r200 driver, brought over by Jon Smirl

src/mesa/drivers/dri/r200/Makefile

@@ -0,0 +1,134 @@
+# $Id: Makefile,v 1.1.2.1 2003/03/10 11:23:04 keithw Exp $
+
+# Mesa 3-D graphics library
+# Version:  5.0
+# Copyright (C) 1995-2002  Brian Paul
+
+
+
+MESA = ../../..
+default: r200_dri.so 
+include $(MESA)/Makefile.include
+
+
+SHARED_INCLUDES= -I$(MESABUILDDIR) -I$(MESA)/include -I. -I../common -Iserver
+MINIGLX_INCLUDES = -I$(MESABUILDDIR)/miniglx 
+DRI_INCLUDES = -I$(MESABUILDDIR)/dri
+
+ifeq ($(FULL_DRIVER),true)
+DEFINES = \
+	-DGLX_DIRECT_RENDERING \
+-Dlinux -D__i386__ -D_POSIX_C_SOURCE=199309L -D_POSIX_SOURCE -D_XOPEN_SOURCE -D_BSD_SOURCE -D_SVID_SOURCE  -D_GNU_SOURCE   -DFUNCPROTO=15 -DNARROWPROTO -DXTHREADS  -D_REENTRANT -DXUSE_MTSAFE_API    -DMALLOC_0_RETURNS_NULL -DGLXEXT -DXF86DRI -DGLX_DIRECT_RENDERING -DGLX_USE_DLOPEN -DGLX_USE_MESA  -DX_BYTE_ORDER=X_LITTLE_ENDIAN \
+	-D_HAVE_SWRAST=1 \
+	-D_HAVE_SWTNL=1 \
+	-D_HAVE_SANITY=1 \
+	-D_HAVE_CODEGEN=1 \
+	-D_HAVE_LIGHTING=1 \
+	-D_HAVE_TEXGEN=1 \
+	-D_HAVE_USERCLIP=1 \
+	-D_HAVE_FULL_GL=1
+else
+DEFINES = \
+	-D_HAVE_SWRAST=0 \
+	-D_HAVE_SWTNL=0 \
+	-D_HAVE_SANITY=0 \
+	-D_HAVE_CODEGEN=0 \
+	-D_HAVE_LIGHTING=0 \
+	-D_HAVE_TEXGEN=0 \
+	-D_HAVE_USERCLIP=0 \
+	-D_HAVE_FULL_GL=0
+endif
+
+# The .a files for each mesa module required by this driver:
+#
+FULL_MESA = 	$(MESABUILDDIR)/swrast_setup/swrast_setup.a \
+		$(MESABUILDDIR)/tnl/tnl.a \
+		$(MESABUILDDIR)/math/math.a \
+		$(MESABUILDDIR)/array_cache/array_cache.a \
+		$(MESABUILDDIR)/swrast/swrast.a \
+		$(MESABUILDDIR)/mesa.a \
+		$(MESABUILDDIR)/math/math.a #kludge
+
+SUBSET_MESA = 	$(MESABUILDDIR)/mesa.a \
+		$(MESABUILDDIR)/math/m_matrix.o
+
+
+
+MINIGLX_SOURCES = server/radeon_dri.c 
+
+DRIVER_SOURCES = r200_context.c \
+		 r200_ioctl.c \
+		 r200_lock.c \
+		 r200_screen.c \
+		 r200_state.c \
+		 r200_state_init.c \
+		 ../common/mm.c 
+
+SUBSET_DRIVER_SOURCES = \
+		r200_subset_bitmap.c \
+		r200_subset_readpix.c \
+		r200_subset_select.c \
+		r200_subset_tex.c \
+		r200_subset_vtx.c 
+
+FULL_DRIVER_SOURCES = 	\
+		 r200_cmdbuf.c \
+		 r200_pixel.c \
+		 r200_tex.c \
+		 r200_texmem.c \
+		 r200_texstate.c \
+		 r200_tcl.c \
+		 r200_swtcl.c \
+		 r200_span.c \
+		 r200_maos.c \
+		 r200_sanity.c \
+		 r200_vtxfmt.c \
+		 r200_vtxfmt_c.c \
+		 r200_vtxfmt_sse.c \
+		 r200_vtxfmt_x86.c 
+
+
+INCLUDES = $(MINIGLX_INCLUDES) \
+	   $(SHARED_INCLUDES)
+
+
+ifeq ($(FULL_DRIVER),true)
+C_SOURCES = $(DRIVER_SOURCES) \
+	    $(FULL_DRIVER_SOURCES) \
+	    $(MINIGLX_SOURCES) 
+MESA_MODULES = $(FULL_MESA)
+else
+C_SOURCES = $(DRIVER_SOURCES) \
+	    $(SUBSET_DRIVER_SOURCES) \
+	    $(MINIGLX_SOURCES) 
+MESA_MODULES = $(SUBSET_MESA)
+endif
+
+
+ifeq ($(WINDOW_SYSTEM),dri)
+WINOBJ=$(MESABUILDDIR)/dri/dri.a
+WINLIB=
+else
+WINOBJ=
+WINLIB=-L$(MESA)/src/miniglx
+endif
+
+ASM_SOURCES = 
+OBJECTS = $(C_SOURCES:.c=.o) \
+	  $(ASM_SOURCES:.S=.o) 
+
+
+
+##### TARGETS #####
+
+
+# Build the subset or full driver?
+#
+r200_dri.so:  $(OBJECTS) $(MESA_MODULES) $(WINOBJ) Makefile
+	rm -f $@ && gcc -o $@ -shared $(OBJECTS) $(MESA_MODULES) $(WINOBJ) $(WINLIB) -lGL -lc -lm -Wl,-rpath,../../../lib
+	rm -f $(MESA)/lib/r200_dri.so && \
+	install r200_dri.so $(MESA)/lib/r200_dri.so
+
+##### DEPENDENCIES #####
+
+-include $(C_SOURCES:.c=.d)

src/mesa/drivers/dri/r200/r200_cmdbuf.c

@@ -0,0 +1,344 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#include "glheader.h"
+#include "imports.h"
+#include "macros.h"
+#include "context.h"
+#include "swrast/swrast.h"
+#include "simple_list.h"
+
+#include "r200_context.h"
+#include "r200_state.h"
+#include "r200_ioctl.h"
+#include "r200_tcl.h"
+#include "r200_sanity.h"
+#include "radeon_reg.h"
+
+static void print_state_atom( struct r200_state_atom *state )
+{
+   int i;
+
+   fprintf(stderr, "emit %s/%d\n", state->name, state->cmd_size);
+
+   if (0 & R200_DEBUG & DEBUG_VERBOSE) 
+      for (i = 0 ; i < state->cmd_size ; i++) 
+	 fprintf(stderr, "\t%s[%d]: %x\n", state->name, i, state->cmd[i]);
+
+}
+
+static void r200_emit_state_list( r200ContextPtr rmesa, 
+				    struct r200_state_atom *list )
+{
+   struct r200_state_atom *state, *tmp;
+   char *dest;
+
+   foreach_s( state, tmp, list ) {
+      if (state->check( rmesa->glCtx, state->idx )) {
+	 dest = r200AllocCmdBuf( rmesa, state->cmd_size * 4, __FUNCTION__);
+	 memcpy( dest, state->cmd, state->cmd_size * 4);
+	 move_to_head( &(rmesa->hw.clean), state );
+	 if (R200_DEBUG & DEBUG_STATE) 
+	    print_state_atom( state );
+      }
+      else if (R200_DEBUG & DEBUG_STATE)
+	 fprintf(stderr, "skip state %s\n", state->name);
+   }
+}
+
+
+void r200EmitState( r200ContextPtr rmesa )
+{
+   struct r200_state_atom *state, *tmp;
+
+   if (R200_DEBUG & (DEBUG_STATE|DEBUG_PRIMS))
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+   /* Somewhat overkill:
+    */
+   if ( rmesa->lost_context) {
+      if (R200_DEBUG & (DEBUG_STATE|DEBUG_PRIMS|DEBUG_IOCTL))
+	 fprintf(stderr, "%s - lost context\n", __FUNCTION__); 
+
+      foreach_s( state, tmp, &(rmesa->hw.clean) ) 
+	 move_to_tail(&(rmesa->hw.dirty), state );
+
+      rmesa->lost_context = 0;
+   }
+   else {
+      move_to_tail( &rmesa->hw.dirty, &rmesa->hw.mtl[0] ); 
+      /* odd bug? -- isosurf, cycle between reflect & lit */
+   }
+
+   r200_emit_state_list( rmesa, &rmesa->hw.dirty );
+}
+
+
+
+/* Fire a section of the retained (indexed_verts) buffer as a regular
+ * primtive.  
+ */
+extern void r200EmitVbufPrim( r200ContextPtr rmesa,
+				GLuint primitive,
+				GLuint vertex_nr )
+{
+   drmRadeonCmdHeader *cmd;
+
+   assert(!(primitive & R200_VF_PRIM_WALK_IND));
+   
+   r200EmitState( rmesa );
+   
+   if (R200_DEBUG & (DEBUG_IOCTL|DEBUG_PRIMS))
+      fprintf(stderr, "%s cmd_used/4: %d prim %x nr %d\n", __FUNCTION__,
+	      rmesa->store.cmd_used/4, primitive, vertex_nr);
+   
+   cmd = (drmRadeonCmdHeader *)r200AllocCmdBuf( rmesa, 3 * sizeof(*cmd),
+						  __FUNCTION__ );
+   cmd[0].i = 0;
+   cmd[0].header.cmd_type = RADEON_CMD_PACKET3_CLIP;
+   cmd[1].i = R200_CP_CMD_3D_DRAW_VBUF_2;
+   cmd[2].i = (primitive | 
+	       R200_VF_PRIM_WALK_LIST |
+	       R200_VF_COLOR_ORDER_RGBA |
+	       (vertex_nr << R200_VF_VERTEX_NUMBER_SHIFT));
+
+
+   if (R200_DEBUG & DEBUG_SYNC) {
+      fprintf(stderr, "\nSyncing\n\n");
+      R200_FIREVERTICES( rmesa );
+      r200Finish( rmesa->glCtx );
+   }
+}
+
+
+void r200FlushElts( r200ContextPtr rmesa )
+{
+   int *cmd = (int *)(rmesa->store.cmd_buf + rmesa->store.elts_start);
+   int dwords;
+   int nr = (rmesa->store.cmd_used - (rmesa->store.elts_start + 12)) / 2;
+
+   if (R200_DEBUG & (DEBUG_IOCTL|DEBUG_PRIMS))
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+   assert( rmesa->dma.flush == r200FlushElts );
+   rmesa->dma.flush = 0;
+
+   /* Cope with odd number of elts:
+    */
+   rmesa->store.cmd_used = (rmesa->store.cmd_used + 2) & ~2;
+   dwords = (rmesa->store.cmd_used - rmesa->store.elts_start) / 4;
+
+   cmd[1] |= (dwords - 3) << 16;
+   cmd[2] |= nr << R200_VF_VERTEX_NUMBER_SHIFT;
+
+   if (R200_DEBUG & DEBUG_SYNC) {
+      fprintf(stderr, "\nSyncing in %s\n\n", __FUNCTION__);
+      R200_FIREVERTICES( rmesa );
+      r200Finish( rmesa->glCtx );
+   }
+}
+
+
+GLushort *r200AllocEltsOpenEnded( r200ContextPtr rmesa,
+				    GLuint primitive,
+				    GLuint min_nr )
+{
+   drmRadeonCmdHeader *cmd;
+   GLushort *retval;
+
+   if (R200_DEBUG & DEBUG_IOCTL)
+      fprintf(stderr, "%s %d prim %x\n", __FUNCTION__, min_nr, primitive);
+
+   assert((primitive & R200_VF_PRIM_WALK_IND));
+   
+   r200EmitState( rmesa );
+   
+   cmd = (drmRadeonCmdHeader *)r200AllocCmdBuf( rmesa, 
+						12 + min_nr*2,
+						__FUNCTION__ );
+   cmd[0].i = 0;
+   cmd[0].header.cmd_type = RADEON_CMD_PACKET3_CLIP;
+   cmd[1].i = R200_CP_CMD_3D_DRAW_INDX_2;
+   cmd[2].i = (primitive | 
+	       R200_VF_PRIM_WALK_IND |
+	       R200_VF_COLOR_ORDER_RGBA);
+
+   
+   retval = (GLushort *)(cmd+3);
+
+   if (R200_DEBUG & DEBUG_PRIMS)
+      fprintf(stderr, "%s: header 0x%x prim %x \n",
+	      __FUNCTION__,
+	      cmd[1].i, primitive);
+
+   assert(!rmesa->dma.flush);
+   rmesa->dma.flush = r200FlushElts;
+
+   rmesa->store.elts_start = ((char *)cmd) - rmesa->store.cmd_buf;
+
+   return retval;
+}
+
+
+
+void r200EmitVertexAOS( r200ContextPtr rmesa,
+			  GLuint vertex_size,
+			  GLuint offset )
+{
+   drmRadeonCmdHeader *cmd;
+
+   if (R200_DEBUG & (DEBUG_PRIMS|DEBUG_IOCTL))
+      fprintf(stderr, "%s:  vertex_size 0x%x offset 0x%x \n",
+	      __FUNCTION__, vertex_size, offset);
+
+   cmd = (drmRadeonCmdHeader *)r200AllocCmdBuf( rmesa, 5 * sizeof(int),
+						  __FUNCTION__ );
+
+   cmd[0].header.cmd_type = RADEON_CMD_PACKET3;
+   cmd[1].i = R200_CP_CMD_3D_LOAD_VBPNTR | (2 << 16);
+   cmd[2].i = 1;
+   cmd[3].i = vertex_size | (vertex_size << 8);
+   cmd[4].i = offset;
+}
+		       
+
+void r200EmitAOS( r200ContextPtr rmesa,
+		    struct r200_dma_region **component,
+		    GLuint nr,
+		    GLuint offset )
+{
+   drmRadeonCmdHeader *cmd;
+   int sz = 3 + ((nr/2)*3) + ((nr&1)*2);
+   int i;
+   int *tmp;
+
+   if (R200_DEBUG & DEBUG_IOCTL)
+      fprintf(stderr, "%s nr arrays: %d\n", __FUNCTION__, nr);
+
+   cmd = (drmRadeonCmdHeader *)r200AllocCmdBuf( rmesa, sz * sizeof(int),
+						  __FUNCTION__ );
+   cmd[0].i = 0;
+   cmd[0].header.cmd_type = RADEON_CMD_PACKET3;
+   cmd[1].i = R200_CP_CMD_3D_LOAD_VBPNTR | ((sz-3) << 16);
+   cmd[2].i = nr;
+   tmp = &cmd[0].i;
+   cmd += 3;
+
+   for (i = 0 ; i < nr ; i++) {
+      if (i & 1) {
+	 cmd[0].i |= ((component[i]->aos_stride << 24) | 
+		      (component[i]->aos_size << 16));
+	 cmd[2].i = (component[i]->aos_start + 
+		     offset * component[i]->aos_stride * 4);
+	 cmd += 3;
+      }
+      else {
+	 cmd[0].i = ((component[i]->aos_stride << 8) | 
+		     (component[i]->aos_size << 0));
+	 cmd[1].i = (component[i]->aos_start + 
+		     offset * component[i]->aos_stride * 4);
+      }
+   }
+
+   if (R200_DEBUG & DEBUG_VERTS) {
+      fprintf(stderr, "%s:\n", __FUNCTION__);
+      for (i = 0 ; i < sz ; i++)
+	 fprintf(stderr, "   %d: %x\n", i, tmp[i]);
+   }
+}
+
+void r200EmitBlit( r200ContextPtr rmesa,
+		   GLuint color_fmt,
+		   GLuint src_pitch,
+		   GLuint src_offset,
+		   GLuint dst_pitch,
+		   GLuint dst_offset,
+		   GLint srcx, GLint srcy,
+		   GLint dstx, GLint dsty,
+		   GLuint w, GLuint h )
+{
+   drmRadeonCmdHeader *cmd;
+
+   if (R200_DEBUG & DEBUG_IOCTL)
+      fprintf(stderr, "%s src %x/%x %d,%d dst: %x/%x %d,%d sz: %dx%d\n",
+	      __FUNCTION__, 
+	      src_pitch, src_offset, srcx, srcy,
+	      dst_pitch, dst_offset, dstx, dsty,
+	      w, h);
+
+   assert( (src_pitch & 63) == 0 );
+   assert( (dst_pitch & 63) == 0 );
+   assert( (src_offset & 1023) == 0 );
+   assert( (dst_offset & 1023) == 0 );
+   assert( w < (1<<16) );
+   assert( h < (1<<16) );
+
+   cmd = (drmRadeonCmdHeader *)r200AllocCmdBuf( rmesa, 8 * sizeof(int),
+						  __FUNCTION__ );
+
+
+   cmd[0].header.cmd_type = RADEON_CMD_PACKET3;
+   cmd[1].i = R200_CP_CMD_BITBLT_MULTI | (5 << 16);
+   cmd[2].i = (RADEON_GMC_SRC_PITCH_OFFSET_CNTL |
+	       RADEON_GMC_DST_PITCH_OFFSET_CNTL |
+	       RADEON_GMC_BRUSH_NONE |
+	       (color_fmt << 8) |
+	       RADEON_GMC_SRC_DATATYPE_COLOR |
+	       RADEON_ROP3_S |
+	       RADEON_DP_SRC_SOURCE_MEMORY |
+	       RADEON_GMC_CLR_CMP_CNTL_DIS |
+	       RADEON_GMC_WR_MSK_DIS );
+
+   cmd[3].i = ((src_pitch/64)<<22) | (src_offset >> 10);
+   cmd[4].i = ((dst_pitch/64)<<22) | (dst_offset >> 10);
+   cmd[5].i = (srcx << 16) | srcy;
+   cmd[6].i = (dstx << 16) | dsty; /* dst */
+   cmd[7].i = (w << 16) | h;
+}
+
+
+void r200EmitWait( r200ContextPtr rmesa, GLuint flags )
+{
+   if (rmesa->dri.drmMinor >= 6) {
+      drmRadeonCmdHeader *cmd;
+
+      assert( !(flags & ~(RADEON_WAIT_2D|RADEON_WAIT_3D)) );
+      
+      cmd = (drmRadeonCmdHeader *)r200AllocCmdBuf( rmesa, 1 * sizeof(int),
+						   __FUNCTION__ );
+      cmd[0].i = 0;
+      cmd[0].wait.cmd_type = RADEON_CMD_WAIT;
+      cmd[0].wait.flags = flags;
+   }
+}

src/mesa/drivers/dri/r200/r200_context.c

@@ -0,0 +1,673 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#include "glheader.h"
+#include "api_arrayelt.h"
+#include "context.h"
+#include "simple_list.h"
+#include "imports.h"
+#include "matrix.h"
+#include "state.h"
+#include "extensions.h"
+#include "state.h"
+
+#include "swrast/swrast.h"
+#include "swrast_setup/swrast_setup.h"
+#include "array_cache/acache.h"
+
+#include "tnl/tnl.h"
+#include "tnl/t_pipeline.h"
+
+#include "r200_context.h"
+#include "r200_ioctl.h"
+#include "r200_state.h"
+#include "r200_span.h"
+#include "r200_pixel.h"
+#include "r200_tex.h"
+#include "r200_swtcl.h"
+#include "r200_tcl.h"
+#include "r200_vtxfmt.h"
+#include "r200_maos.h"
+
+#if defined(USE_X86_ASM)
+#include "X86/common_x86_asm.h"
+#endif
+
+#define R200_DATE	"20021125"
+
+#ifndef R200_DEBUG
+int R200_DEBUG = (0);
+#endif
+
+
+
+/* Return the width and height of the given buffer.
+ */
+static void r200GetBufferSize( GLframebuffer *buffer,
+			       GLuint *width, GLuint *height )
+{
+   GET_CURRENT_CONTEXT(ctx);
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+
+   LOCK_HARDWARE( rmesa );
+   *width  = rmesa->dri.drawable->w;
+   *height = rmesa->dri.drawable->h;
+   UNLOCK_HARDWARE( rmesa );
+}
+
+/* Return various strings for glGetString().
+ */
+static const GLubyte *r200GetString( GLcontext *ctx, GLenum name )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   static char buffer[128];
+
+   switch ( name ) {
+   case GL_VENDOR:
+      return (GLubyte *)"Tungsten Graphics, Inc.";
+
+   case GL_RENDERER:
+      sprintf( buffer, "Mesa DRI R200 " R200_DATE);
+
+      /* Append any chipset-specific information.  None yet.
+       */
+
+      /* Append any AGP-specific information.
+       */
+      switch ( rmesa->r200Screen->AGPMode ) {
+      case 1:
+	 strncat( buffer, " AGP 1x", 7 );
+	 break;
+      case 2:
+	 strncat( buffer, " AGP 2x", 7 );
+	 break;
+      case 4:
+	 strncat( buffer, " AGP 4x", 7 );
+	 break;
+      }
+
+      /* 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
+
+      if ( !(rmesa->TclFallback & R200_TCL_FALLBACK_TCL_DISABLE) ) {
+	 strncat( buffer, " TCL", 4 );
+      }
+      else {
+	 strncat( buffer, " NO-TCL", 7 );
+      }
+
+      return (GLubyte *)buffer;
+
+   default:
+      return NULL;
+   }
+}
+
+
+
+/* Initialize the extensions supported by this driver.
+ */
+static void r200InitExtensions( GLcontext *ctx, r200ScreenPtr r200Screen )
+{
+   _mesa_enable_imaging_extensions( ctx );
+
+   _mesa_enable_extension( ctx, "GL_ARB_multitexture" );
+   if (r200Screen->drmSupportsCubeMaps)
+      _mesa_enable_extension( ctx, "GL_ARB_texture_cube_map" );
+   _mesa_enable_extension( ctx, "GL_ARB_texture_env_add" );
+   _mesa_enable_extension( ctx, "GL_ARB_texture_env_combine" );
+   _mesa_enable_extension( ctx, "GL_ARB_texture_env_dot3" );
+   _mesa_enable_extension( ctx, "GL_ARB_texture_mirrored_repeat" );
+   _mesa_enable_extension( ctx, "GL_ARB_texture_border_clamp" );
+
+   _mesa_enable_extension( ctx, "GL_ATI_texture_env_combine3" );
+   _mesa_enable_extension( ctx, "GL_ATI_texture_mirror_once" );
+
+   _mesa_enable_extension( ctx, "GL_EXT_blend_logic_op" );
+   /*_mesa_enable_extension( ctx, "GL_EXT_fog_coord" );*/
+   _mesa_enable_extension( ctx, "GL_EXT_stencil_wrap" );
+   _mesa_enable_extension( ctx, "GL_EXT_texture_edge_clamp" );
+   _mesa_enable_extension( ctx, "GL_EXT_texture_env_add" );
+   _mesa_enable_extension( ctx, "GL_EXT_texture_env_combine" );
+   _mesa_enable_extension( ctx, "GL_EXT_texture_env_dot3" );
+   _mesa_enable_extension( ctx, "GL_EXT_texture_filter_anisotropic" );
+   _mesa_enable_extension( ctx, "GL_EXT_texture_lod_bias" );
+   _mesa_enable_extension( ctx, "GL_EXT_secondary_color" );
+   _mesa_enable_extension( ctx, "GL_EXT_blend_subtract" );
+   _mesa_enable_extension( ctx, "GL_EXT_blend_minmax" );
+
+/*     _mesa_enable_extension( ctx, "GL_EXT_fog_coord" ); */
+
+   _mesa_enable_extension( ctx, "GL_MESA_pack_invert" );
+   _mesa_enable_extension( ctx, "GL_MESA_ycbcr_texture" );
+   _mesa_enable_extension( ctx, "GL_NV_texture_rectangle" );
+}
+
+extern const struct gl_pipeline_stage _r200_render_stage;
+extern const struct gl_pipeline_stage _r200_tcl_stage;
+
+static const struct gl_pipeline_stage *r200_pipeline[] = {
+
+   /* Try and go straight to t&l
+    */
+   &_r200_tcl_stage,  
+
+   /* 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,
+
+   /* Try again to go to tcl? 
+    *     - no good for asymmetric-twoside (do with multipass)
+    *     - no good for asymmetric-unfilled (do with multipass)
+    *     - good for material
+    *     - good for texgen
+    *     - need to manipulate a bit of state
+    *
+    * - worth it/not worth it?
+    */
+			
+   /* Else do them here.
+    */
+   &_r200_render_stage, 
+   &_tnl_render_stage,		/* FALLBACK:  */
+   0,
+};
+
+
+
+/* Initialize the driver's misc functions.
+ */
+static void r200InitDriverFuncs( GLcontext *ctx )
+{
+    ctx->Driver.GetBufferSize		= r200GetBufferSize;
+    ctx->Driver.ResizeBuffers           = _swrast_alloc_buffers;
+    ctx->Driver.GetString		= r200GetString;
+
+    ctx->Driver.Error			= NULL;
+    ctx->Driver.DrawPixels		= NULL;
+    ctx->Driver.Bitmap			= NULL;
+}
+
+static void add_debug_flags( const char *debug )
+{
+   if (strstr(debug, "fall")) 
+      R200_DEBUG |= DEBUG_FALLBACKS;
+
+   if (strstr(debug, "tex")) 
+      R200_DEBUG |= DEBUG_TEXTURE;
+
+   if (strstr(debug, "ioctl")) 
+      R200_DEBUG |= DEBUG_IOCTL;
+
+   if (strstr(debug, "prim")) 
+      R200_DEBUG |= DEBUG_PRIMS;
+
+   if (strstr(debug, "vert")) 
+      R200_DEBUG |= DEBUG_VERTS;
+
+   if (strstr(debug, "state")) 
+      R200_DEBUG |= DEBUG_STATE;
+
+   if (strstr(debug, "code")) 
+      R200_DEBUG |= DEBUG_CODEGEN;
+
+   if (strstr(debug, "vfmt") || strstr(debug, "vtxf")) 
+      R200_DEBUG |= DEBUG_VFMT;
+
+   if (strstr(debug, "verb")) 
+      R200_DEBUG |= DEBUG_VERBOSE;
+
+   if (strstr(debug, "dri")) 
+      R200_DEBUG |= DEBUG_DRI;
+
+   if (strstr(debug, "dma")) 
+      R200_DEBUG |= DEBUG_DMA;
+
+   if (strstr(debug, "san")) 
+      R200_DEBUG |= DEBUG_SANITY;
+
+   if (strstr(debug, "sync")) 
+      R200_DEBUG |= DEBUG_SYNC;
+
+   if (strstr(debug, "pix")) 
+      R200_DEBUG |= DEBUG_PIXEL;
+
+   if (strstr(debug, "mem")) 
+      R200_DEBUG |= DEBUG_MEMORY;
+}
+
+/* Create the device specific context.
+ */
+GLboolean r200CreateContext( const __GLcontextModes *glVisual,
+			     __DRIcontextPrivate *driContextPriv,
+			     void *sharedContextPrivate)
+{
+   __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv;
+   r200ScreenPtr r200Screen = (r200ScreenPtr)(sPriv->private);
+   r200ContextPtr rmesa;
+   GLcontext *ctx, *shareCtx;
+   int i, memPerUnit;
+
+   assert(glVisual);
+   assert(driContextPriv);
+   assert(r200Screen);
+
+   /* Allocate the R200 context */
+   rmesa = (r200ContextPtr) CALLOC( sizeof(*rmesa) );
+   if ( !rmesa )
+      return GL_FALSE;
+
+   /* Allocate the Mesa context */
+   if (sharedContextPrivate)
+      shareCtx = ((r200ContextPtr) sharedContextPrivate)->glCtx;
+   else
+      shareCtx = NULL;
+   rmesa->glCtx = _mesa_create_context(glVisual, shareCtx, (void *) rmesa, GL_TRUE);
+   if (!rmesa->glCtx) {
+      FREE(rmesa);
+      return GL_FALSE;
+   }
+   driContextPriv->driverPrivate = rmesa;
+
+   /* Init r200 context data */
+   rmesa->dri.context = driContextPriv;
+   rmesa->dri.screen = sPriv;
+   rmesa->dri.drawable = NULL; /* Set by XMesaMakeCurrent */
+   rmesa->dri.hwContext = driContextPriv->hHWContext;
+   rmesa->dri.hwLock = &sPriv->pSAREA->lock;
+   rmesa->dri.fd = sPriv->fd;
+   rmesa->dri.drmMinor = sPriv->drmMinor;
+
+   rmesa->r200Screen = r200Screen;
+   rmesa->sarea = (RADEONSAREAPrivPtr)((GLubyte *)sPriv->pSAREA +
+				       r200Screen->sarea_priv_offset);
+
+
+   rmesa->dma.buf0_address = rmesa->r200Screen->buffers->list[0].address;
+
+   for ( i = 0 ; i < r200Screen->numTexHeaps ; i++ ) {
+      make_empty_list( &rmesa->texture.objects[i] );
+      rmesa->texture.heap[i] = mmInit( 0, r200Screen->texSize[i] );
+      rmesa->texture.age[i] = -1;
+   }
+   rmesa->texture.numHeaps = r200Screen->numTexHeaps;
+   make_empty_list( &rmesa->texture.swapped );
+
+   rmesa->swtcl.RenderIndex = ~0;
+   rmesa->lost_context = 1;
+
+   /* KW: Set the maximum texture size small enough that we can
+    * guarentee that both texture units can bind a maximal texture
+    * and have them both in on-card memory at once.
+    * Test for 2 textures * 4 bytes/texel * size * size.
+    */
+   ctx = rmesa->glCtx;
+   memPerUnit = r200Screen->texSize[RADEON_CARD_HEAP] / 2;
+
+   /* XXX the following code could go into a utility file */
+   if (memPerUnit >= 4 * 2048 * 2048) {
+      ctx->Const.MaxTextureLevels = 12; /* 2048x2048 */
+   }
+   else if (memPerUnit >= 4 * 1024 * 1024) {
+      ctx->Const.MaxTextureLevels = 11; /* 1024x1024 */
+   }
+   else if (memPerUnit >= 4 * 512 * 512) {
+      ctx->Const.MaxTextureLevels = 10; /* 512x512 */
+   }
+   else {
+      ctx->Const.MaxTextureLevels = 9; /* 256x256 */
+   }
+
+#if ENABLE_HW_3D_TEXTURE
+   if (memPerUnit >= 4 * 256 * 256 * 256) {
+      ctx->Const.Max3DTextureLevels = 9; /* 256x256x256 */
+   }
+   else if (memPerUnit >= 4 * 64 * 64 * 64) {
+      ctx->Const.Max3DTextureLevels = 8; /* 128x128x128 */
+   }
+   else if (memPerUnit >= 4 * 32 * 32 * 32) {
+      ctx->Const.Max3DTextureLevels = 7; /* 64x64x64 */
+   }
+   else { /* 256KBytes */
+      ctx->Const.Max3DTextureLevels = 6; /* 32x32x32 */
+   }
+#endif
+
+   if (memPerUnit >= 4 * 6 * 2048 * 2048) {
+      ctx->Const.MaxCubeTextureLevels = 12; /* 2048x2048 */
+   }
+   if (memPerUnit >= 4 * 6 * 1024 * 1024) {
+      ctx->Const.MaxCubeTextureLevels = 11; /* 1024x1024 */
+   }
+   else if (memPerUnit >= 4 * 6 * 512 * 512) {
+      ctx->Const.MaxCubeTextureLevels = 10; /* 512x512 */
+   }
+   else if (memPerUnit >= 4 * 6 * 256 * 256) {
+      ctx->Const.MaxCubeTextureLevels = 9; /* 256x256 */
+   }
+   else { /* 393216 bytes */
+      ctx->Const.MaxCubeTextureLevels = 8; /* 128x128 */
+   }
+
+   ctx->Const.MaxTextureUnits = 2;
+   ctx->Const.MaxTextureMaxAnisotropy = 16.0;
+
+   /* No wide points.
+    */
+   ctx->Const.MinPointSize = 1.0;
+   ctx->Const.MinPointSizeAA = 1.0;
+   ctx->Const.MaxPointSize = 1.0;
+   ctx->Const.MaxPointSizeAA = 1.0;
+
+   ctx->Const.MinLineWidth = 1.0;
+   ctx->Const.MinLineWidthAA = 1.0;
+   ctx->Const.MaxLineWidth = 10.0;
+   ctx->Const.MaxLineWidthAA = 10.0;
+   ctx->Const.LineWidthGranularity = 0.0625;
+
+   /* Initialize the software rasterizer and helper modules.
+    */
+   _swrast_CreateContext( ctx );
+   _ac_CreateContext( ctx );
+   _tnl_CreateContext( ctx );
+   _swsetup_CreateContext( ctx );
+   _ae_create_context( ctx );
+
+   /* Install the customized pipeline:
+    */
+   _tnl_destroy_pipeline( ctx );
+   _tnl_install_pipeline( ctx, r200_pipeline );
+   ctx->Driver.FlushVertices = r200FlushVertices;
+
+   /* Try and keep materials and vertices separate:
+    */
+   _tnl_isolate_materials( ctx, GL_TRUE );
+
+
+   /* Configure swrast to match hardware characteristics:
+    */
+   _swrast_allow_pixel_fog( ctx, GL_FALSE );
+   _swrast_allow_vertex_fog( ctx, GL_TRUE );
+
+
+   _math_matrix_ctr( &rmesa->TexGenMatrix[0] );
+   _math_matrix_ctr( &rmesa->TexGenMatrix[1] );
+   _math_matrix_ctr( &rmesa->tmpmat );
+   _math_matrix_set_identity( &rmesa->TexGenMatrix[0] );
+   _math_matrix_set_identity( &rmesa->TexGenMatrix[1] );
+   _math_matrix_set_identity( &rmesa->tmpmat );
+
+   r200InitExtensions( ctx, r200Screen );
+   r200InitDriverFuncs( ctx );
+   r200InitIoctlFuncs( ctx );
+   r200InitStateFuncs( ctx );
+   r200InitSpanFuncs( ctx );
+   r200InitPixelFuncs( ctx );
+   r200InitTextureFuncs( ctx );
+   r200InitState( rmesa );
+   r200InitSwtcl( ctx );
+
+   rmesa->iw.irq_seq = -1;
+   rmesa->irqsEmitted = 0;
+   rmesa->do_irqs = (rmesa->dri.drmMinor >= 6 && 
+		     !getenv("R200_NO_IRQS") &&
+		     rmesa->r200Screen->irq);
+
+   if (!rmesa->do_irqs)
+      fprintf(stderr, 
+	      "IRQ's not enabled, falling back to busy waits: %d %d %d\n",
+	      rmesa->dri.drmMinor,
+	      !!getenv("R200_NO_IRQS"),
+	      rmesa->r200Screen->irq);
+
+
+   rmesa->do_usleeps = !getenv("R200_NO_USLEEPS");
+   rmesa->prefer_agp_client_texturing = 
+      (getenv("R200_AGP_CLIENT_TEXTURES") != 0);
+   
+   
+#if DO_DEBUG
+   if (getenv("R200_DEBUG"))
+      add_debug_flags( getenv("R200_DEBUG") );
+   if (getenv("RADEON_DEBUG"))
+      add_debug_flags( getenv("RADEON_DEBUG") );
+#endif
+
+   if (getenv("R200_NO_RAST")) {
+      fprintf(stderr, "disabling 3D acceleration\n");
+      FALLBACK(rmesa, R200_FALLBACK_DISABLE, 1); 
+   }
+   else if (getenv("R200_NO_TCL")) {
+      fprintf(stderr, "disabling TCL support\n");
+      TCL_FALLBACK(rmesa->glCtx, R200_TCL_FALLBACK_TCL_DISABLE, 1); 
+   }
+   else {
+      if (!getenv("R200_NO_VTXFMT")) {
+	 r200VtxfmtInit( ctx );
+      }
+      _tnl_need_dlist_norm_lengths( ctx, GL_FALSE );
+   }
+   return GL_TRUE;
+}
+
+
+/* Destroy the device specific context.
+ */
+/* Destroy the Mesa and driver specific context data.
+ */
+void r200DestroyContext( __DRIcontextPrivate *driContextPriv )
+{
+   GET_CURRENT_CONTEXT(ctx);
+   r200ContextPtr rmesa = (r200ContextPtr) driContextPriv->driverPrivate;
+   r200ContextPtr current = ctx ? R200_CONTEXT(ctx) : NULL;
+
+   /* check if we're deleting the currently bound context */
+   if (rmesa == current) {
+      R200_FIREVERTICES( rmesa );
+      _mesa_make_current2(NULL, NULL, NULL);
+   }
+
+   /* Free r200 context resources */
+   assert(rmesa); /* should never be null */
+   if ( rmesa ) {
+      if (rmesa->glCtx->Shared->RefCount == 1) {
+         /* This share group is about to go away, free our private
+          * texture object data.
+          */
+         r200TexObjPtr t, next_t;
+         int i;
+
+         for ( i = 0 ; i < rmesa->texture.numHeaps ; i++ ) {
+            foreach_s ( t, next_t, &rmesa->texture.objects[i] ) {
+               r200DestroyTexObj( rmesa, t );
+            }
+            mmDestroy( rmesa->texture.heap[i] );
+	    rmesa->texture.heap[i] = NULL;
+         }
+
+         foreach_s ( t, next_t, &rmesa->texture.swapped ) {
+            r200DestroyTexObj( rmesa, t );
+         }
+      }
+
+      _swsetup_DestroyContext( rmesa->glCtx );
+      _tnl_DestroyContext( rmesa->glCtx );
+      _ac_DestroyContext( rmesa->glCtx );
+      _swrast_DestroyContext( rmesa->glCtx );
+
+      r200DestroySwtcl( rmesa->glCtx );
+
+      r200ReleaseArrays( rmesa->glCtx, ~0 );
+
+      if (rmesa->dma.current.buf) {
+	 r200ReleaseDmaRegion( rmesa, &rmesa->dma.current, __FUNCTION__ );
+	 r200FlushCmdBuf( rmesa, __FUNCTION__ );
+      }
+
+      if (!rmesa->TclFallback & R200_TCL_FALLBACK_TCL_DISABLE)
+	 if (!getenv("R200_NO_VTXFMT"))
+	    r200VtxfmtDestroy( rmesa->glCtx );
+
+      /* free the Mesa context */
+      rmesa->glCtx->DriverCtx = NULL;
+      _mesa_destroy_context( rmesa->glCtx );
+
+      if (rmesa->state.scissor.pClipRects) {
+	 FREE(rmesa->state.scissor.pClipRects);
+	 rmesa->state.scissor.pClipRects = 0;
+      }
+
+      FREE( rmesa );
+   }
+}
+
+
+
+
+void
+r200SwapBuffers( __DRIdrawablePrivate *dPriv )
+{
+   if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
+      r200ContextPtr rmesa;
+      GLcontext *ctx;
+      rmesa = (r200ContextPtr) dPriv->driContextPriv->driverPrivate;
+      ctx = rmesa->glCtx;
+      if (ctx->Visual.doubleBufferMode) {
+         _mesa_notifySwapBuffers( ctx );  /* flush pending rendering comands */
+         if ( rmesa->doPageFlip ) {
+            r200PageFlip( dPriv );
+         }
+         else {
+            r200CopyBuffer( dPriv );
+         }
+      }
+   }
+   else {
+      /* XXX this shouldn't be an error but we can't handle it for now */
+      _mesa_problem(NULL, "r200SwapBuffers: drawable has no context!\n");
+   }
+}
+
+
+/* Force the context `c' to be the current context and associate with it
+ * buffer `b'.
+ */
+GLboolean
+r200MakeCurrent( __DRIcontextPrivate *driContextPriv,
+                   __DRIdrawablePrivate *driDrawPriv,
+                   __DRIdrawablePrivate *driReadPriv )
+{
+   if ( driContextPriv ) {
+      r200ContextPtr newR200Ctx = 
+	 (r200ContextPtr) driContextPriv->driverPrivate;
+
+      if (R200_DEBUG & DEBUG_DRI)
+	 fprintf(stderr, "%s ctx %p\n", __FUNCTION__, newR200Ctx->glCtx);
+
+      if ( newR200Ctx->dri.drawable != driDrawPriv ) {
+	 newR200Ctx->dri.drawable = driDrawPriv;
+	 r200UpdateWindow( newR200Ctx->glCtx );
+	 r200UpdateViewportOffset( newR200Ctx->glCtx );
+      }
+
+      _mesa_make_current2( newR200Ctx->glCtx,
+			   (GLframebuffer *) driDrawPriv->driverPrivate,
+			   (GLframebuffer *) driReadPriv->driverPrivate );
+
+      if ( !newR200Ctx->glCtx->Viewport.Width ) {
+	 _mesa_set_viewport( newR200Ctx->glCtx, 0, 0,
+			     driDrawPriv->w, driDrawPriv->h );
+      }
+
+      if (newR200Ctx->vb.enabled)
+	 r200VtxfmtMakeCurrent( newR200Ctx->glCtx );
+
+      _mesa_update_state( newR200Ctx->glCtx );
+      r200ValidateState( newR200Ctx->glCtx );
+
+   } else {
+      if (R200_DEBUG & DEBUG_DRI)
+	 fprintf(stderr, "%s ctx is null\n", __FUNCTION__);
+      _mesa_make_current( 0, 0 );
+   }
+
+   if (R200_DEBUG & DEBUG_DRI)
+      fprintf(stderr, "End %s\n", __FUNCTION__);
+   return GL_TRUE;
+}
+
+/* Force the context `c' to be unbound from its buffer.
+ */
+GLboolean
+r200UnbindContext( __DRIcontextPrivate *driContextPriv )
+{
+   r200ContextPtr rmesa = (r200ContextPtr) driContextPriv->driverPrivate;
+
+   if (R200_DEBUG & DEBUG_DRI)
+      fprintf(stderr, "%s ctx %p\n", __FUNCTION__, rmesa->glCtx);
+
+   r200VtxfmtUnbindContext( rmesa->glCtx );
+   return GL_TRUE;
+}
+
+
+
+
+
+

