texture failed to alloc miptree"); - return GL_FALSE; - } - } - - /* Ensure all images are stored in the single main miptree */ - for(face = 0; face < t->mt->faces; ++face) { - for(level = t->mt->firstLevel; level <= t->mt->lastLevel; ++level) { - radeon_texture_image *image = get_radeon_texture_image(texObj->Image[face][level]); - if (RADEON_DEBUG & RADEON_TEXTURE) - fprintf(stderr, " face %i, level %i... %p vs %p ", face, level, t->mt, image->mt); - if (t->mt == image->mt) { - if (RADEON_DEBUG & RADEON_TEXTURE) - fprintf(stderr, "OK\n"); - - continue; - } - - if (RADEON_DEBUG & RADEON_TEXTURE) - fprintf(stderr, "migrating\n"); - migrate_image_to_miptree(t->mt, image, face, level); - } - } - - return GL_TRUE; -} - - /** * Need to map texture image into memory before copying image data, * then unmap it. diff --git a/src/mesa/drivers/dri/radeon/radeon_texture.h b/src/mesa/drivers/dri/radeon/radeon_texture.h index 888a55ba911..906daf12d0a 100644 --- a/src/mesa/drivers/dri/radeon/radeon_texture.h +++ b/src/mesa/drivers/dri/radeon/radeon_texture.h @@ -30,6 +30,11 @@ #ifndef RADEON_TEXTURE_H #define RADEON_TEXTURE_H + +#include "main/formats.h" + +void copy_rows(void* dst, GLuint dststride, const void* src, GLuint srcstride, + GLuint numrows, GLuint rowsize); struct gl_texture_image *radeonNewTextureImage(GLcontext *ctx); void radeonFreeTexImageData(GLcontext *ctx, struct gl_texture_image *timage); @@ -40,14 +45,16 @@ void radeonUnmapTexture(GLcontext *ctx, struct gl_texture_object *texObj); void radeonGenerateMipmap(GLcontext* ctx, GLenum target, struct gl_texture_object *texObj); int radeon_validate_texture_miptree(GLcontext * ctx, struct gl_texture_object *texObj); GLuint radeon_face_for_target(GLenum target); -const struct gl_texture_format *radeonChooseTextureFormat_mesa(GLcontext * ctx, - GLint internalFormat, - GLenum format, - GLenum type); -const struct gl_texture_format *radeonChooseTextureFormat(GLcontext * ctx, - GLint internalFormat, - GLenum format, - GLenum type, GLboolean fbo); + +gl_format radeonChooseTextureFormat_mesa(GLcontext * ctx, + GLint internalFormat, + GLenum format, + GLenum type); + +gl_format radeonChooseTextureFormat(GLcontext * ctx, + GLint internalFormat, + GLenum format, + GLenum type, GLboolean fbo); void radeonTexImage1D(GLcontext * ctx, GLenum target, GLint level, GLint internalFormat, diff --git a/src/mesa/drivers/dri/s3v/Makefile b/src/mesa/drivers/dri/s3v/Makefile deleted file mode 100644 index 9bd79731548..00000000000 --- a/src/mesa/drivers/dri/s3v/Makefile +++ /dev/null @@ -1,36 +0,0 @@ -# src/mesa/drivers/dri/s3v/Makefile - -TOP = ../../../../.. -include $(TOP)/configs/current - -LIBNAME = s3v_dri.so - -# Doesn't exist yet. -#MINIGLX_SOURCES = server/savage_dri.c - -DRIVER_SOURCES = \ - s3v_context.c \ - s3v_dd.c \ - s3v_inithw.c \ - s3v_lock.c \ - s3v_render.c \ - s3v_screen.c \ - s3v_span.c \ - s3v_state.c \ - s3v_tex.c \ - s3v_texmem.c \ - s3v_texstate.c \ - s3v_tris.c \ - s3v_vb.c \ - s3v_xmesa.c - -C_SOURCES = \ - $(COMMON_SOURCES) \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - - -include ../Makefile.template - -symlinks: diff --git a/src/mesa/drivers/dri/s3v/s3v_common.h b/src/mesa/drivers/dri/s3v/s3v_common.h deleted file mode 100644 index b66cdf1df08..00000000000 --- a/src/mesa/drivers/dri/s3v/s3v_common.h +++ /dev/null @@ -1,83 +0,0 @@ -/* - * Author: Max Lingua <[email protected]> - */ - -/* WARNING: If you change any of these defines, make sure to change - * the kernel include file as well (s3v_drm.h) - */ - -#ifndef _XF86DRI_S3V_H_ -#define _XF86DRI_S3V_H_ - -#ifndef _S3V_DEFINES_ -#define _S3V_DEFINES_ -#define S3V_USE_BATCH 1 - -/* #define S3V_BUF_4K 1 */ - -#ifdef S3V_BUF_4K -#define S3V_DMA_BUF_ORDER 12 -#define S3V_DMA_BUF_NR 256 -#else -#define S3V_DMA_BUF_ORDER 16 /* -much- better */ -#define S3V_DMA_BUF_NR 16 -#endif -/* on s3virge you can only choose between * - * 4k (2^12) and 64k (2^16) dma bufs */ -#define S3V_DMA_BUF_SZ (1<<S3V_DMA_BUF_ORDER) - -#define S3V_NR_SAREA_CLIPRECTS 8 - -/* Each region is a minimum of 16k (64*64@4bpp) - * and there are at most 40 of them. - */ -#define S3V_NR_TEX_REGIONS 64 /* was 40 */ -#define S3V_LOG_TEX_GRANULARITY 16 /* was 4 */ -/* 40 * (2 ^ 4) = 640k, that's all we have for tex on 4mb gfx card */ -/* FIXME: will it work with card with less than 4mb? */ -/* FIXME: we should set this at run time */ - -#endif /* _S3V_DEFINES */ - -/* - * WARNING: If you change any of these defines, make sure to change - * the kernel include file as well (gamma_drm.h) - */ - -/* Driver specific DRM command indices - * NOTE: these are not OS specific, but they are driver specific - */ -#define DRM_S3V_INIT_DMA 0x00 -#define DRM_S3V_CLEANUP_DMA 0x01 - -typedef struct _drmS3VInit { - enum { - S3V_INIT_DMA = 0x01, - S3V_CLEANUP_DMA = 0x02 - } func; - - unsigned int pcimode; /* bool: 1=pci 0=agp */ - - unsigned int mmio_offset; - unsigned int buffers_offset; - unsigned int sarea_priv_offset; - - unsigned int front_offset; - unsigned int front_width; - unsigned int front_height; - unsigned int front_pitch; - - unsigned int back_offset; - unsigned int back_width; - unsigned int back_height; - unsigned int back_pitch; - - unsigned int depth_offset; - unsigned int depth_width; - unsigned int depth_height; - unsigned int depth_pitch; - - unsigned int texture_offset; -} drmS3VInit; - -#endif diff --git a/src/mesa/drivers/dri/s3v/s3v_context.c b/src/mesa/drivers/dri/s3v/s3v_context.c deleted file mode 100644 index 0a3bf7258d1..00000000000 --- a/src/mesa/drivers/dri/s3v/s3v_context.c +++ /dev/null @@ -1,260 +0,0 @@ -/* - * Author: Max Lingua <[email protected]> - */ - -#include "s3v_context.h" - -#include "swrast/swrast.h" -#include "swrast_setup/swrast_setup.h" -#include "vbo/vbo.h" - -#include "tnl/tnl.h" -#include "tnl/t_pipeline.h" - -#include "main/context.h" -#include "main/simple_list.h" -#include "main/matrix.h" -#include "main/extensions.h" -#if defined(USE_X86_ASM) -#include "x86/common_x86_asm.h" -#endif -#include "main/simple_list.h" -#include "main/mm.h" - -#include "drivers/common/driverfuncs.h" -#include "s3v_vb.h" -#include "s3v_tris.h" - -#if 0 -extern const struct tnl_pipeline_stage _s3v_render_stage; - -static const struct tnl_pipeline_stage *s3v_pipeline[] = { - &_tnl_vertex_transform_stage, - &_tnl_normal_transform_stage, - &_tnl_lighting_stage, - &_tnl_fog_coordinate_stage, - &_tnl_texgen_stage, - &_tnl_texture_transform_stage, - /* REMOVE: point attenuation stage */ -#if 1 - &_s3v_render_stage, /* ADD: unclipped rastersetup-to-dma */ -#endif - &_tnl_render_stage, - 0, -}; -#endif - -GLboolean s3vCreateContext(const __GLcontextModes *glVisual, - __DRIcontextPrivate *driContextPriv, - void *sharedContextPrivate) -{ - GLcontext *ctx, *shareCtx; - __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; - s3vContextPtr vmesa; - s3vScreenPtr s3vScrn; - S3VSAREAPtr saPriv=(S3VSAREAPtr)(((char*)sPriv->pSAREA) + - sizeof(drm_sarea_t)); - struct dd_function_table functions; - - DEBUG_WHERE(("*** s3vCreateContext ***\n")); - - vmesa = (s3vContextPtr) CALLOC( sizeof(*vmesa) ); - if ( !vmesa ) return GL_FALSE; - - /* Allocate the Mesa context */ - if (sharedContextPrivate) - shareCtx = ((s3vContextPtr) sharedContextPrivate)->glCtx; - else - shareCtx = NULL; - - _mesa_init_driver_functions(&functions); - - vmesa->glCtx = _mesa_create_context(glVisual, shareCtx, &functions, - (void *)vmesa); - if (!vmesa->glCtx) { - FREE(vmesa); - return GL_FALSE; - } - - vmesa->driContext = driContextPriv; - vmesa->driScreen = sPriv; - vmesa->driDrawable = NULL; /* Set by XMesaMakeCurrent */ - - vmesa->hHWContext = driContextPriv->hHWContext; - vmesa->driHwLock = (drmLock *)&sPriv->pSAREA->lock; - vmesa->driFd = sPriv->fd; - vmesa->sarea = saPriv; - - s3vScrn = vmesa->s3vScreen = (s3vScreenPtr)(sPriv->private); - - ctx = vmesa->glCtx; - - ctx->Const.MaxTextureLevels = 11; /* it is (11-1) -> 1024 * 1024 FIXME */ - - ctx->Const.MaxTextureUnits = 1; /* FIXME: or 2 ? */ - - /* No wide points. - */ - ctx->Const.MinPointSize = 1.0; - ctx->Const.MinPointSizeAA = 1.0; - ctx->Const.MaxPointSize = 1.0; - ctx->Const.MaxPointSizeAA = 1.0; - - /* No wide lines. - */ - ctx->Const.MinLineWidth = 1.0; - ctx->Const.MinLineWidthAA = 1.0; - ctx->Const.MaxLineWidth = 1.0; - ctx->Const.MaxLineWidthAA = 1.0; - ctx->Const.LineWidthGranularity = 1.0; - - ctx->Const.MaxDrawBuffers = 1; - - vmesa->texHeap = mmInit( 0, vmesa->s3vScreen->textureSize ); - DEBUG(("vmesa->s3vScreen->textureSize = 0x%x\n", - vmesa->s3vScreen->textureSize)); - - /* NOTE */ - /* mmInit(offset, size); */ - - /* allocates a structure like this: - - struct mem_block_t { - struct mem_block_t *next; - struct mem_block_t *heap; - int ofs,size; - int align; - int free:1; - int reserved:1; - }; - - */ - - make_empty_list(&vmesa->TexObjList); - make_empty_list(&vmesa->SwappedOut); - - vmesa->CurrentTexObj[0] = 0; - vmesa->CurrentTexObj[1] = 0; /* FIXME */ - - vmesa->RenderIndex = ~0; - - /* Initialize the software rasterizer and helper modules. - */ - _swrast_CreateContext( ctx ); - _vbo_CreateContext( ctx ); - _tnl_CreateContext( ctx ); - _swsetup_CreateContext( ctx ); - - /* Install the customized pipeline: - */ -#if 0 - _tnl_destroy_pipeline( ctx ); - _tnl_install_pipeline( ctx, s3v_pipeline ); -#endif - /* Configure swrast to match hardware characteristics: - */ -#if 0 - _swrast_allow_pixel_fog( ctx, GL_FALSE ); - _swrast_allow_vertex_fog( ctx, GL_TRUE ); -#endif - vmesa->_3d_mode = 0; - - /* 3D lines / gouraud tris */ - vmesa->CMD = ( AUTO_EXEC_ON | HW_CLIP_ON | DEST_COL_1555 - | FOG_OFF | ALPHA_OFF | Z_OFF | Z_UPDATE_OFF - | Z_LESS | TEX_WRAP_ON | TEX_MODULATE | LINEAR - | TEX_COL_ARGB1555 | CMD_3D ); - - vmesa->_alpha[0] = vmesa->_alpha[1] = ALPHA_OFF; - vmesa->alpha_cmd = vmesa->_alpha[0]; - vmesa->_tri[0] = DO_GOURAUD_TRI; - vmesa->_tri[1] = DO_TEX_LIT_TRI; - vmesa->prim_cmd = vmesa->_tri[0]; - - /* printf("first vmesa->CMD = 0x%x\n", vmesa->CMD); */ - - vmesa->TexOffset = vmesa->s3vScreen->texOffset; - - s3vInitVB( ctx ); - s3vInitExtensions( ctx ); - s3vInitDriverFuncs( ctx ); - s3vInitStateFuncs( ctx ); - s3vInitSpanFuncs( ctx ); - s3vInitTextureFuncs( ctx ); - s3vInitTriFuncs( ctx ); - s3vInitState( vmesa ); - - driContextPriv->driverPrivate = (void *)vmesa; - - /* HACK */ - vmesa->bufSize = S3V_DMA_BUF_SZ; - - DEBUG(("vmesa->bufSize = %i\n", vmesa->bufSize)); - DEBUG(("vmesa->bufCount = %i\n", vmesa->bufCount)); - - - /* dma init */ - DEBUG_BUFS(("GET_FIRST_DMA\n")); - - vmesa->_bufNum = 0; - - GET_FIRST_DMA(vmesa->driFd, vmesa->hHWContext, - 1, &(vmesa->bufIndex[0]), &(vmesa->bufSize), - &vmesa->_buf[0], &vmesa->bufCount, s3vScrn); - - GET_FIRST_DMA(vmesa->driFd, vmesa->hHWContext, - 1, &(vmesa->bufIndex[1]), &(vmesa->bufSize), - &vmesa->_buf[1], &vmesa->bufCount, s3vScrn); - - vmesa->buf = vmesa->_buf[vmesa->_bufNum]; - -/* - vmesa->CMD = (AUTO_EXEC_ON | HW_CLIP_ON | DEST_COL_1555 - | FOG_OFF | ALPHA_OFF | Z_OFF | Z_UPDATE_OFF - | DO_GOURAUD_TRI | CMD_3D); - - vmesa->TexOffset = vmesa->s3vScreen->texOffset; -*/ - -/* ... but we should support only 15 bit in virge (out of 8/15/24)... */ - - DEBUG(("glVisual->depthBits = %i\n", glVisual->depthBits)); - - switch (glVisual->depthBits) { - case 8: - break; - - case 15: - case 16: - vmesa->depth_scale = 1.0f / 0xffff; - break; - case 24: - vmesa->depth_scale = 1.0f / 0xffffff; - break; - default: - break; - } - - vmesa->cull_zero = 0.0f; - - vmesa->DepthSize = glVisual->depthBits; - vmesa->Flags = S3V_FRONT_BUFFER; - vmesa->Flags |= (glVisual->doubleBufferMode ? S3V_BACK_BUFFER : 0); - vmesa->Flags |= (vmesa->DepthSize > 0 ? S3V_DEPTH_BUFFER : 0); - - vmesa->EnabledFlags = S3V_FRONT_BUFFER; - vmesa->EnabledFlags |= (glVisual->doubleBufferMode ? S3V_BACK_BUFFER : 0); - - - if (vmesa->Flags & S3V_BACK_BUFFER) { - vmesa->readOffset = vmesa->drawOffset = vmesa->s3vScreen->backOffset; - } else { - vmesa->readOffset = vmesa->drawOffset = 0; - } - - s3vInitHW( vmesa ); - - driContextPriv->driverPrivate = (void *)vmesa; - - return GL_TRUE; -} diff --git a/src/mesa/drivers/dri/s3v/s3v_context.h b/src/mesa/drivers/dri/s3v/s3v_context.h deleted file mode 100644 index 671ba90d781..00000000000 --- a/src/mesa/drivers/dri/s3v/s3v_context.h +++ /dev/null @@ -1,443 +0,0 @@ -/* - * Author: Max Lingua <[email protected]> - */ - -#ifndef _S3V_CONTEXT_H_ -#define _S3V_CONTEXT_H_ - -#include "dri_util.h" - -#include "s3v_dri.h" -#include "s3v_regs.h" -#include "s3v_macros.h" -#include "s3v_screen.h" -#include "main/colormac.h" -#include "main/macros.h" -#include "main/mtypes.h" -#include "drm.h" -#include "main/mm.h" -#include "drirenderbuffer.h" - -/* Flags for context */ -#define S3V_FRONT_BUFFER 0x00000001 -#define S3V_BACK_BUFFER 0x00000002 -#define S3V_DEPTH_BUFFER 0x00000004 - - /* FIXME: check */ -#define S3V_MAX_TEXTURE_SIZE 2048 - -/* These are the minimum requirements and should probably be increased */ -#define MAX_MODELVIEW_STACK 16 -#define MAX_PROJECTION_STACK 2 -#define MAX_TEXTURE_STACK 2 - -extern void s3vDDUpdateHWState(GLcontext *ctx); -extern s3vScreenPtr s3vCreateScreen(__DRIscreenPrivate *sPriv); -extern void s3vDestroyScreen(__DRIscreenPrivate *sPriv); -extern GLboolean s3vCreateContext(const __GLcontextModes *glVisual, - __DRIcontextPrivate *driContextPriv, - void *sharedContextPrivate); - -#define S3V_UPLOAD_ALL 0xffffffff -/* #define S3V_UPLOAD_CLIPRECTS 0x00000002 */ -#define S3V_UPLOAD_ALPHA 0x00000004 -#define S3V_UPLOAD_BLEND 0x00000008 -#define S3V_UPLOAD_DEPTH 0x00000010 -#define S3V_UPLOAD_VIEWPORT 0x00000020 -#define S3V_UPLOAD_SHADE 0x00000040 -#define S3V_UPLOAD_CLIP 0x00000080 -#define S3V_UPLOAD_MASKS 0x00000100 -#define S3V_UPLOAD_WINDOW 0x00000200 /* defunct */ -#define S3V_UPLOAD_GEOMETRY 0x00000400 -#define S3V_UPLOAD_POLYGON 0x00000800 -#define S3V_UPLOAD_DITHER 0x00001000 -#define S3V_UPLOAD_LOGICOP 0x00002000 -#define S3V_UPLOAD_FOG 0x00004000 -#define S3V_UPLOAD_LIGHT 0x00008000 -#define S3V_UPLOAD_CONTEXT 0x00010000 -#define S3V_UPLOAD_TEX0 0x00020000 -#define S3V_UPLOAD_STIPPLE 0x00040000 -#define S3V_UPLOAD_TRANSFORM 0x00080000 -#define S3V_UPLOAD_LINEMODE 0x00100000 -#define S3V_UPLOAD_POINTMODE 0x00200000 -#define S3V_UPLOAD_TRIMODE 0x00400000 - -#define S3V_NEW_CLIP 0x00000001 -#define S3V_NEW_WINDOW 0x00000002 -#define S3V_NEW_CONTEXT 0x00000004 -#define S3V_NEW_TEXTURE 0x00000008 /* defunct */ -#define S3V_NEW_ALPHA 0x00000010 -#define S3V_NEW_DEPTH 0x00000020 -#define S3V_NEW_MASKS 0x00000040 -#define S3V_NEW_POLYGON 0x00000080 -#define S3V_NEW_CULL 0x00000100 -#define S3V_NEW_LOGICOP 0x00000200 -#define S3V_NEW_FOG 0x00000400 -#define S3V_NEW_LIGHT 0x00000800 -#define S3V_NEW_STIPPLE 0x00001000 -#define S3V_NEW_ALL 0xffffffff - -#define S3V_FALLBACK_TRI 0x00000001 -#define S3V_FALLBACK_TEXTURE 0x00000002 - -struct s3v_context; -typedef struct s3v_context s3vContextRec; -typedef struct s3v_context *s3vContextPtr; -typedef struct s3v_texture_object_t *s3vTextureObjectPtr; - -#define VALID_S3V_TEXTURE_OBJECT(tobj) (tobj) - -#define S3V_TEX_MAXLEVELS 12 - -/* For shared texture space managment, these texture objects may also - * be used as proxies for regions of texture memory containing other - * client's textures. Such proxy textures (not to be confused with GL - * proxy textures) are subject to the same LRU aging we use for our - * own private textures, and thus we have a mechanism where we can - * fairly decide between kicking out our own textures and those of - * other clients. - * - * Non-local texture objects have a valid MemBlock to describe the - * region managed by the other client, and can be identified by - * 't->globj == 0' - */ -struct s3v_texture_object_t { - struct s3v_texture_object_t *next, *prev; - - GLuint age; - struct gl_texture_object *globj; - - int Pitch; - int Height; - int WidthLog2; - int texelBytes; - int totalSize; - int bound; - - struct mem_block *MemBlock; - GLuint BufAddr; - - GLuint min_level; - GLuint max_level; - GLuint dirty_images; - - GLint firstLevel, lastLevel; /* upload tObj->Image[first .. lastLevel] */ - - struct { - const struct gl_texture_image *image; - int offset; /* into BufAddr */ - int height; - int internalFormat; - } image[S3V_TEX_MAXLEVELS]; - - GLuint TextureCMD; - - GLuint TextureColorMode; - GLuint TextureFilterMode; - GLuint TextureBorderColor; - GLuint TextureWrap; - GLuint TextureMipSize; - - GLuint TextureBaseAddr[S3V_TEX_MAXLEVELS]; - GLuint TextureFormat; - GLuint TextureReadMode; -}; - -#define S3V_NO_PALETTE 0x0 -#define S3V_USE_PALETTE 0x1 -#define S3V_UPDATE_PALETTE 0x2 -#define S3V_FALLBACK_PALETTE 0x4 - -void s3vUpdateTextureState( GLcontext *ctx ); - -void s3vDestroyTexObj( s3vContextPtr vmesa, s3vTextureObjectPtr t); -void s3vUploadTexImages( s3vContextPtr vmesa, s3vTextureObjectPtr t ); - -void s3vResetGlobalLRU( s3vContextPtr vmesa ); -void s3vTexturesGone( s3vContextPtr vmesa, - GLuint start, GLuint end, - GLuint in_use ); - -void s3vEmitHwState( s3vContextPtr vmesa ); -void s3vGetLock( s3vContextPtr vmesa, GLuint flags ); -void s3vInitExtensions( GLcontext *ctx ); -void s3vInitDriverFuncs( GLcontext *ctx ); -void s3vSetSpanFunctions(driRenderbuffer *rb, const GLvisual *vis); -void s3vInitState( s3vContextPtr vmesa ); -void s3vInitHW( s3vContextPtr vmesa ); -void s3vInitStateFuncs( GLcontext *ctx ); -void s3vInitTextureFuncs( GLcontext *ctx ); -void s3vInitTriFuncs( GLcontext *ctx ); - -void s3vUpdateWindow( GLcontext *ctx ); -void s3vUpdateViewportOffset( GLcontext *ctx ); - -void s3vPrintLocalLRU( s3vContextPtr vmesa ); -void s3vPrintGlobalLRU( s3vContextPtr vmesa ); - -extern void s3vFallback( s3vContextPtr vmesa, GLuint bit, GLboolean mode ); -#define FALLBACK( imesa, bit, mode ) s3vFallback( imesa, bit, mode ) - -/* Use the templated vertex formats. Only one of these is used in s3v. - */ -#define TAG(x) s3v##x -#include "tnl_dd/t_dd_vertex.h" -#undef TAG - -typedef void (*s3v_quad_func)( s3vContextPtr, - const s3vVertex *, - const s3vVertex *, - const s3vVertex *, - const s3vVertex * ); -typedef void (*s3v_tri_func)( s3vContextPtr, - const s3vVertex *, - const s3vVertex *, - const s3vVertex * ); -typedef void (*s3v_line_func)( s3vContextPtr, - const s3vVertex *, - const s3vVertex * ); -typedef void (*s3v_point_func)( s3vContextPtr, - const s3vVertex * ); - - -/* static void s3v_lines_emit(GLcontext *ctx, GLuint start, GLuint end); */ -typedef void (*emit_func)( GLcontext *, GLuint, GLuint); - -struct s3v_context { - GLcontext *glCtx; /* Mesa context */ - - __DRIcontextPrivate *driContext; - __DRIscreenPrivate *driScreen; - __DRIdrawablePrivate *driDrawable; - - GLuint new_gl_state; - GLuint new_state; - GLuint dirty; - - S3VSAREAPtr sarea; - - /* Temporaries for translating away float colors - */ - struct gl_client_array UbyteColor; - struct gl_client_array UbyteSecondaryColor; - - /* Mirrors of some DRI state - */ - - drm_context_t hHWContext; - drmLock *driHwLock; - int driFd; - - GLuint numClipRects; /* Cliprects for the draw buffer */ - drm_clip_rect_t *pClipRects; - - GLuint* buf; /* FIXME */ - GLuint* _buf[2]; - int _bufNum; - int bufIndex[2]; - int bufSize; - int bufCount; - - s3vScreenPtr s3vScreen; /* Screen private DRI data */ - - int drawOffset; - int readOffset; - - s3v_point_func draw_point; - s3v_line_func draw_line; - s3v_tri_func draw_tri; - s3v_quad_func draw_quad; - - GLuint Fallback; - GLuint RenderIndex; - GLuint SetupNewInputs; - GLuint SetupIndex; - - GLuint vertex_format; - GLuint vertex_size; - GLuint vertex_stride_shift; - char *verts; - - GLfloat hw_viewport[16]; - GLuint hw_primitive; - GLenum render_primitive; - - GLfloat depth_scale; - - s3vTextureObjectPtr CurrentTexObj[2]; - struct s3v_texture_object_t TexObjList; - struct s3v_texture_object_t SwappedOut; - GLenum TexEnvImageFmt[2]; - - struct mem_block *texHeap; - - int lastSwap; - int texAge; - int ctxAge; - int dirtyAge; - int lastStamp; - - /* max was here: don't touch */ - - unsigned int S3V_REG[S3V_REGS_NUM]; - - GLuint texMode; - GLuint alphaMode; - GLuint lightMode; - - GLuint SrcBase; - GLuint DestBase; - GLuint DestBlit; - GLuint ScissorLR; - GLuint ScissorTB; - GLuint ScissorWH; /* SubScissorWH */ /* RectWH */ - GLuint FrontStride; - GLuint BackStride; - GLuint SrcStride; - GLuint DestStride; - GLuint SrcXY; - GLuint DestXY; - - GLuint ClearColor; - GLuint Color; - GLuint DitherMode; - GLuint ClearDepth; - - GLuint TextureBorderColor; - GLuint TexOffset; - GLuint TexStride; - - GLuint CMD; - GLuint prim_cmd; - GLuint _tri[2]; /* 0 = gouraud; 1 = tex (lit or unlit) */ - GLuint alpha_cmd; /* actual alpha cmd */ - GLuint _alpha[2]; - GLuint _alpha_tex; /* tex alpha type */ - /* (3d_mode) 0 = 3d line/gourad tri; 1 = 3d tex tri */ - GLuint _3d_mode; - - GLfloat backface_sign; - GLfloat cull_zero; - - int restore_primitive; - -/* *** 2check *** */ - - GLuint FogMode; - GLuint AreaStippleMode; - GLuint LBReadFormat; - GLuint LBWriteFormat; - GLuint LineMode; - GLuint PointMode; - GLuint TriangleMode; - GLuint AntialiasMode; - GLfloat ViewportScaleX; - GLfloat ViewportScaleY; - GLfloat ViewportScaleZ; - GLfloat ViewportOffsetX; - GLfloat ViewportOffsetY; - GLfloat ViewportOffsetZ; - int MatrixMode; - int DepthMode; - int TransformMode; - int LBReadMode; - int FBReadMode; - int FBWindowBase; - int LBWindowBase; - int ColorDDAMode; - int GeometryMode; - int AlphaTestMode; - int AlphaBlendMode; - int AB_FBReadMode; - int AB_FBReadMode_Save; - int DeltaMode; - int ColorMaterialMode; - int FBHardwareWriteMask; - int MaterialMode; - int NormalizeMode; - int LightingMode; - int Light0Mode; - int Light1Mode; - int Light2Mode; - int Light3Mode; - int Light4Mode; - int Light5Mode; - int Light6Mode; - int Light7Mode; - int Light8Mode; - int Light9Mode; - int Light10Mode; - int Light11Mode; - int Light12Mode; - int Light13Mode; - int Light14Mode; - int Light15Mode; - int LogicalOpMode; - int ScissorMode; - int ScissorMaxXY; - int ScissorMinXY; - int Window; /* GID part probably should be in draw priv */ - int WindowOrigin; - int x, y, w, h; /* Probably should be in drawable priv */ - int FrameCount; /* Probably should be in drawable priv */ - int NotClipped; /* Probably should be in drawable priv */ - int WindowChanged; /* Probably should be in drawabl... */ - int Flags; - int EnabledFlags; - int DepthSize; - int Begin; - GLenum ErrorValue; - int Texture1DEnabled; - int Texture2DEnabled; - - float ModelView[16]; - float Proj[16]; - float ModelViewProj[16]; - float Texture[16]; - - float ModelViewStack[(MAX_MODELVIEW_STACK-1)*16]; - int ModelViewCount; - float ProjStack[(MAX_PROJECTION_STACK-1)*16]; - int ProjCount; - float TextureStack[(MAX_TEXTURE_STACK-1)*16]; - int TextureCount; -}; - -#define S3VIRGEPACKCOLOR555( r, g, b, a ) \ - ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) | \ - ((a) ? 0x8000 : 0)) - -#define S3VIRGEPACKCOLOR565( r, g, b ) \ - ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3)) - -#define S3VIRGEPACKCOLOR888( r, g, b ) \ - (((r) << 16) | ((g) << 8) | (b)) - -#define S3VIRGEPACKCOLOR8888( r, g, b, a ) \ - (((a) << 24) | ((r) << 16) | ((g) << 8) | (b)) - -#define S3VIRGEPACKCOLOR4444( r, g, b, a ) \ - ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4)) - -static INLINE GLuint s3vPackColor( GLuint cpp, - GLubyte r, GLubyte g, - GLubyte b, GLubyte a ) -{ - unsigned int ret; - DEBUG(("cpp = %i, r=0x%x, g=0x%x, b=0x%x, a=0x%x\n", cpp, r, g, b, a)); - - switch ( cpp ) { - case 2: - ret = S3VIRGEPACKCOLOR555( r, g, b, a ); - DEBUG(("ret = 0x%x\n", ret)); - return ret; - case 4: - return PACK_COLOR_8888( a, r, g, b ); - default: - return 0; - } -} - -#define S3V_CONTEXT(ctx) ((s3vContextPtr)(ctx->DriverCtx)) - -#endif /* _S3V_CONTEXT_H_ */ diff --git a/src/mesa/drivers/dri/s3v/s3v_dd.c b/src/mesa/drivers/dri/s3v/s3v_dd.c deleted file mode 100644 index e340116f5e1..00000000000 --- a/src/mesa/drivers/dri/s3v/s3v_dd.c +++ /dev/null @@ -1,90 +0,0 @@ -/* - * Author: Max Lingua <[email protected]> - */ - -#include "s3v_context.h" -#include "s3v_vb.h" -#include "s3v_lock.h" -#if defined(USE_X86_ASM) -#include "x86/common_x86_asm.h" -#endif - -#include "main/context.h" -#include "main/framebuffer.h" -#include "swrast/swrast.h" - -#define S3V_DATE "20020207" - - -/* Return the width and height of the current color buffer. - */ -static void s3vDDGetBufferSize( GLframebuffer *buffer, - GLuint *width, GLuint *height ) -{ - GET_CURRENT_CONTEXT(ctx); - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - -/* S3VHW_LOCK( vmesa ); */ - *width = vmesa->driDrawable->w; - *height = vmesa->driDrawable->h; -/* S3VHW_UNLOCK( vmesa ); */ -} - - -/* Return various strings for glGetString(). - */ -static const GLubyte *s3vDDGetString( GLcontext *ctx, GLenum name ) -{ - static char buffer[128]; - - switch ( name ) { - case GL_VENDOR: - return (GLubyte *)"Max Lingua (ladybug)"; - - case GL_RENDERER: - sprintf( buffer, "Mesa DRI S3 Virge " S3V_DATE ); - - /* Append any CPU-specific information. - */ -#ifdef USE_X86_ASM - if ( _mesa_x86_cpu_features ) { - strncat( buffer, " x86", 4 ); - -} -#ifdef USE_MMX_ASM - if ( cpu_has_mmx ) { - strncat( buffer, "/MMX", 4 ); - } -#endif -#ifdef USE_3DNOW_ASM - if ( cpu_has_3dnow ) { - strncat( buffer, "/3DNow!", 7 ); - } -#endif -#ifdef USE_SSE_ASM - if ( cpu_has_xmm ) { - strncat( buffer, "/SSE", 4 ); - } -#endif -#endif - return (GLubyte *)buffer; - - default: - return NULL; - } -} - -/* Enable the extensions supported by this driver. - */ -void s3vInitExtensions( GLcontext *ctx ) -{ - /* None... */ -} - -/* Initialize the driver's misc functions. - */ -void s3vInitDriverFuncs( GLcontext *ctx ) -{ - ctx->Driver.GetBufferSize = s3vDDGetBufferSize; - ctx->Driver.GetString = s3vDDGetString; -} diff --git a/src/mesa/drivers/dri/s3v/s3v_dri.h b/src/mesa/drivers/dri/s3v/s3v_dri.h deleted file mode 100644 index 339c579f7f7..00000000000 --- a/src/mesa/drivers/dri/s3v/s3v_dri.h +++ /dev/null @@ -1,143 +0,0 @@ -/* - * Author: Max Lingua <[email protected]> - */ - -#ifndef _S3V_DRI -#define _S3V_DRI - -#include "s3v_common.h" - -#define S3V_MAX_DRAWABLES (S3V_DMA_BUF_NR/2) /* 32 */ /* 256 */ /* FIXME */ - -typedef struct -{ - int deviceID; - int width; - int height; - int mem; - int cpp; - int bitsPerPixel; - - int fbOffset; - int fbStride; - - int logTextureGranularity; - int textureOffset; - - drm_handle_t regs; - drmSize regsSize; - - unsigned int sarea_priv_offset; -/* - drmAddress regsMap; - - drmSize textureSize; - drm_handle_t textures; -*/ - -#if 0 - drm_handle_t agp_buffers; - drmSize agp_buf_size; -#endif - -/* - drmBufMapPtr drmBufs; - int irq; - unsigned int sarea_priv_offset; -*/ - -/* FIXME: cleanup ! */ - - drmSize registerSize; /* == S3V_MMIO_REGSIZE */ - drm_handle_t registerHandle; - - drmSize pciSize; - drm_handle_t pciMemHandle; - - drmSize frontSize; /* == videoRambytes */ -/* drm_handle_t frontHandle; */ - unsigned long frontOffset; /* == fbOffset */ - int frontPitch; -/* unsigned char *front; */ - - unsigned int bufferSize; /* size of depth/back buffer */ - - drmSize backSize; -/* drm_handle_t backHandle; */ - unsigned long backOffset; - int backPitch; -/* unsigned char *back; */ - - drmSize depthSize; -/* drm_handle_t depthHandle; */ - unsigned long depthOffset; - int depthPitch; -/* unsigned char *depth; */ - - drmSize texSize; -/* drm_handle_t texHandle; */ - unsigned long texOffset; - int texPitch; -/* unsigned char *tex; */ - - drmSize dmaBufSize; /* Size of buffers (in bytes) */ - drm_handle_t dmaBufHandle; /* Handle from drmAddMap */ - unsigned long dmaBufOffset; /* Offset/Start */ - int dmaBufPitch; /* Pitch */ - unsigned char *dmaBuf; /* Map */ - int bufNumBufs; /* Number of buffers */ - drmBufMapPtr buffers; /* Buffer map */ - -} S3VDRIRec, *S3VDRIPtr; - -/* WARNING: Do not change the SAREA structure without changing the kernel - * as well */ - -typedef struct { - unsigned char next, prev; /* indices to form a circular LRU */ - unsigned char in_use; /* owned by a client, or free? */ - int age; /* tracked by clients to update local LRU's */ -} S3VTexRegionRec, *S3VTexRegionPtr; - -typedef struct { - unsigned int nbox; - drm_clip_rect_t boxes[S3V_NR_SAREA_CLIPRECTS]; - - /* Maintain an LRU of contiguous regions of texture space. If - * you think you own a region of texture memory, and it has an - * age different to the one you set, then you are mistaken and - * it has been stolen by another client. If global texAge - * hasn't changed, there is no need to walk the list. - * - * These regions can be used as a proxy for the fine-grained - * texture information of other clients - by maintaining them - * in the same lru which is used to age their own textures, - * clients have an approximate lru for the whole of global - * texture space, and can make informed decisions as to which - * areas to kick out. There is no need to choose whether to - * kick out your own texture or someone else's - simply eject - * them all in LRU order. - */ - S3VTexRegionRec texList[S3V_NR_TEX_REGIONS+1]; /* Last elt is sentinal */ - - int texAge; /* last time texture was uploaded */ - - int last_enqueue; /* last time a buffer was enqueued */ - int last_dispatch; /* age of the most recently dispatched buffer */ - int last_quiescent; /* */ - - int ctxOwner; /* last context to upload state */ -} S3VSAREARec, *S3VSAREAPtr; - -typedef struct { - /* Nothing here yet */ - int dummy; -} S3VConfigPrivRec, *S3VConfigPrivPtr; - -typedef struct { - /* Nothing here yet */ - int dummy; -} S3VDRIContextRec, *S3VDRIContextPtr; - - -#endif diff --git a/src/mesa/drivers/dri/s3v/s3v_inithw.c b/src/mesa/drivers/dri/s3v/s3v_inithw.c deleted file mode 100644 index bdc9effb799..00000000000 --- a/src/mesa/drivers/dri/s3v/s3v_inithw.c +++ /dev/null @@ -1,79 +0,0 @@ -/* - * Author: Max Lingua <[email protected]> - */ - -#include <sys/ioctl.h> - -#include "s3v_context.h" - -void s3vInitHW( s3vContextPtr vmesa ) -{ - int i; - static short _reset = 1; - - DEBUG(("vmesa->driDrawable = %p\n", vmesa->driDrawable)); - DEBUG(("stride = %i\n", - vmesa->driScreen->fbWidth*vmesa->s3vScreen->cpp)); - DEBUG(("frontOffset = 0x%x\n", vmesa->s3vScreen->frontOffset)); - DEBUG(("backOffset = 0x%x\n", vmesa->s3vScreen->backOffset)); - DEBUG(("depthOffset = 0x%x\n", vmesa->s3vScreen->depthOffset)); - DEBUG(("textureOffset = 0x%x\n", vmesa->s3vScreen->texOffset)); - -/* if (_reset) { */ -/* ioctl(vmesa->driFd, 0x4a); */ - ioctl(vmesa->driFd, 0x41); /* reset */ - _reset = 0; -/* ioctl(vmesa->driFd, 0x4c); */ -/* } */ - - /* FIXME */ - switch (vmesa->s3vScreen->cpp) { - case 2: - break; - case 4: - break; - } - - /* FIXME for stencil, gid, etc */ - switch (vmesa->DepthSize) { - case 15: - case 16: - break; - case 24: - break; - case 32: - break; - } - - vmesa->FogMode = 1; - vmesa->ClearDepth = 0xffff; - vmesa->x = 0; - vmesa->y = 0; - vmesa->w = 0; - vmesa->h = 0; - vmesa->FrameCount = 0; - vmesa->MatrixMode = GL_MODELVIEW; - vmesa->ModelViewCount = 0; - vmesa->ProjCount = 0; - vmesa->TextureCount = 0; - - - /* FIXME: do we need the following? */ - - for (i = 0; i < 16; i++) - if (i % 5 == 0) - vmesa->ModelView[i] = - vmesa->Proj[i] = - vmesa->ModelViewProj[i] = - vmesa->Texture[i] = 1.0; - else - vmesa->ModelView[i] = - vmesa->Proj[i] = - vmesa->ModelViewProj[i] = - vmesa->Texture[i] = 0.0; - - vmesa->LBWindowBase = vmesa->driScreen->fbWidth * - (vmesa->driScreen->fbHeight - 1); - vmesa->FBWindowBase = vmesa->driScreen->fbWidth * - (vmesa->driScreen->fbHeight - 1); -} diff --git a/src/mesa/drivers/dri/s3v/s3v_lock.c b/src/mesa/drivers/dri/s3v/s3v_lock.c deleted file mode 100644 index 52bb87ecec9..00000000000 --- a/src/mesa/drivers/dri/s3v/s3v_lock.c +++ /dev/null @@ -1,62 +0,0 @@ -/* - * Author: Max Lingua <[email protected]> - */ - -#include "s3v_context.h" - -#if DEBUG_LOCKING -char *prevLockFile = NULL; -int prevLockLine = 0; -#endif - - -/* Update the hardware state. This is called if another context has - * grabbed the hardware lock, which includes the X server. This - * function also updates the driver's window state after the X server - * moves, resizes or restacks a window -- the change will be reflected - * in the drawable position and clip rects. Since the X server grabs - * the hardware lock when it changes the window state, this routine will - * automatically be called after such a change. - */ -void s3vGetLock( s3vContextPtr vmesa, GLuint flags ) -{ - __DRIdrawablePrivate *dPriv = vmesa->driDrawable; -/* __DRIscreenPrivate *sPriv = vmesa->driScreen; */ - - printf("s3vGetLock <- ***\n"); - - drmGetLock( vmesa->driFd, vmesa->hHWContext, flags ); - - /* The window might have moved, so we might need to get new clip - * rects. - * - * NOTE: This releases and regrabs the hw lock to allow the X server - * to respond to the DRI protocol request for new drawable info. - * Since the hardware state depends on having the latest drawable - * clip rects, all state checking must be done _after_ this call. - */ - /* DRI_VALIDATE_DRAWABLE_INFO( vmesa->display, sPriv, dPriv ); */ - - if ( vmesa->lastStamp != dPriv->lastStamp ) { - vmesa->lastStamp = dPriv->lastStamp; - vmesa->new_state |= S3V_NEW_WINDOW | S3V_NEW_CLIP; - } - - vmesa->numClipRects = dPriv->numClipRects; - vmesa->pClipRects = dPriv->pClipRects; - -#if 0 - vmesa->dirty = ~0; - - if ( sarea->ctxOwner != vmesa->hHWContext ) { - sarea->ctxOwner = vmesa->hHWContext; - vmesa->dirty = S3V_UPLOAD_ALL; - } - - for ( i = 0 ; i < vmesa->lastTexHeap ; i++ ) { - if ( sarea->texAge[i] != vmesa->lastTexAge[i] ) { - s3vAgeTextures( vmesa, i ); - } - } -#endif -} diff --git a/src/mesa/drivers/dri/s3v/s3v_lock.h b/src/mesa/drivers/dri/s3v/s3v_lock.h deleted file mode 100644 index c39d24a38ac..00000000000 --- a/src/mesa/drivers/dri/s3v/s3v_lock.h +++ /dev/null @@ -1,99 +0,0 @@ -/* - * Author: Max Lingua <[email protected]> - */ - -#ifndef __S3V_LOCK_H__ -#define __S3V_LOCK_H__ - -#include <sys/ioctl.h> - -extern void s3vGetLock( s3vContextPtr vmesa, GLuint flags ); - -/* Turn DEBUG_LOCKING on to find locking conflicts. - */ -#define DEBUG_LOCKING 0 - -#if DEBUG_LOCKING -extern char *prevLockFile; -extern int prevLockLine; - -#define DEBUG_LOCK() \ - do { \ - prevLockFile = (__FILE__); \ - prevLockLine = (__LINE__); \ - } while (0) - -#define DEBUG_RESET() \ - do { \ - prevLockFile = 0; \ - prevLockLine = 0; \ - } while (0) - -#define DEBUG_CHECK_LOCK() \ - do { \ - if ( prevLockFile ) { \ - fprintf( stderr, \ - "LOCK SET!\n\tPrevious %s:%d\n\tCurrent: %s:%d\n", \ - prevLockFile, prevLockLine, __FILE__, __LINE__ ); \ - exit(1); \ - } \ - } while (0) - -#else - -#define DEBUG_LOCK() -#define DEBUG_RESET() -#define DEBUG_CHECK_LOCK() - -#endif - -/* - * !!! We may want to separate locks from locks with validation. This - * could be used to improve performance for those things commands that - * do not do any drawing !!! - */ - -/* Lock the hardware and validate our state. - */ -#define LOCK_HARDWARE( vmesa ) \ - do { \ - char __ret = 0; \ - DEBUG_CHECK_LOCK(); \ - DRM_CAS( vmesa->driHwLock, vmesa->hHWContext, \ - (DRM_LOCK_HELD | vmesa->hHWContext), __ret ); \ - if ( __ret ) \ - s3vGetLock( vmesa, 0 ); \ - DEBUG_LOCK(); \ - } while (0) - -/* Unlock the hardware. - */ -#define UNLOCK_HARDWARE( vmesa ) \ - do { \ - DRM_UNLOCK( vmesa->driFd, \ - vmesa->driHwLock, \ - vmesa->hHWContext ); \ - DEBUG_RESET(); \ - } while (0) - -#define S3VHW_LOCK( vmesa ) \ - DRM_UNLOCK(vmesa->driFd, vmesa->driHwLock, vmesa->hHWContext); \ - DRM_SPINLOCK(&vmesa->driScreen->pSAREA->drawable_lock, \ - vmesa->driScreen->drawLockID); \ - /* VALIDATE_DRAWABLE_INFO_NO_LOCK(vmesa); */ - -#define S3VHW_UNLOCK( vmesa ) \ - DRM_SPINUNLOCK(&vmesa->driScreen->pSAREA->drawable_lock, \ - vmesa->driScreen->drawLockID); \ - /* VALIDATE_DRAWABLE_INFO_NO_LOCK_POST(vmesa); */ - -#define S3V_SIMPLE_LOCK( vmesa ) \ - ioctl(vmesa->driFd, 0x4a) - -#define S3V_SIMPLE_FLUSH_LOCK( vmesa ) \ - ioctl(vmesa->driFd, 0x4b) - -#define S3V_SIMPLE_UNLOCK( vmesa ) \ - ioctl(vmesa->driFd, 0x4c) - -#endif /* __S3V_LOCK_H__ */ diff --git a/src/mesa/drivers/dri/s3v/s3v_macros.h b/src/mesa/drivers/dri/s3v/s3v_macros.h deleted file mode 100644 index 7e9b4529df4..00000000000 --- a/src/mesa/drivers/dri/s3v/s3v_macros.h +++ /dev/null @@ -1,230 +0,0 @@ -/* - * Author: Max Lingua <[email protected]> - */ - -#ifndef _S3V_MACROS_H_ -#define _S3V_MACROS_H_ - -/**************/ -/* DRI macros */ -/**************/ - -#define GENERIC_DEBUG 0 -#define FLOW_DEBUG 0 -#define DMABUFS_DEBUG 0 - -/* Note: The argument to DEBUG*() _must_ be enclosed in parenthesis */ - -#if (GENERIC_DEBUG || FLOW_DEBUG || DMABUFS_DEBUG) -#include <stdio.h> -#endif - -#undef DEBUG -#if GENERIC_DEBUG -#define DEBUG(str) printf str -#else -#define DEBUG(str) -#endif - -#if FLOW_DEBUG -#define DEBUG_WHERE(str) printf str -#else -#define DEBUG_WHERE(str) -#endif - -#if DMABUFS_DEBUG -#define DEBUG_BUFS(str) printf str -#else -#define DEBUG_BUFS(str) -#endif - - -#if 0 -#define S3V_DMA_SEND_FLAGS DRM_DMA_PRIORITY -#define S3V_DMA_SEND_FLAGS DRM_DMA_BLOCK -#else -#define S3V_DMA_SEND_FLAGS 0 -#endif - -#if 0 -#define S3V_DMA_GET_FLAGS \ - (DRM_DMA_SMALLER_OK | DRM_DMA_LARGER_OK | DRM_DMA_WAIT) -#else -#define S3V_DMA_GET_FLAGS DRM_DMA_WAIT -#endif - - -#define DMAOUT_CHECK(reg,len) \ -do { \ - DEBUG(("DMAOUT_CHECK: reg = 0x%x\n", S3V_##reg##_REG)); \ - DEBUG_BUFS(("DMAOUT_CHECK (was): ")); \ - DEBUG_BUFS(("vmesa->bufCount=%i of vmesa->bufSize=%i\n", \ - vmesa->bufCount, vmesa->bufSize)); \ - /* FIXME: > or >= */ \ - if (vmesa->bufCount+(len+1) >= vmesa->bufSize) \ - DMAFLUSH(); \ -\ - vmesa->bufCount += (len+1); \ - DEBUG_BUFS(("DMAOUT_CHECK (is): vmesa->bufCount=%i len=%i, reg=%x\n", \ - vmesa->bufCount, len, S3V_##reg##_REG)); \ - DMAOUT( ((len & 0xffff) | ((S3V_##reg##_REG & 0xfffc) << 14)) ); \ -} while (0) - -#define DMAOUT(val) \ -do { \ - *(vmesa->buf++)=val; \ - DEBUG_BUFS(("DMAOUT: val=0x%x\n", (unsigned int)val)); \ -} while(0) - -#define DMAFINISH() \ -do { \ - /* NOTE: it does nothing - it just prints some summary infos */ \ - DEBUG(("DMAFINISH: vmesa->bufCount=%i\n", vmesa->bufCount)); \ - DEBUG(("buf: index=%i; addr=%p\n", vmesa->bufIndex[vmesa->_bufNum], \ - vmesa->s3vScreen->bufs->list[vmesa->bufIndex[vmesa->_bufNum]].address)); \ -} while(0) - -#define DMAFLUSH() \ -do { \ - if (vmesa->bufCount) { \ - SEND_DMA(vmesa->driFd, vmesa->hHWContext, 1, \ - &vmesa->bufIndex[vmesa->_bufNum], &vmesa->bufCount); \ -/* - GET_DMA(vmesa->driFd, vmesa->hHWContext, 1, \ - &vmesa->bufIndex, &vmesa->bufSize); \ -*/ \ - vmesa->_bufNum = !(vmesa->_bufNum); \ - vmesa->buf = vmesa->_buf[vmesa->_bufNum]; \ -/* - vmesa->buf = \ - vmesa->s3vScreen->bufs->list[vmesa->bufIndex].address; \ -*/ \ - vmesa->bufCount = 0; \ - } \ -} while (0) - -#define CMDCHANGE() \ -do { \ - DMAOUT_CHECK(3DTRI_CMDSET, 1); /* FIXME: TRI/LINE */ \ - DMAOUT(vmesa->CMD); \ - DMAFINISH(); \ -} while (0) - -#ifdef DONT_SEND_DMA -#define GET_DMA(fd, hHWCtx, n, idx, size) -#define SEND_DMA(fd, hHWCtx,n, idx, cnt) -#else -#define GET_DMA(fd, hHWCtx, n, idx, size) \ -do { \ - drmDMAReq dma; \ - int retcode, i; \ -\ - DEBUG(("GET_DMA: ")); \ - DEBUG(("req_count=%i; req_list[#0]=%i; req_size[#0]=%i\n", \ - n, (idx)[n-1], (size)[n-1])); \ -\ - dma.context = (hHWCtx); \ - dma.send_count = 0; \ - dma.send_list = NULL; \ - dma.send_sizes = NULL; \ - dma.flags = S3V_DMA_GET_FLAGS; \ - dma.request_count = (n); \ - dma.request_size = S3V_DMA_BUF_SZ; \ - dma.request_list = (idx); \ - dma.request_sizes = (size); \ -\ - do { \ - if ((retcode = drmDMA((fd), &dma))) { \ - DEBUG_BUFS(("drmDMA (get) returned %d\n", retcode)); \ - } \ -} while (!(dma).granted_count); \ -\ - for (i = 0; i < (n); i++) { \ - DEBUG(("Got buffer %i (index #%i)\n", (idx)[i], i)); \ - DEBUG(("of %i bytes (%i words) size\n", \ - (size)[i], (size)[i] >>2)); \ - /* Convert from bytes to words */ \ - (size)[i] >>= 2; \ - } \ -} while (0) - -#define SEND_DMA(fd, hHWCtx, n, idx, cnt) \ -do { \ - drmDMAReq dma; \ - int retcode, i; \ -\ - DEBUG(("SEND_DMA: ")); \ - DEBUG(("send_count=%i; send_list[#0]=%i; send_sizes[#0]=%i\n", \ - n, (idx)[n-1], (cnt)[n-1])); \ -\ - for (i = 0; i < (n); i++) { \ - /* Convert from words to bytes */ \ - (cnt)[i] <<= 2; \ - } \ -\ - dma.context = (hHWCtx); \ - dma.send_count = (n); \ - dma.send_list = (idx); \ - dma.send_sizes = (cnt); \ - dma.flags = S3V_DMA_SEND_FLAGS; \ - dma.request_count = 0; \ - dma.request_size = 0; \ - dma.request_list = NULL; \ - dma.request_sizes = NULL; \ -\ - if ((retcode = drmDMA((fd), &dma))) { \ - DEBUG_BUFS(("drmDMA (send) returned %d\n", retcode)); \ - } \ -\ - for (i = 0; i < (n); i++) { \ - DEBUG(("Sent buffer %i (index #%i)\n", (idx)[i], i)); \ - DEBUG(("of %i bytes (%i words) size\n", \ - (cnt)[i], (cnt)[i] >>2)); \ - (cnt)[i] = 0; \ - } \ -} while (0) -#endif /* DONT_SEND_DMA */ - -#define GET_FIRST_DMA(fd, hHWCtx, n, idx, size, buf, cnt, vPriv) \ -do { \ - int i; \ - DEBUG_BUFS(("GET_FIRST_DMA\n")); \ - DEBUG_BUFS(("n=%i idx=%i size=%i\n", n, *idx, *size)); \ - DEBUG_BUFS(("going to GET_DMA\n")); \ - GET_DMA(fd, hHWCtx, n, idx, size); \ - DEBUG_BUFS(("coming from GET_DMA\n")); \ - DEBUG_BUFS(("n=%i idx=%i size=%i\n", n, (idx)[0], (size)[0])); \ - for (i = 0; i < (n); i++) { \ - DEBUG_BUFS(("buf #%i @%p\n", \ - i, (vPriv)->bufs->list[(idx)[i]].address)); \ - (buf)[i] = (vPriv)->bufs->list[(idx)[i]].address; \ - (cnt)[i] = 0; \ - } \ - DEBUG(("GOING HOME\n")); \ -} while (0) - -/**************************/ -/* generic, global macros */ -/**************************/ - -#define CALC_LOG2(l2,s) \ -do { \ - int __s = s; \ - l2 = 0; \ - while (__s > 1) { ++l2; __s >>= 1; } \ -} while (0) - -#define PrimType_Null 0x00000000 -#define PrimType_Points 0x10000000 -#define PrimType_Lines 0x20000000 -#define PrimType_LineLoop 0x30000000 -#define PrimType_LineStrip 0x40000000 -#define PrimType_Triangles 0x50000000 -#define PrimType_TriangleStrip 0x60000000 -#define PrimType_TriangleFan 0x70000000 -#define PrimType_Quads 0x80000000 -#define PrimType_QuadStrip 0x90000000 -#define PrimType_Polygon 0xa0000000 -#define PrimType_Mask 0xf0000000 - -#endif /* _S3V_MACROS_H_ */ diff --git a/src/mesa/drivers/dri/s3v/s3v_regs.h b/src/mesa/drivers/dri/s3v/s3v_regs.h deleted file mode 100644 index 26a7c54af5c..00000000000 --- a/src/mesa/drivers/dri/s3v/s3v_regs.h +++ /dev/null @@ -1,367 +0,0 @@ -/* - * Author: Max Lingua <[email protected]> - */ - -#ifndef _S3V_REG_H -#define _S3V_REG_H - -#define S3V_REGS_NUM 256 - -/************ - * DMA REGS * - ************/ - -#define S3V_DMA_ID 0 -#define S3V_DMA_REG 0x8590 -#define S3V_DMA_WRITEP_ID 1 -#define S3V_DMA_WRITEP_REG 0x8594 -#define S3V_DMA_READP_ID 2 -#define S3V_DMA_READP_REG 0x8598 -#define S3V_DMA_ENABLE_ID 3 -#define S3V_DMA_ENABLE_REG 0x859C -#define S3V_DMA_UPDATE_ID 4 -#define S3V_DMA_UPDATE_REG 0x10000 - -/*************** - * STATUS REGS * - ***************/ - -#define S3V_STAT_ID 10 -#define S3V_STAT_REG 0x8504 -#define S3V_STAT_VSYNC_ID 11 -#define S3V_STAT_VSYNC_REG 0x8505 -#define S3V_STAT_3D_DONE_ID 12 -#define S3V_STAT_3D_DONE_REG 0x8506 -#define S3V_STAT_FIFO_OVER_ID 13 -#define S3V_STAT_FIFO_OVER_REG 0x8508 -#define S3V_STAT_FIFO_EMPTY_ID 14 -#define S3V_STAT_FIFO_EMPTY_REG 0x850C -#define S3V_STAT_HDMA_DONE_ID 15 -#define S3V_STAT_HDMA_DONE_REG 0x8514 -#define S3V_STAT_CDMA_DONE_ID 16 -#define S3V_STAT_CDMA_DONE_REG 0x8524 -#define S3V_STAT_3D_FIFO_EMPTY_ID 17 -#define S3V_STAT_3D_FIFO_EMPTY_REG 0x8544 -#define S3V_STAT_LPB_ID 18 -#define S3V_STAT_LPB_REG 0x8584 -#define S3V_STAT_3D_BUSY_ID 19 -#define S3V_STAT_3D_BUSY_REG 0x8704 - -/*********** - * 2D REGS * - ***********/ - -#define S3V_BITBLT_ID 30 -#define S3V_BITBLT_REG 0xA400 -#define S3V_BITBLT_SRC_BASE_ID 31 -#define S3V_BITBLT_SRC_BASE_REG 0xA4D4 -#define S3V_BITBLT_DEST_BASE_ID 32 -#define S3V_BITBLT_DEST_BASE_REG 0xA4D8 -#define S3V_BITBLT_CLIP_L_R_ID 33 -#define S3V_BITBLT_CLIP_L_R_REG 0xA4DC -#define S3V_BITBLT_CLIP_T_B_ID 34 -#define S3V_BITBLT_CLIP_T_B_REG 0xA4E0 -#define S3V_BITBLT_DEST_SRC_STRIDE_ID 35 -#define S3V_BITBLT_DEST_SRC_STRIDE_REG 0xA4E4 -#define S3V_BITBLT_MONO_PAT0_ID 36 -#define S3V_BITBLT_MONO_PAT0_REG 0xA4E8 -#define S3V_BITBLT_MONO_PAT1_ID 37 -#define S3V_BITBLT_MONO_PAT1_REG 0xA4EC -#define S3V_BITBLT_PAT_BG_COLOR_ID 38 -#define S3V_BITBLT_PAT_BG_COLOR_REG 0xA4F0 -#define S3V_BITBLT_PAT_FG_COLOR_ID 39 -#define S3V_BITBLT_PAT_FG_COLOR_REG 0xA4F4 -#define S3V_BITBLT_CMDSET_ID 40 -#define S3V_BITBLT_CMDSET_REG 0xA500 -#define S3V_BITBLT_WIDTH_HEIGHT_ID 41 -#define S3V_BITBLT_WIDTH_HEIGHT_REG 0xA504 -#define S3V_BITBLT_SRC_X_Y_ID 42 -#define S3V_BITBLT_SRC_X_Y_REG 0xA508 -#define S3V_BITBLT_DEST_X_Y_ID 43 -#define S3V_BITBLT_DEST_X_Y_REG 0xA50C -#define S3V_2DLINE_ID 44 -#define S3V_2DLINE_REG 0xA800 -#define S3V_2DPOLY_ID 45 -#define S3V_2DPOLY_REG 0xAC00 - -/*************** - * 3DLINE REGS * - ***************/ -/* base regs */ -#define S3V_3DLINE_ID 50 -#define S3V_3DLINE_REG 0xB000 -#define S3V_3DLINE_Z_BASE_ID 51 -#define S3V_3DLINE_Z_BASE_REG 0xB0D4 -#define S3V_3DLINE_SRC_BASE_ID 52 /* it is the same reg */ -#define S3V_3DLINE_SRC_BASE_REG 0xB0D4 -#define S3V_3DLINE_DEST_BASE_ID 53 -#define S3V_3DLINE_DEST_BASE_REG 0xB0D8 -#define S3V_3DLINE_CLIP_L_R_ID 54 -#define S3V_3DLINE_CLIP_L_R_REG 0xB0DC -#define S3V_3DLINE_CLIP_T_B_ID 55 -#define S3V_3DLINE_CLIP_T_B_REG 0xB0E0 -#define S3V_3DLINE_DEST_SRC_STRIDE_ID 56 -#define S3V_3DLINE_DEST_SRC_STRIDE_REG 0xB0E4 -#define S3V_3DLINE_Z_STRIDE_ID 57 -#define S3V_3DLINE_Z_STRIDE_REG 0xB0E8 -#define S3V_3DLINE_TEX_BASE_ID 58 -#define S3V_3DLINE_TEX_BASE_REG 0xB0EC -#define S3V_3DLINE_TEX_B_COLOR_ID 59 -#define S3V_3DLINE_TEX_B_COLOR_REG 0xB0F0 -#define S3V_3DLINE_FOG_COLOR_ID 60 -#define S3V_3DLINE_FOG_COLOR_REG 0xB0F4 -#define S3V_3DLINE_COLOR0_ID 61 -#define S3V_3DLINE_COLOR0_REG 0xB0F8 -#define S3V_3DLINE_COLOR1_ID 62 -#define S3V_3DLINE_COLOR1_REG 0xB0FC -#define S3V_3DLINE_CMDSET_ID 63 -#define S3V_3DLINE_CMDSET_REG 0xB100 /* special */ -/* tex regs */ -/* FIXME: shouldn't it be a 1D tex for lines? */ -#define S3V_3DLINE_BASEV_ID 64 -#define S3V_3DLINE_BASEV_REG 0xB104 -#define S3V_3DLINE_BASEU_ID 65 -#define S3V_3DLINE_BASEU_REG 0xB108 -#define S3V_3DLINE_WXD_ID 66 -#define S3V_3DLINE_WXD_REG 0xB10C -#define S3V_3DLINE_WYD_ID 67 -#define S3V_3DLINE_WYD_REG 0xB110 -#define S3V_3DLINE_WSTART_ID 68 -#define S3V_3DLINE_WSTART_REG 0xB114 -#define S3V_3DLINE_DXD_ID 69 -#define S3V_3DLINE_DXD_REG 0xB118 -#define S3V_3DLINE_VXD_ID 70 -#define S3V_3DLINE_VXD_REG 0xB11C -#define S3V_3DLINE_UXD_ID 71 -#define S3V_3DLINE_UXD_REG 0xB120 -#define S3V_3DLINE_DYD_ID 72 -#define S3V_3DLINE_DYD_REG 0xB124 -#define S3V_3DLINE_VYD_ID 73 -#define S3V_3DLINE_VYD_REG 0xB128 -#define S3V_3DLINE_UYD_ID 74 -#define S3V_3DLINE_UYD_REG 0xB12C -#define S3V_3DLINE_DSTART_ID 75 -#define S3V_3DLINE_DSTART_REG 0xB130 -#define S3V_3DLINE_VSTART_ID 76 -#define S3V_3DLINE_VSTART_REG 0xB134 -#define S3V_3DLINE_USTART_ID 77 -#define S3V_3DLINE_USTART_REG 0xB138 -/* gourad regs */ -#define S3V_3DLINE_GBD_ID 78 -#define