src/mesa/drivers/dri/r200/r200_context.h

@@ -0,0 +1,944 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#ifndef __R200_CONTEXT_H__
+#define __R200_CONTEXT_H__
+
+#ifdef GLX_DIRECT_RENDERING
+
+#include "dri_util.h"
+#include "radeon_common.h"
+
+#include "macros.h"
+#include "mtypes.h"
+#include "r200_reg.h"
+
+#define ENABLE_HW_3D_TEXTURE 0  /* XXX this is temporary! */
+
+struct r200_context;
+typedef struct r200_context r200ContextRec;
+typedef struct r200_context *r200ContextPtr;
+
+#include "r200_lock.h"
+#include "r200_screen.h"
+#include "mm.h"
+
+/* Flags for software fallback cases */
+/* See correponding strings in r200_swtcl.c */
+#define R200_FALLBACK_TEXTURE           0x1
+#define R200_FALLBACK_DRAW_BUFFER       0x2
+#define R200_FALLBACK_STENCIL           0x4
+#define R200_FALLBACK_RENDER_MODE       0x8
+#define R200_FALLBACK_BLEND_EQ          0x10
+#define R200_FALLBACK_BLEND_FUNC        0x20
+#define R200_FALLBACK_DISABLE           0x40
+
+/* The blit width for texture uploads
+ */
+#define BLIT_WIDTH_BYTES 1024
+
+/* Use the templated vertex format:
+ */
+#define COLOR_IS_RGBA
+#define TAG(x) r200##x
+#include "tnl_dd/t_dd_vertex.h"
+#undef TAG
+
+typedef void (*r200_tri_func)( r200ContextPtr,
+				 r200Vertex *,
+				 r200Vertex *,
+				 r200Vertex * );
+
+typedef void (*r200_line_func)( r200ContextPtr,
+				  r200Vertex *,
+				  r200Vertex * );
+
+typedef void (*r200_point_func)( r200ContextPtr,
+				   r200Vertex * );
+
+
+struct r200_colorbuffer_state {
+   GLuint clear;
+   GLint drawOffset, drawPitch;
+};
+
+
+struct r200_depthbuffer_state {
+   GLfloat scale;
+};
+
+struct r200_pixel_state {
+   GLint readOffset, readPitch;
+};
+
+struct r200_scissor_state {
+   XF86DRIClipRectRec rect;
+   GLboolean enabled;
+
+   GLuint numClipRects;			/* Cliprects active */
+   GLuint numAllocedClipRects;		/* Cliprects available */
+   XF86DRIClipRectPtr pClipRects;
+};
+
+struct r200_stencilbuffer_state {
+   GLboolean hwBuffer;
+   GLuint clear;			/* rb3d_stencilrefmask value */
+};
+
+struct r200_stipple_state {
+   GLuint mask[32];
+};
+
+
+
+#define TEX_0   0x1
+#define TEX_1   0x2
+#define TEX_ALL 0x3
+
+typedef struct r200_tex_obj r200TexObj, *r200TexObjPtr;
+
+/* Texture object in locally shared texture space.
+ */
+struct r200_tex_obj {
+   r200TexObjPtr next, prev;
+
+   struct gl_texture_object *tObj;	/* Mesa texture object */
+
+   PMemBlock memBlock;			/* Memory block containing texture */
+   GLuint bufAddr;			/* Offset to start of locally
+					   shared texture block */
+
+   GLuint dirty_images[6];		/* Flags for whether or not
+					   images need to be uploaded to
+					   local or AGP texture space.
+					   Six cube faces. */
+
+   GLuint dirty_state;		        /* Flags (1 per texunit) for
+					   whether or not this texobj
+					   has dirty hardware state
+					   (pp_*) that needs to be
+					   brought into the
+					   texunit. */
+
+   GLint heap;				/* Texture heap currently stored in */
+
+   drmRadeonTexImage image[6][RADEON_MAX_TEXTURE_LEVELS];
+					/* Six, for the cube faces */
+
+   GLint totalSize;			/* Total size of the texture
+					   including all mipmap levels,
+					   and all six cube faces */
+
+   GLuint pp_txfilter;		        /* hardware register values */
+   GLuint pp_txformat;
+   GLuint pp_txformat_x;
+   GLuint pp_txoffset;		        /* Image location in texmem.
+					   All cube faces follow. */
+   GLuint pp_txsize;		        /* npot only */
+   GLuint pp_txpitch;		        /* npot only */
+   GLuint pp_border_color;
+   GLuint pp_cubic_faces;	        /* cube face 1,2,3,4 log2 sizes */
+
+   /* texObj->Image[firstLevel] through texObj->Image[lastLevel] are the
+    * images to upload.
+    */
+   GLint firstLevel;     
+   GLint lastLevel;      
+};
+
+
+struct r200_texture_env_state {
+   r200TexObjPtr texobj;
+   GLenum format;
+   GLenum envMode;
+};
+
+#define R200_MAX_TEXTURE_UNITS 3
+
+struct r200_texture_state {
+   struct r200_texture_env_state unit[R200_MAX_TEXTURE_UNITS];
+};
+
+
+struct r200_state_atom {
+   struct r200_state_atom *next, *prev;
+   const char *name;		         /* for debug */
+   int cmd_size;		         /* size in bytes */
+   GLuint idx;
+   int *cmd;			         /* one or more cmd's */
+   int *lastcmd;			 /* one or more cmd's */
+   GLboolean (*check)( GLcontext *, int );    /* is this state active? */
+};
+   
+
+
+/* Trying to keep these relatively short as the variables are becoming
+ * extravagently long.  Drop the R200_ off the front of everything -
+ * I think we know we're in the r200 driver by now, and keep the
+ * prefix to 3 letters unless absolutely impossible.  
+ */
+
+#define CTX_CMD_0             0
+#define CTX_PP_MISC           1
+#define CTX_PP_FOG_COLOR      2
+#define CTX_RE_SOLID_COLOR    3
+#define CTX_RB3D_BLENDCNTL    4
+#define CTX_RB3D_DEPTHOFFSET  5
+#define CTX_RB3D_DEPTHPITCH   6
+#define CTX_RB3D_ZSTENCILCNTL 7
+#define CTX_CMD_1             8
+#define CTX_PP_CNTL           9
+#define CTX_RB3D_CNTL         10
+#define CTX_RB3D_COLOROFFSET  11
+#define CTX_CMD_2             12 /* why */
+#define CTX_RB3D_COLORPITCH   13 /* why */
+#define CTX_STATE_SIZE        14
+
+#define SET_CMD_0               0
+#define SET_SE_CNTL             1
+#define SET_RE_CNTL             2 /* replace se_coord_fmt */
+#define SET_STATE_SIZE          3
+
+#define VTE_CMD_0               0
+#define VTE_SE_VTE_CNTL         1
+#define VTE_STATE_SIZE          2
+
+#define LIN_CMD_0               0
+#define LIN_RE_LINE_PATTERN     1
+#define LIN_RE_LINE_STATE       2
+#define LIN_CMD_1               3
+#define LIN_SE_LINE_WIDTH       4
+#define LIN_STATE_SIZE          5
+
+#define MSK_CMD_0               0
+#define MSK_RB3D_STENCILREFMASK 1
+#define MSK_RB3D_ROPCNTL        2
+#define MSK_RB3D_PLANEMASK      3
+#define MSK_STATE_SIZE          4
+
+#define VPT_CMD_0           0
+#define VPT_SE_VPORT_XSCALE          1
+#define VPT_SE_VPORT_XOFFSET         2
+#define VPT_SE_VPORT_YSCALE          3
+#define VPT_SE_VPORT_YOFFSET         4
+#define VPT_SE_VPORT_ZSCALE          5
+#define VPT_SE_VPORT_ZOFFSET         6
+#define VPT_STATE_SIZE      7
+
+#define ZBS_CMD_0               0
+#define ZBS_SE_ZBIAS_FACTOR     1
+#define ZBS_SE_ZBIAS_CONSTANT   2
+#define ZBS_STATE_SIZE          3
+
+#define MSC_CMD_0               0
+#define MSC_RE_MISC             1
+#define MSC_STATE_SIZE          2
+
+#define TAM_CMD_0               0
+#define TAM_DEBUG3              1
+#define TAM_STATE_SIZE          2
+
+#define TEX_CMD_0                   0
+#define TEX_PP_TXFILTER             1  /*2c00*/
+#define TEX_PP_TXFORMAT             2  /*2c04*/
+#define TEX_PP_TXFORMAT_X           3  /*2c08*/
+#define TEX_PP_TXSIZE               4  /*2c0c*/
+#define TEX_PP_TXPITCH              5  /*2c10*/
+#define TEX_PP_BORDER_COLOR         6  /*2c14*/
+#define TEX_CMD_1                   7
+#define TEX_PP_TXOFFSET             8  /*2d00 */
+#define TEX_STATE_SIZE              9
+
+#define CUBE_CMD_0                  0  /* 1 register follows */
+#define CUBE_PP_CUBIC_FACES         1  /* 0x2c18 */
+#define CUBE_CMD_1                  2  /* 5 registers follow */
+#define CUBE_PP_CUBIC_OFFSET_F1     3  /* 0x2d04 */
+#define CUBE_PP_CUBIC_OFFSET_F2     4  /* 0x2d08 */
+#define CUBE_PP_CUBIC_OFFSET_F3     5  /* 0x2d0c */
+#define CUBE_PP_CUBIC_OFFSET_F4     6  /* 0x2d10 */
+#define CUBE_PP_CUBIC_OFFSET_F5     7  /* 0x2d14 */
+#define CUBE_STATE_SIZE             8
+
+#define PIX_CMD_0                   0
+#define PIX_PP_TXCBLEND             1
+#define PIX_PP_TXCBLEND2            2
+#define PIX_PP_TXABLEND             3
+#define PIX_PP_TXABLEND2            4
+#define PIX_STATE_SIZE              5
+
+#define TF_CMD_0                    0
+#define TF_TFACTOR_0                1
+#define TF_TFACTOR_1                2
+#define TF_TFACTOR_2                3
+#define TF_TFACTOR_3                4
+#define TF_TFACTOR_4                5
+#define TF_TFACTOR_5                6
+#define TF_STATE_SIZE               7
+
+#define TCL_CMD_0                 0
+#define TCL_LIGHT_MODEL_CTL_0     1
+#define TCL_LIGHT_MODEL_CTL_1     2
+#define TCL_PER_LIGHT_CTL_0       3
+#define TCL_PER_LIGHT_CTL_1       4
+#define TCL_PER_LIGHT_CTL_2       5
+#define TCL_PER_LIGHT_CTL_3       6
+#define TCL_CMD_1                 7
+#define TCL_UCP_VERT_BLEND_CTL    8
+#define TCL_STATE_SIZE            9
+
+#define MSL_CMD_0                     0
+#define MSL_MATRIX_SELECT_0           1
+#define MSL_MATRIX_SELECT_1           2
+#define MSL_MATRIX_SELECT_2           3
+#define MSL_MATRIX_SELECT_3           4
+#define MSL_MATRIX_SELECT_4           5
+#define MSL_STATE_SIZE                6
+
+#define TCG_CMD_0                 0
+#define TCG_TEX_PROC_CTL_2            1
+#define TCG_TEX_PROC_CTL_3            2
+#define TCG_TEX_PROC_CTL_0            3
+#define TCG_TEX_PROC_CTL_1            4
+#define TCG_TEX_CYL_WRAP_CTL      5
+#define TCG_STATE_SIZE            6
+
+#define MTL_CMD_0            0	
+#define MTL_EMMISSIVE_RED    1	
+#define MTL_EMMISSIVE_GREEN  2	
+#define MTL_EMMISSIVE_BLUE   3	
+#define MTL_EMMISSIVE_ALPHA  4	
+#define MTL_AMBIENT_RED      5
+#define MTL_AMBIENT_GREEN    6
+#define MTL_AMBIENT_BLUE     7
+#define MTL_AMBIENT_ALPHA    8
+#define MTL_DIFFUSE_RED      9
+#define MTL_DIFFUSE_GREEN    10
+#define MTL_DIFFUSE_BLUE     11
+#define MTL_DIFFUSE_ALPHA    12
+#define MTL_SPECULAR_RED     13
+#define MTL_SPECULAR_GREEN   14
+#define MTL_SPECULAR_BLUE    15
+#define MTL_SPECULAR_ALPHA   16
+#define MTL_CMD_1            17
+#define MTL_SHININESS        18
+#define MTL_STATE_SIZE       19
+
+#define VAP_CMD_0                   0
+#define VAP_SE_VAP_CNTL             1
+#define VAP_STATE_SIZE              2
+
+/* Replaces a lot of packet info from radeon
+ */
+#define VTX_CMD_0                   0
+#define VTX_VTXFMT_0            1
+#define VTX_VTXFMT_1            2
+#define VTX_TCL_OUTPUT_VTXFMT_0 3
+#define VTX_TCL_OUTPUT_VTXFMT_1 4
+#define VTX_CMD_1               5
+#define VTX_TCL_OUTPUT_COMPSEL  6
+#define VTX_CMD_2               7
+#define VTX_STATE_CNTL          8
+#define VTX_STATE_SIZE          9
+
+
+#define VTX_COLOR(v,n)   (((v)>>(R200_VTX_COLOR_0_SHIFT+(n)*2))&\
+                         R200_VTX_COLOR_MASK)
+
+#define MAT_CMD_0              0
+#define MAT_ELT_0              1
+#define MAT_STATE_SIZE         17
+
+#define GRD_CMD_0                  0
+#define GRD_VERT_GUARD_CLIP_ADJ    1
+#define GRD_VERT_GUARD_DISCARD_ADJ 2
+#define GRD_HORZ_GUARD_CLIP_ADJ    3
+#define GRD_HORZ_GUARD_DISCARD_ADJ 4
+#define GRD_STATE_SIZE             5
+
+/* position changes frequently when lighting in modelpos - separate
+ * out to new state item?  
+ */
+#define LIT_CMD_0                  0
+#define LIT_AMBIENT_RED            1
+#define LIT_AMBIENT_GREEN          2
+#define LIT_AMBIENT_BLUE           3
+#define LIT_AMBIENT_ALPHA          4
+#define LIT_DIFFUSE_RED            5
+#define LIT_DIFFUSE_GREEN          6
+#define LIT_DIFFUSE_BLUE           7
+#define LIT_DIFFUSE_ALPHA          8
+#define LIT_SPECULAR_RED           9
+#define LIT_SPECULAR_GREEN         10
+#define LIT_SPECULAR_BLUE          11
+#define LIT_SPECULAR_ALPHA         12
+#define LIT_POSITION_X             13
+#define LIT_POSITION_Y             14
+#define LIT_POSITION_Z             15
+#define LIT_POSITION_W             16
+#define LIT_DIRECTION_X            17
+#define LIT_DIRECTION_Y            18
+#define LIT_DIRECTION_Z            19
+#define LIT_DIRECTION_W            20
+#define LIT_ATTEN_CONST            21
+#define LIT_ATTEN_LINEAR           22
+#define LIT_ATTEN_QUADRATIC        23
+#define LIT_ATTEN_XXX              24
+#define LIT_CMD_1                  25
+#define LIT_SPOT_DCD               26
+#define LIT_SPOT_DCM               27
+#define LIT_SPOT_EXPONENT          28
+#define LIT_SPOT_CUTOFF            29
+#define LIT_SPECULAR_THRESH        30
+#define LIT_RANGE_CUTOFF           31 /* ? */
+#define LIT_RANGE_ATTEN            32 /* ? */
+#define LIT_STATE_SIZE             33
+
+/* Fog
+ */
+#define FOG_CMD_0      0
+#define FOG_R          1
+#define FOG_C          2
+#define FOG_D          3
+#define FOG_PAD        4
+#define FOG_STATE_SIZE 5
+
+/* UCP
+ */
+#define UCP_CMD_0      0
+#define UCP_X          1
+#define UCP_Y          2
+#define UCP_Z          3
+#define UCP_W          4
+#define UCP_STATE_SIZE 5
+
+/* GLT - Global ambient
+ */
+#define GLT_CMD_0      0
+#define GLT_RED        1
+#define GLT_GREEN      2
+#define GLT_BLUE       3
+#define GLT_ALPHA      4
+#define GLT_STATE_SIZE 5
+
+/* EYE
+ */
+#define EYE_CMD_0          0
+#define EYE_X              1
+#define EYE_Y              2
+#define EYE_Z              3
+#define EYE_RESCALE_FACTOR 4
+#define EYE_STATE_SIZE     5
+
+/* CST - constant state
+ */
+#define CST_CMD_0                             0
+#define CST_PP_CNTL_X                         1
+#define CST_CMD_1                             2
+#define CST_RB3D_DEPTHXY_OFFSET               3
+#define CST_CMD_2                             4
+#define CST_RE_AUX_SCISSOR_CNTL               5
+#define CST_CMD_3                             6
+#define CST_RE_SCISSOR_TL_0                   7
+#define CST_RE_SCISSOR_BR_0                   8
+#define CST_CMD_4                             9
+#define CST_SE_VAP_CNTL_STATUS                10
+#define CST_CMD_5                             11
+#define CST_RE_POINTSIZE                      12
+#define CST_CMD_6                             13
+#define CST_SE_TCL_INPUT_VTX_0                14
+#define CST_SE_TCL_INPUT_VTX_1                15
+#define CST_SE_TCL_INPUT_VTX_2                16
+#define CST_SE_TCL_INPUT_VTX_3                17
+#define CST_STATE_SIZE                        18
+
+
+
+
+struct r200_hw_state {
+   /* All state should be on one of these lists:
+    */
+   struct r200_state_atom dirty; /* dirty list head placeholder */
+   struct r200_state_atom clean; /* clean list head placeholder */
+
+   /* Hardware state, stored as cmdbuf commands:  
+    *   -- Need to doublebuffer for
+    *           - reviving state after loss of context
+    *           - eliding noop statechange loops? (except line stipple count)
+    */
+   struct r200_state_atom ctx;
+   struct r200_state_atom set;
+   struct r200_state_atom vte;
+   struct r200_state_atom lin;
+   struct r200_state_atom msk;
+   struct r200_state_atom vpt;
+   struct r200_state_atom vap;
+   struct r200_state_atom vtx;
+   struct r200_state_atom tcl;
+   struct r200_state_atom msl;
+   struct r200_state_atom tcg;
+   struct r200_state_atom msc;
+   struct r200_state_atom cst;
+   struct r200_state_atom tam;
+   struct r200_state_atom tf;
+   struct r200_state_atom tex[2];
+   struct r200_state_atom cube[2];
+   struct r200_state_atom zbs;
+   struct r200_state_atom mtl[2]; 
+   struct r200_state_atom mat[5]; 
+   struct r200_state_atom lit[8]; /* includes vec, scl commands */
+   struct r200_state_atom ucp[6];
+   struct r200_state_atom pix[6]; /* pixshader stages */
+   struct r200_state_atom eye; /* eye pos */
+   struct r200_state_atom grd; /* guard band clipping */
+   struct r200_state_atom fog; 
+   struct r200_state_atom glt; 
+};
+
+struct r200_state {
+   /* Derived state for internal purposes:
+    */
+   struct r200_colorbuffer_state color;
+   struct r200_depthbuffer_state depth;
+   struct r200_pixel_state pixel;
+   struct r200_scissor_state scissor;
+   struct r200_stencilbuffer_state stencil;
+   struct r200_stipple_state stipple;
+   struct r200_texture_state texture;
+};
+
+struct r200_texture {
+   r200TexObj objects[R200_NR_TEX_HEAPS];
+   r200TexObj swapped;
+
+   memHeap_t *heap[R200_NR_TEX_HEAPS];
+   GLint age[R200_NR_TEX_HEAPS];
+
+   GLint numHeaps;
+};
+
+/* Need refcounting on dma buffers:
+ */
+struct r200_dma_buffer {
+   int refcount;		/* the number of retained regions in buf */
+   drmBufPtr buf;
+};
+
+#define GET_START(rvb) (rmesa->r200Screen->agp_buffer_offset +		\
+			(rvb)->address - rmesa->dma.buf0_address +	\
+			(rvb)->start)
+
+/* A retained region, eg vertices for indexed vertices.
+ */
+struct r200_dma_region {
+   struct r200_dma_buffer *buf;
+   char *address;		/* == buf->address */
+   int start, end, ptr;		/* offsets from start of buf */
+   int aos_start;
+   int aos_stride;
+   int aos_size;
+};
+
+
+struct r200_dma {
+   /* Active dma region.  Allocations for vertices and retained
+    * regions come from here.  Also used for emitting random vertices,
+    * these may be flushed by calling flush_current();
+    */
+   struct r200_dma_region current;
+   
+   void (*flush)( r200ContextPtr );
+
+   char *buf0_address;		/* start of buf[0], for index calcs */
+   GLuint nr_released_bufs;	/* flush after so many buffers released */
+};
+
+struct r200_dri_mirror {
+   __DRIcontextPrivate	*context;	/* DRI context */
+   __DRIscreenPrivate	*screen;	/* DRI screen */
+   __DRIdrawablePrivate	*drawable;	/* DRI drawable bound to this ctx */
+
+   drmContext hwContext;
+   drmLock *hwLock;
+   int fd;
+   int drmMinor;
+};
+
+
+#define R200_CMD_BUF_SZ  (8*1024) 
+
+struct r200_store {
+   GLuint statenr;
+   GLuint primnr;
+   char cmd_buf[R200_CMD_BUF_SZ];
+   int cmd_used;   
+   int elts_start;
+};
+
+
+/* r200_tcl.c
+ */
+struct r200_tcl_info {
+   GLuint vertex_format;
+   GLint last_offset;
+   GLuint hw_primitive;
+
+   struct r200_dma_region *aos_components[8];
+   GLuint nr_aos_components;
+
+   GLuint *Elts;
+
+   struct r200_dma_region indexed_verts;
+   struct r200_dma_region obj;
+   struct r200_dma_region rgba;
+   struct r200_dma_region spec;
+   struct r200_dma_region fog;
+   struct r200_dma_region tex[R200_MAX_TEXTURE_UNITS];
+   struct r200_dma_region norm;
+};
+
+
+/* r200_swtcl.c
+ */
+struct r200_swtcl_info {
+   GLuint SetupIndex;
+   GLuint SetupNewInputs;
+   GLuint RenderIndex;
+   GLuint vertex_size;
+   GLuint vertex_stride_shift;
+   GLuint vertex_format;
+   char *verts;
+
+   /* Fallback rasterization functions
+    */
+   r200_point_func draw_point;
+   r200_line_func draw_line;
+   r200_tri_func draw_tri;
+
+   GLuint hw_primitive;
+   GLenum render_primitive;
+   GLuint numverts;
+
+   struct r200_dma_region indexed_verts;
+};
+
+
+struct r200_ioctl {
+   GLuint vertex_offset;
+   GLuint vertex_size;
+};
+
+
+
+#define R200_MAX_PRIMS 64
+
+
+/* Want to keep a cache of these around.  Each is parameterized by
+ * only a single value which has only a small range.  Only expect a
+ * few, so just rescan the list each time?
+ */
+struct dynfn {
+   struct dynfn *next, *prev;
+   int key[2];
+   char *code;
+};
+
+struct dfn_lists {
+   struct dynfn Vertex2f;
+   struct dynfn Vertex2fv;
+   struct dynfn Vertex3f;
+   struct dynfn Vertex3fv;
+   struct dynfn Color4ub;
+   struct dynfn Color4ubv;
+   struct dynfn Color3ub;
+   struct dynfn Color3ubv;
+   struct dynfn Color4f;
+   struct dynfn Color4fv;
+   struct dynfn Color3f;
+   struct dynfn Color3fv;
+   struct dynfn SecondaryColor3ubEXT;
+   struct dynfn SecondaryColor3ubvEXT;
+   struct dynfn SecondaryColor3fEXT;
+   struct dynfn SecondaryColor3fvEXT;
+   struct dynfn Normal3f;
+   struct dynfn Normal3fv;
+   struct dynfn TexCoord2f;
+   struct dynfn TexCoord2fv;
+   struct dynfn TexCoord1f;
+   struct dynfn TexCoord1fv;
+   struct dynfn MultiTexCoord2fARB;
+   struct dynfn MultiTexCoord2fvARB;
+   struct dynfn MultiTexCoord1fARB;
+   struct dynfn MultiTexCoord1fvARB;
+};
+
+struct _vb;
+
+struct dfn_generators {
+   struct dynfn *(*Vertex2f)( GLcontext *, const int * );
+   struct dynfn *(*Vertex2fv)( GLcontext *, const int * );
+   struct dynfn *(*Vertex3f)( GLcontext *, const int * );
+   struct dynfn *(*Vertex3fv)( GLcontext *, const int * );
+   struct dynfn *(*Color4ub)( GLcontext *, const int * );
+   struct dynfn *(*Color4ubv)( GLcontext *, const int * );
+   struct dynfn *(*Color3ub)( GLcontext *, const int * );
+   struct dynfn *(*Color3ubv)( GLcontext *, const int * );
+   struct dynfn *(*Color4f)( GLcontext *, const int * );
+   struct dynfn *(*Color4fv)( GLcontext *, const int * );
+   struct dynfn *(*Color3f)( GLcontext *, const int * );
+   struct dynfn *(*Color3fv)( GLcontext *, const int * );
+   struct dynfn *(*SecondaryColor3ubEXT)( GLcontext *, const int * );
+   struct dynfn *(*SecondaryColor3ubvEXT)( GLcontext *, const int * );
+   struct dynfn *(*SecondaryColor3fEXT)( GLcontext *, const int * );
+   struct dynfn *(*SecondaryColor3fvEXT)( GLcontext *, const int * );
+   struct dynfn *(*Normal3f)( GLcontext *, const int * );
+   struct dynfn *(*Normal3fv)( GLcontext *, const int * );
+   struct dynfn *(*TexCoord2f)( GLcontext *, const int * );
+   struct dynfn *(*TexCoord2fv)( GLcontext *, const int * );
+   struct dynfn *(*TexCoord1f)( GLcontext *, const int * );
+   struct dynfn *(*TexCoord1fv)( GLcontext *, const int * );
+   struct dynfn *(*MultiTexCoord2fARB)( GLcontext *, const int * );
+   struct dynfn *(*MultiTexCoord2fvARB)( GLcontext *, const int * );
+   struct dynfn *(*MultiTexCoord1fARB)( GLcontext *, const int * );
+   struct dynfn *(*MultiTexCoord1fvARB)( GLcontext *, const int * );
+};
+
+
+struct r200_vb {
+   /* Keep these first: referenced from codegen templates:
+    */
+   GLint counter, initial_counter;
+   GLint *dmaptr;
+   void (*notify)( void );
+   GLint vertex_size;
+
+   /* A maximum total of 15 elements per vertex:  3 floats for position, 3
+    * floats for normal, 4 floats for color, 4 bytes for secondary color,
+    * 2 floats for each texture unit (4 floats total).
+    * 
+    * As soon as the 3rd TMU is supported or cube maps (or 3D textures) are
+    * supported, this value will grow.
+    * 
+    * The position data is never actually stored here, so 3 elements could be
+    * trimmed out of the buffer.
+    */
+   union { float f; int i; r200_color_t color; } vertex[15];
+
+   GLfloat *normalptr;
+   GLfloat *floatcolorptr;
+   r200_color_t *colorptr;
+   GLfloat *floatspecptr;
+   r200_color_t *specptr;
+   GLfloat *texcoordptr[2];
+
+   GLcontext *context;		/* current context : Single thread only! */
+};
+
+struct r200_prim {
+   GLuint start;
+   GLuint end;
+   GLuint prim;
+};
+
+struct r200_vbinfo {
+   GLenum *prim;		/* &ctx->Driver.CurrentExecPrimitive */
+   GLuint primflags;
+   GLboolean enabled;		/* R200_NO_VTXFMT//R200_NO_TCL env vars */
+   GLboolean installed;
+   GLboolean fell_back;
+   GLboolean recheck;
+   GLint initial_counter;
+   GLint nrverts;
+   GLuint vtxfmt_0, vtxfmt_1;
+
+   GLuint installed_vertex_format;
+   GLuint installed_color_3f_sz;
+
+   struct r200_prim primlist[R200_MAX_PRIMS];
+   int nrprims;
+
+   struct dfn_lists dfn_cache;
+   struct dfn_generators codegen;
+   GLvertexformat vtxfmt;
+};
+
+
+
+
+struct r200_context {
+   GLcontext *glCtx;			/* Mesa context */
+
+   /* Driver and hardware state management
+    */
+   struct r200_hw_state hw;
+   struct r200_state state;
+
+   /* Texture object bookkeeping
+    */
+   struct r200_texture texture;
+
+
+   /* Rasterization and vertex state:
+    */
+   GLuint TclFallback;
+   GLuint Fallback;
+   GLuint NewGLState;
+
+   
+   /* Temporaries for translating away float colors:
+    */
+   struct gl_client_array UbyteColor;
+   struct gl_client_array UbyteSecondaryColor;
+
+   /* Vertex buffers
+    */
+   struct r200_ioctl ioctl;
+   struct r200_dma dma;
+   struct r200_store store;
+
+   /* Page flipping
+    */
+   GLuint doPageFlip;
+
+   /* Busy waiting
+    */
+   GLuint do_usleeps;
+   GLuint do_irqs;
+   GLuint irqsEmitted;
+   drmRadeonIrqWait iw;
+
+   /* Clientdata textures;
+    */
+   GLuint prefer_agp_client_texturing;
+
+   /* Drawable, cliprect and scissor information
+    */
+   GLuint numClipRects;			/* Cliprects for the draw buffer */
+   XF86DRIClipRectPtr pClipRects;
+   unsigned int lastStamp;
+   GLboolean lost_context;
+   r200ScreenPtr r200Screen;	/* Screen private DRI data */
+   RADEONSAREAPrivPtr sarea;		/* Private SAREA data */
+
+   /* TCL stuff
+    */
+   GLmatrix TexGenMatrix[R200_MAX_TEXTURE_UNITS];
+   GLboolean recheck_texgen[R200_MAX_TEXTURE_UNITS];
+   GLboolean TexGenNeedNormals[R200_MAX_TEXTURE_UNITS];
+   GLuint TexMatEnabled;
+   GLuint TexMatCompSel;
+   GLuint TexGenEnabled;
+   GLuint TexGenInputs;
+   GLuint TexGenCompSel;
+   GLmatrix tmpmat;
+
+   /* VBI
+    */
+   GLuint vbl_seq;
+
+   /* r200_tcl.c
+    */
+   struct r200_tcl_info tcl;
+
+   /* r200_swtcl.c
+    */
+   struct r200_swtcl_info swtcl;
+
+   /* r200_vtxfmt.c
+    */
+   struct r200_vbinfo vb;
+
+   /* Mirrors of some DRI state
+    */
+   struct r200_dri_mirror dri;
+};
+
+#define R200_CONTEXT(ctx)		((r200ContextPtr)(ctx->DriverCtx))
+
+
+static __inline GLuint r200PackColor( GLuint cpp,
+					GLubyte r, GLubyte g,
+					GLubyte b, GLubyte a )
+{
+   switch ( cpp ) {
+   case 2:
+      return PACK_COLOR_565( r, g, b );
+   case 4:
+      return PACK_COLOR_8888( a, r, g, b );
+   default:
+      return 0;
+   }
+}
+
+#define R200_OLD_PACKETS 0
+
+
+extern void r200DestroyContext( __DRIcontextPrivate *driContextPriv );
+extern GLboolean r200CreateContext( const __GLcontextModes *glVisual,
+				    __DRIcontextPrivate *driContextPriv,
+				    void *sharedContextPrivate);
+extern void r200SwapBuffers( __DRIdrawablePrivate *dPriv );
+extern GLboolean r200MakeCurrent( __DRIcontextPrivate *driContextPriv,
+				  __DRIdrawablePrivate *driDrawPriv,
+				  __DRIdrawablePrivate *driReadPriv );
+extern GLboolean r200UnbindContext( __DRIcontextPrivate *driContextPriv );
+
+/* ================================================================
+ * Debugging:
+ */
+#define DO_DEBUG		1
+
+#if DO_DEBUG
+extern int R200_DEBUG;
+#else
+#define R200_DEBUG		0
+#endif
+
+#define DEBUG_TEXTURE	0x001
+#define DEBUG_STATE	0x002
+#define DEBUG_IOCTL	0x004
+#define DEBUG_PRIMS	0x008
+#define DEBUG_VERTS	0x010
+#define DEBUG_FALLBACKS	0x020
+#define DEBUG_VFMT	0x040
+#define DEBUG_CODEGEN	0x080
+#define DEBUG_VERBOSE	0x100
+#define DEBUG_DRI       0x200
+#define DEBUG_DMA       0x400
+#define DEBUG_SANITY    0x800
+#define DEBUG_SYNC      0x1000
+#define DEBUG_PIXEL     0x2000
+#define DEBUG_MEMORY    0x4000
+
+#endif
+#endif /* __R200_CONTEXT_H__ */

src/mesa/drivers/dri/r200/r200_ioctl.c

@@ -0,0 +1,934 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#include "glheader.h"
+#include "imports.h"
+#include "macros.h"
+#include "context.h"
+#include "swrast/swrast.h"
+
+#include "r200_context.h"
+#include "r200_state.h"
+#include "r200_ioctl.h"
+#include "r200_tcl.h"
+#include "r200_sanity.h"
+#include "radeon_reg.h"
+
+#include <unistd.h>  /* for usleep() */
+#include <errno.h>
+
+
+#define R200_TIMEOUT             512
+#define R200_IDLE_RETRY           16
+
+
+static void do_usleep( int nr, const char *caller )
+{
+   //if (1) fprintf(stderr, "usleep %d in %s\n", nr, caller );
+   if (1) usleep( nr );
+}
+
+static void r200WaitForIdle( r200ContextPtr rmesa );
+
+
+int r200FlushCmdBufLocked( r200ContextPtr rmesa, const char * caller )
+{
+   int ret, i;
+   drmRadeonCmdBuffer cmd;
+
+   if (R200_DEBUG & DEBUG_IOCTL) {
+      fprintf(stderr, "%s from %s\n", __FUNCTION__, caller); 
+
+      if (0 & R200_DEBUG & DEBUG_VERBOSE) 
+	 for (i = 0 ; i < rmesa->store.cmd_used ; i += 4 )
+	    fprintf(stderr, "%d: %x\n", i/4, 
+		    *(int *)(&rmesa->store.cmd_buf[i]));
+   }
+
+   if (R200_DEBUG & DEBUG_DMA)
+      fprintf(stderr, "%s: Releasing %d buffers\n", __FUNCTION__,
+	      rmesa->dma.nr_released_bufs);
+
+
+   if (R200_DEBUG & DEBUG_SANITY) {
+      if (rmesa->state.scissor.enabled) 
+	 ret = r200SanityCmdBuffer( rmesa, 
+				    rmesa->state.scissor.numClipRects,
+				    rmesa->state.scissor.pClipRects);
+      else
+	 ret = r200SanityCmdBuffer( rmesa, 
+				    rmesa->numClipRects,
+				    rmesa->pClipRects);
+      if (ret) {
+	 fprintf(stderr, "drmSanityCommandWrite: %d\n", ret);	 
+	 goto out;
+      }
+   }
+
+
+   if (R200_DEBUG & DEBUG_MEMORY) {
+      if (!r200ValidateTexObjs( rmesa )) {
+	 fprintf(stderr, " -- tex memory is inconsistent - expect mangled textures\n");
+      }
+   }
+
+
+   cmd.bufsz = rmesa->store.cmd_used;
+   cmd.buf = rmesa->store.cmd_buf;
+
+   if (rmesa->state.scissor.enabled) {
+      cmd.nbox = rmesa->state.scissor.numClipRects;
+      cmd.boxes = (drmClipRect *)rmesa->state.scissor.pClipRects;
+   } else {
+      cmd.nbox = rmesa->numClipRects;
+      cmd.boxes = (drmClipRect *)rmesa->pClipRects;
+   }
+
+   ret = drmCommandWrite( rmesa->dri.fd,
+			  DRM_RADEON_CMDBUF,
+			  &cmd, sizeof(cmd) );
+
+   if (ret)
+      fprintf(stderr, "drmCommandWrite: %d\n", ret);
+
+   if (R200_DEBUG & DEBUG_SYNC) {
+      fprintf(stderr, "\nSyncing in %s\n\n", __FUNCTION__);
+      r200WaitForIdleLocked( rmesa );
+   }
+
+
+ out:
+   rmesa->store.primnr = 0;
+   rmesa->store.statenr = 0;
+   rmesa->store.cmd_used = 0;
+   rmesa->dma.nr_released_bufs = 0;
+/*    rmesa->lost_context = 0; */
+   rmesa->lost_context = 1;	
+   return ret;
+}
+
+
+/* Note: does not emit any commands to avoid recursion on
+ * r200AllocCmdBuf.
+ */
+void r200FlushCmdBuf( r200ContextPtr rmesa, const char *caller )
+{
+   int ret;
+
+   LOCK_HARDWARE( rmesa );
+
+   ret = r200FlushCmdBufLocked( rmesa, caller );
+
+   UNLOCK_HARDWARE( rmesa );
+
+   if (ret) {
+      fprintf(stderr, "drmRadeonCmdBuffer: %d (exiting)\n", ret);
+      exit(ret);
+   }
+}
+
+
+/* =============================================================
+ * Hardware vertex buffer handling
+ */
+
+
+void r200RefillCurrentDmaRegion( r200ContextPtr rmesa )
+{
+   struct r200_dma_buffer *dmabuf;
+   int fd = rmesa->dri.fd;
+   int index = 0;
+   int size = 0;
+   drmDMAReq dma;
+   int ret;
+
+   if (R200_DEBUG & (DEBUG_IOCTL|DEBUG_DMA))
+      fprintf(stderr, "%s\n", __FUNCTION__);  
+
+   if (rmesa->dma.flush) {
+      rmesa->dma.flush( rmesa );
+   }
+
+   if (rmesa->dma.current.buf)
+      r200ReleaseDmaRegion( rmesa, &rmesa->dma.current, __FUNCTION__ );
+
+   if (rmesa->dma.nr_released_bufs > 4)
+      r200FlushCmdBuf( rmesa, __FUNCTION__ );
+
+   dma.context = rmesa->dri.hwContext;
+   dma.send_count = 0;
+   dma.send_list = NULL;
+   dma.send_sizes = NULL;
+   dma.flags = 0;
+   dma.request_count = 1;
+   dma.request_size = RADEON_BUFFER_SIZE;
+   dma.request_list = &index;
+   dma.request_sizes = &size;
+   dma.granted_count = 0;
+
+   LOCK_HARDWARE(rmesa);	/* no need to validate */
+
+   while (1) {
+      ret = drmDMA( fd, &dma );
+      if (ret == 0)
+	 break;
+   
+      if (rmesa->dma.nr_released_bufs) {
+	 r200FlushCmdBufLocked( rmesa, __FUNCTION__ );
+      }
+      
+      if (rmesa->do_usleeps) {
+	 UNLOCK_HARDWARE( rmesa );
+	 do_usleep(1, __FUNCTION__);
+	 LOCK_HARDWARE( rmesa );
+      }
+   }
+
+   UNLOCK_HARDWARE(rmesa);
+
+   if (R200_DEBUG & DEBUG_DMA)
+      fprintf(stderr, "Allocated buffer %d\n", index);
+
+   dmabuf = CALLOC_STRUCT( r200_dma_buffer );
+   dmabuf->buf = &rmesa->r200Screen->buffers->list[index];
+   dmabuf->refcount = 1;
+
+   rmesa->dma.current.buf = dmabuf;
+   rmesa->dma.current.address = dmabuf->buf->address;
+   rmesa->dma.current.end = dmabuf->buf->total;
+   rmesa->dma.current.start = 0;
+   rmesa->dma.current.ptr = 0;
+}
+
+void r200ReleaseDmaRegion( r200ContextPtr rmesa,
+			     struct r200_dma_region *region,
+			     const char *caller )
+{
+   if (R200_DEBUG & DEBUG_IOCTL)
+      fprintf(stderr, "%s from %s\n", __FUNCTION__, caller); 
+   
+   if (!region->buf)
+      return;
+
+   if (rmesa->dma.flush)
+      rmesa->dma.flush( rmesa );
+
+   if (--region->buf->refcount == 0) {
+      drmRadeonCmdHeader *cmd;
+
+      if (R200_DEBUG & (DEBUG_IOCTL|DEBUG_DMA))
+	 fprintf(stderr, "%s -- DISCARD BUF %d\n", __FUNCTION__,
+		 region->buf->buf->idx);  
+      
+      cmd = (drmRadeonCmdHeader *)r200AllocCmdBuf( rmesa, sizeof(*cmd), 
+						     __FUNCTION__ );
+      cmd->dma.cmd_type = RADEON_CMD_DMA_DISCARD;
+      cmd->dma.buf_idx = region->buf->buf->idx;
+      FREE(region->buf);
+      rmesa->dma.nr_released_bufs++;
+   }
+
+   region->buf = 0;
+   region->start = 0;
+}
+
+/* Allocates a region from rmesa->dma.current.  If there isn't enough
+ * space in current, grab a new buffer (and discard what was left of current)
+ */
+void r200AllocDmaRegion( r200ContextPtr rmesa, 
+			   struct r200_dma_region *region,
+			   int bytes,
+			   int alignment )
+{
+   if (R200_DEBUG & DEBUG_IOCTL)
+      fprintf(stderr, "%s %d\n", __FUNCTION__, bytes);
+
+   if (rmesa->dma.flush)
+      rmesa->dma.flush( rmesa );
+
+   if (region->buf)
+      r200ReleaseDmaRegion( rmesa, region, __FUNCTION__ );
+
+   alignment--;
+   rmesa->dma.current.start = rmesa->dma.current.ptr = 
+      (rmesa->dma.current.ptr + alignment) & ~alignment;
+
+   if ( rmesa->dma.current.ptr + bytes > rmesa->dma.current.end ) 
+      r200RefillCurrentDmaRegion( rmesa );
+
+   region->start = rmesa->dma.current.start;
+   region->ptr = rmesa->dma.current.start;
+   region->end = rmesa->dma.current.start + bytes;
+   region->address = rmesa->dma.current.address;
+   region->buf = rmesa->dma.current.buf;
+   region->buf->refcount++;
+
+   rmesa->dma.current.ptr += bytes; /* bug - if alignment > 7 */
+   rmesa->dma.current.start = 
+      rmesa->dma.current.ptr = (rmesa->dma.current.ptr + 0x7) & ~0x7;  
+}
+
+void r200AllocDmaRegionVerts( r200ContextPtr rmesa, 
+				struct r200_dma_region *region,
+				int numverts,
+				int vertsize,
+				int alignment )
+{
+   r200AllocDmaRegion( rmesa, region, vertsize * numverts, alignment );
+}
+
+/* ================================================================
+ * SwapBuffers with client-side throttling
+ */
+
+static GLuint r200GetLastFrame(r200ContextPtr rmesa)
+{
+   drmRadeonGetParam gp;
+   int ret;
+   GLuint frame;
+
+   gp.param = RADEON_PARAM_LAST_FRAME;
+   gp.value = (int *)&frame;
+   ret = drmCommandWriteRead( rmesa->dri.fd, DRM_RADEON_GETPARAM,
+			      &gp, sizeof(gp) );
+   if ( ret ) {
+      fprintf( stderr, "%s: drmRadeonGetParam: %d\n", __FUNCTION__, ret );
+      exit(1);
+   }
+
+   return frame;
+}
+
+static void r200EmitIrqLocked( r200ContextPtr rmesa )
+{
+   drmRadeonIrqEmit ie;
+   int ret;
+
+   ie.irq_seq = &rmesa->iw.irq_seq;
+   ret = drmCommandWriteRead( rmesa->dri.fd, DRM_RADEON_IRQ_EMIT, 
+			      &ie, sizeof(ie) );
+   if ( ret ) {
+      fprintf( stderr, "%s: drmRadeonIrqEmit: %d\n", __FUNCTION__, ret );
+      exit(1);
+   }
+}
+
+
+static void r200WaitIrq( r200ContextPtr rmesa )
+{
+   int ret;
+
+   do {
+      ret = drmCommandWrite( rmesa->dri.fd, DRM_RADEON_IRQ_WAIT,
+			     &rmesa->iw, sizeof(rmesa->iw) );
+   } while (ret && (errno == EINTR || errno == EAGAIN));
+
+   if ( ret ) {
+      fprintf( stderr, "%s: drmRadeonIrqWait: %d\n", __FUNCTION__, ret );
+      exit(1);
+   }
+}
+
+
+static void r200WaitForFrameCompletion( r200ContextPtr rmesa )
+{
+   RADEONSAREAPrivPtr sarea = rmesa->sarea;
+
+   if (rmesa->do_irqs) {
+      if (r200GetLastFrame(rmesa) < sarea->last_frame) {
+	 if (!rmesa->irqsEmitted) {
+	    while (r200GetLastFrame (rmesa) < sarea->last_frame)
+	       ;
+	 }
+	 else {
+	    UNLOCK_HARDWARE( rmesa ); 
+	    r200WaitIrq( rmesa );	
+	    LOCK_HARDWARE( rmesa ); 
+	 }
+	 rmesa->irqsEmitted = 10;
+      }
+
+      if (rmesa->irqsEmitted) {
+	 r200EmitIrqLocked( rmesa );
+	 rmesa->irqsEmitted--;
+      }
+   } 
+   else {
+      while (r200GetLastFrame (rmesa) < sarea->last_frame) {
+	 UNLOCK_HARDWARE( rmesa ); 
+	 if (rmesa->do_usleeps) 
+	    do_usleep(1, __FUNCTION__); 
+	 LOCK_HARDWARE( rmesa ); 
+      }
+   }
+}
+
+
+
+/* Copy the back color buffer to the front color buffer.
+ */
+void r200CopyBuffer( const __DRIdrawablePrivate *dPriv )
+{
+   r200ContextPtr rmesa;
+   GLint nbox, i, ret;
+
+   assert(dPriv);
+   assert(dPriv->driContextPriv);
+   assert(dPriv->driContextPriv->driverPrivate);
+
+   rmesa = (r200ContextPtr) dPriv->driContextPriv->driverPrivate;
+
+   if ( R200_DEBUG & DEBUG_IOCTL ) {
+      fprintf( stderr, "\n%s( %p )\n\n", __FUNCTION__, rmesa->glCtx );
+   }
+
+   R200_FIREVERTICES( rmesa );
+
+   LOCK_HARDWARE( rmesa );
+
+
+   /* Throttle the frame rate -- only allow one pending swap buffers
+    * request at a time.
+    */
+   r200WaitForFrameCompletion( rmesa );
+
+   r200WaitForVBlank( rmesa );
+
+   nbox = rmesa->dri.drawable->numClipRects; /* must be in locked region */
+
+   for ( i = 0 ; i < nbox ; ) {
+      GLint nr = MIN2( i + RADEON_NR_SAREA_CLIPRECTS , nbox );
+      XF86DRIClipRectPtr box = rmesa->dri.drawable->pClipRects;
+      XF86DRIClipRectPtr b = rmesa->sarea->boxes;
+      GLint n = 0;
+
+      for ( ; i < nr ; i++ ) {
+	 *b++ = box[i];
+	 n++;
+      }
+      rmesa->sarea->nbox = n;
+
+      ret = drmCommandNone( rmesa->dri.fd, DRM_RADEON_SWAP );
+
+      if ( ret ) {
+	 fprintf( stderr, "DRM_R200_SWAP_BUFFERS: return = %d\n", ret );
+	 UNLOCK_HARDWARE( rmesa );
+	 exit( 1 );
+      }
+   }
+
+   UNLOCK_HARDWARE( rmesa );
+   rmesa->lost_context = 1;
+}
+
+void r200PageFlip( const __DRIdrawablePrivate *dPriv )
+{
+   r200ContextPtr rmesa;
+   GLint ret;
+
+   assert(dPriv);
+   assert(dPriv->driContextPriv);
+   assert(dPriv->driContextPriv->driverPrivate);
+
+   rmesa = (r200ContextPtr) dPriv->driContextPriv->driverPrivate;
+
+   if ( R200_DEBUG & DEBUG_IOCTL ) {
+      fprintf(stderr, "%s: pfCurrentPage: %d\n", __FUNCTION__,
+	      rmesa->sarea->pfCurrentPage);
+   }
+
+   R200_FIREVERTICES( rmesa );
+   LOCK_HARDWARE( rmesa );
+
+   if (!rmesa->dri.drawable->numClipRects) {
+      UNLOCK_HARDWARE( rmesa );
+      usleep( 10000 );		/* throttle invisible client 10ms */
+      return;
+   }
+
+   /* Need to do this for the perf box placement:
+    */
+   {
+      XF86DRIClipRectPtr box = rmesa->dri.drawable->pClipRects;
+      XF86DRIClipRectPtr b = rmesa->sarea->boxes;
+      b[0] = box[0];
+      rmesa->sarea->nbox = 1;
+   }
+
+   /* Throttle the frame rate -- only allow a few pending swap buffers
+    * request at a time.
+    */
+   r200WaitForFrameCompletion( rmesa );
+
+   r200WaitForVBlank( rmesa );
+
+   ret = drmCommandNone( rmesa->dri.fd, DRM_RADEON_FLIP );
+
+   UNLOCK_HARDWARE( rmesa );
+
+   if ( ret ) {
+      fprintf( stderr, "DRM_R200_FLIP: return = %d\n", ret );
+      exit( 1 );
+   }
+
+   if ( rmesa->sarea->pfCurrentPage == 1 ) {
+	 rmesa->state.color.drawOffset = rmesa->r200Screen->frontOffset;
+	 rmesa->state.color.drawPitch  = rmesa->r200Screen->frontPitch;
+   } else {
+	 rmesa->state.color.drawOffset = rmesa->r200Screen->backOffset;
+	 rmesa->state.color.drawPitch  = rmesa->r200Screen->backPitch;
+   }
+
+   R200_STATECHANGE( rmesa, ctx );
+   rmesa->hw.ctx.cmd[CTX_RB3D_COLOROFFSET] = rmesa->state.color.drawOffset;
+   rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH]  = rmesa->state.color.drawPitch;
+}
+
+
+/* ================================================================
+ * Buffer clear
+ */
+static void r200Clear( GLcontext *ctx, GLbitfield mask, GLboolean all,
+			 GLint cx, GLint cy, GLint cw, GLint ch )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;
+   GLuint flags = 0;
+   GLuint color_mask = 0;
+   GLint ret, i;
+
+   if ( R200_DEBUG & DEBUG_IOCTL ) {
+      fprintf( stderr, "%s:  all=%d cx=%d cy=%d cw=%d ch=%d\n",
+	       __FUNCTION__, all, cx, cy, cw, ch );
+   }
+
+   {
+      LOCK_HARDWARE( rmesa );
+      UNLOCK_HARDWARE( rmesa );
+      if ( dPriv->numClipRects == 0 ) 
+	 return;
+   }
+
+   r200EmitState( rmesa );
+
+   /* Need to cope with lostcontext here as kernel relies on
+    * some residual state:
+    */
+   R200_FIREVERTICES( rmesa ); 
+
+   if ( mask & DD_FRONT_LEFT_BIT ) {
+      flags |= RADEON_FRONT;
+      color_mask = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK];
+      mask &= ~DD_FRONT_LEFT_BIT;
+   }
+
+   if ( mask & DD_BACK_LEFT_BIT ) {
+      flags |= RADEON_BACK;
+      color_mask = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK];
+      mask &= ~DD_BACK_LEFT_BIT;
+   }
+
+   if ( mask & DD_DEPTH_BIT ) {
+      if ( ctx->Depth.Mask ) flags |= RADEON_DEPTH; /* FIXME: ??? */
+      mask &= ~DD_DEPTH_BIT;
+   }
+
+   if ( (mask & DD_STENCIL_BIT) && rmesa->state.stencil.hwBuffer ) {
+      flags |= RADEON_STENCIL;
+      mask &= ~DD_STENCIL_BIT;
+   }
+
+   if ( mask )
+      _swrast_Clear( ctx, mask, all, cx, cy, cw, ch );
+
+   if ( !flags ) 
+      return;
+
+   /* Flip top to bottom */
+   cx += dPriv->x;
+   cy  = dPriv->y + dPriv->h - cy - ch;
+
+   LOCK_HARDWARE( rmesa );
+
+   /* Throttle the number of clear ioctls we do.
+    */
+   while ( 1 ) {
+      drmRadeonGetParam gp;
+      int ret;
+      int clear;
+
+      gp.param = RADEON_PARAM_LAST_CLEAR;
+      gp.value = (int *)&clear;
+      ret = drmCommandWriteRead( rmesa->dri.fd,
+		      DRM_RADEON_GETPARAM, &gp, sizeof(gp) );
+
+      if ( ret ) {
+	 fprintf( stderr, "%s: drmRadeonGetParam: %d\n", __FUNCTION__, ret );
+	 exit(1);
+      }
+
+      /* Clear throttling needs more thought.
+       */
+      if ( rmesa->sarea->last_clear - clear <= 25 ) {
+	 break;
+      }
+      
+      if (rmesa->do_usleeps) {
+	 UNLOCK_HARDWARE( rmesa );
+	 do_usleep(1, __FUNCTION__);
+	 LOCK_HARDWARE( rmesa );
+      }
+   }
+
+
+   for ( i = 0 ; i < dPriv->numClipRects ; ) {
+      GLint nr = MIN2( i + RADEON_NR_SAREA_CLIPRECTS, dPriv->numClipRects );
+      XF86DRIClipRectPtr box = dPriv->pClipRects;
+      XF86DRIClipRectPtr b = rmesa->sarea->boxes;
+      drmRadeonClearType clear;
+      drmRadeonClearRect depth_boxes[RADEON_NR_SAREA_CLIPRECTS];
+      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++;
+	 }
+      }
+
+      rmesa->sarea->nbox = n;
+
+      clear.flags       = flags;
+      clear.clear_color = rmesa->state.color.clear;
+      clear.clear_depth = 0;	/* not used */
+      clear.color_mask  = rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK];
+      clear.depth_mask  = rmesa->state.stencil.clear;
+      clear.depth_boxes = depth_boxes;
+
+      n--;
+      b = rmesa->sarea->boxes;
+      for ( ; n >= 0 ; n-- ) {
+	 depth_boxes[n].f[RADEON_CLEAR_X1] = (float)b[n].x1;
+	 depth_boxes[n].f[RADEON_CLEAR_Y1] = (float)b[n].y1;
+	 depth_boxes[n].f[RADEON_CLEAR_X2] = (float)b[n].x2;
+	 depth_boxes[n].f[RADEON_CLEAR_Y2] = (float)b[n].y2;
+	 depth_boxes[n].f[RADEON_CLEAR_DEPTH] = ctx->Depth.Clear;
+      }
+
+      ret = drmCommandWrite( rmesa->dri.fd, DRM_RADEON_CLEAR,
+			     &clear, sizeof(drmRadeonClearType));
+
+
+      if ( ret ) {
+	 UNLOCK_HARDWARE( rmesa );
+	 fprintf( stderr, "DRM_RADEON_CLEAR: return = %d\n", ret );
+	 exit( 1 );
+      }
+   }
+
+   UNLOCK_HARDWARE( rmesa );
+   rmesa->lost_context = 1;
+}
+
+
+void r200WaitForIdleLocked( r200ContextPtr rmesa )
+{
+    int ret;
+    int i = 0;
+    
+    do {
+       ret = drmCommandNone( rmesa->dri.fd, DRM_RADEON_CP_IDLE);
+       if (ret) 
+	  do_usleep( 1, __FUNCTION__ );
+    } while (ret && ++i < 100);
+    
+    if ( ret < 0 ) {
+       UNLOCK_HARDWARE( rmesa );
+       fprintf( stderr, "Error: R200 timed out... exiting\n" );
+       exit( -1 );
+    }
+}
+
+static void r200WaitForIdle( r200ContextPtr rmesa )
+{
+   LOCK_HARDWARE(rmesa);
+   r200WaitForIdleLocked( rmesa );
+   UNLOCK_HARDWARE(rmesa);
+}
+
+
+void r200WaitForVBlank( r200ContextPtr rmesa )
+{
+#if 0
+    drmVBlank vbl;
+    int ret;
+
+    if ( !rmesa->r200Screen->irq )
+	return;
+
+    if ( getenv("LIBGL_SYNC_REFRESH") ) {
+	/* Wait for until the next vertical blank */
+	vbl.request.type = DRM_VBLANK_RELATIVE;
+	vbl.request.sequence = 1;
+    } else if ( getenv("LIBGL_THROTTLE_REFRESH") ) {
+	/* Wait for at least one vertical blank since the last call */
+	vbl.request.type = DRM_VBLANK_ABSOLUTE;
+	vbl.request.sequence = rmesa->vbl_seq + 1;
+    } else {
+	return;
+    }
+
+    UNLOCK_HARDWARE( rmesa );
+
+    if ((ret = drmWaitVBlank( rmesa->dri.fd, &vbl ))) {
+	fprintf(stderr, "%s: drmWaitVBlank returned %d, IRQs don't seem to be"
+		" working correctly.\nTry running with LIBGL_THROTTLE_REFRESH"
+		" and LIBL_SYNC_REFRESH unset.\n", __FUNCTION__, ret);
+	exit(1);
+    } else if (R200_DEBUG & DEBUG_IOCTL)
+	fprintf(stderr, "%s: drmWaitVBlank returned %d\n", __FUNCTION__, ret);
+
+    rmesa->vbl_seq = vbl.reply.sequence;
+
+    LOCK_HARDWARE( rmesa );
+#endif
+}
+
+void r200Flush( GLcontext *ctx )
+{
+   r200ContextPtr rmesa = R200_CONTEXT( ctx );
+
+   if (R200_DEBUG & DEBUG_IOCTL)
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+   if (rmesa->dma.flush)
+      rmesa->dma.flush( rmesa );
+
+   if (!is_empty_list(&rmesa->hw.dirty)) 
+      r200EmitState( rmesa );
+   
+   if (rmesa->store.cmd_used)
+      r200FlushCmdBuf( rmesa, __FUNCTION__ );
+}
+
+/* Make sure all commands have been sent to the hardware and have
+ * completed processing.
+ */
+void r200Finish( GLcontext *ctx )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   r200Flush( ctx );
+
+   if (rmesa->do_irqs) {
+      LOCK_HARDWARE( rmesa );
+      r200EmitIrqLocked( rmesa );
+      UNLOCK_HARDWARE( rmesa );
+      r200WaitIrq( rmesa );
+   }
+   else 
+      r200WaitForIdle( rmesa );
+}
+
+
+/* This version of AllocateMemoryNV allocates only agp memory, and
+ * only does so after the point at which the driver has been
+ * initialized.
+ *
+ * Theoretically a valid context isn't required.  However, in this
+ * implementation, it is, as I'm using the hardware lock to protect
+ * the kernel data structures, and the current context to get the
+ * device fd.
+ */
+void *r200AllocateMemoryNV(GLsizei size, GLfloat readfreq,
+			    GLfloat writefreq, GLfloat priority)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   r200ContextPtr rmesa;
+   int region_offset;
+   drmRadeonMemAlloc alloc;
+   int ret;
+
+   if (R200_DEBUG & DEBUG_IOCTL)
+      fprintf(stderr, "%s sz %d %f/%f/%f\n", __FUNCTION__, size, readfreq, 
+	      writefreq, priority);
+
+   if (!ctx || !(rmesa = R200_CONTEXT(ctx)) || rmesa->r200Screen->IsPCI ) 
+      return NULL;
+
+   if (getenv("R200_NO_ALLOC"))
+      return NULL;
+   
+   if (rmesa->dri.drmMinor < 6) 
+      return NULL;
+      
+   alloc.region = RADEON_MEM_REGION_AGP;
+   alloc.alignment = 0;
+   alloc.size = size;
+   alloc.region_offset = &region_offset;
+
+   ret = drmCommandWriteRead( rmesa->r200Screen->driScreen->fd,
+			      DRM_RADEON_ALLOC,
+			      &alloc, sizeof(alloc));
+   
+   if (ret) {
+      fprintf(stderr, "%s: DRM_RADEON_ALLOC ret %d\n", __FUNCTION__, ret);
+      return NULL;
+   }
+   
+   {
+      char *region_start = (char *)rmesa->r200Screen->agpTextures.map;
+      return (void *)(region_start + region_offset);
+   }
+}
+
+
+/* Called via glXFreeMemoryNV() */
+void r200FreeMemoryNV(GLvoid *pointer)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   r200ContextPtr rmesa;
+   int region_offset;
+   drmRadeonMemFree memfree;
+   int ret;
+
+   if (R200_DEBUG & DEBUG_IOCTL)
+      fprintf(stderr, "%s %p\n", __FUNCTION__, pointer);
+
+   if (!ctx || !(rmesa = R200_CONTEXT(ctx)) || rmesa->r200Screen->IsPCI ) {
+      fprintf(stderr, "%s: no context\n", __FUNCTION__);
+      return;
+   }
+
+   if (rmesa->dri.drmMinor < 6) 
+      return;
+
+   region_offset = (char *)pointer - (char *)rmesa->r200Screen->agpTextures.map;
+
+   if (region_offset < 0 || 
+       region_offset > rmesa->r200Screen->agpTextures.size) {
+      fprintf(stderr, "offset %d outside range 0..%d\n", region_offset,
+	      rmesa->r200Screen->agpTextures.size);
+      return;
+   }
+
+   memfree.region = RADEON_MEM_REGION_AGP;
+   memfree.region_offset = region_offset;
+   
+   ret = drmCommandWrite( rmesa->r200Screen->driScreen->fd,
+			  DRM_RADEON_FREE,
+			  &memfree, sizeof(memfree));
+   
+   if (ret) 
+      fprintf(stderr, "%s: DRM_RADEON_FREE ret %d\n", __FUNCTION__, ret);
+}
+
+/* Called via glXGetAGPOffsetMESA() */
+GLuint r200GetAGPOffset(const GLvoid *pointer)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   r200ContextPtr rmesa;
+   GLuint card_offset;
+
+   if (!ctx || !(rmesa = R200_CONTEXT(ctx)) ) {
+      fprintf(stderr, "%s: no context\n", __FUNCTION__);
+      return ~0;
+   }
+
+   if (!r200IsAgpMemory( rmesa, pointer, 0 ))
+      return ~0;
+
+   if (rmesa->dri.drmMinor < 6) 
+      return ~0;
+
+   card_offset = r200AgpOffsetFromVirtual( rmesa, pointer );
+
+   return card_offset - rmesa->r200Screen->agp_base;
+}
+
+
+GLboolean r200IsAgpMemory( r200ContextPtr rmesa, const GLvoid *pointer,
+			   GLint size )
+{
+   int offset = (char *)pointer - (char *)rmesa->r200Screen->agpTextures.map;
+   int valid = (size >= 0 &&
+		offset >= 0 &&
+		offset + size < rmesa->r200Screen->agpTextures.size);
+
+   if (R200_DEBUG & DEBUG_IOCTL)
+      fprintf(stderr, "r200IsAgpMemory( %p ) : %d\n", pointer, valid );
+   
+   return valid;
+}
+
+
+GLuint r200AgpOffsetFromVirtual( r200ContextPtr rmesa, const GLvoid *pointer )
+{
+   int offset = (char *)pointer - (char *)rmesa->r200Screen->agpTextures.map;
+
+   if (offset < 0 || offset > rmesa->r200Screen->agpTextures.size)
+      return ~0;
+   else
+      return rmesa->r200Screen->agp_texture_offset + offset;
+}
+
+
+
+void r200InitIoctlFuncs( GLcontext *ctx )
+{
+    ctx->Driver.Clear = r200Clear;
+    ctx->Driver.Finish = r200Finish;
+    ctx->Driver.Flush = r200Flush;
+}
+