S3V_3DLINE_GBD_REG 0xB144 -#define S3V_3DLINE_ARD_ID 79 -#define S3V_3DLINE_ARD_REG 0xB148 -#define S3V_3DLINE_GS_BS_ID 80 -#define S3V_3DLINE_GS_BS_REG 0xB14C -#define S3V_3DLINE_AS_RS_ID 81 -#define S3V_3DLINE_AS_RS_REG 0xB150 -/* vertex regs */ -#define S3V_3DLINE_DZ_ID 82 -#define S3V_3DLINE_DZ_REG 0xB158 -#define S3V_3DLINE_ZSTART_ID 83 -#define S3V_3DLINE_ZSTART_REG 0xB15C -#define S3V_3DLINE_XEND0_END1_ID 84 -#define S3V_3DLINE_XEND0_END1_REG 0xB16C -#define S3V_3DLINE_DX_ID 85 -#define S3V_3DLINE_DX_REG 0xB170 -#define S3V_3DLINE_XSTART_ID 86 -#define S3V_3DLINE_XSTART_REG 0xB174 -#define S3V_3DLINE_YSTART_ID 87 -#define S3V_3DLINE_YSTART_REG 0xB178 -#define S3V_3DLINE_YCNT_ID 88 -#define S3V_3DLINE_YCNT_REG 0xB17C - -/************** - * 3DTRI REGS * - **************/ -/* base regs */ -#define S3V_3DTRI_ID 100 -#define S3V_3DTRI_REG 0xB400 -#define S3V_3DTRI_Z_BASE_ID 101 -#define S3V_3DTRI_Z_BASE_REG 0xB4D4 -#define S3V_3DTRI_SRC_BASE_ID 102 /* it is the same reg */ -#define S3V_3DTRI_SRC_BASE_REG 0xB4D4 -#define S3V_3DTRI_DEST_BASE_ID 103 -#define S3V_3DTRI_DEST_BASE_REG 0xB4D8 -#define S3V_3DTRI_CLIP_L_R_ID 104 -#define S3V_3DTRI_CLIP_L_R_REG 0xB4DC -#define S3V_3DTRI_CLIP_T_B_ID 105 -#define S3V_3DTRI_CLIP_T_B_REG 0xB4E0 -#define S3V_3DTRI_DEST_SRC_STRIDE_ID 106 -#define S3V_3DTRI_DEST_SRC_STRIDE_REG 0xB4E4 -#define S3V_3DTRI_Z_STRIDE_ID 107 -#define S3V_3DTRI_Z_STRIDE_REG 0xB4E8 -#define S3V_3DTRI_TEX_BASE_ID 108 -#define S3V_3DTRI_TEX_BASE_REG 0xB4EC -#define S3V_3DTRI_TEX_B_COLOR_ID 109 -#define S3V_3DTRI_TEX_B_COLOR_REG 0xB4F0 -#define S3V_3DTRI_FOG_COLOR_ID 110 -#define S3V_3DTRI_FOG_COLOR_REG 0xB4F4 -#define S3V_3DTRI_COLOR0_ID 111 -#define S3V_3DTRI_COLOR0_REG 0xB4F8 -#define S3V_3DTRI_COLOR1_ID 112 -#define S3V_3DTRI_COLOR1_REG 0xB4FC -#define S3V_3DTRI_CMDSET_ID 113 /* special */ -#define S3V_3DTRI_CMDSET_REG 0xB500 -/* tex regs */ -#define S3V_3DTRI_BASEV_ID 114 -#define S3V_3DTRI_BASEV_REG 0xB504 -#define S3V_3DTRI_BASEU_ID 115 -#define S3V_3DTRI_BASEU_REG 0xB508 -#define S3V_3DTRI_WXD_ID 116 -#define S3V_3DTRI_WXD_REG 0xB50C -#define S3V_3DTRI_WYD_ID 117 -#define S3V_3DTRI_WYD_REG 0xB510 -#define S3V_3DTRI_WSTART_ID 118 -#define S3V_3DTRI_WSTART_REG 0xB514 -#define S3V_3DTRI_DXD_ID 119 -#define S3V_3DTRI_DXD_REG 0xB518 -#define S3V_3DTRI_VXD_ID 120 -#define S3V_3DTRI_VXD_REG 0xB51C -#define S3V_3DTRI_UXD_ID 121 -#define S3V_3DTRI_UXD_REG 0xB520 -#define S3V_3DTRI_DYD_ID 122 -#define S3V_3DTRI_DYD_REG 0xB524 -#define S3V_3DTRI_VYD_ID 123 -#define S3V_3DTRI_VYD_REG 0xB528 -#define S3V_3DTRI_UYD_ID 124 -#define S3V_3DTRI_UYD_REG 0xB52C -#define S3V_3DTRI_DSTART_ID 125 -#define S3V_3DTRI_DSTART_REG 0xB530 -#define S3V_3DTRI_VSTART_ID 126 -#define S3V_3DTRI_VSTART_REG 0xB534 -#define S3V_3DTRI_USTART_ID 127 -#define S3V_3DTRI_USTART_REG 0xB538 -/* gourad regs */ -#define S3V_3DTRI_GBX_ID 128 -#define S3V_3DTRI_GBX_REG 0xB53C -#define S3V_3DTRI_ARX_ID 129 -#define S3V_3DTRI_ARX_REG 0xB540 -#define S3V_3DTRI_GBY_ID 130 -#define S3V_3DTRI_GBY_REG 0xB544 -#define S3V_3DTRI_ARY_ID 131 -#define S3V_3DTRI_ARY_REG 0xB548 -#define S3V_3DTRI_GS_BS_ID 132 -#define S3V_3DTRI_GS_BS_REG 0xB54C -#define S3V_3DTRI_AS_RS_ID 133 -#define S3V_3DTRI_AS_RS_REG 0xB550 -/* vertex regs */ -#define S3V_3DTRI_ZXD_ID 134 -#define S3V_3DTRI_ZXD_REG 0xB554 -#define S3V_3DTRI_ZYD_ID 135 -#define S3V_3DTRI_ZYD_REG 0xB558 -#define S3V_3DTRI_ZSTART_ID 136 -#define S3V_3DTRI_ZSTART_REG 0xB55C -#define S3V_3DTRI_TXDELTA12_ID 137 -#define S3V_3DTRI_TXDELTA12_REG 0xB560 -#define S3V_3DTRI_TXEND12_ID 138 -#define S3V_3DTRI_TXEND12_REG 0xB564 -#define S3V_3DTRI_TXDELTA01_ID 139 -#define S3V_3DTRI_TXDELTA01_REG 0xB568 -#define S3V_3DTRI_TXEND01_ID 140 -#define S3V_3DTRI_TXEND01_REG 0xB56C -#define S3V_3DTRI_TXDELTA02_ID 141 -#define S3V_3DTRI_TXDELTA02_REG 0xB570 -#define S3V_3DTRI_TXSTART02_ID 142 -#define S3V_3DTRI_TXSTART02_REG 0xB574 -#define S3V_3DTRI_TYS_ID 143 -#define S3V_3DTRI_TYS_REG 0xB578 -#define S3V_3DTRI_TY01_Y12_ID 144 -#define S3V_3DTRI_TY01_Y12_REG 0xB57C - -/* COMMANDS (to 0xB100 [lines] or 0xB500 [tris]) */ - -/* Auto execute */ -#define AUTO_EXEC_MASK 0x00000001 -#define AUTO_EXEC_OFF (0x0) -#define AUTO_EXEC_ON (0x1) -/* HW clipping */ -#define HW_CLIP_MASK 0x00000002 -#define HW_CLIP_OFF (0x0 << 1) -#define HW_CLIP_ON (0x1 << 1) -/* Destination color */ -#define DEST_COL_MASK 0x0000001c -#define DEST_COL_PAL (0x0 << 2) /* 8 bpp - palettized */ -#define DEST_COL_1555 (0x1 << 2) /* 16 bpp - ZRGB */ -#define DEST_COL_888 (0x2 << 2) /* 24 bpp - RGB */ -/* Texture color */ -#define TEX_COL_MASK 0x000000e0 -#define TEX_COL_ARGB8888 (0x0 << 5) /* 32 bpp - ARGB */ -#define TEX_COL_ARGB4444 (0x1 << 5) /* 16 bpp - ARGB */ -#define TEX_COL_ARGB1555 (0x2 << 5) /* 16 bpp - ARGB */ -#define TEX_COL_ALPHA4 (0x3 << 5) /* 8 bpp - ALPHA4 */ -#define TEX_COL_BLEND4_LOW (0x4 << 5) /* 4 bpp - BLEND4 low nibble */ -#define TEX_COL_BLEND4_HIGH (0x5 << 5) /* 4 bpp - BLEND4 high nibble */ -#define TEX_COL_PAL (0x6 << 5) /* 8 bpp - palettized */ -#define TEX_COL_YUV (0x7 << 5) /* 16 bpp - YUV */ -/* Mipmap level */ -#define MIP_MASK 0x00000f00 -#define MIPMAP_LEVEL(s) (s << 8) /* 8 -> 11 bits */ -/* Texture filtering */ -#define TEX_FILTER_MASK 0x00007000 -#define MIP_NEAREST (0x0 << 12) -#define LINEAR_MIP_NEAREST (0x1 << 12) -#define MIP_LINEAR (0x2 << 12) -#define LINEAR_MIP_LINEAR (0x3 << 12) -#define NEAREST (0x4 << 12) -#define FAST_BILINEAR (0x5 << 12) -#define LINEAR (0x6 << 12) -/* Texture blending */ -#define TEX_BLEND_MAKS 0x00018000 -#define TEX_REFLECT (0x0 << 15) -#define TEX_MODULATE (0x1 << 15) -#define TEX_DECAL (0x2 << 15) -/* Fog */ -#define FOG_MASK 0x00020000 -#define FOG_OFF (0x0 << 17) -#define FOG_ON (0x1 << 17) -/* Alpha blending */ -#define ALPHA_BLEND_MASK 0x000c0000 -#define ALPHA_OFF (0x0 << 18) | (0x0 << 19) -#define ALPHA_TEX (0x2 << 18) -#define ALPHA_SRC (0x3 << 18) -/* Depth compare mode */ -#define Z_MODE_MASK 0x00700000 -#define Z_NEVER (0x0 << 20) -#define Z_GREATER (0x1 << 20) -#define Z_EQUAL (0x2 << 20) -#define Z_GEQUAL (0x3 << 20) -#define Z_LESS (0x4 << 20) -#define Z_NOTEQUAL (0x5 << 20) -#define Z_LEQUAL (0x6 << 20) -#define Z_ALWAYS (0x7 << 20) -/* Depth update */ -#define Z_UPDATE_MASK 0x00800000 -#define Z_UPDATE_OFF (0x0 << 23) /* disable z update */ -#define Z_UPDATE_ON (0x1 << 23) -/* Depth buffering mode */ -#define Z_BUFFER_MASK 0x03000000 -#define Z_BUFFER (0x0 << 24) | (0x0 << 25) -#define Z_MUX_BUF (0x1 << 24) | (0x0 << 25) -#define Z_MUX_DRAW (0x2 << 24) -#define Z_OFF (0x3 << 24) /* no z buffering */ -/* Texture wrapping */ -#define TEX_WRAP_MASK 0x04000000 -#define TEX_WRAP_OFF (0x0 << 26) -#define TEX_WRAP_ON (0x1 << 26) -/* 3d command */ -#define DO_MASK 0x78000000 -#define DO_GOURAUD_TRI (0x0 << 27) -#define DO_TEX_LIT_TRI_OLD (0x1 << 27) -#define DO_TEX_UNLIT_TRI_OLD (0x2 << 27) -#define DO_TEX_LIT_TRI (0x5 << 27) -#define DO_TEX_UNLIT_TRI (0x6 << 27) -#define DO_3D_LINE (0x8 << 27) -#define DO_NOP (0xf << 27) /* turn on autoexec */ -/* status */ -#define CMD_MASK 0x80000000 -#define CMD_2D (0x0 << 31) /* execute a 2d cmd */ -#define CMD_3D (0x1 << 31) /* execute a 3d cmd */ - -/* global masks */ -#define TEX_MASK ( TEX_COL_MASK | TEX_WRAP_MASK | MIP_MASK \ - | TEX_FILTER_MASK | TEX_BLEND_MAKS \ - | TEX_WRAP_MASK ) -#define Z_MASK ( Z_MODE_MASK | Z_UPDATE_MASK | Z_BUFFER_MASK ) - -#endif /* _S3V_REG_H */ diff --git a/src/mesa/drivers/dri/s3v/s3v_render.c b/src/mesa/drivers/dri/s3v/s3v_render.c deleted file mode 100644 index 5023f3c4641..00000000000 --- a/src/mesa/drivers/dri/s3v/s3v_render.c +++ /dev/null @@ -1,203 +0,0 @@ -/* - * Author: Max Lingua <[email protected]> - */ - -#include "main/glheader.h" -#include "main/context.h" -#include "main/macros.h" -#include "main/mtypes.h" - -#include "tnl/t_context.h" - -#include "s3v_context.h" -#include "s3v_tris.h" -#include "s3v_vb.h" - - -#define HAVE_POINTS 0 -#define HAVE_LINES 0 -#define HAVE_LINE_STRIPS 0 -#define HAVE_TRIANGLES 0 -#define HAVE_TRI_STRIPS 0 -#define HAVE_TRI_STRIP_1 0 -#define HAVE_TRI_FANS 0 -#define HAVE_QUADS 0 -#define HAVE_QUAD_STRIPS 0 -#define HAVE_POLYGONS 0 - -#define HAVE_ELTS 0 - -#if 0 -static void VERT_FALLBACK( GLcontext *ctx, - GLuint start, - GLuint count, - GLuint flags ) -{ - TNLcontext *tnl = TNL_CONTEXT(ctx); -/* s3vContextPtr vmesa = S3V_CONTEXT(ctx); */ - int _flags; - - DEBUG(("VERT_FALLBACK: flags & PRIM_MODE_MASK = %i\n", - flags & PRIM_MODE_MASK)); - DEBUG(("VERT_FALLBACK: flags=%i PRIM_MODE_MASK=%i\n", - flags, PRIM_MODE_MASK)); -#if 0 - tnl->Driver.Render.PrimitiveNotify( ctx, flags & PRIM_MODE_MASK ); -#endif - tnl->Driver.Render.BuildVertices( ctx, start, count, ~0 ); - - _flags = flags & PRIM_MODE_MASK; - - tnl->Driver.Render.PrimTabVerts[_flags]( ctx, start, count, flags ); - S3V_CONTEXT(ctx)->SetupNewInputs = VERT_BIT_POS; -} -#endif - -static const GLuint hw_prim[GL_POLYGON+1] = { - PrimType_Points, - PrimType_Lines, - PrimType_LineLoop, - PrimType_LineStrip, - PrimType_Triangles, - PrimType_TriangleStrip, - PrimType_TriangleFan, - PrimType_Quads, - PrimType_QuadStrip, - PrimType_Polygon -}; - -static INLINE void s3vStartPrimitive( s3vContextPtr vmesa, GLenum prim ) -{ - __DRIdrawablePrivate *dPriv = vmesa->driDrawable; - - int _hw_prim = hw_prim[prim]; - - DEBUG(("s3vStartPrimitive (new #%i) ", prim)); - - if (_hw_prim != vmesa->restore_primitive) { - - if (prim == 4) { /* TRI */ - DEBUG(("switching to tri\n")); - vmesa->prim_cmd = vmesa->_tri[vmesa->_3d_mode]; - vmesa->alpha_cmd = vmesa->_alpha[vmesa->_3d_mode]; - DMAOUT_CHECK(3DTRI_Z_BASE, 12); - } else if (prim == 1) { /* LINE */ - DEBUG(("switching to line\n")); - vmesa->prim_cmd = DO_3D_LINE; - vmesa->alpha_cmd = vmesa->_alpha[0]; - DMAOUT_CHECK(3DLINE_Z_BASE, 12); - } else { - DEBUG(("Never mind the bollocks!\n")); - } - - DMAOUT(vmesa->s3vScreen->depthOffset & 0x003FFFF8); - DMAOUT(vmesa->DestBase); - /* DMAOUT(vmesa->ScissorLR); */ - /* DMAOUT(vmesa->ScissorTB); */ - DMAOUT( (0 << 16) | (dPriv->w-1) ); - DMAOUT( (0 << 16) | (dPriv->h-1) ); - DMAOUT( (vmesa->SrcStride << 16) | vmesa->TexStride ); - DMAOUT(vmesa->SrcStride); - DMAOUT(vmesa->TexOffset); - DMAOUT(vmesa->TextureBorderColor); - DMAOUT(0); /* FOG */ - DMAOUT(0); - DMAOUT(0); - DMAOUT(vmesa->CMD | vmesa->prim_cmd | vmesa->alpha_cmd); - DMAFINISH(); - } - - vmesa->restore_primitive = _hw_prim; -} - -static INLINE void s3vEndPrimitive( s3vContextPtr vmesa ) -{ -/* GLcontext *ctx = vmesa->glCtx; */ - DEBUG(("s3vEndPrimitive\n")); -} - -#define LOCAL_VARS s3vContextPtr vmesa = S3V_CONTEXT(ctx) -#define INIT( prim ) s3vStartPrimitive( vmesa, prim ) -#define FINISH s3vEndPrimitive( vmesa ) -#define NEW_PRIMITIVE() (void) vmesa -#define NEW_BUFFER() (void) vmesa -#define FIRE_VERTICES() (void) vmesa -#define GET_CURRENT_VB_MAX_VERTS() \ - (vmesa->bufSize - vmesa->bufCount) / 2 -#define GET_SUBSEQUENT_VB_MAX_VERTS() \ - S3V_DMA_BUF_SZ / 2 -/* XXX */ -#define ALLOC_VERTS(nr) NULL -#define EMIT_VERTS(ctx, start, count, buf) NULL -#define FLUSH() s3vEndPrimitive( vmesa ) - -#define TAG(x) s3v_##x - -#include "tnl_dd/t_dd_dmatmp.h" - -/**********************************************************************/ -/* Render pipeline stage */ -/**********************************************************************/ - - -static GLboolean s3v_run_render( GLcontext *ctx, - struct tnl_pipeline_stage *stage ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - TNLcontext *tnl = TNL_CONTEXT(ctx); - struct vertex_buffer *VB = &tnl->vb; - GLuint i; - tnl_render_func *tab; - - DEBUG(("s3v_run_render\n")); - - /* FIXME: hw clip */ - if (VB->ClipOrMask || vmesa->RenderIndex != 0) { - DEBUG(("*** CLIPPED in render ***\n")); -#if 1 - return GL_TRUE; /* don't handle clipping here */ -#endif - } - - - /* We don't do elts */ - if (VB->Elts) - return GL_TRUE; - - tab = TAG(render_tab_verts); - - tnl->Driver.Render.Start( ctx ); - - for (i = 0 ; i < VB->PrimitiveCount ; i++ ) - { - GLuint prim = _tnl_translate_prim(&VB->Primitive[i]); - GLuint start = VB->Primitive[i].start; - GLuint length = VB->Primitive[i].count; - - DEBUG(("s3v_run_render (loop=%i) (lenght=%i)\n", i, length)); - - if (length) { - tnl->Driver.Render.BuildVertices( ctx, start, - start+length, ~0 /*stage->inputs*/); /* XXX */ - tnl->Driver.Render.PrimTabVerts[prim & PRIM_MODE_MASK] - ( ctx, start, start + length, prim ); - vmesa->SetupNewInputs = VERT_BIT_POS; - } - } - - tnl->Driver.Render.Finish( ctx ); - - return GL_FALSE; /* finished the pipe */ -} - - - -const struct tnl_pipeline_stage _s3v_render_stage = -{ - "s3v render", - NULL, - NULL, - NULL, - NULL, - s3v_run_render /* run */ -}; diff --git a/src/mesa/drivers/dri/s3v/s3v_screen.c b/src/mesa/drivers/dri/s3v/s3v_screen.c deleted file mode 100644 index f1810597e6f..00000000000 --- a/src/mesa/drivers/dri/s3v/s3v_screen.c +++ /dev/null @@ -1,99 +0,0 @@ -/* - * Author: Max Lingua <[email protected]> - */ - -#include "s3v_context.h" -#include "s3v_vb.h" -#include "s3v_dri.h" - -s3vScreenPtr s3vCreateScreen( __DRIscreenPrivate *sPriv ) -{ - s3vScreenPtr s3vScreen; - S3VDRIPtr vDRIPriv = (S3VDRIPtr)sPriv->pDevPriv; - -/* int i; */ - - DEBUG(("s3vCreateScreen\n")); - DEBUG(("sPriv->pDevPriv at %p\n", sPriv->pDevPriv)); - DEBUG(("size = %i\n", sizeof(*vDRIPriv))); - - if (sPriv->devPrivSize != sizeof(S3VDRIRec)) { - fprintf(stderr,"\nERROR! sizeof(S3VDRIRec) does not match passed size from device driver\n"); - return GL_FALSE; - } - - /* Allocate the private area */ - s3vScreen = (s3vScreenPtr) CALLOC( sizeof(*s3vScreen) ); - if ( !s3vScreen ) return NULL; - - s3vScreen->regionCount = 4; /* Magic number. Can we fix this? */ - - s3vScreen->regions = _mesa_malloc(s3vScreen->regionCount * - sizeof(s3vRegion)); - DEBUG(("sPriv->fd = %i\nvDRIPriv->dmaBufHandle = %x\n", - sPriv->fd, vDRIPriv->dmaBufHandle)); - - DEBUG(("vDRIPriv->dmaBufSize=%i\nvDRIPriv->dmaBuf=%p\n", - vDRIPriv->dmaBufSize, vDRIPriv->dmaBuf)); - - - /* Get the list of dma buffers */ - s3vScreen->bufs = drmMapBufs(sPriv->fd); - - if (!s3vScreen->bufs) { - DEBUG(("Helter/skelter with drmMapBufs\n")); - return GL_FALSE; - } - - s3vScreen->textureSize = vDRIPriv->texSize; - s3vScreen->logTextureGranularity = vDRIPriv->logTextureGranularity; - s3vScreen->cpp = vDRIPriv->cpp; - s3vScreen->frontOffset = vDRIPriv->frontOffset; - s3vScreen->frontPitch = vDRIPriv->frontPitch; - s3vScreen->backOffset = vDRIPriv->backOffset; - s3vScreen->backPitch = vDRIPriv->frontPitch; /* FIXME: check */ - s3vScreen->depthOffset = vDRIPriv->depthOffset; - s3vScreen->depthPitch = vDRIPriv->frontPitch; - s3vScreen->texOffset = vDRIPriv->texOffset; - - s3vScreen->driScreen = sPriv; - - DEBUG(("vDRIPriv->width =%i; vDRIPriv->deviceID =%x\n", vDRIPriv->width, - vDRIPriv->deviceID)); - DEBUG(("vDRIPriv->mem =%i\n", vDRIPriv->mem)); - DEBUG(("vDRIPriv->fbOffset =%i\n", vDRIPriv->fbOffset)); - DEBUG((" ps3vDRI->fbStride =%i\n", vDRIPriv->fbStride)); - DEBUG(("s3vScreen->cpp = %i\n", s3vScreen->cpp)); - DEBUG(("s3vScreen->backOffset = %x\n", s3vScreen->backOffset)); - DEBUG(("s3vScreen->depthOffset = %x\n", s3vScreen->depthOffset)); - DEBUG(("s3vScreen->texOffset = %x\n", s3vScreen->texOffset)); - DEBUG(("I will return from s3vCreateScreen now\n")); - - DEBUG(("s3vScreen->bufs = 0x%x\n", s3vScreen->bufs)); - return s3vScreen; -} - -/* Destroy the device specific screen private data struct. - */ -void s3vDestroyScreen( __DRIscreenPrivate *sPriv ) -{ - s3vScreenPtr s3vScreen = (s3vScreenPtr)sPriv->private; - - DEBUG(("s3vDestroyScreen\n")); - - /* First, unmap the dma buffers */ -/* - drmUnmapBufs( s3vScreen->bufs ); -*/ - /* Next, unmap all the regions */ -/* while (s3vScreen->regionCount > 0) { - - (void)drmUnmap(s3vScreen->regions[s3vScreen->regionCount].map, - s3vScreen->regions[s3vScreen->regionCount].size); - s3vScreen->regionCount--; - - } - FREE(s3vScreen->regions); */ - if (s3vScreen) - FREE(s3vScreen); -} diff --git a/src/mesa/drivers/dri/s3v/s3v_screen.h b/src/mesa/drivers/dri/s3v/s3v_screen.h deleted file mode 100644 index c49bc8587da..00000000000 --- a/src/mesa/drivers/dri/s3v/s3v_screen.h +++ /dev/null @@ -1,39 +0,0 @@ -/* - * Author: Max Lingua <[email protected]> - */ - -#include "main/mtypes.h" - -typedef struct _s3vRegion { - drm_handle_t handle; - drmSize size; - drmAddress map; -} s3vRegion, *s3vRegionPtr; - -typedef struct { - - int regionCount; /* Count of register regions */ - s3vRegion *regions; /* Vector of mapped region info */ - - drmBufMapPtr bufs; /* Map of DMA buffers */ - - __DRIscreenPrivate *driScreen; /* Back pointer to DRI screen */ - - int cpp; - int frontPitch; - int frontOffset; - - int backPitch; - int backOffset; - int backX; - int backY; - - int depthOffset; - int depthPitch; - - int texOffset; - int textureOffset; - int textureSize; - int logTextureGranularity; -} s3vScreenRec, *s3vScreenPtr; - diff --git a/src/mesa/drivers/dri/s3v/s3v_span.c b/src/mesa/drivers/dri/s3v/s3v_span.c deleted file mode 100644 index f9f7c0d1eee..00000000000 --- a/src/mesa/drivers/dri/s3v/s3v_span.c +++ /dev/null @@ -1,219 +0,0 @@ -/* - * Author: Max Lingua <[email protected]> - */ - -#include "s3v_context.h" -#include "s3v_lock.h" - -#include "swrast/swrast.h" - -#define _SPANLOCK 1 -#define DBG 0 - -#define LOCAL_VARS \ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); \ - __DRIscreenPrivate *sPriv = vmesa->driScreen; \ - __DRIdrawablePrivate *dPriv = vmesa->driDrawable; \ - driRenderbuffer *drb = (driRenderbuffer *) rb; \ - GLuint cpp = drb->cpp; \ - GLuint pitch = ( (drb->backBuffer) ? \ - ((dPriv->w+31)&~31) * cpp \ - : sPriv->fbWidth * cpp); \ - GLuint height = dPriv->h; \ - char *buf = (char *)(sPriv->pFB + drb->offset \ - + (drb->backBuffer ? 0 : dPriv->x * cpp + dPriv->y * pitch));\ - GLuint p; \ - (void) p - -/* FIXME! Depth/Stencil read/writes don't work ! */ -#define LOCAL_DEPTH_VARS \ - __DRIdrawablePrivate *dPriv = vmesa->driDrawable; \ - __DRIscreenPrivate *sPriv = vmesa->driScreen; \ - driRenderbuffer *drb = (driRenderbuffer *) rb; \ - GLuint pitch = drb->pitch; \ - GLuint height = dPriv->h; \ - char *buf = (char *)(sPriv->pFB + drb->offset); \ - (void) pitch - -#define LOCAL_STENCIL_VARS LOCAL_DEPTH_VARS - -#define Y_FLIP( _y ) (height - _y - 1) - -#if _SPANLOCK /* OK, we lock */ - -#define HW_LOCK() \ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); \ - (void) vmesa; \ - DMAFLUSH(); \ - S3V_SIMPLE_FLUSH_LOCK(vmesa); -#define HW_UNLOCK() S3V_SIMPLE_UNLOCK(vmesa); - -#else /* plz, don't lock */ - -#define HW_LOCK() \ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); \ - (void) vmesa; \ - DMAFLUSH(); -#define HW_UNLOCK() - -#endif - - -/* ================================================================ - * Color buffer - */ - -/* 16 bit, RGB565 color spanline and pixel functions - */ -#define INIT_MONO_PIXEL(p, color) \ - p = S3VIRGEPACKCOLOR555( color[0], color[1], color[2], color[3] ) - -#define WRITE_RGBA( _x, _y, r, g, b, a ) \ -do { \ - *(GLushort *)(buf + _x*2 + _y*pitch) = ((((int)r & 0xf8) << 7) | \ - (((int)g & 0xf8) << 2) | \ - (((int)b & 0xf8) >> 3)); \ - DEBUG(("buf=0x%x drawOffset=0x%x dPriv->x=%i drb->cpp=%i dPriv->y=%i pitch=%i\n", \ - sPriv->pFB, vmesa->drawOffset, dPriv->x, drb->cpp, dPriv->y, pitch)); \ - DEBUG(("dPriv->w = %i\n", dPriv->w)); \ -} while(0) - -#define WRITE_PIXEL( _x, _y, p ) \ - *(GLushort *)(buf + _x*2 + _y*pitch) = p - -#define READ_RGBA( rgba, _x, _y ) \ - do { \ - GLushort p = *(GLushort *)(buf + _x*2 + _y*pitch); \ - rgba[0] = (p >> 7) & 0xf8; \ - rgba[1] = (p >> 2) & 0xf8; \ - rgba[2] = (p << 3) & 0xf8; \ - rgba[3] = 0xff; /* - if ( rgba[0] & 0x08 ) rgba[0] |= 0x07; \ - if ( rgba[1] & 0x04 ) rgba[1] |= 0x03; \ - if ( rgba[2] & 0x08 ) rgba[2] |= 0x07; */ \ - } while (0) - -#define TAG(x) s3v##x##_RGB555 -#include "spantmp.h" - - -/* 32 bit, ARGB8888 color spanline and pixel functions - */ - -#undef INIT_MONO_PIXEL -#define INIT_MONO_PIXEL(p, color) \ - p = PACK_COLOR_8888( color[3], color[0], color[1], color[2] ) - -#define WRITE_RGBA( _x, _y, r, g, b, a ) \ - *(GLuint *)(buf + _x*4 + _y*pitch) = ((b << 0) | \ - (g << 8) | \ - (r << 16) | \ - (a << 24) ) - -#define WRITE_PIXEL( _x, _y, p ) \ - *(GLuint *)(buf + _x*4 + _y*pitch) = p - -#define READ_RGBA( rgba, _x, _y ) \ -do { \ - GLuint p = *(GLuint *)(buf + _x*4 + _y*pitch); \ - rgba[0] = (p >> 16) & 0xff; \ - rgba[1] = (p >> 8) & 0xff; \ - rgba[2] = (p >> 0) & 0xff; \ - rgba[3] = (p >> 24) & 0xff; \ -} while (0) - -#define TAG(x) s3v##x##_ARGB8888 -#include "spantmp.h" - - -/* 16 bit depthbuffer functions. - */ -#define VALUE_TYPE GLushort - -#define WRITE_DEPTH( _x, _y, d ) \ - *(GLushort *)(buf + _x*2 + _y*dPriv->w*2) = d - -#define READ_DEPTH( d, _x, _y ) \ - d = *(GLushort *)(buf + _x*2 + _y*dPriv->w*2); - -#define TAG(x) s3v##x##_z16 -#include "depthtmp.h" - - - - -/* 32 bit depthbuffer functions. - */ -#if 0 -#define VALUE_TYPE GLuint - -#define WRITE_DEPTH( _x, _y, d ) \ - *(GLuint *)(buf + _x*4 + _y*pitch) = d; - -#define READ_DEPTH( d, _x, _y ) \ - d = *(GLuint *)(buf + _x*4 + _y*pitch); - -#define TAG(x) s3v##x##_32 -#include "depthtmp.h" -#endif - - -/* 24/8 bit interleaved depth/stencil functions - */ -#if 0 -#define VALUE_TYPE GLuint - -#define WRITE_DEPTH( _x, _y, d ) { \ - GLuint tmp = *(GLuint *)(buf + _x*4 + _y*pitch); \ - tmp &= 0xff; \ - tmp |= (d) & 0xffffff00; \ - *(GLuint *)(buf + _x*4 + _y*pitch) = tmp; \ -} - -#define READ_DEPTH( d, _x, _y ) \ - d = *(GLuint *)(buf + _x*4 + _y*pitch) & ~0xff - - -#define TAG(x) s3v##x##_24_8 -#include "depthtmp.h" - -#define WRITE_STENCIL( _x, _y, d ) { \ - GLuint tmp = *(GLuint *)(buf + _x*4 + _y*pitch); \ - tmp &= 0xffffff00; \ - tmp |= d & 0xff; \ - *(GLuint *)(buf + _x*4 + _y*pitch) = tmp; \ -} - -#define READ_STENCIL( d, _x, _y ) \ - d = *(GLuint *)(buf + _x*4 + _y*pitch) & 0xff - -#define TAG(x) s3v##x##_24_8 -#include "stenciltmp.h" - -#endif - - -/** - * Plug in the Get/Put routines for the given driRenderbuffer. - */ -void -s3vSetSpanFunctions(driRenderbuffer *drb, const GLvisual *vis) -{ - if (drb->Base.InternalFormat == GL_RGBA) { - if (vis->redBits == 5 && vis->greenBits == 6 && vis->blueBits == 5) { - s3vInitPointers_RGB555(&drb->Base); - } - else { - s3vInitPointers_ARGB8888(&drb->Base); - } - } - else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT16) { - s3vInitDepthPointers_z16(&drb->Base); - } - else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT24) { - /* not done yet */ - } - else if (drb->Base.InternalFormat == GL_STENCIL_INDEX8_EXT) { - /* not done yet */ - } -} diff --git a/src/mesa/drivers/dri/s3v/s3v_state.c b/src/mesa/drivers/dri/s3v/s3v_state.c deleted file mode 100644 index 561f42c705f..00000000000 --- a/src/mesa/drivers/dri/s3v/s3v_state.c +++ /dev/null @@ -1,888 +0,0 @@ -/* - * Author: Max Lingua <[email protected]> - */ - -#include "s3v_context.h" -#include "s3v_macros.h" -#include "s3v_dri.h" -#include "main/macros.h" -#include "main/colormac.h" -#include "swrast/swrast.h" -#include "swrast_setup/swrast_setup.h" -#include "vbo/vbo.h" -#include "tnl/tnl.h" - -/* #define DEBUG(str) printf str */ -#define ENABLELIGHTING 0 - - -/* ============================================================= - * Alpha blending - */ - -static void s3vUpdateAlphaMode( GLcontext *ctx ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - uint32_t cmd = vmesa->CMD; - cmd &= ~ALPHA_BLEND_MASK; - - if ( ctx->Color.BlendEnabled ) { - DEBUG(("ctx->Color.AlphaEnabled = 1")); - vmesa->_alpha[0] = ALPHA_SRC; - vmesa->_alpha[1] = vmesa->_alpha_tex; /* FIXME: not all tex modes - support alpha */ - } else { - DEBUG(("ctx->Color.AlphaEnabled = 0")); - vmesa->_alpha[0] = vmesa->_alpha[1] = ALPHA_OFF; - } -#if 1 - if ((cmd & DO_MASK) & DO_3D_LINE) { /* we are drawing 3d lines */ - /* which don't support tex */ - cmd |= vmesa->_alpha[0]; - } else { - cmd |= vmesa->_alpha[vmesa->_3d_mode]; - } - - vmesa->CMD = cmd; /* FIXME: enough? */ -#else - vmesa->restore_primitive = -1; -#endif - -} - -static void s3vDDAlphaFunc( GLcontext *ctx, GLenum func, GLfloat ref ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - - DEBUG(("s3vDDAlphaFunc\n")); - - vmesa->new_state |= S3V_NEW_ALPHA; -} - -static void s3vDDBlendFunc( GLcontext *ctx, GLenum sfactor, GLenum dfactor ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - - DEBUG(("s3vDDBlendFunc\n")); - - vmesa->new_state |= S3V_NEW_ALPHA; -} - -/* ================================================================ - * Buffer clear - */ - -static void s3vDDClear( GLcontext *ctx, GLbitfield mask ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - unsigned int _stride; - GLint cx = ctx->DrawBuffer->_Xmin; - GLint cy = ctx->DrawBuffer->_Ymin; - GLint cw = ctx->DrawBuffer->_Xmax - cx; - GLint ch = ctx->DrawBuffer->_Ymax - cy; - - /* XXX FIX ME: the cx,cy,cw,ch vars are currently ignored! */ - - vmesa->restore_primitive = -1; - - /* Update and emit any new state. We need to do this here to catch - * changes to the masks. - * FIXME: Just update the masks? - */ - - if ( vmesa->new_state ) - s3vDDUpdateHWState( ctx ); - -/* s3vUpdateMasks( ctx ); */ -/* s3vUpdateClipping( ctx ); */ -/* s3vEmitHwState( vmesa ); */ - - -#if 1 /* soft (0)/hw (1)*/ - - DEBUG(("*** s3vDDClear ***\n")); - - DMAOUT_CHECK(BITBLT_SRC_BASE, 15); - DMAOUT(vmesa->SrcBase); - DMAOUT(vmesa->DestBlit); - DMAOUT( vmesa->ScissorLR ); - DMAOUT( vmesa->ScissorTB ); - DMAOUT( (vmesa->SrcStride << 16) | vmesa->SrcStride ); /* FIXME: unify */ - DMAOUT( (~(0)) ); /* masks */ - DMAOUT( (~(0)) ); - DMAOUT(0); - DMAOUT(vmesa->ClearColor); - DMAOUT(0); - DMAOUT(0); - /* FIXME */ - DMAOUT(0x16000122 | 0x5 | (0xF0 << 17)); /* black magic to me */ - DMAOUT(vmesa->ScissorWH); - DMAOUT(vmesa->SrcXY); - DMAOUT(vmesa->DestXY); - DMAFINISH(); - - if (mask & BUFFER_BIT_DEPTH) { /* depth */ - DEBUG(("BUFFER_BIT_DEPTH\n")); - - _stride = ((cw+31)&~31) * 2; /* XXX cw or Buffer->Width??? */ - - DMAOUT_CHECK(BITBLT_SRC_BASE, 15); - DMAOUT(0); - DMAOUT(vmesa->s3vScreen->depthOffset); - DMAOUT( (0 << 16) | cw ); - DMAOUT( (0 << 16) | ch ); - DMAOUT( (vmesa->SrcStride << 16) | vmesa->DestStride ); - DMAOUT( (~(0)) ); /* masks */ - DMAOUT( (~(0)) ); - DMAOUT(0); - DMAOUT(vmesa->ClearDepth); /* 0x7FFF */ - /* FIXME */ - DMAOUT(0); - DMAOUT(0); - DMAOUT(0x16000122 | 0x5 | (0xF0 << 17)); - DMAOUT( ((cw-1) << 16) | (ch-1) ); - DMAOUT(0); - DMAOUT( (0 << 16) | 0 ); - DMAFINISH(); - - DEBUG(("vmesa->ClearDepth = 0x%x\n", vmesa->ClearDepth)); - mask &= ~BUFFER_BIT_DEPTH; - } - - if (!vmesa->NotClipped) { - DEBUG(("vmesa->NotClipped\n")); /* yes */ - } - - if (!(vmesa->EnabledFlags & S3V_BACK_BUFFER)) { - DEBUG(("!S3V_BACK_BUFFER -> flush\n")); - DMAFLUSH(); - } -/* - if ( mask ) - DEBUG(("still masked ;3(\n")); */ /* yes */ -#else - _swrast_Clear( ctx, mask ); -#endif -} - -/* ============================================================= - * Depth testing - */ - -static void s3vUpdateZMode( GLcontext *ctx ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - uint32_t cmd = vmesa->CMD; - - DEBUG(("Depth.Test = %i\n", ctx->Depth.Test)); - DEBUG(("CMD was = 0x%x ", cmd)); - -/* printf("depth --- CMD was = 0x%x \n", cmd); */ - - cmd &= ~Z_MASK; /* 0xfc0fffff; */ - /* Z_BUFFER */ /* 000 mode */ /* Z_UPDATE_OFF */ - - if (!ctx->Depth.Test) - cmd |= Z_OFF; - - if ( ctx->Depth.Mask ) - cmd |= Z_UPDATE_ON; - - switch ( ctx->Depth.Func ) { - case GL_NEVER: - cmd |= Z_NEVER; - break; - case GL_ALWAYS: - cmd |= Z_ALWAYS; - break; - case GL_LESS: - cmd |= Z_LESS; - break; - case GL_LEQUAL: - cmd |= Z_LEQUAL; - break; - case GL_EQUAL: - cmd |= Z_EQUAL; - break; - case GL_GEQUAL: - cmd |= Z_GEQUAL; - break; - case GL_GREATER: - cmd |= Z_GREATER; - break; - case GL_NOTEQUAL: - cmd |= Z_NOTEQUAL; - break; - } - - DEBUG(("CMD is 0x%x\n", cmd)); - - vmesa->dirty |= S3V_UPLOAD_DEPTH; - vmesa->CMD = cmd; -} - -static void s3vDDDepthFunc( GLcontext *ctx, GLenum func ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - -/* FLUSH_BATCH( vmesa ); */ - DEBUG(("s3vDDDepthFunc\n")); - vmesa->new_state |= S3V_NEW_DEPTH; -} - -static void s3vDDDepthMask( GLcontext *ctx, GLboolean flag ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - - /* FLUSH_BATCH( vmesa ); */ - DEBUG(("s3vDDDepthMask\n")); - vmesa->new_state |= S3V_NEW_DEPTH; -} - -static void s3vDDClearDepth( GLcontext *ctx, GLclampd d ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - - switch ( vmesa->DepthSize ) { - case 15: - case 16: - vmesa->ClearDepth = d * 0x0000ffff; /* 65536 */ - DEBUG(("GLclampd d = %f\n", d)); - DEBUG(("ctx->Depth.Clear = %f\n", ctx->Depth.Clear)); - DEBUG(("(They should be the same)\n")); - break; - case 24: - vmesa->ClearDepth = d * 0x00ffffff; - break; - case 32: - vmesa->ClearDepth = d * 0xffffffff; - break; - } -} - -static void s3vDDFinish( GLcontext *ctx ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - DMAFLUSH(); -} - -static void s3vDDFlush( GLcontext *ctx ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - DMAFLUSH(); -} - -/* ============================================================= - * Fog - */ - -static void s3vUpdateFogAttrib( GLcontext *ctx ) -{ -/* s3vContextPtr vmesa = S3V_CONTEXT(ctx); */ - - if (ctx->Fog.Enabled) { - } else { - } - - switch (ctx->Fog.Mode) { - case GL_LINEAR: - break; - case GL_EXP: - break; - case GL_EXP2: - break; - } -} - -static void s3vDDFogfv( GLcontext *ctx, GLenum pname, const GLfloat *param ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - vmesa->new_state |= S3V_NEW_FOG; -} - -/* ============================================================= - * Lines - */ -static void s3vDDLineWidth( GLcontext *ctx, GLfloat width ) -{ - /* FIXME: on virge you only have one size of 3d lines * - * if we wanted more, we should start using tris instead * - * but virge has problem with some tris when all of the * - * vertices stay on a line */ -} - -/* ============================================================= - * Points - */ -static void s3vDDPointSize( GLcontext *ctx, GLfloat size ) -{ - /* FIXME: we use 3d line to fake points. So same limitations - * as above apply */ -} - -/* ============================================================= - * Polygon - */ - -static void s3vUpdatePolygon( GLcontext *ctx ) -{ - /* FIXME: I don't think we could do much here */ - - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - vmesa->dirty |= S3V_UPLOAD_POLYGON; -} - -/* ============================================================= - * Clipping - */ - -static void s3vUpdateClipping( GLcontext *ctx ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - __DRIdrawablePrivate *dPriv = vmesa->driDrawable; - - int x0,y0,x1,y1; - - DEBUG((">>> s3vUpdateClipping <<<\n")); -/* - if ( vmesa->driDrawable ) { - DEBUG(("s3vUpdateClipping\n")); -*/ - if (vmesa->EnabledFlags & S3V_BACK_BUFFER) { - DEBUG(("S3V_BACK_BUFFER\n")); - - x0 = 0; - y0 = 0; - x1 = dPriv->w - 1; - y1 = dPriv->h - 1; - - vmesa->SrcBase = 0; - vmesa->DestBase = vmesa->s3vScreen->backOffset; - vmesa->DestBlit = vmesa->DestBase; - vmesa->ScissorLR = ( (0 << 16) | (dPriv->w-1) ); - vmesa->ScissorTB = ( (0 << 16) | (dPriv->h-1) ); -/* - vmesa->ScissorLR = ( (x0 << 16) | x1 ); - vmesa->ScissorTB = ( (y0 << 16) | y1 ); -*/ - vmesa->SrcStride = ( ((dPriv->w+31)&~31) * vmesa->s3vScreen->cpp ); - vmesa->DestStride = vmesa->driScreen->fbWidth*vmesa->s3vScreen->cpp; - vmesa->ScissorWH = ( (dPriv->w << 16) | dPriv->h ); - vmesa->SrcXY = 0; -/* vmesa->DestXY = ( (dPriv->x << 16) | dPriv->y ); */ - vmesa->DestXY = ( (0 << 16) | 0 ); - } else { - DEBUG(("S3V_FRONT_BUFFER\n")); - - x0 = dPriv->x; - y0 = dPriv->y; - x1 = x0 + dPriv->w - 1; - y1 = y0 + dPriv->h - 1; - - vmesa->SrcBase = 0; - vmesa->DestBase = 0; - vmesa->ScissorLR = ( (x0 << 16) | x1 ); - vmesa->ScissorTB = ( (y0 << 16) | y1 ); - vmesa->DestStride = vmesa->driScreen->fbWidth*vmesa->s3vScreen->cpp; - vmesa->SrcStride = vmesa->DestStride; - vmesa->DestBase = (y0 * vmesa->DestStride) - + x0*vmesa->s3vScreen->cpp; - vmesa->DestBlit = 0; - vmesa->ScissorWH = ( (x1 << 16) | y1 ); - vmesa->SrcXY = 0; - vmesa->DestXY = ( (0 << 16) | 0 ); -/* vmesa->DestXY = ( (dPriv->x << 16) | dPriv->y ); */ - } - - DEBUG(("x0=%i y0=%i x1=%i y1=%i\n", x0, y0, x1, y1)); - DEBUG(("stride=%i rectWH=0x%x\n\n", vmesa->DestStride, vmesa->ScissorWH)); - - /* FIXME: how could we use the following info? */ - /* if (ctx->Scissor.Enabled) {} */ - - vmesa->dirty |= S3V_UPLOAD_CLIP; -/* } */ -} - -static void s3vDDScissor( GLcontext *ctx, - GLint x, GLint y, GLsizei w, GLsizei h ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - - DEBUG((">>> s3vDDScissor <<<")); - /* FLUSH_BATCH( vmesa ); */ - vmesa->new_state |= S3V_NEW_CLIP; -} - -/* ============================================================= - * Culling - */ - -static void s3vUpdateCull( GLcontext *ctx ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - GLfloat backface_sign = 1; - - DEBUG(("s3vUpdateCull\n")); - /* FIXME: GL_FRONT_AND_BACK */ - - switch ( ctx->Polygon.CullFaceMode ) { - case GL_BACK: - if (ctx->Polygon.FrontFace == GL_CCW) - backface_sign = -1; - break; - - case GL_FRONT: - if (ctx->Polygon.FrontFace != GL_CCW) - backface_sign = -1; - break; - - default: - break; - } - - vmesa->backface_sign = backface_sign; - vmesa->dirty |= S3V_UPLOAD_GEOMETRY; -} - - -static void s3vDDCullFace( GLcontext *ctx, GLenum mode ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - /* FLUSH_BATCH( vmesa ); */ - vmesa->new_state |= S3V_NEW_CULL; -} - -static void s3vDDFrontFace( GLcontext *ctx, GLenum mode ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - /* FLUSH_BATCH( vmesa ); */ - vmesa->new_state |= S3V_NEW_CULL; -} - -/* ============================================================= - * Masks - */ - -static void s3vUpdateMasks( GLcontext *ctx ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - - GLuint mask = s3vPackColor( vmesa->s3vScreen->cpp, - ctx->Color.ColorMask[RCOMP], - ctx->Color.ColorMask[GCOMP], - ctx->Color.ColorMask[BCOMP], - ctx->Color.ColorMask[ACOMP] ); - - if (vmesa->s3vScreen->cpp == 2) mask |= mask << 16; - - /* FIXME: can we do something in virge? */ -} -/* -static void s3vDDColorMask( GLcontext *ctx, GLboolean r, GLboolean g, - GLboolean b, GLboolean a) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - - FLUSH_BATCH( vmesa ); - vmesa->new_state |= S3V_NEW_MASKS; -} -*/ -/* ============================================================= - * Rendering attributes - */ - -/* ============================================================= - * Miscellaneous - */ - -static void s3vDDClearColor( GLcontext *ctx, const GLfloat color[4]) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - - DEBUG(("*** s3vDDClearColor\n")); - - vmesa->ClearColor = s3vPackColor( 2, /* vmesa->s3vScreen->cpp, */ - color[0], color[1], color[2], color[3] ); - -#if 0 - if (vmesa->s3vScreen->cpp == 2) vmesa->ClearColor |= vmesa->ClearColor<<16; -#endif -} - -static void s3vDDSetDrawBuffer( GLcontext *ctx, GLenum mode ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - int found = GL_TRUE; - - DEBUG(("*** s3vDDSetDrawBuffer ***\n")); - - /* FLUSH_BATCH( vmesa ); */ - - switch ( mode ) { - case GL_FRONT_LEFT: - vmesa->drawOffset = vmesa->s3vScreen->frontOffset; - break; - case GL_BACK_LEFT: - vmesa->drawOffset = vmesa->s3vScreen->backOffset; - /* vmesa->driScreen->fbHeight * - * vmesa->driScreen->fbWidth * - * vmesa->s3vScreen->cpp; */ - break; - default: - found = GL_FALSE; - break; - } - - DEBUG(("vmesa->drawOffset = 0x%x\n", vmesa->drawOffset)); -/* return GL_TRUE; */ -} - -/* ============================================================= - * Window position and viewport transformation - */ - -void s3vUpdateWindow( GLcontext *ctx ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - __DRIdrawablePrivate *dPriv = vmesa->driDrawable; - GLfloat xoffset = (GLfloat)dPriv->x; - GLfloat yoffset = - vmesa->driScreen->fbHeight - (GLfloat)dPriv->y - dPriv->h; - const GLfloat *v = ctx->Viewport._WindowMap.m; - - GLfloat sx = v[MAT_SX]; - GLfloat tx = v[MAT_TX] + xoffset; - GLfloat sy = v[MAT_SY]; - GLfloat ty = v[MAT_TY] + yoffset; - GLfloat sz = v[MAT_SZ] * vmesa->depth_scale; - GLfloat tz = v[MAT_TZ] * vmesa->depth_scale; - - vmesa->dirty |= S3V_UPLOAD_VIEWPORT; - - vmesa->ViewportScaleX = sx; - vmesa->ViewportScaleY = sy; - vmesa->ViewportScaleZ = sz; - vmesa->ViewportOffsetX = tx; - vmesa->ViewportOffsetY = ty; - vmesa->ViewportOffsetZ = tz; -} - - -/* -static void s3vDDViewport( GLcontext *ctx, GLint x, GLint y, - GLsizei width, GLsizei height ) -{ - s3vUpdateWindow( ctx ); -} - -static void s3vDDDepthRange( GLcontext *ctx, GLclampd nearval, - GLclampd farval ) -{ - s3vUpdateWindow( ctx ); -} -*/ -void s3vUpdateViewportOffset( GLcontext *ctx ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - __DRIdrawablePrivate *dPriv = vmesa->driDrawable; - GLfloat xoffset = (GLfloat)dPriv->x; - GLfloat yoffset = - vmesa->driScreen->fbHeight - (GLfloat)dPriv->y - dPriv->h; - const GLfloat *v = ctx->Viewport._WindowMap.m; - - GLfloat tx = v[MAT_TX] + xoffset; - GLfloat ty = v[MAT_TY] + yoffset; - - DEBUG(("*** s3vUpdateViewportOffset ***\n")); - - if ( vmesa->ViewportOffsetX != tx || - vmesa->ViewportOffsetY != ty ) - { - vmesa->ViewportOffsetX = tx; - vmesa->ViewportOffsetY = ty; - - vmesa->new_state |= S3V_NEW_WINDOW; - } - -/* vmesa->new_state |= S3V_NEW_CLIP; */ -} - -/* ============================================================= - * State enable/disable - */ - -static void s3vDDEnable( GLcontext *ctx, GLenum cap, GLboolean state ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - - switch ( cap ) { - case GL_ALPHA_TEST: - case GL_BLEND: - vmesa->new_state |= S3V_NEW_ALPHA; - DEBUG(("s3vDDEnable: GL_BLEND\n")); - break; - - case GL_CULL_FACE: - vmesa->new_state |= S3V_NEW_CULL; - DEBUG(("s3vDDEnable: GL_CULL_FACE\n")); - break; - - case GL_DEPTH_TEST: - vmesa->new_state |= S3V_NEW_DEPTH; - DEBUG(("s3vDDEnable: GL_DEPTH\n")); - break; -#if 0 - case GL_FOG: - vmesa->new_state |= S3V_NEW_FOG; - break; -#endif - - case GL_SCISSOR_TEST: - vmesa->new_state |= S3V_NEW_CLIP; - break; - - case GL_TEXTURE_2D: - DEBUG(("*** GL_TEXTURE_2D: %i\n", state)); - vmesa->_3d_mode = state; - vmesa->restore_primitive = -1; - break; - - default: - return; - } -} - -/* ============================================================= - * State initialization, management - */ - - -/* - * Load the current context's state into the hardware. - * - * NOTE: Be VERY careful about ensuring the context state is marked for - * upload, the only place it shouldn't be uploaded is when the setup - * state has changed in ReducedPrimitiveChange as this comes right after - * a state update. - * - * Blits of any type should always upload the context and masks after - * they are done. - */ -void s3vEmitHwState( s3vContextPtr vmesa ) -{ - if (!vmesa->driDrawable) return; - if (!vmesa->dirty) return; - - DEBUG(("**********************\n")); - DEBUG(("*** s3vEmitHwState ***\n")); - DEBUG(("**********************\n")); - - if (vmesa->dirty & S3V_UPLOAD_VIEWPORT) { - vmesa->dirty &= ~S3V_UPLOAD_VIEWPORT; - DEBUG(("S3V_UPLOAD_VIEWPORT\n")); - } - - if ( (vmesa->dirty & S3V_UPLOAD_POINTMODE) || - (vmesa->dirty & S3V_UPLOAD_LINEMODE) || - (vmesa->dirty & S3V_UPLOAD_TRIMODE) ) { - - } - - if (vmesa->dirty & S3V_UPLOAD_POINTMODE) { - vmesa->dirty &= ~S3V_UPLOAD_POINTMODE; - } - - if (vmesa->dirty & S3V_UPLOAD_LINEMODE) { - vmesa->dirty &= ~S3V_UPLOAD_LINEMODE; - } - - if (vmesa->dirty & S3V_UPLOAD_TRIMODE) { - vmesa->dirty &= ~S3V_UPLOAD_TRIMODE; - } - - if (vmesa->dirty & S3V_UPLOAD_FOG) { - GLchan c[3], col; - UNCLAMPED_FLOAT_TO_RGB_CHAN( c, vmesa->glCtx->Fog.Color ); - DEBUG(("uploading ** FOG **\n")); - col = s3vPackColor(2, c[0], c[1], c[2], 0); - vmesa->dirty &= ~S3V_UPLOAD_FOG; - } - - if (vmesa->dirty & S3V_UPLOAD_DITHER) { - vmesa->dirty &= ~S3V_UPLOAD_DITHER; - } - - if (vmesa->dirty & S3V_UPLOAD_LOGICOP) { - vmesa->dirty &= ~S3V_UPLOAD_LOGICOP; - } - - if (vmesa->dirty & S3V_UPLOAD_CLIP) { - vmesa->dirty &= ~S3V_UPLOAD_CLIP; - DEBUG(("S3V_UPLOAD_CLIP\n")); - DEBUG(("vmesa->ScissorLR: %i\n", vmesa->ScissorLR)); - DEBUG(("vmesa->ScissorTB: %i\n", vmesa->ScissorTB)); - } - - if (vmesa->dirty & S3V_UPLOAD_MASKS) { - vmesa->dirty &= ~S3V_UPLOAD_MASKS; - DEBUG(("S3V_UPLOAD_BLEND\n")); - } - - if (vmesa->dirty & S3V_UPLOAD_ALPHA) { - vmesa->dirty &= ~S3V_UPLOAD_ALPHA; - DEBUG(("S3V_UPLOAD_ALPHA\n")); - } - - if (vmesa->dirty & S3V_UPLOAD_SHADE) { - vmesa->dirty &= ~S3V_UPLOAD_SHADE; - } - - if (vmesa->dirty & S3V_UPLOAD_POLYGON) { - vmesa->dirty &= ~S3V_UPLOAD_POLYGON; - } - - if (vmesa->dirty & S3V_UPLOAD_DEPTH) { - vmesa->dirty &= ~S3V_UPLOAD_DEPTH; - DEBUG(("S3V_UPLOAD_DEPTH: DepthMode = 0x%x08\n", vmesa->DepthMode)); - } - - if (vmesa->dirty & S3V_UPLOAD_GEOMETRY) { - vmesa->dirty &= ~S3V_UPLOAD_GEOMETRY; - } - - if (vmesa->dirty & S3V_UPLOAD_TRANSFORM) { - vmesa->dirty &= ~S3V_UPLOAD_TRANSFORM; - } - - if (vmesa->dirty & S3V_UPLOAD_TEX0) { - s3vTextureObjectPtr curTex = vmesa->CurrentTexObj[0]; - vmesa->dirty &= ~S3V_UPLOAD_TEX0; - DEBUG(("S3V_UPLOAD_TEX0\n")); - if (curTex) { - DEBUG(("S3V_UPLOAD_TEX0: curTex\n")); - } else { - DEBUG(("S3V_UPLOAD_TEX0: !curTex\n")); - } - } -} - -void s3vDDUpdateHWState( GLcontext *ctx ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - - int new_state = vmesa->new_state; - - /* s3vUpdateClipping( ctx ); */ - - if ( new_state ) - { - - vmesa->new_state = 0; - - /* Update the various parts of the context's state. - */ - if ( new_state & S3V_NEW_ALPHA ) - s3vUpdateAlphaMode( ctx ); - - if ( new_state & S3V_NEW_DEPTH ) - s3vUpdateZMode( ctx ); - - if ( new_state & S3V_NEW_FOG ) - s3vUpdateFogAttrib( ctx ); - - if ( new_state & S3V_NEW_CLIP ) - { - DEBUG(("---> going to s3vUpdateClipping\n")); - s3vUpdateClipping( ctx ); - } - - if ( new_state & S3V_NEW_POLYGON ) - s3vUpdatePolygon( ctx ); - - if ( new_state & S3V_NEW_CULL ) - s3vUpdateCull( ctx ); - - if ( new_state & S3V_NEW_MASKS ) - s3vUpdateMasks( ctx ); - - if ( new_state & S3V_NEW_WINDOW ) - s3vUpdateWindow( ctx ); -/* - if ( new_state & S3_NEW_TEXTURE ) - s3vUpdateTextureState( ctx ); -*/ - CMDCHANGE(); - } - - /* HACK ! */ - s3vEmitHwState( vmesa ); -} - - -static void s3vDDUpdateState( GLcontext *ctx, GLuint new_state ) -{ - _swrast_InvalidateState( ctx, new_state ); - _swsetup_InvalidateState( ctx, new_state ); - _vbo_InvalidateState( ctx, new_state ); - _tnl_InvalidateState( ctx, new_state ); - S3V_CONTEXT(ctx)->new_gl_state |= new_state; -} - - -/* Initialize the context's hardware state. - */ -void s3vInitState( s3vContextPtr vmesa ) -{ - vmesa->new_state = 0; -} - -/* Initialize the driver's state functions. - */ -void s3vInitStateFuncs( GLcontext *ctx ) -{ - ctx->Driver.UpdateState = s3vDDUpdateState; - - ctx->Driver.Clear = s3vDDClear; - ctx->Driver.ClearIndex = NULL; - ctx->Driver.ClearColor = s3vDDClearColor; - ctx->Driver.DrawBuffer = s3vDDSetDrawBuffer; - ctx->Driver.ReadBuffer = NULL; /* XXX */ - - ctx->Driver.IndexMask = NULL; - ctx->Driver.ColorMask = NULL; /* s3vDDColorMask; */ /* FIXME */ - - ctx->Driver.AlphaFunc = s3vDDAlphaFunc; /* FIXME */ -#if 0 - ctx->Driver.BlendEquation = NULL; /* s3vDDBlendEquation; */ - ctx->Driver.BlendFunc = s3vDDBlendFunc; /* FIXME */ -#endif - ctx->Driver.BlendFuncSeparate = NULL; /* s3vDDBlendFuncSeparate; */ - ctx->Driver.ClearDepth = s3vDDClearDepth; - ctx->Driver.CullFace = s3vDDCullFace; - ctx->Driver.FrontFace = s3vDDFrontFace; - ctx->Driver.DepthFunc = s3vDDDepthFunc; /* FIXME */ - ctx->Driver.DepthMask = s3vDDDepthMask; /* FIXME */ - ctx->Driver.DepthRange = NULL; /* s3vDDDepthRange; */ - ctx->Driver.Enable = s3vDDEnable; /* FIXME */ - ctx->Driver.Finish = s3vDDFinish; - ctx->Driver.Flush = s3vDDFlush; -#if 1 - ctx->Driver.Fogfv = NULL; /* s3vDDFogfv; */ -#endif - ctx->Driver.Hint = NULL; - ctx->Driver.LineWidth = NULL; /* s3vDDLineWidth; */ - ctx->Driver.LineStipple = NULL; /* s3vDDLineStipple; */ -#if ENABLELIGHTING - ctx->Driver.Lightfv = NULL; /* s3vDDLightfv; */ - - ctx->Driver.LightModelfv = NULL; /* s3vDDLightModelfv; */ -#endif - ctx->Driver.LogicOpcode = NULL; /* s3vDDLogicalOpcode; */ - ctx->Driver.PointSize = NULL; /* s3vDDPointSize; */ - ctx->Driver.PolygonMode = NULL; /* s3vDDPolygonMode; */ - ctx->Driver.PolygonStipple = NULL; /* s3vDDPolygonStipple; */ - ctx->Driver.Scissor = s3vDDScissor; /* ScissorLR / ScissorTB */ - ctx->Driver.ShadeModel = NULL; /* s3vDDShadeModel; */ - ctx->Driver.Viewport = NULL; /* s3vDDViewport; */ -} diff --git a/src/mesa/drivers/dri/s3v/s3v_tex.c b/src/mesa/drivers/dri/s3v/s3v_tex.c deleted file mode 100644 index ec1182f34f7..00000000000 --- a/src/mesa/drivers/dri/s3v/s3v_tex.c +++ /dev/null @@ -1,548 +0,0 @@ -/* - * Author: Max Lingua <[email protected]> - */ - -#include <stdlib.h> -#include <stdio.h> - -#include "main/glheader.h" -#include "main/mtypes.h" -#include "main/simple_list.h" -#include "main/enums.h" -#include "main/mm.h" -#include "main/texstore.h" -#include "main/texformat.h" -#include "main/teximage.h" -#include "swrast/swrast.h" - -#include "s3v_context.h" -#include "s3v_tex.h" - - -extern void s3vSwapOutTexObj(s3vContextPtr vmesa, s3vTextureObjectPtr t); -extern void s3vDestroyTexObj(s3vContextPtr vmesa, s3vTextureObjectPtr t); - -/* -static GLuint s3vComputeLodBias(GLfloat bias) -{ -#if TEX_DEBUG_ON - DEBUG_TEX(("*** s3vComputeLodBias ***\n")); -#endif - return bias; -} -*/ - -static void s3vSetTexWrapping(s3vContextPtr vmesa, - s3vTextureObjectPtr t, - GLenum wraps, GLenum wrapt) -{ - GLuint t0 = t->TextureCMD; - GLuint cmd = vmesa->CMD; -#if TEX_DEBUG_ON - static unsigned int times=0; - DEBUG_TEX(("*** s3vSetTexWrapping: #%i ***\n", ++times)); -#endif - - - t0 &= ~TEX_WRAP_MASK; - cmd &= ~TEX_WRAP_MASK; - - if ((wraps != GL_CLAMP) || (wrapt != GL_CLAMP)) { - DEBUG(("TEX_WRAP_ON\n")); - t0 |= TEX_WRAP_ON; - cmd |= TEX_WRAP_ON; - } - - cmd |= TEX_WRAP_ON; /* FIXME: broken if off */ - t->TextureCMD = t0; - vmesa->CMD = cmd; -} - - -static void s3vSetTexFilter(s3vContextPtr vmesa, - s3vTextureObjectPtr t, - GLenum minf, GLenum magf) -{ - GLuint t0 = t->TextureCMD; - GLuint cmd = vmesa->CMD; -#if TEX_DEBUG_ON - static unsigned int times=0; - DEBUG_TEX(("*** s3vSetTexFilter: #%i ***\n", ++times)); -#endif - - t0 &= ~TEX_FILTER_MASK; - cmd &= ~TEX_FILTER_MASK; - - switch (minf) { - case GL_NEAREST: - DEBUG(("GL_NEAREST\n")); - t0 |= NEAREST; - cmd |= NEAREST; - break; - case GL_LINEAR: - DEBUG(("GL_LINEAR\n")); - t0 |= LINEAR; - cmd |= LINEAR; - break; - case GL_NEAREST_MIPMAP_NEAREST: - DEBUG(("GL_MIPMAP_NEAREST\n")); - t0 |= MIP_NEAREST; - cmd |= MIP_NEAREST; - break; - case GL_LINEAR_MIPMAP_NEAREST: - DEBUG(("GL_LINEAR_MIPMAP_NEAREST\n")); - t0 |= LINEAR_MIP_NEAREST; - cmd |= LINEAR_MIP_NEAREST; - break; - case GL_NEAREST_MIPMAP_LINEAR: - DEBUG(("GL_NEAREST_MIPMAP_LINEAR\n")); - t0 |= MIP_LINEAR; - cmd |= MIP_LINEAR; - break; - case GL_LINEAR_MIPMAP_LINEAR: - DEBUG(("GL_LINEAR_MIPMAP_LINEAR\n")); - t0 |= LINEAR_MIP_LINEAR; - cmd |= LINEAR_MIP_LINEAR; - break; - default: - break; - } - /* FIXME: bilinear? */ - -#if 0 - switch (magf) { - case GL_NEAREST: - break; - case GL_LINEAR: - break; - default: - break; - } -#endif - - t->TextureCMD = t0; - - DEBUG(("CMD was = 0x%x\n", vmesa->CMD)); - DEBUG(("CMD is = 0x%x\n", cmd)); - - vmesa->CMD = cmd; - /* CMDCHANGE(); */ -} - - -static void s3vSetTexBorderColor(s3vContextPtr vmesa, - s3vTextureObjectPtr t, - const GLfloat color[4]) -{ - GLubyte c[4]; - CLAMPED_FLOAT_TO_UBYTE(c[0], color[0]); - CLAMPED_FLOAT_TO_UBYTE(c[1], color[1]); - CLAMPED_FLOAT_TO_UBYTE(c[2], color[2]); - CLAMPED_FLOAT_TO_UBYTE(c[3], color[3]); - -#if TEX_DEBUG_ON - static unsigned int times=0; - DEBUG_TEX(("*** s3vSetTexBorderColor: #%i ***\n", ++times)); -#endif - - /*FIXME: it should depend on tex col format */ - /* switch(t0 ... t->TextureColorMode) */ - - /* case TEX_COL_ARGB1555: */ - t->TextureBorderColor = S3VIRGEPACKCOLOR555(c[0], c[1], c[2], c[3]); - - DEBUG(("TextureBorderColor = 0x%x\n", t->TextureBorderColor)); - - vmesa->TextureBorderColor = t->TextureBorderColor; -} - -static void s3vTexParameter( GLcontext *ctx, GLenum target, - struct gl_texture_object *tObj, - GLenum pname, const GLfloat *params ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - s3vTextureObjectPtr t = (s3vTextureObjectPtr) tObj->DriverData; -#if TEX_DEBUG_ON - static unsigned int times=0; - DEBUG_TEX(("*** s3vTexParameter: #%i ***\n", ++times)); -#endif - - if (!t) return; - - /* Can't do the update now as we don't know whether to flush - * vertices or not. Setting vmesa->new_state means that - * s3vUpdateTextureState() will be called before any triangles are - * rendered. If a statechange has occurred, it will be detected at - * that point, and buffered vertices flushed. - */ - switch (pname) { - case GL_TEXTURE_MIN_FILTER: - case GL_TEXTURE_MAG_FILTER: - s3vSetTexFilter( vmesa, t, tObj->MinFilter, tObj->MagFilter ); - break; - - case GL_TEXTURE_WRAP_S: - case GL_TEXTURE_WRAP_T: - s3vSetTexWrapping( vmesa, t, tObj->WrapS, tObj->WrapT ); - break; - - case GL_TEXTURE_BORDER_COLOR: - s3vSetTexBorderColor( vmesa, t, tObj->BorderColor ); - break; - - case GL_TEXTURE_BASE_LEVEL: - case GL_TEXTURE_MAX_LEVEL: - case GL_TEXTURE_MIN_LOD: - case GL_TEXTURE_MAX_LOD: - /* This isn't the most efficient solution but there doesn't appear to - * be a nice alternative for Virge. Since there's no LOD clamping, - * we just have to rely on loading the right subset of mipmap levels - * to simulate a clamped LOD. - */ - s3vSwapOutTexObj( vmesa, t ); - break; - - default: - return; - } - - if (t == vmesa->CurrentTexObj[0]) - vmesa->dirty |= S3V_UPLOAD_TEX0; - -#if 0 - if (t == vmesa->CurrentTexObj[1]) { - vmesa->dirty |= S3V_UPLOAD_TEX1; - } -#endif -} - - -static void s3vTexEnv( GLcontext *ctx, GLenum target, - GLenum pname, const GLfloat *param ) -{ - s3vContextPtr vmesa = S3V_CONTEXT( ctx ); - GLuint unit = ctx->Texture.CurrentUnit; -#if TEX_DEBUG_ON - static unsigned int times=0; - DEBUG_TEX(("*** s3vTexEnv: #%i ***\n", ++times)); -#endif - - /* Only one env color. Need a fallback if env colors are different - * and texture setup references env color in both units. - */ - switch (pname) { - case GL_TEXTURE_ENV_COLOR: { - struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; - GLfloat *fc = texUnit->EnvColor; - GLuint r, g, b, a, col; - CLAMPED_FLOAT_TO_UBYTE(r, fc[0]); - CLAMPED_FLOAT_TO_UBYTE(g, fc[1]); - CLAMPED_FLOAT_TO_UBYTE(b, fc[2]); - CLAMPED_FLOAT_TO_UBYTE(a, fc[3]); - - col = ((a << 24) | - (r << 16) | - (g << 8) | - (b << 0)); - - break; - } - case GL_TEXTURE_ENV_MODE: - vmesa->TexEnvImageFmt[unit] = 0; /* force recalc of env state */ - break; - case GL_TEXTURE_LOD_BIAS_EXT: { -/* - struct gl_texture_object *tObj = - ctx->Texture.Unit[unit]._Current; - - s3vTextureObjectPtr t = (s3vTextureObjectPtr) tObj->DriverData; -*/ - break; - } - default: - break; - } -} - -static void s3vTexImage1D( GLcontext *ctx, GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint border, - GLenum format, GLenum type, - const GLvoid *pixels, - const struct gl_pixelstore_attrib *pack, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ) -{ - s3vContextPtr vmesa = S3V_CONTEXT( ctx ); - s3vTextureObjectPtr t = (s3vTextureObjectPtr) texObj->DriverData; -#if TEX_DEBUG_ON - static unsigned int times=0; - DEBUG_TEX(("*** s3vTexImage1D: #%i ***\n", ++times)); -#endif - -#if 1 - if (t) { -#if _TEXFLUSH - DMAFLUSH(); -#endif - s3vSwapOutTexObj( vmesa, t ); -/* - s3vDestroyTexObj( vmesa, t ); - texObj->DriverData = 0; -*/ - } -#endif - _mesa_store_teximage1d( ctx, target, level, internalFormat, - width, border, format, type, - pixels, pack, texObj, texImage ); -} - -static void s3vTexSubImage1D( GLcontext *ctx, - GLenum target, - GLint level, - GLint xoffset, - GLsizei width, - GLenum format, GLenum type, - const GLvoid *pixels, - const struct gl_pixelstore_attrib *pack, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ) -{ - s3vContextPtr vmesa = S3V_CONTEXT( ctx ); - s3vTextureObjectPtr t = (s3vTextureObjectPtr) texObj->DriverData; -#if TEX_DEBUG_ON - static unsigned int times=0; - DEBUG_TEX(("*** s3vTexSubImage1D: #%i ***\n", ++times)); -#endif - -#if 1 - if (t) { -#if _TEXFLUSH - DMAFLUSH(); -#endif - s3vSwapOutTexObj( vmesa, t ); -/* - s3vDestroyTexObj( vmesa, t ); - texObj->DriverData = 0; -*/ - } -#endif - _mesa_store_texsubimage1d(ctx, target, level, xoffset, width, - format, type, pixels, pack, texObj, - texImage); -} - -static void s3vTexImage2D( GLcontext *ctx, GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint border, - GLenum format, GLenum type, const GLvoid *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ) -{ - s3vContextPtr vmesa = S3V_CONTEXT( ctx ); - s3vTextureObjectPtr t = (s3vTextureObjectPtr) texObj->DriverData; - -#if TEX_DEBUG_ON - static unsigned int times=0; - DEBUG_TEX(("*** s3vTexImage2D: #%i ***\n", ++times)); -#endif - -#if 1 - if (t) { -#if _TEXFLUSH - DMAFLUSH(); -#endif - s3vSwapOutTexObj( vmesa, t ); -/* - s3vDestroyTexObj( vmesa, t ); - texObj->DriverData = 0; -*/ - } -#endif - _mesa_store_teximage2d( ctx, target, level, internalFormat, - width, height, border, format, type, - pixels, packing, texObj, texImage ); -} - -static void s3vTexSubImage2D( GLcontext *ctx, - GLenum target, - GLint level, - GLint xoffset, GLint yoffset, - GLsizei width, GLsizei height, - GLenum format, GLenum type, - const GLvoid *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ) -{ - s3vContextPtr vmesa = S3V_CONTEXT( ctx ); - s3vTextureObjectPtr t = (s3vTextureObjectPtr) texObj->DriverData; -#if TEX_DEBUG_ON - static unsigned int times=0; - DEBUG_TEX(("*** s3vTexSubImage2D: #%i ***\n", ++times)); -#endif - -#if 1 - if (t) { -#if _TEXFLUSH - DMAFLUSH(); -#endif - s3vSwapOutTexObj( vmesa, t ); -/* - s3vDestroyTexObj( vmesa, t ); - texObj->DriverData = 0; -*/ - } -#endif - _mesa_store_texsubimage2d(ctx, target, level, xoffset, yoffset, width, - height, format, type, pixels, packing, texObj, - texImage); -} - - -static void s3vBindTexture( GLcontext *ctx, GLenum target, - struct gl_texture_object *tObj ) -{ - s3vContextPtr vmesa = S3V_CONTEXT( ctx ); - s3vTextureObjectPtr t = (s3vTextureObjectPtr) tObj->DriverData; - GLuint cmd = vmesa->CMD; -#if TEX_DEBUG_ON - static unsigned int times=0; - DEBUG_TEX(("*** s3vBindTexture: #%i ***\n", ++times)); -#endif - - if (!t) { -/* - GLfloat bias = ctx->Texture.Unit[ctx->Texture.CurrentUnit].LodBias; -*/ - t = CALLOC_STRUCT(s3v_texture_object_t); - - /* Initialize non-image-dependent parts of the state: - */ - t->globj = tObj; -#if 0 - if (target == GL_TEXTURE_2D) { - } else - if (target == GL_TEXTURE_1D) { - } - -#if X_BYTE_ORDER == X_LITTLE_ENDIAN - t->TextureFormat = (TF_LittleEndian | -#else - t->TextureFormat = (TF_BigEndian | -#endif -#endif - t->dirty_images = ~0; - - tObj->DriverData = t; - make_empty_list( t ); -#if 0 - s3vSetTexWrapping( vmesa, t, tObj->WrapS, tObj->WrapT ); - s3vSetTexFilter( vmesa, t, tObj->MinFilter, tObj->MagFilter ); - s3vSetTexBorderColor( vmesa, t, tObj->BorderColor ); -#endif - } - - cmd = vmesa->CMD & ~MIP_MASK; - vmesa->dirty |= S3V_UPLOAD_TEX0; - vmesa->TexOffset = t->TextureBaseAddr[tObj->BaseLevel]; - vmesa->TexStride = t->Pitch; - cmd |= MIPMAP_LEVEL(t->WidthLog2); - vmesa->CMD = cmd; - vmesa->restore_primitive = -1; -#if 0 - printf("t->TextureBaseAddr[0] = 0x%x\n", t->TextureBaseAddr[0]); - printf("t->TextureBaseAddr[1] = 0x%x\n", t->TextureBaseAddr[1]); - printf("t->TextureBaseAddr[2] = 0x%x\n", t->TextureBaseAddr[2]); -#endif -} - - -static void s3vDeleteTexture( GLcontext *ctx, struct gl_texture_object *tObj ) -{ - s3vTextureObjectPtr t = (s3vTextureObjectPtr)tObj->DriverData; -#if TEX_DEBUG_ON - static unsigned int times=0; - DEBUG_TEX(("*** s3vDeleteTexture: #%i ***\n", ++times)); -#endif - - if (t) { - s3vContextPtr vmesa = S3V_CONTEXT( ctx ); - -#if _TEXFLUSH - if (vmesa) { - DMAFLUSH(); - } -#endif - - s3vDestroyTexObj( vmesa, t ); - tObj->DriverData = 0; - - } -} - -static GLboolean s3vIsTextureResident( GLcontext *ctx, - struct gl_texture_object *tObj ) -{ - s3vTextureObjectPtr t = (s3vTextureObjectPtr)tObj->DriverData; -#if TEX_DEBUG_ON - static unsigned int times=0; - DEBUG_TEX(("*** s3vIsTextureResident: #%i ***\n", ++times)); -#endif - - return (t && t->MemBlock); -} - -static void s3vInitTextureObjects( GLcontext *ctx ) -{ - /* s3vContextPtr vmesa = S3V_CONTEXT(ctx); */ - struct gl_texture_object *texObj; - GLuint tmp = ctx->Texture.CurrentUnit; -#if TEX_DEBUG_ON - static unsigned int times=0; - DEBUG_TEX(("*** s3vInitTextureObjects: #%i ***\n", ++times)); -#endif - -#if 1 - ctx->Texture.CurrentUnit = 0; - - texObj = ctx->Texture.Unit[0].CurrentTex[TEXTURE_1D_INDEX]; - s3vBindTexture( ctx, GL_TEXTURE_1D, texObj ); - - texObj = ctx->Texture.Unit[0].CurrentTex[TEXTURE_2D_INDEX]; - s3vBindTexture( ctx, GL_TEXTURE_2D, texObj ); -#endif - -#if 0 - ctx->Texture.CurrentUnit = 1; - - texObj = ctx->Texture.Unit[1].CurrentTex[TEXTURE_1D_INDEX]; - s3vBindTexture( ctx, GL_TEXTURE_1D, texObj ); - - texObj = ctx->Texture.Unit[1].CurrentTex[TEXTURE_2D_INDEX]; - s3vBindTexture( ctx, GL_TEXTURE_2D, texObj ); -#endif - - ctx->Texture.CurrentUnit = tmp; -} - - -void s3vInitTextureFuncs( GLcontext *ctx ) -{ -#if TEX_DEBUG_ON - static unsigned int times=0; - DEBUG_TEX(("*** s3vInitTextureFuncs: #%i ***\n", ++times)); -#endif - - ctx->Driver.TexEnv = s3vTexEnv; - ctx->Driver.TexImage2D = s3vTexImage2D; - ctx->Driver.TexSubImage2D = s3vTexSubImage2D; - ctx->Driver.BindTexture = s3vBindTexture; - ctx->Driver.DeleteTexture = s3vDeleteTexture; - ctx->Driver.TexParameter = s3vTexParameter; - ctx->Driver.UpdateTexturePalette = 0; - ctx->Driver.IsTextureResident = s3vIsTextureResident; - - s3vInitTextureObjects( ctx ); -} diff --git a/src/mesa/drivers/dri/s3v/s3v_tex.h b/src/mesa/drivers/dri/s3v/s3v_tex.h deleted file mode 100644 index a823fe24531..00000000000 --- a/src/mesa/drivers/dri/s3v/s3v_tex.h +++ /dev/null @@ -1,28 +0,0 @@ -/* - * Author: Max Lingua <[email protected]> - */ - -#ifndef _S3V_TEX_H -#define _S3V_TEX_H - -#define TEX_DEBUG_ON 0 - -extern void s3vUpdateTexLRU( s3vContextPtr vmesa, s3vTextureObjectPtr t ); - -#if TEX_DEBUG_ON -#define DEBUG_TEX(str) printf str -#else -#define DEBUG_TEX(str) /* str */ -#endif - -#define _TEXFLUSH 1 /* flush before uploading */ -#define _TEXLOCK 1 /* lock before writing new texures to card mem */ - /* if you turn it on you will gain stability and image - quality, but you will loose performance (~10%) */ -#define _TEXFALLBACK 0 /* fallback to software for -big- textures (slow) */ - /* turning this off, you will lose some tex (e.g. mountains - on tuxracer) but you will increase average playability */ - -#define _TEXALIGN 0x00000007 - -#endif diff --git a/src/mesa/drivers/dri/s3v/s3v_texmem.c b/src/mesa/drivers/dri/s3v/s3v_texmem.c deleted file mode 100644 index 705d105f55f..00000000000 --- a/src/mesa/drivers/dri/s3v/s3v_texmem.c +++ /dev/null @@ -1,582 +0,0 @@ -/* - * Author: Max Lingua <[email protected]> - */ - -#include <stdlib.h> -#include <stdio.h> - -#include "main/glheader.h" -#include "main/macros.h" -#include "main/mtypes.h" -#include "main/simple_list.h" -#include "main/enums.h" - -#include "main/mm.h" -#include "s3v_context.h" -#include "s3v_lock.h" -#include "s3v_tex.h" - -void s3vSwapOutTexObj(s3vContextPtr vmesa, s3vTextureObjectPtr t); -void s3vUpdateTexLRU( s3vContextPtr vmesa, s3vTextureObjectPtr t ); - - -void s3vDestroyTexObj(s3vContextPtr vmesa, s3vTextureObjectPtr t) -{ -#if TEX_DEBUG_ON - static unsigned int times=0; - DEBUG_TEX(("*** s3vDestroyTexObj: #%i ***\n", ++times)); -#endif - - if (!t) return; - -/* FIXME: useful? */ -#if _TEXFLUSH - if (vmesa) - DMAFLUSH(); -#endif - - /* This is sad - need to sync *in case* we upload a texture - * to this newly free memory... - */ - if (t->MemBlock) { - mmFreeMem(t->MemBlock); - t->MemBlock = 0; - - if (vmesa && t->age > vmesa->dirtyAge) - vmesa->dirtyAge = t->age; - } - - if (t->globj) - t->globj->DriverData = NULL; - - if (vmesa) { - if (vmesa->CurrentTexObj[0] == t) { - vmesa->CurrentTexObj[0] = 0; - vmesa->dirty &= ~S3V_UPLOAD_TEX0; - } - -#if 0 - if (vmesa->CurrentTexObj[1] == t) { - vmesa->CurrentTexObj[1] = 0; - vmesa->dirty &= ~S3V_UPLOAD_TEX1; - } -#endif - } - - remove_from_list(t); - FREE(t); -} - - -void s3vSwapOutTexObj(s3vContextPtr vmesa, s3vTextureObjectPtr t) -{ -/* int i; */ -#if TEX_DEBUG_ON - static unsigned int times=0; - DEBUG_TEX(("*** s3vSwapOutTexObj: #%i ***\n", ++times)); -#endif - - if (t->MemBlock) { - - mmFreeMem(t->MemBlock); - t->MemBlock = 0; - - if (t->age > vmesa->dirtyAge) - vmesa->dirtyAge = t->age; - - t->dirty_images = ~0; - move_to_tail(&(vmesa->SwappedOut), t); - } -} - - -/* Upload an image from mesa's internal copy. - */ - -static void s3vUploadTexLevel( s3vContextPtr vmesa, s3vTextureObjectPtr t, - int level ) -{ - __DRIscreenPrivate *sPriv = vmesa->driScreen; - const struct gl_texture_image *image = t->image[level].image; - int i,j; - int l2d; - /* int offset = 0; */ - int words; - GLuint* dest; -#if TEX_DEBUG_ON - static unsigned int times=0; -#endif - if ( !image ) return; - if (image->Data == 0) return; - - DEBUG_TEX(("*** s3vUploadTexLevel: #%i ***\n", ++times)); - DEBUG_TEX(("level = %i\n", level)); - - l2d = 5; /* 32bits per texel == 1<<5 */ -/* - if (level == 0) - ; -*/ - DEBUG_TEX(("t->image[%i].offset = 0x%x\n", - level, t->image[level].offset)); - - t->TextureBaseAddr[level] = (GLuint)(t->BufAddr + t->image[level].offset - + _TEXALIGN) & (GLuint)(~_TEXALIGN); - dest = (GLuint*)(sPriv->pFB + t->TextureBaseAddr[level]); - - DEBUG_TEX(("sPriv->pFB = 0x%x\n", sPriv->pFB)); - DEBUG_TEX(("dest = 0x%x\n", dest)); - DEBUG_TEX(("dest - sPriv->pFB = 0x%x\n", ((int)dest - (int)sPriv->pFB))); - - /* NOTE: we implicitly suppose t->texelBytes == 2 */ - - words = (image->Width * image->Height) >> 1; - - DEBUG_TEX(("\n\n")); - - switch (t->image[level].internalFormat) { - case GL_RGB: - case 3: - { - GLubyte *src = (GLubyte *)image->Data; - - DEBUG_TEX(("GL_RGB:\n")); -/* - if (level == 0) - ; -*/ - /* The UGLY way, and SLOW : use DMA FIXME ! */ - - for (i = 0; i < words; i++) { - unsigned int data; - /* data = PACK_COLOR_565(src[0],src[1],src[2]); */ - data = S3VIRGEPACKCOLOR555(src[0],src[1],src[2],255) - |(S3VIRGEPACKCOLOR555(src[3],src[4],src[5],255)<<16); - - *dest++ = data; - /* src += 3; */ - src +=6; - } - } - break; - - case GL_RGBA: - case 4: - { - GLubyte *src = (GLubyte *)image->Data; - - DEBUG_TEX(("GL_RGBA:\n")); -/* - if (level == 0) - ; -*/ - for (i = 0; i < words; i++) { - unsigned int data; - - /* data = PACK_COLOR_8888(src[0],src[1],src[2],src[3]); */ - data = S3VIRGEPACKCOLOR4444(src[0], src[1],src[2], src[3]) - | (S3VIRGEPACKCOLOR4444(src[4], src[5], src[6], src[7]) << 16); - - *dest++ = data; - /* src += 4; */ - src += 8; - } - } - break; - - case GL_LUMINANCE: - { - GLubyte *src = (GLubyte *)image->Data; - - DEBUG_TEX(("GL_LUMINANCE:\n")); -/* - if (level == 0) - ; -*/ - for (i = 0; i < words; i++) { - unsigned int data; - - /* data = PACK_COLOR_888(src[0],src[0],src[0]); */ - data = S3VIRGEPACKCOLOR4444(src[0],src[0],src[0],src[0]) - | (S3VIRGEPACKCOLOR4444(src[1],src[1],src[1],src[1]) << 16); - - *dest++ = data; - /* src ++; */ - src +=2; - } - } - break; - - case GL_INTENSITY: - { - GLubyte *src = (GLubyte *)image->Data; - - DEBUG_TEX(("GL_INTENSITY:\n")); -/* - if (level == 0) - ; -*/ - for (i = 0; i < words; i++) { - unsigned int data; - - /* data = PACK_COLOR_8888(src[0],src[0],src[0],src[0]); */ - data = S3VIRGEPACKCOLOR4444(src[0],src[0],src[0],src[0]) - | (S3VIRGEPACKCOLOR4444(src[1],src[1],src[1],src[1]) << 16); - - *dest++ = data; - /* src ++; */ - src += 2; - } - } - break; - - case GL_LUMINANCE_ALPHA: - { - GLubyte *src = (GLubyte *)image->Data; - - DEBUG_TEX(("GL_LUMINANCE_ALPHA:\n")); -/* - if (level == 0) - ; -*/ - for (i = 0; i < words; i++) { - unsigned int data; - - /* data = PACK_COLOR_8888(src[0],src[0],src[0],src[1]); */ - data = S3VIRGEPACKCOLOR4444(src[0],src[0],src[0],src[1]) - | (S3VIRGEPACKCOLOR4444(src[2],src[2],src[2],src[3]) << 16); - - *dest++ = data; - /* src += 2; */ - src += 4; - } - } - break; - - case GL_ALPHA: - { - GLubyte *src = (GLubyte *)image->Data; - - DEBUG_TEX(("GL_ALPHA:\n")); -/* - if (level == 0) - ; -*/ - for (i = 0; i < words; i++) { - unsigned int data; - - /* data = PACK_COLOR_8888(255,255,255,src[0]); */ - data = S3VIRGEPACKCOLOR4444(255,255,255,src[0]) - | (S3VIRGEPACKCOLOR4444(255,255,255,src[1]) << 16); - - *dest++ = data; - /* src += 1; */ - src += 2; - } - } - break; - - /* TODO: Translate color indices *now*: - */ - case GL_COLOR_INDEX: - { - - GLubyte *dst = (GLubyte *)(t->BufAddr + t->image[level].offset); - GLubyte *src = (GLubyte *)image->Data; - - DEBUG_TEX(("GL_COLOR_INDEX:\n")); - - for (j = 0 ; j < image->Height ; j++, dst += t->Pitch) { - for (i = 0 ; i < image->Width ; i++) { - dst[i] = src[0]; - src += 1; - } - } - } - break; - - default: - fprintf(stderr, "Not supported texture format %s\n", - _mesa_lookup_enum_by_nr(image->_BaseFormat)); - } - - DEBUG_TEX(("words = %i\n\n", words)); -} - -void s3vPrintLocalLRU( s3vContextPtr vmesa ) -{ - s3vTextureObjectPtr t; - int sz = 1 << (vmesa->s3vScreen->logTextureGranularity); - -#if TEX_DEBUG_ON - static unsigned int times=0; - DEBUG_TEX(("*** s3vPrintLocalLRU: #%i ***\n", ++times)); -#endif - - foreach( t, &vmesa->TexObjList ) { - if (!t->globj) - fprintf(stderr, "Placeholder %d at %x sz %x\n", - t->MemBlock->ofs / sz, - t->MemBlock->ofs, - t->MemBlock->size); - else - fprintf(stderr, "Texture at %x sz %x\n", - t->MemBlock->ofs, - t->MemBlock->size); - - } -} - -void s3vPrintGlobalLRU( s3vContextPtr vmesa ) -{ - int i, j; - S3VTexRegionPtr list = vmesa->sarea->texList; -#if TEX_DEBUG_ON - static unsigned int times=0; - DEBUG_TEX(("*** s3vPrintGlobalLRU: #%i ***\n", ++times)); -#endif - - for (i = 0, j = S3V_NR_TEX_REGIONS ; i < S3V_NR_TEX_REGIONS ; i++) { - fprintf(stderr, "list[%d] age %d next %d prev %d\n", - j, list[j].age, list[j].next, list[j].prev); - j = list[j].next; - if (j == S3V_NR_TEX_REGIONS) break; - } - - if (j != S3V_NR_TEX_REGIONS) - fprintf(stderr, "Loop detected in global LRU\n"); -} - - -void s3vResetGlobalLRU( s3vContextPtr vmesa ) -{ - S3VTexRegionPtr list = vmesa->sarea->texList; - int sz = 1 << vmesa->s3vScreen->logTextureGranularity; - int i; - -#if TEX_DEBUG_ON - static unsigned int times=0; - DEBUG_TEX(("*** s3vResetGlobalLRU: #%i ***\n", ++times)); -#endif - - /* (Re)initialize the global circular LRU list. The last element - * in the array (S3V_NR_TEX_REGIONS) is the sentinal. Keeping it - * at the end of the array allows it to be addressed rationally - * when looking up objects at a particular location in texture - * memory. - */ - for (i = 0 ; (i+1) * sz <= vmesa->s3vScreen->textureSize ; i++) { - list[i].prev = i-1; - list[i].next = i+1; - list[i].age = 0; - } - - i--; - list[0].prev = S3V_NR_TEX_REGIONS; - list[i].prev = i-1; - list[i].next = S3V_NR_TEX_REGIONS; - list[S3V_NR_TEX_REGIONS].prev = i; - list[S3V_NR_TEX_REGIONS].next = 0; - vmesa->sarea->texAge = 0; -} - - -void s3vUpdateTexLRU( s3vContextPtr vmesa, s3vTextureObjectPtr t ) -{ -/* - int i; - int logsz = vmesa->s3vScreen->logTextureGranularity; - int start = t->MemBlock->ofs >> logsz; - int end = (t->MemBlock->ofs + t->MemBlock->size - 1) >> logsz; - S3VTexRegionPtr list = vmesa->sarea->texList; -*/ - -#if TEX_DEBUG_ON - static unsigned int times=0; - DEBUG_TEX(("*** s3vUpdateTexLRU: #%i ***\n", ++times)); -#endif - - vmesa->texAge = ++vmesa->sarea->texAge; - - /* Update our local LRU - */ - move_to_head( &(vmesa->TexObjList), t ); - - /* Update the global LRU - */ -#if 0 - for (i = start ; i <= end ; i++) { - - list[i].in_use = 1; - list[i].age = vmesa->texAge; - - /* remove_from_list(i) - */ - list[(unsigned)list[i].next].prev = list[i].prev; - list[(unsigned)list[i].prev].next = list[i].next; - - /* insert_at_head(list, i) - */ - list[i].prev = S3V_NR_TEX_REGIONS; - list[i].next = list[S3V_NR_TEX_REGIONS].next; - list[(unsigned)list[S3V_NR_TEX_REGIONS].next].prev = i; - list[S3V_NR_TEX_REGIONS].next = i; - } -#endif -} - - -/* Called for every shared texture region which has increased in age - * since we last held the lock. - * - * Figures out which of our textures have been ejected by other clients, - * and pushes a placeholder texture onto the LRU list to represent - * the other client's textures. - */ -void s3vTexturesGone( s3vContextPtr vmesa, - GLuint offset, - GLuint size, - GLuint in_use ) -{ - s3vTextureObjectPtr t, tmp; -#if TEX_DEBUG_ON - static unsigned int times=0; - DEBUG_TEX(("*** s3vTexturesGone: #%i ***\n", ++times)); -#endif - - foreach_s ( t, tmp, &vmesa->TexObjList ) { - - if (t->MemBlock->ofs >= offset + size || - t->MemBlock->ofs + t->MemBlock->size <= offset) - continue; - - /* It overlaps - kick it off. Need to hold onto the currently bound - * objects, however. - */ - s3vSwapOutTexObj( vmesa, t ); - } - - if (in_use) { - t = (s3vTextureObjectPtr) calloc(1,sizeof(*t)); - if (!t) return; - - t->MemBlock = mmAllocMem( vmesa->texHeap, size, 0, offset); - insert_at_head( &vmesa->TexObjList, t ); - } - - /* Reload any lost textures referenced by current vertex buffer. - */ -#if 0 - if (vmesa->vertex_buffer) { - int i, j; - - fprintf(stderr, "\n\nreload tex\n"); - - for (i = 0 ; i < vmesa->statenr ; i++) { - for (j = 0 ; j < 2 ; j++) { - s3vTextureObjectPtr t = vmesa->state_tex[j][i]; - if (t) { - if (t->MemBlock == 0) - s3vUploadTexImages( vmesa, t ); - } - } - } - - /* Hard to do this with the lock held: - */ - /* S3V_FIREVERTICES( vmesa ); */ - } -#endif -} - - -/* This is called with the lock held. May have to eject our own and/or - * other client's texture objects to make room for the upload. - */ -void s3vUploadTexImages( s3vContextPtr vmesa, s3vTextureObjectPtr t ) -{ - int i; - int ofs; - int numLevels; -#if TEX_DEBUG_ON - static unsigned int times=0; - static unsigned int try=0; - - DEBUG_TEX(("*** s3vUploadTexImages: #%i ***\n", ++times)); - DEBUG_TEX(("vmesa->texHeap = 0x%x; t->totalSize = %i\n", - (unsigned int)vmesa->texHeap, t->totalSize)); -#endif - - /* Do we need to eject LRU texture objects? - */ - if (!t->MemBlock) { - - while (1) - { - /* int try = 0; */ - DEBUG_TEX(("trying to alloc mem for tex (try %i)\n", ++try)); - - t->MemBlock = mmAllocMem( vmesa->texHeap, t->totalSize, 12, 0 ); - - if (t->MemBlock) - break; - - if (vmesa->TexObjList.prev == vmesa->CurrentTexObj[0]) { -/* || vmesa->TexObjList.prev == vmesa->CurrentTexObj[1]) { - fprintf(stderr, "Hit bound texture in upload\n"); - s3vPrintLocalLRU( vmesa ); */ - return; - } - - if (vmesa->TexObjList.prev == &(vmesa->TexObjList)) { -/* fprintf(stderr, "Failed to upload texture, sz %d\n", - t->totalSize); - mmDumpMemInfo( vmesa->texHeap ); */ - return; - } - - DEBUG_TEX(("swapping out: %p\n", vmesa->TexObjList.prev)); - s3vSwapOutTexObj( vmesa, vmesa->TexObjList.prev ); - } - - ofs = t->MemBlock->ofs; - - t->BufAddr = vmesa->s3vScreen->texOffset + ofs; - - DEBUG_TEX(("ofs = 0x%x\n", ofs)); - DEBUG_TEX(("t->BufAddr = 0x%x\n", t->BufAddr)); - -/* FIXME: check if we need it */ -#if 0 - if (t == vmesa->CurrentTexObj[0]) { - vmesa->dirty |= S3V_UPLOAD_TEX0; - vmesa->restore_primitive = -1; - } -#endif - -#if 0 - if (t == vmesa->CurrentTexObj[1]) - vmesa->dirty |= S3V_UPLOAD_TEX1; -#endif - - s3vUpdateTexLRU( vmesa, t ); - } - -#if 0 - if (vmesa->dirtyAge >= GET_DISPATCH_AGE(vmesa)) - s3vWaitAgeLocked( vmesa, vmesa->dirtyAge ); -#endif - -#if _TEXLOCK - S3V_SIMPLE_FLUSH_LOCK(vmesa); -#endif - numLevels = t->lastLevel - t->firstLevel + 1; - for (i = 0 ; i < numLevels ; i++) - if (t->dirty_images & (1<<i)) - s3vUploadTexLevel( vmesa, t, i ); - - t->dirty_images = 0; -#if _TEXLOCK - S3V_SIMPLE_UNLOCK(vmesa); -#endif -} diff --git a/src/mesa/drivers/dri/s3v/s3v_texstate.c b/src/mesa/drivers/dri/s3v/s3v_texstate.c deleted file mode 100644 index 455bae6301d..00000000000 --- a/src/mesa/drivers/dri/s3v/s3v_texstate.c +++ /dev/null @@ -1,300 +0,0 @@ -/* - * Author: Max Lingua <[email protected]> - */ - -#include <stdlib.h> -#include <stdio.h> - -#include "main/glheader.h" -#include "main/macros.h" -#include "main/mtypes.h" -#include "main/simple_list.h" -#include "main/enums.h" - -#include "main/mm.h" -#include "s3v_context.h" -#include "s3v_tex.h" - - -static void s3vSetTexImages( s3vContextPtr vmesa, - struct gl_texture_object *tObj ) -{ - GLuint height, width, pitch, i, /*textureFormat,*/ log_pitch; - s3vTextureObjectPtr t = (s3vTextureObjectPtr) tObj->DriverData; - const struct gl_texture_image *baseImage = tObj->Image[0][tObj->BaseLevel]; - GLint firstLevel, lastLevel, numLevels; - GLint log2Width, log2Height; -#if TEX_DEBUG_ON - static unsigned int times=0; - DEBUG_TEX(("*** s3vSetTexImages: #%i ***\n", ++times)); -#endif - - t->texelBytes = 2; /* FIXME: always 2 ? */ - - /* Compute which mipmap levels we really want to send to the hardware. - * This depends on the base image size, GL_TEXTURE_MIN_LOD, - * GL_TEXTURE_MAX_LOD, GL_TEXTURE_BASE_LEVEL, and GL_TEXTURE_MAX_LEVEL. - * Yes, this looks overly complicated, but it's all needed. - */ - if (tObj->MinFilter == GL_LINEAR || tObj->MinFilter == GL_NEAREST) { - firstLevel = lastLevel = tObj->BaseLevel; - } - else { - firstLevel = tObj->BaseLevel + (GLint) (tObj->MinLod + 0.5); - firstLevel = MAX2(firstLevel, tObj->BaseLevel); - lastLevel = tObj->BaseLevel + (GLint) (tObj->MaxLod + 0.5); - lastLevel = MAX2(lastLevel, tObj->BaseLevel); - lastLevel = MIN2(lastLevel, tObj->BaseLevel + baseImage->MaxLog2); - lastLevel = MIN2(lastLevel, tObj->MaxLevel); - lastLevel = MAX2(firstLevel, lastLevel); /* need at least one level */ - } - - /* save these values */ - t->firstLevel = firstLevel; - t->lastLevel = lastLevel; - - numLevels = lastLevel - firstLevel + 1; - - log2Width = tObj->Image[0][firstLevel]->WidthLog2; - log2Height = tObj->Image[0][firstLevel]->HeightLog2; - - - /* Figure out the amount of memory required to hold all the mipmap - * levels. Choose the smallest pitch to accomodate the largest - * mipmap: - */ - width = tObj->Image[0][firstLevel]->Width * t->texelBytes; - for (pitch = 32, log_pitch=2 ; pitch < width ; pitch *= 2 ) - log_pitch++; - - /* All images must be loaded at this pitch. Count the number of - * lines required: - */ - for ( height = i = 0 ; i < numLevels ; i++ ) { - t->image[i].image = tObj->Image[0][firstLevel + i]; - t->image[i].offset = height * pitch; - t->image[i].internalFormat = baseImage->_BaseFormat; - height += t->image[i].image->Height; - t->TextureBaseAddr[i] = (t->BufAddr + t->image[i].offset + - _TEXALIGN) & (GLuint)(~_TEXALIGN); - } - - t->Pitch = pitch; - t->WidthLog2 = log2Width; - t->totalSize = height*pitch; - t->max_level = i-1; - vmesa->dirty |= S3V_UPLOAD_TEX0 /* | S3V_UPLOAD_TEX1*/; - vmesa->restore_primitive = -1; - DEBUG(("<><>pitch = TexStride = %i\n", pitch)); - DEBUG(("log2Width = %i\n", log2Width)); - - s3vUploadTexImages( vmesa, t ); -} - -static void s3vUpdateTexEnv( GLcontext *ctx, GLuint unit ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - const struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; - const struct gl_texture_object *tObj = texUnit->_Current; - const GLuint format = tObj->Image[0][tObj->BaseLevel]->_BaseFormat; -/* - s3vTextureObjectPtr t = (s3vTextureObjectPtr)tObj->DriverData; - GLuint tc; -*/ - GLuint alpha = 0; - GLuint cmd = vmesa->CMD; -#if TEX_DEBUG_ON - static unsigned int times=0; - DEBUG_TEX(("*** s3vUpdateTexEnv: %i ***\n", ++times)); -#endif - - cmd &= ~TEX_COL_MASK; - cmd &= ~TEX_BLEND_MAKS; -/* cmd &= ~ALPHA_BLEND_MASK; */ - - DEBUG(("format = ")); - - switch (format) { - case GL_RGB: - DEBUG_TEX(("GL_RGB\n")); - cmd |= TEX_COL_ARGB1555; - break; - case GL_LUMINANCE: - DEBUG_TEX(("GL_LUMINANCE\n")); - cmd |= TEX_COL_ARGB4444; - alpha = 1; /* FIXME: check */ - break; - case GL_ALPHA: - DEBUG_TEX(("GL_ALPHA\n")); - cmd |= TEX_COL_ARGB4444; - alpha = 1; - break; - case GL_LUMINANCE_ALPHA: - DEBUG_TEX(("GL_LUMINANCE_ALPHA\n")); - cmd |= TEX_COL_ARGB4444; - alpha = 1; - break; - case GL_INTENSITY: - DEBUG_TEX(("GL_INTENSITY\n")); - cmd |= TEX_COL_ARGB4444; - alpha = 1; - break; - case GL_RGBA: - DEBUG_TEX(("GL_RGBA\n")); - cmd |= TEX_COL_ARGB4444; - alpha = 1; - break; - case GL_COLOR_INDEX: - DEBUG_TEX(("GL_COLOR_INDEX\n")); - cmd |= TEX_COL_PAL; - break; - } - - DEBUG_TEX(("EnvMode = ")); - - switch (texUnit->EnvMode) { - case GL_REPLACE: - DEBUG_TEX(("GL_REPLACE\n")); - cmd |= TEX_REFLECT; /* FIXME */ - vmesa->_tri[1] = DO_TEX_UNLIT_TRI; /* FIXME: white tri hack */ - vmesa->_alpha_tex = ALPHA_TEX /* * alpha */; - break; - case GL_MODULATE: - DEBUG_TEX(("GL_MODULATE\n")); - cmd |= TEX_MODULATE; - vmesa->_tri[1] = DO_TEX_LIT_TRI; -#if 0 - if (alpha) - vmesa->_alpha_tex = ALPHA_TEX /* * alpha */; - else - vmesa->_alpha_tex = ALPHA_SRC /* * alpha */; -#else - vmesa->_alpha_tex = ALPHA_TEX ; -#endif - break; - case GL_ADD: - DEBUG_TEX(("DEBUG_TEX\n")); - /* do nothing ???*/ - break; - case GL_DECAL: - DEBUG_TEX(("GL_DECAL\n")); - cmd |= TEX_DECAL; - vmesa->_tri[1] = DO_TEX_LIT_TRI; - vmesa->_alpha_tex = ALPHA_OFF; - break; - case GL_BLEND: - DEBUG_TEX(("GL_BLEND\n")); - cmd |= TEX_DECAL; - vmesa->_tri[1] = DO_TEX_LIT_TRI; - vmesa->_alpha_tex = ALPHA_OFF; /* FIXME: sure? */ - break; - default: - fprintf(stderr, "unknown tex env mode"); - return; - } - - DEBUG_TEX(("\n\n vmesa->CMD was 0x%x\n", vmesa->CMD)); - DEBUG_TEX(( " vmesa->CMD is 0x%x\n\n", cmd )); - - vmesa->_alpha[1] = vmesa->_alpha_tex; - vmesa->CMD = cmd; /* | MIPMAP_LEVEL(8); */ - vmesa->restore_primitive = -1; -} - -static void s3vUpdateTexUnit( GLcontext *ctx, GLuint unit ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; - GLuint cmd = vmesa->CMD; -#if TEX_DEBUG_ON - static unsigned int times=0; - DEBUG_TEX(("*** s3vUpdateTexUnit: %i ***\n", ++times)); - DEBUG_TEX(("and vmesa->CMD was 0x%x\n", vmesa->CMD)); -#endif - - if (texUnit->_ReallyEnabled == TEXTURE_2D_BIT) - { - struct gl_texture_object *tObj = texUnit->_Current; - s3vTextureObjectPtr t = (s3vTextureObjectPtr)tObj->DriverData; - - /* Upload teximages (not pipelined) - */ - if (t->dirty_images) { -#if _TEXFLUSH - DMAFLUSH(); -#endif - s3vSetTexImages( vmesa, tObj ); - if (!t->MemBlock) { -#if _TEXFALLBACK - FALLBACK( vmesa, S3V_FALLBACK_TEXTURE, GL_TRUE ); -#endif - return; - } - } - - /* Update state if this is a different texture object to last - * time. - */ -#if 1 - if (vmesa->CurrentTexObj[unit] != t) { - vmesa->dirty |= S3V_UPLOAD_TEX0 /* << unit */; - vmesa->CurrentTexObj[unit] = t; - s3vUpdateTexLRU( vmesa, t ); /* done too often */ - } -#endif - - /* Update texture environment if texture object image format or - * texture environment state has changed. - */ - if (tObj->Image[0][tObj->BaseLevel]->_BaseFormat != - vmesa->TexEnvImageFmt[unit]) { - vmesa->TexEnvImageFmt[unit] = tObj->Image[0][tObj->BaseLevel]->_BaseFormat; - s3vUpdateTexEnv( ctx, unit ); - } -#if 1 - cmd = vmesa->CMD & ~MIP_MASK; - vmesa->dirty |= S3V_UPLOAD_TEX0 /* << unit */; - vmesa->CurrentTexObj[unit] = t; - vmesa->TexOffset = t->TextureBaseAddr[tObj->BaseLevel]; - vmesa->TexStride = t->Pitch; - cmd |= MIPMAP_LEVEL(t->WidthLog2); - - DEBUG_TEX(("\n\n>> vmesa->CMD was 0x%x\n", vmesa->CMD)); - DEBUG_TEX(( ">> vmesa->CMD is 0x%x\n\n", cmd )); - DEBUG_TEX(("t->WidthLog2 = %i\n", t->WidthLog2)); - DEBUG_TEX(("MIPMAP_LEVEL(t->WidthLog2) = 0x%x\n", MIPMAP_LEVEL(t->WidthLog2))); - - vmesa->CMD = cmd; - vmesa->restore_primitive = -1; -#endif - } - else if (texUnit->_ReallyEnabled) { /* _ReallyEnabled but != TEXTURE0_2D */ -#if _TEXFALLBACK - FALLBACK( vmesa, S3V_FALLBACK_TEXTURE, GL_TRUE ); -#endif - } - else /*if (vmesa->CurrentTexObj[unit])*/ { /* !_ReallyEnabled */ - vmesa->CurrentTexObj[unit] = 0; - vmesa->TexEnvImageFmt[unit] = 0; - vmesa->dirty &= ~(S3V_UPLOAD_TEX0<<unit); - } -} - - -void s3vUpdateTextureState( GLcontext *ctx ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - (void) vmesa; -#if TEX_DEBUG_ON - static unsigned int times=0; - DEBUG_TEX(("*** s3vUpdateTextureState: #%i ***\n", ++times)); -#endif - -#if _TEXFALLBACK - FALLBACK( vmesa, S3V_FALLBACK_TEXTURE, GL_FALSE ); -#endif - s3vUpdateTexUnit( ctx, 0 ); -#if 0 - s3vUpdateTexUnit( ctx, 1 ); -#endif -} diff --git a/src/mesa/drivers/dri/s3v/s3v_tris.c b/src/mesa/drivers/dri/s3v/s3v_tris.c deleted file mode 100644 index fafd38480c9..00000000000 --- a/src/mesa/drivers/dri/s3v/s3v_tris.c +++ /dev/null @@ -1,850 +0,0 @@ -/* - * Author: Max Lingua <[email protected]> - */ - -#include <stdio.h> -#include <stdlib.h> - -#include <sys/ioctl.h> - -#include "s3v_context.h" -#include "s3v_vb.h" -#include "s3v_tris.h" - -#include "main/glheader.h" -#include "main/mtypes.h" -#include "main/macros.h" -#include "main/colormac.h" - -#include "swrast/swrast.h" -#include "swrast_setup/swrast_setup.h" -#include "tnl/tnl.h" -#include "tnl/t_context.h" -#include "tnl/t_pipeline.h" - - -/*********************************************************************** - * Build hardware rasterization functions * - ***********************************************************************/ - -#define DO_TRI 1 -#define HAVE_RGBA 1 -#define HAVE_SPEC 0 -#define HAVE_BACK_COLORS 0 -#define HAVE_HW_FLATSHADE 1 -#define VERTEX s3vVertex -#define TAB rast_tab - -#define VERT_SET_RGBA( v, c ) \ -do { \ - UNCLAMPED_FLOAT_TO_RGBA_CHAN( v->ub4[4], c); \ -/* *(v->ub4[4]) = c; \ */ \ -} while (0) -#define VERT_COPY_RGBA( v0, v1 ) v0->ui[4] = v1->ui[4] -/* -#define VERT_COPY_RGBA1( v0, v1 ) v0->ui[4] = v1->ui[4] -*/ -#define VERT_SAVE_RGBA( idx ) color[idx] = v[idx]->ui[4] -#define VERT_RESTORE_RGBA( idx ) v[idx]->ui[4] = color[idx] - -#define S3V_OFFSET_BIT 0x01 -#define S3V_TWOSIDE_BIT 0x02 -#define S3V_UNFILLED_BIT 0x04 -#define S3V_FALLBACK_BIT 0x08 -#define S3V_MAX_TRIFUNC 0x10 - - -static struct { - tnl_points_func points; - tnl_line_func line; - tnl_triangle_func triangle; - tnl_quad_func quad; -} rast_tab[S3V_MAX_TRIFUNC]; - -#define S3V_RAST_CULL_BIT 0x01 -#define S3V_RAST_FLAT_BIT 0x02 -#define S3V_RAST_TEX_BIT 0x04 - -static s3v_point_func s3v_point_tab[0x8]; -static s3v_line_func s3v_line_tab[0x8]; -static s3v_tri_func s3v_tri_tab[0x8]; -static s3v_quad_func s3v_quad_tab[0x8]; - -#define IND (0) -#define TAG(x) x -#include "s3v_tritmp.h" - -#define IND (S3V_RAST_CULL_BIT) -#define TAG(x) x##_cull -#include "s3v_tritmp.h" - -#define IND (S3V_RAST_FLAT_BIT) -#define TAG(x) x##_flat -#include "s3v_tritmp.h" - -#define IND (S3V_RAST_CULL_BIT|S3V_RAST_FLAT_BIT) -#define TAG(x) x##_cull_flat -#include "s3v_tritmp.h" - -#define IND (S3V_RAST_TEX_BIT) -#define TAG(x) x##_tex -#include "s3v_tritmp.h" - -#define IND (S3V_RAST_CULL_BIT|S3V_RAST_TEX_BIT) -#define TAG(x) x##_cull_tex -#include "s3v_tritmp.h" - -#define IND (S3V_RAST_FLAT_BIT|S3V_RAST_TEX_BIT) -#define TAG(x) x##_flat_tex -#include "s3v_tritmp.h" - -#define IND (S3V_RAST_CULL_BIT|S3V_RAST_FLAT_BIT|S3V_RAST_TEX_BIT) -#define TAG(x) x##_cull_flat_tex -#include "s3v_tritmp.h" - -static void init_rast_tab( void ) -{ - DEBUG(("*** init_rast_tab ***\n")); - - s3v_init(); - s3v_init_cull(); - s3v_init_flat(); - s3v_init_cull_flat(); - s3v_init_tex(); - s3v_init_cull_tex(); - s3v_init_flat_tex(); - s3v_init_cull_flat_tex(); -} - -/*********************************************************************** - * Rasterization fallback helpers * - ***********************************************************************/ - - -/* This code is hit only when a mix of accelerated and unaccelerated - * primitives are being drawn, and only for the unaccelerated - * primitives. - */ - -#if 0 -static void -s3v_fallback_quad( s3vContextPtr vmesa, - const s3vVertex *v0, - const s3vVertex *v1, - const s3vVertex *v2, - const s3vVertex *v3 ) -{ - GLcontext *ctx = vmesa->glCtx; - SWvertex v[4]; - s3v_translate_vertex( ctx, v0, &v[0] ); - s3v_translate_vertex( ctx, v1, &v[1] ); - s3v_translate_vertex( ctx, v2, &v[2] ); - s3v_translate_vertex( ctx, v3, &v[3] ); - DEBUG(("s3v_fallback_quad\n")); -/* _swrast_Quad( ctx, &v[0], &v[1], &v[2], &v[3] ); */ -} - -static void -s3v_fallback_tri( s3vContextPtr vmesa, - const s3vVertex *v0, - const s3vVertex *v1, - const s3vVertex *v2 ) -{ - GLcontext *ctx = vmesa->glCtx; - SWvertex v[3]; - s3v_translate_vertex( ctx, v0, &v[0] ); - s3v_translate_vertex( ctx, v1, &v[1] ); - s3v_translate_vertex( ctx, v2, &v[2] ); - DEBUG(("s3v_fallback_tri\n")); -/* _swrast_Triangle( ctx, &v[0], &v[1], &v[2] ); */ -} - -static void -s3v_fallback_line( s3vContextPtr vmesa, - const s3vVertex *v0, - const s3vVertex *v1 ) -{ - GLcontext *ctx = vmesa->glCtx; - SWvertex v[2]; - s3v_translate_vertex( ctx, v0, &v[0] ); - s3v_translate_vertex( ctx, v1, &v[1] ); - DEBUG(("s3v_fallback_line\n")); - _swrast_Line( ctx, &v[0], &v[1] ); -} - -/* -static void -s3v_fallback_point( s3vContextPtr vmesa, - const s3vVertex *v0 ) -{ - GLcontext *ctx = vmesa->glCtx; - SWvertex v[1]; - s3v_translate_vertex( ctx, v0, &v[0] ); - _swrast_Point( ctx, &v[0] ); -} -*/ -#endif - -/*********************************************************************** - * Choose rasterization functions * - ***********************************************************************/ - -#define _S3V_NEW_RASTER_STATE (_NEW_FOG | \ - _NEW_TEXTURE | \ - _DD_NEW_TRI_SMOOTH | \ - _DD_NEW_LINE_SMOOTH | \ - _DD_NEW_POINT_SMOOTH | \ - _DD_NEW_TRI_STIPPLE | \ - _DD_NEW_LINE_STIPPLE) - -#define LINE_FALLBACK (0) -#define TRI_FALLBACK (0) - -static void s3v_nodraw_triangle(GLcontext *ctx, s3vVertex *v0, - s3vVertex *v1, s3vVertex *v2) -{ - (void) (ctx && v0 && v1 && v2); -} - -static void s3v_nodraw_quad(GLcontext *ctx, - s3vVertex *v0, s3vVertex *v1, - s3vVertex *v2, s3vVertex *v3) -{ - (void) (ctx && v0 && v1 && v2 && v3); -} - -void s3vChooseRasterState(GLcontext *ctx); - -void s3vChooseRasterState(GLcontext *ctx) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - GLuint flags = ctx->_TriangleCaps; - GLuint ind = 0; - - DEBUG(("*** s3vChooseRasterState ***\n")); - - if (ctx->Polygon.CullFlag) { - if (ctx->Polygon.CullFaceMode == GL_FRONT_AND_BACK) { - vmesa->draw_tri = (s3v_tri_func)s3v_nodraw_triangle; - vmesa->draw_quad = (s3v_quad_func)s3v_nodraw_quad; - return; - } - ind |= S3V_RAST_CULL_BIT; - /* s3v_update_cullsign(ctx); */ - } /* else vmesa->backface_sign = 0; */ - - if ( flags & DD_FLATSHADE ) - ind |= S3V_RAST_FLAT_BIT; - - if ( ctx->Texture.Unit[0]._ReallyEnabled ) { - ind |= S3V_RAST_TEX_BIT; - } - - DEBUG(("ind = %i\n", ind)); - - vmesa->draw_line = s3v_line_tab[ind]; - vmesa->draw_tri = s3v_tri_tab[ind]; - vmesa->draw_quad = s3v_quad_tab[ind]; - vmesa->draw_point = s3v_point_tab[ind]; - -#if 0 - /* Hook in fallbacks for specific primitives. CURRENTLY DISABLED - */ - - if (flags & LINE_FALLBACK) - vmesa->draw_line = s3v_fallback_line; - - if (flags & TRI_FALLBACK) { - DEBUG(("TRI_FALLBACK\n")); - vmesa->draw_tri = s3v_fallback_tri; - vmesa->draw_quad = s3v_fallback_quad; - } -#endif -} - - - - -/*********************************************************************** - * Macros for t_dd_tritmp.h to draw basic primitives * - ***********************************************************************/ - -#define TRI( v0, v1, v2 ) \ -do { \ - /* - if (DO_FALLBACK) \ - vmesa->draw_tri( vmesa, v0, v1, v2 ); \ - else */ \ - DEBUG(("TRI: max was here\n")); /* \ - s3v_draw_tex_triangle( vmesa, v0, v1, v2 ); */ \ - vmesa->draw_tri( vmesa, v0, v1, v2 ); \ -} while (0) - -#define QUAD( v0, v1, v2, v3 ) \ -do { \ - DEBUG(("QUAD: max was here\n")); \ - vmesa->draw_quad( vmesa, v0, v1, v2, v3 ); \ -} while (0) - -#define LINE( v0, v1 ) \ -do { \ - DEBUG(("LINE: max was here\n")); \ - vmesa->draw_line( vmesa, v0, v1 ); \ -} while (0) - -#define POINT( v0 ) \ -do { \ - vmesa->draw_point( vmesa, v0 ); \ -} while (0) - - -/*********************************************************************** - * Build render functions from dd templates * - ***********************************************************************/ - -/* -#define S3V_OFFSET_BIT 0x01 -#define S3V_TWOSIDE_BIT 0x02 -#define S3V_UNFILLED_BIT 0x04 -#define S3V_FALLBACK_BIT 0x08 -#define S3V_MAX_TRIFUNC 0x10 - - -static struct { - points_func points; - line_func line; - triangle_func triangle; - quad_func quad; -} rast_tab[S3V_MAX_TRIFUNC]; -*/ - -#define DO_FALLBACK (IND & S3V_FALLBACK_BIT) -#define DO_OFFSET (IND & S3V_OFFSET_BIT) -#define DO_UNFILLED (IND & S3V_UNFILLED_BIT) -#define DO_TWOSIDE (IND & S3V_TWOSIDE_BIT) -#define DO_FLAT 0 -#define DO_TRI 1 -#define DO_QUAD 1 -#define DO_LINE 1 -#define DO_POINTS 1 -#define DO_FULL_QUAD 1 - -#define HAVE_RGBA 1 -#define HAVE_SPEC 0 -#define HAVE_BACK_COLORS 0 -#define HAVE_HW_FLATSHADE 1 -#define VERTEX s3vVertex -#define TAB rast_tab - -#define DEPTH_SCALE 1.0 -#define UNFILLED_TRI unfilled_tri -#define UNFILLED_QUAD unfilled_quad -#define VERT_X(_v) _v->v.x -#define VERT_Y(_v) _v->v.y -#define VERT_Z(_v) _v->v.z -#define AREA_IS_CCW( a ) (a > 0) -#define GET_VERTEX(e) (vmesa->verts + (e<<vmesa->vertex_stride_shift)) - -#if 0 -#define VERT_SET_RGBA( v, c ) \ -do { \ -/* UNCLAMPED_FLOAT_TO_RGBA_CHAN( v->ub4[4], c) */ \ -} while (0) - -#define VERT_COPY_RGBA( v0, v1 ) v0->ui[4] = v1->ui[4] -/* -#define VERT_COPY_RGBA1( v0, v1 ) v0->ui[4] = v1->ui[4] -*/ -#define VERT_SAVE_RGBA( idx ) color[idx] = v[idx]->ui[4] -#define VERT_RESTORE_RGBA( idx ) v[idx]->ui[4] = color[idx] -#endif - -#define LOCAL_VARS(n) \ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); \ - GLuint color[n]; \ - (void) color; - - -/*********************************************************************** - * Helpers for rendering unfilled primitives * - ***********************************************************************/ - -static const GLuint hw_prim[GL_POLYGON+1] = { - PrimType_Points, - PrimType_Lines, - PrimType_Lines, - PrimType_Lines, - PrimType_Triangles, - PrimType_Triangles, - PrimType_Triangles, - PrimType_Triangles, - PrimType_Triangles, - PrimType_Triangles -}; - -static void s3vResetLineStipple( GLcontext *ctx ); -static void s3vRasterPrimitive( GLcontext *ctx, GLuint hwprim ); -static void s3vRenderPrimitive( GLcontext *ctx, GLenum prim ); -/* -extern static void s3v_lines_emit(GLcontext *ctx, GLuint start, GLuint end); -extern static void s3v_tris_emit(GLcontext *ctx, GLuint start, GLuint end); -*/ -#define RASTERIZE(x) if (vmesa->hw_primitive != hw_prim[x]) \ - s3vRasterPrimitive( ctx, hw_prim[x] ) -#define RENDER_PRIMITIVE vmesa->render_primitive -#define TAG(x) x -#define IND S3V_FALLBACK_BIT -#include "tnl_dd/t_dd_unfilled.h" -#undef IND - -/*********************************************************************** - * Generate GL render functions * - ***********************************************************************/ - -#define IND (0) -#define TAG(x) x -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (S3V_OFFSET_BIT) -#define TAG(x) x##_offset -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (S3V_TWOSIDE_BIT) -#define TAG(x) x##_twoside -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (S3V_TWOSIDE_BIT|S3V_OFFSET_BIT) -#define TAG(x) x##_twoside_offset -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (S3V_UNFILLED_BIT) -#define TAG(x) x##_unfilled -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (S3V_OFFSET_BIT|S3V_UNFILLED_BIT) -#define TAG(x) x##_offset_unfilled -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (S3V_TWOSIDE_BIT|S3V_UNFILLED_BIT) -#define TAG(x) x##_twoside_unfilled -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (S3V_TWOSIDE_BIT|S3V_OFFSET_BIT|S3V_UNFILLED_BIT) -#define TAG(x) x##_twoside_offset_unfilled -#include "tnl_dd/t_dd_tritmp.h" - - -static void init_render_tab( void ) -{ - DEBUG(("*** init_render_tab ***\n")); - - init(); - init_offset(); - init_twoside(); - init_twoside_offset(); - init_unfilled(); - init_offset_unfilled(); - init_twoside_unfilled(); - init_twoside_offset_unfilled(); -} - - -/**********************************************************************/ -/* Render unclipped begin/end objects */ -/**********************************************************************/ - -#define VERT(x) (s3vVertex *)(s3vverts + (x << shift)) - -#define RENDER_POINTS( start, count ) \ - DEBUG(("RENDER_POINTS...(ok)\n")); \ - for ( ; start < count ; start++) \ - vmesa->draw_line( vmesa, VERT(start), VERT(start) ) - /* vmesa->draw_point( vmesa, VERT(start) ) */ - -#define RENDER_LINE( v0, v1 ) \ - /* DEBUG(("RENDER_LINE...(ok)\n")); \ */ \ - vmesa->draw_line( vmesa, VERT(v0), VERT(v1) ); \ - DEBUG(("RENDER_LINE...(ok)\n")) - -#define RENDER_TRI( v0, v1, v2 ) \ - DEBUG(("RENDER_TRI...(ok)\n")); \ - vmesa->draw_tri( vmesa, VERT(v0), VERT(v1), VERT(v2) ) - -#define RENDER_QUAD( v0, v1, v2, v3 ) \ - DEBUG(("RENDER_QUAD...