src/mesa/drivers/dri/r200/r200_ioctl.h

@@ -0,0 +1,188 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#ifndef __R200_IOCTL_H__
+#define __R200_IOCTL_H__
+
+#ifdef GLX_DIRECT_RENDERING
+
+#include "simple_list.h"
+#include "radeon_dri.h"
+#include "r200_lock.h"
+
+#include "xf86drm.h"
+#include "radeon_common.h"
+
+extern void r200EmitState( r200ContextPtr rmesa );
+extern void r200EmitVertexAOS( r200ContextPtr rmesa,
+				 GLuint vertex_size,
+				 GLuint offset );
+
+extern void r200EmitVbufPrim( r200ContextPtr rmesa,
+				GLuint primitive,
+				GLuint vertex_nr );
+
+extern void r200FlushElts( r200ContextPtr rmesa );
+
+extern GLushort *r200AllocEltsOpenEnded( r200ContextPtr rmesa,
+					   GLuint primitive,
+					   GLuint min_nr );
+
+extern void r200EmitAOS( r200ContextPtr rmesa,
+			   struct r200_dma_region **regions,
+			   GLuint n,
+			   GLuint offset );
+
+extern void r200EmitBlit( r200ContextPtr rmesa,
+			  GLuint color_fmt,
+			  GLuint src_pitch,
+			  GLuint src_offset,
+			  GLuint dst_pitch,
+			  GLuint dst_offset,
+			  GLint srcx, GLint srcy,
+			  GLint dstx, GLint dsty,
+			  GLuint w, GLuint h );
+
+extern void r200EmitWait( r200ContextPtr rmesa, GLuint flags );
+
+extern void r200FlushCmdBuf( r200ContextPtr rmesa, const char * );
+extern int r200FlushCmdBufLocked( r200ContextPtr rmesa, const char * caller );
+
+extern void r200RefillCurrentDmaRegion( r200ContextPtr rmesa );
+
+extern void r200AllocDmaRegion( r200ContextPtr rmesa,
+				  struct r200_dma_region *region,
+				  int bytes, 
+				  int alignment );
+
+extern void r200AllocDmaRegionVerts( r200ContextPtr rmesa,
+				       struct r200_dma_region *region,
+				       int numverts,
+				       int vertsize, 
+				       int alignment );
+
+extern void r200ReleaseDmaRegion( r200ContextPtr rmesa,
+				    struct r200_dma_region *region,
+				    const char *caller );
+
+extern void r200CopyBuffer( const __DRIdrawablePrivate *drawable );
+extern void r200PageFlip( const __DRIdrawablePrivate *drawable );
+extern void r200Flush( GLcontext *ctx );
+extern void r200Finish( GLcontext *ctx );
+extern void r200WaitForIdleLocked( r200ContextPtr rmesa );
+extern void r200WaitForVBlank( r200ContextPtr rmesa );
+extern void r200InitIoctlFuncs( GLcontext *ctx );
+
+extern void *r200AllocateMemoryNV( GLsizei size, GLfloat readfreq,
+				   GLfloat writefreq, GLfloat priority );
+extern void r200FreeMemoryNV( GLvoid *pointer );
+extern GLuint r200GetAGPOffset( const GLvoid *pointer );
+extern GLboolean r200IsAgpMemory( r200ContextPtr rmesa, const GLvoid *pointer,
+				  GLint size );
+
+extern GLuint r200AgpOffsetFromVirtual( r200ContextPtr rmesa, 
+					const GLvoid *pointer );
+
+/* ================================================================
+ * Helper macros:
+ */
+
+/* Close off the last primitive, if it exists.
+ */
+#define R200_NEWPRIM( rmesa )			\
+do {						\
+   if ( rmesa->dma.flush )			\
+      rmesa->dma.flush( rmesa );	\
+} while (0)
+
+/* Can accomodate several state changes and primitive changes without
+ * actually firing the buffer.
+ */
+#define R200_STATECHANGE( rmesa, ATOM )			\
+do {								\
+   R200_NEWPRIM( rmesa );					\
+   move_to_head( &(rmesa->hw.dirty), &(rmesa->hw.ATOM));	\
+} while (0)
+
+#define R200_DB_STATE( ATOM )			        \
+   memcpy( rmesa->hw.ATOM.lastcmd, rmesa->hw.ATOM.cmd,	\
+	   rmesa->hw.ATOM.cmd_size * 4)
+
+static __inline int R200_DB_STATECHANGE( 
+   r200ContextPtr rmesa,
+   struct r200_state_atom *atom )
+{
+   if (memcmp(atom->cmd, atom->lastcmd, atom->cmd_size*4)) {
+      int *tmp;
+      R200_NEWPRIM( rmesa );
+      move_to_head( &(rmesa->hw.dirty), atom );
+      tmp = atom->cmd; 
+      atom->cmd = atom->lastcmd;
+      atom->lastcmd = tmp;
+      return 1;
+   }
+   else
+      return 0;
+}
+
+
+/* Fire the buffered vertices no matter what.
+ */
+#define R200_FIREVERTICES( rmesa )			\
+do {							\
+   if ( rmesa->store.cmd_used || rmesa->dma.flush ) {	\
+      r200Flush( rmesa->glCtx );			\
+   }							\
+} while (0)
+
+/* Alloc space in the command buffer
+ */
+static __inline char *r200AllocCmdBuf( r200ContextPtr rmesa,
+					 int bytes, const char *where )
+{
+   if (rmesa->store.cmd_used + bytes > R200_CMD_BUF_SZ)
+      r200FlushCmdBuf( rmesa, __FUNCTION__ );
+   
+   {
+      char *head = rmesa->store.cmd_buf + rmesa->store.cmd_used;
+      rmesa->store.cmd_used += bytes;
+      return head;
+   }
+}
+
+
+
+
+#endif
+#endif /* __R200_IOCTL_H__ */

src/mesa/drivers/dri/r200/r200_lock.c

@@ -0,0 +1,118 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#include "r200_context.h"
+#include "r200_lock.h"
+#include "r200_tex.h"
+#include "r200_state.h"
+#include "r200_ioctl.h"
+
+#if DEBUG_LOCKING
+char *prevLockFile = NULL;
+int prevLockLine = 0;
+#endif
+
+/* Turn on/off page flipping according to the flags in the sarea:
+ */
+static void
+r200UpdatePageFlipping( r200ContextPtr rmesa )
+{
+   int use_back;
+   rmesa->doPageFlip = rmesa->sarea->pfAllowPageFlip;
+
+   use_back = (rmesa->glCtx->Color._DrawDestMask == BACK_LEFT_BIT);
+   use_back ^= (rmesa->sarea->pfCurrentPage == 1);
+   
+   if (use_back) {
+	 rmesa->state.color.drawOffset = rmesa->r200Screen->backOffset;
+	 rmesa->state.color.drawPitch  = rmesa->r200Screen->backPitch;
+   } else {
+	 rmesa->state.color.drawOffset = rmesa->r200Screen->frontOffset;
+	 rmesa->state.color.drawPitch  = rmesa->r200Screen->frontPitch;
+   }
+
+   R200_STATECHANGE( rmesa, ctx );
+   rmesa->hw.ctx.cmd[CTX_RB3D_COLOROFFSET] = rmesa->state.color.drawOffset;
+   rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH]  = rmesa->state.color.drawPitch;
+}
+
+
+
+/* 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 r200GetLock( r200ContextPtr rmesa, GLuint flags )
+{
+   __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;
+   __DRIscreenPrivate *sPriv = rmesa->dri.screen;
+   RADEONSAREAPrivPtr sarea = rmesa->sarea;
+   int i;
+
+   drmGetLock( rmesa->dri.fd, rmesa->dri.hwContext, 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( sPriv, dPriv );
+
+   if ( rmesa->lastStamp != dPriv->lastStamp ) {
+      r200UpdatePageFlipping( rmesa );
+      if (rmesa->glCtx->Color._DrawDestMask == BACK_LEFT_BIT)
+         r200SetCliprects( rmesa, GL_BACK_LEFT );
+      else
+         r200SetCliprects( rmesa, GL_FRONT_LEFT );
+      r200UpdateViewportOffset( rmesa->glCtx );
+      rmesa->lastStamp = dPriv->lastStamp;
+   }
+
+   if ( sarea->ctxOwner != rmesa->dri.hwContext ) {
+      sarea->ctxOwner = rmesa->dri.hwContext;
+   }
+
+   for ( i = 0 ; i < rmesa->texture.numHeaps ; i++ ) {
+      if ( sarea->texAge[i] != rmesa->texture.age[i] ) {
+	 r200AgeTextures( rmesa, i );
+      }
+   }
+}

src/mesa/drivers/dri/r200/r200_lock.h

@@ -0,0 +1,111 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#ifndef __R200_LOCK_H__
+#define __R200_LOCK_H__
+
+#ifdef GLX_DIRECT_RENDERING
+
+extern void r200GetLock( r200ContextPtr rmesa, 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( rmesa )					\
+   do {								\
+      char __ret = 0;						\
+      DEBUG_CHECK_LOCK();					\
+      DRM_CAS( rmesa->dri.hwLock, rmesa->dri.hwContext,		\
+	       (DRM_LOCK_HELD | rmesa->dri.hwContext), __ret );	\
+      if ( __ret )						\
+	 r200GetLock( rmesa, 0 );				\
+      DEBUG_LOCK();						\
+   } while (0)
+
+/* Unlock the hardware.
+ */
+#define UNLOCK_HARDWARE( rmesa )					\
+   do {									\
+      DRM_UNLOCK( rmesa->dri.fd,					\
+		  rmesa->dri.hwLock,					\
+		  rmesa->dri.hwContext );				\
+      DEBUG_RESET();							\
+   } while (0)
+
+#endif
+#endif /* __R200_LOCK_H__ */

src/mesa/drivers/dri/r200/r200_maos.c

@@ -0,0 +1,12 @@
+
+
+/* If using new packets, can choose either verts or arrays.
+ * Otherwise, must use verts.
+ */
+#include "r200_context.h"
+#define R200_MAOS_VERTS 0
+#if (R200_MAOS_VERTS) || (R200_OLD_PACKETS)
+#include "r200_maos_verts.c"
+#else
+#include "r200_maos_arrays.c"
+#endif

src/mesa/drivers/dri/r200/r200_maos.h

@@ -0,0 +1,46 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#ifndef __R200_MAOS_H__
+#define __R200_MAOS_H__
+
+#ifdef GLX_DIRECT_RENDERING
+
+#include "r200_context.h"
+
+extern void r200EmitArrays( GLcontext *ctx, GLuint inputs );
+extern void r200ReleaseArrays( GLcontext *ctx, GLuint newinputs );
+
+#endif
+#endif

src/mesa/drivers/dri/r200/r200_maos_arrays.c

@@ -0,0 +1,481 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#include "glheader.h"
+#include "mtypes.h"
+#include "colormac.h"
+#include "imports.h"
+#include "mmath.h"
+#include "macros.h"
+
+#include "swrast_setup/swrast_setup.h"
+#include "math/m_translate.h"
+#include "tnl/tnl.h"
+#include "tnl/t_context.h"
+#include "tnl/t_imm_debug.h"
+
+#include "r200_context.h"
+#include "r200_ioctl.h"
+#include "r200_state.h"
+#include "r200_swtcl.h"
+#include "r200_maos.h"
+
+/* Usage:
+ *   - from r200_tcl_render
+ *   - call r200EmitArrays to ensure uptodate arrays in dma
+ *   - emit primitives (new type?) which reference the data
+ *       -- need to use elts for lineloop, quads, quadstrip/flat
+ *       -- other primitives are all well-formed (need tristrip-1,fake-poly)
+ *
+ */
+static void emit_ubyte_rgba3( GLcontext *ctx,
+		       struct r200_dma_region *rvb,
+		       char *data,
+		       int stride,
+		       int count )
+{
+   int i;
+   r200_color_t *out = (r200_color_t *)(rvb->start + rvb->address);
+
+   if (R200_DEBUG & DEBUG_VERTS)
+      fprintf(stderr, "%s count %d stride %d out %p\n",
+	      __FUNCTION__, count, stride, out);
+
+   for (i = 0; i < count; i++) {
+      out->red   = *data;
+      out->green = *(data+1);
+      out->blue  = *(data+2);
+      out->alpha = 0xFF;
+      out++;
+      data += stride;
+   }
+}
+
+
+#if defined(USE_X86_ASM)
+#define COPY_DWORDS( dst, src, nr )					\
+do {									\
+	int __tmp;							\
+	__asm__ __volatile__( "rep ; movsl"				\
+			      : "=%c" (__tmp), "=D" (dst), "=S" (__tmp)	\
+			      : "0" (nr),				\
+			        "D" ((long)dst),			\
+			        "S" ((long)src) );			\
+} while (0)
+#else
+#define COPY_DWORDS( dst, src, nr )		\
+do {						\
+   int j;					\
+   for ( j = 0 ; j < nr ; j++ )			\
+      dst[j] = ((int *)src)[j];			\
+   dst += nr;					\
+} while (0)
+#endif
+
+
+
+static void emit_ubyte_rgba4( GLcontext *ctx,
+			      struct r200_dma_region *rvb,
+			      char *data,
+			      int stride,
+			      int count )
+{
+   int i;
+   int *out = (int *)(rvb->address + rvb->start);
+
+   if (R200_DEBUG & DEBUG_VERTS)
+      fprintf(stderr, "%s count %d stride %d\n",
+	      __FUNCTION__, count, stride);
+
+   if (stride == 4) {
+      for (i = 0; i < count; i++)
+	 ((int *)out)[i] = LE32_TO_CPU(((int *)data)[i]);
+   } else {
+      for (i = 0; i < count; i++) {
+	 *(int *)out++ = LE32_TO_CPU(*(int *)data);
+	 data += stride;
+      }
+   }
+}
+
+
+static void emit_ubyte_rgba( GLcontext *ctx,
+			     struct r200_dma_region *rvb,
+			     char *data,
+			     int size,
+			     int stride,
+			     int count )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+
+   if (R200_DEBUG & DEBUG_VERTS)
+      fprintf(stderr, "%s %d/%d\n", __FUNCTION__, count, size);
+
+   assert (!rvb->buf);
+
+   if (stride == 0) {
+      r200AllocDmaRegion( rmesa, rvb, 4, 4 );
+      count = 1;
+      rvb->aos_start = GET_START(rvb);
+      rvb->aos_stride = 0;
+      rvb->aos_size = 1;
+   }
+   else {
+      r200AllocDmaRegion( rmesa, rvb, 4 * count, 4 );	/* alignment? */
+      rvb->aos_start = GET_START(rvb);
+      rvb->aos_stride = 1;
+      rvb->aos_size = 1;
+   }
+
+   /* Emit the data
+    */
+   switch (size) {
+   case 3:
+      emit_ubyte_rgba3( ctx, rvb, data, stride, count );
+      break;
+   case 4:
+      emit_ubyte_rgba4( ctx, rvb, data, stride, count );
+      break;
+   default:
+      assert(0);
+      exit(1);
+      break;
+   }
+}
+
+
+
+
+static void emit_vec8( GLcontext *ctx,
+		       struct r200_dma_region *rvb,
+		       char *data,
+		       int stride,
+		       int count )
+{
+   int i;
+   int *out = (int *)(rvb->address + rvb->start);
+
+   if (R200_DEBUG & DEBUG_VERTS)
+      fprintf(stderr, "%s count %d stride %d\n",
+	      __FUNCTION__, count, stride);
+
+   if (stride == 8)
+      COPY_DWORDS( out, data, count*2 );
+   else
+      for (i = 0; i < count; i++) {
+	 out[0] = *(int *)data;
+	 out[1] = *(int *)(data+4);
+	 out += 2;
+	 data += stride;
+      }
+}
+
+static void emit_vec12( GLcontext *ctx,
+		       struct r200_dma_region *rvb,
+		       char *data,
+		       int stride,
+		       int count )
+{
+   int i;
+   int *out = (int *)(rvb->address + rvb->start);
+
+   if (R200_DEBUG & DEBUG_VERTS)
+      fprintf(stderr, "%s count %d stride %d out %p data %p\n",
+	      __FUNCTION__, count, stride, out, data);
+
+   if (stride == 12)
+      COPY_DWORDS( out, data, count*3 );
+   else
+      for (i = 0; i < count; i++) {
+	 out[0] = *(int *)data;
+	 out[1] = *(int *)(data+4);
+	 out[2] = *(int *)(data+8);
+	 out += 3;
+	 data += stride;
+      }
+}
+
+static void emit_vec16( GLcontext *ctx,
+			struct r200_dma_region *rvb,
+			char *data,
+			int stride,
+			int count )
+{
+   int i;
+   int *out = (int *)(rvb->address + rvb->start);
+
+   if (R200_DEBUG & DEBUG_VERTS)
+      fprintf(stderr, "%s count %d stride %d\n",
+	      __FUNCTION__, count, stride);
+
+   if (stride == 16)
+      COPY_DWORDS( out, data, count*4 );
+   else
+      for (i = 0; i < count; i++) {
+	 out[0] = *(int *)data;
+	 out[1] = *(int *)(data+4);
+	 out[2] = *(int *)(data+8);
+	 out[3] = *(int *)(data+12);
+	 out += 4;
+	 data += stride;
+      }
+}
+
+
+static void emit_vector( GLcontext *ctx,
+			 struct r200_dma_region *rvb,
+			 char *data,
+			 int size,
+			 int stride,
+			 int count )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+
+   if (R200_DEBUG & DEBUG_VERTS)
+      fprintf(stderr, "%s count %d size %d stride %d\n",
+	      __FUNCTION__, count, size, stride);
+
+   assert (!rvb->buf);
+
+   if (stride == 0) {
+      r200AllocDmaRegion( rmesa, rvb, size * 4, 4 );
+      count = 1;
+      rvb->aos_start = GET_START(rvb);
+      rvb->aos_stride = 0;
+      rvb->aos_size = size;
+   }
+   else {
+      r200AllocDmaRegion( rmesa, rvb, size * count * 4, 4 );	/* alignment? */
+      rvb->aos_start = GET_START(rvb);
+      rvb->aos_stride = size;
+      rvb->aos_size = size;
+   }
+
+   /* Emit the data
+    */
+   switch (size) {
+   case 2:
+      emit_vec8( ctx, rvb, data, stride, count );
+      break;
+   case 3:
+      emit_vec12( ctx, rvb, data, stride, count );
+      break;
+   case 4:
+      emit_vec16( ctx, rvb, data, stride, count );
+      break;
+   default:
+      assert(0);
+      exit(1);
+      break;
+   }
+
+}
+
+
+
+/* Emit any changed arrays to new agp memory, re-emit a packet to
+ * update the arrays.  
+ */
+void r200EmitArrays( GLcontext *ctx, GLuint inputs )
+{
+   r200ContextPtr rmesa = R200_CONTEXT( ctx );
+   struct vertex_buffer *VB = &TNL_CONTEXT( ctx )->vb;
+   struct r200_dma_region **component = rmesa->tcl.aos_components;
+   GLuint nr = 0;
+   GLuint vfmt0 = 0, vfmt1 = 0;
+   GLuint count = VB->Count;
+   
+   if (R200_DEBUG & DEBUG_VERTS) 
+      _tnl_print_vert_flags( __FUNCTION__, inputs );
+
+   if (1) {
+      if (!rmesa->tcl.obj.buf) 
+	 emit_vector( ctx, 
+		      &rmesa->tcl.obj, 
+		      (char *)VB->ObjPtr->data,
+		      VB->ObjPtr->size,
+		      VB->ObjPtr->stride,
+		      count);
+
+      switch( VB->ObjPtr->size ) {
+      case 4: vfmt0 |= R200_VTX_W0;
+      case 3: vfmt0 |= R200_VTX_Z0;
+      case 2: 
+      default:
+	break;
+      }
+      component[nr++] = &rmesa->tcl.obj;
+   }
+   
+
+   if (inputs & VERT_BIT_NORMAL) {
+      if (!rmesa->tcl.norm.buf)
+	 emit_vector( ctx, 
+		      &(rmesa->tcl.norm), 
+		      (char *)VB->NormalPtr->data,
+		      3,
+		      VB->NormalPtr->stride,
+		      count);
+
+      vfmt0 |= R200_VTX_N0;
+      component[nr++] = &rmesa->tcl.norm;
+   }
+
+   if (inputs & VERT_BIT_COLOR0) {
+      if (VB->ColorPtr[0]->Type == GL_UNSIGNED_BYTE) {
+	 if (!rmesa->tcl.rgba.buf)
+	    emit_ubyte_rgba( ctx, 
+			     &rmesa->tcl.rgba, 
+			     (char *)VB->ColorPtr[0]->Ptr,
+			     VB->ColorPtr[0]->Size,
+			     VB->ColorPtr[0]->StrideB,
+			     count);
+
+	 vfmt0 |= R200_VTX_PK_RGBA << R200_VTX_COLOR_0_SHIFT; 
+      }
+      else {
+	 int emitsize;
+
+	 if (VB->ColorPtr[0]->Size == 4 &&
+	     (VB->ColorPtr[0]->StrideB != 0 ||
+	      ((GLfloat *)VB->ColorPtr[0]->Ptr)[3] != 1.0)) { 
+	    vfmt0 |= R200_VTX_FP_RGBA << R200_VTX_COLOR_0_SHIFT; 
+	    emitsize = 4;
+	 }
+	 else { 
+	    vfmt0 |= R200_VTX_FP_RGB << R200_VTX_COLOR_0_SHIFT; 
+	    emitsize = 3;
+	 }
+
+	 if (!rmesa->tcl.rgba.buf)
+	    emit_vector( ctx, 
+			 &(rmesa->tcl.rgba), 
+			 (char *)VB->ColorPtr[0]->Ptr,
+			 emitsize,
+			 VB->ColorPtr[0]->StrideB,
+			 count);
+      }
+
+      component[nr++] = &rmesa->tcl.rgba;
+   }
+
+
+   if (inputs & VERT_BIT_COLOR1) {
+      if (!rmesa->tcl.spec.buf) {
+	 if (VB->SecondaryColorPtr[0]->Type != GL_UNSIGNED_BYTE)
+	    r200_import_float_spec_colors( ctx );
+
+	 emit_ubyte_rgba( ctx, 
+			  &rmesa->tcl.spec, 
+			  (char *)VB->SecondaryColorPtr[0]->Ptr,
+			  3,
+			  VB->SecondaryColorPtr[0]->StrideB,
+			  count);
+      }
+
+      /* How does this work?
+       */
+      vfmt0 |= R200_VTX_PK_RGBA << R200_VTX_COLOR_1_SHIFT; 
+      component[nr++] = &rmesa->tcl.spec;
+   }
+
+/*    vtx = (rmesa->hw.tcl.cmd[TCL_OUTPUT_VTXFMT] & */
+/* 	  ~(R200_TCL_VTX_Q0|R200_TCL_VTX_Q1)); */
+      
+   if (inputs & VERT_BIT_TEX0) {
+      if (!rmesa->tcl.tex[0].buf)
+	 emit_vector( ctx, 
+		      &(rmesa->tcl.tex[0]), 
+		      (char *)VB->TexCoordPtr[0]->data,
+		      VB->TexCoordPtr[0]->size,
+		      VB->TexCoordPtr[0]->stride,
+		      count );
+
+      vfmt1 |= VB->TexCoordPtr[0]->size << R200_VTX_TEX0_COMP_CNT_SHIFT;
+      component[nr++] = &rmesa->tcl.tex[0];
+   }
+
+   if (inputs & VERT_BIT_TEX1) {
+      if (!rmesa->tcl.tex[1].buf)
+	 emit_vector( ctx, 
+		      &(rmesa->tcl.tex[1]), 
+		      (char *)VB->TexCoordPtr[1]->data,
+		      VB->TexCoordPtr[1]->size,
+		      VB->TexCoordPtr[1]->stride,
+		      count );
+	 
+      vfmt1 |= VB->TexCoordPtr[1]->size << R200_VTX_TEX1_COMP_CNT_SHIFT;
+      component[nr++] = &rmesa->tcl.tex[1];
+   }
+
+   if (vfmt0 != rmesa->hw.vtx.cmd[VTX_VTXFMT_0] ||
+       vfmt1 != rmesa->hw.vtx.cmd[VTX_VTXFMT_1]) { 
+      R200_STATECHANGE( rmesa, vtx ); 
+      rmesa->hw.vtx.cmd[VTX_VTXFMT_0] = vfmt0;
+      rmesa->hw.vtx.cmd[VTX_VTXFMT_1] = vfmt1;
+   } 
+
+
+/*    fprintf(stderr, "VTXFMT_0: %x VTXFMT_1: %x\n", vfmt0, vfmt1); */
+    
+   rmesa->tcl.nr_aos_components = nr;
+   rmesa->tcl.vertex_format = vfmt0;
+}
+
+
+void r200ReleaseArrays( GLcontext *ctx, GLuint newinputs )
+{
+   r200ContextPtr rmesa = R200_CONTEXT( ctx );
+
+   if (R200_DEBUG & DEBUG_VERTS) 
+      _tnl_print_vert_flags( __FUNCTION__, newinputs );
+
+   if (newinputs & VERT_BIT_POS) 
+     r200ReleaseDmaRegion( rmesa, &rmesa->tcl.obj, __FUNCTION__ );
+
+   if (newinputs & VERT_BIT_NORMAL) 
+      r200ReleaseDmaRegion( rmesa, &rmesa->tcl.norm, __FUNCTION__ );
+
+   if (newinputs & VERT_BIT_COLOR0) 
+      r200ReleaseDmaRegion( rmesa, &rmesa->tcl.rgba, __FUNCTION__ );
+
+   if (newinputs & VERT_BIT_COLOR1) 
+      r200ReleaseDmaRegion( rmesa, &rmesa->tcl.spec, __FUNCTION__ );
+
+   if (newinputs & VERT_BIT_TEX0)
+      r200ReleaseDmaRegion( rmesa, &rmesa->tcl.tex[0], __FUNCTION__ );
+
+   if (newinputs & VERT_BIT_TEX1)
+      r200ReleaseDmaRegion( rmesa, &rmesa->tcl.tex[1], __FUNCTION__ );
+}

src/mesa/drivers/dri/r200/r200_maos_vbtmp.h

@@ -0,0 +1,378 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#ifndef LOCALVARS
+#define LOCALVARS
+#endif
+
+#undef TCL_DEBUG
+#ifndef TCL_DEBUG
+#define TCL_DEBUG 0
+#endif
+
+static void TAG(emit)( GLcontext *ctx,
+		       GLuint start, GLuint end,
+		       void *dest )
+{
+   LOCALVARS
+      struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+   GLuint (*tc0)[4], (*tc1)[4];
+   GLfloat *fog;
+   GLuint (*tc2)[4], (*norm)[3];
+   GLubyte (*col)[4], (*spec)[4];
+   GLuint tc0_stride, tc1_stride, col_stride, spec_stride, fog_stride;
+   GLuint tc2_stride, norm_stride;
+   GLuint (*coord)[4];
+   GLuint coord_stride;
+   GLubyte dummy[4];
+   int i;
+
+   union emit_union *v = (union emit_union *)dest;
+
+
+   if (R200_DEBUG & DEBUG_VERTS)
+      fprintf(stderr, "%s\n", __FUNCTION__); 
+
+   /* The vertex code expects Obj to be clean to element 3.  To fix
+    * this, add more vertex code (for obj-2, obj-3) or preferably move
+    * to maos.  
+    */
+   if (VB->ObjPtr->size < 3) {
+      if (VB->ObjPtr->flags & VEC_NOT_WRITEABLE) {
+	 VB->import_data( ctx, VERT_OBJ, VEC_NOT_WRITEABLE );
+      }
+      _mesa_vector4f_clean_elem( VB->ObjPtr, VB->Count, 2 );
+   }
+
+   if (DO_W && VB->ObjPtr->size < 4) {
+      if (VB->ObjPtr->flags & VEC_NOT_WRITEABLE) {
+	 VB->import_data( ctx, VERT_OBJ, VEC_NOT_WRITEABLE );
+      }
+      _mesa_vector4f_clean_elem( VB->ObjPtr, VB->Count, 3 );
+   }
+
+   coord = (GLuint (*)[4])VB->ObjPtr->data;
+   coord_stride = VB->ObjPtr->stride;
+
+   if (DO_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 < 4) {
+	 if (VB->TexCoordPtr[t2]->flags & VEC_NOT_WRITEABLE) {
+	    VB->import_data( ctx, VERT_TEX2, VEC_NOT_WRITEABLE );
+	 }
+	 _mesa_vector4f_clean_elem( VB->TexCoordPtr[t2], VB->Count, 3 );
+      }
+   }
+
+   if (DO_TEX1) {
+      if (VB->TexCoordPtr[1]) {
+	 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 < 4) {
+	    if (VB->TexCoordPtr[t1]->flags & VEC_NOT_WRITEABLE) {
+	       VB->import_data( ctx, VERT_TEX1, VEC_NOT_WRITEABLE );
+	    }
+	    _mesa_vector4f_clean_elem( VB->TexCoordPtr[t1], VB->Count, 3 );
+	 }
+      } else {
+	 tc1 = (GLuint (*)[4])&ctx->Current.Texcoord[1]; /* could be anything, really */
+	 tc1_stride = 0;
+      }
+   }
+
+   if (DO_TEX0) {
+      if (VB->TexCoordPtr[0]) {
+	 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 < 4) {
+	    if (VB->TexCoordPtr[t0]->flags & VEC_NOT_WRITEABLE) {
+	       VB->import_data( ctx, VERT_TEX0, VEC_NOT_WRITEABLE );
+	    }
+	    _mesa_vector4f_clean_elem( VB->TexCoordPtr[t0], VB->Count, 3 );
+	 }
+      } else {
+	 tc0 = (GLuint (*)[4])&ctx->Current.Texcoord[0]; /* could be anything, really */
+	 tc0_stride = 0;
+      }
+	 
+   }
+
+   if (DO_NORM) {
+      if (VB->NormalPtr) {
+	 norm_stride = VB->NormalPtr->stride;
+	 norm = (GLuint (*)[3])VB->NormalPtr->data;
+      } else {
+	 norm_stride = 0;
+	 norm = (GLuint (*)[3])&ctx->Current.Normal;
+      }
+   }
+
+   if (DO_RGBA) {
+      if (VB->ColorPtr[0]) {
+	 /* This is incorrect when colormaterial is enabled:
+	  */
+	 if (VB->ColorPtr[0]->Type != GL_UNSIGNED_BYTE) {
+	    if (0) fprintf(stderr, "IMPORTING FLOAT COLORS\n");
+	    IMPORT_FLOAT_COLORS( ctx );
+	 }
+	 col = (GLubyte (*)[4])VB->ColorPtr[0]->Ptr;
+	 col_stride = VB->ColorPtr[0]->StrideB;
+      } else {
+	 col = &dummy; /* any old memory is fine */
+	 col_stride = 0;
+      }
+      
+   }
+
+   if (DO_SPEC) {
+      if (VB->SecondaryColorPtr[0]) {
+	 if (VB->SecondaryColorPtr[0]->Type != GL_UNSIGNED_BYTE)
+	    IMPORT_FLOAT_SPEC_COLORS( ctx );
+	 spec = (GLubyte (*)[4])VB->SecondaryColorPtr[0]->Ptr;
+	 spec_stride = VB->SecondaryColorPtr[0]->StrideB;
+      } else {
+	 spec = &dummy;
+	 spec_stride = 0;
+      }
+	 
+   }
+
+   if (DO_FOG) {
+      if (VB->FogCoordPtr) {
+	 fog = VB->FogCoordPtr->data;
+	 fog_stride = VB->FogCoordPtr->stride;
+      } else {
+	 fog = (GLfloat *)&dummy; *fog = 0;
+	 fog_stride = 0;
+      }
+	      
+   }
+   
+   
+   if (VB->importable_data) {
+      if (start) {
+	 coord =  (GLuint (*)[4])((GLubyte *)coord + start * coord_stride);
+	 if (DO_TEX0)
+	    tc0 =  (GLuint (*)[4])((GLubyte *)tc0 + start * tc0_stride);
+	 if (DO_TEX1) 
+	    tc1 =  (GLuint (*)[4])((GLubyte *)tc1 + start * tc1_stride);
+	 if (DO_TEX2) 
+	    tc2 =  (GLuint (*)[4])((GLubyte *)tc2 + start * tc2_stride);
+	 if (DO_NORM) 
+	    norm =  (GLuint (*)[3])((GLubyte *)norm + start * norm_stride);
+	 if (DO_RGBA) 
+	    STRIDE_4UB(col, start * col_stride);
+	 if (DO_SPEC)
+	    STRIDE_4UB(spec, start * spec_stride);
+	 if (DO_FOG)
+	    STRIDE_F(fog, start * fog_stride);
+      }
+
+      for (i=start; i < end; i++) {
+	 v[0].ui = coord[0][0];
+	 v[1].ui = coord[0][1];
+	 v[2].ui = coord[0][2];
+	 if (TCL_DEBUG) fprintf(stderr, "%d: %.2f %.2f %.2f ", i, v[0].f, v[1].f, v[2].f);
+	 if (DO_W) {
+	    v[3].ui = coord[0][3];
+	    if (TCL_DEBUG) fprintf(stderr, "%.2f ", v[3].f);
+	    v += 4;
+	 } 
+	 else
+	    v += 3;
+	 coord =  (GLuint (*)[4])((GLubyte *)coord +  coord_stride);
+
+	 if (DO_NORM) {
+	    v[0].ui = norm[0][0];
+	    v[1].ui = norm[0][1];
+	    v[2].ui = norm[0][2];
+	    if (TCL_DEBUG) fprintf(stderr, "norm: %.2f %.2f %.2f ", v[0].f, v[1].f, v[2].f);
+	    v += 3;
+	    norm =  (GLuint (*)[3])((GLubyte *)norm +  norm_stride);
+	 }
+	 if (DO_RGBA) {
+	    v[0].ui = LE32_TO_CPU(*(GLuint *)&col[0]);
+	    STRIDE_4UB(col, col_stride);
+	    if (TCL_DEBUG) fprintf(stderr, "%x ", v[0].ui);
+	    v++;
+	 }
+	 if (DO_SPEC || DO_FOG) {
+	    if (DO_SPEC) {
+	       v[0].ub[0] = spec[0][0];
+	       v[0].ub[1] = spec[0][1];
+	       v[0].ub[2] = spec[0][2];
+	       STRIDE_4UB(spec, spec_stride);
+	    }
+	    if (DO_FOG) {
+	       v[0].ub[3] = fog[0] * 255.0;
+	       STRIDE_F(fog, fog_stride);
+	    }
+	    if (TCL_DEBUG) fprintf(stderr, "%x ", v[0].ui);
+	    v++;
+	 }
+	 if (DO_TEX0) {
+	    v[0].ui = tc0[0][0];
+	    v[1].ui = tc0[0][1];
+	    if (TCL_DEBUG) fprintf(stderr, "t0: %.2f %.2f ", v[0].f, v[1].f);
+	    if (DO_PTEX) {
+	       v[2].ui = tc0[0][3];
+	       if (TCL_DEBUG) fprintf(stderr, "%.2f ", v[2].f);
+	       v += 3;
+	    } 
+	    else
+	       v += 2;
+	    tc0 =  (GLuint (*)[4])((GLubyte *)tc0 +  tc0_stride);
+	 }
+	 if (DO_TEX1) {
+	    v[0].ui = tc1[0][0];
+	    v[1].ui = tc1[0][1];
+	    if (TCL_DEBUG) fprintf(stderr, "t1: %.2f %.2f ", v[0].f, v[1].f);
+	    if (DO_PTEX) {
+	       v[2].ui = tc1[0][3];
+	       if (TCL_DEBUG) fprintf(stderr, "%.2f ", v[2].f);
+	       v += 3;
+	    } 
+	    else
+	       v += 2;
+	    tc1 =  (GLuint (*)[4])((GLubyte *)tc1 +  tc1_stride);
+	 } 
+	 if (DO_TEX2) {
+	    v[0].ui = tc2[0][0];
+	    v[1].ui = tc2[0][1];
+	    if (DO_PTEX) {
+	       v[2].ui = tc2[0][3];
+	       v += 3;
+	    } 
+	    else
+	       v += 2;
+	    tc2 =  (GLuint (*)[4])((GLubyte *)tc2 +  tc2_stride);
+	 } 
+	 if (TCL_DEBUG) fprintf(stderr, "\n");
+      }
+   } else {
+      for (i=start; i < end; i++) {
+	 v[0].ui = coord[i][0];
+	 v[1].ui = coord[i][1];
+	 v[2].ui = coord[i][2];
+	 if (DO_W) {
+	    v[3].ui = coord[i][3];
+	    v += 4;
+	 } 
+	 else
+	    v += 3;
+
+	 if (DO_NORM) {
+	    v[0].ui = norm[i][0];
+	    v[1].ui = norm[i][1];
+	    v[2].ui = norm[i][2];
+	    v += 3;
+	 }
+	 if (DO_RGBA) {
+	    v[0].ui = LE32_TO_CPU(*(GLuint *)&col[i]);
+	    v++;
+	 }
+	 if (DO_SPEC || DO_FOG) {
+	    if (DO_SPEC) {
+	       v[0].ub[0] = spec[i][0];
+	       v[0].ub[1] = spec[i][1];
+	       v[0].ub[2] = spec[i][2];
+	    }
+	    if (DO_FOG) {
+	       v[0].ub[3] = fog[i] * 255.0;
+	    }
+	    v++;
+	 }
+	 if (DO_TEX0) {
+	    v[0].ui = tc0[i][0];
+	    v[1].ui = tc0[i][1];
+	    if (DO_PTEX) {
+	       v[2].ui = tc0[i][3];
+	       v += 3;
+	    } 
+	    else
+	       v += 2;
+	 }
+	 if (DO_TEX1) {
+	    v[0].ui = tc1[i][0];
+	    v[1].ui = tc1[i][1];
+	    if (DO_PTEX) {
+	       v[2].ui = tc1[i][3];
+	       v += 3;
+	    } 
+	    else
+	       v += 2;
+	 } 
+	 if (DO_TEX2) {
+	    v[0].ui = tc2[i][0];
+	    v[1].ui = tc2[i][1];
+	    if (DO_PTEX) {
+	       v[2].ui = tc2[i][3];
+	       v += 3;
+	    } 
+	    else
+	       v += 2;
+	 } 
+      }
+   }
+}
+
+
+
+static void TAG(init)( void )
+{
+   int sz = 3;
+   if (DO_W) sz++;
+   if (DO_NORM) sz += 3;
+   if (DO_RGBA) sz++;
+   if (DO_SPEC || DO_FOG) sz++;
+   if (DO_TEX0) sz += 2;
+   if (DO_TEX0 && DO_PTEX) sz++;
+   if (DO_TEX1) sz += 2;
+   if (DO_TEX1 && DO_PTEX) sz++;
+   if (DO_TEX2) sz += 2;
+   if (DO_TEX2 && DO_PTEX) sz++;
+
+   setup_tab[IDX].emit = TAG(emit);
+   setup_tab[IDX].vertex_format = IND;
+   setup_tab[IDX].vertex_size = sz;
+}
+
+
+#undef IND
+#undef TAG
+#undef IDX

src/mesa/drivers/dri/r200/r200_maos_verts.c

@@ -0,0 +1,338 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#include "glheader.h"
+#include "imports.h"
+#include "mmath.h"
+#include "mtypes.h"
+#include "enums.h"
+#include "colormac.h"
+#include "light.h"
+
+#include "array_cache/acache.h"
+#include "tnl/tnl.h"
+#include "tnl/t_pipeline.h"
+#include "tnl/t_imm_debug.h"
+
+#include "r200_context.h"
+#include "r200_state.h"
+#include "r200_ioctl.h"
+#include "r200_tex.h"
+#include "r200_tcl.h"
+#include "r200_swtcl.h"
+#include "r200_maos.h"
+
+
+#define R200_TCL_MAX_SETUP 13
+
+union emit_union { float f; GLuint ui; GLubyte ub[4]; };
+
+static struct {
+   void   (*emit)( GLcontext *, GLuint, GLuint, void * );
+   GLuint vertex_size;
+   GLuint vertex_format;
+} setup_tab[R200_TCL_MAX_SETUP];
+
+#define DO_W    (IND & R200_CP_VC_FRMT_W0)
+#define DO_RGBA (IND & R200_CP_VC_FRMT_PKCOLOR)
+#define DO_SPEC (IND & R200_CP_VC_FRMT_PKSPEC)
+#define DO_FOG  (IND & R200_CP_VC_FRMT_PKSPEC)
+#define DO_TEX0 (IND & R200_CP_VC_FRMT_ST0)
+#define DO_TEX1 (IND & R200_CP_VC_FRMT_ST1)
+#define DO_PTEX (IND & R200_CP_VC_FRMT_Q0)
+#define DO_NORM (IND & R200_CP_VC_FRMT_N0)
+
+#define DO_TEX2 0
+#define DO_TEX3 0
+
+#define GET_TEXSOURCE(n)  n
+#define GET_UBYTE_COLOR_STORE() &R200_CONTEXT(ctx)->UbyteColor
+#define GET_UBYTE_SPEC_COLOR_STORE() &R200_CONTEXT(ctx)->UbyteSecondaryColor
+
+#define IMPORT_FLOAT_COLORS r200_import_float_colors
+#define IMPORT_FLOAT_SPEC_COLORS r200_import_float_spec_colors
+
+/***********************************************************************
+ *             Generate vertex emit functions               *
+ ***********************************************************************/
+
+
+/* Defined in order of increasing vertex size:
+ */
+#define IDX 0
+#define IND (R200_CP_VC_FRMT_XY|		\
+	     R200_CP_VC_FRMT_Z|		\
+	     R200_CP_VC_FRMT_PKCOLOR)
+#define TAG(x) x##_rgba
+#include "r200_maos_vbtmp.h"
+
+#define IDX 1
+#define IND (R200_CP_VC_FRMT_XY|		\
+	     R200_CP_VC_FRMT_Z|		\
+	     R200_CP_VC_FRMT_N0)
+#define TAG(x) x##_n
+#include "r200_maos_vbtmp.h"
+
+#define IDX 2
+#define IND (R200_CP_VC_FRMT_XY|		\
+	     R200_CP_VC_FRMT_Z|		\
+	     R200_CP_VC_FRMT_PKCOLOR|		\
+	     R200_CP_VC_FRMT_ST0)
+#define TAG(x) x##_rgba_st
+#include "r200_maos_vbtmp.h"
+
+#define IDX 3
+#define IND (R200_CP_VC_FRMT_XY|		\
+	     R200_CP_VC_FRMT_Z|		\
+	     R200_CP_VC_FRMT_PKCOLOR|		\
+	     R200_CP_VC_FRMT_N0)
+#define TAG(x) x##_rgba_n
+#include "r200_maos_vbtmp.h"
+
+#define IDX 4
+#define IND (R200_CP_VC_FRMT_XY|		\
+	     R200_CP_VC_FRMT_Z|		\
+	     R200_CP_VC_FRMT_ST0|		\
+	     R200_CP_VC_FRMT_N0)
+#define TAG(x) x##_st_n
+#include "r200_maos_vbtmp.h"
+
+#define IDX 5
+#define IND (R200_CP_VC_FRMT_XY|		\
+	     R200_CP_VC_FRMT_Z|		\
+	     R200_CP_VC_FRMT_PKCOLOR|		\
+	     R200_CP_VC_FRMT_ST0|		\
+	     R200_CP_VC_FRMT_ST1)
+#define TAG(x) x##_rgba_st_st
+#include "r200_maos_vbtmp.h"
+
+#define IDX 6
+#define IND (R200_CP_VC_FRMT_XY|		\
+	     R200_CP_VC_FRMT_Z|		\
+	     R200_CP_VC_FRMT_PKCOLOR|		\
+	     R200_CP_VC_FRMT_ST0|		\
+	     R200_CP_VC_FRMT_N0)
+#define TAG(x) x##_rgba_st_n
+#include "r200_maos_vbtmp.h"
+
+#define IDX 7
+#define IND (R200_CP_VC_FRMT_XY|		\
+	     R200_CP_VC_FRMT_Z|		\
+	     R200_CP_VC_FRMT_PKCOLOR|		\
+	     R200_CP_VC_FRMT_PKSPEC|		\
+	     R200_CP_VC_FRMT_ST0|		\
+	     R200_CP_VC_FRMT_ST1)
+#define TAG(x) x##_rgba_spec_st_st
+#include "r200_maos_vbtmp.h"
+
+#define IDX 8
+#define IND (R200_CP_VC_FRMT_XY|		\
+	     R200_CP_VC_FRMT_Z|		\
+	     R200_CP_VC_FRMT_ST0|		\
+	     R200_CP_VC_FRMT_ST1|		\
+	     R200_CP_VC_FRMT_N0)
+#define TAG(x) x##_st_st_n
+#include "r200_maos_vbtmp.h"
+
+#define IDX 9
+#define IND (R200_CP_VC_FRMT_XY|		\
+	     R200_CP_VC_FRMT_Z|		\
+	     R200_CP_VC_FRMT_PKCOLOR|		\
+	     R200_CP_VC_FRMT_PKSPEC|		\
+	     R200_CP_VC_FRMT_ST0|		\
+	     R200_CP_VC_FRMT_ST1|		\
+	     R200_CP_VC_FRMT_N0)
+#define TAG(x) x##_rgpa_spec_st_st_n
+#include "r200_maos_vbtmp.h"
+
+#define IDX 10
+#define IND (R200_CP_VC_FRMT_XY|		\
+	     R200_CP_VC_FRMT_Z|		\
+	     R200_CP_VC_FRMT_PKCOLOR|		\
+	     R200_CP_VC_FRMT_ST0|		\
+	     R200_CP_VC_FRMT_Q0)
+#define TAG(x) x##_rgba_stq
+#include "r200_maos_vbtmp.h"
+
+#define IDX 11
+#define IND (R200_CP_VC_FRMT_XY|		\
+	     R200_CP_VC_FRMT_Z|		\
+	     R200_CP_VC_FRMT_PKCOLOR|		\
+	     R200_CP_VC_FRMT_ST1|		\
+	     R200_CP_VC_FRMT_Q1|		\
+	     R200_CP_VC_FRMT_ST0|		\
+	     R200_CP_VC_FRMT_Q0)
+#define TAG(x) x##_rgba_stq_stq
+#include "r200_maos_vbtmp.h"
+
+#define IDX 12
+#define IND (R200_CP_VC_FRMT_XY|		\
+	     R200_CP_VC_FRMT_Z|		\
+	     R200_CP_VC_FRMT_W0|		\
+	     R200_CP_VC_FRMT_PKCOLOR|		\
+	     R200_CP_VC_FRMT_PKSPEC|		\
+	     R200_CP_VC_FRMT_ST0|		\
+	     R200_CP_VC_FRMT_Q0|		\
+	     R200_CP_VC_FRMT_ST1|		\
+	     R200_CP_VC_FRMT_Q1|		\
+	     R200_CP_VC_FRMT_N0)
+#define TAG(x) x##_w_rgpa_spec_stq_stq_n
+#include "r200_maos_vbtmp.h"
+
+
+
+
+
+/***********************************************************************
+ *                         Initialization 
+ ***********************************************************************/
+
+
+static void init_tcl_verts( void )
+{
+   init_rgba();
+   init_n();
+   init_rgba_n();
+   init_rgba_st();
+   init_st_n();
+   init_rgba_st_st();
+   init_rgba_st_n();
+   init_rgba_spec_st_st();
+   init_st_st_n();
+   init_rgpa_spec_st_st_n();
+   init_rgba_stq();
+   init_rgba_stq_stq();
+   init_w_rgpa_spec_stq_stq_n();
+}
+
+
+void r200EmitArrays( GLcontext *ctx, GLuint inputs )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+   GLuint req = 0;
+   GLuint vtx = (rmesa->hw.tcl.cmd[TCL_OUTPUT_VTXFMT] &
+		 ~(R200_TCL_VTX_Q0|R200_TCL_VTX_Q1));
+   int i;
+   static int firsttime = 1;
+
+   if (firsttime) {
+      init_tcl_verts();
+      firsttime = 0;
+   }
+		     
+   if (1) {
+      req |= R200_CP_VC_FRMT_Z;
+      if (VB->ObjPtr->size == 4) {
+	 req |= R200_CP_VC_FRMT_W0;
+      }
+   }
+
+   if (inputs & VERT_BIT_NORMAL) {
+      req |= R200_CP_VC_FRMT_N0;
+   }
+   
+   if (inputs & VERT_BIT_COLOR0) {
+      req |= R200_CP_VC_FRMT_PKCOLOR;
+   }
+
+   if (inputs & VERT_BIT_COLOR1) {
+      req |= R200_CP_VC_FRMT_PKSPEC;
+   }
+
+   if (inputs & VERT_BIT_TEX0) {
+      req |= R200_CP_VC_FRMT_ST0;
+
+      if (VB->TexCoordPtr[0]->size == 4) {
+	 req |= R200_CP_VC_FRMT_Q0;
+	 vtx |= R200_TCL_VTX_Q0;
+      }
+   }
+
+   if (inputs & VERT_BIT_TEX1) {
+      req |= R200_CP_VC_FRMT_ST1;
+
+      if (VB->TexCoordPtr[1]->size == 4) {
+	 req |= R200_CP_VC_FRMT_Q1;
+	 vtx |= R200_TCL_VTX_Q1;
+      }
+   }
+
+   if (vtx != rmesa->hw.tcl.cmd[TCL_OUTPUT_VTXFMT]) {
+      R200_STATECHANGE( rmesa, tcl );
+      rmesa->hw.tcl.cmd[TCL_OUTPUT_VTXFMT] = vtx;
+   }
+
+   for (i = 0 ; i < R200_TCL_MAX_SETUP ; i++) 
+      if ((setup_tab[i].vertex_format & req) == req) 
+	 break;
+
+   if (rmesa->tcl.vertex_format == setup_tab[i].vertex_format &&
+       rmesa->tcl.indexed_verts.buf)
+      return;
+
+   if (rmesa->tcl.indexed_verts.buf)
+      r200ReleaseArrays( ctx, ~0 );
+
+   r200AllocDmaRegionVerts( rmesa, 
+			      &rmesa->tcl.indexed_verts, 
+			      VB->Count,
+			      setup_tab[i].vertex_size * 4, 
+			      4);
+
+   setup_tab[i].emit( ctx, 0, VB->Count, 
+		      rmesa->tcl.indexed_verts.address + 
+		      rmesa->tcl.indexed_verts.start );
+
+   rmesa->tcl.vertex_format = setup_tab[i].vertex_format;
+   rmesa->tcl.indexed_verts.aos_start = GET_START( &rmesa->tcl.indexed_verts );
+   rmesa->tcl.indexed_verts.aos_size = setup_tab[i].vertex_size;
+   rmesa->tcl.indexed_verts.aos_stride = setup_tab[i].vertex_size;
+
+   rmesa->tcl.aos_components[0] = &rmesa->tcl.indexed_verts;
+   rmesa->tcl.nr_aos_components = 1;
+}
+
+
+
+void r200ReleaseArrays( GLcontext *ctx, GLuint newinputs )
+{
+   r200ContextPtr rmesa = R200_CONTEXT( ctx );
+
+   if (R200_DEBUG & DEBUG_VERTS) 
+      _tnl_print_vert_flags( __FUNCTION__, newinputs );
+
+   if (newinputs) 
+     r200ReleaseDmaRegion( rmesa, &rmesa->tcl.indexed_verts, __FUNCTION__ );
+}

src/mesa/drivers/dri/r200/r200_pixel.c

@@ -0,0 +1,494 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#include "glheader.h"
+#include "enums.h"
+#include "mtypes.h"
+#include "macros.h"
+#include "texutil.h"
+#include "swrast/swrast.h"
+
+#include "r200_context.h"
+#include "r200_ioctl.h"
+#include "r200_pixel.h"
+#include "r200_swtcl.h"
+
+
+
+static GLboolean
+check_color( const GLcontext *ctx, GLenum type, GLenum format,
+	     const struct gl_pixelstore_attrib *packing,
+	     const void *pixels, GLint sz, GLint pitch )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   GLuint cpp = rmesa->r200Screen->cpp;
+
+   if (R200_DEBUG & DEBUG_PIXEL)
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+   if (	(pitch & 63) ||
+	ctx->_ImageTransferState ||
+	packing->SwapBytes ||
+	packing->LsbFirst) {
+      if (R200_DEBUG & DEBUG_PIXEL)
+	 fprintf(stderr, "%s: failed 1\n", __FUNCTION__);
+      return GL_FALSE;
+   }
+
+   if ( type == GL_UNSIGNED_INT_8_8_8_8_REV && 
+	cpp == 4 && 
+	format == GL_BGRA ) {
+      if (R200_DEBUG & DEBUG_PIXEL)
+	 fprintf(stderr, "%s: passed 2\n", __FUNCTION__);
+      return GL_TRUE;
+   }
+
+   if (R200_DEBUG & DEBUG_PIXEL)
+      fprintf(stderr, "%s: failed\n", __FUNCTION__);
+
+   return GL_FALSE;
+}
+
+static GLboolean
+check_color_per_fragment_ops( const GLcontext *ctx )
+{
+   int result;
+   result = (!(     ctx->Color.AlphaEnabled || 
+		    ctx->Depth.Test ||
+		    ctx->Fog.Enabled ||
+		    ctx->Scissor.Enabled ||
+		    ctx->Stencil.Enabled ||
+		    !ctx->Color.ColorMask[0] ||
+		    !ctx->Color.ColorMask[1] ||
+		    !ctx->Color.ColorMask[2] ||
+		    !ctx->Color.ColorMask[3] ||
+		    ctx->Color.ColorLogicOpEnabled ||
+		    ctx->Texture._EnabledUnits ||
+		    ctx->Depth.OcclusionTest
+           ) &&
+	   ctx->Current.RasterPosValid);
+   
+   return result;
+}
+
+
+
+static GLboolean
+clip_pixelrect( const GLcontext *ctx,
+		const GLframebuffer *buffer,
+		GLint *x, GLint *y,
+		GLsizei *width, GLsizei *height,
+		GLint *size )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+
+   /* left clipping */
+   if (*x < buffer->_Xmin) {
+      *width -= (buffer->_Xmin - *x);
+      *x = buffer->_Xmin;
+   }
+
+   /* right clipping */
+   if (*x + *width > buffer->_Xmax)
+      *width -= (*x + *width - buffer->_Xmax - 1);
+
+   if (*width <= 0)
+      return GL_FALSE;
+
+   /* bottom clipping */
+   if (*y < buffer->_Ymin) {
+      *height -= (buffer->_Ymin - *y);
+      *y = buffer->_Ymin;
+   }
+
+   /* top clipping */
+   if (*y + *height > buffer->_Ymax)
+      *height -= (*y + *height - buffer->_Ymax - 1);
+
+   if (*height <= 0)
+      return GL_FALSE;
+
+   *size = ((*y + *height - 1) * rmesa->r200Screen->frontPitch +
+	    (*x + *width - 1) * rmesa->r200Screen->cpp);
+
+   return GL_TRUE;
+}
+
+static GLboolean
+r200TryReadPixels( GLcontext *ctx,
+		  GLint x, GLint y, GLsizei width, GLsizei height,
+		  GLenum format, GLenum type,
+		  const struct gl_pixelstore_attrib *pack,
+		  GLvoid *pixels )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   GLint size;
+   GLint pitch = pack->RowLength ? pack->RowLength : width;
+   GLint blit_format;
+
+   if (R200_DEBUG & DEBUG_PIXEL)
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+   /* Only accelerate reading to agp buffers.
+    */
+   if ( !r200IsAgpMemory(rmesa, pixels, 
+			 pitch * height * rmesa->r200Screen->cpp ) ) {
+      if (R200_DEBUG & DEBUG_PIXEL)
+	 fprintf(stderr, "%s: dest not agp\n", __FUNCTION__);
+      return GL_FALSE;
+   }
+
+   /* Need GL_PACK_INVERT_MESA to cope with upsidedown results from
+    * blitter:
+    */
+   if (!pack->Invert) {
+      if (R200_DEBUG & DEBUG_PIXEL)
+	 fprintf(stderr, "%s: MESA_PACK_INVERT not set\n", __FUNCTION__);
+      return GL_FALSE;
+   }
+
+   if (!check_color(ctx, type, format, pack, pixels, size, pitch))
+      return GL_FALSE;
+
+   switch ( rmesa->r200Screen->cpp ) {
+   case 4:
+      blit_format = R200_CP_COLOR_FORMAT_ARGB8888;
+      break;
+   default:
+      return GL_FALSE;
+   }
+
+
+   /* Although the blits go on the command buffer, need to do this and
+    * fire with lock held to guarentee cliprects and drawOffset are
+    * correct.
+    *
+    * This is an unusual situation however, as the code which flushes
+    * a full command buffer expects to be called unlocked.  As a
+    * workaround, immediately flush the buffer on aquiring the lock.
+    */
+   LOCK_HARDWARE( rmesa );
+
+   if (rmesa->store.cmd_used)
+      r200FlushCmdBufLocked( rmesa, __FUNCTION__ );
+
+   if (!clip_pixelrect(ctx, ctx->ReadBuffer, &x, &y, &width, &height,
+		       &size)) {
+      UNLOCK_HARDWARE( rmesa );
+      if (R200_DEBUG & DEBUG_PIXEL)
+	 fprintf(stderr, "%s totally clipped -- nothing to do\n",
+		 __FUNCTION__);
+      return GL_TRUE;
+   }
+
+   {
+      __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;
+      int nbox = dPriv->numClipRects;
+      int src_offset = rmesa->state.color.drawOffset;
+      int src_pitch = rmesa->state.color.drawPitch * rmesa->r200Screen->cpp;
+      int dst_offset = r200AgpOffsetFromVirtual( rmesa, pixels);
+      int dst_pitch = pitch * rmesa->r200Screen->cpp;
+      XF86DRIClipRectRec *box = dPriv->pClipRects;
+      int i;
+
+      r200EmitWait( rmesa, RADEON_WAIT_3D ); 
+
+      y = dPriv->h - y - height;
+      x += dPriv->x;
+      y += dPriv->y;
+
+
+      if (R200_DEBUG & DEBUG_PIXEL)
+	 fprintf(stderr, "readpixel blit src_pitch %d dst_pitch %d\n",
+		 src_pitch, dst_pitch);
+
+      for (i = 0 ; i < nbox ; i++)
+      {
+	 GLint bx = box[i].x1;
+	 GLint by = box[i].y1;
+	 GLint bw = box[i].x2 - bx;
+	 GLint bh = box[i].y2 - by;
+	 
+	 if (bx < x) bw -= x - bx, bx = x;
+	 if (by < y) bh -= y - by, by = y;
+	 if (bx + bw > x + width) bw = x + width - bx;
+	 if (by + bh > y + height) bh = y + height - by;
+	 if (bw <= 0) continue;
+	 if (bh <= 0) continue;
+
+	 r200EmitBlit( rmesa,
+		       blit_format,
+		       src_pitch, src_offset,
+		       dst_pitch, dst_offset,
+		       bx, by,
+		       bx - x, by - y,
+		       bw, bh );
+      }
+
+      r200FlushCmdBufLocked( rmesa, __FUNCTION__ );
+   }
+   UNLOCK_HARDWARE( rmesa );
+
+   r200Finish( ctx ); /* required by GL */
+
+   return GL_TRUE;
+}
+
+static void
+r200ReadPixels( GLcontext *ctx,
+		 GLint x, GLint y, GLsizei width, GLsizei height,
+		 GLenum format, GLenum type,
+		 const struct gl_pixelstore_attrib *pack,
+		 GLvoid *pixels )
+{
+   if (R200_DEBUG & DEBUG_PIXEL)
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+   if (!r200TryReadPixels( ctx, x, y, width, height, format, type, pack, 
+			   pixels))
+      _swrast_ReadPixels( ctx, x, y, width, height, format, type, pack, 
+			  pixels);
+}
+
+
+
+
+static void do_draw_pix( GLcontext *ctx,
+			 GLint x, GLint y, GLsizei width, GLsizei height,
+			 GLint pitch,
+			 const void *pixels,
+			 GLuint dest, GLuint planemask)
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;
+   XF86DRIClipRectPtr box = dPriv->pClipRects;
+   int nbox = dPriv->numClipRects;
+   int i;
+   int blit_format;
+   int size;
+   int src_offset = r200AgpOffsetFromVirtual( rmesa, pixels);
+   int src_pitch = pitch * rmesa->r200Screen->cpp;
+
+   if (R200_DEBUG & DEBUG_PIXEL)
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+   switch ( rmesa->r200Screen->cpp ) {
+   case 2:
+      blit_format = R200_CP_COLOR_FORMAT_RGB565;
+      break;
+   case 4:
+      blit_format = R200_CP_COLOR_FORMAT_ARGB8888;
+      break;
+   default:
+      return;
+   }
+
+
+   LOCK_HARDWARE( rmesa );
+
+   if (rmesa->store.cmd_used)
+      r200FlushCmdBufLocked( rmesa, __FUNCTION__ );
+
+   y -= height;			/* cope with pixel zoom */
+   
+   if (!clip_pixelrect(ctx, ctx->DrawBuffer,
+		       &x, &y, &width, &height,
+		       &size)) {
+      UNLOCK_HARDWARE( rmesa );
+      return;
+   }
+
+   y = dPriv->h - y - height; 	/* convert from gl to hardware coords */
+   x += dPriv->x;
+   y += dPriv->y;
+
+
+   r200EmitWait( rmesa, RADEON_WAIT_3D );
+
+   for (i = 0 ; i < nbox ; i++ )
+   {
+      GLint bx = box[i].x1;
+      GLint by = box[i].y1;
+      GLint bw = box[i].x2 - bx;
+      GLint bh = box[i].y2 - by;
+
+      if (bx < x) bw -= x - bx, bx = x;
+      if (by < y) bh -= y - by, by = y;
+      if (bx + bw > x + width) bw = x + width - bx;
+      if (by + bh > y + height) bh = y + height - by;
+      if (bw <= 0) continue;
+      if (bh <= 0) continue;
+
+      r200EmitBlit( rmesa,
+		    blit_format,
+		    src_pitch, src_offset,
+		    rmesa->state.color.drawPitch * rmesa->r200Screen->cpp,
+		    rmesa->state.color.drawOffset,
+		    bx - x, by - y,
+		    bx, by,
+		    bw, bh );
+   }
+
+   r200FlushCmdBufLocked( rmesa, __FUNCTION__ );
+   r200WaitForIdleLocked( rmesa ); /* required by GL */
+   UNLOCK_HARDWARE( rmesa );
+}
+
+
+
+
+static GLboolean
+r200TryDrawPixels( GLcontext *ctx,
+		  GLint x, GLint y, GLsizei width, GLsizei height,
+		  GLenum format, GLenum type,
+		  const struct gl_pixelstore_attrib *unpack,
+		  const GLvoid *pixels )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   GLint pitch = unpack->RowLength ? unpack->RowLength : width;
+   GLuint dest, planemask;
+   GLuint cpp = rmesa->r200Screen->cpp;
+   GLint size = width * pitch * cpp;
+
+   if (R200_DEBUG & DEBUG_PIXEL)
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+   switch (format) {
+   case GL_RGB:
+   case GL_RGBA:
+   case GL_BGRA:
+      dest = rmesa->state.color.drawOffset;
+
+      planemask = r200PackColor(cpp,
+				ctx->Color.ColorMask[RCOMP],
+				ctx->Color.ColorMask[GCOMP],
+				ctx->Color.ColorMask[BCOMP],
+				ctx->Color.ColorMask[ACOMP]);
+
+      if (cpp == 2)
+	 planemask |= planemask << 16;
+
+      if (planemask != ~0)
+	 return GL_FALSE;	/* fix me -- should be possible */
+
+      /* Can't do conversions on agp reads/draws. 
+       */
+      if ( !r200IsAgpMemory( rmesa, pixels, size ) ) {
+	 if (R200_DEBUG & DEBUG_PIXEL)
+	    fprintf(stderr, "%s: not agp memory\n", __FUNCTION__);
+	 return GL_FALSE;
+      }
+
+      if (!check_color(ctx, type, format, unpack, pixels, size, pitch)) {
+	 return GL_FALSE;
+      }
+      if (!check_color_per_fragment_ops(ctx)) {
+	 return GL_FALSE;
+      }
+
+      if (ctx->Pixel.ZoomX != 1.0F ||
+	  ctx->Pixel.ZoomY != -1.0F)
+	 return GL_FALSE;
+      break;
+
+   default:
+      return GL_FALSE;
+   }
+
+   if ( r200IsAgpMemory(rmesa, pixels, size) )
+   {
+      do_draw_pix( ctx, x, y, width, height, pitch, pixels,
+		   dest, planemask );
+      return GL_TRUE;
+   }
+   else if (0)
+   {
+      /* Pixels is in regular memory -- get dma buffers and perform
+       * upload through them.
+       */
+   }
+   else
+      return GL_FALSE;
+}
+
+static void
+r200DrawPixels( GLcontext *ctx,
+		 GLint x, GLint y, GLsizei width, GLsizei height,
+		 GLenum format, GLenum type,
+		 const struct gl_pixelstore_attrib *unpack,
+		 const GLvoid *pixels )
+{
+   if (R200_DEBUG & DEBUG_PIXEL)
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+   if (!r200TryDrawPixels( ctx, x, y, width, height, format, type,
+			  unpack, pixels ))
+      _swrast_DrawPixels( ctx, x, y, width, height, format, type,
+			  unpack, pixels );
+}
+
+
+static void
+r200Bitmap( GLcontext *ctx, GLint px, GLint py,
+		  GLsizei width, GLsizei height,
+		  const struct gl_pixelstore_attrib *unpack,
+		  const GLubyte *bitmap )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+
+   if (rmesa->Fallback)
+      _swrast_Bitmap( ctx, px, py, width, height, unpack, bitmap );
+   else
+      r200PointsBitmap( ctx, px, py, width, height, unpack, bitmap );
+}
+
+
+
+void r200InitPixelFuncs( GLcontext *ctx )
+{
+   /* Pixel path fallbacks.
+    */
+   ctx->Driver.Accum = _swrast_Accum;
+   ctx->Driver.Bitmap = _swrast_Bitmap;
+   ctx->Driver.CopyPixels = _swrast_CopyPixels;
+   ctx->Driver.DrawPixels = _swrast_DrawPixels;
+   ctx->Driver.ReadPixels = _swrast_ReadPixels;
+
+   if (!getenv("R200_NO_BLITS") && R200_CONTEXT(ctx)->dri.drmMinor >= 6) {
+      ctx->Driver.ReadPixels = r200ReadPixels;  
+      ctx->Driver.DrawPixels = r200DrawPixels; 
+      if (getenv("R200_HW_BITMAP")) 
+	 ctx->Driver.Bitmap = r200Bitmap;
+   }
+}