(ok)\n")); \ - /* s3v_draw_quad( vmesa, VERT(v0), VERT(v1), VERT(v2),VERT(v3) ) */\ - /* s3v_draw_triangle( vmesa, VERT(v0), VERT(v1), VERT(v2) ); \ - s3v_draw_triangle( vmesa, VERT(v0), VERT(v2), VERT(v3) ) */ \ - vmesa->draw_quad( vmesa, VERT(v0), VERT(v1), VERT(v2), VERT(v3) ) - -#define INIT(x) s3vRenderPrimitive( ctx, x ); -#undef LOCAL_VARS -#define LOCAL_VARS \ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); \ - const GLuint shift = vmesa->vertex_stride_shift; \ - const char *s3vverts = (char *)vmesa->verts; \ - const GLboolean stipple = ctx->Line.StippleFlag; \ - (void) stipple; -#define RESET_STIPPLE if ( stipple ) s3vResetLineStipple( ctx ); -#define RESET_OCCLUSION -#define PRESERVE_VB_DEFS -#define ELT(x) (x) -#define TAG(x) s3v_##x##_verts -#include "tnl_dd/t_dd_rendertmp.h" - - -/**********************************************************************/ -/* Render clipped primitives */ -/**********************************************************************/ - -static void s3vRenderClippedPoly( GLcontext *ctx, const GLuint *elts, - GLuint n ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - TNLcontext *tnl = TNL_CONTEXT(ctx); - GLuint prim = vmesa->render_primitive; - - DEBUG(("I AM in: s3vRenderClippedPoly\n")); - - /* Render the new vertices as an unclipped polygon. - */ - if (1) - { - GLuint *tmp = VB->Elts; - VB->Elts = (GLuint *)elts; - tnl->Driver.Render.PrimTabElts[GL_POLYGON] - ( ctx, 0, n, PRIM_BEGIN|PRIM_END ); - - VB->Elts = tmp; - } - - /* Restore the render primitive - */ -#if 1 - if (prim != GL_POLYGON) { - DEBUG(("and prim != GL_POLYGON\n")); - tnl->Driver.Render.PrimitiveNotify( ctx, prim ); - } - -#endif -} - -static void s3vRenderClippedLine( GLcontext *ctx, GLuint ii, GLuint jj ) -{ - TNLcontext *tnl = TNL_CONTEXT(ctx); - /*tnl->Driver.LineFunc = s3v_line_tab[2];*/ /* _swsetup_Line; */ - - DEBUG(("I AM in: s3vRenderClippedLine\n")); - tnl->Driver.Render.Line( ctx, ii, jj ); -} - - -/**********************************************************************/ -/* Choose render functions */ -/**********************************************************************/ - - - -#define _S3V_NEW_RENDERSTATE (_DD_NEW_TRI_UNFILLED | \ - _DD_NEW_TRI_LIGHT_TWOSIDE | \ - _DD_NEW_TRI_OFFSET) - -#define ANY_RASTER_FLAGS (DD_TRI_LIGHT_TWOSIDE|DD_TRI_OFFSET|DD_TRI_UNFILLED) - -static void s3vChooseRenderState(GLcontext *ctx) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - TNLcontext *tnl = TNL_CONTEXT(ctx); - GLuint flags = ctx->_TriangleCaps; - GLuint index = 0; - - DEBUG(("s3vChooseRenderState\n")); - - if (flags & ANY_RASTER_FLAGS) { - if (flags & DD_TRI_LIGHT_TWOSIDE) index |= S3V_TWOSIDE_BIT; - if (flags & DD_TRI_OFFSET) index |= S3V_OFFSET_BIT; - if (flags & DD_TRI_UNFILLED) index |= S3V_UNFILLED_BIT; - } - - DEBUG(("vmesa->RenderIndex = %i\n", vmesa->RenderIndex)); - DEBUG(("index = %i\n", index)); - - if (vmesa->RenderIndex != index) { - vmesa->RenderIndex = index; - - tnl->Driver.Render.Points = rast_tab[index].points; - tnl->Driver.Render.Line = rast_tab[index].line; - tnl->Driver.Render.Triangle = rast_tab[index].triangle; - tnl->Driver.Render.Quad = rast_tab[index].quad; - - if (vmesa->RenderIndex == 0) - tnl->Driver.Render.PrimTabVerts = s3v_render_tab_verts; - else - tnl->Driver.Render.PrimTabVerts = _tnl_render_tab_verts; - tnl->Driver.Render.PrimTabElts = _tnl_render_tab_elts; - tnl->Driver.Render.ClippedLine = s3vRenderClippedLine; - tnl->Driver.Render.ClippedPolygon = s3vRenderClippedPoly; - } -} - - -/**********************************************************************/ -/* High level hooks for t_vb_render.c */ -/**********************************************************************/ - - - -/* Determine the rasterized primitive when not drawing unfilled - * polygons. - * - * Used only for the default render stage which always decomposes - * primitives to trianges/lines/points. For the accelerated stage, - * which renders strips as strips, the equivalent calculations are - * performed in s3v_render.c. - */ - -static void s3vRasterPrimitive( GLcontext *ctx, GLuint hwprim ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); -/* __DRIdrawablePrivate *dPriv = vmesa->driDrawable; */ - GLuint cmd = vmesa->CMD; - - unsigned int _hw_prim = hwprim; - - DEBUG(("s3vRasterPrimitive: hwprim = 0x%x ", _hw_prim)); - -/* printf("* vmesa->CMD = 0x%x\n", vmesa->CMD); */ - - if (vmesa->hw_primitive != _hw_prim) - { - DEBUG(("(new one) ***\n")); - cmd &= ~DO_MASK; - cmd &= ~ALPHA_BLEND_MASK; - vmesa->hw_primitive = _hw_prim; - - if (_hw_prim == PrimType_Triangles) { - /* TRI */ - DEBUG(("->switching to tri\n")); - cmd |= (vmesa->_tri[vmesa->_3d_mode] | vmesa->_alpha[vmesa->_3d_mode]); - } else if (_hw_prim == PrimType_Lines - || _hw_prim == PrimType_Points) { - /* LINE */ - DEBUG(("->switching to line\n")); - cmd |= (DO_3D_LINE | vmesa->_alpha[0]); - } else { - /* ugh? */ - DEBUG(("->switching to your sis'ass\n")); - } - - DEBUG(("\n")); - - vmesa->restore_primitive = _hw_prim; - /* 0xacc16827: good value -> lightened newave!!! */ - vmesa->CMD = cmd; - CMDCHANGE(); - } -} - -static void s3vRenderPrimitive( GLcontext *ctx, GLenum prim ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - __DRIdrawablePrivate *dPriv = vmesa->driDrawable; - GLuint cmd = vmesa->CMD; - - unsigned int _hw_prim = hw_prim[prim]; - - vmesa->render_primitive = prim; - vmesa->hw_primitive = _hw_prim; - - DEBUG(("s3vRenderPrimitive #%i ", prim)); - DEBUG(("_hw_prim = 0x%x\n", _hw_prim)); - -/* printf(" vmesa->CMD = 0x%x\n", vmesa->CMD); */ - - if (_hw_prim != vmesa->restore_primitive) { - DEBUG(("_hw_prim != vmesa->restore_primitive (was 0x%x)\n", - vmesa->restore_primitive)); -#if 1 - cmd &= ~DO_MASK; - cmd &= ~ALPHA_BLEND_MASK; -/* - printf(" cmd = 0x%x\n", cmd); - printf(" vmesa->_3d_mode=%i; vmesa->_tri[vmesa->_3d_mode]=0x%x\n", - vmesa->_3d_mode, vmesa->_tri[vmesa->_3d_mode]); - printf("vmesa->alpha[0] = 0x%x; vmesa->alpha[1] = 0x%x\n", - vmesa->_alpha[0], vmesa->_alpha[1]); -*/ - if (_hw_prim == PrimType_Triangles) { /* TRI */ - DEBUG(("->switching to tri\n")); - cmd |= (vmesa->_tri[vmesa->_3d_mode] | vmesa->_alpha[vmesa->_3d_mode]); - DEBUG(("vmesa->TexStride = %i\n", vmesa->TexStride)); - DEBUG(("vmesa->TexOffset = %i\n", vmesa->TexOffset)); - DMAOUT_CHECK(3DTRI_Z_BASE, 12); - } else { /* LINE */ - DEBUG(("->switching to line\n")); - cmd |= (DO_3D_LINE | vmesa->_alpha[0]); - DMAOUT_CHECK(3DLINE_Z_BASE, 12); - } - - DMAOUT(vmesa->s3vScreen->depthOffset & 0x003FFFF8); - DMAOUT(vmesa->DestBase); - /* DMAOUT(vmesa->ScissorLR); */ - /* DMAOUT(vmesa->ScissorTB); */ - - /* NOTE: we need to restore all these values since we - * are coming back from a vmesa->restore_primitive */ - DMAOUT( (0 << 16) | (dPriv->w-1) ); - DMAOUT( (0 << 16) | (dPriv->h-1) ); - DMAOUT( (vmesa->SrcStride << 16) | vmesa->TexStride ); - DMAOUT(vmesa->SrcStride); - DMAOUT(vmesa->TexOffset); - DMAOUT(vmesa->TextureBorderColor); - DMAOUT(0); /* FOG */ - DMAOUT(0); - DMAOUT(0); - DMAOUT(cmd); - /* 0xacc16827: good value -> lightened newave!!! */ - DMAFINISH(); - - vmesa->CMD = cmd; -#endif - } - - DEBUG(("\n")); - - vmesa->restore_primitive = _hw_prim; -} - -static void s3vRunPipeline( GLcontext *ctx ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - - DEBUG(("*** s3vRunPipeline ***\n")); - - if ( vmesa->new_state ) - s3vDDUpdateHWState( ctx ); - - if (vmesa->new_gl_state) { - - if (vmesa->new_gl_state & _NEW_TEXTURE) { - s3vUpdateTextureState( ctx ); - } - - if (!vmesa->Fallback) { - if (vmesa->new_gl_state & _S3V_NEW_VERTEX) - s3vChooseVertexState( ctx ); - - if (vmesa->new_gl_state & _S3V_NEW_RASTER_STATE) - s3vChooseRasterState( ctx ); - - if (vmesa->new_gl_state & _S3V_NEW_RENDERSTATE) - s3vChooseRenderState( ctx ); - } - - vmesa->new_gl_state = 0; - - } - - _tnl_run_pipeline( ctx ); -} - -static void s3vRenderStart( GLcontext *ctx ) -{ - /* Check for projective texturing. Make sure all texcoord - * pointers point to something. (fix in mesa?) - */ - - DEBUG(("s3vRenderStart\n")); - /* s3vCheckTexSizes( ctx ); */ -} - -static void s3vRenderFinish( GLcontext *ctx ) -{ - if (0) - _swrast_flush( ctx ); /* never needed */ -} - -static void s3vResetLineStipple( GLcontext *ctx ) -{ -/* s3vContextPtr vmesa = S3V_CONTEXT(ctx); */ - - /* Reset the hardware stipple counter. - */ -/* - CHECK_DMA_BUFFER(vmesa, 1); - WRITE(vmesa->buf, UpdateLineStippleCounters, 0); -*/ -} - - -/**********************************************************************/ -/* Transition to/from hardware rasterization. */ -/**********************************************************************/ - - -void s3vFallback( s3vContextPtr vmesa, GLuint bit, GLboolean mode ) -{ - GLcontext *ctx = vmesa->glCtx; - TNLcontext *tnl = TNL_CONTEXT(ctx); - GLuint oldfallback = vmesa->Fallback; - - DEBUG(("*** s3vFallback: ")); - - if (mode) { - vmesa->Fallback |= bit; - if (oldfallback == 0) { - DEBUG(("oldfallback == 0 ***\n")); - _swsetup_Wakeup( ctx ); - _tnl_need_projected_coords( ctx, GL_TRUE ); - vmesa->RenderIndex = ~0; - } - } - else { - DEBUG(("***\n")); - vmesa->Fallback &= ~bit; - if (oldfallback == bit) { - _swrast_flush( ctx ); - tnl->Driver.Render.Start = s3vRenderStart; - tnl->Driver.Render.PrimitiveNotify = s3vRenderPrimitive; - tnl->Driver.Render.Finish = s3vRenderFinish; - tnl->Driver.Render.BuildVertices = s3vBuildVertices; - tnl->Driver.Render.ResetLineStipple = s3vResetLineStipple; - vmesa->new_gl_state |= (_S3V_NEW_RENDERSTATE| - _S3V_NEW_RASTER_STATE| - _S3V_NEW_VERTEX); - } - } -} - - -/**********************************************************************/ -/* Initialization. */ -/**********************************************************************/ - - -void s3vInitTriFuncs( GLcontext *ctx ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - TNLcontext *tnl = TNL_CONTEXT(ctx); - static int firsttime = 1; - - if (firsttime) { - init_rast_tab(); - init_render_tab(); - firsttime = 0; - } - - vmesa->RenderIndex = ~0; - - tnl->Driver.RunPipeline = s3vRunPipeline; - tnl->Driver.Render.Start = s3vRenderStart; - tnl->Driver.Render.Finish = s3vRenderFinish; - tnl->Driver.Render.PrimitiveNotify = s3vRenderPrimitive; - tnl->Driver.Render.ResetLineStipple = s3vResetLineStipple; -/* - tnl->Driver.RenderInterp = _swsetup_RenderInterp; - tnl->Driver.RenderCopyPV = _swsetup_RenderCopyPV; -*/ - tnl->Driver.Render.BuildVertices = s3vBuildVertices; -} diff --git a/src/mesa/drivers/dri/s3v/s3v_tris.h b/src/mesa/drivers/dri/s3v/s3v_tris.h deleted file mode 100644 index 0010a7fe0a7..00000000000 --- a/src/mesa/drivers/dri/s3v/s3v_tris.h +++ /dev/null @@ -1,11 +0,0 @@ -/* - * Author: Max Lingua <[email protected]> - */ - -#ifndef _S3V_TRIS_H -#define _S3V_TRIS_H - -extern void s3vDDTrifuncInit(void); -extern void s3vDDChooseTriRenderState(GLcontext *); - -#endif /* !(_S3V_TRIS_H) */ diff --git a/src/mesa/drivers/dri/s3v/s3v_tritmp.h b/src/mesa/drivers/dri/s3v/s3v_tritmp.h deleted file mode 100644 index 2321bd414ff..00000000000 --- a/src/mesa/drivers/dri/s3v/s3v_tritmp.h +++ /dev/null @@ -1,958 +0,0 @@ -/* - * Author: Max Lingua <[email protected]> - */ - -/**** MACROS start ****/ - -/* point/line macros */ - -#define LINE_VERT_VARS \ - SWvertex v[3]; \ - s3vVertex* vvv[2]; \ - int x[3], y[3], z[3]; \ - int idx[3]; \ - int dx01, dy01; \ - int delt02; \ - int deltzy, zstart; \ - int start02, end01; \ - int ystart, y01y12; \ - int i, tmp, tmp2, tmp3; \ - GLfloat ydiff, fy[3] -#define LINE_VERT_VARS_VOIDS \ - (void) v; (void) vvv; (void) x; (void) y; (void) z; (void) idx; \ - (void) dx01; (void) dy01; (void) delt02; (void) deltzy; \ - (void) zstart; (void) start02; (void) ystart; (void) y01y12; \ - (void) i; (void) tmp; (void) tmp2; (void) tmp3; (void) ydiff; (void) fy - -#define LINE_FLAT_VARS \ - int arstart, gbstart; \ - int deltarx, deltgbx, deltary, deltgby; \ - GLubyte *(col)[3] -#define LINE_FLAT_VARS_VOIDS \ - (void) arstart; (void) gbstart; (void) deltarx; (void) deltgbx; \ - (void) deltary; (void) deltgby; (void) col - -#define LINE_GOURAUD_VARS \ - int arstart, gbstart; \ - int deltary, deltgby; \ - int ctmp, ctmp2, ctmp3, ctmp4; \ - GLubyte *(col)[3] -#define LINE_GOURAUD_VARS_VOIDS \ - (void) arstart; (void) gbstart; (void) deltary; (void) deltgby; \ - (void) ctmp; (void) ctmp2; (void) ctmp3; (void) ctmp4; (void) col - -#define SORT_LINE_VERT() \ -do { \ - if(v[0].attrib[FRAG_ATTRIB_WPOS][1] <= v[1].attrib[FRAG_ATTRIB_WPOS][1]) { \ -\ - idx[0] = 0; \ - idx[1] = 1; \ -\ - } else if (v[0].attrib[FRAG_ATTRIB_WPOS][1] > v[1].attrib[FRAG_ATTRIB_WPOS][1]) { \ -\ - idx[0] = 1; \ - idx[1] = 0; \ -\ - } \ -} while(0) - -#define SET_LINE_VERT() \ -do { \ - x[0] = (v[idx[0]].attrib[FRAG_ATTRIB_WPOS][0] * 1024.0f * 1024.0f); /* 0x100000 */ \ - y[0] = fy[0] = dPriv->h - v[idx[0]].attrib[FRAG_ATTRIB_WPOS][1]; \ - z[0] = (v[idx[0]].attrib[FRAG_ATTRIB_WPOS][2]) * 1024.0f * 32.0f; /* 0x8000; */ \ -\ - x[1] = (v[idx[1]].attrib[FRAG_ATTRIB_WPOS][0] * 1024.0f * 1024.0f); /* 0x100000 */ \ - y[1] = dPriv->h - v[idx[1]].attrib[FRAG_ATTRIB_WPOS][1]; \ - z[1] = (v[idx[1]].attrib[FRAG_ATTRIB_WPOS][2]) * 1024.0f * 32.0f; /* 0x8000 */ \ -} while(0) - -#define SET_LINE_XY() \ -do { \ - tmp = v[idx[0]].attrib[FRAG_ATTRIB_WPOS][0]; \ - tmp2 = v[idx[1]].attrib[FRAG_ATTRIB_WPOS][0]; \ -\ - dx01 = x[0] - x[1]; \ - dy01 = y[0] - y[1]; \ -\ - ydiff = fy[0] - (float)y[0]; \ - ystart = y[0]; \ - y01y12 = dy01 + 1; \ -} while (0) - -#define SET_LINE_DIR() \ -do { \ - if (tmp2 > tmp) { \ - y01y12 |= 0x80000000; \ - tmp3 = tmp2-tmp; \ - } else { \ - tmp3 = tmp-tmp2; \ - } \ -\ - end01 = ((tmp << 16) | tmp2); \ -\ - if (dy01) \ - delt02 = -(dx01/dy01); \ - else \ - delt02 = 0; \ -\ - if (dy01 > tmp3) { /* Y MAJ */ \ - /* NOTE: tmp3 always >=0 */ \ - start02 = x[0]; \ - } else if (delt02 >= 0){ /* X MAJ - positive delta */ \ - start02 = x[0] + delt02/2; \ - dy01 = tmp3; /* could be 0 */ \ - } else { /* X MAJ - negative delta */ \ - start02 = x[0] + delt02/2 + ((1 << 20) - 1); \ - dy01 = tmp3; /* could be 0 */ \ - } \ -} while(0) - -#define SET_LINE_Z() \ -do { \ - zstart = z[0]; \ -\ - if (dy01) { \ - deltzy = (z[1] - z[0])/dy01; \ - } else { \ - deltzy = 0; /* dy01 = tmp3 = 0 (it's a point)*/ \ - } \ -} while (0) - -#define SET_LINE_FLAT_COL() \ -do { \ - col[0] = &(v[idx[0]].color[0]); \ - deltarx = deltary = deltgbx = deltgby = 0; \ - gbstart = (((col[0][1]) << 23) | ((col[0][2]) << 7)); \ - arstart = (((col[0][3]) << 23) | ((col[0][0]) << 7)); \ -} while(0) - -#define SET_LINE_GOURAUD_COL() \ -do { \ - col[0] = &(v[idx[0]].color[0]); \ - col[1] = &(v[idx[1]].color[0]); \ -\ - vvv[0] = _v0; \ - vvv[1] = _v1; \ -\ - for (i=0; i<2; i++) { \ - /* FIXME: swapped ! */ \ - col[i][0] = vvv[!idx[i]]->v.color.red; \ - col[i][1] = vvv[!idx[i]]->v.color.green; \ - col[i][2] = vvv[!idx[i]]->v.color.blue; \ - col[i][3] = vvv[!idx[i]]->v.color.alpha; \ - } \ -\ - if (dy01) { \ -\ - ctmp = ((col[0][1] - col[1][1]) << 7) / dy01; \ - ctmp2 = ((col[0][2] - col[1][2]) << 7) / dy01; \ - deltgby = ((ctmp << 16) & 0xFFFF0000) | (ctmp2 & 0xFFFF); \ -\ - ctmp3 = ((col[0][3] - col[1][3]) << 7) / dy01; \ - ctmp4 = ((col[0][0] - col[1][0]) << 7) / dy01; \ - deltary = ((ctmp3 << 16) & 0xFFFF0000) | (ctmp4 & 0xFFFF); \ - } else { \ - ctmp = ((col[1][1] - col[0][1]) << 7); \ - ctmp2 = ((col[1][2] - col[0][2]) << 7); \ - deltgby = ((ctmp << 16) & 0xFFFF0000) | (ctmp2 & 0xFFFF); \ -\ - ctmp3 = ((col[1][3] - col[0][3]) << 7); \ - ctmp4 = ((col[1][0] - col[0][0]) << 7); \ - deltary = ((ctmp3 << 16) & 0xFFFF0000) | (ctmp4 & 0xFFFF); \ - deltgby = deltary = 0; \ - } \ -\ - idx[0] = 1; /* FIXME: swapped */ \ -\ - gbstart = \ - (((int)((ydiff * ctmp) + (col[idx[0]][1] << 7)) << 16) & 0x7FFF0000) \ - | ((int)((ydiff * ctmp2) + (col[idx[0]][2] << 7)) & 0x7FFF); \ - arstart = \ - (((int)((ydiff * ctmp3) + (col[idx[0]][3] << 7)) << 16) & 0x7FFF0000) \ - | ((int)((ydiff * ctmp4) + (col[idx[0]][0] << 7)) & 0x7FFF); \ -} while(0) - -#define SEND_LINE_COL() \ -do { \ - DMAOUT(deltgby); \ - DMAOUT(deltary); \ - DMAOUT(gbstart); \ - DMAOUT(arstart); \ -} while (0) - -#define SEND_LINE_VERT() \ -do { \ - DMAOUT(deltzy); \ - DMAOUT(zstart); \ - DMAOUT(0); \ - DMAOUT(0); \ - DMAOUT(0); \ - DMAOUT(end01); \ - DMAOUT(delt02); \ - DMAOUT(start02); \ - DMAOUT(ystart); \ - DMAOUT(y01y12); \ -} while (0) - - -/* tri macros (mostly stolen from utah-glx...) */ - -#define VERT_VARS \ - SWvertex v[3]; \ - int x[3], y[3], z[3]; \ - int idx[3]; \ - int dx01, dy01; \ - int dx02, dy02; \ - int dx12, dy12; \ - int delt01, delt02, delt12; \ - int deltzx, deltzy, zstart; \ - int start02, end01, end12; \ - int ystart, y01y12; \ - int i, tmp, lr; \ - GLfloat ydiff, fy[3] -#define VERT_VARS_VOIDS \ - (void) v; (void) x; (void) y; (void) z; (void) idx; (void) dx01; \ - (void) dy01; (void) dx02; (void) dy02; (void) dx12; (void) dy12; \ - (void) delt01; (void) delt02; (void) delt12; (void) deltzx; \ - (void) deltzy; (void) zstart; (void) start02; (void) end01; \ - (void) end12; (void) ystart; (void) y01y12; (void) i; (void) tmp; \ - (void) lr; (void) ydiff; (void) fy - -#define GOURAUD_VARS \ - int arstart, gbstart; \ - int deltarx, deltgbx, deltary, deltgby; \ - int ctmp, ctmp2, ctmp3, ctmp4; \ - GLubyte *(col)[3] -#define GOURAUD_VARS_VOIDS \ - (void) arstart; (void) gbstart; (void) deltarx; (void) deltgbx; \ - (void) deltary; (void) deltgby; (void) ctmp; (void) ctmp2; \ - (void) ctmp3; (void) ctmp4; (void) col - -#define FLAT_VARS \ - int arstart, gbstart; \ - int deltarx, deltgbx, deltary, deltgby; \ - GLubyte *(col)[3] -#define FLAT_VARS_VOIDS \ - (void) arstart; (void) gbstart; (void) deltarx; (void) deltgbx; \ - (void) deltary; (void) deltgby; (void) col - -#define TEX_VARS \ - int u0, u1, u2; \ - GLfloat ru0, ru1, ru2; \ - int v0, v1, v2; \ - GLfloat rv0, rv1, rv2; \ - GLfloat w0, w1, w2; \ - GLfloat rw0, rw1, rw2; \ - int baseu, basev; \ - int d0, d1, d2; \ - int deltdx, deltvx, deltux, deltdy, deltvy, deltuy; \ - int deltwx, deltwy; \ - int rbaseu, rbasev; \ - int dstart, ustart, wstart, vstart; \ - static int stmp = 0; \ - s3vTextureObjectPtr t -#define TEX_VARS_VOIDS \ - (void) u0; (void) u1; (void) u2; (void) ru0; (void) ru1; (void) ru2; \ - (void) v0; (void) v1; (void) v2; (void) rv0; (void) rv1; (void) rv2; \ - (void) w0; (void) w1; (void) w2; (void) rw0; (void) rw1; (void) rw2; \ - (void) baseu; (void) basev; (void) d0; (void) d1; (void) d2; \ - (void) deltdx; (void) deltvx; (void) deltux; (void) deltdy; \ - (void) deltuy; (void) deltwx; (void) deltwy; (void) rbaseu; \ - (void) rbasev; (void) dstart; (void) ustart; (void) wstart; \ - (void) vstart; (void) stmp; (void) t - -#define SORT_VERT() \ -do { \ - for (i=0; i<3; i++) \ - fy[i] = v[i].attrib[FRAG_ATTRIB_WPOS][1]; \ -\ - if (fy[1] > fy[0]) { /* (fy[1] > fy[0]) */ \ -\ - if (fy[2] > fy[0]) { \ - idx[0] = 0; \ - if (fy[1] > fy[2]) { \ - idx[1] = 2; \ - idx[2] = 1; \ - } else { \ - idx[1] = 1; \ - idx[2] = 2; \ - } \ - } else { \ - idx[0] = 2; \ - idx[1] = 0; \ - idx[2] = 1; \ - } \ - } else { /* (fy[1] < y[0]) */ \ - if (fy[2] > fy[0]) { \ - idx[0] = 1; \ - idx[1] = 0; \ - idx[2] = 2; \ - } else { \ - idx[2] = 0; \ - if (fy[2] > fy[1]) { \ - idx[0] = 1; \ - idx[1] = 2; \ - } else { \ - idx[0] = 2; \ - idx[1] = 1; \ - } \ - } \ - } \ -} while(0) - -#define SET_VERT() \ -do { \ - for (i=0; i<3; i++) \ - { \ - x[i] = ((v[idx[i]].attrib[FRAG_ATTRIB_WPOS][0]) * /* 0x100000*/ 1024.0 * 1024.0); \ - y[i] = fy[i] = (dPriv->h - v[idx[i]].attrib[FRAG_ATTRIB_WPOS][1]); \ - z[i] = ((v[idx[i]].attrib[FRAG_ATTRIB_WPOS][2]) * /* 0x8000 */ 1024.0 * 32.0); \ - } \ -\ - ydiff = fy[0] - (float)y[0]; \ -\ - ystart = y[0]; \ -\ - dx12 = x[2] - x[1]; \ - dy12 = y[1] - y[2]; \ - dx01 = x[1] - x[0]; \ - dy01 = y[0] - y[1]; \ - dx02 = x[2] - x[0]; \ - dy02 = y[0] - y[2]; \ -\ - delt01 = delt02 = delt12 = 0; \ -} while (0) - - -#define SET_XY() \ -do { \ - if (dy01) delt01 = dx01 / dy01; \ - if (dy12) delt12 = dx12 / dy12; \ - delt02 = dx02 / dy02; \ -\ - start02 = x[0] + (ydiff * delt02); \ - end01 = x[0] + (ydiff * delt01); \ - end12 = x[1] + ((fy[1] - (GLfloat)y[1]) * delt12); \ -} while (0) - -#define SET_DIR() \ -do { \ - tmp = x[1] - (dy01 * delt02 + x[0]); \ - if (tmp > 0) { \ - lr = 0x80000000; \ - } else { \ - tmp *= -1; \ - lr = 0; \ - } \ - tmp >>= 20; \ -\ - y01y12 = ((((y[0] - y[1]) & 0x7FF) << 16) \ - | ((y[1] - y[2]) & 0x7FF) | lr); \ -} while (0) - -#define SET_Z() \ -do { \ - deltzy = (z[2] - z[0]) / dy02; \ - if (tmp) { \ - deltzx = (z[1] - (dy01 * deltzy + z[0])) / tmp; \ - } else { \ - deltzx = 0; \ - } \ - zstart = (deltzy * ydiff) + z[0]; \ -} while (0) - -#define SET_FLAT_COL() \ -do { \ - col[0] = &(v[0].color[0]); \ - deltarx = deltary = deltgbx = deltgby = 0; \ - gbstart = (((col[0][1]) << 23) | ((col[0][2]) << 7)); \ - arstart = (((col[0][3]) << 23) | ((col[0][0]) << 7)); \ -} while(0) - -#define SET_GOURAUD_COL() \ -do { \ - col[0] = &(v[idx[0]].color[0]); \ - col[1] = &(v[idx[1]].color[0]); \ - col[2] = &(v[idx[2]].color[0]); \ -\ - ctmp = ((col[2][3] - col[0][3]) << 7) / dy02; \ - ctmp2 = ((col[2][0] - col[0][0]) << 7) / dy02; \ - deltary = ((ctmp << 16) & 0xFFFF0000) | (ctmp2 & 0xFFFF); \ - ctmp3 = ((col[2][1] - col[0][1]) << 7) / dy02; \ - ctmp4 = ((col[2][2] - col[0][2]) << 7) / dy02; \ - deltgby = ((ctmp3 << 16) & 0xFFFF0000) | (ctmp4 & 0xFFFF); \ - gbstart = \ - (((int)((ydiff * ctmp3) + (col[0][1] << 7)) << 16) & 0x7FFF0000) \ - | ((int)((ydiff * ctmp4) + (col[0][2] << 7)) & 0x7FFF); \ - arstart = \ - (((int)((ydiff * ctmp) + (col[0][3] << 7)) << 16) & 0x7FFF0000) \ - | ((int)((ydiff * ctmp2) + (col[0][0] << 7)) & 0x7FFF); \ - if (tmp) { \ - int ax, rx, gx, bx; \ - ax = ((col[1][3] << 7) - (dy01 * ctmp + (col[0][3] << 7))) / tmp; \ - rx = ((col[1][0] << 7) - (dy01 * ctmp2 + (col[0][0] << 7))) / tmp; \ - gx = ((col[1][1] << 7) - (dy01 * ctmp3 + (col[0][1] << 7))) / tmp; \ - bx = ((col[1][2] << 7) - (dy01 * ctmp4 + (col[0][2] << 7))) / tmp; \ - deltarx = ((ax << 16) & 0xFFFF0000) | (rx & 0xFFFF); \ - deltgbx = ((gx << 16) & 0xFFFF0000) | (bx & 0xFFFF); \ - } else { \ - deltgbx = deltarx = 0; \ - } \ -} while (0) - -#define SET_TEX_VERT() \ -do { \ - t = ((s3vTextureObjectPtr) \ - ctx->Texture.Unit[0]._Current->DriverData); \ - deltwx = deltwy = wstart = deltdx = deltdy = dstart = 0; \ -\ - u0 = (v[idx[0]].attrib[FRAG_ATTRIB_TEX0][0] \ - * (GLfloat)(t->image[0].image->Width) * 256.0); \ - u1 = (v[idx[1]].attrib[FRAG_ATTRIB_TEX0][0] \ - * (GLfloat)(t->globj->Image[0][0]->Width) * 256.0); \ - u2 = (v[idx[2]].attrib[FRAG_ATTRIB_TEX0][0] \ - * (GLfloat)(t->globj->Image[0][0]->Width) * 256.0); \ - v0 = (v[idx[0]].attrib[FRAG_ATTRIB_TEX0][1] \ - * (GLfloat)(t->globj->Image[0][0]->Height) * 256.0); \ - v1 = (v[idx[1]].attrib[FRAG_ATTRIB_TEX0][1] \ - * (GLfloat)(t->globj->Image[0][0]->Height) * 256.0); \ - v2 = (v[idx[2]].attrib[FRAG_ATTRIB_TEX0][1] \ - * (GLfloat)(t->globj->Image[0][0]->Height) * 256.0); \ -\ - w0 = (v[idx[0]].attrib[FRAG_ATTRIB_WPOS][3]); \ - w1 = (v[idx[1]].attrib[FRAG_ATTRIB_WPOS][3]); \ - w2 = (v[idx[2]].attrib[FRAG_ATTRIB_WPOS][3]); \ -} while (0) - -#define SET_BASEUV() \ -do { \ - if (u0 < u1) { \ - if (u0 < u2) { \ - baseu = u0; \ - } else { \ - baseu = u2; \ - } \ - } else { \ - if (u1 < u2) { \ - baseu = u1; \ - } else { \ - baseu = u2; \ - } \ - } \ -\ - if (v0 < v1) { \ - if (v0 < v2) { \ - basev = v0; \ - } else { \ - basev = v2; \ - } \ - } else { \ - if (v1 < v2) { \ - basev = v1; \ - } else { \ - basev = v2; \ - } \ - } \ -} while (0) - - -#define SET_RW() \ -do { \ - /* GLfloat minW; \ -\ - if (w0 < w1) { \ - if (w0 < w2) { \ - minW = w0; \ - } else { \ - minW = w2; \ - } \ - } else { \ - if (w1 < w2) { \ - minW = w1; \ - } else { \ - minW = w2; \ - } \ - } */ \ -\ - rw0 = (512.0 * w0); \ - rw1 = (512.0 * w1); \ - rw2 = (512.0 * w2); \ -} while (0) - -#define SET_D() \ -do { \ - GLfloat sxy, suv; \ - int lev; \ -\ - suv = (v[idx[0]].attrib[FRAG_ATTRIB_TEX0][0] - \ - v[idx[2]].attrib[FRAG_ATTRIB_TEX0][0]) * \ - (v[idx[1]].attrib[FRAG_ATTRIB_TEX0][1] - \ - v[idx[2]].attrib[FRAG_ATTRIB_TEX0][1]) - \ - (v[idx[1]].attrib[FRAG_ATTRIB_TEX0][0] - \ - v[idx[2]].attrib[FRAG_ATTRIB_TEX0][0]) * \ - (v[idx[0]].attrib[FRAG_ATTRIB_TEX0][1] - \ - v[idx[2]].attrib[FRAG_ATTRIB_TEX0][2]); \ -\ - sxy = (v[idx[0]].attrib[FRAG_ATTRIB_TEX0][0] - \ - v[idx[2]].attrib[FRAG_ATTRIB_TEX0][0]) * \ - (v[idx[1]].attrib[FRAG_ATTRIB_TEX0][1] - \ - v[idx[2]].attrib[FRAG_ATTRIB_TEX0][1]) - \ - (v[idx[1]].attrib[FRAG_ATTRIB_TEX0][0] - \ - v[idx[2]].attrib[FRAG_ATTRIB_TEX0][0]) * \ - (v[idx[0]].attrib[FRAG_ATTRIB_TEX0][1] - \ - v[idx[2]].attrib[FRAG_ATTRIB_TEX0][2]); \ -\ - if (sxy < 0) sxy *= -1.0; \ - if (suv < 0) suv *= -1.0; \ -\ - lev = *(int*)&suv - *(int *)&sxy; \ - if (lev < 0) \ - lev = 0; \ - else \ - lev >>=23; \ - dstart = (lev << 27); \ -} while (0) - -#define SET_UVWD() \ -do { \ - SET_BASEUV(); \ - SET_RW(); \ - SET_D(); \ - ru0 = (((u0 - baseu) * rw0)); \ - ru1 = (((u1 - baseu) * rw1)); \ - ru2 = (((u2 - baseu) * rw2)); \ - rv0 = (((v0 - basev) * rw0)); \ - rv1 = (((v1 - basev) * rw1)); \ - rv2 = (((v2 - basev) * rw2)); \ -\ - while (baseu < 0) { baseu += (t->globj->Image[0][0]->Width << 8); } \ - while (basev < 0) { basev += (t->globj->Image[0][0]->Height << 8); } \ -\ - if (!(baseu & 0xFF)) \ - { baseu = (baseu >> 8); } \ - else \ - { baseu = (baseu >> 8) + 1; } \ -\ - if ((basev & 0x80) || !(basev & 0xFF)) \ - { basev = (basev >> 8); } \ - else \ - { basev = (basev >> 8) - 1; } \ -\ - rbaseu = (baseu) << (16 - t->globj->Image[0][0]->WidthLog2); \ - rbasev = (basev) << (16 - t->globj->Image[0][0]->WidthLog2); \ - deltuy = (((ru2 - ru0) / dy02)); \ - deltvy = (((rv2 - rv0) / dy02)); \ - rw0 *= (1024.0 * 512.0); \ - rw1 *= (1024.0 * 512.0); \ - rw2 *= (1024.0 * 512.0); \ - deltwy = ((rw2 - rw0) / dy02); \ - if (tmp) { \ - deltux = ((ru1 - (dy01 * deltuy + ru0)) / tmp); \ - deltvx = ((rv1 - (dy01 * deltvy + rv0)) / tmp); \ - deltwx = ((rw1 - (dy01 * deltwy + rw0)) / tmp); \ - } else { deltux = deltvx = deltwx = 0; } \ - ustart = (deltuy * ydiff) + (ru0); \ - vstart = (deltvy * ydiff) + (rv0); \ - wstart = (deltwy * ydiff) + (rw0); \ -} while (0) - -#define SEND_UVWD() \ -do { \ - DMAOUT((rbasev & 0xFFFF)); \ - DMAOUT((0xa0000000 | (rbaseu & 0xFFFF))); \ - DMAOUT(deltwx); \ - DMAOUT(deltwy); \ - DMAOUT(wstart); \ - DMAOUT(deltdx); \ - DMAOUT(deltvx); \ - DMAOUT(deltux); \ - DMAOUT(deltdy); \ - DMAOUT(deltvy); \ - DMAOUT(deltuy); \ - DMAOUT(dstart); \ - DMAOUT(vstart); \ - DMAOUT(ustart); \ -} while (0) - -#define SEND_VERT() \ -do { \ - DMAOUT(deltzx); \ - DMAOUT(deltzy); \ - DMAOUT(zstart); \ - DMAOUT(delt12); \ - DMAOUT(end12); \ - DMAOUT(delt01); \ - DMAOUT(end01); \ - DMAOUT(delt02); \ - DMAOUT(start02); \ - DMAOUT(ystart); \ - DMAOUT(y01y12); \ -} while (0) - -#define SEND_COL() \ -do { \ - DMAOUT(deltgbx); \ - DMAOUT(deltarx); \ - DMAOUT(deltgby); \ - DMAOUT(deltary); \ - DMAOUT(gbstart); \ - DMAOUT(arstart); \ -} while (0) - -/**** MACROS end ****/ - - - - -static void TAG(s3v_point)( s3vContextPtr vmesa, - const s3vVertex *_v0 ) -{ -} - -static void TAG(s3v_line)( s3vContextPtr vmesa, - const s3vVertex *_v0, - const s3vVertex *_v1 ) -{ - GLcontext *ctx = vmesa->glCtx; - __DRIdrawablePrivate *dPriv = vmesa->driDrawable; - - LINE_VERT_VARS; -#if (IND & S3V_RAST_FLAT_BIT) - LINE_FLAT_VARS; -#else - LINE_GOURAUD_VARS; -#endif -#if (IND & S3V_RAST_CULL_BIT) - GLfloat cull; - (void) cull; -#endif - LINE_VERT_VARS_VOIDS; -#if (IND & S3V_RAST_FLAT_BIT) - LINE_FLAT_VARS_VOIDS; -#else - LINE_GOURAUD_VARS_VOIDS; -#endif - - DEBUG(("*** s3v_line: ")); -#if (IND & S3V_RAST_CULL_BIT) - DEBUG(("cull ")); -#endif -#if (IND & S3V_RAST_FLAT_BIT) - DEBUG(("flat ")); -#endif - - DEBUG(("***\n")); - -#if 0 - s3v_print_vertex(ctx, _v0); - s3v_print_vertex(ctx, _v1); -#endif - - s3v_translate_vertex( ctx, _v0, &v[0] ); - s3v_translate_vertex( ctx, _v1, &v[1] ); - -#if (IND & S3V_RAST_CULL_BIT) - /* FIXME: should we cull lines too? */ -#endif - (void)v; /* v[0]; v[1]; */ - - SORT_LINE_VERT(); - SET_LINE_VERT(); - - SET_LINE_XY(); - SET_LINE_DIR(); - SET_LINE_Z(); - -#if (IND & S3V_RAST_FLAT_BIT) - SET_LINE_FLAT_COL(); -#else - SET_LINE_GOURAUD_COL(); -#endif - - DMAOUT_CHECK(3DLINE_GBD, 15); - SEND_LINE_COL(); - DMAOUT(0); - SEND_LINE_VERT(); - DMAFINISH(); -} - -static void TAG(s3v_triangle)( s3vContextPtr vmesa, - const s3vVertex *_v0, - const s3vVertex *_v1, - const s3vVertex *_v2 ) -{ - GLcontext *ctx = vmesa->glCtx; - __DRIdrawablePrivate *dPriv = vmesa->driDrawable; - - VERT_VARS; -#if (IND & S3v_RAST_FLAT_BIT) - FLAT_VARS; -#else - GOURAUD_VARS; -#endif -#if (IND & S3V_RAST_TEX_BIT) - TEX_VARS; -#endif -#if (IND & S3V_RAST_CULL_BIT) - GLfloat cull; -#endif - VERT_VARS_VOIDS; -#if (IND & S3v_RAST_FLAT_BIT) - FLAT_VARS_VOIDS; -#else - GOURAUD_VARS_VOIDS; -#endif -#if (IND & S3V_RAST_TEX_BIT) - TEX_VARS_VOIDS; -#endif - - DEBUG(("*** s3v_triangle: ")); -#if (IND & S3V_RAST_CULL_BIT) - DEBUG(("cull ")); -#endif -#if (IND & S3V_RAST_FLAT_BIT) - DEBUG(("flat ")); -#endif -#if (IND & S3V_RAST_TEX_BIT) - DEBUG(("tex ")); -#endif - -DEBUG(("***\n")); - -#if 0 - s3v_print_vertex(ctx, _v0); - s3v_print_vertex(ctx, _v1); - s3v_print_vertex(ctx, _v2); -#endif - - s3v_translate_vertex( ctx, _v0, &v[0] ); - s3v_translate_vertex( ctx, _v1, &v[1] ); - s3v_translate_vertex( ctx, _v2, &v[2] ); - -#if (IND & S3V_RAST_CULL_BIT) - cull = vmesa->backface_sign * - ((v[1].attrib[FRAG_ATTRIB_WPOS][0] - v[0].attrib[FRAG_ATTRIB_WPOS][0]) * (v[0].attrib[FRAG_ATTRIB_WPOS][1] - v[2].attrib[FRAG_ATTRIB_WPOS][1]) + - (v[1].attrib[FRAG_ATTRIB_WPOS][1] - v[0].attrib[FRAG_ATTRIB_WPOS][1]) * (v[2].attrib[FRAG_ATTRIB_WPOS][0] - v[0].attrib[FRAG_ATTRIB_WPOS][0])); - - if (cull < vmesa->cull_zero /* -0.02f */) return; -#endif - - (void)v; /* v[0]; v[1]; v[2]; */ - - SORT_VERT(); - SET_VERT(); - - if (dy02 == 0) return; - - SET_XY(); - SET_DIR(); - SET_Z(); - -#if (IND & S3V_RAST_TEX_BIT) - SET_TEX_VERT(); - SET_UVWD(); -#endif - -#if (IND & S3V_RAST_FLAT_BIT) - SET_FLAT_COL(); -#else - SET_GOURAUD_COL(); -#endif - -#if (IND & S3V_RAST_TEX_BIT) - DMAOUT_CHECK(3DTRI_BASEV, 31); - SEND_UVWD(); - SEND_COL(); - SEND_VERT(); - DMAFINISH(); -#else - DMAOUT_CHECK(3DTRI_GBX, 17); - SEND_COL(); - SEND_VERT(); - DMAFINISH(); -#endif -} - -static void TAG(s3v_quad)( s3vContextPtr vmesa, - const s3vVertex *_v0, - const s3vVertex *_v1, - const s3vVertex *_v2, - const s3vVertex *_v3 ) -{ - GLcontext *ctx = vmesa->glCtx; - __DRIdrawablePrivate *dPriv = vmesa->driDrawable; - - SWvertex temp_v[4]; - VERT_VARS; -#if (IND & S3v_RAST_FLAT_BIT) - FLAT_VARS; -#else - GOURAUD_VARS; -#endif -#if (IND & S3V_RAST_TEX_BIT) - TEX_VARS; -#endif -#if (IND & S3V_RAST_CULL_BIT) - GLfloat cull; -#endif - VERT_VARS_VOIDS; -#if (IND & S3v_RAST_FLAT_BIT) - FLAT_VARS_VOIDS; -#else - GOURAUD_VARS_VOIDS; -#endif -#if (IND & S3V_RAST_TEX_BIT) - TEX_VARS_VOIDS; -#endif - - DEBUG(("*** s3v_quad: ")); -#if (IND & S3V_RAST_CULL_BIT) - DEBUG(("cull ")); - /* printf(""); */ /* speed trick */ -#endif -#if (IND & S3V_RAST_FLAT_BIT) - DEBUG(("flat ")); -#endif -#if (IND & S3V_RAST_TEX_BIT) - DEBUG(("tex ")); -#endif - - DEBUG(("***\n")); - -#if 0 - s3v_print_vertex(ctx, _v0); - s3v_print_vertex(ctx, _v1); - s3v_print_vertex(ctx, _v2); - s3v_print_vertex(ctx, _v3); -#endif - s3v_translate_vertex( ctx, _v0, &temp_v[0] ); - s3v_translate_vertex( ctx, _v1, &temp_v[1] ); - s3v_translate_vertex( ctx, _v2, &temp_v[2] ); - s3v_translate_vertex( ctx, _v3, &temp_v[3] ); - - /* FIRST TRI (0,1,2) */ - - /* ROMEO */ - /* printf(""); */ /* speed trick (a) [turn on if (a) is return]*/ - - v[0] = temp_v[0]; - v[1] = temp_v[1]; - v[2] = temp_v[2]; - -#if (IND & S3V_RAST_CULL_BIT) - cull = vmesa->backface_sign * - ((v[1].attrib[FRAG_ATTRIB_WPOS][0] - v[0].attrib[FRAG_ATTRIB_WPOS][0]) * (v[0].attrib[FRAG_ATTRIB_WPOS][1] - v[2].attrib[FRAG_ATTRIB_WPOS][1]) + - (v[1].attrib[FRAG_ATTRIB_WPOS][1] - v[0].attrib[FRAG_ATTRIB_WPOS][1]) * (v[2].attrib[FRAG_ATTRIB_WPOS][0] - v[0].attrib[FRAG_ATTRIB_WPOS][0])); - - if (cull < vmesa->cull_zero /* -0.02f */) goto second; /* return; */ /* (a) */ -#endif - -#if 0 - v[0] = temp_v[0]; - v[1] = temp_v[1]; - v[2] = temp_v[2]; -#else - (void) v; -#endif - SORT_VERT(); - SET_VERT(); - - if (dy02 == 0) goto second; - - SET_XY(); - SET_DIR(); - SET_Z(); - -#if (IND & S3V_RAST_TEX_BIT) - SET_TEX_VERT(); - SET_UVWD(); -#endif - -#if (IND & S3V_RAST_FLAT_BIT) - SET_FLAT_COL(); -#else - SET_GOURAUD_COL(); -#endif - -#if (IND & S3V_RAST_TEX_BIT) - DMAOUT_CHECK(3DTRI_BASEV, 31); - SEND_UVWD(); - SEND_COL(); - SEND_VERT(); - DMAFINISH(); -#else - DMAOUT_CHECK(3DTRI_GBX, 17); - SEND_COL(); - SEND_VERT(); - DMAFINISH(); -#endif - - /* SECOND TRI (0,2,3) */ - -second: - v[0] = temp_v[0]; - v[1] = temp_v[2]; - v[2] = temp_v[3]; - -#if (IND & S3V_RAST_CULL_BIT) - cull = vmesa->backface_sign * - ((v[1].attrib[FRAG_ATTRIB_WPOS][0] - v[0].attrib[FRAG_ATTRIB_WPOS][0]) * (v[0].attrib[FRAG_ATTRIB_WPOS][1] - v[2].attrib[FRAG_ATTRIB_WPOS][1]) + - (v[1].attrib[FRAG_ATTRIB_WPOS][1] - v[0].attrib[FRAG_ATTRIB_WPOS][1]) * (v[2].attrib[FRAG_ATTRIB_WPOS][0] - v[0].attrib[FRAG_ATTRIB_WPOS][0])); - - if (cull < /* -0.02f */ vmesa->cull_zero) return; -#endif - -/* second: */ - - /* ROMEO */ - /* printf(""); */ /* speed trick */ - - v[0] = temp_v[0]; - v[1] = temp_v[2]; - v[2] = temp_v[3]; - - SORT_VERT(); - SET_VERT(); - - if (dy02 == 0) return; - - SET_XY(); - SET_DIR(); - SET_Z(); - -#if (IND & S3V_RAST_TEX_BIT) - SET_TEX_VERT(); - SET_UVWD(); -#endif - -#if (IND & S3V_RAST_FLAT_BIT) - SET_FLAT_COL(); -#else - SET_GOURAUD_COL(); -#endif - -#if (IND & S3V_RAST_TEX_BIT) - DMAOUT_CHECK(3DTRI_BASEV, 31); - SEND_UVWD(); - SEND_COL(); - SEND_VERT(); - DMAFINISH(); -#else - DMAOUT_CHECK(3DTRI_GBX, 17); - SEND_COL(); - SEND_VERT(); - DMAFINISH(); -#endif -} - -static void TAG(s3v_init)(void) -{ - s3v_point_tab[IND] = TAG(s3v_point); - s3v_line_tab[IND] = TAG(s3v_line); - s3v_tri_tab[IND] = TAG(s3v_triangle); - s3v_quad_tab[IND] = TAG(s3v_quad); -} - -#undef IND -#undef TAG diff --git a/src/mesa/drivers/dri/s3v/s3v_vb.c b/src/mesa/drivers/dri/s3v/s3v_vb.c deleted file mode 100644 index 00e375c6c43..00000000000 --- a/src/mesa/drivers/dri/s3v/s3v_vb.c +++ /dev/null @@ -1,339 +0,0 @@ -/* - * Author: Max Lingua <[email protected]> - */ - -#include "main/glheader.h" -#include "main/mtypes.h" -#include "main/macros.h" -#include "main/colormac.h" - -#include "swrast_setup/swrast_setup.h" -#include "tnl/t_context.h" -#include "tnl/tnl.h" - -#include "s3v_context.h" -#include "s3v_vb.h" -#include "s3v_tris.h" - -#define S3V_XYZW_BIT 0x1 -#define S3V_RGBA_BIT 0x2 -#define S3V_TEX0_BIT 0x4 -#define S3V_PTEX_BIT 0x8 -#define S3V_FOG_BIT 0x10 -#define S3V_MAX_SETUP 0x20 - -static struct { - void (*emit)( GLcontext *, GLuint, GLuint, void *, GLuint ); - tnl_interp_func interp; - tnl_copy_pv_func copy_pv; - GLboolean (*check_tex_sizes)( GLcontext *ctx ); - GLuint vertex_size; - GLuint vertex_stride_shift; - GLuint vertex_format; -} setup_tab[S3V_MAX_SETUP]; - - -/* Only one vertex format, atm, so no need to give them names: - */ -#define TINY_VERTEX_FORMAT 1 -#define NOTEX_VERTEX_FORMAT 0 -#define TEX0_VERTEX_FORMAT 0 -#define TEX1_VERTEX_FORMAT 0 -#define PROJ_TEX1_VERTEX_FORMAT 0 -#define TEX2_VERTEX_FORMAT 0 -#define TEX3_VERTEX_FORMAT 0 -#define PROJ_TEX3_VERTEX_FORMAT 0 - -#define DO_XYZW (IND & S3V_XYZW_BIT) -#define DO_RGBA (IND & S3V_RGBA_BIT) -#define DO_SPEC 0 -#define DO_FOG (IND & S3V_FOG_BIT) -#define DO_TEX0 (IND & S3V_TEX0_BIT) -#define DO_TEX1 0 -#define DO_TEX2 0 -#define DO_TEX3 0 -#define DO_PTEX (IND & S3V_PTEX_BIT) - -#define VERTEX s3vVertex -#define LOCALVARS /* s3vContextPtr vmesa = S3V_CONTEXT(ctx); */ -#define GET_VIEWPORT_MAT() 0 /* vmesa->hw_viewport */ -#define GET_TEXSOURCE(n) n -#define GET_VERTEX_FORMAT() 0 -#define GET_VERTEX_SIZE() S3V_CONTEXT(ctx)->vertex_size * sizeof(GLuint) -#define GET_VERTEX_STORE() S3V_CONTEXT(ctx)->verts -#define GET_VERTEX_STRIDE_SHIFT() S3V_CONTEXT(ctx)->vertex_stride_shift -#define INVALIDATE_STORED_VERTICES() -#define GET_UBYTE_COLOR_STORE() &S3V_CONTEXT(ctx)->UbyteColor -#define GET_UBYTE_SPEC_COLOR_STORE() &S3V_CONTEXT(ctx)->UbyteSecondaryColor - -#define HAVE_HW_VIEWPORT 1 /* FIXME */ -#define HAVE_HW_DIVIDE 1 -#define HAVE_RGBA_COLOR 0 /* we're BGRA */ -#define HAVE_TINY_VERTICES 1 -#define HAVE_NOTEX_VERTICES 0 -#define HAVE_TEX0_VERTICES 0 -#define HAVE_TEX1_VERTICES 0 -#define HAVE_TEX2_VERTICES 0 -#define HAVE_TEX3_VERTICES 0 -#define HAVE_PTEX_VERTICES 1 - -/* -#define SUBPIXEL_X -.5 -#define SUBPIXEL_Y -.5 -#define UNVIEWPORT_VARS GLfloat h = S3V_CONTEXT(ctx)->driDrawable->h -#define UNVIEWPORT_X(x) x - SUBPIXEL_X -#define UNVIEWPORT_Y(y) - y + h + SUBPIXEL_Y -#define UNVIEWPORT_Z(z) z / vmesa->depth_scale -*/ - -#define PTEX_FALLBACK() /* never needed */ - -#define IMPORT_QUALIFIER -#define IMPORT_FLOAT_COLORS s3v_import_float_colors -#define IMPORT_FLOAT_SPEC_COLORS s3v_import_float_spec_colors - -#define INTERP_VERTEX setup_tab[S3V_CONTEXT(ctx)->SetupIndex].interp -#define COPY_PV_VERTEX setup_tab[S3V_CONTEXT(ctx)->SetupIndex].copy_pv - - - -/*********************************************************************** - * Generate pv-copying and translation functions * - ***********************************************************************/ - -#define TAG(x) s3v_##x -#include "tnl_dd/t_dd_vb.c" - -/*********************************************************************** - * Generate vertex emit and interp functions * - ***********************************************************************/ - - -#define IND (S3V_XYZW_BIT|S3V_RGBA_BIT) -#define TAG(x) x##_wg -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (S3V_XYZW_BIT|S3V_RGBA_BIT|S3V_TEX0_BIT) -#define TAG(x) x##_wgt0 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (S3V_XYZW_BIT|S3V_RGBA_BIT|S3V_TEX0_BIT|S3V_PTEX_BIT) -#define TAG(x) x##_wgpt0 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (S3V_TEX0_BIT) -#define TAG(x) x##_t0 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (S3V_RGBA_BIT) -#define TAG(x) x##_g -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (S3V_RGBA_BIT|S3V_TEX0_BIT) -#define TAG(x) x##_gt0 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (S3V_XYZW_BIT|S3V_RGBA_BIT|S3V_FOG_BIT) -#define TAG(x) x##_wgf -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (S3V_XYZW_BIT|S3V_RGBA_BIT|S3V_FOG_BIT|S3V_TEX0_BIT) -#define TAG(x) x##_wgft0 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (S3V_XYZW_BIT|S3V_RGBA_BIT|S3V_FOG_BIT|S3V_TEX0_BIT|S3V_PTEX_BIT) -#define TAG(x) x##_wgfpt0 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (S3V_FOG_BIT) -#define TAG(x) x##_f -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (S3V_RGBA_BIT | S3V_FOG_BIT) -#define TAG(x) x##_gf -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (S3V_RGBA_BIT | S3V_FOG_BIT | S3V_TEX0_BIT) -#define TAG(x) x##_gft0 -#include "tnl_dd/t_dd_vbtmp.h" - -static void init_setup_tab( void ) -{ - init_wg(); /* pos + col */ - init_wgt0(); /* pos + col + tex0 */ - init_wgpt0(); /* pos + col + p-tex0 (?) */ - init_t0(); /* tex0 */ - init_g(); /* col */ - init_gt0(); /* col + tex */ - init_wgf(); - init_wgft0(); - init_wgfpt0(); - init_f(); - init_gf(); - init_gft0(); -} - - -#if 0 -void s3vPrintSetupFlags(char *msg, GLuint flags ) -{ - fprintf(stderr, "%s(%x): %s%s%s%s%s%s\n", - msg, - (int)flags, - (flags & S3V_XYZW_BIT) ? " xyzw," : "", - (flags & S3V_RGBA_BIT) ? " rgba," : "", - (flags & S3V_SPEC_BIT) ? " spec," : "", - (flags & S3V_FOG_BIT) ? " fog," : "", - (flags & S3V_TEX0_BIT) ? " tex-0," : "", - (flags & S3V_TEX1_BIT) ? " tex-1," : ""); -} -#endif - - -void s3vCheckTexSizes( GLcontext *ctx ) -{ - TNLcontext *tnl = TNL_CONTEXT(ctx); - s3vContextPtr vmesa = S3V_CONTEXT( ctx ); - - if (!setup_tab[vmesa->SetupIndex].check_tex_sizes(ctx)) { - - vmesa->SetupIndex |= (S3V_PTEX_BIT|S3V_RGBA_BIT); - - if (1 || !(ctx->_TriangleCaps & (DD_TRI_LIGHT_TWOSIDE|DD_TRI_UNFILLED))) { - tnl->Driver.Render.Interp = setup_tab[vmesa->SetupIndex].interp; - tnl->Driver.Render.CopyPV = setup_tab[vmesa->SetupIndex].copy_pv; - } - } -} - -void s3vBuildVertices( GLcontext *ctx, - GLuint start, - GLuint count, - GLuint newinputs ) -{ - s3vContextPtr vmesa = S3V_CONTEXT( ctx ); - GLubyte *v = ((GLubyte *)vmesa->verts + - (start<<vmesa->vertex_stride_shift)); - GLuint stride = 1<<vmesa->vertex_stride_shift; - - DEBUG(("*** s3vBuildVertices ***\n")); - DEBUG(("vmesa->SetupNewInputs = 0x%x\n", vmesa->SetupNewInputs)); - DEBUG(("vmesa->SetupIndex = 0x%x\n", vmesa->SetupIndex)); - -#if 1 - setup_tab[vmesa->SetupIndex].emit( ctx, start, count, v, stride ); -#else - newinputs |= vmesa->SetupNewInputs; - vmesa->SetupNewInputs = 0; - - DEBUG(("newinputs is 0x%x\n", newinputs)); - - if (!newinputs) { - DEBUG(("!newinputs\n")); - return; - } - - if (newinputs & VERT_CLIP) { - setup_tab[vmesa->SetupIndex].emit( ctx, start, count, v, stride ); - DEBUG(("newinputs & VERT_CLIP\n")); - return; - } /* else { */ -/* GLuint ind = 0; */ - - if (newinputs & VERT_RGBA) { - DEBUG(("newinputs & VERT_RGBA\n")); - ind |= S3V_RGBA_BIT; - } - - if (newinputs & VERT_TEX0) { - DEBUG(("newinputs & VERT_TEX0\n")); - ind |= S3V_TEX0_BIT; - } - - if (newinputs & VERT_FOG_COORD) - ind |= S3V_FOG_BIT; - - if (vmesa->SetupIndex & S3V_PTEX_BIT) - ind = ~0; - - ind &= vmesa->SetupIndex; - - DEBUG(("vmesa->SetupIndex = 0x%x\n", vmesa->SetupIndex)); - DEBUG(("ind = 0x%x\n", ind)); - DEBUG(("ind & vmesa->SetupIndex = 0x%x\n", (ind & vmesa->SetupIndex))); - - if (ind) { - setup_tab[ind].emit( ctx, start, count, v, stride ); - } -#endif -} - -void s3vChooseVertexState( GLcontext *ctx ) -{ - s3vContextPtr vmesa = S3V_CONTEXT( ctx ); - TNLcontext *tnl = TNL_CONTEXT(ctx); - - GLuint ind = S3V_XYZW_BIT | S3V_RGBA_BIT; - - /* FIXME: will segv in tnl_dd/t_dd_vbtmp.h (line 196) on some demos */ -/* - if (ctx->Fog.Enabled) - ind |= S3V_FOG_BIT; -*/ - - - if (ctx->Texture.Unit[0]._ReallyEnabled) { - _tnl_need_projected_coords( ctx, GL_FALSE ); - ind |= S3V_TEX0_BIT; - } else { - _tnl_need_projected_coords( ctx, GL_TRUE ); - } - - vmesa->SetupIndex = ind; - - if (ctx->_TriangleCaps & (DD_TRI_LIGHT_TWOSIDE|DD_TRI_UNFILLED)) { - tnl->Driver.Render.Interp = s3v_interp_extras; - tnl->Driver.Render.CopyPV = s3v_copy_pv_extras; - } else { - tnl->Driver.Render.Interp = setup_tab[ind].interp; - tnl->Driver.Render.CopyPV = setup_tab[ind].copy_pv; - } -} - - -void s3vInitVB( GLcontext *ctx ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - GLuint size = TNL_CONTEXT(ctx)->vb.Size; - - vmesa->verts = (char *)ALIGN_MALLOC(size * 4 * 16, 32); - - { - static int firsttime = 1; - if (firsttime) { - init_setup_tab(); - firsttime = 0; - vmesa->vertex_stride_shift = 6 /* 4 */; /* FIXME - only one vertex setup */ - } - } -} - - -void s3vFreeVB( GLcontext *ctx ) -{ - s3vContextPtr vmesa = S3V_CONTEXT(ctx); - if (vmesa->verts) { - ALIGN_FREE(vmesa->verts); - vmesa->verts = 0; - } - - if (vmesa->UbyteSecondaryColor.Ptr) { - ALIGN_FREE((void *)vmesa->UbyteSecondaryColor.Ptr); - vmesa->UbyteSecondaryColor.Ptr = 0; - } - - if (vmesa->UbyteColor.Ptr) { - ALIGN_FREE((void *)vmesa->UbyteColor.Ptr); - vmesa->UbyteColor.Ptr = 0; - } -} diff --git a/src/mesa/drivers/dri/s3v/s3v_vb.h b/src/mesa/drivers/dri/s3v/s3v_vb.h deleted file mode 100644 index 0fd54373805..00000000000 --- a/src/mesa/drivers/dri/s3v/s3v_vb.h +++ /dev/null @@ -1,39 +0,0 @@ -/* - * Author: Max Lingua <[email protected]> - */ - -#ifndef S3VVB_INC -#define S3VVB_INC - -#include "main/mtypes.h" -#include "swrast/swrast.h" - -#define _S3V_NEW_VERTEX (_NEW_TEXTURE | \ - _DD_NEW_TRI_UNFILLED | \ - _DD_NEW_TRI_LIGHT_TWOSIDE) - - -extern void s3vChooseVertexState( GLcontext *ctx ); -extern void s3vCheckTexSizes( GLcontext *ctx ); -extern void s3vBuildVertices( GLcontext *ctx, - GLuint start, - GLuint count, - GLuint newinputs ); - - -extern void s3v_import_float_colors( GLcontext *ctx ); -extern void s3v_import_float_spec_colors( GLcontext *ctx ); - -extern void s3v_translate_vertex( GLcontext *ctx, - const s3vVertex *src, - SWvertex *dst ); - -extern void s3vInitVB( GLcontext *ctx ); -extern void s3vFreeVB( GLcontext *ctx ); - -extern void s3v_print_vertex( GLcontext *ctx, const s3vVertex *v ); -#if 0 -extern void s3vPrintSetupFlags(char *msg, GLuint flags ); -#endif - -#endif diff --git a/src/mesa/drivers/dri/s3v/s3v_xmesa.c b/src/mesa/drivers/dri/s3v/s3v_xmesa.c deleted file mode 100644 index 85f14817696..00000000000 --- a/src/mesa/drivers/dri/s3v/s3v_xmesa.c +++ /dev/null @@ -1,341 +0,0 @@ -/* - * Author: Max Lingua <[email protected]> - */ - -#include "s3v_context.h" -#include "s3v_vb.h" -#include "s3v_dri.h" -#include "main/context.h" -#include "main/matrix.h" -#include "main/framebuffer.h" -#include "main/renderbuffer.h" -#include "main/viewport.h" - -#include "swrast/swrast.h" -#include "swrast_setup/swrast_setup.h" -#include "tnl/tnl.h" -#include "vbo/vbo.h" - -/* #define DEBUG(str) printf str */ - -static const __DRIconfig ** -s3vInitScreen(__DRIscreen *sPriv) -{ - sPriv->private = (void *) s3vCreateScreen( sPriv ); - - if (!sPriv->private) { - s3vDestroyScreen( sPriv ); - return GL_FALSE; - } - - return NULL; -} - -static void -s3vDestroyContext(__DRIcontextPrivate *driContextPriv) -{ - s3vContextPtr vmesa = (s3vContextPtr)driContextPriv->driverPrivate; - - if (vmesa) { - _swsetup_DestroyContext( vmesa->glCtx ); - _tnl_DestroyContext( vmesa->glCtx ); - _vbo_DestroyContext( vmesa->glCtx ); - _swrast_DestroyContext( vmesa->glCtx ); - - s3vFreeVB( vmesa->glCtx ); - - /* free the Mesa context */ - vmesa->glCtx->DriverCtx = NULL; - _mesa_destroy_context(vmesa->glCtx); - - _mesa_free(vmesa); - driContextPriv->driverPrivate = NULL; - } -} - - -static GLboolean -s3vCreateBuffer( __DRIscreenPrivate *driScrnPriv, - __DRIdrawablePrivate *driDrawPriv, - const __GLcontextModes *mesaVis, - GLboolean isPixmap ) -{ - s3vScreenPtr screen = (s3vScreenPtr) driScrnPriv->private; - - if (isPixmap) { - return GL_FALSE; /* not implemented */ - } - else { - struct gl_framebuffer *fb = _mesa_create_framebuffer(mesaVis); - - { - driRenderbuffer *frontRb - = driNewRenderbuffer(GL_RGBA, NULL, screen->cpp, - screen->frontOffset, screen->frontPitch, - driDrawPriv); - s3vSetSpanFunctions(frontRb, mesaVis); - _mesa_add_renderbuffer(fb, BUFFER_FRONT_LEFT, &frontRb->Base); - } - - if (mesaVis->doubleBufferMode) { - driRenderbuffer *backRb - = driNewRenderbuffer(GL_RGBA, NULL, screen->cpp, - screen->backOffset, screen->backPitch, - driDrawPriv); - s3vSetSpanFunctions(backRb, mesaVis); - _mesa_add_renderbuffer(fb, BUFFER_BACK_LEFT, &backRb->Base); - backRb->backBuffer = GL_TRUE; - } - - if (mesaVis->depthBits == 16) { - driRenderbuffer *depthRb - = driNewRenderbuffer(GL_DEPTH_COMPONENT16, NULL, screen->cpp, - screen->depthOffset, screen->depthPitch, - driDrawPriv); - s3vSetSpanFunctions(depthRb, mesaVis); - _mesa_add_renderbuffer(fb, BUFFER_DEPTH, &depthRb->Base); - } - else if (mesaVis->depthBits == 24) { - driRenderbuffer *depthRb - = driNewRenderbuffer(GL_DEPTH_COMPONENT24, NULL, screen->cpp, - screen->depthOffset, screen->depthPitch, - driDrawPriv); - s3vSetSpanFunctions(depthRb, mesaVis); - _mesa_add_renderbuffer(fb, BUFFER_DEPTH, &depthRb->Base); - } - - /* no h/w stencil yet? - if (mesaVis->stencilBits > 0) { - driRenderbuffer *stencilRb - = driNewRenderbuffer(GL_STENCIL_INDEX8_EXT, NULL, - screen->cpp, screen->depthOffset, - screen->depthPitch, driDrawPriv); - s3vSetSpanFunctions(stencilRb, mesaVis); - _mesa_add_renderbuffer(fb, BUFFER_STENCIL, &stencilRb->Base); - } - */ - - _mesa_add_soft_renderbuffers(fb, - GL_FALSE, /* color */ - GL_FALSE, /* depth */ - mesaVis->stencilBits > 0, - mesaVis->accumRedBits > 0, - GL_FALSE, /* alpha */ - GL_FALSE /* aux */); - driDrawPriv->driverPrivate = (void *) fb; - - return (driDrawPriv->driverPrivate != NULL); - } -} - - -static void -s3vDestroyBuffer(__DRIdrawablePrivate *driDrawPriv) -{ - _mesa_reference_framebuffer((GLframebuffer **)(&(driDrawPriv->driverPrivate)), NULL); -} - -static void -s3vSwapBuffers(__DRIdrawablePrivate *drawablePrivate) -{ - __DRIdrawablePrivate *dPriv = (__DRIdrawablePrivate *) drawablePrivate; - __DRIscreenPrivate *sPriv; - GLcontext *ctx; - s3vContextPtr vmesa; - s3vScreenPtr s3vscrn; - - vmesa = (s3vContextPtr) dPriv->driContextPriv->driverPrivate; - sPriv = vmesa->driScreen; - s3vscrn = vmesa->s3vScreen; - ctx = vmesa->glCtx; - - DEBUG(("*** s3vSwapBuffers ***\n")); - -/* DMAFLUSH(); */ - - _mesa_notifySwapBuffers( ctx ); - - vmesa = (s3vContextPtr) dPriv->driContextPriv->driverPrivate; -/* driScrnPriv = vmesa->driScreen; */ - -/* if (vmesa->EnabledFlags & S3V_BACK_BUFFER) */ - -/* _mesa_notifySwapBuffers( ctx ); */ -#if 1 -{ - int x0, y0, x1, y1; -/* - int nRect = dPriv->numClipRects; - XF86DRIClipRectPtr pRect = dPriv->pClipRects; - - __DRIscreenPrivate *driScrnPriv = vmesa->driScreen; -*/ - -/* - DEBUG(("s3vSwapBuffers: S3V_BACK_BUFFER = 1 - nClip = %i\n", nRect)); -*/ -/* vmesa->drawOffset=vmesa->s3vScreen->backOffset; */ - - x0 = dPriv->x; - y0 = dPriv->y; - - x1 = x0 + dPriv->w - 1; - y1 = y0 + dPriv->h - 1; - - DMAOUT_CHECK(BITBLT_SRC_BASE, 15); - DMAOUT(vmesa->s3vScreen->backOffset); - DMAOUT(0); /* 0xc0000000 */ - DMAOUT( ((x0 << 16) | x1) ); - DMAOUT( ((y0 << 16) | y1) ); - DMAOUT( (vmesa->DestStride << 16) | vmesa->SrcStride ); - DMAOUT( (~(0)) ); - DMAOUT( (~(0)) ); - DMAOUT(0); - DMAOUT(0); - /* FIXME */ - DMAOUT(0); - DMAOUT(0); - DMAOUT( (0x01 | /* Autoexecute */ - 0x02 | /* clip */ - 0x04 | /* 16 bit */ - 0x20 | /* draw */ - 0x400 | /* word alignment (bit 10=1) */ - (0x2 << 11) | /* offset = 1 byte */ - (0xCC << 17) | /* rop #204 */ - (0x3 << 25)) ); /* l-r, t-b */ - DMAOUT(vmesa->ScissorWH); - DMAOUT( /* 0 */ vmesa->SrcXY ); - DMAOUT( (dPriv->x << 16) | dPriv->y ); - DMAFINISH(); - - DMAFLUSH(); - - vmesa->restore_primitive = -1; - -} -#endif -} - -static GLboolean -s3vMakeCurrent(__DRIcontextPrivate *driContextPriv, - __DRIdrawablePrivate *driDrawPriv, - __DRIdrawablePrivate *driReadPriv) -{ - int x1,x2,y1,y2; - int cx, cy, cw, ch; - unsigned int src_stride, dest_stride; - int cl; - - s3vContextPtr vmesa; - __DRIdrawablePrivate *dPriv = driDrawPriv; - vmesa = (s3vContextPtr) dPriv->driContextPriv->driverPrivate; - - DEBUG(("s3vMakeCurrent\n")); - - DEBUG(("dPriv->x=%i y=%i w=%i h=%i\n", dPriv->x, dPriv->y, - dPriv->w, dPriv->h)); - - if (driContextPriv) { - GET_CURRENT_CONTEXT(ctx); - s3vContextPtr oldVirgeCtx = ctx ? S3V_CONTEXT(ctx) : NULL; - s3vContextPtr newVirgeCtx = (s3vContextPtr) driContextPriv->driverPrivate; - - if ( newVirgeCtx != oldVirgeCtx ) { - - newVirgeCtx->dirty = ~0; - cl = 1; - DEBUG(("newVirgeCtx != oldVirgeCtx\n")); -/* s3vUpdateClipping(newVirgeCtx->glCtx ); */ - } - - if (newVirgeCtx->driDrawable != driDrawPriv) { - newVirgeCtx->driDrawable = driDrawPriv; - DEBUG(("driDrawable != driDrawPriv\n")); - s3vUpdateWindow ( newVirgeCtx->glCtx ); - s3vUpdateViewportOffset( newVirgeCtx->glCtx ); -/* s3vUpdateClipping(newVirgeCtx->glCtx ); */ - } -/* - s3vUpdateWindow ( newVirgeCtx->glCtx ); - s3vUpdateViewportOffset( newVirgeCtx->glCtx ); -*/ - -/* - _mesa_make_current( newVirgeCtx->glCtx, - (GLframebuffer *) driDrawPriv->driverPrivate, - (GLframebuffer *) driReadPriv->driverPrivate ); - - _mesa_set_viewport(newVirgeCtx->glCtx, 0, 0, - newVirgeCtx->driDrawable->w, - newVirgeCtx->driDrawable->h); -*/ - -#if 0 - newVirgeCtx->Window &= ~W_GIDMask; - newVirgeCtx->Window |= (driDrawPriv->index << 5); - CHECK_DMA_BUFFER(newVirgeCtx,1); - WRITE(newVirgeCtx->buf, S3VWindow, newVirgeCtx->Window); -#endif - - newVirgeCtx->new_state |= S3V_NEW_WINDOW; /* FIXME */ - - _mesa_make_current( newVirgeCtx->glCtx, - (GLframebuffer *) driDrawPriv->driverPrivate, - (GLframebuffer *) driReadPriv->driverPrivate ); - - if (!newVirgeCtx->glCtx->Viewport.Width) { - _mesa_set_viewport(newVirgeCtx->glCtx, 0, 0, - driDrawPriv->w, driDrawPriv->h); - -/* s3vUpdateClipping(newVirgeCtx->glCtx ); */ - } - -/* - if (cl) { - s3vUpdateClipping(newVirgeCtx->glCtx ); - cl =0; - } -*/ - - newVirgeCtx->new_state |= S3V_NEW_CLIP; - - if (1) { - cx = dPriv->x; - cw = dPriv->w; - cy = dPriv->y; - ch = dPriv->h; - } - - x1 = y1 = 0; - x2 = cw-1; - y2 = ch-1; - - /* src_stride = vmesa->s3vScreen->w * vmesa->s3vScreen->cpp; - dest_stride = ((x2+31)&~31) * vmesa->s3vScreen->cpp; */ - src_stride = vmesa->driScreen->fbWidth * 2; - dest_stride = ((x2+31)&~31) * 2; - } else { - _mesa_make_current( NULL, NULL, NULL ); - } - - return GL_TRUE; -} - - -static GLboolean -s3vUnbindContext( __DRIcontextPrivate *driContextPriv ) -{ - return GL_TRUE; -} - -const struct __DriverAPIRec driDriverAPI = { - .InitScreen = s3vInitScreen, - .DestroyScreen = s3vDestroyScreen, - .CreateContext = s3vCreateContext, - .DestroyContext = s3vDestroyContext, - .CreateBuffer = s3vCreateBuffer, - .DestroyBuffer = s3vDestroyBuffer, - .SwapBuffers = s3vSwapBuffers, - .MakeCurrent = s3vMakeCurrent, - .UnbindContext = s3vUnbindContext, -}; diff --git a/src/mesa/drivers/dri/s3v/s3virgetri.h b/src/mesa/drivers/dri/s3v/s3virgetri.h deleted file mode 100644 index 5519cfd7416..00000000000 --- a/src/mesa/drivers/dri/s3v/s3virgetri.h +++ /dev/null @@ -1,383 +0,0 @@ -/* - * Author: Max Lingua <[email protected]> - */ - -#define LOCAL_VARS \ - int vert0, vert1, vert2; \ - GLfloat y0, y1, y2, ydiff; \ - int iy0, iy1, iy2; \ - int x0, x1, x2, z0, z1, z2; \ - int dy01, dy02, dy12, dx01, dx02, dx12; \ - int delt02, delt01, delt12, end01, end12, start02; \ - int zstart, arstart, gbstart; \ - int deltzy, deltzx, deltarx, deltgbx, deltary, deltgby; \ - GLubyte (*colours)[4]; \ - GLubyte (*scolours)[4]; \ - static int tp = 0; \ - int tmp, lr - -#define LOCAL_TEX_VARS \ - int u0, u1, u2; \ - GLfloat ru0, ru1, ru2; \ - int v0, v1, v2; \ - GLfloat rv0, rv1, rv2; \ - GLfloat w0, w1, w2; \ - GLfloat rw0, rw1, rw2; \ - int baseu, basev; \ - int d0, d1, d2; \ - int deltdx, deltvx, deltux, deltdy, deltvy, deltuy; \ - int deltwx, deltwy; \ - int rbaseu, rbasev; \ - int dstart, ustart, wstart, vstart; \ - static int stmp = 0; \ - s3virgeTextureObject_t *t - -#define CULL_BACKFACE() \ - do { \ - GLfloat *w0 = VB->Win.data[e0]; \ - GLfloat *w1 = VB->Win.data[e1]; \ - GLfloat *w2 = VB->Win.data[e2]; \ - float cull; \ - cull = ctx->backface_sign * ((w1[0] - w0[0]) * (w0[1] - w2[1]) + \ - (w1[1] - w0[1]) * (w2[0] - w0[0])); \ - if (cull < 0) \ - return; \ - } while (0) - -#define SORT_VERTICES() \ - do { \ - y0 = VB->Win.data[e0][1]; \ - y1 = VB->Win.data[e1][1]; \ - y2 = VB->Win.data[e2][1]; \ - if (y1 > y0) { \ - if (y2 > y0) { \ - vert0 = e0; \ - if (y1 > y2) { vert2 = e1; vert1 = e2; } else { vert2 = e2; vert1 = e1; } \ - } else { vert0 = e2; vert1 = e0; vert2 = e1; } \ - } else { \ - if (y2 > y0) { vert0 = e1; vert1 = e0; vert2 = e2; } else { \ - vert2 = e0; \ - if (y2 > y1) { vert0 = e1; vert1 = e2; } else { vert0 = e2; vert1 = e1; } \ - } \ - } \ - } while (0) - -#define SET_VARIABLES() \ - do { \ - iy0 = y0 = ((s3virgeDB->height - (VB->Win.data[vert0][1]))); \ - iy1 = y1 = ((s3virgeDB->height - (VB->Win.data[vert1][1]))); \ - iy2 = y2 = ((s3virgeDB->height - (VB->Win.data[vert2][1]))); \ - if (iy0 == iy2) { return; } \ - ydiff = y0 - (float)iy0; \ - x0 = ((VB->Win.data[vert0][0]) * 1024.0 * 1024.0); \ - x1 = ((VB->Win.data[vert1][0]) * 1024.0 * 1024.0); \ - x2 = ((VB->Win.data[vert2][0]) * 1024.0 * 1024.0); \ - z0 = (VB->Win.data[vert0][2] * 1024.0 * 32.0); \ - z1 = (VB->Win.data[vert1][2] * 1024.0 * 32.0); \ - z2 = (VB->Win.data[vert2][2] * 1024.0 * 32.0); \ - dx12 = x2 - x1; \ - dy12 = iy1 - iy2; \ - dx01 = x1 - x0; \ - dy01 = iy0 - iy1; \ - dx02 = x2 - x0; \ - dy02 = iy0 - iy2; \ - delt12 = delt02 = delt01 = 0; \ - } while (0) - -#define SET_TEX_VARIABLES() \ - do { \ - t = ((s3virgeTextureObject_t *)ctx->Texture.Unit[0].Current->DriverData); \ - deltwx = deltwy = wstart = deltdx = deltdy = dstart = 0; \ - u0 = (VB->TexCoordPtr[0]->data[vert0][0] * (GLfloat)(t->tObj->Image[0]->Width) * 256.0); \ - u1 = (VB->TexCoordPtr[0]->data[vert1][0] * (GLfloat)(t->tObj->Image[0]->Width) * 256.0); \ - u2 = (VB->TexCoordPtr[0]->data[vert2][0] * (GLfloat)(t->tObj->Image[0]->Width) * 256.0); \ - v0 = (VB->TexCoordPtr[0]->data[vert0][1] * (GLfloat)(t->tObj->Image[0]->Height) * 256.0); \ - v1 = (VB->TexCoordPtr[0]->data[vert1][1] * (GLfloat)(t->tObj->Image[0]->Height) * 256.0); \ - v2 = (VB->TexCoordPtr[0]->data[vert2][1] * (GLfloat)(t->tObj->Image[0]->Height) * 256.0); \ - w0 = (VB->Win.data[vert0][3]); \ - w1 = (VB->Win.data[vert1][3]); \ - w2 = (VB->Win.data[vert2][3]); \ - } while (0) - -#define FLATSHADE_COLORS() \ - do { \ - GLubyte *col = &(colours[pv][0]); \ - deltarx = deltary = deltgbx = deltgby = 0; \ - gbstart = (((col[1]) << 23) | ((col[2]) << 7)); \ - arstart = (((col[3]) << 23) | ((col[0]) << 7)); \ - } while (0) - -#define GOURAUD_COLORS() \ - do { \ - int ctmp, ctmp2, ctmp3, ctmp4; \ - GLubyte *col0, *col1, *col2; \ - col0 = &(colours[vert0][0]); \ - col1 = &(colours[vert1][0]); \ - col2 = &(colours[vert2][0]); \ - ctmp = ((col2[3] - col0[3]) << 7) / dy02; \ - ctmp2 = ((col2[0] - col0[0]) << 7) / dy02; \ - deltary = ((ctmp << 16) & 0xFFFF0000) | (ctmp2 & 0xFFFF); \ - ctmp3 = ((col2[1] - col0[1]) << 7) / dy02; \ - ctmp4 = ((col2[2] - col0[2]) << 7) / dy02; \ - deltgby = ((ctmp3 << 16) & 0xFFFF0000) | (ctmp4 & 0xFFFF); \ - gbstart = (((int)((ydiff * ctmp3) + (col0[1] << 7)) << 16) & 0x7FFF0000) | \ - ((int)((ydiff * ctmp4) + (col0[2] << 7)) & 0x7FFF); \ - arstart = (((int)((ydiff * ctmp) + (col0[3] << 7)) << 16) & 0x7FFF0000) | \ - ((int)((ydiff * ctmp2) + (col0[0] << 7)) & 0x7FFF); \ - if (tmp) { \ - int ax, rx, gx, bx; \ - ax = ((col1[3] << 7) - (dy01 * ctmp + (col0[3] << 7))) / tmp; \ - rx = ((col1[0] << 7) - (dy01 * ctmp2 + (col0[0] << 7))) / tmp; \ - gx = ((col1[1] << 7) - (dy01 * ctmp3 + (col0[1] << 7))) / tmp; \ - bx = ((col1[2] << 7) - (dy01 * ctmp4 + (col0[2] << 7))) / tmp; \ - deltarx = ((ax << 16) & 0xFFFF0000) | (rx & 0xFFFF); \ - deltgbx = ((gx << 16) & 0xFFFF0000) | (bx & 0xFFFF); \ - } else { \ - deltgbx = deltarx = 0; \ - } \ - } while (0) - -#define SET_XY() \ - do { \ - delt02 = dx02 / dy02; \ - if (dy12) delt12 = dx12 / dy12; \ - if (dy01) delt01 = dx01 / dy01; \ - start02 = (ydiff * delt02) + x0; \ - end01 = (ydiff * delt01) + x0; \ - end12 = ((y1 - (GLfloat)iy1) * delt12) + x1; \ - } while (0) - -#define SET_DIR() \ - do { \ - tmp = x1 - (dy01 * delt02 + x0); \ - if (tmp > 0) { \ - lr = 0x80000000; \ - } else { \ - tmp *= -1; \ - lr = 0; \ - } \ - tmp >>= 20; \ - } while (0) - -#define SET_Z() \ - do { \ - deltzy = (z2 - z0) / dy02; \ - if (tmp) { \ - deltzx = (z1 - (dy01 * deltzy + z0)) / tmp; \ - } else { deltzx = 0; } \ - zstart = (deltzy * ydiff) + z0; \ - } while (0) - -#define SET_BASEUV() \ - do { \ - if (u0 < u1) { \ - if (u0 < u2) { \ - baseu = u0; \ - } else { \ - baseu = u2; \ - } \ - } else { \ - if (u1 < u2) { \ - baseu = u1; \ - } else { \ - baseu = u2; \ - } \ - } \ - if (v0 < v1) { \ - if (v0 < v2) { \ - basev = v0; \ - } else { \ - basev = v2; \ - } \ - } else { \ - if (v1 < v2) { \ - basev = v1; \ - } else { \ - basev = v2; \ - } \ - } \ - } while (0) - -#define SET_RW() \ - do { \ - /* GLfloat minW; \ - if (w0 < w1) { \ - if (w0 < w2) { \ - minW = w0; \ - } else { \ - minW = w2; \ - } \ - } else { \ - if (w1 < w2) { \ - minW = w1; \ - } else { \ - minW = w2; \ - } \ - } */ \ - rw0 = (512.0 * w0); \ - rw1 = (512.0 * w1); \ - rw2 = (512.0 * w2); \ - } while (0) - - -#define SET_D() \ - do { \ - GLfloat sxy, suv; \ - int lev; \ - suv = (VB->TexCoordPtr[0]->data[vert0][0] - \ - VB->TexCoordPtr[0]->data[vert2][0]) * \ - (VB->TexCoordPtr[0]->data[vert1][1] - \ - VB->TexCoordPtr[0]->data[vert2][1]) - \ - (VB->TexCoordPtr[0]->data[vert1][0] - \ - VB->TexCoordPtr[0]->data[vert2][0]) * \ - (VB->TexCoordPtr[0]->data[vert0][1] - \ - VB->TexCoordPtr[0]->data[vert2][2]); \ - sxy = (VB->Win.data[vert0][0] - \ - VB->Win.data[vert2][0]) * \ - (VB->Win.data[vert1][1] - \ - VB->Win.data[vert2][1]) - \ - (VB->Win.data[vert1][0] - \ - VB->Win.data[vert2][0]) * \ - (VB->Win.data[vert0][1] - \ - VB->Win.data[vert2][2]); \ - if (sxy < 0) sxy *= -1.0; \ - if (suv < 0) suv *= -1.0; \ - lev = *(int*)&suv - *(int *)&sxy; \ - if (lev < 0) \ - lev = 0; \ - else \ - lev >>=23; \ - dstart = (lev << 27); \ - } while (0) - - - -#define SET_UVWD() \ - do { \ - SET_BASEUV(); \ - SET_RW(); \ - SET_D(); \ - ru0 = (((u0 - baseu) * rw0)); \ - ru1 = (((u1 - baseu) * rw1)); \ - ru2 = (((u2 - baseu) * rw2)); \ - rv0 = (((v0 - basev) * rw0)); \ - rv1 = (((v1 - basev) * rw1)); \ - rv2 = (((v2 - basev) * rw2)); \ - while (baseu < 0) { baseu += (t->tObj->Image[0]->Width << 8); } \ - while (basev < 0) { basev += (t->tObj->Image[0]->Height << 8); } \ - if (!(baseu & 0xFF)) { baseu = (baseu >> 8); } else { baseu = (baseu >> 8) + 1; } \ - if ((basev & 0x80) || !(basev & 0xFF)) { basev = (basev >> 8); } else { basev = (basev >> 8) - 1; } \ - rbaseu = (baseu) << (16 - t->widthLog2); \ - rbasev = (basev) << (16 - t->widthLog2); \ - deltuy = (((ru2 - ru0) / dy02)); \ - deltvy = (((rv2 - rv0) / dy02)); \ - rw0 *= (1024.0 * 512.0); \ - rw1 *= (1024.0 * 512.0); \ - rw2 *= (1024.0 * 512.0); \ - deltwy = ((rw2 - rw0) / dy02); \ - if (tmp) { \ - deltux = ((ru1 - (dy01 * deltuy + ru0)) / tmp); \ - deltvx = ((rv1 - (dy01 * deltvy + rv0)) / tmp); \ - deltwx = ((rw1 - (dy01 * deltwy + rw0)) / tmp); \ - } else { deltux = deltvx = deltwx = 0; } \ - ustart = (deltuy * ydiff) + (ru0); \ - vstart = (deltvy * ydiff) + (rv0); \ - wstart = (deltwy * ydiff) + (rw0); \ - } while (0) - - -#define SEND_COLORS() \ - do { \ - WAITFIFOEMPTY(6); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_GBX), deltgbx); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_ARX), deltarx); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_GBY), deltgby); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_ARY), deltary); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_GS_BS), gbstart); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_AS_RS), arstart); \ - } while (0) - -#define SEND_VERTICES() \ - do { \ - WAITFIFOEMPTY(6); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_ZSTART), zstart); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_ZXD), deltzx); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_ZYD), deltzy); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_TXDELTA12), delt12); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_TXEND12), end12); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_TXDELTA01), delt01); \ - WAITFIFOEMPTY(5); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_TXEND01), end01); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_TXDELTA02), delt02); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_TXSTART02), start02); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_TYS), iy0); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_TY01_Y12), \ - ((((iy0 - iy1) & 0x7FF) << 16) | \ - ((iy1 - iy2) & 0x7FF) | lr)); \ - } while (0) - -#define SEND_UVWD() \ - do { \ - WAITFIFOEMPTY(7); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_BASEV), (rbasev & 0xFFFF)); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_BASEU), (0xa0000000 | (rbaseu & 0xFFFF))); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_WXD), deltwx); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_WYD), deltwy); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_WSTART), wstart); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_DXD), deltdx); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_VXD), deltvx); \ - WAITFIFOEMPTY(7); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_UXD), deltux); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_DYD), deltdy); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_VYD), deltvy); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_UYD), deltuy); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_DSTART), dstart); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_VSTART), vstart); \ - OUTREG( (S3VIRGE_3DTRI_REG | S3VIRGE_3DTRI_USTART), ustart); \ - } while (0) - -#define DMA_SEND_UVWD() \ - do { \ - DMAOUT((rbasev & 0xFFFF)); \ - DMAOUT((0xa0000000 | (rbaseu & 0xFFFF))); \ - DMAOUT(deltwx); \ - DMAOUT(deltwy); \ - DMAOUT(wstart); \ - DMAOUT(deltdx); \ - DMAOUT(deltvx); \ - DMAOUT(deltux); \ - DMAOUT(deltdy); \ - DMAOUT(deltvy); \ - DMAOUT(deltuy); \ - DMAOUT(dstart); \ - DMAOUT(vstart); \ - DMAOUT(ustart); \ - } while (0) - - -#define DMA_SEND_COLORS() \ - do { \ - DMAOUT(deltgbx); \ - DMAOUT(deltarx); \ - DMAOUT(deltgby); \ - DMAOUT(deltary); \ - DMAOUT(gbstart); \ - DMAOUT(arstart); \ - } while (0) - -#define DMA_SEND_VERTICES() \ - do { \ - DMAOUT(deltzx); \ - DMAOUT(deltzy); \ - DMAOUT(zstart); \ - DMAOUT(delt12); \ - DMAOUT(end12); \ - DMAOUT(delt01); \ - DMAOUT(end01); \ - DMAOUT(delt02); \ - DMAOUT(start02); \ - DMAOUT(iy0); \ - DMAOUT(((((iy0 - iy1) & 0x7FF) << 16) | \ - ((iy1 - iy2) & 0x7FF) | lr)); \ - } while (0) - diff --git a/src/mesa/drivers/dri/savage/Makefile b/src/mesa/drivers/dri/savage/Makefile index 018482f66be..2e5c40802c1 100644 --- a/src/mesa/drivers/dri/savage/Makefile +++ b/src/mesa/drivers/dri/savage/Makefile @@ -27,4 +27,3 @@ ASM_SOURCES = include ../Makefile.template -symlinks: diff --git a/src/mesa/drivers/dri/savage/savage_xmesa.c b/src/mesa/drivers/dri/savage/savage_xmesa.c index 931ceff0a81..d307b81e8ec 100644 --- a/src/mesa/drivers/dri/savage/savage_xmesa.c +++ b/src/mesa/drivers/dri/savage/savage_xmesa.c @@ -59,7 +59,7 @@ #include "texmem.h" #define need_GL_EXT_secondary_color -#include "extension_helper.h" +#include "main/remap_helper.h" #include "xmlpool.h" @@ -436,7 +436,7 @@ savageCreateContext( const __GLcontextModes *mesaVis, if (ctx->Const.MaxTextureLevels <= 6) { /*spec requires at least 64x64*/ __driUtilMessage("Not enough texture memory. " "Falling back to indirect rendering."); - Xfree(imesa); + _mesa_free(imesa); return GL_FALSE; } @@ -574,7 +574,7 @@ savageDestroyContext(__DRIcontextPrivate *driContextPriv) _mesa_destroy_context(imesa->glCtx); /* no longer use vertex_dma_buf*/ - Xfree(imesa); + _mesa_free(imesa); } } @@ -602,7 +602,7 @@ savageCreateBuffer( __DRIscreenPrivate *driScrnPriv, { driRenderbuffer *frontRb - = driNewRenderbuffer(GL_RGBA, + = driNewRenderbuffer(MESA_FORMAT_ARGB8888, (GLubyte *) screen->aperture.map + 0x01000000 * TARGET_FRONT, screen->cpp, @@ -615,7 +615,7 @@ savageCreateBuffer( __DRIscreenPrivate *driScrnPriv, if (mesaVis->doubleBufferMode) { driRenderbuffer *backRb - = driNewRenderbuffer(GL_RGBA, + = driNewRenderbuffer(MESA_FORMAT_ARGB8888, (GLubyte *) screen->aperture.map + 0x01000000 * TARGET_BACK, screen->cpp, @@ -628,7 +628,7 @@ savageCreateBuffer( __DRIscreenPrivate *driScrnPriv, if (mesaVis->depthBits == 16) { driRenderbuffer *depthRb - = driNewRenderbuffer(GL_DEPTH_COMPONENT16, + = driNewRenderbuffer(MESA_FORMAT_Z16, (GLubyte *) screen->aperture.map + 0x01000000 * TARGET_DEPTH, screen->zpp, @@ -639,7 +639,7 @@ savageCreateBuffer( __DRIscreenPrivate *driScrnPriv, } else if (mesaVis->depthBits == 24) { driRenderbuffer *depthRb - = driNewRenderbuffer(GL_DEPTH_COMPONENT24, + = driNewRenderbuffer(MESA_FORMAT_S8_Z24, (GLubyte *) screen->aperture.map + 0x01000000 * TARGET_DEPTH, screen->zpp, @@ -651,7 +651,7 @@ savageCreateBuffer( __DRIscreenPrivate *driScrnPriv, if (mesaVis->stencilBits > 0 && !swStencil) { driRenderbuffer *stencilRb - = driNewRenderbuffer(GL_STENCIL_INDEX8_EXT, + = driNewRenderbuffer(MESA_FORMAT_S8, (GLubyte *) screen->aperture.map + 0x01000000 * TARGET_DEPTH, screen->zpp, @@ -980,18 +980,6 @@ savageInitScreen(__DRIscreenPrivate *psp) &psp->drm_version, & drm_expected ) ) return NULL; - /* Calling driInitExtensions here, with a NULL context pointer, - * does not actually enable the extensions. It just makes sure - * that all the dispatch offsets for all the extensions that - * *might* be enables are known. This is needed because the - * dispatch offsets need to be known when _mesa_context_create is - * called, but we can't enable the extensions until we have a - * context pointer. - * - * Hello chicken. Hello egg. How are you two today? - */ - driInitExtensions( NULL, card_extensions, GL_FALSE ); - if (!savageInitDriver(psp)) return NULL; diff --git a/src/mesa/drivers/dri/savage/savagerender.c b/src/mesa/drivers/dri/savage/savagerender.c index 32c74f9467e..8221edf3872 100644 --- a/src/mesa/drivers/dri/savage/savagerender.c +++ b/src/mesa/drivers/dri/savage/savagerender.c @@ -252,13 +252,13 @@ static GLboolean run_texnorm_stage( GLcontext *ctx, const GLboolean normalizeS = (texObj->WrapS == GL_REPEAT); const GLboolean normalizeT = (reallyEnabled & TEXTURE_2D_BIT) && (texObj->WrapT == GL_REPEAT); - const GLfloat *in = (GLfloat *)VB->TexCoordPtr[i]->data; - const GLint instride = VB->TexCoordPtr[i]->stride; + const GLfloat *in = (GLfloat *)VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->data; + const GLint instride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->stride; GLfloat (*out)[4] = store->texcoord[i].data; GLint j; if (!ctx->Texture.Unit[i]._ReallyEnabled || - VB->TexCoordPtr[i]->size == 4) + VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]->size == 4) /* Never try to normalize homogenous tex coords! */ continue; @@ -297,7 +297,7 @@ static GLboolean run_texnorm_stage( GLcontext *ctx, } if (normalizeS || normalizeT) - VB->AttribPtr[VERT_ATTRIB_TEX0+i] = VB->TexCoordPtr[i] = &store->texcoord[i]; + VB->AttribPtr[_TNL_ATTRIB_TEX0 + i] = &store->texcoord[i]; } } diff --git a/src/mesa/drivers/dri/savage/savagespan.c b/src/mesa/drivers/dri/savage/savagespan.c index 9615e34013c..3bb6fbcc63c 100644 --- a/src/mesa/drivers/dri/savage/savagespan.c +++ b/src/mesa/drivers/dri/savage/savagespan.c @@ -255,15 +255,13 @@ void savageSetSpanFunctions(driRenderbuffer *drb, const GLvisual *vis, GLboolean float_depth) { - if (drb->Base.InternalFormat == GL_RGBA) { - if (vis->redBits == 5 && vis->greenBits == 6 && vis->blueBits == 5) { - savageInitPointers_565(&drb->Base); - } - else { - savageInitPointers_8888(&drb->Base); - } + if (drb->Base.Format == MESA_FORMAT_RGB565) { + savageInitPointers_565(&drb->Base); + } + else if (drb->Base.Format == MESA_FORMAT_ARGB8888) { + savageInitPointers_8888(&drb->Base); } - else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT16) { + else if (drb->Base.Format == MESA_FORMAT_Z16) { if (float_depth) { savageInitDepthPointers_z16f(&drb->Base); } @@ -271,7 +269,7 @@ savageSetSpanFunctions(driRenderbuffer *drb, const GLvisual *vis, savageInitDepthPointers_z16(&drb->Base); } } - else if (drb->Base.InternalFormat == GL_DEPTH_COMPONENT24) { + else if (drb->Base.Format == MESA_FORMAT_S8_Z24) { if (float_depth) { savageInitDepthPointers_s8_z24f(&drb->Base); } @@ -279,7 +277,7 @@ savageSetSpanFunctions(driRenderbuffer *drb, const GLvisual *vis, savageInitDepthPointers_s8_z24(&drb->Base); } } - else if (drb->Base.InternalFormat == GL_STENCIL_INDEX8_EXT) { + else if (drb->Base.Format == MESA_FORMAT_S8) { savageInitStencilPointers_s8_z24(&drb->Base); } } diff --git a/src/mesa/drivers/dri/savage/savagetex.c b/src/mesa/drivers/dri/savage/savagetex.c index fe239e1b05f..6c97bb6c702 100644 --- a/src/mesa/drivers/dri/savage/savagetex.c +++ b/src/mesa/drivers/dri/savage/savagetex.c @@ -23,29 +23,24 @@ */ -#include <stdlib.h> -#include <stdio.h> - -#include <GL/gl.h> - +#include "main/context.h" #include "main/mm.h" -#include "savagecontext.h" -#include "savagetex.h" -#include "savagetris.h" -#include "savageioctl.h" -#include "main/simple_list.h" -#include "main/enums.h" -#include "savage_bci.h" - #include "main/macros.h" -#include "main/texformat.h" #include "main/texstore.h" #include "main/texobj.h" #include "main/convolve.h" #include "main/colormac.h" +#include "main/simple_list.h" +#include "main/enums.h" #include "swrast/swrast.h" +#include "savagecontext.h" +#include "savagetex.h" +#include "savagetris.h" +#include "savageioctl.h" +#include "savage_bci.h" + #include "xmlpool.h" #define TILE_INDEX_DXT1 0 @@ -527,6 +522,11 @@ savageAllocTexObj( struct gl_texture_object *texObj ) * components to white. This way we get the correct result. */ +#if 0 +/* Using MESA_FORMAT_RGBA8888 to store alpha-only textures should + * work but is space inefficient. + */ + static GLboolean _savage_texstore_a1114444(TEXSTORE_PARAMS); @@ -590,10 +590,11 @@ _savage_texstore_a1114444(TEXSTORE_PARAMS) return GL_FALSE; _mesa_adjust_image_for_convolution(ctx, dims, &srcWidth, &srcHeight); for (img = 0; img < srcDepth; img++) { + GLuint texelBytes = _mesa_get_format_bytes(dstFormat); GLubyte *dstRow = (GLubyte *) dstAddr - + dstImageOffsets[dstZoffset + img] * dstFormat->TexelBytes + + dstImageOffsets[dstZoffset + img] * texelBytes + dstYoffset * dstRowStride - + dstXoffset * dstFormat->TexelBytes; + + dstXoffset * texelBytes; for (row = 0; row < srcHeight; row++) { GLushort *dstUI = (GLushort *) dstRow; for (col = 0; col < srcWidth; col++) { @@ -629,10 +630,11 @@ _savage_texstore_a1118888(TEXSTORE_PARAMS) return GL_FALSE; _mesa_adjust_image_for_convolution(ctx, dims, &srcWidth, &srcHeight); for (img = 0; img < srcDepth; img++) { + GLuint texelBytes = _mesa_get_format_bytes(dstFormat); GLubyte *dstRow = (GLubyte *) dstAddr - + dstImageOffsets[dstZoffset + img] * dstFormat->TexelBytes + + dstImageOffsets[dstZoffset + img] * texelBytes + dstYoffset * dstRowStride - + dstXoffset * dstFormat->TexelBytes; + + dstXoffset * texelBytes; for (row = 0; row < srcHeight; row++) { GLuint *dstUI = (GLuint *) dstRow; for (col = 0; col < srcWidth; col++) { @@ -647,10 +649,11 @@ _savage_texstore_a1118888(TEXSTORE_PARAMS) return GL_TRUE; } +#endif /* Called by the _mesa_store_teximage[123]d() functions. */ -static const struct gl_texture_format * +static gl_format savageChooseTextureFormat( GLcontext *ctx, GLint internalFormat, GLenum format, GLenum type ) { @@ -669,15 +672,15 @@ savageChooseTextureFormat( GLcontext *ctx, GLint internalFormat, switch ( type ) { case GL_UNSIGNED_INT_10_10_10_2: case GL_UNSIGNED_INT_2_10_10_10_REV: - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb1555; + return do32bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB1555; case GL_UNSIGNED_SHORT_4_4_4_4: case GL_UNSIGNED_SHORT_4_4_4_4_REV: - return &_mesa_texformat_argb4444; + return MESA_FORMAT_ARGB4444; case GL_UNSIGNED_SHORT_5_5_5_1: case GL_UNSIGNED_SHORT_1_5_5_5_REV: - return &_mesa_texformat_argb1555; + return MESA_FORMAT_ARGB1555; default: - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + return do32bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB4444; } case 3: @@ -686,129 +689,152 @@ savageChooseTextureFormat( GLcontext *ctx, GLint internalFormat, switch ( type ) { case GL_UNSIGNED_SHORT_4_4_4_4: case GL_UNSIGNED_SHORT_4_4_4_4_REV: - return &_mesa_texformat_argb4444; + return MESA_FORMAT_ARGB4444; case GL_UNSIGNED_SHORT_5_5_5_1: case GL_UNSIGNED_SHORT_1_5_5_5_REV: - return &_mesa_texformat_argb1555; + return MESA_FORMAT_ARGB1555; case GL_UNSIGNED_SHORT_5_6_5: case GL_UNSIGNED_SHORT_5_6_5_REV: - return &_mesa_texformat_rgb565; + return MESA_FORMAT_RGB565; default: - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565; + return do32bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_RGB565; } case GL_RGBA8: case GL_RGBA12: case GL_RGBA16: return !force16bpt ? - &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB4444; case GL_RGB10_A2: return !force16bpt ? - &_mesa_texformat_argb8888 : &_mesa_texformat_argb1555; + MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB1555; case GL_RGBA4: case GL_RGBA2: - return &_mesa_texformat_argb4444; + return MESA_FORMAT_ARGB4444; case GL_RGB5_A1: - return &_mesa_texformat_argb1555; + return MESA_FORMAT_ARGB1555; case GL_RGB8: case GL_RGB10: case GL_RGB12: case GL_RGB16: - return !force16bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565; + return !force16bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_RGB565; case GL_RGB5: case GL_RGB4: case GL_R3_G3_B2: - return &_mesa_texformat_rgb565; + return MESA_FORMAT_RGB565; case GL_ALPHA: case GL_COMPRESSED_ALPHA: - return isSavage4 ? &_mesa_texformat_a8 : ( +#if 0 + return isSavage4 ? MESA_FORMAT_a8 : ( do32bpt ? &_savage_texformat_a1118888 : &_savage_texformat_a1114444); +#else + if (isSavage4) + return MESA_FORMAT_A8; + else if (do32bpt) + return MESA_FORMAT_ARGB8888; + else + return MESA_FORMAT_ARGB4444; +#endif case GL_ALPHA4: - return isSavage4 ? &_mesa_texformat_a8 : &_savage_texformat_a1114444; +#if 0 + return isSavage4 ? MESA_FORMAT_a8 : &_savage_texformat_a1114444; +#else + if (isSavage4) + return MESA_FORMAT_A8; + else + return MESA_FORMAT_ARGB4444; +#endif case GL_ALPHA8: case GL_ALPHA12: case GL_ALPHA16: - return isSavage4 ? &_mesa_texformat_a8 : ( +#if 0 + return isSavage4 ? MESA_FORMAT_a8 : ( !force16bpt ? &_savage_texformat_a1118888 : &_savage_texformat_a1114444); - +#else + if (isSavage4) + return MESA_FORMAT_A8; + else if (force16bpt) + return MESA_FORMAT_ARGB4444; + else + return MESA_FORMAT_ARGB8888; +#endif case 1: case GL_LUMINANCE: case GL_COMPRESSED_LUMINANCE: /* no alpha, but use argb1555 in 16bit case to get pure grey values */ - return isSavage4 ? &_mesa_texformat_l8 : ( - do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb1555); + return isSavage4 ? MESA_FORMAT_L8 : ( + do32bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB1555); case GL_LUMINANCE4: - return isSavage4 ? &_mesa_texformat_l8 : &_mesa_texformat_argb1555; + return isSavage4 ? MESA_FORMAT_L8 : MESA_FORMAT_ARGB1555; case GL_LUMINANCE8: case GL_LUMINANCE12: case GL_LUMINANCE16: - return isSavage4 ? &_mesa_texformat_l8 : ( - !force16bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb1555); + return isSavage4 ? MESA_FORMAT_L8 : ( + !force16bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB1555); case 2: case GL_LUMINANCE_ALPHA: case GL_COMPRESSED_LUMINANCE_ALPHA: /* Savage4 has a al44 texture format. But it's not supported by Mesa. */ - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + return do32bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB4444; case GL_LUMINANCE4_ALPHA4: case GL_LUMINANCE6_ALPHA2: - return &_mesa_texformat_argb4444; + return MESA_FORMAT_ARGB4444; case GL_LUMINANCE8_ALPHA8: case GL_LUMINANCE12_ALPHA4: case GL_LUMINANCE12_ALPHA12: case GL_LUMINANCE16_ALPHA16: - return !force16bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + return !force16bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB4444; #if 0 /* TFT_I8 produces garbage on ProSavageDDR and subsequent texture * disable keeps rendering garbage. Disabled for now. */ case GL_INTENSITY: case GL_COMPRESSED_INTENSITY: - return isSavage4 ? &_mesa_texformat_i8 : ( - do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444); + return isSavage4 ? MESA_FORMAT_i8 : ( + do32bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB4444); case GL_INTENSITY4: - return isSavage4 ? &_mesa_texformat_i8 : &_mesa_texformat_argb4444; + return isSavage4 ? MESA_FORMAT_i8 : MESA_FORMAT_ARGB4444; case GL_INTENSITY8: case GL_INTENSITY12: case GL_INTENSITY16: - return isSavage4 ? &_mesa_texformat_i8 : ( - !force16bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444); + return isSavage4 ? MESA_FORMAT_i8 : ( + !force16bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB4444); #else case GL_INTENSITY: case GL_COMPRESSED_INTENSITY: - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + return do32bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB4444; case GL_INTENSITY4: - return &_mesa_texformat_argb4444; + return MESA_FORMAT_ARGB4444; case GL_INTENSITY8: case GL_INTENSITY12: case GL_INTENSITY16: - return !force16bpt ? &_mesa_texformat_argb8888 : - &_mesa_texformat_argb4444; + return !force16bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB4444; #endif case GL_RGB_S3TC: case GL_RGB4_S3TC: case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: - return &_mesa_texformat_rgb_dxt1; + return MESA_FORMAT_RGB_DXT1; case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: - return &_mesa_texformat_rgba_dxt1; + return MESA_FORMAT_RGBA_DXT1; case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: - return &_mesa_texformat_rgba_dxt3; + return MESA_FORMAT_RGBA_DXT3; case GL_RGBA_S3TC: case GL_RGBA4_S3TC: if (!isSavage4) /* Not the best choice but Savage3D/MX/IX don't support DXT3 or DXT5. */ - return &_mesa_texformat_rgba_dxt1; + return MESA_FORMAT_RGBA_DXT1; /* fall through */ case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: - return &_mesa_texformat_rgba_dxt5; + return MESA_FORMAT_RGBA_DXT5; /* case GL_COLOR_INDEX: @@ -822,7 +848,7 @@ savageChooseTextureFormat( GLcontext *ctx, GLint internalFormat, */ default: _mesa_problem(ctx, "unexpected texture format in %s", __FUNCTION__); - return NULL; + return MESA_FORMAT_NONE; } } @@ -837,7 +863,7 @@ static void savageSetTexImages( savageContextPtr imesa, assert(t); assert(image); - switch (image->TexFormat->MesaFormat) { + switch (image->TexFormat) { case MESA_FORMAT_ARGB8888: textureFormat = TFT_ARGB8888; t->texelBytes = tileIndex = 4; @@ -2083,6 +2109,7 @@ void savageDDInitTextureFuncs( struct dd_function_table *functions ) /* Texel fetching with our custom texture formats works just like * the standard argb formats. */ +#if 0 _savage_texformat_a1114444.FetchTexel1D = _mesa_texformat_argb4444.FetchTexel1D; _savage_texformat_a1114444.FetchTexel2D = _mesa_texformat_argb4444.FetchTexel2D; _savage_texformat_a1114444.FetchTexel3D = _mesa_texformat_argb4444.FetchTexel3D; @@ -2096,4 +2123,5 @@ void savageDDInitTextureFuncs( struct dd_function_table *functions ) _savage_texformat_a1118888.FetchTexel1Df= _mesa_texformat_argb8888.FetchTexel1Df; _savage_texformat_a1118888.FetchTexel2Df= _mesa_texformat_argb8888.FetchTexel2Df; _savage_texformat_a1118888.FetchTexel3Df= _mesa_texformat_argb8888.FetchTexel3Df; +#endif } diff --git a/src/mesa/drivers/dri/savage/savagetris.c b/src/mesa/drivers/dri/savage/savagetris.c index c04763b40e1..e9529d19391 100644 --- a/src/mesa/drivers/dri/savage/savagetris.c +++ b/src/mesa/drivers/dri/savage/savagetris.c @@ -879,13 +879,13 @@ static GLboolean savageCheckPTexHack( GLcontext *ctx ) RENDERINPUTS_COPY( index_bitset, tnl->render_inputs_bitset ); - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 ) && VB->TexCoordPtr[0]->size == 4) { + if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 ) && VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 4) { if (!RENDERINPUTS_TEST_RANGE( index_bitset, _TNL_ATTRIB_TEX1, _TNL_LAST_TEX )) return GL_TRUE; /* apply ptex hack */ else FALLBACK(ctx, SAVAGE_FALLBACK_PROJ_TEXTURE, GL_TRUE); } - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX1 ) && VB->TexCoordPtr[1]->size == 4) + if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX1 ) && VB->AttribPtr[_TNL_ATTRIB_TEX1]->size == 4) FALLBACK(ctx, SAVAGE_FALLBACK_PROJ_TEXTURE, GL_TRUE); return GL_FALSE; /* don't apply ptex hack */ @@ -976,13 +976,13 @@ static INLINE GLuint savageChooseVertexFormat_s3d( GLcontext *ctx ) if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 )) { if (imesa->ptexHack) EMIT_ATTR( _TNL_ATTRIB_TEX0, EMIT_3F_XYW, SAVAGE_EMIT_STQ0, SAVAGE_SKIP_ST0); - else if (VB->TexCoordPtr[0]->size == 4) + else if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 4) assert (0); /* should be caught by savageCheckPTexHack */ - else if (VB->TexCoordPtr[0]->size >= 2) + else if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size >= 2) /* The chromium menu emits some 3D tex coords even though no * 3D texture is enabled. Ignore the 3rd coordinate. */ EMIT_ATTR( _TNL_ATTRIB_TEX0, EMIT_2F, SAVAGE_EMIT_ST0, SAVAGE_SKIP_ST0 ); - else if (VB->TexCoordPtr[0]->size == 1) { + else if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 1) { EMIT_ATTR( _TNL_ATTRIB_TEX0, EMIT_1F, SAVAGE_EMIT_S0, SAVAGE_SKIP_S0 ); EMIT_PAD( 4 ); } else @@ -1025,9 +1025,9 @@ static INLINE GLuint savageChooseVertexFormat_s4( GLcontext *ctx ) if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 )) { if (imesa->ptexHack) NEED_ATTR( SAVAGE_EMIT_STQ0, SAVAGE_SKIP_ST0); - else if (VB->TexCoordPtr[0]->size == 4) + else if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 4) assert (0); /* should be caught by savageCheckPTexHack */ - else if (VB->TexCoordPtr[0]->size >= 2) + else if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size >= 2) /* The chromium menu emits some 3D tex coords even though no * 3D texture is enabled. Ignore the 3rd coordinate. */ NEED_ATTR( SAVAGE_EMIT_ST0, SAVAGE_SKIP_ST0 ); @@ -1035,10 +1035,10 @@ static INLINE GLuint savageChooseVertexFormat_s4( GLcontext *ctx ) NEED_ATTR( SAVAGE_EMIT_S0, SAVAGE_SKIP_S0 ); } if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX1 )) { - if (VB->TexCoordPtr[1]->size == 4) + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]->size == 4) /* projective textures are not supported by the hardware */ assert (0); /* should be caught by savageCheckPTexHack */ - else if (VB->TexCoordPtr[1]->size >= 2) + else if (VB->AttribPtr[_TNL_ATTRIB_TEX1]->size >= 2) NEED_ATTR( SAVAGE_EMIT_ST1, SAVAGE_SKIP_ST1 ); else NEED_ATTR( SAVAGE_EMIT_S1, SAVAGE_SKIP_S1 ); diff --git a/src/mesa/drivers/dri/sis/Makefile b/src/mesa/drivers/dri/sis/Makefile index d2354e67764..ad009fc2398 100644 --- a/src/mesa/drivers/dri/sis/Makefile +++ b/src/mesa/drivers/dri/sis/Makefile @@ -34,4 +34,3 @@ ASM_SOURCES = include ../Makefile.template -symlinks: diff --git a/src/mesa/drivers/dri/sis/sis_context.c b/src/mesa/drivers/dri/sis/sis_context.c index a070fe3d79f..f501e7ad2e9 100644 --- a/src/mesa/drivers/dri/sis/sis_context.c +++ b/src/mesa/drivers/dri/sis/sis_context.c @@ -59,7 +59,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #define need_GL_EXT_fog_coord #define need_GL_EXT_secondary_color -#include "extension_helper.h" +#include "main/remap_helper.h" #ifndef SIS_DEBUG int SIS_DEBUG = 0; @@ -69,7 +69,7 @@ int GlobalCurrentHwcx = -1; int GlobalHwcxCountBase = 1; int GlobalCmdQueueLen = 0; -struct dri_extension card_extensions[] = +static struct dri_extension card_extensions[] = { { "GL_ARB_multitexture", NULL }, { "GL_ARB_texture_border_clamp", NULL }, @@ -83,7 +83,7 @@ struct dri_extension card_extensions[] = { NULL, NULL } }; -struct dri_extension card_extensions_6326[] = +static struct dri_extension card_extensions_6326[] = { /*{ "GL_ARB_texture_border_clamp", NULL },*/ /*{ "GL_ARB_texture_mirrored_repeat", NULL },*/ diff --git a/src/mesa/drivers/dri/sis/sis_dd.c b/src/mesa/drivers/dri/sis/sis_dd.c index bddc4a9285c..217d77557f3 100644 --- a/src/mesa/drivers/dri/sis/sis_dd.c +++ b/src/mesa/drivers/dri/sis/sis_dd.c @@ -41,6 +41,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #include "sis_tris.h" #include "swrast/swrast.h" +#include "main/formats.h" #include "main/framebuffer.h" #include "main/renderbuffer.h" @@ -142,25 +143,25 @@ sisInitRenderbuffer(struct gl_renderbuffer *rb, GLenum format) if (format == GL_RGBA) { /* Color */ - rb->_BaseFormat = GL_RGBA; + rb->Format = MESA_FORMAT_ARGB8888; rb->DataType = GL_UNSIGNED_BYTE; } else if (format == GL_DEPTH_COMPONENT16) { /* Depth */ - rb->_BaseFormat = GL_DEPTH_COMPONENT; /* we always Get/Put 32-bit Z values */ + rb->Format = MESA_FORMAT_Z16; rb->DataType = GL_UNSIGNED_INT; } else if (format == GL_DEPTH_COMPONENT24) { /* Depth */ - rb->_BaseFormat = GL_DEPTH_COMPONENT; /* we always Get/Put 32-bit Z values */ + rb->Format = MESA_FORMAT_Z32; rb->DataType = GL_UNSIGNED_INT; } else { /* Stencil */ ASSERT(format == GL_STENCIL_INDEX8_EXT); - rb->_BaseFormat = GL_STENCIL_INDEX; + rb->Format = MESA_FORMAT_S8; rb->DataType = GL_UNSIGNED_BYTE; } diff --git a/src/mesa/drivers/dri/sis/sis_screen.c b/src/mesa/drivers/dri/sis/sis_screen.c index b5f04ae28d0..fec91582369 100644 --- a/src/mesa/drivers/dri/sis/sis_screen.c +++ b/src/mesa/drivers/dri/sis/sis_screen.c @@ -298,18 +298,6 @@ sisInitScreen(__DRIscreenPrivate *psp) &psp->drm_version, &drm_expected)) return NULL; - /* Calling driInitExtensions here, with a NULL context pointer, - * does not actually enable the extensions. It just makes sure - * that all the dispatch offsets for all the extensions that - * *might* be enables are known. This is needed because the - * dispatch offsets need to be known when _mesa_context_create is - * called, but we can't enable the extensions until we have a - * context pointer. - * - * Hello chicken. Hello egg. How are you two today? - */ - driInitExtensions( NULL, card_extensions, GL_FALSE ); - psp->private = sisCreateScreen(psp); if (!psp->private) { diff --git a/src/mesa/drivers/dri/sis/sis_span.c b/src/mesa/drivers/dri/sis/sis_span.c index 9e9a509755d..cfbb51007d1 100644 --- a/src/mesa/drivers/dri/sis/sis_span.c +++ b/src/mesa/drivers/dri/sis/sis_span.c @@ -176,24 +176,22 @@ sisDDInitSpanFuncs( GLcontext *ctx ) void sisSetSpanFunctions(struct sis_renderbuffer *srb, const GLvisual *vis) { - if (srb->Base.InternalFormat == GL_RGBA) { - if (vis->redBits == 5 && vis->greenBits == 6 && vis->blueBits == 5) { - sisInitPointers_RGB565( &srb->Base ); - } - else { - sisInitPointers_ARGB8888( &srb->Base ); - } + if (srb->Base.Format == MESA_FORMAT_RGB565) { + sisInitPointers_RGB565( &srb->Base ); } - else if (srb->Base.InternalFormat == GL_DEPTH_COMPONENT16) { + else if (srb->Base.Format == MESA_FORMAT_ARGB8888) { + sisInitPointers_ARGB8888( &srb->Base ); + } + else if (srb->Base.Format == MESA_FORMAT_Z16) { sisInitDepthPointers_z16(&srb->Base); } - else if (srb->Base.InternalFormat == GL_DEPTH_COMPONENT24) { + else if (srb->Base.Format == MESA_FORMAT_S8_Z24) { sisInitDepthPointers_z24_s8(&srb->Base); } - else if (srb->Base.InternalFormat == GL_DEPTH_COMPONENT32) { + else if (srb->Base.Format == MESA_FORMAT_Z32) { sisInitDepthPointers_z32(&srb->Base); } - else if (srb->Base.InternalFormat == GL_STENCIL_INDEX8_EXT) { + else if (srb->Base.Format == MESA_FORMAT_S8) { sisInitStencilPointers_z24_s8(&srb->Base); } } diff --git a/src/mesa/drivers/dri/sis/sis_tex.c b/src/mesa/drivers/dri/sis/sis_tex.c index 28ced6cfd5f..951c470dadf 100644 --- a/src/mesa/drivers/dri/sis/sis_tex.c +++ b/src/mesa/drivers/dri/sis/sis_tex.c @@ -28,17 +28,15 @@ CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. * Eric Anholt <[email protected]> */ -#include "sis_context.h" -#include "sis_alloc.h" -#include "sis_tex.h" - #include "swrast/swrast.h" #include "main/imports.h" -#include "main/texformat.h" #include "main/texstore.h" #include "main/teximage.h" #include "main/texobj.h" +#include "sis_context.h" +#include "sis_alloc.h" +#include "sis_tex.h" #include "xmlpool.h" #define ALIGN(value, align) (GLubyte *)((long)(value + align - 1) & ~(align - 1)) @@ -65,7 +63,7 @@ sisAllocTexImage( sisContextPtr smesa, sisTexObjPtr t, int level, if (t->format == 0) { t->format = image->_BaseFormat; - switch (image->TexFormat->MesaFormat) + switch (image->TexFormat) { case MESA_FORMAT_ARGB8888: t->hwformat = TEXEL_ARGB_8888_32; @@ -101,13 +99,12 @@ sisAllocTexImage( sisContextPtr smesa, sisTexObjPtr t, int level, t->hwformat = TEXEL_VUY422; break; default: - sis_fatal_error("Bad texture format 0x%x.\n", - image->TexFormat->MesaFormat); + sis_fatal_error("Bad texture format 0x%x.