src/mesa/drivers/dri/r200/r200_pixel.h

@@ -0,0 +1,43 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#ifndef __R200_PIXEL_H__
+#define __R200_PIXEL_H__
+
+#ifdef GLX_DIRECT_RENDERING
+
+extern void r200InitPixelFuncs( GLcontext *ctx );
+
+#endif
+#endif

src/mesa/drivers/dri/r200/r200_sanity.h

@@ -0,0 +1,11 @@
+#ifndef R200_SANITY_H
+#define R200_SANITY_H
+
+extern int r200SanityCmdBuffer( r200ContextPtr rmesa,
+				int nbox,
+				XF86DRIClipRectRec *boxes );
+
+
+extern int r200ValidateTexObjs( r200ContextPtr rmesa );
+
+#endif

src/mesa/drivers/dri/r200/r200_screen.c

@@ -0,0 +1,434 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#include <dlfcn.h>
+
+#include "glheader.h"
+#include "imports.h"
+#include "context.h"
+
+#include "r200_screen.h"
+#include "r200_context.h"
+#include "r200_ioctl.h"
+
+
+#if 1
+/* Including xf86PciInfo.h introduces a bunch of errors...
+ */
+#define PCI_CHIP_R200_QD	0x5144
+#define PCI_CHIP_R200_QE	0x5145
+#define PCI_CHIP_R200_QF	0x5146
+#define PCI_CHIP_R200_QG	0x5147
+#define PCI_CHIP_R200_QY	0x5159
+#define PCI_CHIP_R200_QZ	0x515A
+#define PCI_CHIP_R200_LW	0x4C57 
+#define PCI_CHIP_R200_LY	0x4C59
+#define PCI_CHIP_R200_LZ	0x4C5A
+#define PCI_CHIP_RV200_QW	0x5157 
+#endif
+
+static r200ScreenPtr __r200Screen;
+
+/* Create the device specific screen private data struct.
+ */
+static r200ScreenPtr 
+r200CreateScreen( __DRIscreenPrivate *sPriv )
+{
+   r200ScreenPtr r200Screen;
+   RADEONDRIPtr r200DRIPriv = (RADEONDRIPtr)sPriv->pDevPriv;
+
+   /* Check the DRI extension version */
+   if ( sPriv->driMajor != 4 || sPriv->driMinor < 0 ) {
+       __driUtilMessage( "R200 DRI driver expected DRI version 4.0.x "
+                         "but got version %d.%d.%d",
+                         sPriv->driMajor, sPriv->driMinor, sPriv->driPatch );
+       return NULL;
+   }
+
+   /* Check that the DDX driver version is compatible */
+   if ( sPriv->ddxMajor != 4 ||
+	sPriv->ddxMinor < 0 ) {
+      __driUtilMessage( "R200 DRI driver expected DDX driver version 4.0.x "
+			"but got version %d.%d.%d", 
+			sPriv->ddxMajor, sPriv->ddxMinor, sPriv->ddxPatch );
+      return NULL;
+   }
+
+   /* Check that the DRM driver version is compatible 
+    *    -- R200 support added at 1.5.0.
+    */
+   if ( sPriv->drmMajor != 1 ||
+	sPriv->drmMinor < 5) {
+      __driUtilMessage( "R200 DRI driver expected DRM driver version 1.5.x "
+			"but got version %d.%d.%d", 
+			sPriv->drmMajor, sPriv->drmMinor, sPriv->drmPatch );
+      return NULL;
+   }
+
+   /* Allocate the private area */
+   r200Screen = (r200ScreenPtr) CALLOC( sizeof(*r200Screen) );
+   if ( !r200Screen ) {
+      __driUtilMessage("%s: CALLOC r200Screen struct failed",
+		       __FUNCTION__);
+      return NULL;
+   }
+
+   /* Check if kernel module is new enough to support cube maps */
+   if (sPriv->drmMajor * 100 + sPriv->drmMinor >= 107)
+      r200Screen->drmSupportsCubeMaps = GL_TRUE;
+   else
+      r200Screen->drmSupportsCubeMaps = GL_FALSE;
+
+   switch ( r200DRIPriv->deviceID ) {
+   case PCI_CHIP_R200_QD:
+   case PCI_CHIP_R200_QE:
+   case PCI_CHIP_R200_QF:
+   case PCI_CHIP_R200_QG:
+   case PCI_CHIP_R200_QY:
+   case PCI_CHIP_R200_QZ:
+   case PCI_CHIP_RV200_QW:
+   case PCI_CHIP_R200_LW:
+   case PCI_CHIP_R200_LY:
+   case PCI_CHIP_R200_LZ:
+      __driUtilMessage("r200CreateScreen(): Device isn't an r200!\n");
+      FREE( r200Screen );
+      return NULL;      
+   default:
+      r200Screen->chipset = R200_CHIPSET_R200;
+      break;
+   }
+
+
+   /* This is first since which regions we map depends on whether or
+    * not we are using a PCI card.
+    */
+   r200Screen->IsPCI = r200DRIPriv->IsPCI;
+
+   {
+      int ret;
+      drmRadeonGetParam gp;
+
+      gp.param = RADEON_PARAM_AGP_BUFFER_OFFSET;
+      gp.value = &r200Screen->agp_buffer_offset;
+
+      ret = drmCommandWriteRead( sPriv->fd, DRM_RADEON_GETPARAM,
+				 &gp, sizeof(gp));
+      if (ret) {
+	 FREE( r200Screen );
+	 fprintf(stderr, "drmR200GetParam: %d\n", ret);
+	 return NULL;
+      }
+
+      r200Screen->agp_texture_offset = 
+	 r200Screen->agp_buffer_offset + 2*1024*1024;
+
+
+      if (sPriv->drmMinor >= 6) {
+	 gp.param = RADEON_PARAM_AGP_BASE;
+	 gp.value = &r200Screen->agp_base;
+
+	 ret = drmCommandWriteRead( sPriv->fd, DRM_RADEON_GETPARAM,
+				    &gp, sizeof(gp));
+	 if (ret) {
+	    FREE( r200Screen );
+	    fprintf(stderr, "drmR200GetParam (RADEON_PARAM_AGP_BUFFER_OFFSET): %d\n", ret);
+	    return NULL;
+	 }
+      }
+
+      if (sPriv->drmMinor >= 6) {
+	 gp.param = RADEON_PARAM_IRQ_NR;
+	 gp.value = &r200Screen->irq;
+
+	 ret = drmCommandWriteRead( sPriv->fd, DRM_RADEON_GETPARAM,
+				    &gp, sizeof(gp));
+	 if (ret) {
+	    FREE( r200Screen );
+	    fprintf(stderr, "drmR200GetParam (RADEON_PARAM_IRQ_NR): %d\n", ret);
+	    return NULL;
+	 }
+      }
+
+   }
+
+   r200Screen->mmio.handle = r200DRIPriv->registerHandle;
+   r200Screen->mmio.size   = r200DRIPriv->registerSize;
+   if ( drmMap( sPriv->fd,
+		r200Screen->mmio.handle,
+		r200Screen->mmio.size,
+		&r200Screen->mmio.map ) ) {
+      FREE( r200Screen );
+      __driUtilMessage("r200CreateScreen(): drmMap failed\n");
+      return NULL;
+   }
+
+   r200Screen->status.handle = r200DRIPriv->statusHandle;
+   r200Screen->status.size   = r200DRIPriv->statusSize;
+   if ( drmMap( sPriv->fd,
+		r200Screen->status.handle,
+		r200Screen->status.size,
+		&r200Screen->status.map ) ) {
+      drmUnmap( r200Screen->mmio.map, r200Screen->mmio.size );
+      FREE( r200Screen );
+      __driUtilMessage("r200CreateScreen(): drmMap (2) failed\n");
+      return NULL;
+   }
+   r200Screen->scratch = (__volatile__ GLuint *)
+      ((GLubyte *)r200Screen->status.map + RADEON_SCRATCH_REG_OFFSET);
+
+   r200Screen->buffers = drmMapBufs( sPriv->fd );
+   if ( !r200Screen->buffers ) {
+      drmUnmap( r200Screen->status.map, r200Screen->status.size );
+      drmUnmap( r200Screen->mmio.map, r200Screen->mmio.size );
+      FREE( r200Screen );
+      __driUtilMessage("r200CreateScreen(): drmMapBufs failed\n");
+      return NULL;
+   }
+
+   if ( !r200Screen->IsPCI ) {
+      r200Screen->agpTextures.handle = r200DRIPriv->agpTexHandle;
+      r200Screen->agpTextures.size   = r200DRIPriv->agpTexMapSize;
+      if ( drmMap( sPriv->fd,
+		   r200Screen->agpTextures.handle,
+		   r200Screen->agpTextures.size,
+		   (drmAddressPtr)&r200Screen->agpTextures.map ) ) {
+	 drmUnmapBufs( r200Screen->buffers );
+	 drmUnmap( r200Screen->status.map, r200Screen->status.size );
+	 drmUnmap( r200Screen->mmio.map, r200Screen->mmio.size );
+	 FREE( r200Screen );
+         __driUtilMessage("r200CreateScreen(): IsPCI failed\n");
+	 return NULL;
+      }
+   }
+
+
+
+   r200Screen->cpp = r200DRIPriv->bpp / 8;
+   r200Screen->AGPMode = r200DRIPriv->AGPMode;
+
+   r200Screen->frontOffset	= r200DRIPriv->frontOffset;
+   r200Screen->frontPitch	= r200DRIPriv->frontPitch;
+   r200Screen->backOffset	= r200DRIPriv->backOffset;
+   r200Screen->backPitch	= r200DRIPriv->backPitch;
+   r200Screen->depthOffset	= r200DRIPriv->depthOffset;
+   r200Screen->depthPitch	= r200DRIPriv->depthPitch;
+
+   r200Screen->texOffset[RADEON_CARD_HEAP] = r200DRIPriv->textureOffset;
+   r200Screen->texSize[RADEON_CARD_HEAP] = r200DRIPriv->textureSize;
+   r200Screen->logTexGranularity[RADEON_CARD_HEAP] =
+      r200DRIPriv->log2TexGran;
+
+   if ( r200Screen->IsPCI ) {
+      r200Screen->numTexHeaps = RADEON_NR_TEX_HEAPS - 1;
+      r200Screen->texOffset[RADEON_AGP_HEAP] = 0;
+      r200Screen->texSize[RADEON_AGP_HEAP] = 0;
+      r200Screen->logTexGranularity[RADEON_AGP_HEAP] = 0;
+   } else {
+      r200Screen->numTexHeaps = RADEON_NR_TEX_HEAPS;
+      r200Screen->texOffset[RADEON_AGP_HEAP] =
+	 r200DRIPriv->agpTexOffset + R200_AGP_TEX_OFFSET;
+      r200Screen->texSize[RADEON_AGP_HEAP] = r200DRIPriv->agpTexMapSize;
+      r200Screen->logTexGranularity[RADEON_AGP_HEAP] =
+	 r200DRIPriv->log2AGPTexGran;
+   }
+
+
+   r200Screen->driScreen = sPriv;
+   r200Screen->sarea_priv_offset = r200DRIPriv->sarea_priv_offset;
+   return r200Screen;
+}
+
+/* Destroy the device specific screen private data struct.
+ */
+static void 
+r200DestroyScreen( __DRIscreenPrivate *sPriv )
+{
+   r200ScreenPtr r200Screen = (r200ScreenPtr)sPriv->private;
+
+   if (!r200Screen)
+      return;
+
+   if ( !r200Screen->IsPCI ) {
+      drmUnmap( r200Screen->agpTextures.map,
+		r200Screen->agpTextures.size );
+   }
+   drmUnmapBufs( r200Screen->buffers );
+   drmUnmap( r200Screen->status.map, r200Screen->status.size );
+   drmUnmap( r200Screen->mmio.map, r200Screen->mmio.size );
+
+   FREE( r200Screen );
+   sPriv->private = NULL;
+}
+
+
+/* Initialize the driver specific screen private data.
+ */
+static GLboolean
+r200InitDriver( __DRIscreenPrivate *sPriv )
+{
+   __r200Screen = r200CreateScreen( sPriv );
+
+   sPriv->private = (void *) __r200Screen;
+
+   return sPriv->private ? GL_TRUE : GL_FALSE;
+}
+
+
+
+/* Create and initialize the Mesa and driver specific pixmap buffer
+ * data.
+ */
+static GLboolean
+r200CreateBuffer( __DRIscreenPrivate *driScrnPriv,
+                  __DRIdrawablePrivate *driDrawPriv,
+                  const __GLcontextModes *mesaVis,
+                  GLboolean isPixmap )
+{
+   if (isPixmap) {
+      return GL_FALSE; /* not implemented */
+   }
+   else {
+      const GLboolean swDepth = GL_FALSE;
+      const GLboolean swAlpha = GL_FALSE;
+      const GLboolean swAccum = mesaVis->accumRedBits > 0;
+      const GLboolean swStencil = mesaVis->stencilBits > 0 &&
+         mesaVis->depthBits != 24;
+      driDrawPriv->driverPrivate = (void *)
+         _mesa_create_framebuffer( mesaVis,
+                                   swDepth,
+                                   swStencil,
+                                   swAccum,
+                                   swAlpha );
+      return (driDrawPriv->driverPrivate != NULL);
+   }
+}
+
+
+static void
+r200DestroyBuffer(__DRIdrawablePrivate *driDrawPriv)
+{
+   _mesa_destroy_framebuffer((GLframebuffer *) (driDrawPriv->driverPrivate));
+}
+
+
+
+
+/* Fullscreen mode isn't used for much -- could be a way to shrink
+ * front/back buffers & get more texture memory if the client has
+ * changed the video resolution.
+ * 
+ * Pageflipping is now done automatically whenever there is a single
+ * 3d client.
+ */
+static GLboolean
+r200OpenCloseFullScreen( __DRIcontextPrivate *driContextPriv )
+{
+   return GL_TRUE;
+}
+
+static struct __DriverAPIRec r200API = {
+   r200InitDriver,
+   r200DestroyScreen,
+   r200CreateContext,
+   r200DestroyContext,
+   r200CreateBuffer,
+   r200DestroyBuffer,
+   r200SwapBuffers,
+   r200MakeCurrent,
+   r200UnbindContext,
+   r200OpenCloseFullScreen,
+   r200OpenCloseFullScreen
+};
+
+
+
+/*
+ * This is the bootstrap function for the driver.
+ * The __driCreateScreen name is the symbol that libGL.so fetches.
+ * Return:  pointer to a __DRIscreenPrivate.
+ *
+ */
+void *__driCreateScreen(Display *dpy, int scrn, __DRIscreen *psc,
+                        int numConfigs, __GLXvisualConfig *config)
+{
+   __DRIscreenPrivate *psp;
+   psp = __driUtilCreateScreen(dpy, scrn, psc, numConfigs, config, &r200API);
+   return (void *) psp;
+}
+
+
+/* This function is called by libGL.so to allow the driver to dynamically
+ * extend libGL.  We can add new GLX functions and/or new GL functions.
+ * Note that _mesa_create_context() will probably add most of the newer
+ * OpenGL extension functions into the dispatcher.
+ */
+void
+__driRegisterExtensions( void )
+{
+   /* dlopen ourself */
+   void *dll = dlopen(NULL, RTLD_GLOBAL);
+   if (dll) {
+      typedef void *(*registerFunc)(const char *funcName, void *funcAddr);
+      typedef void (*registerString)(const char *extName);
+
+      /* Get pointers to libGL's __glXRegisterGLXFunction
+       * and __glXRegisterGLXExtensionString, if they exist.
+       */
+      registerFunc regFunc = (registerFunc) dlsym(dll, "__glXRegisterGLXFunction");
+      registerString regString = (registerString) dlsym(dll, "__glXRegisterGLXExtensionString");
+
+      if (regFunc) {
+         /* register our GLX extensions with libGL */
+         void *p;
+         p = regFunc("glXAllocateMemoryNV", (void *) r200AllocateMemoryNV);
+         if (p)
+            ;  /* XXX already registered - what to do, wrap? */
+
+         p = regFunc("glXFreeMemoryNV", (void *) r200FreeMemoryNV);
+         if (p)
+            ;  /* XXX already registered - what to do, wrap? */
+
+         p = regFunc("glXGetAGPOffsetMESA", (void *) r200GetAGPOffset);
+         if (p)
+            ;  /* XXX already registered - what to do, wrap? */
+      }
+
+      if (regString) {
+         regString("GLX_NV_vertex_array_range");
+         regString("GLX_MESA_agp_offset");
+      }
+
+      dlclose(dll);
+   }
+}

src/mesa/drivers/dri/r200/r200_screen.h

@@ -0,0 +1,96 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#ifndef __R200_SCREEN_H__
+#define __R200_SCREEN_H__
+
+#ifdef GLX_DIRECT_RENDERING
+
+#include "dri_util.h"
+#include "radeon_common.h"
+#include "radeon_sarea.h"
+
+typedef struct {
+   drmHandle handle;			/* Handle to the DRM region */
+   drmSize size;			/* Size of the DRM region */
+   drmAddress map;			/* Mapping of the DRM region */
+} r200RegionRec, *r200RegionPtr;
+
+#define R200_CHIPSET_R200   1
+#define R200_CHIPSET_MOBILITY 2
+
+
+#define R200_NR_TEX_HEAPS 2
+
+typedef struct {
+
+   int chipset;
+   int cpp;
+   int IsPCI;				/* Current card is a PCI card */
+   int AGPMode;
+   unsigned int irq;			/* IRQ number (0 means none) */
+
+   unsigned int frontOffset;
+   unsigned int frontPitch;
+   unsigned int backOffset;
+   unsigned int backPitch;
+
+   unsigned int depthOffset;
+   unsigned int depthPitch;
+
+    /* Shared texture data */
+   int numTexHeaps;
+   int texOffset[R200_NR_TEX_HEAPS];
+   int texSize[R200_NR_TEX_HEAPS];
+   int logTexGranularity[R200_NR_TEX_HEAPS];
+
+   r200RegionRec mmio;
+   r200RegionRec status;
+   r200RegionRec agpTextures;
+
+   drmBufMapPtr buffers;
+
+   __volatile__ GLuint *scratch;
+
+   __DRIscreenPrivate *driScreen;
+   unsigned int sarea_priv_offset;
+   unsigned int agp_buffer_offset;	/* offset in card memory space */
+   unsigned int agp_texture_offset;	/* offset in card memory space */
+   unsigned int agp_base;
+
+   GLboolean drmSupportsCubeMaps;       /* need radeon kernel module >=1.7 */
+} r200ScreenRec, *r200ScreenPtr;
+
+#endif
+#endif /* __R200_SCREEN_H__ */

src/mesa/drivers/dri/r200/r200_span.c

@@ -0,0 +1,430 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#include "glheader.h"
+#include "imports.h"
+#include "swrast/swrast.h"
+
+#include "r200_context.h"
+#include "r200_ioctl.h"
+#include "r200_state.h"
+#include "r200_span.h"
+#include "r200_tex.h"
+
+#define DBG 0
+
+#define LOCAL_VARS							\
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);			\
+   r200ScreenPtr r200Screen = rmesa->r200Screen;			\
+   __DRIscreenPrivate *sPriv = rmesa->dri.screen;			\
+   __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;			\
+   GLuint pitch = r200Screen->frontPitch * r200Screen->cpp;		\
+   GLuint height = dPriv->h;						\
+   char *buf = (char *)(sPriv->pFB +					\
+			rmesa->state.color.drawOffset +			\
+			(dPriv->x * r200Screen->cpp) +		\
+			(dPriv->y * pitch));				\
+   char *read_buf = (char *)(sPriv->pFB +				\
+			     rmesa->state.pixel.readOffset +		\
+			     (dPriv->x * r200Screen->cpp) +		\
+			     (dPriv->y * pitch));			\
+   GLuint p;								\
+   (void) read_buf; (void) buf; (void) p
+
+#define LOCAL_DEPTH_VARS						\
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);			\
+   r200ScreenPtr r200Screen = rmesa->r200Screen;			\
+   __DRIscreenPrivate *sPriv = rmesa->dri.screen;			\
+   __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;			\
+   GLuint height = dPriv->h;						\
+   GLuint xo = dPriv->x;						\
+   GLuint yo = dPriv->y;						\
+   char *buf = (char *)(sPriv->pFB + r200Screen->depthOffset);	\
+   (void) buf
+
+#define LOCAL_STENCIL_VARS	LOCAL_DEPTH_VARS
+
+
+#define CLIPPIXEL( _x, _y )						\
+   ((_x >= minx) && (_x < maxx) && (_y >= miny) && (_y < maxy))
+
+
+#define CLIPSPAN( _x, _y, _n, _x1, _n1, _i )				\
+   if ( _y < miny || _y >= maxy ) {					\
+      _n1 = 0, _x1 = x;							\
+   } else {								\
+      _n1 = _n;								\
+      _x1 = _x;								\
+      if ( _x1 < minx ) _i += (minx-_x1), n1 -= (minx-_x1), _x1 = minx; \
+      if ( _x1 + _n1 >= maxx ) n1 -= (_x1 + n1 - maxx);		        \
+   }
+
+#define Y_FLIP( _y )		(height - _y - 1)
+
+
+#define HW_LOCK() 
+
+#define HW_CLIPLOOP()							\
+   do {									\
+      __DRIdrawablePrivate *dPriv = rmesa->dri.drawable;		\
+      int _nc = dPriv->numClipRects;					\
+									\
+      while ( _nc-- ) {							\
+	 int minx = dPriv->pClipRects[_nc].x1 - dPriv->x;		\
+	 int miny = dPriv->pClipRects[_nc].y1 - dPriv->y;		\
+	 int maxx = dPriv->pClipRects[_nc].x2 - dPriv->x;		\
+	 int maxy = dPriv->pClipRects[_nc].y2 - dPriv->y;
+
+#define HW_ENDCLIPLOOP()						\
+      }									\
+   } while (0)
+
+#define HW_UNLOCK()							
+
+
+
+/* ================================================================
+ * Color buffer
+ */
+
+/* 16 bit, RGB565 color spanline and pixel functions
+ */
+#define INIT_MONO_PIXEL(p, color) \
+  p = PACK_COLOR_565( color[0], color[1], color[2] )
+
+#define WRITE_RGBA( _x, _y, r, g, b, a )				\
+   *(GLushort *)(buf + _x*2 + _y*pitch) = ((((int)r & 0xf8) << 8) |	\
+					   (((int)g & 0xfc) << 3) |	\
+					   (((int)b & 0xf8) >> 3))
+
+#define WRITE_PIXEL( _x, _y, p )					\
+   *(GLushort *)(buf + _x*2 + _y*pitch) = p
+
+#define READ_RGBA( rgba, _x, _y )					\
+   do {									\
+      GLushort p = *(GLushort *)(read_buf + _x*2 + _y*pitch);		\
+      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)
+
+#define TAG(x) r200##x##_RGB565
+#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 )			\
+do {								\
+   *(GLuint *)(buf + _x*4 + _y*pitch) = ((b <<  0) |		\
+					 (g <<  8) |		\
+					 (r << 16) |		\
+					 (a << 24) );		\
+} while (0)
+
+#define WRITE_PIXEL( _x, _y, p ) 			\
+do {							\
+   *(GLuint *)(buf + _x*4 + _y*pitch) = p;		\
+} while (0)
+
+#define READ_RGBA( rgba, _x, _y )				\
+do {								\
+   volatile GLuint *ptr = (volatile GLuint *)(read_buf + _x*4 + _y*pitch); \
+   GLuint p = *ptr;					\
+   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) r200##x##_ARGB8888
+#include "spantmp.h"
+
+
+
+/* ================================================================
+ * Depth buffer
+ */
+
+/* The R200 has depth tiling on all the time, so we have to convert
+ * the x,y coordinates into the memory bus address (mba) in the same
+ * manner as the engine.  In each case, the linear block address (ba)
+ * is calculated, and then wired with x and y to produce the final
+ * memory address.
+ */
+
+#define BIT(x,b) ((x & (1<<b))>>b)
+static GLuint r200_mba_z32( r200ContextPtr rmesa,
+				       GLint x, GLint y )
+{
+   GLuint pitch = rmesa->r200Screen->frontPitch;
+   GLuint b = ((y & 0x3FF) >> 4) * ((pitch & 0xFFF) >> 5) + ((x & 0x3FF) >> 5);
+   GLuint a = 
+      (BIT(x,0) << 2) |
+      (BIT(y,0) << 3) |
+      (BIT(x,1) << 4) |
+      (BIT(y,1) << 5) |
+      (BIT(x,3) << 6) |
+      (BIT(x,4) << 7) |
+      (BIT(x,2) << 8) |
+      (BIT(y,2) << 9) |
+      (BIT(y,3) << 10) |
+      (((pitch & 0x20) ? (b & 0x01) : ((b & 0x01) ^ (BIT(y,4)))) << 11) |
+      ((b >> 1) << 12);
+   return a;
+}
+
+static GLuint r200_mba_z16( r200ContextPtr rmesa, GLint x, GLint y )
+{
+   GLuint pitch = rmesa->r200Screen->frontPitch;
+   GLuint b = ((y & 0x3FF) >> 4) * ((pitch & 0xFFF) >> 6) + ((x & 0x3FF) >> 6);
+   GLuint a = 
+      (BIT(x,0) << 1) |
+      (BIT(y,0) << 2) |
+      (BIT(x,1) << 3) |
+      (BIT(y,1) << 4) |
+      (BIT(x,2) << 5) |
+      (BIT(x,4) << 6) |
+      (BIT(x,5) << 7) |
+      (BIT(x,3) << 8) |
+      (BIT(y,2) << 9) |
+      (BIT(y,3) << 10) |
+      (((pitch & 0x40) ? (b & 0x01) : ((b & 0x01) ^ (BIT(y,4)))) << 11) |
+      ((b >> 1) << 12);
+   return a;
+}
+
+
+/* 16-bit depth buffer functions
+ */
+#define WRITE_DEPTH( _x, _y, d )					\
+   *(GLushort *)(buf + r200_mba_z16( rmesa, _x + xo, _y + yo )) = d;
+
+#define READ_DEPTH( d, _x, _y )						\
+   d = *(GLushort *)(buf + r200_mba_z16( rmesa, _x + xo, _y + yo ));
+
+#define TAG(x) r200##x##_16
+#include "depthtmp.h"
+
+/* 24 bit depth, 8 bit stencil depthbuffer functions
+ */
+#define WRITE_DEPTH( _x, _y, d )					\
+do {									\
+   GLuint offset = r200_mba_z32( rmesa, _x + xo, _y + yo );		\
+   GLuint tmp = *(GLuint *)(buf + offset);				\
+   tmp &= 0xff000000;							\
+   tmp |= ((d) & 0x00ffffff);						\
+   *(GLuint *)(buf + offset) = tmp;					\
+} while (0)
+
+#define READ_DEPTH( d, _x, _y )						\
+   d = *(GLuint *)(buf + r200_mba_z32( rmesa, _x + xo,		\
+					 _y + yo )) & 0x00ffffff;
+
+#define TAG(x) r200##x##_24_8
+#include "depthtmp.h"
+
+
+/* ================================================================
+ * Stencil buffer
+ */
+
+/* 24 bit depth, 8 bit stencil depthbuffer functions
+ */
+#define WRITE_STENCIL( _x, _y, d )					\
+do {									\
+   GLuint offset = r200_mba_z32( rmesa, _x + xo, _y + yo );		\
+   GLuint tmp = *(GLuint *)(buf + offset);				\
+   tmp &= 0x00ffffff;							\
+   tmp |= (((d) & 0xff) << 24);						\
+   *(GLuint *)(buf + offset) = tmp;					\
+} while (0)
+
+#define READ_STENCIL( d, _x, _y )					\
+do {									\
+   GLuint offset = r200_mba_z32( rmesa, _x + xo, _y + yo );		\
+   GLuint tmp = *(GLuint *)(buf + offset);				\
+   tmp &= 0xff000000;							\
+   d = tmp >> 24;							\
+} while (0)
+
+#define TAG(x) r200##x##_24_8
+#include "stenciltmp.h"
+
+
+/*
+ * This function is called to specify which buffer to read and write
+ * for software rasterization (swrast) fallbacks.  This doesn't necessarily
+ * correspond to glDrawBuffer() or glReadBuffer() calls.
+ */
+static void r200SetBuffer( GLcontext *ctx,
+                           GLframebuffer *colorBuffer,
+                           GLuint bufferBit )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+
+   switch ( bufferBit ) {
+   case FRONT_LEFT_BIT:
+      if ( rmesa->doPageFlip && rmesa->sarea->pfCurrentPage == 1 ) {
+        rmesa->state.pixel.readOffset = rmesa->r200Screen->backOffset;
+        rmesa->state.pixel.readPitch  = rmesa->r200Screen->backPitch;
+        rmesa->state.color.drawOffset = rmesa->r200Screen->backOffset;
+        rmesa->state.color.drawPitch  = rmesa->r200Screen->backPitch;
+      } else {
+      	rmesa->state.pixel.readOffset = rmesa->r200Screen->frontOffset;
+      	rmesa->state.pixel.readPitch  = rmesa->r200Screen->frontPitch;
+      	rmesa->state.color.drawOffset = rmesa->r200Screen->frontOffset;
+      	rmesa->state.color.drawPitch  = rmesa->r200Screen->frontPitch;
+      }
+      break;
+   case BACK_LEFT_BIT:
+      if ( rmesa->doPageFlip && rmesa->sarea->pfCurrentPage == 1 ) {
+      	rmesa->state.pixel.readOffset = rmesa->r200Screen->frontOffset;
+      	rmesa->state.pixel.readPitch  = rmesa->r200Screen->frontPitch;
+      	rmesa->state.color.drawOffset = rmesa->r200Screen->frontOffset;
+      	rmesa->state.color.drawPitch  = rmesa->r200Screen->frontPitch;
+      } else {
+        rmesa->state.pixel.readOffset = rmesa->r200Screen->backOffset;
+        rmesa->state.pixel.readPitch  = rmesa->r200Screen->backPitch;
+        rmesa->state.color.drawOffset = rmesa->r200Screen->backOffset;
+        rmesa->state.color.drawPitch  = rmesa->r200Screen->backPitch;
+      }
+      break;
+   default:
+      _mesa_problem(ctx, "Bad bufferBit in r200SetBuffer()");
+      break;
+   }
+}
+
+/* Move locking out to get reasonable span performance (10x better
+ * than doing this in HW_LOCK above).  WaitForIdle() is the main
+ * culprit.
+ */
+
+static void r200SpanRenderStart( GLcontext *ctx )
+{
+   r200ContextPtr rmesa = R200_CONTEXT( ctx );
+
+   R200_FIREVERTICES( rmesa );
+   LOCK_HARDWARE( rmesa );
+   r200WaitForIdleLocked( rmesa );
+
+   /* Read & rewrite the first pixel in the frame buffer.  This should
+    * be a noop, right?  In fact without this conform fails as reading
+    * from the framebuffer sometimes produces old results -- the
+    * on-card read cache gets mixed up and doesn't notice that the
+    * framebuffer has been updated.
+    *
+    * In the worst case this is buggy too as p might get the wrong
+    * value first time, so really need a hidden pixel somewhere for this.
+    */
+   {
+      int p;
+      volatile int *read_buf = (volatile int *)(rmesa->dri.screen->pFB + 
+						rmesa->state.pixel.readOffset);
+      p = *read_buf;
+      *read_buf = p;
+   }
+}
+
+static void r200SpanRenderFinish( GLcontext *ctx )
+{
+   r200ContextPtr rmesa = R200_CONTEXT( ctx );
+   _swrast_flush( ctx );
+   UNLOCK_HARDWARE( rmesa );
+}
+
+void r200InitSpanFuncs( GLcontext *ctx )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   struct swrast_device_driver *swdd = _swrast_GetDeviceDriverReference(ctx);
+
+   swdd->SetBuffer = r200SetBuffer;
+
+   switch ( rmesa->r200Screen->cpp ) {
+   case 2:
+      swdd->WriteRGBASpan	= r200WriteRGBASpan_RGB565;
+      swdd->WriteRGBSpan	= r200WriteRGBSpan_RGB565;
+      swdd->WriteMonoRGBASpan	= r200WriteMonoRGBASpan_RGB565;
+      swdd->WriteRGBAPixels	= r200WriteRGBAPixels_RGB565;
+      swdd->WriteMonoRGBAPixels	= r200WriteMonoRGBAPixels_RGB565;
+      swdd->ReadRGBASpan	= r200ReadRGBASpan_RGB565;
+      swdd->ReadRGBAPixels      = r200ReadRGBAPixels_RGB565;
+      break;
+
+   case 4:
+      swdd->WriteRGBASpan	= r200WriteRGBASpan_ARGB8888;
+      swdd->WriteRGBSpan	= r200WriteRGBSpan_ARGB8888;
+      swdd->WriteMonoRGBASpan   = r200WriteMonoRGBASpan_ARGB8888;
+      swdd->WriteRGBAPixels     = r200WriteRGBAPixels_ARGB8888;
+      swdd->WriteMonoRGBAPixels = r200WriteMonoRGBAPixels_ARGB8888;
+      swdd->ReadRGBASpan	= r200ReadRGBASpan_ARGB8888;
+      swdd->ReadRGBAPixels      = r200ReadRGBAPixels_ARGB8888;
+      break;
+
+   default:
+      break;
+   }
+
+   switch ( rmesa->glCtx->Visual.depthBits ) {
+   case 16:
+      swdd->ReadDepthSpan	= r200ReadDepthSpan_16;
+      swdd->WriteDepthSpan	= r200WriteDepthSpan_16;
+      swdd->ReadDepthPixels	= r200ReadDepthPixels_16;
+      swdd->WriteDepthPixels	= r200WriteDepthPixels_16;
+      break;
+
+   case 24:
+      swdd->ReadDepthSpan	= r200ReadDepthSpan_24_8;
+      swdd->WriteDepthSpan	= r200WriteDepthSpan_24_8;
+      swdd->ReadDepthPixels	= r200ReadDepthPixels_24_8;
+      swdd->WriteDepthPixels	= r200WriteDepthPixels_24_8;
+
+      swdd->ReadStencilSpan	= r200ReadStencilSpan_24_8;
+      swdd->WriteStencilSpan	= r200WriteStencilSpan_24_8;
+      swdd->ReadStencilPixels	= r200ReadStencilPixels_24_8;
+      swdd->WriteStencilPixels	= r200WriteStencilPixels_24_8;
+      break;
+
+   default:
+      break;
+   }
+
+   swdd->SpanRenderStart          = r200SpanRenderStart;
+   swdd->SpanRenderFinish         = r200SpanRenderFinish; 
+}

src/mesa/drivers/dri/r200/r200_span.h

@@ -0,0 +1,43 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#ifndef __R200_SPAN_H__
+#define __R200_SPAN_H__
+
+#ifdef GLX_DIRECT_RENDERING
+
+extern void r200InitSpanFuncs( GLcontext *ctx );
+
+#endif
+#endif

src/mesa/drivers/dri/r200/r200_state.h

@@ -0,0 +1,68 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#ifndef __R200_STATE_H__
+#define __R200_STATE_H__
+
+#ifdef GLX_DIRECT_RENDERING
+
+#include "r200_context.h"
+
+extern void r200InitState( r200ContextPtr rmesa );
+extern void r200InitStateFuncs( GLcontext *ctx );
+
+extern void r200UpdateMaterial( GLcontext *ctx );
+
+extern void r200SetCliprects( r200ContextPtr rmesa, GLenum mode );
+extern void r200RecalcScissorRects( r200ContextPtr rmesa );
+extern void r200UpdateViewportOffset( GLcontext *ctx );
+extern void r200UpdateWindow( GLcontext *ctx );
+
+extern void r200ValidateState( GLcontext *ctx );
+
+extern void r200PrintDirty( r200ContextPtr rmesa,
+			      const char *msg );
+
+
+extern void r200Fallback( GLcontext *ctx, GLuint bit, GLboolean mode );
+#define FALLBACK( rmesa, bit, mode ) do {				\
+   if ( 0 ) fprintf( stderr, "FALLBACK in %s: #%d=%d\n",		\
+		     __FUNCTION__, bit, mode );				\
+   r200Fallback( rmesa->glCtx, bit, mode );				\
+} while (0)
+
+extern void r200LightingSpaceChange( GLcontext *ctx );
+
+#endif
+#endif

src/mesa/drivers/dri/r200/r200_state_init.c

@@ -0,0 +1,692 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#include "glheader.h"
+#include "imports.h"
+#include "mmath.h"
+#include "enums.h"
+#include "colormac.h"
+#include "api_arrayelt.h"
+
+#include "swrast/swrast.h"
+#include "array_cache/acache.h"
+#include "tnl/tnl.h"
+#include "tnl/t_pipeline.h"
+#include "swrast_setup/swrast_setup.h"
+
+#include "r200_context.h"
+#include "r200_ioctl.h"
+#include "r200_state.h"
+#include "r200_tcl.h"
+#include "r200_tex.h"
+#include "r200_swtcl.h"
+#include "r200_vtxfmt.h"
+
+/* =============================================================
+ * State initialization
+ */
+
+void r200PrintDirty( r200ContextPtr rmesa, const char *msg )
+{
+   struct r200_state_atom *l;
+
+   fprintf(stderr, msg);
+   fprintf(stderr, ": ");
+
+   foreach(l, &(rmesa->hw.dirty)) {
+      fprintf(stderr, "%s, ", l->name);
+   }
+
+   fprintf(stderr, "\n");
+}
+
+static int cmdpkt( int id ) 
+{
+   drmRadeonCmdHeader h;
+   h.i = 0;
+   h.packet.cmd_type = RADEON_CMD_PACKET;
+   h.packet.packet_id = id;
+   return h.i;
+}
+
+static int cmdvec( int offset, int stride, int count ) 
+{
+   drmRadeonCmdHeader h;
+   h.i = 0;
+   h.vectors.cmd_type = RADEON_CMD_VECTORS;
+   h.vectors.offset = offset;
+   h.vectors.stride = stride;
+   h.vectors.count = count;
+   return h.i;
+}
+
+static int cmdscl( int offset, int stride, int count ) 
+{
+   drmRadeonCmdHeader h;
+   h.i = 0;
+   h.scalars.cmd_type = RADEON_CMD_SCALARS;
+   h.scalars.offset = offset;
+   h.scalars.stride = stride;
+   h.scalars.count = count;
+   return h.i;
+}
+
+static int cmdscl2( int offset, int stride, int count ) 
+{
+   drmRadeonCmdHeader h;
+   h.i = 0;
+   h.scalars.cmd_type = RADEON_CMD_SCALARS2;
+   h.scalars.offset = offset - 0x100;
+   h.scalars.stride = stride;
+   h.scalars.count = count;
+   return h.i;
+}
+
+#define CHECK( NM, FLAG )				\
+static GLboolean check_##NM( GLcontext *ctx, int idx )	\
+{							\
+   (void) idx;						\
+   return FLAG;						\
+}
+
+#define TCL_CHECK( NM, FLAG )				\
+static GLboolean check_##NM( GLcontext *ctx, int idx )	\
+{							\
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);		\
+   (void) idx;						\
+   return !rmesa->TclFallback && (FLAG);		\
+}
+
+
+
+CHECK( always, GL_TRUE )
+CHECK( tex_any, ctx->Texture._EnabledUnits )
+CHECK( tex, ctx->Texture.Unit[idx]._ReallyEnabled )
+CHECK( fog, ctx->Fog.Enabled )
+TCL_CHECK( tcl, GL_TRUE )
+TCL_CHECK( tcl_tex_any, ctx->Texture._EnabledUnits )
+TCL_CHECK( tcl_tex, ctx->Texture.Unit[idx]._ReallyEnabled )
+TCL_CHECK( tcl_lighting, ctx->Light.Enabled )
+TCL_CHECK( tcl_eyespace_or_lighting, ctx->_NeedEyeCoords || ctx->Light.Enabled )
+TCL_CHECK( tcl_light, ctx->Light.Enabled && ctx->Light.Light[idx].Enabled )
+TCL_CHECK( tcl_ucp, (ctx->Transform.ClipPlanesEnabled & (1 << idx)) )
+/* TCL_CHECK( tcl_eyespace_or_fog, ctx->_NeedEyeCoords || ctx->Fog.Enabled )  */
+
+
+static GLboolean check_tcl_eyespace_or_fog( GLcontext *ctx, int idx )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   int res;
+   (void) idx;
+   res = !rmesa->TclFallback && (ctx->_NeedEyeCoords || ctx->Fog.Enabled);
+   fprintf(stderr, "%s: %d\n", __FUNCTION__, res);
+   return res;
+}
+
+
+/* Initialize the context's hardware state.
+ */
+void r200InitState( r200ContextPtr rmesa )
+{
+   GLcontext *ctx = rmesa->glCtx;
+   GLuint color_fmt, depth_fmt, i;
+
+   switch ( rmesa->r200Screen->cpp ) {
+   case 2:
+      color_fmt = R200_COLOR_FORMAT_RGB565;
+      break;
+   case 4:
+      color_fmt = R200_COLOR_FORMAT_ARGB8888;
+      break;
+   default:
+      fprintf( stderr, "Error: Unsupported pixel depth... exiting\n" );
+      exit( -1 );
+   }
+
+   rmesa->state.color.clear = 0x00000000;
+
+   switch ( ctx->Visual.depthBits ) {
+   case 16:
+      rmesa->state.depth.scale = 1.0 / (GLfloat)0xffff;
+      depth_fmt = R200_DEPTH_FORMAT_16BIT_INT_Z;
+      rmesa->state.stencil.clear = 0x00000000;
+      break;
+   case 24:
+      rmesa->state.depth.scale = 1.0 / (GLfloat)0xffffff;
+      depth_fmt = R200_DEPTH_FORMAT_24BIT_INT_Z;
+      rmesa->state.stencil.clear = 0xff000000;
+      break;
+   default:
+      fprintf( stderr, "Error: Unsupported depth %d... exiting\n",
+	       ctx->Visual.depthBits );
+      exit( -1 );
+   }
+
+   /* Only have hw stencil when depth buffer is 24 bits deep */
+   rmesa->state.stencil.hwBuffer = ( ctx->Visual.stencilBits > 0 &&
+				     ctx->Visual.depthBits == 24 );
+
+   rmesa->Fallback = 0;
+
+   if ( ctx->Visual.doubleBufferMode && rmesa->sarea->pfCurrentPage == 0 ) {
+      rmesa->state.color.drawOffset = rmesa->r200Screen->backOffset;
+      rmesa->state.color.drawPitch  = rmesa->r200Screen->backPitch;
+   } else {
+      rmesa->state.color.drawOffset = rmesa->r200Screen->frontOffset;
+      rmesa->state.color.drawPitch  = rmesa->r200Screen->frontPitch;
+   }
+
+   rmesa->state.pixel.readOffset = rmesa->state.color.drawOffset;
+   rmesa->state.pixel.readPitch  = rmesa->state.color.drawPitch;
+
+   /* Initialize lists:
+    */
+   make_empty_list(&(rmesa->hw.dirty)); rmesa->hw.dirty.name = "DIRTY";
+   make_empty_list(&(rmesa->hw.clean)); rmesa->hw.clean.name = "CLEAN";
+
+
+#define ALLOC_STATE( ATOM, CHK, SZ, NM, IDX )				\
+   do {								\
+      rmesa->hw.ATOM.cmd_size = SZ;				\
+      rmesa->hw.ATOM.cmd = (int *)CALLOC(SZ * sizeof(int));	\
+      rmesa->hw.ATOM.lastcmd = (int *)CALLOC(SZ * sizeof(int));	\
+      rmesa->hw.ATOM.name = NM;					\
+      rmesa->hw.ATOM.idx = IDX;					\
+      rmesa->hw.ATOM.check = check_##CHK;				\
+      insert_at_head(&(rmesa->hw.dirty), &(rmesa->hw.ATOM));	\
+   } while (0)
+      
+      
+   /* Allocate state buffers:
+    */
+   ALLOC_STATE( ctx, always, CTX_STATE_SIZE, "CTX/context", 0 );
+   ALLOC_STATE( set, always, SET_STATE_SIZE, "SET/setup", 0 );
+   ALLOC_STATE( lin, always, LIN_STATE_SIZE, "LIN/line", 0 );
+   ALLOC_STATE( msk, always, MSK_STATE_SIZE, "MSK/mask", 0 );
+   ALLOC_STATE( vpt, always, VPT_STATE_SIZE, "VPT/viewport", 0 );
+   ALLOC_STATE( vtx, always, VTX_STATE_SIZE, "VTX/vertex", 0 );
+   ALLOC_STATE( vap, always, VAP_STATE_SIZE, "VAP/vap", 0 );
+   ALLOC_STATE( vte, always, VTE_STATE_SIZE, "VTE/vte", 0 );
+   ALLOC_STATE( msc, always, MSC_STATE_SIZE, "MSC/misc", 0 );
+   ALLOC_STATE( cst, always, CST_STATE_SIZE, "CST/constant", 0 );
+   ALLOC_STATE( zbs, always, ZBS_STATE_SIZE, "ZBS/zbias", 0 );
+   ALLOC_STATE( tam, tex_any, TAM_STATE_SIZE, "TAM/tam", 0 );
+   ALLOC_STATE( tf, tex_any, TF_STATE_SIZE, "TF/tfactor", 0 );
+   ALLOC_STATE( tex[0], tex_any, TEX_STATE_SIZE, "TEX/tex-0", 0 );
+   ALLOC_STATE( tex[1], tex_any, TEX_STATE_SIZE, "TEX/tex-1", 1 );
+   ALLOC_STATE( cube[0], tex_any, CUBE_STATE_SIZE, "CUBE/tex-0", 0 );
+   ALLOC_STATE( cube[1], tex_any, CUBE_STATE_SIZE, "CUBE/tex-1", 1 );
+
+   ALLOC_STATE( tcl, tcl, TCL_STATE_SIZE, "TCL/tcl", 0 );
+   ALLOC_STATE( msl, tcl, MSL_STATE_SIZE, "MSL/matrix-select", 0 );
+   ALLOC_STATE( tcg, tcl, TCG_STATE_SIZE, "TCG/texcoordgen", 0 );
+   ALLOC_STATE( mtl[0], tcl_lighting, MTL_STATE_SIZE, "MTL0/material0", 0 );
+   ALLOC_STATE( grd, tcl, GRD_STATE_SIZE, "GRD/guard-band", 0 );
+   ALLOC_STATE( fog, fog, FOG_STATE_SIZE, "FOG/fog", 0 );
+   ALLOC_STATE( glt, tcl_lighting, GLT_STATE_SIZE, "GLT/light-global", 0 );
+   ALLOC_STATE( eye, tcl_lighting, EYE_STATE_SIZE, "EYE/eye-vector", 0 );
+   ALLOC_STATE( mat[R200_MTX_MV], tcl, MAT_STATE_SIZE, "MAT/modelview", 0 );
+   ALLOC_STATE( mat[R200_MTX_IMV], tcl, MAT_STATE_SIZE, "MAT/it-modelview", 0 );
+   ALLOC_STATE( mat[R200_MTX_MVP], tcl, MAT_STATE_SIZE, "MAT/modelproject", 0 );
+   ALLOC_STATE( mat[R200_MTX_TEX0], tcl_tex, MAT_STATE_SIZE, "MAT/texmat0", 0 );
+   ALLOC_STATE( mat[R200_MTX_TEX1], tcl_tex, MAT_STATE_SIZE, "MAT/texmat1", 1 );
+   ALLOC_STATE( ucp[0], tcl_ucp, UCP_STATE_SIZE, "UCP/userclip-0", 0 );
+   ALLOC_STATE( ucp[1], tcl_ucp, UCP_STATE_SIZE, "UCP/userclip-1", 1 );
+   ALLOC_STATE( ucp[2], tcl_ucp, UCP_STATE_SIZE, "UCP/userclip-2", 2 );
+   ALLOC_STATE( ucp[3], tcl_ucp, UCP_STATE_SIZE, "UCP/userclip-3", 3 );
+   ALLOC_STATE( ucp[4], tcl_ucp, UCP_STATE_SIZE, "UCP/userclip-4", 4 );
+   ALLOC_STATE( ucp[5], tcl_ucp, UCP_STATE_SIZE, "UCP/userclip-5", 5 );
+   ALLOC_STATE( lit[0], tcl_light, LIT_STATE_SIZE, "LIT/light-0", 0 );
+   ALLOC_STATE( lit[1], tcl_light, LIT_STATE_SIZE, "LIT/light-1", 1 );
+   ALLOC_STATE( lit[2], tcl_light, LIT_STATE_SIZE, "LIT/light-2", 2 );
+   ALLOC_STATE( lit[3], tcl_light, LIT_STATE_SIZE, "LIT/light-3", 3 );
+   ALLOC_STATE( lit[4], tcl_light, LIT_STATE_SIZE, "LIT/light-4", 4 );
+   ALLOC_STATE( lit[5], tcl_light, LIT_STATE_SIZE, "LIT/light-5", 5 );
+   ALLOC_STATE( lit[6], tcl_light, LIT_STATE_SIZE, "LIT/light-6", 6 );
+   ALLOC_STATE( lit[7], tcl_light, LIT_STATE_SIZE, "LIT/light-7", 7 );
+   ALLOC_STATE( pix[0], always, PIX_STATE_SIZE, "PIX/pixstage-0", 0 );
+   ALLOC_STATE( pix[1], tex, PIX_STATE_SIZE, "PIX/pixstage-1", 1 );
+
+
+   /* Fill in the packet headers:
+    */
+   rmesa->hw.ctx.cmd[CTX_CMD_0] = cmdpkt(RADEON_EMIT_PP_MISC);
+   rmesa->hw.ctx.cmd[CTX_CMD_1] = cmdpkt(RADEON_EMIT_PP_CNTL);
+   rmesa->hw.ctx.cmd[CTX_CMD_2] = cmdpkt(RADEON_EMIT_RB3D_COLORPITCH);
+   rmesa->hw.lin.cmd[LIN_CMD_0] = cmdpkt(RADEON_EMIT_RE_LINE_PATTERN);
+   rmesa->hw.lin.cmd[LIN_CMD_1] = cmdpkt(RADEON_EMIT_SE_LINE_WIDTH);
+   rmesa->hw.msk.cmd[MSK_CMD_0] = cmdpkt(RADEON_EMIT_RB3D_STENCILREFMASK);
+   rmesa->hw.vpt.cmd[VPT_CMD_0] = cmdpkt(RADEON_EMIT_SE_VPORT_XSCALE);
+   rmesa->hw.set.cmd[SET_CMD_0] = cmdpkt(RADEON_EMIT_SE_CNTL);
+   rmesa->hw.msc.cmd[MSC_CMD_0] = cmdpkt(RADEON_EMIT_RE_MISC);
+   rmesa->hw.cst.cmd[CST_CMD_0] = cmdpkt(R200_EMIT_PP_CNTL_X);
+   rmesa->hw.cst.cmd[CST_CMD_1] = cmdpkt(R200_EMIT_RB3D_DEPTHXY_OFFSET);
+   rmesa->hw.cst.cmd[CST_CMD_2] = cmdpkt(R200_EMIT_RE_AUX_SCISSOR_CNTL);
+   rmesa->hw.cst.cmd[CST_CMD_3] = cmdpkt(R200_EMIT_RE_SCISSOR_TL_0);
+   rmesa->hw.cst.cmd[CST_CMD_4] = cmdpkt(R200_EMIT_SE_VAP_CNTL_STATUS);
+   rmesa->hw.cst.cmd[CST_CMD_5] = cmdpkt(R200_EMIT_RE_POINTSIZE);
+   rmesa->hw.cst.cmd[CST_CMD_6] = cmdpkt(R200_EMIT_TCL_INPUT_VTX_VECTOR_ADDR_0);
+   rmesa->hw.tam.cmd[TAM_CMD_0] = cmdpkt(R200_EMIT_PP_TAM_DEBUG3);
+   rmesa->hw.tf.cmd[TF_CMD_0]   = cmdpkt(R200_EMIT_TFACTOR_0);
+   rmesa->hw.tex[0].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXFILTER_0);
+   rmesa->hw.tex[0].cmd[TEX_CMD_1] = cmdpkt(R200_EMIT_PP_TXOFFSET_0);
+   rmesa->hw.tex[1].cmd[TEX_CMD_0] = cmdpkt(R200_EMIT_PP_TXFILTER_1);
+   rmesa->hw.tex[1].cmd[TEX_CMD_1] = cmdpkt(R200_EMIT_PP_TXOFFSET_1);
+   rmesa->hw.cube[0].cmd[CUBE_CMD_0] = cmdpkt(R200_EMIT_PP_CUBIC_FACES_0);
+   rmesa->hw.cube[0].cmd[CUBE_CMD_1] = cmdpkt(R200_EMIT_PP_CUBIC_OFFSETS_0);
+   rmesa->hw.cube[1].cmd[CUBE_CMD_0] = cmdpkt(R200_EMIT_PP_CUBIC_FACES_1);
+   rmesa->hw.cube[1].cmd[CUBE_CMD_1] = cmdpkt(R200_EMIT_PP_CUBIC_OFFSETS_1);
+   rmesa->hw.pix[0].cmd[PIX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCBLEND_0);
+   rmesa->hw.pix[1].cmd[PIX_CMD_0] = cmdpkt(R200_EMIT_PP_TXCBLEND_1);
+   rmesa->hw.zbs.cmd[ZBS_CMD_0] = cmdpkt(RADEON_EMIT_SE_ZBIAS_FACTOR);
+   rmesa->hw.tcl.cmd[TCL_CMD_0] = cmdpkt(R200_EMIT_TCL_LIGHT_MODEL_CTL_0);
+   rmesa->hw.tcl.cmd[TCL_CMD_1] = cmdpkt(R200_EMIT_TCL_UCP_VERT_BLEND_CTL);
+   rmesa->hw.tcg.cmd[TCG_CMD_0] = cmdpkt(R200_EMIT_TEX_PROC_CTL_2);
+   rmesa->hw.msl.cmd[MSL_CMD_0] = cmdpkt(R200_EMIT_MATRIX_SELECT_0);
+   rmesa->hw.vap.cmd[VAP_CMD_0] = cmdpkt(R200_EMIT_VAP_CTL);
+   rmesa->hw.vtx.cmd[VTX_CMD_0] = cmdpkt(R200_EMIT_VTX_FMT_0);
+   rmesa->hw.vtx.cmd[VTX_CMD_1] = cmdpkt(R200_EMIT_OUTPUT_VTX_COMP_SEL);
+   rmesa->hw.vtx.cmd[VTX_CMD_2] = cmdpkt(R200_EMIT_SE_VTX_STATE_CNTL);
+   rmesa->hw.vte.cmd[VTE_CMD_0] = cmdpkt(R200_EMIT_VTE_CNTL);
+   rmesa->hw.mtl[0].cmd[MTL_CMD_0] = 
+      cmdvec( R200_VS_MAT_0_EMISS, 1, 16 );
+   rmesa->hw.mtl[0].cmd[MTL_CMD_1] = 
+      cmdscl2( R200_SS_MAT_0_SHININESS, 1, 1 );
+   rmesa->hw.grd.cmd[GRD_CMD_0] = 
+      cmdscl( R200_SS_VERT_GUARD_CLIP_ADJ_ADDR, 1, 4 );
+   rmesa->hw.fog.cmd[FOG_CMD_0] = 
+      cmdvec( R200_VS_FOG_PARAM_ADDR, 1, 4 );
+   rmesa->hw.glt.cmd[GLT_CMD_0] = 
+      cmdvec( R200_VS_GLOBAL_AMBIENT_ADDR, 1, 4 );
+   rmesa->hw.eye.cmd[EYE_CMD_0] = 
+      cmdvec( R200_VS_EYE_VECTOR_ADDR, 1, 4 );
+
+   rmesa->hw.mat[R200_MTX_MV].cmd[MAT_CMD_0] = 
+      cmdvec( R200_VS_MATRIX_0_MV, 1, 16);
+   rmesa->hw.mat[R200_MTX_IMV].cmd[MAT_CMD_0] = 
+      cmdvec( R200_VS_MATRIX_1_INV_MV, 1, 16);
+   rmesa->hw.mat[R200_MTX_MVP].cmd[MAT_CMD_0] = 
+      cmdvec( R200_VS_MATRIX_2_MVP, 1, 16);
+   rmesa->hw.mat[R200_MTX_TEX0].cmd[MAT_CMD_0] = 
+      cmdvec( R200_VS_MATRIX_3_TEX0, 1, 16);
+   rmesa->hw.mat[R200_MTX_TEX1].cmd[MAT_CMD_0] = 
+      cmdvec( R200_VS_MATRIX_4_TEX1, 1, 16);
+
+   for (i = 0 ; i < 8; i++) {
+      rmesa->hw.lit[i].cmd[LIT_CMD_0] = 
+	 cmdvec( R200_VS_LIGHT_AMBIENT_ADDR + i, 8, 24 );
+      rmesa->hw.lit[i].cmd[LIT_CMD_1] = 
+	 cmdscl( R200_SS_LIGHT_DCD_ADDR + i, 8, 7 );
+   }
+
+   for (i = 0 ; i < 6; i++) {
+      rmesa->hw.ucp[i].cmd[UCP_CMD_0] = 
+	 cmdvec( R200_VS_UCP_ADDR + i, 1, 4 );
+   }
+
+   /* Initial Harware state:
+    */
+   rmesa->hw.ctx.cmd[CTX_PP_MISC] = (R200_ALPHA_TEST_PASS
+				     /* | R200_RIGHT_HAND_CUBE_OGL*/);
+
+   rmesa->hw.ctx.cmd[CTX_PP_FOG_COLOR] = (R200_FOG_VERTEX |
+					  R200_FOG_USE_SPEC_ALPHA);
+
+   rmesa->hw.ctx.cmd[CTX_RE_SOLID_COLOR] = 0x00000000;
+
+   rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCNTL] = (R200_COMB_FCN_ADD_CLAMP |
+					    R200_SRC_BLEND_GL_ONE |
+					    R200_DST_BLEND_GL_ZERO );
+
+   rmesa->hw.ctx.cmd[CTX_RB3D_DEPTHOFFSET] =
+      rmesa->r200Screen->depthOffset;
+
+   rmesa->hw.ctx.cmd[CTX_RB3D_DEPTHPITCH] = 
+      ((rmesa->r200Screen->depthPitch &
+	R200_DEPTHPITCH_MASK) |
+       R200_DEPTH_ENDIAN_NO_SWAP);
+
+   rmesa->hw.ctx.cmd[CTX_RB3D_ZSTENCILCNTL] = (depth_fmt |
+  					       R200_Z_TEST_LESS |  
+					       R200_STENCIL_TEST_ALWAYS |
+					       R200_STENCIL_FAIL_KEEP |
+					       R200_STENCIL_ZPASS_KEEP |
+					       R200_STENCIL_ZFAIL_KEEP |
+					       R200_Z_WRITE_ENABLE);
+
+   rmesa->hw.ctx.cmd[CTX_PP_CNTL] = (R200_ANTI_ALIAS_NONE 
+ 				     | R200_TEX_BLEND_0_ENABLE);
+
+   rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] = color_fmt;
+   rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |=  R200_DITHER_ENABLE;
+
+   rmesa->hw.ctx.cmd[CTX_RB3D_COLOROFFSET] = (rmesa->state.color.drawOffset &
+					      R200_COLOROFFSET_MASK);
+
+   rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] = ((rmesa->state.color.drawPitch &
+					      R200_COLORPITCH_MASK) |
+					     R200_COLOR_ENDIAN_NO_SWAP);
+
+   rmesa->hw.set.cmd[SET_SE_CNTL] = (R200_FFACE_CULL_CCW |
+				     R200_BFACE_SOLID |
+				     R200_FFACE_SOLID |
+				     R200_FLAT_SHADE_VTX_LAST |
+				     R200_DIFFUSE_SHADE_GOURAUD |
+				     R200_ALPHA_SHADE_GOURAUD |
+				     R200_SPECULAR_SHADE_GOURAUD |
+				     R200_FOG_SHADE_GOURAUD |
+				     R200_VTX_PIX_CENTER_OGL |
+				     R200_ROUND_MODE_TRUNC |
+				     R200_ROUND_PREC_8TH_PIX);
+
+   rmesa->hw.set.cmd[SET_RE_CNTL] = (R200_PERSPECTIVE_ENABLE |
+				     R200_SCISSOR_ENABLE);
+
+   rmesa->hw.lin.cmd[LIN_RE_LINE_PATTERN] = ((1 << 16) | 0xffff);
+
+   rmesa->hw.lin.cmd[LIN_RE_LINE_STATE] = 
+      ((0 << R200_LINE_CURRENT_PTR_SHIFT) |
+       (1 << R200_LINE_CURRENT_COUNT_SHIFT));
+
+   rmesa->hw.lin.cmd[LIN_SE_LINE_WIDTH] = (1 << 4);
+
+   rmesa->hw.msk.cmd[MSK_RB3D_STENCILREFMASK] = 
+      ((0x00 << R200_STENCIL_REF_SHIFT) |
+       (0xff << R200_STENCIL_MASK_SHIFT) |
+       (0xff << R200_STENCIL_WRITEMASK_SHIFT));
+
+   rmesa->hw.msk.cmd[MSK_RB3D_ROPCNTL] = R200_ROP_COPY;
+   rmesa->hw.msk.cmd[MSK_RB3D_PLANEMASK] = 0xffffffff;
+
+   rmesa->hw.tam.cmd[TAM_DEBUG3] = 0;
+
+   rmesa->hw.msc.cmd[MSC_RE_MISC] = 
+      ((0 << R200_STIPPLE_X_OFFSET_SHIFT) |
+       (0 << R200_STIPPLE_Y_OFFSET_SHIFT) |
+       R200_STIPPLE_BIG_BIT_ORDER);
+
+
+   rmesa->hw.cst.cmd[CST_PP_CNTL_X] = 0;
+   rmesa->hw.cst.cmd[CST_RB3D_DEPTHXY_OFFSET] = 0;
+   rmesa->hw.cst.cmd[CST_RE_AUX_SCISSOR_CNTL] = 0x0;
+   rmesa->hw.cst.cmd[CST_RE_SCISSOR_TL_0] = 0;
+   rmesa->hw.cst.cmd[CST_RE_SCISSOR_BR_0] = 0;
+   rmesa->hw.cst.cmd[CST_SE_VAP_CNTL_STATUS] =
+#ifdef MESA_BIG_ENDIAN
+						R200_VC_32BIT_SWAP;
+#else
+						R200_VC_NO_SWAP;
+#endif
+   rmesa->hw.cst.cmd[CST_RE_POINTSIZE] = 0x100010;
+   rmesa->hw.cst.cmd[CST_SE_TCL_INPUT_VTX_0] =
+      (0x0 << R200_VERTEX_POSITION_ADDR__SHIFT);
+   rmesa->hw.cst.cmd[CST_SE_TCL_INPUT_VTX_1] =
+      (0x02 << R200_VTX_COLOR_0_ADDR__SHIFT) |
+      (0x03 << R200_VTX_COLOR_1_ADDR__SHIFT);
+   rmesa->hw.cst.cmd[CST_SE_TCL_INPUT_VTX_2] =
+      (0x06 << R200_VTX_TEX_0_ADDR__SHIFT) |
+      (0x07 << R200_VTX_TEX_1_ADDR__SHIFT) |
+      (0x08 << R200_VTX_TEX_2_ADDR__SHIFT) |
+      (0x09 << R200_VTX_TEX_3_ADDR__SHIFT);
+   rmesa->hw.cst.cmd[CST_SE_TCL_INPUT_VTX_3] =
+      (0x0A << R200_VTX_TEX_4_ADDR__SHIFT) |
+      (0x0B << R200_VTX_TEX_5_ADDR__SHIFT);
+  
+
+   rmesa->hw.vpt.cmd[VPT_SE_VPORT_XSCALE]  = 0x00000000;
+   rmesa->hw.vpt.cmd[VPT_SE_VPORT_XOFFSET] = 0x00000000;
+   rmesa->hw.vpt.cmd[VPT_SE_VPORT_YSCALE]  = 0x00000000;
+   rmesa->hw.vpt.cmd[VPT_SE_VPORT_YOFFSET] = 0x00000000;
+   rmesa->hw.vpt.cmd[VPT_SE_VPORT_ZSCALE]  = 0x00000000;
+   rmesa->hw.vpt.cmd[VPT_SE_VPORT_ZOFFSET] = 0x00000000;
+
+   for (i = 0; i < 2; i++) { /* 2 texture units for now */
+      rmesa->hw.tex[i].cmd[TEX_PP_TXFILTER] = R200_BORDER_MODE_OGL;
+      rmesa->hw.tex[i].cmd[TEX_PP_TXFORMAT] = 
+         ((i << R200_TXFORMAT_ST_ROUTE_SHIFT) |  /* <-- note i */
+          (2 << R200_TXFORMAT_WIDTH_SHIFT) |
+          (2 << R200_TXFORMAT_HEIGHT_SHIFT));
+      rmesa->hw.tex[i].cmd[TEX_PP_TXOFFSET] = 0;
+      rmesa->hw.tex[i].cmd[TEX_PP_BORDER_COLOR] = 0;
+      rmesa->hw.tex[i].cmd[TEX_PP_TXFORMAT_X] =
+         (/* R200_TEXCOORD_PROJ | */
+          0x100000);	/* Small default bias */
+
+      rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_FACES] = 0;
+      rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_F1] = 0;
+      rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_F2] = 0;
+      rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_F3] = 0;
+      rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_F4] = 0;
+      rmesa->hw.cube[i].cmd[CUBE_PP_CUBIC_OFFSET_F5] = 0;
+   }
+
+   rmesa->hw.pix[0].cmd[PIX_PP_TXCBLEND] =  
+      (R200_TXC_ARG_A_ZERO |
+       R200_TXC_ARG_B_ZERO |
+       R200_TXC_ARG_C_DIFFUSE_COLOR |
+       R200_TXC_OP_MADD);
+
+   rmesa->hw.pix[0].cmd[PIX_PP_TXCBLEND2] =  
+      ((0 << R200_TXC_TFACTOR_SEL_SHIFT) | 
+       R200_TXC_SCALE_1X |
+       R200_TXC_CLAMP_0_1 |
+       R200_TXC_OUTPUT_REG_R0);
+
+   rmesa->hw.pix[0].cmd[PIX_PP_TXABLEND] =  
+      (R200_TXA_ARG_A_ZERO |
+       R200_TXA_ARG_B_ZERO |
+       R200_TXA_ARG_C_DIFFUSE_ALPHA |
+       R200_TXA_OP_MADD);
+
+   rmesa->hw.pix[0].cmd[PIX_PP_TXABLEND2] =  
+      ((0 << R200_TXA_TFACTOR_SEL_SHIFT) | 
+       R200_TXA_SCALE_1X |
+       R200_TXA_CLAMP_0_1 |
+       R200_TXA_OUTPUT_REG_R0);
+
+   rmesa->hw.pix[1].cmd[PIX_PP_TXCBLEND] =  
+      (R200_TXC_ARG_A_ZERO |
+       R200_TXC_ARG_B_ZERO |
+       R200_TXC_ARG_C_DIFFUSE_COLOR |
+       R200_TXC_OP_MADD);
+
+   rmesa->hw.pix[1].cmd[PIX_PP_TXCBLEND2] =  
+      ((0 << R200_TXC_TFACTOR_SEL_SHIFT) | 
+       R200_TXC_SCALE_1X |
+       R200_TXC_CLAMP_0_1 |
+       R200_TXC_OUTPUT_REG_R0);
+
+   rmesa->hw.pix[1].cmd[PIX_PP_TXABLEND] =  
+      (R200_TXA_ARG_A_ZERO |
+       R200_TXA_ARG_B_ZERO |
+       R200_TXA_ARG_C_DIFFUSE_ALPHA |
+       R200_TXA_OP_MADD);
+
+   rmesa->hw.pix[1].cmd[PIX_PP_TXABLEND2] =  
+      ((0 << R200_TXA_TFACTOR_SEL_SHIFT) | 
+       R200_TXA_SCALE_1X |
+       R200_TXA_CLAMP_0_1 |
+       R200_TXA_OUTPUT_REG_R0);
+
+   rmesa->hw.tf.cmd[TF_TFACTOR_0] = 0;
+   rmesa->hw.tf.cmd[TF_TFACTOR_1] = 0;
+   rmesa->hw.tf.cmd[TF_TFACTOR_2] = 0;
+   rmesa->hw.tf.cmd[TF_TFACTOR_3] = 0;
+   rmesa->hw.tf.cmd[TF_TFACTOR_4] = 0;
+   rmesa->hw.tf.cmd[TF_TFACTOR_5] = 0;
+
+   rmesa->hw.vap.cmd[VAP_SE_VAP_CNTL] = 
+      (R200_VAP_TCL_ENABLE | 
+       (0x9 << R200_VAP_VF_MAX_VTX_NUM__SHIFT));
+
+   rmesa->hw.vte.cmd[VTE_SE_VTE_CNTL] = 
+      (R200_VPORT_X_SCALE_ENA |
+       R200_VPORT_Y_SCALE_ENA |
+       R200_VPORT_Z_SCALE_ENA |
+       R200_VPORT_X_OFFSET_ENA |
+       R200_VPORT_Y_OFFSET_ENA |
+       R200_VPORT_Z_OFFSET_ENA |
+/* FIXME: Turn on for tex rect only */
+       R200_VTX_ST_DENORMALIZED |  
+       R200_VTX_W0_FMT); 
+
+
+   rmesa->hw.vtx.cmd[VTX_VTXFMT_0] = 0;
+   rmesa->hw.vtx.cmd[VTX_VTXFMT_1] = 0;
+   rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_0] = 
+      ((R200_VTX_Z0 | R200_VTX_W0 |
+       (R200_VTX_FP_RGBA << R200_VTX_COLOR_0_SHIFT)));	
+   rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_VTXFMT_1] = 0;
+   rmesa->hw.vtx.cmd[VTX_TCL_OUTPUT_COMPSEL] = (R200_OUTPUT_XYZW);
+   rmesa->hw.vtx.cmd[VTX_STATE_CNTL] = R200_VSC_UPDATE_USER_COLOR_0_ENABLE;
+						   
+
+   /* Matrix selection */
+   rmesa->hw.msl.cmd[MSL_MATRIX_SELECT_0] = 
+      (R200_MTX_MV << R200_MODELVIEW_0_SHIFT);
+   
+   rmesa->hw.msl.cmd[MSL_MATRIX_SELECT_1] = 
+       (R200_MTX_IMV << R200_IT_MODELVIEW_0_SHIFT);
+
+   rmesa->hw.msl.cmd[MSL_MATRIX_SELECT_2] = 
+      (R200_MTX_MVP << R200_MODELPROJECT_0_SHIFT);
+
+   rmesa->hw.msl.cmd[MSL_MATRIX_SELECT_3] = 
+      ((R200_MTX_TEX0 << R200_TEXMAT_0_SHIFT) |
+       (R200_MTX_TEX1 << R200_TEXMAT_1_SHIFT) |
+       (R200_MTX_TEX2 << R200_TEXMAT_2_SHIFT) |
+       (R200_MTX_TEX3 << R200_TEXMAT_3_SHIFT));
+
+   rmesa->hw.msl.cmd[MSL_MATRIX_SELECT_4] = 
+      ((R200_MTX_TEX4 << R200_TEXMAT_4_SHIFT) |
+       (R200_MTX_TEX5 << R200_TEXMAT_5_SHIFT));
+
+
+   /* General TCL state */
+   rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_0] = 
+      (R200_SPECULAR_LIGHTS |
+       R200_DIFFUSE_SPECULAR_COMBINE |
+       R200_LOCAL_LIGHT_VEC_GL);
+
+   rmesa->hw.tcl.cmd[TCL_LIGHT_MODEL_CTL_1] = 
+      ((R200_LM1_SOURCE_LIGHT_PREMULT << R200_FRONT_EMISSIVE_SOURCE_SHIFT) |
+       (R200_LM1_SOURCE_LIGHT_PREMULT << R200_FRONT_AMBIENT_SOURCE_SHIFT) |
+       (R200_LM1_SOURCE_LIGHT_PREMULT << R200_FRONT_DIFFUSE_SOURCE_SHIFT) |
+       (R200_LM1_SOURCE_LIGHT_PREMULT << R200_FRONT_SPECULAR_SOURCE_SHIFT) |
+       (R200_LM1_SOURCE_LIGHT_PREMULT << R200_BACK_EMISSIVE_SOURCE_SHIFT) |
+       (R200_LM1_SOURCE_LIGHT_PREMULT << R200_BACK_AMBIENT_SOURCE_SHIFT) |
+       (R200_LM1_SOURCE_LIGHT_PREMULT << R200_BACK_DIFFUSE_SOURCE_SHIFT) |
+       (R200_LM1_SOURCE_LIGHT_PREMULT << R200_BACK_SPECULAR_SOURCE_SHIFT)); 
+
+   rmesa->hw.tcl.cmd[TCL_PER_LIGHT_CTL_0] = 0; /* filled in via callbacks */
+   rmesa->hw.tcl.cmd[TCL_PER_LIGHT_CTL_1] = 0;
+   rmesa->hw.tcl.cmd[TCL_PER_LIGHT_CTL_2] = 0;
+   rmesa->hw.tcl.cmd[TCL_PER_LIGHT_CTL_3] = 0;
+   
+   rmesa->hw.tcl.cmd[TCL_UCP_VERT_BLEND_CTL] = 
+      (R200_UCP_IN_CLIP_SPACE |
+       R200_CULL_FRONT_IS_CCW);
+
+   /* Texgen/Texmat state */
+   rmesa->hw.tcg.cmd[TCG_TEX_PROC_CTL_2] = 0x0; /* masks??? */
+   rmesa->hw.tcg.cmd[TCG_TEX_PROC_CTL_3] = 
+      ((0 << R200_TEXGEN_0_INPUT_TEX_SHIFT) |
+       (1 << R200_TEXGEN_1_INPUT_TEX_SHIFT) |
+       (2 << R200_TEXGEN_2_INPUT_TEX_SHIFT) |
+       (3 << R200_TEXGEN_3_INPUT_TEX_SHIFT) |
+       (4 << R200_TEXGEN_4_INPUT_TEX_SHIFT) |
+       (5 << R200_TEXGEN_5_INPUT_TEX_SHIFT)); 
+   rmesa->hw.tcg.cmd[TCG_TEX_PROC_CTL_0] = 0; 
+   rmesa->hw.tcg.cmd[TCG_TEX_PROC_CTL_1] =  
+      ((0 << R200_TEXGEN_0_INPUT_SHIFT) |
+       (1 << R200_TEXGEN_1_INPUT_SHIFT) |
+       (2 << R200_TEXGEN_2_INPUT_SHIFT) |
+       (3 << R200_TEXGEN_3_INPUT_SHIFT) |
+       (4 << R200_TEXGEN_4_INPUT_SHIFT) |
+       (5 << R200_TEXGEN_5_INPUT_SHIFT)); 
+   rmesa->hw.tcg.cmd[TCG_TEX_CYL_WRAP_CTL] = 0;
+
+   rmesa->TexGenInputs = rmesa->hw.tcg.cmd[TCG_TEX_PROC_CTL_1];
+
+
+   for (i = 0 ; i < 8; i++) {
+      struct gl_light *l = &ctx->Light.Light[i];
+      GLenum p = GL_LIGHT0 + i;
+      *(float *)&(rmesa->hw.lit[i].cmd[LIT_RANGE_CUTOFF]) = FLT_MAX;
+
+      ctx->Driver.Lightfv( ctx, p, GL_AMBIENT, l->Ambient );
+      ctx->Driver.Lightfv( ctx, p, GL_DIFFUSE, l->Diffuse );
+      ctx->Driver.Lightfv( ctx, p, GL_SPECULAR, l->Specular );
+      ctx->Driver.Lightfv( ctx, p, GL_POSITION, 0 );
+      ctx->Driver.Lightfv( ctx, p, GL_SPOT_DIRECTION, 0 );
+      ctx->Driver.Lightfv( ctx, p, GL_SPOT_EXPONENT, &l->SpotExponent );
+      ctx->Driver.Lightfv( ctx, p, GL_SPOT_CUTOFF, &l->SpotCutoff );
+      ctx->Driver.Lightfv( ctx, p, GL_CONSTANT_ATTENUATION,
+			   &l->ConstantAttenuation );
+      ctx->Driver.Lightfv( ctx, p, GL_LINEAR_ATTENUATION, 
+			   &l->LinearAttenuation );
+      ctx->Driver.Lightfv( ctx, p, GL_QUADRATIC_ATTENUATION, 
+		     &l->QuadraticAttenuation );
+   }
+
+   ctx->Driver.LightModelfv( ctx, GL_LIGHT_MODEL_AMBIENT, 
+			     ctx->Light.Model.Ambient );
+
+   TNL_CONTEXT(ctx)->Driver.NotifyMaterialChange( ctx );
+
+   for (i = 0 ; i < 6; i++) {
+      ctx->Driver.ClipPlane( ctx, GL_CLIP_PLANE0 + i, NULL );
+   }
+
+   ctx->Driver.Fogfv( ctx, GL_FOG_MODE, 0 );
+   ctx->Driver.Fogfv( ctx, GL_FOG_DENSITY, &ctx->Fog.Density );
+   ctx->Driver.Fogfv( ctx, GL_FOG_START, &ctx->Fog.Start );
+   ctx->Driver.Fogfv( ctx, GL_FOG_END, &ctx->Fog.End );
+   ctx->Driver.Fogfv( ctx, GL_FOG_COLOR, ctx->Fog.Color );
+   ctx->Driver.Fogfv( ctx, GL_FOG_COORDINATE_SOURCE_EXT, 0 );
+   
+   rmesa->hw.grd.cmd[GRD_VERT_GUARD_CLIP_ADJ] = IEEE_ONE;
+   rmesa->hw.grd.cmd[GRD_VERT_GUARD_DISCARD_ADJ] = IEEE_ONE;
+   rmesa->hw.grd.cmd[GRD_HORZ_GUARD_CLIP_ADJ] = IEEE_ONE;
+   rmesa->hw.grd.cmd[GRD_HORZ_GUARD_DISCARD_ADJ] = IEEE_ONE;
+
+   rmesa->hw.eye.cmd[EYE_X] = 0;
+   rmesa->hw.eye.cmd[EYE_Y] = 0;
+   rmesa->hw.eye.cmd[EYE_Z] = IEEE_ONE;
+   rmesa->hw.eye.cmd[EYE_RESCALE_FACTOR] = IEEE_ONE;
+
+   r200LightingSpaceChange( ctx );
+   
+   rmesa->lost_context = 1;
+}