\n", image->TexFormat); } } assert(t->format == image->_BaseFormat); - texel_size = image->TexFormat->TexelBytes; + texel_size = _mesa_get_format_bytes(image->TexFormat); size = image->Width * image->Height * texel_size + TEXTURE_HW_PLUS; addr = sisAllocFB( smesa, size, &t->image[level].handle ); @@ -230,7 +227,7 @@ static GLboolean sisIsTextureResident( GLcontext * ctx, return (texObj->DriverData != NULL); } -static const struct gl_texture_format * +static gl_format sisChooseTextureFormat( GLcontext *ctx, GLint internalFormat, GLenum format, GLenum type ) { @@ -248,15 +245,15 @@ sisChooseTextureFormat( GLcontext *ctx, GLint internalFormat, switch ( type ) { case GL_UNSIGNED_INT_10_10_10_2: case GL_UNSIGNED_INT_2_10_10_10_REV: - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb1555; + return do32bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB1555; case GL_UNSIGNED_SHORT_4_4_4_4: case GL_UNSIGNED_SHORT_4_4_4_4_REV: - return &_mesa_texformat_argb4444; + return MESA_FORMAT_ARGB4444; case GL_UNSIGNED_SHORT_5_5_5_1: case GL_UNSIGNED_SHORT_1_5_5_5_REV: - return &_mesa_texformat_argb1555; + return MESA_FORMAT_ARGB1555; default: - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + return do32bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB4444; } case 3: @@ -265,46 +262,46 @@ sisChooseTextureFormat( GLcontext *ctx, GLint internalFormat, switch ( type ) { case GL_UNSIGNED_SHORT_4_4_4_4: case GL_UNSIGNED_SHORT_4_4_4_4_REV: - return &_mesa_texformat_argb4444; + return MESA_FORMAT_ARGB4444; case GL_UNSIGNED_SHORT_5_5_5_1: case GL_UNSIGNED_SHORT_1_5_5_5_REV: - return &_mesa_texformat_argb1555; + return MESA_FORMAT_ARGB1555; case GL_UNSIGNED_SHORT_5_6_5: case GL_UNSIGNED_SHORT_5_6_5_REV: - return &_mesa_texformat_rgb565; + return MESA_FORMAT_RGB565; default: - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565; + return do32bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_RGB565; } case GL_RGBA8: case GL_RGBA12: case GL_RGBA16: return !force16bpt ? - &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB4444; case GL_RGB10_A2: return !force16bpt ? - &_mesa_texformat_argb8888 : &_mesa_texformat_argb1555; + MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB1555; case GL_RGBA4: case GL_RGBA2: - return &_mesa_texformat_argb4444; + return MESA_FORMAT_ARGB4444; case GL_RGB5_A1: - return &_mesa_texformat_argb1555; + return MESA_FORMAT_ARGB1555; case GL_RGB8: case GL_RGB10: case GL_RGB12: case GL_RGB16: - return !force16bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565; + return !force16bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_RGB565; case GL_RGB5: case GL_RGB4: - return &_mesa_texformat_rgb565; + return MESA_FORMAT_RGB565; case GL_R3_G3_B2: - return &_mesa_texformat_rgb332; + return MESA_FORMAT_RGB332; case GL_ALPHA: case GL_ALPHA4: /* FIXME: This could use its own texstore */ @@ -312,7 +309,7 @@ sisChooseTextureFormat( GLcontext *ctx, GLint internalFormat, case GL_ALPHA12: case GL_ALPHA16: case GL_COMPRESSED_ALPHA: - return &_mesa_texformat_a8; + return MESA_FORMAT_A8; case 1: case GL_LUMINANCE: @@ -321,7 +318,7 @@ sisChooseTextureFormat( GLcontext *ctx, GLint internalFormat, case GL_LUMINANCE12: case GL_LUMINANCE16: case GL_COMPRESSED_LUMINANCE: - return &_mesa_texformat_l8; + return MESA_FORMAT_L8; case 2: case GL_LUMINANCE_ALPHA: @@ -332,7 +329,7 @@ sisChooseTextureFormat( GLcontext *ctx, GLint internalFormat, case GL_LUMINANCE12_ALPHA12: case GL_LUMINANCE16_ALPHA16: case GL_COMPRESSED_LUMINANCE_ALPHA: - return &_mesa_texformat_al88; + return MESA_FORMAT_AL88; case GL_INTENSITY: case GL_INTENSITY4: @@ -340,19 +337,19 @@ sisChooseTextureFormat( GLcontext *ctx, GLint internalFormat, case GL_INTENSITY12: case GL_INTENSITY16: case GL_COMPRESSED_INTENSITY: - return &_mesa_texformat_i8; + return MESA_FORMAT_I8; case GL_YCBCR_MESA: if (type == GL_UNSIGNED_SHORT_8_8_APPLE || type == GL_UNSIGNED_BYTE) - return &_mesa_texformat_ycbcr; + return MESA_FORMAT_YCBCR; else - return &_mesa_texformat_ycbcr_rev; + return MESA_FORMAT_YCBCR_REV; default: _mesa_problem(ctx, "unexpected format in sisDDChooseTextureFormat: %d", internalFormat); - return NULL; + return MESA_FORMAT_NONE; } } @@ -425,7 +422,7 @@ static void sisTexSubImage1D( GLcontext *ctx, /* Upload the texture */ WaitEngIdle(smesa); - texelBytes = texImage->TexFormat->TexelBytes; + texelBytes = _mesa_get_format_bytes(texImage->TexFormat); copySize = width * texelBytes; src = (char *)texImage->Data + xoffset * texelBytes; @@ -513,7 +510,7 @@ static void sisTexSubImage2D( GLcontext *ctx, /* Upload the texture */ WaitEngIdle(smesa); - texelBytes = texImage->TexFormat->TexelBytes; + texelBytes = _mesa_get_format_bytes(texImage->TexFormat); copySize = width * texelBytes; src = (char *)texImage->Data + (xoffset + yoffset * texImage->Width) * diff --git a/src/mesa/drivers/dri/sis/sis_texstate.c b/src/mesa/drivers/dri/sis/sis_texstate.c index 46417ce414c..a507173b218 100644 --- a/src/mesa/drivers/dri/sis/sis_texstate.c +++ b/src/mesa/drivers/dri/sis/sis_texstate.c @@ -36,7 +36,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/colormac.h" #include "main/context.h" #include "main/macros.h" -#include "main/texformat.h" #include "sis_context.h" #include "sis_state.h" diff --git a/src/mesa/drivers/dri/sis/sis_tris.c b/src/mesa/drivers/dri/sis/sis_tris.c index 76d12d07b3c..3cf10007b58 100644 --- a/src/mesa/drivers/dri/sis/sis_tris.c +++ b/src/mesa/drivers/dri/sis/sis_tris.c @@ -903,14 +903,14 @@ static void sisRenderStart( GLcontext *ctx ) /* projective textures are not supported by the hardware */ if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 )) { - if (VB->TexCoordPtr[0]->size > 2) + if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size > 2) tex_fallback = GL_TRUE; EMIT_ATTR(_TNL_ATTRIB_TEX0, EMIT_2F); AGPParseSet |= SiS_PS_HAS_UV0; } /* Will only hit tex1 on SiS300 */ if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX1 )) { - if (VB->TexCoordPtr[1]->size > 2) + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]->size > 2) tex_fallback = GL_TRUE; EMIT_ATTR(_TNL_ATTRIB_TEX1, EMIT_2F); AGPParseSet |= SiS_PS_HAS_UV1; diff --git a/src/mesa/drivers/dri/swrast/Makefile b/src/mesa/drivers/dri/swrast/Makefile index 5f3a4f2191b..771169c1ff9 100644 --- a/src/mesa/drivers/dri/swrast/Makefile +++ b/src/mesa/drivers/dri/swrast/Makefile @@ -21,4 +21,3 @@ SWRAST_COMMON_SOURCES = \ include ../Makefile.template -symlinks: diff --git a/src/mesa/drivers/dri/swrast/swrast.c b/src/mesa/drivers/dri/swrast/swrast.c index 3016987d561..8340861aff8 100644 --- a/src/mesa/drivers/dri/swrast/swrast.c +++ b/src/mesa/drivers/dri/swrast/swrast.c @@ -33,6 +33,7 @@ #include "main/context.h" #include "main/extensions.h" +#include "main/formats.h" #include "main/framebuffer.h" #include "main/imports.h" #include "main/renderbuffer.h" @@ -43,82 +44,12 @@ #include "tnl/t_pipeline.h" #include "vbo/vbo.h" #include "drivers/common/driverfuncs.h" +#include "drivers/common/meta.h" #include "utils.h" #include "swrast_priv.h" -#define need_GL_VERSION_1_3 -#define need_GL_VERSION_1_4 -#define need_GL_VERSION_1_5 -#define need_GL_VERSION_2_0 -#define need_GL_VERSION_2_1 - -/* sw extensions for imaging */ -#define need_GL_EXT_blend_color -#define need_GL_EXT_blend_minmax -#define need_GL_EXT_convolution -#define need_GL_EXT_histogram -#define need_GL_SGI_color_table - -/* sw extensions not associated with some GL version */ -#define need_GL_ARB_draw_elements_base_vertex -#define need_GL_ARB_shader_objects -#define need_GL_ARB_vertex_array_object -#define need_GL_ARB_vertex_program -#define need_GL_ARB_sync -#define need_GL_APPLE_vertex_array_object -#define need_GL_ATI_fragment_shader -#define need_GL_ATI_separate_stencil -#define need_GL_EXT_depth_bounds_test -#define need_GL_EXT_framebuffer_object -#define need_GL_EXT_framebuffer_blit -#define need_GL_EXT_gpu_program_parameters -#define need_GL_EXT_paletted_texture -#define need_GL_EXT_stencil_two_side -#define need_GL_MESA_resize_buffers -#define need_GL_NV_vertex_program -#define need_GL_NV_fragment_program - -#include "extension_helper.h" - -const struct dri_extension card_extensions[] = -{ - { "GL_VERSION_1_3", GL_VERSION_1_3_functions }, - { "GL_VERSION_1_4", GL_VERSION_1_4_functions }, - { "GL_VERSION_1_5", GL_VERSION_1_5_functions }, - { "GL_VERSION_2_0", GL_VERSION_2_0_functions }, - { "GL_VERSION_2_1", GL_VERSION_2_1_functions }, - - { "GL_EXT_blend_color", GL_EXT_blend_color_functions }, - { "GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions }, - { "GL_EXT_convolution", GL_EXT_convolution_functions }, - { "GL_EXT_histogram", GL_EXT_histogram_functions }, - { "GL_SGI_color_table", GL_SGI_color_table_functions }, - - { "GL_ARB_depth_clamp", NULL }, - { "GL_ARB_draw_elements_base_vertex", GL_ARB_draw_elements_base_vertex_functions }, - { "GL_ARB_shader_objects", GL_ARB_shader_objects_functions }, - { "GL_ARB_vertex_array_object", GL_ARB_vertex_array_object_functions }, - { "GL_ARB_vertex_program", GL_ARB_vertex_program_functions }, - { "GL_ARB_sync", GL_ARB_sync_functions }, - { "GL_APPLE_vertex_array_object", GL_APPLE_vertex_array_object_functions }, - { "GL_ATI_fragment_shader", GL_ATI_fragment_shader_functions }, - { "GL_ATI_separate_stencil", GL_ATI_separate_stencil_functions }, - { "GL_EXT_depth_bounds_test", GL_EXT_depth_bounds_test_functions }, - { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions }, - { "GL_EXT_framebuffer_blit", GL_EXT_framebuffer_blit_functions }, - { "GL_EXT_gpu_program_parameters", GL_EXT_gpu_program_parameters_functions }, - { "GL_EXT_paletted_texture", GL_EXT_paletted_texture_functions }, - { "GL_EXT_stencil_two_side", GL_EXT_stencil_two_side_functions }, - { "GL_MESA_resize_buffers", GL_MESA_resize_buffers_functions }, - { "GL_NV_depth_clamp", NULL }, - { "GL_NV_vertex_program", GL_NV_vertex_program_functions }, - { "GL_NV_fragment_program", GL_NV_fragment_program_functions }, - { NULL, NULL } -}; - - /** * Screen and config-related functions */ @@ -242,7 +173,7 @@ driCreateNewScreen(int scrn, const __DRIextension **extensions, *driver_configs = (const __DRIconfig **) driConcatConfigs(configs24, configs32); - driInitExtensions( NULL, card_extensions, GL_FALSE ); + driInitExtensions( NULL, NULL, GL_FALSE ); return psp; } @@ -376,50 +307,38 @@ swrast_new_renderbuffer(const GLvisual *visual, GLboolean front) switch (pixel_format) { case PF_A8R8G8B8: + xrb->Base.Format = MESA_FORMAT_ARGB8888; xrb->Base.InternalFormat = GL_RGBA; xrb->Base._BaseFormat = GL_RGBA; xrb->Base.DataType = GL_UNSIGNED_BYTE; - xrb->Base.RedBits = 8 * sizeof(GLubyte); - xrb->Base.GreenBits = 8 * sizeof(GLubyte); - xrb->Base.BlueBits = 8 * sizeof(GLubyte); - xrb->Base.AlphaBits = 8 * sizeof(GLubyte); xrb->bpp = 32; break; case PF_X8R8G8B8: + xrb->Base.Format = MESA_FORMAT_ARGB8888; /* XXX */ xrb->Base.InternalFormat = GL_RGB; xrb->Base._BaseFormat = GL_RGB; xrb->Base.DataType = GL_UNSIGNED_BYTE; - xrb->Base.RedBits = 8 * sizeof(GLubyte); - xrb->Base.GreenBits = 8 * sizeof(GLubyte); - xrb->Base.BlueBits = 8 * sizeof(GLubyte); - xrb->Base.AlphaBits = 0; xrb->bpp = 32; break; case PF_R5G6B5: + xrb->Base.Format = MESA_FORMAT_RGB565; xrb->Base.InternalFormat = GL_RGB; xrb->Base._BaseFormat = GL_RGB; xrb->Base.DataType = GL_UNSIGNED_BYTE; - xrb->Base.RedBits = 5 * sizeof(GLubyte); - xrb->Base.GreenBits = 6 * sizeof(GLubyte); - xrb->Base.BlueBits = 5 * sizeof(GLubyte); - xrb->Base.AlphaBits = 0; xrb->bpp = 16; break; case PF_R3G3B2: + xrb->Base.Format = MESA_FORMAT_RGB332; xrb->Base.InternalFormat = GL_RGB; xrb->Base._BaseFormat = GL_RGB; xrb->Base.DataType = GL_UNSIGNED_BYTE; - xrb->Base.RedBits = 3 * sizeof(GLubyte); - xrb->Base.GreenBits = 3 * sizeof(GLubyte); - xrb->Base.BlueBits = 2 * sizeof(GLubyte); - xrb->Base.AlphaBits = 0; xrb->bpp = 8; break; case PF_CI8: + xrb->Base.Format = MESA_FORMAT_CI8; xrb->Base.InternalFormat = GL_COLOR_INDEX8_EXT; xrb->Base._BaseFormat = GL_COLOR_INDEX; xrb->Base.DataType = GL_UNSIGNED_BYTE; - xrb->Base.IndexBits = 8 * sizeof(GLubyte); xrb->bpp = 8; break; default: @@ -649,6 +568,8 @@ driCreateNewContext(__DRIscreen *screen, const __DRIconfig *config, _mesa_enable_2_0_extensions(mesaCtx); _mesa_enable_2_1_extensions(mesaCtx); + _mesa_meta_init(mesaCtx); + return ctx; } @@ -660,6 +581,7 @@ driDestroyContext(__DRIcontext *ctx) if (ctx) { mesaCtx = &ctx->Base; + _mesa_meta_free(mesaCtx); _swsetup_DestroyContext( mesaCtx ); _swrast_DestroyContext( mesaCtx ); _tnl_DestroyContext( mesaCtx ); diff --git a/src/mesa/drivers/dri/swrast/swrast_span.c b/src/mesa/drivers/dri/swrast/swrast_span.c index 2d3c25dcbef..f8e503463fa 100644 --- a/src/mesa/drivers/dri/swrast/swrast_span.c +++ b/src/mesa/drivers/dri/swrast/swrast_span.c @@ -63,56 +63,42 @@ static const GLubyte kernel[16] = { /* 32-bit BGRA */ #define STORE_PIXEL_A8R8G8B8(DST, X, Y, VALUE) \ - DST[3] = VALUE[ACOMP]; \ - DST[2] = VALUE[RCOMP]; \ - DST[1] = VALUE[GCOMP]; \ - DST[0] = VALUE[BCOMP] + *DST = VALUE[ACOMP] << 24 | VALUE[RCOMP] << 16 | VALUE[GCOMP] << 8 | VALUE[BCOMP] #define STORE_PIXEL_RGB_A8R8G8B8(DST, X, Y, VALUE) \ - DST[3] = 0xff; \ - DST[2] = VALUE[RCOMP]; \ - DST[1] = VALUE[GCOMP]; \ - DST[0] = VALUE[BCOMP] + *DST = 0xff << 24 | VALUE[RCOMP] << 16 | VALUE[GCOMP] << 8 | VALUE[BCOMP] #define FETCH_PIXEL_A8R8G8B8(DST, SRC) \ - DST[ACOMP] = SRC[3]; \ - DST[RCOMP] = SRC[2]; \ - DST[GCOMP] = SRC[1]; \ - DST[BCOMP] = SRC[0] + DST[ACOMP] = *SRC >> 24; \ + DST[RCOMP] = (*SRC >> 16) & 0xff; \ + DST[GCOMP] = (*SRC >> 8) & 0xff; \ + DST[BCOMP] = *SRC & 0xff /* 32-bit BGRX */ #define STORE_PIXEL_X8R8G8B8(DST, X, Y, VALUE) \ - DST[3] = 0xff; \ - DST[2] = VALUE[RCOMP]; \ - DST[1] = VALUE[GCOMP]; \ - DST[0] = VALUE[BCOMP] + *DST = 0xff << 24 | VALUE[RCOMP] << 16 | VALUE[GCOMP] << 8 | VALUE[BCOMP] #define STORE_PIXEL_RGB_X8R8G8B8(DST, X, Y, VALUE) \ - DST[3] = 0xff; \ - DST[2] = VALUE[RCOMP]; \ - DST[1] = VALUE[GCOMP]; \ - DST[0] = VALUE[BCOMP] + *DST = 0xff << 24 | VALUE[RCOMP] << 16 | VALUE[GCOMP] << 8 | VALUE[BCOMP] #define FETCH_PIXEL_X8R8G8B8(DST, SRC) \ - DST[ACOMP] = 0xff; \ - DST[RCOMP] = SRC[2]; \ - DST[GCOMP] = SRC[1]; \ - DST[BCOMP] = SRC[0] + DST[ACOMP] = 0xff; \ + DST[RCOMP] = (*SRC >> 16) & 0xff; \ + DST[GCOMP] = (*SRC >> 8) & 0xff; \ + DST[BCOMP] = *SRC & 0xff /* 16-bit BGR */ #define STORE_PIXEL_R5G6B5(DST, X, Y, VALUE) \ do { \ int d = DITHER_COMP(X, Y) >> 6; \ - GLushort *p = (GLushort *)DST; \ - *p = ( ((DITHER_CLAMP((VALUE[RCOMP]) + d) & 0xf8) << 8) | \ - ((DITHER_CLAMP((VALUE[GCOMP]) + d) & 0xfc) << 3) | \ - ((DITHER_CLAMP((VALUE[BCOMP]) + d) & 0xf8) >> 3) ); \ + *DST = ( ((DITHER_CLAMP((VALUE[RCOMP]) + d) & 0xf8) << 8) | \ + ((DITHER_CLAMP((VALUE[GCOMP]) + d) & 0xfc) << 3) | \ + ((DITHER_CLAMP((VALUE[BCOMP]) + d) & 0xf8) >> 3) ); \ } while(0) #define FETCH_PIXEL_R5G6B5(DST, SRC) \ do { \ - GLushort p = *(GLushort *)SRC; \ DST[ACOMP] = 0xff; \ - DST[RCOMP] = ((p >> 8) & 0xf8) * 255 / 0xf8; \ - DST[GCOMP] = ((p >> 3) & 0xfc) * 255 / 0xfc; \ - DST[BCOMP] = ((p << 3) & 0xf8) * 255 / 0xf8; \ + DST[RCOMP] = ((*SRC >> 8) & 0xf8) * 255 / 0xf8; \ + DST[GCOMP] = ((*SRC >> 3) & 0xfc) * 255 / 0xfc; \ + DST[BCOMP] = ((*SRC << 3) & 0xf8) * 255 / 0xf8; \ } while(0) @@ -145,8 +131,8 @@ static const GLubyte kernel[16] = { #define SPAN_VARS \ struct swrast_renderbuffer *xrb = swrast_renderbuffer(rb); #define INIT_PIXEL_PTR(P, X, Y) \ - GLubyte *P = (GLubyte *)xrb->Base.Data + YFLIP(xrb, Y) * xrb->pitch + (X) * 4; -#define INC_PIXEL_PTR(P) P += 4 + GLuint *P = (GLuint *)xrb->Base.Data + YFLIP(xrb, Y) * xrb->pitch / 4 + (X) +#define INC_PIXEL_PTR(P) P++ #define STORE_PIXEL(DST, X, Y, VALUE) \ STORE_PIXEL_A8R8G8B8(DST, X, Y, VALUE) #define STORE_PIXEL_RGB(DST, X, Y, VALUE) \ @@ -163,8 +149,8 @@ static const GLubyte kernel[16] = { #define SPAN_VARS \ struct swrast_renderbuffer *xrb = swrast_renderbuffer(rb); #define INIT_PIXEL_PTR(P, X, Y) \ - GLubyte *P = (GLubyte *)xrb->Base.Data + YFLIP(xrb, Y) * xrb->pitch + (X) * 4; -#define INC_PIXEL_PTR(P) P += 4 + GLuint *P = (GLuint *)xrb->Base.Data + YFLIP(xrb, Y) * xrb->pitch / 4 + (X); +#define INC_PIXEL_PTR(P) P++ #define STORE_PIXEL(DST, X, Y, VALUE) \ STORE_PIXEL_X8R8G8B8(DST, X, Y, VALUE) #define STORE_PIXEL_RGB(DST, X, Y, VALUE) \ @@ -181,8 +167,8 @@ static const GLubyte kernel[16] = { #define SPAN_VARS \ struct swrast_renderbuffer *xrb = swrast_renderbuffer(rb); #define INIT_PIXEL_PTR(P, X, Y) \ - GLubyte *P = (GLubyte *)xrb->Base.Data + YFLIP(xrb, Y) * xrb->pitch + (X) * 2; -#define INC_PIXEL_PTR(P) P += 2 + GLushort *P = (GLushort *)xrb->Base.Data + YFLIP(xrb, Y) * xrb->pitch / 2 + (X); +#define INC_PIXEL_PTR(P) P++ #define STORE_PIXEL(DST, X, Y, VALUE) \ STORE_PIXEL_R5G6B5(DST, X, Y, VALUE) #define FETCH_PIXEL(DST, SRC) \ @@ -234,8 +220,8 @@ static const GLubyte kernel[16] = { #define SPAN_VARS \ struct swrast_renderbuffer *xrb = swrast_renderbuffer(rb); #define INIT_PIXEL_PTR(P, X, Y) \ - GLubyte *P = (GLubyte *)row; -#define INC_PIXEL_PTR(P) P += 4 + GLuint *P = (GLuint *)row; +#define INC_PIXEL_PTR(P) P++ #define STORE_PIXEL(DST, X, Y, VALUE) \ STORE_PIXEL_A8R8G8B8(DST, X, Y, VALUE) #define STORE_PIXEL_RGB(DST, X, Y, VALUE) \ @@ -252,8 +238,8 @@ static const GLubyte kernel[16] = { #define SPAN_VARS \ struct swrast_renderbuffer *xrb = swrast_renderbuffer(rb); #define INIT_PIXEL_PTR(P, X, Y) \ - GLubyte *P = (GLubyte *)row; -#define INC_PIXEL_PTR(P) P += 4 + GLuint *P = (GLuint *)row; +#define INC_PIXEL_PTR(P) P++ #define STORE_PIXEL(DST, X, Y, VALUE) \ STORE_PIXEL_X8R8G8B8(DST, X, Y, VALUE) #define STORE_PIXEL_RGB(DST, X, Y, VALUE) \ @@ -270,7 +256,7 @@ static const GLubyte kernel[16] = { #define SPAN_VARS \ struct swrast_renderbuffer *xrb = swrast_renderbuffer(rb); #define INIT_PIXEL_PTR(P, X, Y) \ - GLubyte *P = (GLubyte *)row; + GLushort *P = (GLushort *)row; #define INC_PIXEL_PTR(P) P += 2 #define STORE_PIXEL(DST, X, Y, VALUE) \ STORE_PIXEL_R5G6B5(DST, X, Y, VALUE) diff --git a/src/mesa/drivers/dri/tdfx/Makefile b/src/mesa/drivers/dri/tdfx/Makefile index 092c580feac..b9f25db4fe8 100644 --- a/src/mesa/drivers/dri/tdfx/Makefile +++ b/src/mesa/drivers/dri/tdfx/Makefile @@ -32,6 +32,4 @@ ASM_SOURCES = include ../Makefile.template -symlinks: - diff --git a/src/mesa/drivers/dri/tdfx/tdfx_context.c b/src/mesa/drivers/dri/tdfx/tdfx_context.c index 68b5027561b..e742d414a59 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_context.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_context.c @@ -68,13 +68,13 @@ #define need_GL_EXT_paletted_texture /* #define need_GL_EXT_secondary_color */ /* #define need_GL_NV_vertex_program */ -#include "extension_helper.h" +#include "main/remap_helper.h" /** * Common extension strings exported by all cards */ -const struct dri_extension card_extensions[] = +static const struct dri_extension card_extensions[] = { { "GL_ARB_occlusion_query", GL_ARB_occlusion_query_functions }, { "GL_ARB_texture_mirrored_repeat", NULL }, @@ -107,7 +107,7 @@ const struct dri_extension card_extensions[] = /** * Extension strings exported only by Naplam (e.g., Voodoo4 & Voodoo5) cards. */ -const struct dri_extension napalm_extensions[] = +static const struct dri_extension napalm_extensions[] = { { "GL_ARB_texture_env_combine", NULL }, { "GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions }, diff --git a/src/mesa/drivers/dri/tdfx/tdfx_screen.c b/src/mesa/drivers/dri/tdfx/tdfx_screen.c index 58bd48b294c..2eb0024d40b 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_screen.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_screen.c @@ -69,9 +69,6 @@ static const __DRIextension *tdfxExtensions[] = { static const GLuint __driNConfigOptions = 1; -extern const struct dri_extension card_extensions[]; -extern const struct dri_extension napalm_extensions[]; - static GLboolean tdfxCreateScreen( __DRIscreenPrivate *sPriv ) { @@ -173,7 +170,7 @@ tdfxCreateBuffer( __DRIscreenPrivate *driScrnPriv, { driRenderbuffer *frontRb - = driNewRenderbuffer(GL_RGBA, NULL, screen->cpp, + = driNewRenderbuffer(MESA_FORMAT_ARGB8888, NULL, screen->cpp, screen->fbOffset, screen->width, driDrawPriv); tdfxSetSpanFunctions(frontRb, mesaVis); _mesa_add_renderbuffer(fb, BUFFER_FRONT_LEFT, &frontRb->Base); @@ -181,7 +178,7 @@ tdfxCreateBuffer( __DRIscreenPrivate *driScrnPriv, if (mesaVis->doubleBufferMode) { driRenderbuffer *backRb - = driNewRenderbuffer(GL_RGBA, NULL, screen->cpp, + = driNewRenderbuffer(MESA_FORMAT_ARGB8888, NULL, screen->cpp, screen->backOffset, screen->width, driDrawPriv); tdfxSetSpanFunctions(backRb, mesaVis); @@ -191,7 +188,7 @@ tdfxCreateBuffer( __DRIscreenPrivate *driScrnPriv, if (mesaVis->depthBits == 16) { driRenderbuffer *depthRb - = driNewRenderbuffer(GL_DEPTH_COMPONENT16, NULL, screen->cpp, + = driNewRenderbuffer(MESA_FORMAT_Z16, NULL, screen->cpp, screen->depthOffset, screen->width, driDrawPriv); tdfxSetSpanFunctions(depthRb, mesaVis); @@ -199,7 +196,7 @@ tdfxCreateBuffer( __DRIscreenPrivate *driScrnPriv, } else if (mesaVis->depthBits == 24) { driRenderbuffer *depthRb - = driNewRenderbuffer(GL_DEPTH_COMPONENT24, NULL, screen->cpp, + = driNewRenderbuffer(MESA_FORMAT_Z24_S8, NULL, screen->cpp, screen->depthOffset, screen->width, driDrawPriv); tdfxSetSpanFunctions(depthRb, mesaVis); @@ -208,7 +205,7 @@ tdfxCreateBuffer( __DRIscreenPrivate *driScrnPriv, if (mesaVis->stencilBits > 0) { driRenderbuffer *stencilRb - = driNewRenderbuffer(GL_STENCIL_INDEX8_EXT, NULL, screen->cpp, + = driNewRenderbuffer(MESA_FORMAT_S8, NULL, screen->cpp, screen->depthOffset, screen->width, driDrawPriv); tdfxSetSpanFunctions(stencilRb, mesaVis); @@ -418,19 +415,6 @@ tdfxInitScreen(__DRIscreen *psp) &psp->drm_version, & drm_expected ) ) return NULL; - /* Calling driInitExtensions here, with a NULL context pointer, - * does not actually enable the extensions. It just makes sure - * that all the dispatch offsets for all the extensions that - * *might* be enables are known. This is needed because the - * dispatch offsets need to be known when _mesa_context_create is - * called, but we can't enable the extensions until we have a - * context pointer. - * - * Hello chicken. Hello egg. How are you two today? - */ - driInitExtensions( NULL, card_extensions, GL_FALSE ); - driInitExtensions( NULL, napalm_extensions, GL_FALSE ); - if (!tdfxInitDriver(psp)) return NULL; diff --git a/src/mesa/drivers/dri/tdfx/tdfx_state.c b/src/mesa/drivers/dri/tdfx/tdfx_state.c index 591df8a905a..cf2712720f3 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_state.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_state.c @@ -40,7 +40,6 @@ #include "main/mtypes.h" #include "main/colormac.h" -#include "main/texformat.h" #include "main/texstore.h" #include "main/teximage.h" diff --git a/src/mesa/drivers/dri/tdfx/tdfx_tex.c b/src/mesa/drivers/dri/tdfx/tdfx_tex.c index f6a48b3ae12..0aa09e733be 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_tex.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_tex.c @@ -42,7 +42,6 @@ #include "main/image.h" #include "main/mipmap.h" #include "main/texcompress.h" -#include "main/texformat.h" #include "main/teximage.h" #include "main/texstore.h" #include "main/texobj.h" @@ -72,13 +71,13 @@ _mesa_halve2x2_teximage2d ( GLcontext *ctx, GLubyte *_d = NULL; GLenum _t = 0; - if (texImage->TexFormat->MesaFormat == MESA_FORMAT_RGB565) { + if (texImage->TexFormat == MESA_FORMAT_RGB565) { _t = GL_UNSIGNED_SHORT_5_6_5_REV; bpt = bytesPerPixel; - } else if (texImage->TexFormat->MesaFormat == MESA_FORMAT_ARGB4444) { + } else if (texImage->TexFormat == MESA_FORMAT_ARGB4444) { _t = GL_UNSIGNED_SHORT_4_4_4_4_REV; bpt = bytesPerPixel; - } else if (texImage->TexFormat->MesaFormat == MESA_FORMAT_ARGB1555) { + } else if (texImage->TexFormat == MESA_FORMAT_ARGB1555) { _t = GL_UNSIGNED_SHORT_1_5_5_5_REV; bpt = bytesPerPixel; } @@ -93,13 +92,13 @@ _mesa_halve2x2_teximage2d ( GLcontext *ctx, } _s = src = MALLOC(srcRowStride * srcHeight); _d = dst = MALLOC(dstWidth * bytesPerPixel * dstHeight); - _mesa_texstore_rgba8888(ctx, 2, GL_RGBA, - &_mesa_texformat_rgba8888_rev, src, - 0, 0, 0, /* dstX/Y/Zoffset */ - srcRowStride, /* dstRowStride */ - &dstImageOffsets, - srcWidth, srcHeight, 1, - texImage->_BaseFormat, _t, srcImage, &ctx->DefaultPacking); + _mesa_texstore(ctx, 2, GL_RGBA, + MESA_FORMAT_RGBA8888_REV, src, + 0, 0, 0, /* dstX/Y/Zoffset */ + srcRowStride, /* dstRowStride */ + &dstImageOffsets, + srcWidth, srcHeight, 1, + texImage->_BaseFormat, _t, srcImage, &ctx->DefaultPacking); } if (srcHeight == 1) { @@ -140,13 +139,13 @@ _mesa_halve2x2_teximage2d ( GLcontext *ctx, if (bpt) { src = _s; dst = _d; - texImage->TexFormat->StoreImage(ctx, 2, texImage->_BaseFormat, - texImage->TexFormat, dstImage, - 0, 0, 0, /* dstX/Y/Zoffset */ - dstWidth * bpt, - &dstImageOffsets, - dstWidth, dstHeight, 1, - GL_BGRA, CHAN_TYPE, dst, &ctx->DefaultPacking); + _mesa_texstore(ctx, 2, texImage->_BaseFormat, + texImage->TexFormat, dstImage, + 0, 0, 0, /* dstX/Y/Zoffset */ + dstWidth * bpt, + &dstImageOffsets, + dstWidth, dstHeight, 1, + GL_BGRA, CHAN_TYPE, dst, &ctx->DefaultPacking); FREE(dst); FREE(src); } @@ -190,7 +189,7 @@ tdfxGenerateMipmap(GLcontext *ctx, GLenum target, const tdfxMipMapLevel *mml; texImage = _mesa_get_tex_image(ctx, texObj, target, level); - assert(!texImage->IsCompressed); + texelBytes = _mesa_get_format_bytes(texImage->TexFormat); mml = TDFX_TEXIMAGE_DATA(texImage); @@ -760,7 +759,7 @@ fxTexusError(const char *string, FxBool fatal) #endif -static const struct gl_texture_format * +static gl_format tdfxChooseTextureFormat( GLcontext *ctx, GLint internalFormat, GLenum srcFormat, GLenum srcType ) { @@ -774,7 +773,7 @@ tdfxChooseTextureFormat( GLcontext *ctx, GLint internalFormat, case GL_ALPHA12: case GL_ALPHA16: case GL_COMPRESSED_ALPHA: - return &_mesa_texformat_a8; + return MESA_FORMAT_A8; case 1: case GL_LUMINANCE: case GL_LUMINANCE4: @@ -782,7 +781,7 @@ tdfxChooseTextureFormat( GLcontext *ctx, GLint internalFormat, case GL_LUMINANCE12: case GL_LUMINANCE16: case GL_COMPRESSED_LUMINANCE: - return &_mesa_texformat_l8; + return MESA_FORMAT_L8; case 2: case GL_LUMINANCE_ALPHA: case GL_LUMINANCE4_ALPHA4: @@ -792,48 +791,47 @@ tdfxChooseTextureFormat( GLcontext *ctx, GLint internalFormat, case GL_LUMINANCE12_ALPHA12: case GL_LUMINANCE16_ALPHA16: case GL_COMPRESSED_LUMINANCE_ALPHA: - return &_mesa_texformat_al88; + return MESA_FORMAT_AL88; case GL_INTENSITY: case GL_INTENSITY4: case GL_INTENSITY8: case GL_INTENSITY12: case GL_INTENSITY16: case GL_COMPRESSED_INTENSITY: - return &_mesa_texformat_i8; + return MESA_FORMAT_I8; case GL_R3_G3_B2: case GL_RGB4: case GL_RGB5: - return &_mesa_texformat_rgb565; + return MESA_FORMAT_RGB565; case GL_COMPRESSED_RGB: /* intentional fall-through */ case 3: case GL_RGB: if ( srcFormat == GL_RGB && srcType == GL_UNSIGNED_SHORT_5_6_5 ) { - return &_mesa_texformat_rgb565; + return MESA_FORMAT_RGB565; } /* intentional fall through */ case GL_RGB8: case GL_RGB10: case GL_RGB12: case GL_RGB16: - return (allow32bpt) ? &_mesa_texformat_argb8888 - : &_mesa_texformat_rgb565; + return (allow32bpt) ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_RGB565; case GL_RGBA2: case GL_RGBA4: - return &_mesa_texformat_argb4444; + return MESA_FORMAT_ARGB4444; case GL_COMPRESSED_RGBA: /* intentional fall-through */ case 4: case GL_RGBA: if ( srcFormat == GL_BGRA ) { if ( srcType == GL_UNSIGNED_INT_8_8_8_8_REV ) { - return &_mesa_texformat_argb8888; + return MESA_FORMAT_ARGB8888; } else if ( srcType == GL_UNSIGNED_SHORT_4_4_4_4_REV ) { - return &_mesa_texformat_argb4444; + return MESA_FORMAT_ARGB4444; } else if ( srcType == GL_UNSIGNED_SHORT_1_5_5_5_REV ) { - return &_mesa_texformat_argb1555; + return MESA_FORMAT_ARGB1555; } } /* intentional fall through */ @@ -841,10 +839,9 @@ tdfxChooseTextureFormat( GLcontext *ctx, GLint internalFormat, case GL_RGB10_A2: case GL_RGBA12: case GL_RGBA16: - return allow32bpt ? &_mesa_texformat_argb8888 - : &_mesa_texformat_argb4444; + return allow32bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB4444; case GL_RGB5_A1: - return &_mesa_texformat_argb1555; + return MESA_FORMAT_ARGB1555; case GL_COLOR_INDEX: case GL_COLOR_INDEX1_EXT: case GL_COLOR_INDEX2_EXT: @@ -852,29 +849,29 @@ tdfxChooseTextureFormat( GLcontext *ctx, GLint internalFormat, case GL_COLOR_INDEX8_EXT: case GL_COLOR_INDEX12_EXT: case GL_COLOR_INDEX16_EXT: - return &_mesa_texformat_ci8; + return MESA_FORMAT_CI8; /* GL_EXT_texture_compression_s3tc */ /* GL_S3_s3tc */ case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: case GL_RGB_S3TC: case GL_RGB4_S3TC: - return &_mesa_texformat_rgb_dxt1; + return MESA_FORMAT_RGB_DXT1; case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: - return &_mesa_texformat_rgba_dxt1; + return MESA_FORMAT_RGBA_DXT1; case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: case GL_RGBA_S3TC: case GL_RGBA4_S3TC: - return &_mesa_texformat_rgba_dxt3; + return MESA_FORMAT_RGBA_DXT3; case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: - return &_mesa_texformat_rgba_dxt5; + return MESA_FORMAT_RGBA_DXT5; /* GL_3DFX_texture_compression_FXT1 */ case GL_COMPRESSED_RGB_FXT1_3DFX: - return &_mesa_texformat_rgb_fxt1; + return MESA_FORMAT_RGB_FXT1; case GL_COMPRESSED_RGBA_FXT1_3DFX: - return &_mesa_texformat_rgba_fxt1; + return MESA_FORMAT_RGBA_FXT1; default: _mesa_problem(ctx, "unexpected format in tdfxChooseTextureFormat"); - return NULL; + return MESA_FORMAT_NONE; } } @@ -1126,7 +1123,9 @@ fetch_rgb_dxt1(const struct gl_texture_image *texImage, i = i * mml->wScale; j = j * mml->hScale; + /* XXX Get fetch func from _mesa_get_texel_fetch_func() _mesa_texformat_rgb_dxt1.FetchTexel2D(texImage, i, j, k, rgba); + */ } @@ -1139,7 +1138,9 @@ fetch_rgba_dxt1(const struct gl_texture_image *texImage, i = i * mml->wScale; j = j * mml->hScale; + /* XXX Get fetch func from _mesa_get_texel_fetch_func() _mesa_texformat_rgba_dxt1.FetchTexel2D(texImage, i, j, k, rgba); + */ } @@ -1152,7 +1153,9 @@ fetch_rgba_dxt3(const struct gl_texture_image *texImage, i = i * mml->wScale; j = j * mml->hScale; + /* XXX Get fetch func from _mesa_get_texel_fetch_func() _mesa_texformat_rgba_dxt3.FetchTexel2D(texImage, i, j, k, rgba); + */ } @@ -1165,7 +1168,9 @@ fetch_rgba_dxt5(const struct gl_texture_image *texImage, i = i * mml->wScale; j = j * mml->hScale; + /* XXX Get fetch func from _mesa_get_texel_fetch_func() _mesa_texformat_rgba_dxt5.FetchTexel2D(texImage, i, j, k, rgba); + */ } @@ -1226,20 +1231,21 @@ adjust2DRatio (GLcontext *ctx, GLvoid *tempImage; GLuint dstImageOffsets = 0; - if (!texImage->IsCompressed) { + if (!_mesa_is_format_compressed(texImage->TexFormat)) { GLubyte *destAddr; + tempImage = MALLOC(width * height * texelBytes); if (!tempImage) { return GL_FALSE; } - texImage->TexFormat->StoreImage(ctx, 2, texImage->_BaseFormat, - texImage->TexFormat, tempImage, - 0, 0, 0, /* dstX/Y/Zoffset */ - width * texelBytes, /* dstRowStride */ - &dstImageOffsets, - width, height, 1, - format, type, pixels, packing); + _mesa_texstore(ctx, 2, texImage->_BaseFormat, + texImage->TexFormat, tempImage, + 0, 0, 0, /* dstX/Y/Zoffset */ + width * texelBytes, /* dstRowStride */ + &dstImageOffsets, + width, height, 1, + format, type, pixels, packing); /* now rescale */ /* compute address of dest subimage within the overal tex image */ @@ -1256,6 +1262,7 @@ adjust2DRatio (GLcontext *ctx, } else { const GLint rawBytes = 4; GLvoid *rawImage = MALLOC(width * height * rawBytes); + if (!rawImage) { return GL_FALSE; } @@ -1265,26 +1272,26 @@ adjust2DRatio (GLcontext *ctx, return GL_FALSE; } /* unpack image, apply transfer ops and store in rawImage */ - _mesa_texstore_rgba8888(ctx, 2, GL_RGBA, - &_mesa_texformat_rgba8888_rev, rawImage, - 0, 0, 0, /* dstX/Y/Zoffset */ - width * rawBytes, /* dstRowStride */ - &dstImageOffsets, - width, height, 1, - format, type, pixels, packing); + _mesa_texstore(ctx, 2, GL_RGBA, + MESA_FORMAT_RGBA8888_REV, rawImage, + 0, 0, 0, /* dstX/Y/Zoffset */ + width * rawBytes, /* dstRowStride */ + &dstImageOffsets, + width, height, 1, + format, type, pixels, packing); _mesa_rescale_teximage2d(rawBytes, width, newWidth * rawBytes, /* dst stride */ width, height, /* src */ newWidth, newHeight, /* dst */ rawImage /*src*/, tempImage /*dst*/ ); - texImage->TexFormat->StoreImage(ctx, 2, texImage->_BaseFormat, - texImage->TexFormat, texImage->Data, - xoffset * mml->wScale, yoffset * mml->hScale, 0, /* dstX/Y/Zoffset */ - dstRowStride, - &dstImageOffsets, - newWidth, newHeight, 1, - GL_RGBA, CHAN_TYPE, tempImage, &ctx->DefaultPacking); + _mesa_texstore(ctx, 2, texImage->_BaseFormat, + texImage->TexFormat, texImage->Data, + xoffset * mml->wScale, yoffset * mml->hScale, 0, /* dstX/Y/Zoffset */ + dstRowStride, + &dstImageOffsets, + newWidth, newHeight, 1, + GL_RGBA, CHAN_TYPE, tempImage, &ctx->DefaultPacking); FREE(rawImage); } @@ -1357,7 +1364,7 @@ tdfxTexImage2D(GLcontext *ctx, GLenum target, GLint level, * be correct, since it would mess with "compressedSize". * Ditto for GL_RGBA[4]_S3TC, which is always mapped to DXT3. */ - if (texImage->IsCompressed) { + if (_mesa_is_format_compressed(texImage->TexFormat)) { switch (internalFormat) { case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: case GL_RGB_S3TC: @@ -1384,30 +1391,22 @@ tdfxTexImage2D(GLcontext *ctx, GLenum target, GLint level, } if (texNapalm) { texImage->InternalFormat = internalFormat = texNapalm; - texImage->IsCompressed = GL_TRUE; + _mesa_is_format_compressed(texImage->TexFormat) = GL_TRUE; } } #endif - /* choose the texture format */ - assert(ctx->Driver.ChooseTextureFormat); - texImage->TexFormat = (*ctx->Driver.ChooseTextureFormat)(ctx, - internalFormat, format, type); - assert(texImage->TexFormat); - mesaFormat = texImage->TexFormat->MesaFormat; + mesaFormat = texImage->TexFormat; mml->glideFormat = fxGlideFormat(mesaFormat); ti->info.format = mml->glideFormat; texImage->FetchTexelc = fxFetchFunction(mesaFormat); - texelBytes = texImage->TexFormat->TexelBytes; - - if (texImage->IsCompressed) { - texImage->CompressedSize = _mesa_compressed_texture_size(ctx, - mml->width, - mml->height, - 1, - mesaFormat); - dstRowStride = _mesa_compressed_row_stride(texImage->TexFormat->MesaFormat, mml->width); - texImage->Data = _mesa_alloc_texmemory(texImage->CompressedSize); + texelBytes = _mesa_get_format_bytes(texImage->TexFormat); + + if (_mesa_is_format_compressed(texImage->TexFormat)) { + GLuint compressedSize = _mesa_format_image_size(mesaFormat, mml->width, + mml->height, 1); + dstRowStride = _mesa_format_row_stride(texImage->TexFormat, mml->width); + texImage->Data = _mesa_alloc_texmemory(compressedSize); } else { dstRowStride = mml->width * texelBytes; texImage->Data = _mesa_alloc_texmemory(mml->width * mml->height * texelBytes); @@ -1437,13 +1436,13 @@ tdfxTexImage2D(GLcontext *ctx, GLenum target, GLint level, else { /* no rescaling needed */ /* unpack image, apply transfer ops and store in texImage->Data */ - texImage->TexFormat->StoreImage(ctx, 2, texImage->_BaseFormat, - texImage->TexFormat, texImage->Data, - 0, 0, 0, /* dstX/Y/Zoffset */ - dstRowStride, - texImage->ImageOffsets, - width, height, 1, - format, type, pixels, packing); + _mesa_texstore(ctx, 2, texImage->_BaseFormat, + texImage->TexFormat, texImage->Data, + 0, 0, 0, /* dstX/Y/Zoffset */ + dstRowStride, + texImage->ImageOffsets, + width, height, 1, + format, type, pixels, packing); } } @@ -1482,9 +1481,9 @@ tdfxTexSubImage2D(GLcontext *ctx, GLenum target, GLint level, assert(texImage->Data); /* must have an existing texture image! */ assert(texImage->_BaseFormat); - texelBytes = texImage->TexFormat->TexelBytes; - if (texImage->IsCompressed) { - dstRowStride = _mesa_compressed_row_stride(texImage->TexFormat->MesaFormat, mml->width); + texelBytes = _mesa_get_format_bytes(texImage->TexFormat); + if (_mesa_is_format_compressed(texImage->TexFormat)) { + dstRowStride = _mesa_format_row_stride(texImage->TexFormat, mml->width); } else { dstRowStride = mml->width * texelBytes; } @@ -1507,13 +1506,13 @@ tdfxTexSubImage2D(GLcontext *ctx, GLenum target, GLint level, } else { /* no rescaling needed */ - texImage->TexFormat->StoreImage(ctx, 2, texImage->_BaseFormat, - texImage->TexFormat, texImage->Data, - xoffset, yoffset, 0, - dstRowStride, - texImage->ImageOffsets, - width, height, 1, - format, type, pixels, packing); + _mesa_texstore(ctx, 2, texImage->_BaseFormat, + texImage->TexFormat, texImage->Data, + xoffset, yoffset, 0, + dstRowStride, + texImage->ImageOffsets, + width, height, 1, + format, type, pixels, packing); } ti->reloadImages = GL_TRUE; /* signal the image needs to be reloaded */ @@ -1572,7 +1571,8 @@ tdfxCompressedTexImage2D (GLcontext *ctx, GLenum target, tdfxContextPtr fxMesa = TDFX_CONTEXT(ctx); tdfxTexInfo *ti; tdfxMipMapLevel *mml; - GLuint mesaFormat; + gl_format mesaFormat; + GLuint compressedSize; if (TDFX_DEBUG & DEBUG_VERBOSE_DRI) { fprintf(stderr, "tdfxCompressedTexImage2D: id=%d int 0x%x %dx%d\n", @@ -1585,8 +1585,6 @@ tdfxCompressedTexImage2D (GLcontext *ctx, GLenum target, return; } - assert(texImage->IsCompressed); - ti = TDFX_TEXTURE_DATA(texObj); if (!ti) { texObj->DriverData = fxAllocTexObjData(fxMesa); @@ -1615,28 +1613,19 @@ tdfxCompressedTexImage2D (GLcontext *ctx, GLenum target, mml->height = height * mml->hScale; - /* choose the texture format */ - assert(ctx->Driver.ChooseTextureFormat); - texImage->TexFormat = (*ctx->Driver.ChooseTextureFormat)(ctx, - internalFormat, -1/*format*/, -1/*type*/); - assert(texImage->TexFormat); - /* Determine the appropriate Glide texel format, * given the user's internal texture format hint. */ - mesaFormat = texImage->TexFormat->MesaFormat; + mesaFormat = texImage->TexFormat; mml->glideFormat = fxGlideFormat(mesaFormat); ti->info.format = mml->glideFormat; texImage->FetchTexelc = fxFetchFunction(mesaFormat); /* allocate new storage for texture image, if needed */ if (!texImage->Data) { - texImage->CompressedSize = _mesa_compressed_texture_size(ctx, - mml->width, - mml->height, - 1, - mesaFormat); - texImage->Data = _mesa_alloc_texmemory(texImage->CompressedSize); + compressedSize = _mesa_format_image_size(mesaFormat, mml->width, + mml->height, 1); + texImage->Data = _mesa_alloc_texmemory(compressedSize); if (!texImage->Data) { _mesa_error(ctx, GL_OUT_OF_MEMORY, "glCompressedTexImage2D"); return; @@ -1659,10 +1648,10 @@ tdfxCompressedTexImage2D (GLcontext *ctx, GLenum target, * we replicate the data over the padded area. * For now, we take 2) + 3) but texelfetchers will be wrong! */ - const GLuint mesaFormat = texImage->TexFormat->MesaFormat; - GLuint srcRowStride = _mesa_compressed_row_stride(mesaFormat, width); + const GLuint mesaFormat = texImage->TexFormat; + GLuint srcRowStride = _mesa_format_row_stride(mesaFormat, width); - GLuint destRowStride = _mesa_compressed_row_stride(mesaFormat, + GLuint destRowStride = _mesa_format_row_stride(mesaFormat, mml->width); _mesa_upscale_teximage2d(srcRowStride, (height+3) / 4, @@ -1671,7 +1660,7 @@ tdfxCompressedTexImage2D (GLcontext *ctx, GLenum target, texImage->Data); ti->padded = GL_TRUE; } else { - MEMCPY(texImage->Data, data, texImage->CompressedSize); + MEMCPY(texImage->Data, data, compressedSize); } RevalidateTexture(ctx, texObj); @@ -1696,7 +1685,7 @@ tdfxCompressedTexSubImage2D( GLcontext *ctx, GLenum target, GLint destRowStride, srcRowStride; GLint i, rows; GLubyte *dest; - const GLuint mesaFormat = texImage->TexFormat->MesaFormat; + const GLuint mesaFormat = texImage->TexFormat; if (TDFX_DEBUG & DEBUG_VERBOSE_DRI) { fprintf(stderr, "tdfxCompressedTexSubImage2D: id=%d\n", texObj->Name); @@ -1707,9 +1696,9 @@ tdfxCompressedTexSubImage2D( GLcontext *ctx, GLenum target, mml = TDFX_TEXIMAGE_DATA(texImage); assert(mml); - srcRowStride = _mesa_compressed_row_stride(mesaFormat, width); + srcRowStride = _mesa_format_row_stride(mesaFormat, width); - destRowStride = _mesa_compressed_row_stride(mesaFormat, mml->width); + destRowStride = _mesa_format_row_stride(mesaFormat, mml->width); dest = _mesa_compressed_image_address(xoffset, yoffset, 0, mesaFormat, mml->width, @@ -1727,9 +1716,9 @@ tdfxCompressedTexSubImage2D( GLcontext *ctx, GLenum target, * see fxDDCompressedTexImage2D for caveats */ if (mml->wScale != 1 || mml->hScale != 1) { - srcRowStride = _mesa_compressed_row_stride(mesaFormat, texImage->Width); + srcRowStride = _mesa_format_row_stride(mesaFormat, texImage->Width); - destRowStride = _mesa_compressed_row_stride(mesaFormat, mml->width); + destRowStride = _mesa_format_row_stride(mesaFormat, mml->width); _mesa_upscale_teximage2d(srcRowStride, texImage->Height / 4, destRowStride, mml->height / 4, 1, texImage->Data, destRowStride, diff --git a/src/mesa/drivers/dri/tdfx/tdfx_vb.c b/src/mesa/drivers/dri/tdfx/tdfx_vb.c index 49288022328..c200ba3255a 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_vb.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_vb.c @@ -69,11 +69,11 @@ static void interp_extras( GLcontext *ctx, /*fprintf(stderr, "%s\n", __FUNCTION__);*/ - if (VB->ColorPtr[1]) { + if (VB->BackfaceColorPtr) { INTERP_4F( t, - GET_COLOR(VB->ColorPtr[1], dst), - GET_COLOR(VB->ColorPtr[1], out), - GET_COLOR(VB->ColorPtr[1], in) ); + GET_COLOR(VB->BackfaceColorPtr, dst), + GET_COLOR(VB->BackfaceColorPtr, out), + GET_COLOR(VB->BackfaceColorPtr, in) ); } if (VB->EdgeFlag) { @@ -88,9 +88,9 @@ static void copy_pv_extras( GLcontext *ctx, GLuint dst, GLuint src ) { struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - if (VB->ColorPtr[1]) { - COPY_4FV( GET_COLOR(VB->ColorPtr[1], dst), - GET_COLOR(VB->ColorPtr[1], src) ); + if (VB->BackfaceColorPtr) { + COPY_4FV( GET_COLOR(VB->BackfaceColorPtr, dst), + GET_COLOR(VB->BackfaceColorPtr, src) ); } setup_tab[TDFX_CONTEXT(ctx)->SetupIndex].copy_pv(ctx, dst, src); diff --git a/src/mesa/drivers/dri/tdfx/tdfx_vbtmp.h b/src/mesa/drivers/dri/tdfx/tdfx_vbtmp.h index 9b780761f42..19baf7d0d25 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_vbtmp.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_vbtmp.h @@ -58,32 +58,32 @@ static void TAG(emit)( GLcontext *ctx, /* fprintf(stderr, "%s\n", __FUNCTION__); */ if (IND & TDFX_TEX0_BIT) { - tc0_stride = VB->TexCoordPtr[tmu0_source]->stride; - tc0 = VB->TexCoordPtr[tmu0_source]->data; + tc0_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu0_source]->stride; + tc0 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu0_source]->data; u0scale = fxMesa->sScale0; v0scale = fxMesa->tScale0; if (IND & TDFX_PTEX_BIT) - tc0_size = VB->TexCoordPtr[tmu0_source]->size; + tc0_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu0_source]->size; } if (IND & TDFX_TEX1_BIT) { - tc1 = VB->TexCoordPtr[tmu1_source]->data; - tc1_stride = VB->TexCoordPtr[tmu1_source]->stride; + tc1 = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu1_source]->data; + tc1_stride = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu1_source]->stride; u1scale = fxMesa->sScale1; v1scale = fxMesa->tScale1; if (IND & TDFX_PTEX_BIT) - tc1_size = VB->TexCoordPtr[tmu1_source]->size; + tc1_size = VB->AttribPtr[_TNL_ATTRIB_TEX0 + tmu1_source]->size; } if (IND & TDFX_RGBA_BIT) { - col = VB->ColorPtr[0]->data; - col_stride = VB->ColorPtr[0]->stride; - col_size = VB->ColorPtr[0]->size; + col = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->data; + col_stride = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->stride; + col_size = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->size; } if (IND & TDFX_FOGC_BIT) { - fog = VB->FogCoordPtr->data; - fog_stride = VB->FogCoordPtr->stride; + fog = VB->AttribPtr[_TNL_ATTRIB_FOG]->data; + fog_stride = VB->AttribPtr[_TNL_ATTRIB_FOG]->stride; } { @@ -168,14 +168,14 @@ static GLboolean TAG(check_tex_sizes)( GLcontext *ctx ) struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; if (IND & TDFX_TEX1_BIT) { - if (VB->TexCoordPtr[0] == 0) - VB->TexCoordPtr[0] = VB->TexCoordPtr[1]; + if (VB->AttribPtr[_TNL_ATTRIB_TEX0] == 0) + VB->AttribPtr[_TNL_ATTRIB_TEX0] = VB->AttribPtr[_TNL_ATTRIB_TEX1]; - if (VB->TexCoordPtr[1]->size == 4) + if (VB->AttribPtr[_TNL_ATTRIB_TEX1]->size == 4) return GL_FALSE; } - if (VB->TexCoordPtr[0]->size == 4) + if (VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 4) return GL_FALSE; } diff --git a/src/mesa/drivers/dri/trident/Makefile b/src/mesa/drivers/dri/trident/Makefile deleted file mode 100644 index 9ee24c504c1..00000000000 --- a/src/mesa/drivers/dri/trident/Makefile +++ /dev/null @@ -1,26 +0,0 @@ -# src/mesa/drivers/dri/trident/Makefile - -TOP = ../../../../.. -include $(TOP)/configs/current - -LIBNAME = trident_dri.so - -# Not yet -# MINIGLX_SOURCES = server/trident_dri.c - -DRIVER_SOURCES = \ - trident_context.c \ - trident_state.c \ - trident_vb.c \ - trident_dd.c \ - trident_tris.c - -C_SOURCES = \ - $(COMMON_SOURCES) \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - -include ../Makefile.template - -symlinks: diff --git a/src/mesa/drivers/dri/trident/trident_context.c b/src/mesa/drivers/dri/trident/trident_context.c deleted file mode 100644 index b5126b07ea6..00000000000 --- a/src/mesa/drivers/dri/trident/trident_context.c +++ /dev/null @@ -1,465 +0,0 @@ -/* - * Copyright 2002 by Alan Hourihane, Sychdyn, North Wales, UK. - * - * Permission to use, copy, modify, distribute, and sell this software and its - * documentation for any purpose is hereby granted without fee, provided that - * the above copyright notice appear in all copies and that both that - * copyright notice and this permission notice appear in supporting - * documentation, and that the name of Alan Hourihane not be used in - * advertising or publicity pertaining to distribution of the software without - * specific, written prior permission. Alan Hourihane makes no representations - * about the suitability of this software for any purpose. It is provided - * "as is" without express or implied warranty. - * - * ALAN HOURIHANE DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, - * INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO - * EVENT SHALL ALAN HOURIHANE BE LIABLE FOR ANY SPECIAL, INDIRECT OR - * CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, - * DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER - * TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR - * PERFORMANCE OF THIS SOFTWARE. - * - * Authors: Alan Hourihane, <[email protected]> - * - * Trident CyberBladeXP driver. - * - */ -#include "trident_dri.h" -#include "trident_context.h" -#include "trident_lock.h" - -#include "swrast/swrast.h" -#include "swrast_setup/swrast_setup.h" -#include "vbo/vbo.h" - -#include "tnl/tnl.h" -#include "tnl/t_pipeline.h" - -#include "main/context.h" -#include "main/simple_list.h" -#include "main/matrix.h" -#include "main/extensions.h" -#include "main/framebuffer.h" -#include "main/renderbuffer.h" -#include "main/viewport.h" -#if defined(USE_X86_ASM) -#include "x86/common_x86_asm.h" -#endif -#include "main/simple_list.h" -#include "main/mm.h" -#include "drirenderbuffer.h" - -#include "drivers/common/driverfuncs.h" -#include "dri_util.h" -#include "utils.h" - -static const struct tnl_pipeline_stage *trident_pipeline[] = { - &_tnl_vertex_transform_stage, - &_tnl_normal_transform_stage, - &_tnl_lighting_stage, - &_tnl_texgen_stage, - &_tnl_texture_transform_stage, - &_tnl_render_stage, - 0, -}; - - -static GLboolean -tridentCreateContext( const __GLcontextModes *glVisual, - __DRIcontextPrivate *driContextPriv, - void *sharedContextPrivate) -{ - GLcontext *ctx, *shareCtx; - __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; - tridentContextPtr tmesa; - tridentScreenPtr tridentscrn; - struct dd_function_table functions; -#if 0 - drm_trident_sarea_t *saPriv=(drm_trident_sarea_t *)(((char*)sPriv->pSAREA)+ - sizeof(XF86DRISAREARec)); -#endif - - tmesa = (tridentContextPtr) CALLOC( sizeof(*tmesa) ); - if ( !tmesa ) return GL_FALSE; - - /* Allocate the Mesa context */ - if (sharedContextPrivate) - shareCtx = ((tridentContextPtr) sharedContextPrivate)->glCtx; - else - shareCtx = NULL; - - _mesa_init_driver_functions(&functions); - - tmesa->glCtx = - _mesa_create_context(glVisual, shareCtx, &functions, (void *)tmesa); - - if (!tmesa->glCtx) { - FREE(tmesa); - return GL_FALSE; - } - - tmesa->driContext = driContextPriv; - tmesa->driScreen = sPriv; - tmesa->driDrawable = NULL; /* Set by XMesaMakeCurrent */ - - tmesa->hHWContext = driContextPriv->hHWContext; - tmesa->driHwLock = (drmLock *)&sPriv->pSAREA->lock; - tmesa->driFd = sPriv->fd; -#if 0 - tmesa->sarea = saPriv; -#endif - - tridentscrn = tmesa->tridentScreen = (tridentScreenPtr)(sPriv->private); - - ctx = tmesa->glCtx; - - ctx->Const.MaxTextureLevels = 13; /* 4K by 4K? Is that right? */ - ctx->Const.MaxTextureUnits = 1; /* Permedia 3 */ - - ctx->Const.MinLineWidth = 0.0; - ctx->Const.MaxLineWidth = 255.0; - - ctx->Const.MinLineWidthAA = 0.0; - ctx->Const.MaxLineWidthAA = 65536.0; - - ctx->Const.MinPointSize = 0.0; - ctx->Const.MaxPointSize = 255.0; - - ctx->Const.MinPointSizeAA = 0.5; /* 4x4 quality mode */ - ctx->Const.MaxPointSizeAA = 16.0; - ctx->Const.PointSizeGranularity = 0.25; - - ctx->Const.MaxDrawBuffers = 1; - -#if 0 - tmesa->texHeap = mmInit( 0, tmesa->tridentScreen->textureSize ); - - make_empty_list(&tmesa->TexObjList); - make_empty_list(&tmesa->SwappedOut); - - tmesa->CurrentTexObj[0] = 0; - tmesa->CurrentTexObj[1] = 0; /* Permedia 3, second texture */ - - tmesa->RenderIndex = ~0; -#endif - - /* Initialize the software rasterizer and helper modules. - */ - _swrast_CreateContext( ctx ); - _vbo_CreateContext( ctx ); - _tnl_CreateContext( ctx ); - _swsetup_CreateContext( ctx ); - - /* Install the customized pipeline: - */ - _tnl_destroy_pipeline( ctx ); - _tnl_install_pipeline( ctx, trident_pipeline ); - - /* Configure swrast to match hardware characteristics: - */ - _swrast_allow_pixel_fog( ctx, GL_FALSE ); - _swrast_allow_vertex_fog( ctx, GL_TRUE ); - - tridentInitVB( ctx ); - tridentDDInitExtensions( ctx ); - tridentDDInitDriverFuncs( ctx ); - tridentDDInitStateFuncs( ctx ); -#if 0 - tridentDDInitSpanFuncs( ctx ); - tridentDDInitTextureFuncs( ctx ); -#endif - tridentDDInitTriFuncs( ctx ); - tridentDDInitState( tmesa ); - - driContextPriv->driverPrivate = (void *)tmesa; - - UNLOCK_HARDWARE(tmesa); - - return GL_TRUE; -} - -static void -tridentDestroyContext(__DRIcontextPrivate *driContextPriv) -{ - tridentContextPtr tmesa = (tridentContextPtr)driContextPriv->driverPrivate; - - if (tmesa) { - _swsetup_DestroyContext( tmesa->glCtx ); - _tnl_DestroyContext( tmesa->glCtx ); - _vbo_DestroyContext( tmesa->glCtx ); - _swrast_DestroyContext( tmesa->glCtx ); - - /* free the Mesa context */ - tmesa->glCtx->DriverCtx = NULL; - _mesa_destroy_context(tmesa->glCtx); - - _mesa_free(tmesa); - driContextPriv->driverPrivate = NULL; - } -} - - -static GLboolean -tridentCreateBuffer( __DRIscreenPrivate *driScrnPriv, - __DRIdrawablePrivate *driDrawPriv, - const __GLcontextModes *mesaVis, - GLboolean isPixmap ) -{ - tridentScreenPtr screen = (tridentScreenPtr) driScrnPriv->private; - - if (isPixmap) { - return GL_FALSE; /* not implemented */ - } - else { - struct gl_framebuffer *fb = _mesa_create_framebuffer(mesaVis); - - { - driRenderbuffer *frontRb - = driNewRenderbuffer(GL_RGBA, NULL, screen->cpp, - screen->frontOffset, screen->frontPitch, - driDrawPriv); - /* - tridentSetSpanFunctions(frontRb, mesaVis); - */ - _mesa_add_renderbuffer(fb, BUFFER_FRONT_LEFT, &frontRb->Base); - } - - if (mesaVis->doubleBufferMode) { - driRenderbuffer *backRb - = driNewRenderbuffer(GL_RGBA, NULL, screen->cpp, - screen->backOffset, screen->backPitch, - driDrawPriv); - /* - tridentSetSpanFunctions(backRb, mesaVis); - */ - _mesa_add_renderbuffer(fb, BUFFER_BACK_LEFT, &backRb->Base); - } - - if (mesaVis->depthBits == 16) { - driRenderbuffer *depthRb - = driNewRenderbuffer(GL_DEPTH_COMPONENT16, NULL, screen->cpp, - screen->depthOffset, screen->depthPitch, - driDrawPriv); - /* - tridentSetSpanFunctions(depthRb, mesaVis); - */ - _mesa_add_renderbuffer(fb, BUFFER_DEPTH, &depthRb->Base); - } - else if (mesaVis->depthBits == 24) { - driRenderbuffer *depthRb - = driNewRenderbuffer(GL_DEPTH_COMPONENT24, NULL, screen->cpp, - screen->depthOffset, screen->depthPitch, - driDrawPriv); - /* - tridentSetSpanFunctions(depthRb, mesaVis); - */ - _mesa_add_renderbuffer(fb, BUFFER_DEPTH, &depthRb->Base); - } - - /* no h/w stencil? - if (mesaVis->stencilBits > 0 && !swStencil) { - driRenderbuffer *stencilRb - = driNewRenderbuffer(GL_STENCIL_INDEX8_EXT); - tridentSetSpanFunctions(stencilRb, mesaVis); - _mesa_add_renderbuffer(fb, BUFFER_STENCIL, &stencilRb->Base); - } - */ - - _mesa_add_soft_renderbuffers(fb, - GL_FALSE, /* color */ - GL_FALSE, /* depth */ - mesaVis->stencilBits > 0, - mesaVis->accumRedBits > 0, - GL_FALSE, /* alpha */ - GL_FALSE /* aux */); - driDrawPriv->driverPrivate = (void *) fb; - - return (driDrawPriv->driverPrivate != NULL); - } -} - - -static void -tridentDestroyBuffer(__DRIdrawablePrivate *driDrawPriv) -{ - _mesa_reference_framebuffer((GLframebuffer **)(&(driDrawPriv->driverPrivate)), NULL); -} - -static void -tridentSwapBuffers(__DRIdrawablePrivate *drawablePrivate) -{ - __DRIdrawablePrivate *dPriv = (__DRIdrawablePrivate *) drawablePrivate; - - if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { - tridentContextPtr tmesa; - GLcontext *ctx; - tmesa = (tridentContextPtr) dPriv->driContextPriv->driverPrivate; - ctx = tmesa->glCtx; - if (ctx->Visual.doubleBufferMode) { - _mesa_notifySwapBuffers( ctx ); /* flush pending rendering comands */ - tridentCopyBuffer( dPriv ); - } - } - else { - /* XXX this shouldn't be an error but we can't handle it for now */ - _mesa_problem(NULL, "tridentSwapBuffers: drawable has no context!\n"); - } -} - -static GLboolean -tridentMakeCurrent(__DRIcontextPrivate *driContextPriv, - __DRIdrawablePrivate *driDrawPriv, - __DRIdrawablePrivate *driReadPriv) -{ - if (driContextPriv) { - GET_CURRENT_CONTEXT(ctx); - tridentContextPtr oldCtx = ctx ? TRIDENT_CONTEXT(ctx) : NULL; - tridentContextPtr newCtx = (tridentContextPtr) driContextPriv->driverPrivate; - - if ( newCtx != oldCtx ) { - newCtx->dirty = ~0; - } - - if (newCtx->driDrawable != driDrawPriv) { - newCtx->driDrawable = driDrawPriv; -#if 0 - tridentUpdateWindow ( newCtx->glCtx ); - tridentUpdateViewportOffset( newCtx->glCtx ); -#endif - } - - newCtx->drawOffset = newCtx->tridentScreen->backOffset; - newCtx->drawPitch = newCtx->tridentScreen->backPitch; - - _mesa_make_current( newCtx->glCtx, - (GLframebuffer *) driDrawPriv->driverPrivate, - (GLframebuffer *) driReadPriv->driverPrivate ); - - if (!newCtx->glCtx->Viewport.Width) { - _mesa_set_viewport(newCtx->glCtx, 0, 0, - driDrawPriv->w, driDrawPriv->h); - } - } else { - _mesa_make_current( NULL, NULL, NULL ); - } - return GL_TRUE; -} - - -static GLboolean -tridentUnbindContext( __DRIcontextPrivate *driContextPriv ) -{ - return GL_TRUE; -} - - -static tridentScreenPtr -tridentCreateScreen( __DRIscreenPrivate *sPriv ) -{ - TRIDENTDRIPtr tDRIPriv = (TRIDENTDRIPtr)sPriv->pDevPriv; - tridentScreenPtr tridentScreen; - - if (sPriv->devPrivSize != sizeof(TRIDENTDRIRec)) { - fprintf(stderr,"\nERROR! sizeof(TRIDENTDRIRec) does not match passed size from device driver\n"); - return GL_FALSE; - } - - /* Allocate the private area */ - tridentScreen = (tridentScreenPtr) CALLOC( sizeof(*tridentScreen) ); - if ( !tridentScreen ) return NULL; - - tridentScreen->driScreen = sPriv; - - tridentScreen->frontOffset = tDRIPriv->frontOffset; - tridentScreen->backOffset = tDRIPriv->backOffset; - tridentScreen->depthOffset = tDRIPriv->depthOffset; - tridentScreen->frontPitch = tDRIPriv->frontPitch; - tridentScreen->backPitch = tDRIPriv->backPitch; - tridentScreen->depthPitch = tDRIPriv->depthPitch; - tridentScreen->width = tDRIPriv->width; - tridentScreen->height = tDRIPriv->height; - -printf("%d %d\n",tridentScreen->width,tridentScreen->height); -printf("%d %d\n",tridentScreen->frontPitch,tridentScreen->backPitch); -printf("offset 0x%x 0x%x\n",tridentScreen->backOffset,tridentScreen->depthOffset); - - tridentScreen->mmio.handle = tDRIPriv->regs; - tridentScreen->mmio.size = 0x20000; - - if (drmMap(sPriv->fd, - tridentScreen->mmio.handle, tridentScreen->mmio.size, - (drmAddressPtr)&tridentScreen->mmio.map)) { - FREE(tridentScreen); - return GL_FALSE; - } -printf("MAPPED at %p\n", tridentScreen->mmio.map); - - return tridentScreen; -} - -/* Destroy the device specific screen private data struct. - */ -static void -tridentDestroyScreen( __DRIscreenPrivate *sPriv ) -{ - tridentScreenPtr tridentScreen = (tridentScreenPtr)sPriv->private; - - FREE(tridentScreen); -} - -static GLboolean -tridentInitDriver(__DRIscreenPrivate *sPriv) -{ - sPriv->private = (void *) tridentCreateScreen( sPriv ); - - if (!sPriv->private) { - tridentDestroyScreen( sPriv ); - return GL_FALSE; - } - - return GL_TRUE; -} - -/** - * This is the driver specific part of the createNewScreen entry point. - * - * \todo maybe fold this into intelInitDriver - * - * \return the __GLcontextModes supported by this driver - */ -const __DRIconfig **tridentInitScreen(__DRIscreenPrivate *psp) -{ - static const __DRIversion ddx_expected = { 4, 0, 0 }; - static const __DRIversion dri_expected = { 3, 1, 0 }; - static const __DRIversion drm_expected = { 1, 0, 0 }; - - if ( ! driCheckDriDdxDrmVersions2( "Trident", - &psp->dri_version, & dri_expected, - &psp->ddx_version, & ddx_expected, - &psp->drm_version, & drm_expected ) ) - return NULL; - - if (!tridentInitDriver(psp)) - return NULL; - - /* Wait... what? This driver doesn't report any modes... */ -#if 0 - TRIDENTDRIPtr dri_priv = (TRIDENTDRIPtr) psp->pDevPriv; - *driver_modes = tridentFillInModes( dri_priv->bytesPerPixel * 8, - GL_TRUE ); -#endif - - return NULL; -} - -const struct __DriverAPIRec driDriverAPI = { - tridentInitScreen, - tridentDestroyScreen, - tridentCreateContext, - tridentDestroyContext, - tridentCreateBuffer, - tridentDestroyBuffer, - tridentSwapBuffers, - tridentMakeCurrent, - tridentUnbindContext, -}; diff --git a/src/mesa/drivers/dri/trident/trident_context.h b/src/mesa/drivers/dri/trident/trident_context.h deleted file mode 100644 index fbbb4a96e78..00000000000 --- a/src/mesa/drivers/dri/trident/trident_context.h +++ /dev/null @@ -1,242 +0,0 @@ -/* - * Copyright 2002 by Alan Hourihane, Sychdyn, North Wales, UK. - * - * Permission to use, copy, modify, distribute, and sell this software and its - * documentation for any purpose is hereby granted without fee, provided that - * the above copyright notice appear in all copies and that both that - * copyright notice and this permission notice appear in supporting - * documentation, and that the name of Alan Hourihane not be used in - * advertising or publicity pertaining to distribution of the software without - * specific, written prior permission. Alan Hourihane makes no representations - * about the suitability of this software for any purpose. It is provided - * "as is" without express or implied warranty. - * - * ALAN HOURIHANE DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, - * INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO - * EVENT SHALL ALAN HOURIHANE BE LIABLE FOR ANY SPECIAL, INDIRECT OR - * CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, - * DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER - * TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR - * PERFORMANCE OF THIS SOFTWARE. - * - * Authors: Alan Hourihane, <[email protected]> - * - * Trident CyberBladeXP driver. - * - */ -#ifndef _TRIDENT_CONTEXT_H_ -#define _TRIDENT_CONTEXT_H_ - -#include "dri_util.h" -#include "main/macros.h" -#include "main/mtypes.h" -#include "drm.h" -#include "main/mm.h" - -#define SUBPIXEL_X (0.0F) -#define SUBPIXEL_Y (0.125F) - -#define _TRIDENT_NEW_VERTEX (_NEW_TEXTURE | \ - _DD_NEW_TRI_UNFILLED | \ - _DD_NEW_TRI_LIGHT_TWOSIDE) - -#define TRIDENT_FALLBACK_TEXTURE 0x01 -#define TRIDENT_FALLBACK_DRAW_BUFFER 0x02 - -#define TRIDENT_NEW_CLIP 0x01 - -#define TRIDENT_UPLOAD_COMMAND_D 0x00000001 -#define TRIDENT_UPLOAD_CONTEXT 0x04000000 -#define TRIDENT_UPLOAD_CLIPRECTS 0x80000000 - -#define TAG(x) trident##x -#include "tnl_dd/t_dd_vertex.h" -#undef TAG - -/* these require that base be dword-aligned */ -static INLINE void MMIO_OUT32(unsigned char *base, unsigned int offset, - unsigned int val) -{ - unsigned int *addr = (unsigned int *)(base + offset); - *addr = val; -} - -static INLINE unsigned int MMIO_IN32(unsigned char *base, unsigned int offset) -{ - unsigned int *addr = (unsigned int *)(base + offset); - return *addr; -} - -#define MMIO_OUT8(base, offset, val) *((base) + (offset)) = (val) -#define MMIO_IN8(base, offset) *((base) + (offset)) - -struct trident_context; -typedef struct trident_context tridentContextRec; -typedef struct trident_context *tridentContextPtr; - -typedef void (*trident_quad_func)( tridentContextPtr, - const tridentVertex *, - const tridentVertex *, - const tridentVertex *, - const tridentVertex * ); -typedef void (*trident_tri_func)( tridentContextPtr, - const tridentVertex *, - const tridentVertex *, - const tridentVertex * ); -typedef void (*trident_line_func)( tridentContextPtr, - const tridentVertex *, - const tridentVertex * ); -typedef void (*trident_point_func)( tridentContextPtr, - const tridentVertex * ); - -typedef struct { - drm_handle_t handle; /* Handle to the DRM region */ - drmSize size; /* Size of the DRM region */ - unsigned char *map; /* Mapping of the DRM region */ -} tridentRegionRec, *tridentRegionPtr; - -typedef struct { - __DRIscreenPrivate *driScreen; /* Back pointer to DRI screen */ - - drmBufMapPtr buffers; - - unsigned int frontOffset; - unsigned int frontPitch; - unsigned int backOffset; - unsigned int backPitch; - unsigned int depthOffset; - unsigned int depthPitch; - unsigned int width; - unsigned int height; - unsigned int cpp; - -#if 0 - unsigned int sarea_priv_offset; -#endif - - tridentRegionRec mmio; -} tridentScreenRec, *tridentScreenPtr; - -/** - * tridentRenderbuffer, derived from Mesa's gl_renderbuffer - */ -typedef struct { - struct gl_renderbuffer Base; - /* XXX per-window info should go here */ - int foo, bar; -} tridentRenderbuffer; - - -struct trident_context { - GLcontext *glCtx; /* Mesa context */ - - __DRIcontextPrivate *driContext; - __DRIscreenPrivate *driScreen; - __DRIdrawablePrivate *driDrawable; - - GLuint new_gl_state; - GLuint new_state; - GLuint dirty; - -#if 0 - drm_trident_sarea_t *sarea; -#endif - - /* Temporaries for translating away float colors: - */ - struct gl_client_array UbyteColor; - struct gl_client_array UbyteSecondaryColor; - - /* Mirrors of some DRI state - */ - int lastStamp; /* mirror driDrawable->lastStamp */ - - drm_context_t hHWContext; - drmLock *driHwLock; - int driFd; - - tridentScreenPtr tridentScreen; /* Screen private DRI data */ - - /* Visual, drawable, cliprect and scissor information - */ - GLenum DrawBuffer; - GLint drawOffset, drawPitch; - GLint drawX, drawY; /* origin of drawable in draw buffer */ - GLint readOffset, readPitch; - - GLuint numClipRects; /* Cliprects for the draw buffer */ - drm_clip_rect_t *pClipRects; - - GLint scissor; - drm_clip_rect_t ScissorRect; /* Current software scissor */ - - GLuint Fallback; - GLuint RenderIndex; - GLuint SetupNewInputs; - GLuint SetupIndex; - GLfloat hw_viewport[16]; - GLfloat depth_scale; - GLuint vertex_format; - GLuint vertex_size; - GLuint vertex_stride_shift; - GLubyte *verts; - - GLint tmu_source[2]; - - GLuint hw_primitive; - GLenum render_primitive; - - trident_point_func draw_point; - trident_line_func draw_line; - trident_tri_func draw_tri; - trident_quad_func draw_quad; - -#if 0 - gammaTextureObjectPtr CurrentTexObj[2]; - struct gamma_texture_object_t TexObjList; - struct gamma_texture_object_t SwappedOut; - GLenum TexEnvImageFmt[2]; - - struct mem_block *texHeap; - - int lastSwap; - int texAge; - int ctxAge; - int dirtyAge; - int lastStamp; -#endif - - /* Chip state */ - - int commandD; - - /* Context State */ - - int ClearColor; -}; - -void tridentDDInitExtensions( GLcontext *ctx ); -void tridentDDInitDriverFuncs( GLcontext *ctx ); -void tridentDDInitSpanFuncs( GLcontext *ctx ); -void tridentDDInitState( tridentContextPtr tmesa ); -void tridentInitHW( tridentContextPtr tmesa ); -void tridentDDInitStateFuncs( GLcontext *ctx ); -void tridentDDInitTextureFuncs( GLcontext *ctx ); -void tridentDDInitTriFuncs( GLcontext *ctx ); - -extern void tridentBuildVertices( GLcontext *ctx, - GLuint start, - GLuint count, - GLuint newinputs ); -extern void tridentInitVB( GLcontext *ctx ); -extern void tridentCopyBuffer( const __DRIdrawablePrivate *dPriv ); -extern void tridentFallback( tridentContextPtr tmesa, GLuint bit, - GLboolean mode ); -extern void tridentCheckTexSizes( GLcontext *ctx ); -extern void tridentChooseVertexState( GLcontext *ctx ); -extern void tridentDDUpdateHWState( GLcontext *ctx ); -extern void tridentUploadHwStateLocked( tridentContextPtr tmesa ); - -#define TRIDENT_CONTEXT(ctx) ((tridentContextPtr)(ctx->DriverCtx)) - -#endif /* _TRIDENT_CONTEXT_H_ */ diff --git a/src/mesa/drivers/dri/trident/trident_dd.c b/src/mesa/drivers/dri/trident/trident_dd.c deleted file mode 100644 index faa40c36a26..00000000000 --- a/src/mesa/drivers/dri/trident/trident_dd.c +++ /dev/null @@ -1,110 +0,0 @@ -/* - * Copyright 2002 by Alan Hourihane, Sychdyn, North Wales, UK. - * - * Permission to use, copy, modify, distribute, and sell this software and its - * documentation for any purpose is hereby granted without fee, provided that - * the above copyright notice appear in all copies and that both that - * copyright notice and this permission notice appear in supporting - * documentation, and that the name of Alan Hourihane not be used in - * advertising or publicity pertaining to distribution of the software without - * specific, written prior permission. Alan Hourihane makes no representations - * about the suitability of this software for any purpose. It is provided - * "as is" without express or implied warranty. - * - * ALAN HOURIHANE DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, - * INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO - * EVENT SHALL ALAN HOURIHANE BE LIABLE FOR ANY SPECIAL, INDIRECT OR - * CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, - * DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER - * TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR - * PERFORMANCE OF THIS SOFTWARE. - * - * Authors: Alan Hourihane, <[email protected]> - * - * Trident CyberBladeXP driver. - * - */ -#include "trident_context.h" -#include "trident_lock.h" -#if defined(USE_X86_ASM) -#include "x86/common_x86_asm.h" -#endif - -#include "swrast/swrast.h" -#include "main/context.h" -#include "main/framebuffer.h" - -#define TRIDENT_DATE "20041223" - -/* Return the width and height of the current color buffer. - */ -static void tridentDDGetBufferSize( GLframebuffer *framebuffer, - GLuint *width, GLuint *height ) -{ - GET_CURRENT_CONTEXT(ctx); - tridentContextPtr tmesa = TRIDENT_CONTEXT(ctx); - - LOCK_HARDWARE(tmesa); - *width = tmesa->driDrawable->w; - *height = tmesa->driDrawable->h; - UNLOCK_HARDWARE(tmesa); -} - - -/* Return various strings for glGetString(). - */ -static const GLubyte *tridentDDGetString( GLcontext *ctx, GLenum name ) -{ - static char buffer[128]; - - switch ( name ) { - case GL_VENDOR: - return (GLubyte *)"Alan Hourihane"; - - case GL_RENDERER: - sprintf( buffer, "Mesa DRI Trident " TRIDENT_DATE ); - - /* Append any CPU-specific information. - */ -#ifdef USE_X86_ASM - if ( _mesa_x86_cpu_features ) { - strncat( buffer, " x86", 4 ); - } -#ifdef USE_MMX_ASM - if ( cpu_has_mmx ) { - strncat( buffer, "/MMX", 4 ); - } -#endif -#ifdef USE_3DNOW_ASM - if ( cpu_has_3dnow ) { - strncat( buffer, "/3DNow!", 7 ); - } -#endif -#ifdef USE_SSE_ASM - if ( cpu_has_xmm ) { - strncat( buffer, "/SSE", 4 ); - } -#endif -#endif - return (GLubyte *)buffer; - - default: - return NULL; - } -} - -/* Enable the extensions supported by this driver. - */ -void tridentDDInitExtensions( GLcontext *ctx ) -{ - /* None... */ -} - -/* Initialize the driver's misc functions. - */ -void tridentDDInitDriverFuncs( GLcontext *ctx ) -{ - ctx->Driver.GetBufferSize = tridentDDGetBufferSize; - ctx->Driver.GetString = tridentDDGetString; - ctx->Driver.Error = NULL; -} diff --git a/src/mesa/drivers/dri/trident/trident_dri.h b/src/mesa/drivers/dri/trident/trident_dri.h deleted file mode 100644 index c1ce3c46827..00000000000 --- a/src/mesa/drivers/dri/trident/trident_dri.h +++ /dev/null @@ -1,28 +0,0 @@ -#ifndef _TRIDENT_DRI_ -#define _TRIDENT_DRI_ - -#include "xf86drm.h" - -typedef struct { - drm_handle_t regs; - drmSize regsSize; - drmAddress regsMap; - int deviceID; - int width; - int height; - int mem; - int frontOffset; - int frontPitch; - int backOffset; - int backPitch; - int depthOffset; - int depthPitch; - int cpp; -#if 0 - int textureOffset; - int textureSize; -#endif - unsigned int sarea_priv_offset; -} TRIDENTDRIRec, *TRIDENTDRIPtr; - -#endif diff --git a/src/mesa/drivers/dri/trident/trident_lock.h b/src/mesa/drivers/dri/trident/trident_lock.h deleted file mode 100644 index ee0819f5caa..00000000000 --- a/src/mesa/drivers/dri/trident/trident_lock.h +++ /dev/null @@ -1,11 +0,0 @@ -/* XXX tridentGetLock doesn't exist... */ - -#define LOCK_HARDWARE(tmesa) \ - do { \ - char __ret = 0; \ - DRM_CAS(tmesa->driHwLock, tmesa->hHWContext, \ - DRM_LOCK_HELD | tmesa->hHWContext, __ret); \ - } while (0) - -#define UNLOCK_HARDWARE(tmesa) \ - DRM_UNLOCK(tmesa->driFd, tmesa->driHwLock, tmesa->hHWContext) diff --git a/src/mesa/drivers/dri/trident/trident_state.c b/src/mesa/drivers/dri/trident/trident_state.c deleted file mode 100644 index e68d3a73c64..00000000000 --- a/src/mesa/drivers/dri/trident/trident_state.c +++ /dev/null @@ -1,548 +0,0 @@ -/* - * Copyright 2002 by Alan Hourihane, Sychdyn, North Wales, UK. - * - * Permission to use, copy, modify, distribute, and sell this software and its - * documentation for any purpose is hereby granted without fee, provided that - * the above copyright notice appear in all copies and that both that - * copyright notice and this permission notice appear in supporting - * documentation, and that the name of Alan Hourihane not be used in - * advertising or publicity pertaining to distribution of the software without - * specific, written prior permission. Alan Hourihane makes no representations - * about the suitability of this software for any purpose. It is provided - * "as is" without express or implied warranty. - * - * ALAN HOURIHANE DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, - * INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO - * EVENT SHALL ALAN HOURIHANE BE LIABLE FOR ANY SPECIAL, INDIRECT OR - * CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, - * DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER - * TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR - * PERFORMANCE OF THIS SOFTWARE. - * - * Authors: Alan Hourihane, <[email protected]> - * - * Trident CyberBladeXP driver. - * - */ -#include "trident_context.h" -#include "trident_lock.h" -#include "vbo/vbo.h" -#include "swrast/swrast.h" -#include "swrast_setup/swrast_setup.h" -#include "tnl/tnl.h" -#include "main/framebuffer.h" - -#define TRIDENTPACKCOLOR332(r, g, b) \ - (((r) & 0xe0) | (((g) & 0xe0) >> 3) | (((b) & 0xc0) >> 6)) - -#define TRIDENTPACKCOLOR1555(r, g, b, a) \ - ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) | \ - ((a) ? 0x8000 : 0)) - -#define TRIDENTPACKCOLOR565(r, g, b) \ - ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3)) - -#define TRIDENTPACKCOLOR888(r, g, b) \ - (((r) << 16) | ((g) << 8) | (b)) - -#define TRIDENTPACKCOLOR8888(r, g, b, a) \ - (((a) << 24) | ((r) << 16) | ((g) << 8) | (b)) - -#define TRIDENTPACKCOLOR4444(r, g, b, a) \ - ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4)) - -static INLINE GLuint tridentPackColor( GLuint cpp, - GLubyte r, GLubyte g, - GLubyte b, GLubyte a ) -{ - switch ( cpp ) { - case 2: - return TRIDENTPACKCOLOR565( r, g, b ); - case 4: - return TRIDENTPACKCOLOR8888( r, g, b, a ); - default: - return 0; - } -} - -void tridentUploadHwStateLocked( tridentContextPtr tmesa ) -{ - unsigned char *MMIO = tmesa->tridentScreen->mmio.map; -#if 0 - ATISAREAPrivPtr sarea = tmesa->sarea; - trident_context_regs_t *regs = &(sarea->ContextState); -#endif - - if ( tmesa->dirty & TRIDENT_UPLOAD_COMMAND_D ) { - MMIO_OUT32(MMIO, 0x00281C, tmesa->commandD ); - tmesa->dirty &= ~TRIDENT_UPLOAD_COMMAND_D; - } - - if ( tmesa->dirty & TRIDENT_UPLOAD_CLIPRECTS ) { - /* XXX FIX ME ! */ - MMIO_OUT32(MMIO, 0x002C80 , 0x20008000 | tmesa->tridentScreen->height ); - MMIO_OUT32(MMIO, 0x002C84 , 0x20000000 | tmesa->tridentScreen->width ); - tmesa->dirty &= ~TRIDENT_UPLOAD_CLIPRECTS; - } - - tmesa->dirty = 0; -} - -/* Copy the back color buffer to the front color buffer. - */ -void tridentCopyBuffer( const __DRIdrawablePrivate *dPriv ) -{ - unsigned char *MMIO; - tridentContextPtr tmesa; - GLint nbox, i; - int busy; - drm_clip_rect_t *pbox; - - assert(dPriv); - assert(dPriv->driContextPriv); - assert(dPriv->driContextPriv->driverPrivate); - - tmesa = (tridentContextPtr) dPriv->driContextPriv->driverPrivate; - MMIO = tmesa->tridentScreen->mmio.map; - - LOCK_HARDWARE( tmesa ); - - /* use front buffer cliprects */ - nbox = dPriv->numClipRects; - pbox = dPriv->pClipRects; - - for ( i = 0 ; i < nbox ; i++ ) { -#if 0 - GLint nr = MIN2( i + MACH64_NR_SAREA_CLIPRECTS , nbox ); - drm_clip_rect_t *b = tmesa->sarea->boxes; - GLint n = 0; - - for ( ; i < nr ; i++ ) { - *b++ = pbox[i]; - n++; - } - tmesa->sarea->nbox = n; -#endif - - MMIO_OUT32(MMIO, 0x2150, tmesa->tridentScreen->frontPitch << 20 | tmesa->tridentScreen->frontOffset>>4); - MMIO_OUT32(MMIO, 0x2154, tmesa->tridentScreen->backPitch << 20 | tmesa->tridentScreen->backOffset>>4); - MMIO_OUT8(MMIO, 0x2127, 0xCC); /* Copy Rop */ - MMIO_OUT32(MMIO, 0x2128, 0x4); /* scr2scr */ - MMIO_OUT32(MMIO, 0x2138, (pbox->x1 << 16) | pbox->y1); - MMIO_OUT32(MMIO, 0x213C, (pbox->x1 << 16) | pbox->y1); - MMIO_OUT32(MMIO, 0x2140, (pbox->x2 - pbox->x1) << 16 | (pbox->y2 - pbox->y1) ); - MMIO_OUT8(MMIO, 0x2124, 0x01); /* BLT */ -#define GE_BUSY 0x80 - for (;;) { - busy = MMIO_IN8(MMIO, 0x2120); - if ( !(busy & GE_BUSY) ) - break; - } - } - - UNLOCK_HARDWARE( tmesa ); - -#if 0 - tmesa->dirty |= (MACH64_UPLOAD_CONTEXT | - MACH64_UPLOAD_MISC | - MACH64_UPLOAD_CLIPRECTS); -#endif -} - - -static void tridentDDClear( GLcontext *ctx, GLbitfield mask ) -{ - tridentContextPtr tmesa = TRIDENT_CONTEXT(ctx); - unsigned char *MMIO = tmesa->tridentScreen->mmio.map; - int busy; - GLuint flags = 0; - GLint i; - GLint cx, cy, cw, ch; - -#define DRM_TRIDENT_FRONT 0x01 -#define DRM_TRIDENT_BACK 0x02 -#define DRM_TRIDENT_DEPTH 0x04 - - if ( tmesa->new_state ) - tridentDDUpdateHWState( ctx ); - - if ( mask & BUFFER_BIT_FRONT_LEFT ) { - flags |= DRM_TRIDENT_FRONT; - mask &= ~BUFFER_BIT_FRONT_LEFT; - } - - if ( mask & BUFFER_BIT_BACK_LEFT ) { - flags |= DRM_TRIDENT_BACK; - mask &= ~BUFFER_BIT_BACK_LEFT; - } - - if ( ( mask & BUFFER_BIT_DEPTH ) && ctx->Depth.Mask ) { - flags |= DRM_TRIDENT_DEPTH; - mask &= ~BUFFER_BIT_DEPTH; - } - - LOCK_HARDWARE(tmesa); - - /* get region after locking: */ - cx = ctx->DrawBuffer->_Xmin; - cy = ctx->DrawBuffer->_Ymin; - cw = ctx->DrawBuffer->_Xmax - cx; - ch = ctx->DrawBuffer->_Ymax - cy; - - if ( flags ) { - - cx += tmesa->drawX; - cy += tmesa->drawY; - - /* HACK!!! - */ - if ( tmesa->dirty & ~TRIDENT_UPLOAD_CLIPRECTS ) { - tridentUploadHwStateLocked( tmesa ); - } - - for ( i = 0 ; i < tmesa->numClipRects ; i++ ) { -#if 0 - int nr = MIN2( i + TRIDENT_NR_SAREA_CLIPRECTS, tmesa->numClipRects ); - drm_clip_rect_t *box = tmesa->pClipRects; - drm_clip_rect_t *b = tmesa->sarea->boxes; - GLint n = 0; - - if ( !all ) { - for ( ; i < nr ; i++ ) { - GLint x = box[i].x1; - GLint y = box[i].y1; - GLint w = box[i].x2 - x; - GLint h = box[i].y2 - y; - - if ( x < cx ) w -= cx - x, x = cx; - if ( y < cy ) h -= cy - y, y = cy; - if ( x + w > cx + cw ) w = cx + cw - x; - if ( y + h > cy + ch ) h = cy + ch - y; - if ( w <= 0 ) continue; - if ( h <= 0 ) continue; - - b->x1 = x; - b->y1 = y; - b->x2 = x + w; - b->y2 = y + h; - b++; - n++; - } - } else { - for ( ; i < nr ; i++ ) { - *b++ = box[i]; - n++; - } - } - - tmesa->sarea->nbox = n; -#endif - -if (flags & DRM_TRIDENT_BACK) { - MMIO_OUT32(MMIO, 0x2150, tmesa->tridentScreen->backPitch << 20 | tmesa->tridentScreen->backOffset>>4); - MMIO_OUT8(MMIO, 0x2127, 0xF0); /* Pat Rop */ - MMIO_OUT32(MMIO, 0x2158, tmesa->ClearColor); - MMIO_OUT32(MMIO, 0x2128, 0x4000); /* solidfill */ - MMIO_OUT32(MMIO, 0x2138, cx << 16 | cy); - MMIO_OUT32(MMIO, 0x2140, cw << 16 | ch); - MMIO_OUT8(MMIO, 0x2124, 0x01); /* BLT */ -#define GE_BUSY 0x80 - for (;;) { - busy = MMIO_IN8(MMIO, 0x2120); - if ( !(busy & GE_BUSY) ) - break; - } -} -if (flags & DRM_TRIDENT_DEPTH) { - MMIO_OUT32(MMIO, 0x2150, tmesa->tridentScreen->depthPitch << 20 | tmesa->tridentScreen->depthOffset>>4); - MMIO_OUT8(MMIO, 0x2127, 0xF0); /* Pat Rop */ - MMIO_OUT32(MMIO, 0x2158, tmesa->ClearColor); - MMIO_OUT32(MMIO, 0x2128, 0x4000); /* solidfill */ - MMIO_OUT32(MMIO, 0x2138, cx << 16 | cy); - MMIO_OUT32(MMIO, 0x2140, cw << 16 | ch); - MMIO_OUT8(MMIO, 0x2124, 0x01); /* BLT */ -#define GE_BUSY 0x80 - for (;;) { - busy = MMIO_IN8(MMIO, 0x2120); - if ( !(busy & GE_BUSY) ) - break; - } -} - MMIO_OUT32(MMIO, 0x2150, tmesa->tridentScreen->frontPitch << 20 | tmesa->tridentScreen->frontOffset>>4); -if (flags & DRM_TRIDENT_FRONT) { - MMIO_OUT8(MMIO, 0x2127, 0xF0); /* Pat Rop */ - MMIO_OUT32(MMIO, 0x2158, tmesa->ClearColor); - MMIO_OUT32(MMIO, 0x2128, 0x4000); /* solidfill */ - MMIO_OUT32(MMIO, 0x2138, cx << 16 | cy); - MMIO_OUT32(MMIO, 0x2140, cw << 16 | ch); - MMIO_OUT8(MMIO, 0x2124, 0x01); /* BLT */ -#define GE_BUSY 0x80 - for (;;) { - busy = MMIO_IN8(MMIO, 0x2120); - if ( !(busy & GE_BUSY) ) - break; - } -} - - } - -#if 0 - tmesa->dirty |= (TRIDENT_UPLOAD_CONTEXT | - TRIDENT_UPLOAD_MISC | - TRIDENT_UPLOAD_CLIPRECTS); -#endif - } - - UNLOCK_HARDWARE(tmesa); - - if ( mask ) - _swrast_Clear( ctx, mask ); -} - -static void tridentDDShadeModel( GLcontext *ctx, GLenum mode ) -{ - tridentContextPtr tmesa = TRIDENT_CONTEXT(ctx); - GLuint s = tmesa->commandD; - -#define TRIDENT_FLAT_SHADE 0x000000E0 -#define TRIDENT_FLAT_SHADE_VERTEX_C 0x00000060 -#define TRIDENT_FLAT_SHADE_GOURAUD 0x00000080 - - s &= ~TRIDENT_FLAT_SHADE; - - switch ( mode ) { - case GL_FLAT: - s |= TRIDENT_FLAT_SHADE_VERTEX_C; - break; - case GL_SMOOTH: - s |= TRIDENT_FLAT_SHADE_GOURAUD; - break; - default: - return; - } - - if ( tmesa->commandD != s ) { - tmesa->commandD = s; - - tmesa->dirty |= TRIDENT_UPLOAD_COMMAND_D; - } -} - -static void -tridentCalcViewport( GLcontext *ctx ) -{ - tridentContextPtr tmesa = TRIDENT_CONTEXT(ctx); - const GLfloat *v = ctx->Viewport._WindowMap.m; - GLfloat *m = tmesa->hw_viewport; - - /* See also trident_translate_vertex. - */ - m[MAT_SX] = v[MAT_SX]; - m[MAT_TX] = v[MAT_TX] + tmesa->drawX + SUBPIXEL_X; - m[MAT_SY] = - v[MAT_SY]; - m[MAT_TY] = - v[MAT_TY] + tmesa->driDrawable->h + tmesa->drawY + SUBPIXEL_Y; -#if 0 - m[MAT_SZ] = v[MAT_SZ] * tmesa->depth_scale; - m[MAT_TZ] = v[MAT_TZ] * tmesa->depth_scale; -#else - m[MAT_SZ] = v[MAT_SZ]; - m[MAT_TZ] = v[MAT_TZ]; -#endif - - tmesa->SetupNewInputs = ~0; -} - -static void tridentDDViewport( GLcontext *ctx, - GLint x, GLint y, - GLsizei width, GLsizei height ) -{ - tridentCalcViewport( ctx ); -} - -static void tridentDDDepthRange( GLcontext *ctx, - GLclampd nearval, GLclampd farval ) -{ - tridentCalcViewport( ctx ); -} - -static void -tridentSetCliprects( tridentContextPtr tmesa, GLenum mode ) -{ - __DRIdrawablePrivate *dPriv = tmesa->driDrawable; - - switch ( mode ) { - case GL_FRONT_LEFT: - if (dPriv->numClipRects == 0) { - static drm_clip_rect_t zeroareacliprect = {0,0,0,0}; - tmesa->numClipRects = 1; - tmesa->pClipRects = &zeroareacliprect; - } else { - tmesa->numClipRects = dPriv->numClipRects; - tmesa->pClipRects = (drm_clip_rect_t *)dPriv->pClipRects; - } - tmesa->drawX = dPriv->x; - tmesa->drawY = dPriv->y; - break; - case GL_BACK_LEFT: - if ( dPriv->numBackClipRects == 0 ) { - if (dPriv->numClipRects == 0) { - static drm_clip_rect_t zeroareacliprect = {0,0,0,0}; - tmesa->numClipRects = 1; - tmesa->pClipRects = &zeroareacliprect; - } else { - tmesa->numClipRects = dPriv->numClipRects; - tmesa->pClipRects = (drm_clip_rect_t *)dPriv->pClipRects; - tmesa->drawX = dPriv->x; - tmesa->drawY = dPriv->y; - } - } - else { - tmesa->numClipRects = dPriv->numBackClipRects; - tmesa->pClipRects = (drm_clip_rect_t *)dPriv->pBackClipRects; - tmesa->drawX = dPriv->backX; - tmesa->drawY = dPriv->backY; - } - break; - default: - return; - } - -#if 0 - tmesa->dirty |= TRIDENT_UPLOAD_CLIPRECTS; -#endif -} - -#if 0 -static GLboolean tridentDDSetDrawBuffer( GLcontext *ctx, GLenum mode ) -{ - tridentContextPtr tmesa = TRIDENT_CONTEXT(ctx); - int found = GL_TRUE; - - if ( tmesa->DrawBuffer != mode ) { - tmesa->DrawBuffer = mode; - - switch ( mode ) { - case GL_FRONT_LEFT: - tridentFallback( tmesa, TRIDENT_FALLBACK_DRAW_BUFFER, GL_FALSE ); - tmesa->drawOffset = tmesa->tridentScreen->frontOffset; - tmesa->drawPitch = tmesa->tridentScreen->frontPitch; - tridentSetCliprects( tmesa, GL_FRONT_LEFT ); - break; - case GL_BACK_LEFT: - tridentFallback( tmesa, TRIDENT_FALLBACK_DRAW_BUFFER, GL_FALSE ); - tmesa->drawOffset = tmesa->tridentScreen->backOffset; - tmesa->drawPitch = tmesa->tridentScreen->backPitch; - tridentSetCliprects( tmesa, GL_BACK_LEFT ); - break; - default: - tridentFallback( tmesa, TRIDENT_FALLBACK_DRAW_BUFFER, GL_TRUE ); - found = GL_FALSE; - break; - } - -#if 0 - tmesa->setup.dst_off_pitch = (((tmesa->drawPitch/8) << 22) | - (tmesa->drawOffset >> 3)); - - tmesa->dirty |= MACH64_UPLOAD_DST_OFF_PITCH | MACH64_UPLOAD_CONTEXT; -#endif - - } - - return found; -} - -static void tridentDDClearColor( GLcontext *ctx, - const GLchan color[4] ) -{ - tridentContextPtr tmesa = TRIDENT_CONTEXT(ctx); - - tmesa->ClearColor = tridentPackColor( tmesa->tridentScreen->cpp, - color[0], color[1], - color[2], color[3] ); -} -#endif - -static void -tridentDDUpdateState( GLcontext *ctx, GLuint new_state ) -{ - _swrast_InvalidateState( ctx, new_state ); - _swsetup_InvalidateState( ctx, new_state ); - _vbo_InvalidateState( ctx, new_state ); - _tnl_InvalidateState( ctx, new_state ); - TRIDENT_CONTEXT(ctx)->new_gl_state |= new_state; -} - - -/* Initialize the context's hardware state. - */ -void tridentDDInitState( tridentContextPtr tmesa ) -{ - tmesa->new_state = 0; - - switch ( tmesa->glCtx->Visual.depthBits ) { - case 16: - tmesa->depth_scale = 1.0 / (GLfloat)0xffff; - break; - case 24: - tmesa->depth_scale = 1.0 / (GLfloat)0xffffff; - break; - } -} - -void tridentDDUpdateHWState( GLcontext *ctx ) -{ - tridentContextPtr tmesa = TRIDENT_CONTEXT(ctx); - int new_state = tmesa->new_state; - - if ( new_state ) - { - tmesa->new_state = 0; - -#if 0 - /* Update the various parts of the context's state. - */ - if ( new_state & GAMMA_NEW_ALPHA ) - tridentUpdateAlphaMode( ctx ); - - if ( new_state & GAMMA_NEW_DEPTH ) - tridentUpdateZMode( ctx ); - - if ( new_state & GAMMA_NEW_FOG ) - gammaUpdateFogAttrib( ctx ); - - if ( new_state & GAMMA_NEW_CLIP ) - gammaUpdateClipping( ctx ); - - if ( new_state & GAMMA_NEW_POLYGON ) - gammaUpdatePolygon( ctx ); - - if ( new_state & GAMMA_NEW_CULL ) - gammaUpdateCull( ctx ); - - if ( new_state & GAMMA_NEW_MASKS ) - gammaUpdateMasks( ctx ); - - if ( new_state & GAMMA_NEW_STIPPLE ) - gammaUpdateStipple( ctx ); -#endif - } - - /* HACK ! */ - -#if 0 - gammaEmitHwState( tmesa ); -#endif -} - -/* Initialize the driver's state functions. - */ -void tridentDDInitStateFuncs( GLcontext *ctx ) -{ - ctx->Driver.UpdateState = tridentDDUpdateState; - - ctx->Driver.Clear = tridentDDClear; - ctx->Driver.DepthRange = tridentDDDepthRange; - ctx->Driver.ShadeModel = tridentDDShadeModel; - ctx->Driver.Viewport = tridentDDViewport; -} diff --git a/src/mesa/drivers/dri/trident/trident_tris.c b/src/mesa/drivers/dri/trident/trident_tris.c deleted file mode 100644 index ee85ab482cd..00000000000 --- a/src/mesa/drivers/dri/trident/trident_tris.c +++ /dev/null @@ -1,1125 +0,0 @@ -/* - * Copyright 2002 by Alan Hourihane, Sychdyn, North Wales, UK. - * - * Permission to use, copy, modify, distribute, and sell this software and its - * documentation for any purpose is hereby granted without fee, provided that - * the above copyright notice appear in all copies and that both that - * copyright notice and this permission notice appear in supporting - * documentation, and that the name of Alan Hourihane not be used in - * advertising or publicity pertaining to distribution of the software without - * specific, written prior permission. Alan Hourihane makes no representations - * about the suitability of this software for any purpose. It is provided - * "as is" without express or implied warranty. - * - * ALAN HOURIHANE DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, - * INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO - * EVENT SHALL ALAN HOURIHANE BE LIABLE FOR ANY SPECIAL, INDIRECT OR - * CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, - * DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER - * TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR - * PERFORMANCE OF THIS SOFTWARE. - * - * Authors: Alan Hourihane, <[email protected]> - * - * Trident CyberBladeXP driver. - * - */ - -#include "trident_context.h" -#include "trident_lock.h" -#include "tnl/tnl.h" -#include "tnl/t_context.h" -#include "tnl/t_pipeline.h" -#include "swrast/swrast.h" -#include "swrast_setup/swrast_setup.h" - -static int first = 1; - -typedef struct reg { - int addr; - int data; -} RegData; - -RegData initRegData[]={ - {0x2804, 0x19980824}, - {0x2F70, 0x46455858}, - {0x2F74, 0x41584998}, - {0x2F00, 0x00000000}, - {0x2F04, 0x80000800}, - {0x2F08, 0x00550200}, - {0x2F40, 0x00000001}, - {0x2F40, 0x00000001}, - {0x2F44, 0x00830097}, - {0x2F48, 0x0087009F}, - {0x2F4C, 0x00BF0003}, - {0x2F50, 0xF00B6C1B}, - {0x2C04, 0x00000000}, - {0x2D00, 0x00000080}, - {0x2D00, 0x00000000}, - {0x2DD4, 0x00100000}, - {0x2DD4, 0x00100010}, - {0x2DD8, 0x00100000}, - {0x2DD8, 0x00100010}, - {0x2C88, 0xFFFFFFFF}, - {0x2C94 , 0xFFFFFFFF}, - {0x281C, 0x00008000}, - {0x2C80, 0x00000000}, - {0x2C80, 0x00000000}, - {0x2C80 , 0x00008000}, - {0x2C00 , 0x00000000}, - {0x2C04 , 0x00000000}, - {0x2C08 , 0x00000000}, - {0x2C0C , 0x00000000}, - {0x2C10 , 0x00000000}, - {0x2C14 , 0x00000000}, - {0x2C18 , 0x00000000}, - {0x2C1C , 0x00000000}, - {0x2C20 , 0x00000000}, - {0x2C24 , 0x00000000}, - {0x2C2C , 0x00000000}, - {0x2C30 , 0x00000000}, - {0x2C34 , 0x00000000}, - {0x2C38 , 0x00000000}, - {0x2C3C , 0x00000000}, - {0x2C40 , 0x00000000}, - {0x2C44 , 0x00000000}, - {0x2C48 , 0x00000000}, - {0x2C4C , 0x00000000}, - {0x2C50 , 0x00000000}, - {0x2C54 , 0x00000000}, - {0x2C58 , 0x00000000}, - {0x2C5C , 0x00000000}, - {0x2C60 , 0x00000000}, - {0x2C64 , 0x00000000}, - {0x2C68 , 0x00000000}, - {0x2C6C , 0x00000000}, - {0x2C70 , 0x00000000}, - {0x2C74 , 0x00000000}, - {0x2C78 , 0x00000000}, - {0x2C7C , 0x00000000}, - {0x2C80 , 0x00008000}, - {0x2C84 , 0x00000000}, - {0x2C88 , 0xFFFFFFFF}, - {0x2C8C , 0x00000000}, - {0x2C90 , 0x00000000}, - {0x2C94 , 0xFFFFFFFF}, - {0x2C98 , 0x00000000}, - {0x2C9C , 0x00000000}, - {0x2CA0 , 0x00000000}, - {0x2CA4 , 0x00000000}, - {0x2CA8 , 0x00000000}, - {0x2CAC , 0x00000000}, - {0x2CB0 , 0x00000000}, - {0x2CB4 , 0x00000000}, - {0x2CB8 , 0x00000000}, - {0x2CBC , 0x00000000}, - {0x2CC0 , 0x00000000}, - {0x2CC4 , 0x00000000}, - {0x2CC8 , 0x00000000}, - {0x2CCC , 0x00000000}, - {0x2CD0 , 0x00000000}, - {0x2CD4 , 0x00000000}, - {0x2CD8 , 0x00000000}, - {0x2CDC , 0x00000000}, - {0x2CE0 , 0x00000000}, - {0x2CE4 , 0x00000000}, - {0x2CE8 , 0x00000000}, - {0x2CEC , 0x00000000}, - {0x2CF0 , 0x00000000}, - {0x2CF4 , 0x00000000}, - {0x2CF8 , 0x00000000}, - {0x2CFC , 0x00000000}, - {0x2D00 , 0x00000000}, - {0x2D04 , 0x00000000}, - {0x2D08 , 0x00000000}, - {0x2D0C , 0x00000000}, - {0x2D10 , 0x00000000}, - {0x2D14 , 0x00000000}, - {0x2D18 , 0x00000000}, - {0x2D1C , 0x00000000}, - {0x2D20 , 0x00000000}, - {0x2D24 , 0x00000000}, - {0x2D28 , 0x00000000}, - {0x2D2C , 0x00000000}, - {0x2D30 , 0x00000000}, - {0x2D34 , 0x00000000}, - {0x2D38 , 0x00000000}, - {0x2D3C , 0x00000000}, - {0x2D40 , 0x00000000}, - {0x2D44 , 0x00000000}, - {0x2D48 , 0x00000000}, - {0x2D4C , 0x00000000}, - {0x2D50 , 0x00000000}, - {0x2D54 , 0x00000000}, - {0x2D58 , 0x00000000}, - {0x2D5C , 0x00000000}, - {0x2D60 , 0x00000000}, - {0x2D64 , 0x00000000}, - {0x2D68 , 0x00000000}, - {0x2D6C , 0x00000000}, - {0x2D70 , 0x00000000}, - {0x2D74 , 0x00000000}, - {0x2D78 , 0x00000000}, - {0x2D7C , 0x00000000}, - {0x2D80 , 0x00000000}, - {0x2D84 , 0x00000000}, - {0x2D88 , 0x00000000}, - {0x2D8C , 0x00000000}, - {0x2D90 , 0x00000000}, - {0x2D94 , 0x00000000}, - {0x2D98 , 0x00000000}, - {0x2D9C , 0x00000000}, - {0x2DA0 , 0x00000000}, - {0x2DA4 , 0x00000000}, - {0x2DA8 , 0x00000000}, - {0x2DAC , 0x00000000}, - {0x2DB0 , 0x00000000}, - {0x2DB4 , 0x00000000}, - {0x2DB8 , 0x00000000}, - {0x2DBC , 0x00000000}, - {0x2DC0 , 0x00000000}, - {0x2DC4 , 0x00000000}, - {0x2DC8 , 0x00000000}, - {0x2DCC , 0x00000000}, - {0x2DD0 , 0x00000000}, - {0x2DD4 , 0x00100010}, - {0x2DD8 , 0x00100010}, - {0x2DDC , 0x00000000}, - {0x2DE0 , 0x00000000}, - {0x2DE4 , 0x00000000}, - {0x2DE8 , 0x00000000}, - {0x2DEC , 0x00000000}, - {0x2DF0 , 0x00000000}, - {0x2DF4 , 0x00000000}, - {0x2DF8 , 0x00000000}, - {0x2DFC , 0x00000000}, - {0x2E00 , 0x00000000}, - {0x2E04 , 0x00000000}, - {0x2E08 , 0x00000000}, - {0x2E0C , 0x00000000}, - {0x2E10 , 0x00000000}, - {0x2E14 , 0x00000000}, - {0x2E18 , 0x00000000}, - {0x2E1C , 0x00000000}, - {0x2E20 , 0x00000000}, - {0x2E24 , 0x00000000}, - {0x2E28 , 0x00000000}, - {0x2E2C , 0x00000000}, - {0x2E30 , 0x00000000}, - {0x2E34 , 0x00000000}, - {0x2E38 , 0x00000000}, - {0x2E3C , 0x00000000}, - {0x2E40 , 0x00000000}, - {0x2E44 , 0x00000000}, - {0x2E48 , 0x00000000}, - {0x2E4C , 0x00000000}, - {0x2E50 , 0x00000000}, - {0x2E54 , 0x00000000}, - {0x2E58 , 0x00000000}, - {0x2E5C , 0x00000000}, - {0x2E60 , 0x00000000}, - {0x2E64 , 0x00000000}, - {0x2E68 , 0x00000000}, - {0x2E6C , 0x00000000}, - {0x2E70 , 0x00000000}, - {0x2E74 , 0x00000000}, - {0x2E78 , 0x00000000}, - {0x2E7C , 0x00000000}, - {0x2E80 , 0x00000000}, - {0x2E84 , 0x00000000}, - {0x2E88 , 0x00000000}, - {0x2E8C , 0x00000000}, - {0x2E90 , 0x00000000}, - {0x2E94 , 0x00000000}, - {0x2E98 , 0x00000000}, - {0x2E9C , 0x00000000}, - {0x2EA0 , 0x00000000}, - {0x2EA4 , 0x00000000}, - {0x2EA8 , 0x00000000}, - {0x2EAC , 0x00000000}, - {0x2EB0 , 0x00000000}, - {0x2EB4 , 0x00000000}, - {0x2EB8 , 0x00000000}, - {0x2EBC , 0x00000000}, - {0x2EC0 , 0x00000000}, - {0x2EC4 , 0x00000000}, - {0x2EC8 , 0x00000000}, - {0x2ECC , 0x00000000}, - {0x2ED0 , 0x00000000}, - {0x2ED4 , 0x00000000}, - {0x2ED8 , 0x00000000}, - {0x2EDC , 0x00000000}, - {0x2EE0 , 0x00000000}, - {0x2EE4 ,0x00000000}, - {0x2EE8 ,0x00000000}, - {0x2EEC , 0x00000000}, - {0x2EF0 , 0x00000000}, - {0x2EF4 , 0x00000000}, - {0x2EF8 , 0x00000000}, - {0x2EFC , 0x00000000}, - /*{0x2F60 , 0x00000000},*/ -}; - -int initRegDataNum=sizeof(initRegData)/sizeof(RegData); - -typedef union { - unsigned int i; - float f; -} dmaBufRec, *dmaBuf; - -void Init3D( tridentContextPtr tmesa ) -{ - unsigned char *MMIO = tmesa->tridentScreen->mmio.map; - int i; - - for(i=0;i<initRegDataNum;++i) - MMIO_OUT32(MMIO, initRegData[i].addr, initRegData[i].data); -} - -int DrawTriangle( tridentContextPtr tmesa) -{ - unsigned char *MMIO = tmesa->tridentScreen->mmio.map; - dmaBufRec clr; - -printf("DRAW TRI\n"); - Init3D(tmesa); - -printf("ENGINE STATUS 0x%x\n",MMIO_IN32(MMIO, 0x2800)); - MMIO_OUT32(MMIO, 0x002800, 0x00000000 ); -#if 0 - MMIO_OUT32(MMIO, 0x002368 , MMIO_IN32(MMIO,0x002368)|1 ); -#endif - - MMIO_OUT32(MMIO, 0x002C00 , 0x00000014 ); -#if 0 - MMIO_OUT32(MMIO, 0x002C04 , 0x0A8004C0 ); -#else - MMIO_OUT32(MMIO, 0x002C04 , 0x0A8000C0 ); -#endif - -#if 0 - MMIO_OUT32(MMIO, 0x002C08 , 0x00000000 ); - MMIO_OUT32(MMIO, 0x002C0C , 0xFFCCCCCC ); - MMIO_OUT32(MMIO, 0x002C10 , 0x3F800000 ); - MMIO_OUT32(MMIO, 0x002C14 , 0x3D0D3DCB ); - MMIO_OUT32(MMIO, 0x002C2C , 0x70000000 ); - MMIO_OUT32(MMIO, 0x002C24 , 0x00202C00 ); - MMIO_OUT32(MMIO, 0x002C28 , 0xE0002500 ); - MMIO_OUT32(MMIO, 0x002C30 , 0x00000000 ); - MMIO_OUT32(MMIO, 0x002C34 , 0xE0000000 ); - MMIO_OUT32(MMIO, 0x002C38 , 0x00000000 ); -#endif - - MMIO_OUT32(MMIO, 0x002C50 , 0x00000000 ); - MMIO_OUT32(MMIO, 0x002C54 , 0x0C320C80 ); - MMIO_OUT32(MMIO, 0x002C50 , 0x00000000 ); - MMIO_OUT32(MMIO, 0x002C54 , 0x0C320C80 ); - MMIO_OUT32(MMIO, 0x002C80 , 0x20008258 ); - MMIO_OUT32(MMIO, 0x002C84 , 0x20000320 ); - MMIO_OUT32(MMIO, 0x002C94 , 0xFFFFFFFF ); - -#if 0 - MMIO_OUT32(MMIO, 0x002D00 , 0x00009009 ); - MMIO_OUT32(MMIO, 0x002D38 , 0x00000000 ); - MMIO_OUT32(MMIO, 0x002D94 , 0x20002000 ); - MMIO_OUT32(MMIO, 0x002D50 , 0xf0000000 ); - MMIO_OUT32(MMIO, 0x002D80 , 0x24002000 ); - MMIO_OUT32(MMIO, 0x002D98 , 0x81000000 ); - MMIO_OUT32(MMIO, 0x002DB0 , 0x81000000 ); - MMIO_OUT32(MMIO, 0x002DC8 , 0x808000FF ); - MMIO_OUT32(MMIO, 0x002DD4 , 0x02000200 ); - MMIO_OUT32(MMIO, 0x002DD8 , 0x02000200 ); - MMIO_OUT32(MMIO, 0x002D30 , 0x02092400 ); - MMIO_OUT32(MMIO, 0x002D04 , 0x00102120 ); - MMIO_OUT32(MMIO, 0x002D08 , 0xFFFFFFFF ); - MMIO_OUT32(MMIO, 0x002D0C , 0xF00010D0 ); - MMIO_OUT32(MMIO, 0x002D10 , 0xC0000400 ); -#endif - - MMIO_OUT32(MMIO, 0x002814, 0x00000000 ); -#if 0 - MMIO_OUT32(MMIO, 0x002818 , 0x00036C20 ); -#else - MMIO_OUT32(MMIO, 0x002818 , 0x00036020 ); -#endif - MMIO_OUT32(MMIO, 0x00281C , 0x00098081 ); - -printf("first TRI\n"); - clr.f = 5.0; - MMIO_OUT32(MMIO, 0x002820 , clr.i ); - clr.f = 595.0; - MMIO_OUT32(MMIO, 0x002824 , clr.i ); - clr.f = 1.0; - MMIO_OUT32(MMIO, 0x002828 , clr.i ); - MMIO_OUT32(MMIO, 0x00282C , 0x00FF00 ); -#if 0 - clr.f = 0.0; - MMIO_OUT32(MMIO, 0x002830 , clr.i ); - clr.f = 1.0; - MMIO_OUT32(MMIO, 0x002834 , clr.i ); -#endif - - clr.f = 5.0; - MMIO_OUT32(MMIO, 0x002820 , clr.i ); - clr.f = 5.0; - MMIO_OUT32(MMIO, 0x002824 , clr.i ); - clr.f = 1.0; - MMIO_OUT32(MMIO, 0x002828 , clr.i ); - MMIO_OUT32(MMIO, 0x00282C , 0xFF0000 ); -#if 0 - clr.f = 0.0; - MMIO_OUT32(MMIO, 0x002830 , clr.i ); - clr.f = 0.0; - MMIO_OUT32(MMIO, 0x002834 , clr.i ); -#endif - - clr.f = 395.0; -printf("0x%x\n",clr.i); - MMIO_OUT32(MMIO, 0x002820 , clr.i ); - clr.f = 5.0; - MMIO_OUT32(MMIO, 0x002824 , clr.i ); - clr.f = 1.0; - MMIO_OUT32(MMIO, 0x002828 , clr.i ); - MMIO_OUT32(MMIO, 0x00282C , 0xFF ); -#if 0 - clr.f = 1.0; - MMIO_OUT32(MMIO, 0x002830 , clr.i ); - clr.f = 0.0; - MMIO_OUT32(MMIO, 0x002834 , clr.i ); -#endif - -printf("sec TRI\n"); - MMIO_OUT32(MMIO, 0x00281C , 0x00093980 ); - clr.f = 395.0; - MMIO_OUT32(MMIO, 0x002820 , clr.i ); - clr.f = 595.0; - MMIO_OUT32(MMIO, 0x002824 , clr.i ); - clr.f = 1.0; - MMIO_OUT32(MMIO, 0x002828 , clr.i ); - MMIO_OUT32(MMIO, 0x00282C , 0x00FF00 ); -#if 0 - clr.f = 1.0; - MMIO_OUT32(MMIO, 0x002830 , clr.i ); - clr.f = 1.0; - MMIO_OUT32(MMIO, 0x002834 , clr.i ); -#endif - -#if 0 - MMIO_OUT32(MMIO, 0x002368 , MMIO_IN32(MMIO,0x002368)&0xfffffffe ); -#endif - -printf("fin TRI\n"); - - return 0; -} - -static INLINE void trident_draw_point(tridentContextPtr tmesa, - const tridentVertex *v0 ) -{ - unsigned char *MMIO = tmesa->tridentScreen->mmio.map; - (void) MMIO; -} - -static INLINE void trident_draw_line( tridentContextPtr tmesa, - const tridentVertex *v0, - const tridentVertex *v1 ) -{ - unsigned char *MMIO = tmesa->tridentScreen->mmio.map; - (void) MMIO; -} - -static INLINE void trident_draw_triangle( tridentContextPtr tmesa, - const tridentVertex *v0, - const tridentVertex *v1, - const tridentVertex *v2 ) -{ -} - -static INLINE void trident_draw_quad( tridentContextPtr tmesa, - const tridentVertex *v0, - const tridentVertex *v1, - const tridentVertex *v2, - const tridentVertex *v3 ) -{ - GLuint vertsize = tmesa->vertex_size; - GLint coloridx = (vertsize > 4) ? 4 : 3; - unsigned char *MMIO = tmesa->tridentScreen->mmio.map; - int clr; - float *ftmp = (float *)(&clr); - - if (tmesa->dirty) - tridentUploadHwStateLocked( tmesa ); -#if 0 - DrawTriangle(tmesa); - exit(0); -#else -#if 1 - if (first) { - Init3D(tmesa); -#if 0 - DrawTriangle(tmesa); -#endif - first = 0; - } -#endif - - LOCK_HARDWARE( tmesa ); - - MMIO_OUT32(MMIO, 0x002C00 , 0x00000010 ); - MMIO_OUT32(MMIO, 0x002C04 , 0x029C00C0 ); - - /* Z buffer */ - MMIO_OUT32(MMIO, 0x002C24 , 0x00100000 /*| (tmesa->tridentScreen->depthOffset)*/ ); - MMIO_OUT32(MMIO, 0x002C28 , 0xE0000000 | (tmesa->tridentScreen->depthPitch * 4) ); - - /* front buffer */ - MMIO_OUT32(MMIO, 0x002C50 , 0x00000000 | (tmesa->drawOffset) ); - MMIO_OUT32(MMIO, 0x002C54 , 0x0C320000 | (tmesa->drawPitch * 4) ); - - /* clipper */ - MMIO_OUT32(MMIO, 0x002C80 , 0x20008000 | tmesa->tridentScreen->height ); - MMIO_OUT32(MMIO, 0x002C84 , 0x20000000 | tmesa->tridentScreen->width ); - - /* writemask */ - MMIO_OUT32(MMIO, 0x002C94 , 0xFFFFFFFF ); - -if (vertsize == 4) { - MMIO_OUT32(MMIO, 0x002818 , 0x0003A020 ); - MMIO_OUT32(MMIO, 0x00281C , 0x00098021 ); - - *ftmp = v0->v.x; - MMIO_OUT32(MMIO, 0x002820 , clr ); - *ftmp = v0->v.y; - MMIO_OUT32(MMIO, 0x002824 , clr ); - *ftmp = v0->v.z; - MMIO_OUT32(MMIO, 0x002828 , clr ); -#if 0 - *ftmp = v0->v.w; - MMIO_OUT32(MMIO, 0x00282C , clr ); -#endif - MMIO_OUT32(MMIO, 0x00282C , v0->ui[coloridx] ); - - *ftmp = v1->v.x; - MMIO_OUT32(MMIO, 0x002820 , clr ); - *ftmp = v1->v.y; - MMIO_OUT32(MMIO, 0x002824 , clr ); - *ftmp = v1->v.z; - MMIO_OUT32(MMIO, 0x002828 , clr ); -#if 0 - *ftmp = v1->v.w; - MMIO_OUT32(MMIO, 0x00282C , clr ); -#endif - MMIO_OUT32(MMIO, 0x00282C , v1->ui[coloridx] ); - - *ftmp = v2->v.x; - MMIO_OUT32(MMIO, 0x002820 , clr ); - *ftmp = v2->v.y; - MMIO_OUT32(MMIO, 0x002824 , clr ); - *ftmp = v2->v.z; - MMIO_OUT32(MMIO, 0x002828 , clr ); -#if 0 - *ftmp = v2->v.w; - MMIO_OUT32(MMIO, 0x00282C , clr ); -#endif - MMIO_OUT32(MMIO, 0x00282C , v2->ui[coloridx] ); - - MMIO_OUT32(MMIO, 0x00281C , 0x00093020 ); - *ftmp = v3->v.x; - MMIO_OUT32(MMIO, 0x002820 , clr ); - *ftmp = v3->v.y; - MMIO_OUT32(MMIO, 0x002824 , clr ); - *ftmp = v3->v.z; - MMIO_OUT32(MMIO, 0x002828 , clr ); -#if 0 - *ftmp = v3->v.w; - MMIO_OUT32(MMIO, 0x00282C , clr ); -#endif - MMIO_OUT32(MMIO, 0x00282C , v3->ui[coloridx] ); - -} -#endif - - UNLOCK_HARDWARE( tmesa ); -} -/*********************************************************************** - * Rasterization fallback helpers * - ***********************************************************************/ - - -/* This code is hit only when a mix of accelerated and unaccelerated - * primitives are being drawn, and only for the unaccelerated - * primitives. - */ -#if 0 -static void -trident_fallback_quad( tridentContextPtr tmesa, - const tridentVertex *v0, - const tridentVertex *v1, - const tridentVertex *v2, - const tridentVertex *v3 ) -{ - GLcontext *ctx = tmesa->glCtx; - SWvertex v[4]; - trident_translate_vertex( ctx, v0, &v[0] ); - trident_translate_vertex( ctx, v1, &v[1] ); - trident_translate_vertex( ctx, v2, &v[2] ); - trident_translate_vertex( ctx, v3, &v[3] ); - _swrast_Quad( ctx, &v[0], &v[1], &v[2], &v[3] ); -} -#endif - -/* XXX hack to get the prototype defined in time... */ -void trident_translate_vertex(GLcontext *ctx, const tridentVertex *src, - SWvertex *dst); - -static void -trident_fallback_tri( tridentContextPtr tmesa, - const tridentVertex *v0, - const tridentVertex *v1, - const tridentVertex *v2 ) -{ - GLcontext *ctx = tmesa->glCtx; - SWvertex v[3]; - trident_translate_vertex( ctx, v0, &v[0] ); - trident_translate_vertex( ctx, v1, &v[1] ); - trident_translate_vertex( ctx, v2, &v[2] ); - _swrast_Triangle( ctx, &v[0], &v[1], &v[2] ); -} - -static void -trident_fallback_line( tridentContextPtr tmesa, - const tridentVertex *v0, - const tridentVertex *v1 ) -{ - GLcontext *ctx = tmesa->glCtx; - SWvertex v[2]; - trident_translate_vertex( ctx, v0, &v[0] ); - trident_translate_vertex( ctx, v1, &v[1] ); - _swrast_Line( ctx, &v[0], &v[1] ); -} - - -static void -trident_fallback_point( tridentContextPtr tmesa, - const tridentVertex *v0 ) -{ - GLcontext *ctx = tmesa->glCtx; - SWvertex v[1]; - trident_translate_vertex( ctx, v0, &v[0] ); - _swrast_Point( ctx, &v[0] ); -} - -/*********************************************************************** - * Macros for t_dd_tritmp.h to draw basic primitives * - ***********************************************************************/ - -#define TRI( a, b, c ) \ -do { \ - if (DO_FALLBACK) \ - tmesa->draw_tri( tmesa, a, b, c ); \ - else \ - trident_draw_triangle( tmesa, a, b, c ); \ -} while (0) - -#define QUAD( a, b, c, d ) \ -do { \ - if (DO_FALLBACK) { \ - tmesa->draw_tri( tmesa, a, b, d ); \ - tmesa->draw_tri( tmesa, b, c, d ); \ - } else \ - trident_draw_quad( tmesa, a, b, c, d ); \ -} while (0) - -#define LINE( v0, v1 ) \ -do { \ - if (DO_FALLBACK) \ - tmesa->draw_line( tmesa, v0, v1 ); \ - else \ - trident_draw_line( tmesa, v0, v1 ); \ -} while (0) - -#define POINT( v0 ) \ -do { \ - if (DO_FALLBACK) \ - tmesa->draw_point( tmesa, v0 ); \ - else \ - trident_draw_point( tmesa, v0 ); \ -} while (0) - -/*********************************************************************** - * Build render functions from dd templates * - ***********************************************************************/ - -#define TRIDENT_OFFSET_BIT 0x01 -#define TRIDENT_TWOSIDE_BIT 0x02 -#define TRIDENT_UNFILLED_BIT 0x04 -#define TRIDENT_FALLBACK_BIT 0x08 -#define TRIDENT_MAX_TRIFUNC 0x10 - - -static struct { - tnl_points_func points; - tnl_line_func line; - tnl_triangle_func triangle; - tnl_quad_func quad; -} rast_tab[TRIDENT_MAX_TRIFUNC]; - - -#define DO_FALLBACK (IND & TRIDENT_FALLBACK_BIT) -#define DO_OFFSET (IND & TRIDENT_OFFSET_BIT) -#define DO_UNFILLED (IND & TRIDENT_UNFILLED_BIT) -#define DO_TWOSIDE (IND & TRIDENT_TWOSIDE_BIT) -#define DO_FLAT 0 -#define DO_TRI 1 -#define DO_QUAD 1 -#define DO_LINE 1 -#define DO_POINTS 1 -#define DO_FULL_QUAD 1 - -#define HAVE_RGBA 1 -#define HAVE_SPEC 1 -#define HAVE_BACK_COLORS 0 -#define HAVE_HW_FLATSHADE 1 -#define VERTEX tridentVertex -#define TAB rast_tab - -#define DEPTH_SCALE 1.0 -#define UNFILLED_TRI unfilled_tri -#define UNFILLED_QUAD unfilled_quad -#define VERT_X(_v) _v->v.x -#define VERT_Y(_v) _v->v.y -#define VERT_Z(_v) _v->v.z -#define AREA_IS_CCW( a ) (a > 0) -#define GET_VERTEX(e) (tmesa->verts + (e<<tmesa->vertex_stride_shift)) - -#define TRIDENT_COLOR( dst, src ) \ -do { \ - dst[0] = src[2]; \ - dst[1] = src[1]; \ - dst[2] = src[0]; \ - dst[3] = src[3]; \ -} while (0) - -#define TRIDENT_SPEC( dst, src ) \ -do { \ - dst[0] = src[2]; \ - dst[1] = src[1]; \ - dst[2] = src[0]; \ -} while (0) - -#define VERT_SET_RGBA( v, c ) TRIDENT_COLOR( v->ub4[coloroffset], c ) -#define VERT_COPY_RGBA( v0, v1 ) v0->ui[coloroffset] = v1->ui[coloroffset] -#define VERT_SAVE_RGBA( idx ) color[idx] = v[idx]->ui[coloroffset] -#define VERT_RESTORE_RGBA( idx ) v[idx]->ui[coloroffset] = color[idx] - -#define VERT_SET_SPEC( v, c ) if (havespec) TRIDENT_SPEC( v->ub4[5], c ) -#define VERT_COPY_SPEC( v0, v1 ) if (havespec) COPY_3V(v0->ub4[5], v1->ub4[5]) -#define VERT_SAVE_SPEC( idx ) if (havespec) spec[idx] = v[idx]->ui[5] -#define VERT_RESTORE_SPEC( idx ) if (havespec) v[idx]->ui[5] = spec[idx] - -#define LOCAL_VARS(n) \ - tridentContextPtr tmesa = TRIDENT_CONTEXT(ctx); \ - GLuint color[n], spec[n]; \ - GLuint coloroffset = (tmesa->vertex_size == 4 ? 3 : 4); \ - GLboolean havespec = (tmesa->vertex_size == 4 ? 0 : 1); \ - (void) color; (void) spec; (void) coloroffset; (void) havespec; -/*********************************************************************** - * Helpers for rendering unfilled primitives * - ***********************************************************************/ -#if 0 -static const GLuint hw_prim[GL_POLYGON+1] = { - B_PrimType_Points, - B_PrimType_Lines, - B_PrimType_Lines, - B_PrimType_Lines, - B_PrimType_Triangles, - B_PrimType_Triangles, - B_PrimType_Triangles, - B_PrimType_Triangles, - B_PrimType_Triangles, - B_PrimType_Triangles -}; -#endif - -static void tridentResetLineStipple( GLcontext *ctx ); -#if 0 -static void tridentRasterPrimitive( GLcontext *ctx, GLuint hwprim ); -#endif -static void tridentRenderPrimitive( GLcontext *ctx, GLenum prim ); - -#define RASTERIZE(x) /*if (tmesa->hw_primitive != hw_prim[x]) \ - tridentRasterPrimitive( ctx, hw_prim[x] ) */ -#define RENDER_PRIMITIVE tmesa->render_primitive -#define TAG(x) x -#define IND TRIDENT_FALLBACK_BIT -#include "tnl_dd/t_dd_unfilled.h" -#undef IND - -/*********************************************************************** - * Generate GL render functions * - ***********************************************************************/ - -#define IND (0) -#define TAG(x) x -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (TRIDENT_OFFSET_BIT) -#define TAG(x) x##_offset -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (TRIDENT_TWOSIDE_BIT) -#define TAG(x) x##_twoside -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (TRIDENT_TWOSIDE_BIT|TRIDENT_OFFSET_BIT) -#define TAG(x) x##_twoside_offset -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (TRIDENT_UNFILLED_BIT) -#define TAG(x) x##_unfilled -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (TRIDENT_OFFSET_BIT|TRIDENT_UNFILLED_BIT) -#define TAG(x) x##_offset_unfilled -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (TRIDENT_TWOSIDE_BIT|TRIDENT_UNFILLED_BIT) -#define TAG(x) x##_twoside_unfilled -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (TRIDENT_TWOSIDE_BIT|TRIDENT_OFFSET_BIT|TRIDENT_UNFILLED_BIT) -#define TAG(x) x##_twoside_offset_unfilled -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (TRIDENT_FALLBACK_BIT) -#define TAG(x) x##_fallback -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (TRIDENT_OFFSET_BIT|TRIDENT_FALLBACK_BIT) -#define TAG(x) x##_offset_fallback -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (TRIDENT_TWOSIDE_BIT|TRIDENT_FALLBACK_BIT) -#define TAG(x) x##_twoside_fallback -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (TRIDENT_TWOSIDE_BIT|TRIDENT_OFFSET_BIT|TRIDENT_FALLBACK_BIT) -#define TAG(x) x##_twoside_offset_fallback -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (TRIDENT_UNFILLED_BIT|TRIDENT_FALLBACK_BIT) -#define TAG(x) x##_unfilled_fallback -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (TRIDENT_OFFSET_BIT|TRIDENT_UNFILLED_BIT|TRIDENT_FALLBACK_BIT) -#define TAG(x) x##_offset_unfilled_fallback -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (TRIDENT_TWOSIDE_BIT|TRIDENT_UNFILLED_BIT|TRIDENT_FALLBACK_BIT) -#define TAG(x) x##_twoside_unfilled_fallback -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (TRIDENT_TWOSIDE_BIT|TRIDENT_OFFSET_BIT|TRIDENT_UNFILLED_BIT|TRIDENT_FALLBACK_BIT) -#define TAG(x) x##_twoside_offset_unfilled_fallback -#include "tnl_dd/t_dd_tritmp.h" - -static void init_rast_tab( void ) -{ - init(); - init_offset(); - init_twoside(); - init_twoside_offset(); - init_unfilled(); - init_offset_unfilled(); - init_twoside_unfilled(); - init_twoside_offset_unfilled(); - init_fallback(); - init_offset_fallback(); - init_twoside_fallback(); - init_twoside_offset_fallback(); - init_unfilled_fallback(); - init_offset_unfilled_fallback(); - init_twoside_unfilled_fallback(); - init_twoside_offset_unfilled_fallback(); -} - - -/**********************************************************************/ -/* Render unclipped begin/end objects */ -/**********************************************************************/ - -#define VERT(x) (tridentVertex *)(tridentverts + (x << shift)) -#define RENDER_POINTS( start, count ) \ - for ( ; start < count ; start++) \ - trident_draw_point( tmesa, VERT(start) ) -#define RENDER_LINE( v0, v1 ) \ - trident_draw_line( tmesa, VERT(v0), VERT(v1) ) -#define RENDER_TRI( v0, v1, v2 ) \ - trident_draw_triangle( tmesa, VERT(v0), VERT(v1), VERT(v2) ) -#define RENDER_QUAD( v0, v1, v2, v3 ) \ - trident_draw_quad( tmesa, VERT(v0), VERT(v1), VERT(v2), VERT(v3) ) -#define INIT(x) tridentRenderPrimitive( ctx, x ); -#undef LOCAL_VARS -#define LOCAL_VARS \ - tridentContextPtr tmesa = TRIDENT_CONTEXT(ctx); \ - const GLuint shift = tmesa->vertex_stride_shift; \ - const char *tridentverts = (char *)tmesa->verts; \ - const GLboolean stipple = ctx->Line.StippleFlag; \ - const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts; \ - (void) elt; -#define RESET_STIPPLE if ( stipple ) tridentResetLineStipple( ctx ); -#define RESET_OCCLUSION -#define PRESERVE_VB_DEFS -#define ELT(x) (x) -#define TAG(x) trident_##x##_verts -#include "tnl/t_vb_rendertmp.h" -#undef ELT -#undef TAG -#define TAG(x) trident_##x##_elts -#define ELT(x) elt[x] -#include "tnl/t_vb_rendertmp.h" - -/**********************************************************************/ -/* Render clipped primitives */ -/**********************************************************************/ - -static void tridentRenderClippedPoly( GLcontext *ctx, const GLuint *elts, - GLuint n ) -{ - tridentContextPtr tmesa = TRIDENT_CONTEXT(ctx); - struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - TNLcontext *tnl = TNL_CONTEXT(ctx); - GLuint prim = tmesa->render_primitive; - - /* Render the new vertices as an unclipped polygon. - */ - { - GLuint *tmp = VB->Elts; - VB->Elts = (GLuint *)elts; - tnl->Driver.Render.PrimTabElts[GL_POLYGON]( ctx, 0, n, PRIM_BEGIN|PRIM_END ); - VB->Elts = tmp; - } - - /* Restore the render primitive - */ - if (prim != GL_POLYGON) - tnl->Driver.Render.PrimitiveNotify( ctx, prim ); -} - -static void tridentRenderClippedLine( GLcontext *ctx, GLuint ii, GLuint jj ) -{ - TNLcontext *tnl = TNL_CONTEXT(ctx); - tnl->Driver.Render.Line( ctx, ii, jj ); -} - - -/**********************************************************************/ -/* Choose render functions */ -/**********************************************************************/ - -#define _TRIDENT_NEW_RENDER_STATE (_DD_NEW_LINE_STIPPLE | \ - _DD_NEW_LINE_SMOOTH | \ - _DD_NEW_POINT_SMOOTH | \ - _DD_NEW_TRI_SMOOTH | \ - _DD_NEW_TRI_UNFILLED | \ - _DD_NEW_TRI_LIGHT_TWOSIDE | \ - _DD_NEW_TRI_OFFSET) \ - - -#define POINT_FALLBACK (DD_POINT_SMOOTH) -#define LINE_FALLBACK (DD_LINE_STIPPLE|DD_LINE_SMOOTH) -#define TRI_FALLBACK (DD_TRI_SMOOTH) -#define ANY_FALLBACK_FLAGS (POINT_FALLBACK|LINE_FALLBACK|TRI_FALLBACK) -#define ANY_RASTER_FLAGS (DD_TRI_LIGHT_TWOSIDE|DD_TRI_OFFSET|DD_TRI_UNFILLED) - - -static void tridentChooseRenderState(GLcontext *ctx) -{ - tridentContextPtr tmesa = TRIDENT_CONTEXT(ctx); - TNLcontext *tnl = TNL_CONTEXT(ctx); - GLuint flags = ctx->_TriangleCaps; - GLuint index = 0; - - if (flags & (ANY_RASTER_FLAGS|ANY_FALLBACK_FLAGS)) { - tmesa->draw_point = trident_draw_point; - tmesa->draw_line = trident_draw_line; - tmesa->draw_tri = trident_draw_triangle; - - if (flags & ANY_RASTER_FLAGS) { - if (flags & DD_TRI_LIGHT_TWOSIDE) index |= TRIDENT_TWOSIDE_BIT; - if (flags & DD_TRI_OFFSET) index |= TRIDENT_OFFSET_BIT; - if (flags & DD_TRI_UNFILLED) index |= TRIDENT_UNFILLED_BIT; - } - - /* Hook in fallbacks for specific primitives. - */ - if (flags & (POINT_FALLBACK|LINE_FALLBACK|TRI_FALLBACK)) { - if (flags & POINT_FALLBACK) tmesa->draw_point = trident_fallback_point; - if (flags & LINE_FALLBACK) tmesa->draw_line = trident_fallback_line; - if (flags & TRI_FALLBACK) tmesa->draw_tri = trident_fallback_tri; - index |= TRIDENT_FALLBACK_BIT; - } - } - - if (tmesa->RenderIndex != index) { - tmesa->RenderIndex = index; - - tnl->Driver.Render.Points = rast_tab[index].points; - tnl->Driver.Render.Line = rast_tab[index].line; - tnl->Driver.Render.Triangle = rast_tab[index].triangle; - tnl->Driver.Render.Quad = rast_tab[index].quad; - - if (tmesa->RenderIndex == 0) { - tnl->Driver.Render.PrimTabVerts = trident_render_tab_verts; - tnl->Driver.Render.PrimTabElts = trident_render_tab_elts; - } else { - tnl->Driver.Render.PrimTabVerts = _tnl_render_tab_verts; - tnl->Driver.Render.PrimTabElts = _tnl_render_tab_elts; - } - tnl->Driver.Render.ClippedLine = tridentRenderClippedLine; - tnl->Driver.Render.ClippedPolygon = tridentRenderClippedPoly; - } -} - - -/**********************************************************************/ -/* High level hooks for t_vb_render.c */ -/**********************************************************************/ - - - -/* Determine the rasterized primitive when not drawing unfilled - * polygons. - * - * Used only for the default render stage which always decomposes - * primitives to trianges/lines/points. For the accelerated stage, - * which renders strips as strips, the equivalent calculations are - * performed in tridentrender.c. - */ -#if 0 -static void tridentRasterPrimitive( GLcontext *ctx, GLuint hwprim ) -{ - tridentContextPtr tmesa = TRIDENT_CONTEXT(ctx); - if (tmesa->hw_primitive != hwprim) - tmesa->hw_primitive = hwprim; -} -#endif - -static void tridentRenderPrimitive( GLcontext *ctx, GLenum prim ) -{ - tridentContextPtr tmesa = TRIDENT_CONTEXT(ctx); - tmesa->render_primitive = prim; -} - -static void tridentRunPipeline( GLcontext *ctx ) -{ - tridentContextPtr tmesa = TRIDENT_CONTEXT(ctx); - - if ( tmesa->new_state ) - tridentDDUpdateHWState( ctx ); - - if (tmesa->new_gl_state) { -#if 0 - if (tmesa->new_gl_state & _NEW_TEXTURE) - tridentUpdateTextureState( ctx ); -#endif - - if (!tmesa->Fallback) { - if (tmesa->new_gl_state & _TRIDENT_NEW_VERTEX) - tridentChooseVertexState( ctx ); - - if (tmesa->new_gl_state & _TRIDENT_NEW_RENDER_STATE) - tridentChooseRenderState( ctx ); - } - - tmesa->new_gl_state = 0; - } - - _tnl_run_pipeline( ctx ); -} - -static void tridentRenderStart( GLcontext *ctx ) -{ - /* Check for projective texturing. Make sure all texcoord - * pointers point to something. (fix in mesa?) - */ - tridentCheckTexSizes( ctx ); -} - -static void tridentRenderFinish( GLcontext *ctx ) -{ - if (0) - _swrast_flush( ctx ); /* never needed */ -} - -static void tridentResetLineStipple( GLcontext *ctx ) -{ - tridentContextPtr tmesa = TRIDENT_CONTEXT(ctx); - (void) tmesa; - - /* Reset the hardware stipple counter. - */ -} - - -/**********************************************************************/ -/* Transition to/from hardware rasterization. */ -/**********************************************************************/ - - -void tridentFallback( tridentContextPtr tmesa, GLuint bit, GLboolean mode ) -{ - GLcontext *ctx = tmesa->glCtx; - TNLcontext *tnl = TNL_CONTEXT(ctx); - GLuint oldfallback = tmesa->Fallback; - - _tnl_need_projected_coords( ctx, GL_FALSE ); - - if (mode) { - tmesa->Fallback |= bit; - if (oldfallback == 0) { - _swsetup_Wakeup( ctx ); - tmesa->RenderIndex = ~0; - } - } - else { - tmesa->Fallback &= ~bit; - if (oldfallback == bit) { - _swrast_flush( ctx ); - tnl->Driver.Render.Start = tridentRenderStart; - tnl->Driver.Render.PrimitiveNotify = tridentRenderPrimitive; - tnl->Driver.Render.Finish = tridentRenderFinish; - tnl->Driver.Render.BuildVertices = tridentBuildVertices; - tnl->Driver.Render.ResetLineStipple = tridentResetLineStipple; - tmesa->new_gl_state |= (_TRIDENT_NEW_RENDER_STATE| - _TRIDENT_NEW_VERTEX); - } - } -} - - -/**********************************************************************/ -/* Initialization. */ -/**********************************************************************/ - - -void tridentDDInitTriFuncs( GLcontext *ctx ) -{ - tridentContextPtr tmesa = TRIDENT_CONTEXT(ctx); - TNLcontext *tnl = TNL_CONTEXT(ctx); - static int firsttime = 1; - - if (firsttime) { - init_rast_tab(); - firsttime = 0; - } - - tmesa->RenderIndex = ~0; - - tnl->Driver.RunPipeline = tridentRunPipeline; - tnl->Driver.Render.Start = tridentRenderStart; - tnl->Driver.Render.Finish = tridentRenderFinish; - tnl->Driver.Render.PrimitiveNotify = tridentRenderPrimitive; - tnl->Driver.Render.ResetLineStipple = tridentResetLineStipple; - tnl->Driver.Render.BuildVertices = tridentBuildVertices; -} diff --git a/src/mesa/drivers/dri/trident/trident_vb.c b/src/mesa/drivers/dri/trident/trident_vb.c deleted file mode 100644 index 055a9145953..00000000000 --- a/src/mesa/drivers/dri/trident/trident_vb.c +++ /dev/null @@ -1,434 +0,0 @@ -/* - * Copyright 2002 by Alan Hourihane, Sychdyn, North Wales, UK. - * - * Permission to use, copy, modify, distribute, and sell this software and its - * documentation for any purpose is hereby granted without fee, provided that - * the above copyright notice appear in all copies and that both that - * copyright notice and this permission notice appear in supporting - * documentation, and that the name of Alan Hourihane not be used in - * advertising or publicity pertaining to distribution of the software without - * specific, written prior permission. Alan Hourihane makes no representations - * about the suitability of this software for any purpose. It is provided - * "as is" without express or implied warranty. - * - * ALAN HOURIHANE DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, - * INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO - * EVENT SHALL ALAN HOURIHANE BE LIABLE FOR ANY SPECIAL, INDIRECT OR - * CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, - * DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER - * TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR - * PERFORMANCE OF THIS SOFTWARE. - * - * Authors: Alan Hourihane, <[email protected]> - * - * Trident CyberBladeXP driver. - * - */ -#include "main/glheader.h" -#include "main/mtypes.h" -#include "main/macros.h" -#include "main/colormac.h" - -#include "swrast_setup/swrast_setup.h" -#include "swrast/swrast.h" -#include "tnl/t_context.h" -#include "tnl/tnl.h" - -#include "trident_context.h" - -#define TRIDENT_TEX1_BIT 0x1 -#define TRIDENT_TEX0_BIT 0x2 -#define TRIDENT_RGBA_BIT 0x4 -#define TRIDENT_SPEC_BIT 0x8 -#define TRIDENT_FOG_BIT 0x10 -#define TRIDENT_XYZW_BIT 0x20 -#define TRIDENT_PTEX_BIT 0x40 -#define TRIDENT_MAX_SETUP 0x80 - -static struct { - void (*emit)( GLcontext *, GLuint, GLuint, void *, GLuint ); - tnl_interp_func interp; - tnl_copy_pv_func copy_pv; - GLboolean (*check_tex_sizes)( GLcontext *ctx ); - GLuint vertex_size; - GLuint vertex_stride_shift; - GLuint vertex_format; -} setup_tab[TRIDENT_MAX_SETUP]; - -#define TINY_VERTEX_FORMAT 1 -#define NOTEX_VERTEX_FORMAT 2 -#define TEX0_VERTEX_FORMAT 3 -#define TEX1_VERTEX_FORMAT 4 -#define PROJ_TEX1_VERTEX_FORMAT 5 -#define TEX2_VERTEX_FORMAT 6 -#define TEX3_VERTEX_FORMAT 7 -#define PROJ_TEX3_VERTEX_FORMAT 8 - -#define DO_XYZW (IND & TRIDENT_XYZW_BIT) -#define DO_RGBA (IND & TRIDENT_RGBA_BIT) -#define DO_SPEC (IND & TRIDENT_SPEC_BIT) -#define DO_FOG (IND & TRIDENT_FOG_BIT) -#define DO_TEX0 (IND & TRIDENT_TEX0_BIT) -#define DO_TEX1 (IND & TRIDENT_TEX1_BIT) -#define DO_TEX2 0 -#define DO_TEX3 0 -#define DO_PTEX (IND & TRIDENT_PTEX_BIT) - -#define VERTEX tridentVertex -#define VERTEX_COLOR trident_color_t -#define LOCALVARS tridentContextPtr tmesa = TRIDENT_CONTEXT(ctx); -#define GET_VIEWPORT_MAT() tmesa->hw_viewport -#define GET_TEXSOURCE(n) tmesa->tmu_source[n] -#define GET_VERTEX_FORMAT() tmesa->vertex_format -#define GET_VERTEX_SIZE() tmesa->vertex_size -#define GET_VERTEX_STORE() tmesa->verts -#define GET_VERTEX_STRIDE_SHIFT() tmesa->vertex_stride_shift -#define GET_UBYTE_COLOR_STORE() &tmesa->UbyteColor -#define GET_UBYTE_SPEC_COLOR_STORE() &tmesa->UbyteSecondaryColor - -#define HAVE_HW_VIEWPORT 0 -#define HAVE_HW_DIVIDE 0 -#define HAVE_RGBA_COLOR 0 -#define HAVE_TINY_VERTICES 1 -#define HAVE_NOTEX_VERTICES 1 -#define HAVE_TEX0_VERTICES 1 -#define HAVE_TEX1_VERTICES 1 -#define HAVE_TEX2_VERTICES 0 -#define HAVE_TEX3_VERTICES 0 -#define HAVE_PTEX_VERTICES 0 - -#define UNVIEWPORT_VARS \ - const GLfloat dx = - tmesa->drawX - SUBPIXEL_X; \ - const GLfloat dy = (tmesa->driDrawable->h + \ - tmesa->drawY + SUBPIXEL_Y); \ - const GLfloat sz = 1.0 / tmesa->depth_scale - -#define UNVIEWPORT_X(x) x + dx; -#define UNVIEWPORT_Y(y) - y + dy; -#define UNVIEWPORT_Z(z) z * sz; - -#define PTEX_FALLBACK() tridentFallback(TRIDENT_CONTEXT(ctx), TRIDENT_FALLBACK_TEXTURE, 1) - -#define IMPORT_FLOAT_COLORS trident_import_float_colors -#define IMPORT_FLOAT_SPEC_COLORS trident_import_float_spec_colors - -#define INTERP_VERTEX setup_tab[tmesa->SetupIndex].interp -#define COPY_PV_VERTEX setup_tab[tmesa->SetupIndex].copy_pv - -/*********************************************************************** - * Generate pv-copying and translation functions * - ***********************************************************************/ - -#define TAG(x) trident_##x -#include "tnl_dd/t_dd_vb.c" - -/*********************************************************************** - * Generate vertex emit and interp functions * - ***********************************************************************/ - -#define IND (TRIDENT_XYZW_BIT|TRIDENT_RGBA_BIT) -#define TAG(x) x##_wg -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_XYZW_BIT|TRIDENT_RGBA_BIT|TRIDENT_SPEC_BIT) -#define TAG(x) x##_wgs -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_XYZW_BIT|TRIDENT_RGBA_BIT|TRIDENT_TEX0_BIT) -#define TAG(x) x##_wgt0 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_XYZW_BIT|TRIDENT_RGBA_BIT|TRIDENT_TEX0_BIT|TRIDENT_TEX1_BIT) -#define TAG(x) x##_wgt0t1 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_XYZW_BIT|TRIDENT_RGBA_BIT|TRIDENT_TEX0_BIT|TRIDENT_PTEX_BIT) -#define TAG(x) x##_wgpt0 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_XYZW_BIT|TRIDENT_RGBA_BIT|TRIDENT_SPEC_BIT|TRIDENT_TEX0_BIT) -#define TAG(x) x##_wgst0 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_XYZW_BIT|TRIDENT_RGBA_BIT|TRIDENT_SPEC_BIT|TRIDENT_TEX0_BIT|TRIDENT_TEX1_BIT) -#define TAG(x) x##_wgst0t1 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_XYZW_BIT|TRIDENT_RGBA_BIT|TRIDENT_SPEC_BIT|TRIDENT_TEX0_BIT|TRIDENT_PTEX_BIT) -#define TAG(x) x##_wgspt0 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_XYZW_BIT|TRIDENT_RGBA_BIT|TRIDENT_FOG_BIT) -#define TAG(x) x##_wgf -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_XYZW_BIT|TRIDENT_RGBA_BIT|TRIDENT_FOG_BIT|TRIDENT_SPEC_BIT) -#define TAG(x) x##_wgfs -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_XYZW_BIT|TRIDENT_RGBA_BIT|TRIDENT_FOG_BIT|TRIDENT_TEX0_BIT) -#define TAG(x) x##_wgft0 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_XYZW_BIT|TRIDENT_RGBA_BIT|TRIDENT_FOG_BIT|TRIDENT_TEX0_BIT|TRIDENT_TEX1_BIT) -#define TAG(x) x##_wgft0t1 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_XYZW_BIT|TRIDENT_RGBA_BIT|TRIDENT_FOG_BIT|TRIDENT_TEX0_BIT|TRIDENT_PTEX_BIT) -#define TAG(x) x##_wgfpt0 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_XYZW_BIT|TRIDENT_RGBA_BIT|TRIDENT_FOG_BIT|TRIDENT_SPEC_BIT|TRIDENT_TEX0_BIT) -#define TAG(x) x##_wgfst0 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_XYZW_BIT|TRIDENT_RGBA_BIT|TRIDENT_FOG_BIT|TRIDENT_SPEC_BIT|TRIDENT_TEX0_BIT|TRIDENT_TEX1_BIT) -#define TAG(x) x##_wgfst0t1 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_XYZW_BIT|TRIDENT_RGBA_BIT|TRIDENT_FOG_BIT|TRIDENT_SPEC_BIT|TRIDENT_TEX0_BIT|TRIDENT_PTEX_BIT) -#define TAG(x) x##_wgfspt0 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_TEX0_BIT) -#define TAG(x) x##_t0 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_TEX0_BIT|TRIDENT_TEX1_BIT) -#define TAG(x) x##_t0t1 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_FOG_BIT) -#define TAG(x) x##_f -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_FOG_BIT|TRIDENT_TEX0_BIT) -#define TAG(x) x##_ft0 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_FOG_BIT|TRIDENT_TEX0_BIT|TRIDENT_TEX1_BIT) -#define TAG(x) x##_ft0t1 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_RGBA_BIT) -#define TAG(x) x##_g -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_RGBA_BIT|TRIDENT_SPEC_BIT) -#define TAG(x) x##_gs -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_RGBA_BIT|TRIDENT_TEX0_BIT) -#define TAG(x) x##_gt0 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_RGBA_BIT|TRIDENT_TEX0_BIT|TRIDENT_TEX1_BIT) -#define TAG(x) x##_gt0t1 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_RGBA_BIT|TRIDENT_SPEC_BIT|TRIDENT_TEX0_BIT) -#define TAG(x) x##_gst0 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_RGBA_BIT|TRIDENT_SPEC_BIT|TRIDENT_TEX0_BIT|TRIDENT_TEX1_BIT) -#define TAG(x) x##_gst0t1 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_RGBA_BIT|TRIDENT_FOG_BIT) -#define TAG(x) x##_gf -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_RGBA_BIT|TRIDENT_FOG_BIT|TRIDENT_SPEC_BIT) -#define TAG(x) x##_gfs -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_RGBA_BIT|TRIDENT_FOG_BIT|TRIDENT_TEX0_BIT) -#define TAG(x) x##_gft0 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_RGBA_BIT|TRIDENT_FOG_BIT|TRIDENT_TEX0_BIT|TRIDENT_TEX1_BIT) -#define TAG(x) x##_gft0t1 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_RGBA_BIT|TRIDENT_FOG_BIT|TRIDENT_SPEC_BIT|TRIDENT_TEX0_BIT) -#define TAG(x) x##_gfst0 -#include "tnl_dd/t_dd_vbtmp.h" - -#define IND (TRIDENT_RGBA_BIT|TRIDENT_FOG_BIT|TRIDENT_SPEC_BIT|TRIDENT_TEX0_BIT|TRIDENT_TEX1_BIT) -#define TAG(x) x##_gfst0t1 -#include "tnl_dd/t_dd_vbtmp.h" - -static void init_setup_tab( void ) -{ - init_wg(); - init_wgs(); - init_wgt0(); - init_wgt0t1(); - init_wgpt0(); - init_wgst0(); - init_wgst0t1(); - init_wgspt0(); - init_wgf(); - init_wgfs(); - init_wgft0(); - init_wgft0t1(); - init_wgfpt0(); - init_wgfst0(); - init_wgfst0t1(); - init_wgfspt0(); - init_t0(); - init_t0t1(); - init_f(); - init_ft0(); - init_ft0t1(); - init_g(); - init_gs(); - init_gt0(); - init_gt0t1(); - init_gst0(); - init_gst0t1(); - init_gf(); - init_gfs(); - init_gft0(); - init_gft0t1(); - init_gfst0(); - init_gfst0t1(); -} - -void tridentBuildVertices( GLcontext *ctx, - GLuint start, - GLuint count, - GLuint newinputs ) -{ - tridentContextPtr tmesa = TRIDENT_CONTEXT( ctx ); - GLubyte *v = ((GLubyte *)tmesa->verts + (start<<tmesa->vertex_stride_shift)); - GLuint stride = 1<<tmesa->vertex_stride_shift; - - newinputs |= tmesa->SetupNewInputs; - tmesa->SetupNewInputs = 0; - - if (!newinputs) - return; - - if (newinputs & VERT_BIT_POS) { - setup_tab[tmesa->SetupIndex].emit( ctx, start, count, v, stride ); - } else { - GLuint ind = 0; - - if (newinputs & VERT_BIT_COLOR0) - ind |= TRIDENT_RGBA_BIT; - - if (newinputs & VERT_BIT_COLOR1) - ind |= TRIDENT_SPEC_BIT; - - if (newinputs & VERT_BIT_TEX0) - ind |= TRIDENT_TEX0_BIT; - - if (newinputs & VERT_BIT_TEX1) - ind |= TRIDENT_TEX1_BIT; - - if (newinputs & VERT_BIT_FOG) - ind |= TRIDENT_FOG_BIT; - - if (tmesa->SetupIndex & TRIDENT_PTEX_BIT) - ind = ~0; - - ind &= tmesa->SetupIndex; - - if (ind) { - setup_tab[ind].emit( ctx, start, count, v, stride ); - } - } -} - -void tridentCheckTexSizes( GLcontext *ctx ) -{ - tridentContextPtr tmesa = TRIDENT_CONTEXT( ctx ); - - if (!setup_tab[tmesa->SetupIndex].check_tex_sizes(ctx)) { - TNLcontext *tnl = TNL_CONTEXT(ctx); - - /* Invalidate stored verts - */ - tmesa->SetupNewInputs = ~0; - tmesa->SetupIndex |= TRIDENT_PTEX_BIT; - - if (!tmesa->Fallback && - !(ctx->_TriangleCaps & (DD_TRI_LIGHT_TWOSIDE|DD_TRI_UNFILLED))) { - tnl->Driver.Render.Interp = setup_tab[tmesa->SetupIndex].interp; - tnl->Driver.Render.CopyPV = setup_tab[tmesa->SetupIndex].copy_pv; - } - } -} - -void tridentChooseVertexState( GLcontext *ctx ) -{ - tridentContextPtr tmesa = TRIDENT_CONTEXT( ctx ); - TNLcontext *tnl = TNL_CONTEXT(ctx); - GLuint ind = TRIDENT_XYZW_BIT|TRIDENT_RGBA_BIT; - - if (ctx->_TriangleCaps & DD_SEPARATE_SPECULAR) - ind |= TRIDENT_SPEC_BIT; - - if (ctx->Fog.Enabled) - ind |= TRIDENT_FOG_BIT; - - if (ctx->Texture.Unit[0]._ReallyEnabled) { - ind |= TRIDENT_TEX0_BIT; - if (ctx->Texture.Unit[1]._ReallyEnabled) { - ind |= TRIDENT_TEX1_BIT; - } - } - - tmesa->SetupIndex = ind; - - if (ctx->_TriangleCaps & (DD_TRI_LIGHT_TWOSIDE|DD_TRI_UNFILLED)) { - tnl->Driver.Render.Interp = trident_interp_extras; - tnl->Driver.Render.CopyPV = trident_copy_pv_extras; - } else { - tnl->Driver.Render.Interp = setup_tab[ind].interp; - tnl->Driver.Render.CopyPV = setup_tab[ind].copy_pv; - } - - if (setup_tab[ind].vertex_format != tmesa->vertex_format) { - tmesa->vertex_format = setup_tab[ind].vertex_format; - tmesa->vertex_size = setup_tab[ind].vertex_size; - tmesa->vertex_stride_shift = setup_tab[ind].vertex_stride_shift; - } -} - -void tridentInitVB( GLcontext *ctx ) -{ - tridentContextPtr tmesa = TRIDENT_CONTEXT(ctx); - GLuint size = TNL_CONTEXT(ctx)->vb.Size; - - tmesa->verts = (GLubyte *)ALIGN_MALLOC( size * 16 * 4, 32 ); - - { - static int firsttime = 1; - if (firsttime) { - init_setup_tab(); - firsttime = 0; - } - } -} - -void tridentFreeVB( GLcontext *ctx ) -{ - tridentContextPtr tmesa = TRIDENT_CONTEXT(ctx); - - if (tmesa->verts) { - ALIGN_FREE(tmesa->verts); - tmesa->verts = 0; - } - - if (tmesa->UbyteSecondaryColor.Ptr) { - ALIGN_FREE((void *)tmesa->UbyteSecondaryColor.Ptr); - tmesa->UbyteSecondaryColor.Ptr = 0; - } - - if (tmesa->UbyteColor.Ptr) { - ALIGN_FREE((void *)tmesa->UbyteColor.Ptr); - tmesa->UbyteColor.Ptr = 0; - } -} diff --git a/src/mesa/drivers/dri/unichrome/Makefile b/src/mesa/drivers/dri/unichrome/Makefile index 5fe00c1bd1f..344d34fce36 100644 --- a/src/mesa/drivers/dri/unichrome/Makefile +++ b/src/mesa/drivers/dri/unichrome/Makefile @@ -29,4 +29,3 @@ ASM_SOURCES = include ../Makefile.template -symlinks: diff --git a/src/mesa/drivers/dri/unichrome/via_context.c b/src/mesa/drivers/dri/unichrome/via_context.c index 6eb19ac0790..0524becf3ed 100644 --- a/src/mesa/drivers/dri/unichrome/via_context.c +++ b/src/mesa/drivers/dri/unichrome/via_context.c @@ -32,6 +32,7 @@ #include "main/glheader.h" #include "main/context.h" +#include "main/formats.h" #include "main/matrix.h" #include "main/state.h" #include "main/simple_list.h" @@ -65,7 +66,7 @@ #define need_GL_ARB_point_parameters #define need_GL_EXT_fog_coord #define need_GL_EXT_secondary_color -#include "extension_helper.h" +#include "main/remap_helper.h" #define DRIVER_DATE "20060710" @@ -163,24 +164,28 @@ viaInitRenderbuffer(struct via_renderbuffer *vrb, GLenum format, if (format == GL_RGBA) { /* Color */ rb->_BaseFormat = GL_RGBA; + rb->Format = MESA_FORMAT_ARGB8888; rb->DataType = GL_UNSIGNED_BYTE; } else if (format == GL_DEPTH_COMPONENT16) { /* Depth */ rb->_BaseFormat = GL_DEPTH_COMPONENT; /* we always Get/Put 32-bit Z values */ + rb->Format = MESA_FORMAT_Z16; rb->DataType = GL_UNSIGNED_INT; } else if (format == GL_DEPTH_COMPONENT24) { /* Depth */ rb->_BaseFormat = GL_DEPTH_COMPONENT; /* we always Get/Put 32-bit Z values */ + rb->Format = MESA_FORMAT_Z32; rb->DataType = GL_UNSIGNED_INT; } else { /* Stencil */ ASSERT(format == GL_STENCIL_INDEX8_EXT); rb->_BaseFormat = GL_STENCIL_INDEX; + rb->Format = MESA_FORMAT_S8; rb->DataType = GL_UNSIGNED_BYTE; } @@ -362,7 +367,7 @@ void viaReAllocateBuffers(GLcontext *ctx, GLframebuffer *drawbuffer, /* Extension strings exported by the Unichrome driver. */ -const struct dri_extension card_extensions[] = +static const struct dri_extension card_extensions[] = { { "GL_ARB_multitexture", NULL }, { "GL_ARB_point_parameters", GL_ARB_point_parameters_functions }, diff --git a/src/mesa/drivers/dri/unichrome/via_ioctl.c b/src/mesa/drivers/dri/unichrome/via_ioctl.c index 6746f552ae6..b34c133600a 100644 --- a/src/mesa/drivers/dri/unichrome/via_ioctl.c +++ b/src/mesa/drivers/dri/unichrome/via_ioctl.c @@ -886,8 +886,6 @@ void viaFlushDmaLocked(struct via_context *vmesa, GLuint flags) else if (vmesa->numClipRects) { drm_clip_rect_t *pbox = vmesa->pClipRects; __DRIdrawablePrivate *dPriv = vmesa->driDrawable; - struct via_renderbuffer *const vrb = - (struct via_renderbuffer *) dPriv->driverPrivate; for (i = 0; i < vmesa->numClipRects; i++) { drm_clip_rect_t b; diff --git a/src/mesa/drivers/dri/unichrome/via_screen.c b/src/mesa/drivers/dri/unichrome/via_screen.c index 3dbb5705710..e0bf58ca9a2 100644 --- a/src/mesa/drivers/dri/unichrome/via_screen.c +++ b/src/mesa/drivers/dri/unichrome/via_screen.c @@ -62,8 +62,6 @@ DRI_CONF_BEGIN DRI_CONF_END; static const GLuint __driNConfigOptions = 3; -extern const struct dri_extension card_extensions[]; - static drmBufMapPtr via_create_empty_buffers(void) { drmBufMapPtr retval; @@ -210,7 +208,9 @@ viaCreateBuffer(__DRIscreenPrivate *driScrnPriv, const __GLcontextModes *mesaVis, GLboolean isPixmap) { +#if 0 viaScreenPrivate *screen = (viaScreenPrivate *) driScrnPriv->private; +#endif GLboolean swStencil = (mesaVis->stencilBits > 0 && mesaVis->depthBits != 24); @@ -248,7 +248,7 @@ viaCreateBuffer(__DRIscreenPrivate *driScrnPriv, /* XXX check/fix the offset/pitch parameters! */ { driRenderbuffer *frontRb - = driNewRenderbuffer(GL_RGBA, NULL, + = driNewRenderbuffer(MESA_FORMAT_ARGB8888, NULL, screen->bytesPerPixel, 0, screen->width, driDrawPriv); viaSetSpanFunctions(frontRb, mesaVis); @@ -257,7 +257,7 @@ viaCreateBuffer(__DRIscreenPrivate *driScrnPriv, if (mesaVis->doubleBufferMode) { driRenderbuffer *backRb - = driNewRenderbuffer(GL_RGBA, NULL, + = driNewRenderbuffer(MESA_FORMAT_ARGB8888, NULL, screen->bytesPerPixel, 0, screen->width, driDrawPriv); viaSetSpanFunctions(backRb, mesaVis); @@ -266,7 +266,7 @@ viaCreateBuffer(__DRIscreenPrivate *driScrnPriv, if (mesaVis->depthBits == 16) { driRenderbuffer *depthRb - = driNewRenderbuffer(GL_DEPTH_COMPONENT16, NULL, + = driNewRenderbuffer(MESA_FORMAT_Z16, NULL, screen->bytesPerPixel, 0, screen->width, driDrawPriv); viaSetSpanFunctions(depthRb, mesaVis); @@ -274,7 +274,7 @@ viaCreateBuffer(__DRIscreenPrivate *driScrnPriv, } else if (mesaVis->depthBits == 24) { driRenderbuffer *depthRb - = driNewRenderbuffer(GL_DEPTH_COMPONENT24, NULL, + = driNewRenderbuffer(MESA_FORMAT_Z24_S8, NULL, screen->bytesPerPixel, 0, screen->width, driDrawPriv); viaSetSpanFunctions(depthRb, mesaVis); @@ -282,7 +282,7 @@ viaCreateBuffer(__DRIscreenPrivate *driScrnPriv, } else if (mesaVis->depthBits == 32) { driRenderbuffer *depthRb - = driNewRenderbuffer(GL_DEPTH_COMPONENT32, NULL, + = driNewRenderbuffer(MESA_FORMAT_Z32, NULL, screen->bytesPerPixel, 0, screen->width, driDrawPriv); viaSetSpanFunctions(depthRb, mesaVis); @@ -291,7 +291,7 @@ viaCreateBuffer(__DRIscreenPrivate *driScrnPriv, if (mesaVis->stencilBits > 0 && !swStencil) { driRenderbuffer *stencilRb - = driNewRenderbuffer(GL_STENCIL_INDEX8_EXT, NULL, + = driNewRenderbuffer(MESA_FORMAT_S8, NULL, screen->bytesPerPixel, 0, screen->width, driDrawPriv); viaSetSpanFunctions(stencilRb, mesaVis); @@ -393,18 +393,6 @@ viaInitScreen(__DRIscreenPrivate *psp) &psp->drm_version, & drm_expected) ) return NULL; - /* Calling driInitExtensions here, with a NULL context pointer, - * does not actually enable the extensions. It just makes sure - * that all the dispatch offsets for all the extensions that - * *might* be enables are known. This is needed because the - * dispatch offsets need to be known when _mesa_context_create is - * called, but we can't enable the extensions until we have a - * context pointer. - * - * Hello chicken. Hello egg. How are you two today? - */ - driInitExtensions( NULL, card_extensions, GL_FALSE ); - if (!viaInitDriver(psp)) return NULL; diff --git a/src/mesa/drivers/dri/unichrome/via_span.c b/src/mesa/drivers/dri/unichrome/via_span.c index b908f0fb23a..e847164cd02 100644 --- a/src/mesa/drivers/dri/unichrome/via_span.c +++ b/src/mesa/drivers/dri/unichrome/via_span.c @@ -23,6 +23,7 @@ */ #include "main/glheader.h" +#include "main/formats.h" #include "main/macros.h" #include "main/mtypes.h" #include "main/colormac.h" @@ -177,24 +178,22 @@ void viaInitSpanFuncs(GLcontext *ctx) void viaSetSpanFunctions(struct via_renderbuffer *vrb, const GLvisual *vis) { - if (vrb->Base.InternalFormat == GL_RGBA) { - if (vis->redBits == 5 && vis->greenBits == 6 && vis->blueBits == 5) { - viaInitPointers_565(&vrb->Base); - } - else { - viaInitPointers_8888(&vrb->Base); - } + if (vrb->Base.Format == MESA_FORMAT_RGB565) { + viaInitPointers_565(&vrb->Base); } - else if (vrb->Base.InternalFormat == GL_DEPTH_COMPONENT16) { + else if (vrb->Base.Format == MESA_FORMAT_ARGB8888) { + viaInitPointers_8888(&vrb->Base); + } + else if (vrb->Base.Format == MESA_FORMAT_Z16) { viaInitDepthPointers_z16(&vrb->Base); } - else if (vrb->Base.InternalFormat == GL_DEPTH_COMPONENT24) { + else if (vrb->Base.Format == MESA_FORMAT_Z24_S8) { viaInitDepthPointers_z24_s8(&vrb->Base); } - else if (vrb->Base.InternalFormat == GL_DEPTH_COMPONENT32) { + else if (vrb->Base.Format == MESA_FORMAT_Z32) { viaInitDepthPointers_z32(&vrb->Base); } - else if (vrb->Base.InternalFormat == GL_STENCIL_INDEX8_EXT) { + else if (vrb->Base.Format == MESA_FORMAT_S8) { viaInitStencilPointers_z24_s8(&vrb->Base); } } diff --git a/src/mesa/drivers/dri/unichrome/via_tex.c b/src/mesa/drivers/dri/unichrome/via_tex.c index 54073e76918..24924d26138 100644 --- a/src/mesa/drivers/dri/unichrome/via_tex.c +++ b/src/mesa/drivers/dri/unichrome/via_tex.c @@ -30,17 +30,17 @@ #include "main/macros.h" #include "main/mtypes.h" #include "main/enums.h" +#include "main/formats.h" #include "main/colortab.h" #include "main/convolve.h" #include "main/context.h" #include "main/mipmap.h" +#include "main/mm.h" #include "main/simple_list.h" #include "main/texcompress.h" -#include "main/texformat.h" #include "main/texobj.h" #include "main/texstore.h" -#include "main/mm.h" #include "via_context.h" #include "via_fb.h" #include "via_tex.h" @@ -48,7 +48,7 @@ #include "via_ioctl.h" #include "via_3d_reg.h" -static const struct gl_texture_format * +static gl_format viaChooseTexFormat( GLcontext *ctx, GLint internalFormat, GLenum format, GLenum type ) { @@ -65,56 +65,56 @@ viaChooseTexFormat( GLcontext *ctx, GLint internalFormat, if ( format == GL_BGRA ) { if ( type == GL_UNSIGNED_INT_8_8_8_8_REV || type == GL_UNSIGNED_BYTE ) { - return &_mesa_texformat_argb8888; + return MESA_FORMAT_ARGB8888; } else if ( type == GL_UNSIGNED_SHORT_4_4_4_4_REV ) { - return &_mesa_texformat_argb4444; + return MESA_FORMAT_ARGB4444; } else if ( type == GL_UNSIGNED_SHORT_1_5_5_5_REV ) { - return &_mesa_texformat_argb1555; + return MESA_FORMAT_ARGB1555; } } else if ( type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV || type == GL_UNSIGNED_INT_8_8_8_8 ) { - return &_mesa_texformat_argb8888; + return MESA_FORMAT_ARGB8888; } - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + return do32bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_ARGB4444; case 3: case GL_RGB: case GL_COMPRESSED_RGB: if ( format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5 ) { - return &_mesa_texformat_rgb565; + return MESA_FORMAT_RGB565; } else if ( type == GL_UNSIGNED_BYTE ) { - return &_mesa_texformat_argb8888; + return MESA_FORMAT_ARGB8888; } - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565; + return do32bpt ? MESA_FORMAT_ARGB8888 : MESA_FORMAT_RGB565; case GL_RGBA8: case GL_RGB10_A2: case GL_RGBA12: case GL_RGBA16: - return &_mesa_texformat_argb8888; + return MESA_FORMAT_ARGB8888; case GL_RGBA4: case GL_RGBA2: - return &_mesa_texformat_argb4444; + return MESA_FORMAT_ARGB4444; case GL_RGB5_A1: - return &_mesa_texformat_argb1555; + return MESA_FORMAT_ARGB1555; case GL_RGB8: case GL_RGB10: case GL_RGB12: case GL_RGB16: - return &_mesa_texformat_argb8888; + return MESA_FORMAT_ARGB8888; case GL_RGB5: case GL_RGB4: case GL_R3_G3_B2: - return &_mesa_texformat_rgb565; + return MESA_FORMAT_RGB565; case GL_ALPHA: case GL_ALPHA4: @@ -122,7 +122,7 @@ viaChooseTexFormat( GLcontext *ctx, GLint internalFormat, case GL_ALPHA12: case GL_ALPHA16: case GL_COMPRESSED_ALPHA: - return &_mesa_texformat_a8; + return MESA_FORMAT_A8; case 1: case GL_LUMINANCE: @@ -131,7 +131,7 @@ viaChooseTexFormat( GLcontext *ctx, GLint internalFormat, case GL_LUMINANCE12: case GL_LUMINANCE16: case GL_COMPRESSED_LUMINANCE: - return &_mesa_texformat_l8; + return MESA_FORMAT_L8; case 2: case GL_LUMINANCE_ALPHA: @@ -142,7 +142,7 @@ viaChooseTexFormat( GLcontext *ctx, GLint internalFormat, case GL_LUMINANCE12_ALPHA12: case GL_LUMINANCE16_ALPHA16: case GL_COMPRESSED_LUMINANCE_ALPHA: - return &_mesa_texformat_al88; + return MESA_FORMAT_AL88; case GL_INTENSITY: case GL_INTENSITY4: @@ -150,35 +150,35 @@ viaChooseTexFormat( GLcontext *ctx, GLint internalFormat, case GL_INTENSITY12: case GL_INTENSITY16: case GL_COMPRESSED_INTENSITY: - return &_mesa_texformat_i8; + return MESA_FORMAT_I8; case GL_YCBCR_MESA: if (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE) - return &_mesa_texformat_ycbcr; + return MESA_FORMAT_YCBCR; else - return &_mesa_texformat_ycbcr_rev; + return MESA_FORMAT_YCBCR_REV; case GL_COMPRESSED_RGB_FXT1_3DFX: - return &_mesa_texformat_rgb_fxt1; + return MESA_FORMAT_RGB_FXT1; case GL_COMPRESSED_RGBA_FXT1_3DFX: - return &_mesa_texformat_rgba_fxt1; + return MESA_FORMAT_RGBA_FXT1; case GL_RGB_S3TC: case GL_RGB4_S3TC: case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: - return &_mesa_texformat_rgb_dxt1; + return MESA_FORMAT_RGB_DXT1; case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: - return &_mesa_texformat_rgba_dxt1; + return MESA_FORMAT_RGBA_DXT1; case GL_RGBA_S3TC: case GL_RGBA4_S3TC: case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: - return &_mesa_texformat_rgba_dxt3; + return MESA_FORMAT_RGBA_DXT3; case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: - return &_mesa_texformat_rgba_dxt5; + return MESA_FORMAT_RGBA_DXT5; case GL_COLOR_INDEX: case GL_COLOR_INDEX1_EXT: @@ -187,16 +187,16 @@ viaChooseTexFormat( GLcontext *ctx, GLint internalFormat, case GL_COLOR_INDEX8_EXT: case GL_COLOR_INDEX12_EXT: case GL_COLOR_INDEX16_EXT: - return &_mesa_texformat_ci8; + return MESA_FORMAT_CI8; default: fprintf(stderr, "unexpected texture format %s in %s\n", _mesa_lookup_enum_by_nr(internalFormat), __FUNCTION__); - return NULL; + return MESA_FORMAT_NONE; } - return NULL; /* never get here */ + return MESA_FORMAT_NONE; /* never get here */ } static int logbase2(int n) @@ -457,7 +457,7 @@ static GLboolean viaSetTexImages(GLcontext *ctx, GLuint widthExp = 0; GLuint heightExp = 0; - switch (baseImage->image.TexFormat->MesaFormat) { + switch (baseImage->image.TexFormat) { case MESA_FORMAT_ARGB8888: texFormat = HC_HTXnFM_ARGB8888; break; @@ -689,24 +689,7 @@ static void viaTexImage(GLcontext *ctx, assert(texImage->TexFormat); - if (dims == 1) { - texImage->FetchTexelc = texImage->TexFormat->FetchTexel1D; - texImage->FetchTexelf = texImage->TexFormat->FetchTexel1Df; - } - else { - texImage->FetchTexelc = texImage->TexFormat->FetchTexel2D; - texImage->FetchTexelf = texImage->TexFormat->FetchTexel2Df; - } - texelBytes = texImage->TexFormat->TexelBytes; - - if (texelBytes == 0) { - /* compressed format */ - texImage->IsCompressed = GL_TRUE; - texImage->CompressedSize = - ctx->Driver.CompressedTextureSize(ctx, texImage->Width, - texImage->Height, texImage->Depth, - texImage->TexFormat->MesaFormat); - } + texelBytes = _mesa_get_format_bytes(texImage->TexFormat); /* Minimum pitch of 32 bytes */ if (postConvWidth * texelBytes < 32) { @@ -718,8 +701,11 @@ static void viaTexImage(GLcontext *ctx, viaImage->pitchLog2 = logbase2(postConvWidth * texelBytes); /* allocate memory */ - if (texImage->IsCompressed) - sizeInBytes = texImage->CompressedSize; + if (_mesa_is_format_compressed(texImage->TexFormat)) + sizeInBytes = _mesa_format_image_size(texImage->TexFormat, + texImage->Width, + texImage->Height, + texImage->Depth); else sizeInBytes = postConvWidth * postConvHeight * texelBytes; @@ -797,22 +783,22 @@ static void viaTexImage(GLcontext *ctx, else { GLint dstRowStride; GLboolean success; - if (texImage->IsCompressed) { - dstRowStride = _mesa_compressed_row_stride(texImage->TexFormat->MesaFormat, width); + + if (_mesa_is_format_compressed(texImage->TexFormat)) { + dstRowStride = _mesa_format_row_stride(texImage->TexFormat, width); } else { - dstRowStride = postConvWidth * texImage->TexFormat->TexelBytes; + dstRowStride = postConvWidth * _mesa_get_format_bytes(texImage->TexFormat); } - ASSERT(texImage->TexFormat->StoreImage); - success = texImage->TexFormat->StoreImage(ctx, dims, - texImage->_BaseFormat, - texImage->TexFormat, - texImage->Data, - 0, 0, 0, /* dstX/Y/Zoffset */ - dstRowStride, - texImage->ImageOffsets, - width, height, 1, - format, type, pixels, packing); + success = _mesa_texstore(ctx, dims, + texImage->_BaseFormat, + texImage->TexFormat, + texImage->Data, + 0, 0, 0, /* dstX/Y/Zoffset */ + dstRowStride, + texImage->ImageOffsets, + width, height, 1, + format, type, pixels, packing); if (!success) { _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); } diff --git a/src/mesa/drivers/dri/unichrome/via_tris.c b/src/mesa/drivers/dri/unichrome/via_tris.c index 79e67620c9e..ab457d41dc8 100644 --- a/src/mesa/drivers/dri/unichrome/via_tris.c +++ b/src/mesa/drivers/dri/unichrome/via_tris.c @@ -832,13 +832,13 @@ static GLboolean viaCheckPTexHack( GLcontext *ctx ) RENDERINPUTS_COPY( index_bitset, tnl->render_inputs_bitset ); - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 ) && VB->TexCoordPtr[0]->size == 4) { + if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 ) && VB->AttribPtr[_TNL_ATTRIB_TEX0]->size == 4) { if (!RENDERINPUTS_TEST_RANGE( index_bitset, _TNL_ATTRIB_TEX1, _TNL_LAST_TEX )) ptexHack = GL_TRUE; else fallback = GL_TRUE; } - if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX1 ) && VB->TexCoordPtr[1]->size == 4) + if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX1 ) && VB->AttribPtr[_TNL_ATTRIB_TEX1]->size == 4) fallback = GL_TRUE; FALLBACK(VIA_CONTEXT(ctx), VIA_FALLBACK_PROJ_TEXTURE, fallback); |