src/mesa/drivers/dri/r200/r200_swtcl.h

@@ -0,0 +1,80 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#ifndef __R200_SWTCL_H__
+#define __R200_SWTCL_H__
+
+#include "mtypes.h"
+#include "swrast/swrast.h"
+#include "r200_context.h"
+
+extern void r200InitSwtcl( GLcontext *ctx );
+extern void r200DestroySwtcl( GLcontext *ctx );
+
+extern void r200FlushVertices( GLcontext *ctx, GLuint flags );
+extern void r200ChooseRenderState( GLcontext *ctx );
+extern void r200ChooseVertexState( GLcontext *ctx );
+
+extern void r200CheckTexSizes( GLcontext *ctx );
+
+extern void r200BuildVertices( GLcontext *ctx, GLuint start, GLuint count,
+				 GLuint newinputs );
+
+extern void r200PrintSetupFlags(char *msg, GLuint flags );
+
+
+extern void r200_emit_contiguous_verts( GLcontext *ctx,
+					  GLuint start,
+					  GLuint count );
+
+extern void r200_emit_indexed_verts( GLcontext *ctx,
+				       GLuint start,
+				       GLuint count );
+
+extern void r200_translate_vertex( GLcontext *ctx, 
+				     const r200Vertex *src, 
+				     SWvertex *dst );
+
+extern void r200_print_vertex( GLcontext *ctx, const r200Vertex *v );
+
+extern void r200_import_float_colors( GLcontext *ctx );
+extern void r200_import_float_spec_colors( GLcontext *ctx );
+
+extern void r200PointsBitmap( GLcontext *ctx, GLint px, GLint py,
+			      GLsizei width, GLsizei height,
+			      const struct gl_pixelstore_attrib *unpack,
+			      const GLubyte *bitmap );
+
+
+#endif

src/mesa/drivers/dri/r200/r200_tcl.c

@@ -0,0 +1,555 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#include "glheader.h"
+#include "imports.h"
+#include "mmath.h"
+#include "mtypes.h"
+#include "enums.h"
+#include "colormac.h"
+#include "light.h"
+
+#include "array_cache/acache.h"
+#include "tnl/tnl.h"
+#include "tnl/t_pipeline.h"
+
+#include "r200_context.h"
+#include "r200_state.h"
+#include "r200_ioctl.h"
+#include "r200_tex.h"
+#include "r200_tcl.h"
+#include "r200_swtcl.h"
+#include "r200_maos.h"
+
+
+
+#define HAVE_POINTS      1
+#define HAVE_LINES       1
+#define HAVE_LINE_LOOP   0
+#define HAVE_LINE_STRIPS 1
+#define HAVE_TRIANGLES   1
+#define HAVE_TRI_STRIPS  1
+#define HAVE_TRI_STRIP_1 0
+#define HAVE_TRI_FANS    1
+#define HAVE_QUADS       0	/* hw quad verts in wrong order??? */
+#define HAVE_QUAD_STRIPS 1
+#define HAVE_POLYGONS    1
+#define HAVE_ELTS        1
+
+
+#define HW_POINTS           R200_VF_PRIM_POINTS
+#define HW_LINES            R200_VF_PRIM_LINES
+#define HW_LINE_LOOP        0
+#define HW_LINE_STRIP       R200_VF_PRIM_LINE_STRIP
+#define HW_TRIANGLES        R200_VF_PRIM_TRIANGLES
+#define HW_TRIANGLE_STRIP_0 R200_VF_PRIM_TRIANGLE_STRIP
+#define HW_TRIANGLE_STRIP_1 0
+#define HW_TRIANGLE_FAN     R200_VF_PRIM_TRIANGLE_FAN
+#define HW_QUADS            R200_VF_PRIM_QUADS
+#define HW_QUAD_STRIP       R200_VF_PRIM_QUAD_STRIP
+#define HW_POLYGON          R200_VF_PRIM_POLYGON
+
+
+static GLboolean discrete_prim[0x10] = {
+   0,				/* 0 none */
+   1,				/* 1 points */
+   1,				/* 2 lines */
+   0,				/* 3 line_strip */
+   1,				/* 4 tri_list */
+   0,				/* 5 tri_fan */
+   0,				/* 6 tri_strip */
+   0,				/* 7 tri_w_flags */
+   1,				/* 8 rect list (unused) */
+   1,				/* 9 3vert point */
+   1,				/* a 3vert line */
+   0,				/* b point sprite */
+   0,				/* c line loop */
+   1,				/* d quads */
+   0,				/* e quad strip */
+   0,				/* f polygon */
+};
+   
+
+#define LOCAL_VARS r200ContextPtr rmesa = R200_CONTEXT(ctx)
+#define ELTS_VARS  GLushort *dest
+
+#define ELT_INIT(prim, hw_prim) \
+   r200TclPrimitive( ctx, prim, hw_prim | R200_VF_PRIM_WALK_IND )
+
+#define GET_ELTS() rmesa->tcl.Elts
+
+
+#define NEW_PRIMITIVE()  R200_NEWPRIM( rmesa )
+#define NEW_BUFFER()  r200RefillCurrentDmaRegion( rmesa )
+
+/* Don't really know how many elts will fit in what's left of cmdbuf,
+ * as there is state to emit, etc:
+ */
+
+#if 0
+#define GET_CURRENT_VB_MAX_ELTS() \
+   ((R200_CMD_BUF_SZ - (rmesa->store.cmd_used + 16)) / 2) 
+#define GET_SUBSEQUENT_VB_MAX_ELTS() ((R200_CMD_BUF_SZ - 16) / 2) 
+#else
+/* Testing on isosurf shows a maximum around here.  Don't know if it's
+ * the card or driver or kernel module that is causing the behaviour.
+ */
+#define GET_CURRENT_VB_MAX_ELTS() 300
+#define GET_SUBSEQUENT_VB_MAX_ELTS() 300
+#endif
+
+#define RESET_STIPPLE() do {			\
+   R200_STATECHANGE( rmesa, lin );		\
+   r200EmitState( rmesa );			\
+} while (0)
+
+#define AUTO_STIPPLE( mode )  do {		\
+   R200_STATECHANGE( rmesa, lin );		\
+   if (mode)					\
+      rmesa->hw.lin.cmd[LIN_RE_LINE_PATTERN] |=	\
+	 R200_LINE_PATTERN_AUTO_RESET;	\
+   else						\
+      rmesa->hw.lin.cmd[LIN_RE_LINE_PATTERN] &=	\
+	 ~R200_LINE_PATTERN_AUTO_RESET;	\
+   r200EmitState( rmesa );			\
+} while (0)
+
+
+/* How do you extend an existing primitive?
+ */
+#define ALLOC_ELTS(nr)							\
+do {									\
+   if (rmesa->dma.flush == r200FlushElts &&				\
+       rmesa->store.cmd_used + nr*2 < R200_CMD_BUF_SZ) {		\
+									\
+      dest = (GLushort *)(rmesa->store.cmd_buf + 			\
+			  rmesa->store.cmd_used);			\
+      rmesa->store.cmd_used += nr*2;					\
+   }									\
+   else {								\
+      if (rmesa->dma.flush)						\
+	 rmesa->dma.flush( rmesa );					\
+									\
+      r200EmitAOS( rmesa,						\
+	  	     rmesa->tcl.aos_components,				\
+		     rmesa->tcl.nr_aos_components,			\
+		     0 );						\
+									\
+      dest = r200AllocEltsOpenEnded( rmesa,				\
+				       rmesa->tcl.hw_primitive,		\
+				       nr );				\
+   }									\
+} while (0) 
+
+
+
+/* TODO: Try to extend existing primitive if both are identical,
+ * discrete and there are no intervening state changes.  (Somewhat
+ * duplicates changes to DrawArrays code)
+ */
+static void EMIT_PRIM( GLcontext *ctx, 
+		       GLenum prim, 
+		       GLuint hwprim, 
+		       GLuint start, 
+		       GLuint count)	
+{
+   r200ContextPtr rmesa = R200_CONTEXT( ctx );
+   r200TclPrimitive( ctx, prim, hwprim );
+   
+   r200EmitAOS( rmesa,
+		  rmesa->tcl.aos_components,
+		  rmesa->tcl.nr_aos_components,
+		  start );
+   
+   /* Why couldn't this packet have taken an offset param?
+    */
+   r200EmitVbufPrim( rmesa,
+		     rmesa->tcl.hw_primitive,
+		     count - start );
+}
+
+
+
+/* Try & join small primitives
+ */
+#if 0
+#define PREFER_DISCRETE_ELT_PRIM( NR, PRIM ) 0
+#else
+#define PREFER_DISCRETE_ELT_PRIM( NR, PRIM )			\
+  ((NR) < 20 ||							\
+   ((NR) < 40 &&						\
+    rmesa->tcl.hw_primitive == (PRIM|				\
+			    R200_VF_TCL_OUTPUT_VTX_ENABLE|	\
+			        R200_VF_PRIM_WALK_IND)))
+#endif
+
+#ifdef MESA_BIG_ENDIAN
+/* We could do without (most of) this ugliness if dest was always 32 bit word aligned... */
+#define EMIT_ELT(offset, x) do {                                \
+        int off = offset + ( ( (GLuint)dest & 0x2 ) >> 1 );     \
+        GLushort *des = (GLushort *)( (GLuint)dest & ~0x2 );    \
+        (des)[ off + 1 - 2 * ( off & 1 ) ] = (GLushort)(x); } while (0)
+#else
+#define EMIT_ELT(offset, x) (dest)[offset] = (GLushort) (x)
+#endif
+#define EMIT_TWO_ELTS(offset, x, y)  *(GLuint *)(dest+offset) = ((y)<<16)|(x);
+#define INCR_ELTS( nr ) dest += nr
+#define RELEASE_ELT_VERTS() \
+   r200ReleaseArrays( ctx, ~0 )
+
+
+
+#define TAG(x) tcl_##x
+#include "tnl_dd/t_dd_dmatmp2.h"
+
+/**********************************************************************/
+/*                          External entrypoints                     */
+/**********************************************************************/
+
+void r200EmitPrimitive( GLcontext *ctx, 
+			  GLuint first,
+			  GLuint last,
+			  GLuint flags )
+{
+   tcl_render_tab_verts[flags&PRIM_MODE_MASK]( ctx, first, last, flags );
+}
+
+void r200EmitEltPrimitive( GLcontext *ctx, 
+			     GLuint first,
+			     GLuint last,
+			     GLuint flags )
+{
+   tcl_render_tab_elts[flags&PRIM_MODE_MASK]( ctx, first, last, flags );
+}
+
+void r200TclPrimitive( GLcontext *ctx, 
+			 GLenum prim,
+			 int hw_prim )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   GLuint newprim = hw_prim | R200_VF_TCL_OUTPUT_VTX_ENABLE;
+
+   if (newprim != rmesa->tcl.hw_primitive ||
+       !discrete_prim[hw_prim&0xf]) {
+      R200_NEWPRIM( rmesa );
+      rmesa->tcl.hw_primitive = newprim;
+   }
+}
+
+
+/**********************************************************************/
+/*                          Render pipeline stage                     */
+/**********************************************************************/
+
+
+/* TCL render.
+ */
+static GLboolean r200_run_tcl_render( GLcontext *ctx,
+					struct gl_pipeline_stage *stage )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+   GLuint i,flags = 0,length;
+
+   /* TODO: separate this from the swtnl pipeline 
+    */
+   if (rmesa->TclFallback)
+      return GL_TRUE;	/* fallback to software t&l */
+
+   if (R200_DEBUG & DEBUG_PRIMS)
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+   if (VB->Count == 0)
+      return GL_FALSE;
+
+   r200ReleaseArrays( ctx, stage->changed_inputs );
+   r200EmitArrays( ctx, stage->inputs );
+
+   rmesa->tcl.Elts = VB->Elts;
+
+   for (i = VB->FirstPrimitive ; !(flags & PRIM_LAST) ; i += length)
+   {
+      flags = VB->Primitive[i];
+      length = VB->PrimitiveLength[i];
+
+      if (R200_DEBUG & DEBUG_PRIMS)
+	 fprintf(stderr, "%s: prim %s %d..%d\n", 
+		 __FUNCTION__,
+		 _mesa_lookup_enum_by_nr(flags & PRIM_MODE_MASK), 
+		 i, i+length);
+
+      if (!length)
+	 continue;
+
+      if (rmesa->tcl.Elts)
+	 r200EmitEltPrimitive( ctx, i, i+length, flags );
+      else
+	 r200EmitPrimitive( ctx, i, i+length, flags );
+   }
+
+   return GL_FALSE;		/* finished the pipe */
+}
+
+
+
+static void r200_check_tcl_render( GLcontext *ctx,
+				     struct gl_pipeline_stage *stage )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   GLuint inputs = VERT_BIT_POS;
+
+   /* Validate state:
+    */
+   if (rmesa->NewGLState)
+      r200ValidateState( ctx );
+
+   if (0)
+      fprintf(stderr, "%s: RE %d TGE %d NN %d\n",
+	   __FUNCTION__,
+	   ctx->Texture.Unit[0]._ReallyEnabled,
+	   ctx->Texture.Unit[0].TexGenEnabled,
+	   rmesa->TexGenNeedNormals[0]);
+
+   if (ctx->RenderMode == GL_RENDER) {
+      /* Make all this event-driven:
+       */
+      if (ctx->Light.Enabled) {
+	 inputs |= VERT_BIT_NORMAL;
+
+	 if (ctx->Light.ColorMaterialEnabled) {
+	    inputs |= VERT_BIT_COLOR0;
+	 }
+      }
+      else {
+	 inputs |= VERT_BIT_COLOR0;
+	 
+	 if (ctx->_TriangleCaps & DD_SEPARATE_SPECULAR) {
+	    inputs |= VERT_BIT_COLOR1;
+	 }
+      }
+
+      if (ctx->Texture.Unit[0]._ReallyEnabled) {
+	 if (ctx->Texture.Unit[0].TexGenEnabled) {
+	    if (rmesa->TexGenNeedNormals[0]) {
+	       inputs |= VERT_BIT_NORMAL;
+	    }
+	 } else {
+	    inputs |= VERT_BIT_TEX0;
+	 }
+      }
+
+      if (ctx->Texture.Unit[1]._ReallyEnabled) {
+	 if (ctx->Texture.Unit[1].TexGenEnabled) {
+	    if (rmesa->TexGenNeedNormals[1]) {
+	       inputs |= VERT_BIT_NORMAL;
+	    }
+	 } else {
+	    inputs |= VERT_BIT_TEX1;
+	 }
+      }
+
+      stage->inputs = inputs;
+      stage->active = 1;
+   }
+   else
+      stage->active = 0;
+}
+
+static void r200_init_tcl_render( GLcontext *ctx,
+				    struct gl_pipeline_stage *stage )
+{
+   stage->check = r200_check_tcl_render;
+   stage->check( ctx, stage );
+}
+
+static void dtr( struct gl_pipeline_stage *stage )
+{
+   (void)stage;
+}
+
+
+/* Initial state for tcl stage.  
+ */
+const struct gl_pipeline_stage _r200_tcl_stage =
+{
+   "r200 render",
+   (_DD_NEW_SEPARATE_SPECULAR |
+    _NEW_LIGHT|
+    _NEW_TEXTURE|
+    _NEW_FOG|
+    _NEW_RENDERMODE),		/* re-check (new inputs) */
+   0,				/* re-run (always runs) */
+   GL_TRUE,			/* active */
+   0, 0,			/* inputs (set in check_render), outputs */
+   0, 0,			/* changed_inputs, private */
+   dtr,				/* destructor */
+   r200_init_tcl_render,	/* check - initially set to alloc data */
+   r200_run_tcl_render	/* run */
+};
+
+
+
+/**********************************************************************/
+/*                 Validate state at pipeline start                   */
+/**********************************************************************/
+
+
+/*-----------------------------------------------------------------------
+ * Manage TCL fallbacks
+ */
+
+
+static void transition_to_swtnl( GLcontext *ctx )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+
+   R200_NEWPRIM( rmesa );
+   rmesa->swtcl.vertex_format = 0;
+
+   r200ChooseVertexState( ctx );
+   r200ChooseRenderState( ctx );
+
+   _mesa_validate_all_lighting_tables( ctx ); 
+
+   tnl->Driver.NotifyMaterialChange = 
+      _mesa_validate_all_lighting_tables;
+
+   r200ReleaseArrays( ctx, ~0 );
+
+   /* Still using the D3D based hardware-rasterizer from the radeon;
+    * need to put the card into D3D mode to make it work:
+    */
+   R200_STATECHANGE( rmesa, vap );
+   rmesa->hw.vap.cmd[VAP_SE_VAP_CNTL] &= ~R200_VAP_TCL_ENABLE;
+   rmesa->hw.vap.cmd[VAP_SE_VAP_CNTL] |= R200_VAP_D3D_TEX_DEFAULT;
+
+   R200_STATECHANGE( rmesa, vte );
+   rmesa->hw.vte.cmd[VTE_SE_VTE_CNTL] &= ~R200_VTX_W0_FMT;
+
+   R200_STATECHANGE( rmesa, set );
+   rmesa->hw.set.cmd[SET_RE_CNTL] |= (R200_VTX_STQ0_D3D |
+				      R200_VTX_STQ1_D3D);
+}
+
+
+static void transition_to_hwtnl( GLcontext *ctx )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+
+   _tnl_need_projected_coords( ctx, GL_FALSE );
+
+   r200UpdateMaterial( ctx );
+
+   tnl->Driver.NotifyMaterialChange = r200UpdateMaterial;
+
+   if ( rmesa->dma.flush )			
+      rmesa->dma.flush( rmesa );	
+
+   rmesa->dma.flush = 0;
+   rmesa->swtcl.vertex_format = 0;
+   
+   if (rmesa->swtcl.indexed_verts.buf) 
+      r200ReleaseDmaRegion( rmesa, &rmesa->swtcl.indexed_verts, 
+			      __FUNCTION__ );
+
+   R200_STATECHANGE( rmesa, vap );
+   rmesa->hw.vap.cmd[VAP_SE_VAP_CNTL] |= R200_VAP_TCL_ENABLE;
+   rmesa->hw.vap.cmd[VAP_SE_VAP_CNTL] &= ~(R200_VAP_FORCE_W_TO_ONE |
+					   R200_VAP_D3D_TEX_DEFAULT);
+
+   R200_STATECHANGE( rmesa, vte );
+   rmesa->hw.vte.cmd[VTE_SE_VTE_CNTL] &= ~(R200_VTX_XY_FMT|R200_VTX_Z_FMT);
+   rmesa->hw.vte.cmd[VTE_SE_VTE_CNTL] |= R200_VTX_W0_FMT;
+
+   R200_STATECHANGE( rmesa, set );
+   rmesa->hw.set.cmd[SET_RE_CNTL] &= ~(R200_VTX_STQ0_D3D |
+				       R200_VTX_STQ1_D3D);
+
+
+   if (R200_DEBUG & DEBUG_FALLBACKS) 
+      fprintf(stderr, "R200 end tcl fallback\n");
+}
+
+
+static char *fallbackStrings[] = {
+   "Rasterization fallback",
+   "Unfilled triangles",
+   "Twosided lighting, differing materials",
+   "Materials in VB (maybe between begin/end)",
+   "Texgen unit 0",
+   "Texgen unit 1",
+   "Texgen unit 2",
+   "User disable"
+};
+
+
+static char *getFallbackString(GLuint bit)
+{
+   int i = 0;
+   while (bit > 1) {
+      i++;
+      bit >>= 1;
+   }
+   return fallbackStrings[i];
+}
+
+
+
+void r200TclFallback( GLcontext *ctx, GLuint bit, GLboolean mode )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   GLuint oldfallback = rmesa->TclFallback;
+
+   if (mode) {
+      rmesa->TclFallback |= bit;
+      if (oldfallback == 0) {
+	 if (R200_DEBUG & DEBUG_FALLBACKS) 
+	    fprintf(stderr, "R200 begin tcl fallback %s\n",
+		    getFallbackString( bit ));
+	 transition_to_swtnl( ctx );
+      }
+   }
+   else {
+      rmesa->TclFallback &= ~bit;
+      if (oldfallback == bit) {
+	 if (R200_DEBUG & DEBUG_FALLBACKS) 
+	    fprintf(stderr, "R200 end tcl fallback %s\n",
+		    getFallbackString( bit ));
+	 transition_to_hwtnl( ctx );
+      }
+   }
+}

src/mesa/drivers/dri/r200/r200_tcl.h

@@ -0,0 +1,66 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#ifndef __R200_TCL_H__
+#define __R200_TCL_H__
+
+#ifdef GLX_DIRECT_RENDERING
+
+#include "r200_context.h"
+
+extern void r200TclPrimitive( GLcontext *ctx, GLenum prim, int hw_prim );
+extern void r200EmitEltPrimitive( GLcontext *ctx, GLuint first, GLuint last,
+				    GLuint flags );
+extern void r200EmitPrimitive( GLcontext *ctx, GLuint first, GLuint last,
+				 GLuint flags );
+
+extern void r200TclFallback( GLcontext *ctx, GLuint bit, GLboolean mode );
+					      
+#define R200_TCL_FALLBACK_RASTER            0x1 /* rasterization */
+#define R200_TCL_FALLBACK_UNFILLED          0x2 /* unfilled tris */
+#define R200_TCL_FALLBACK_LIGHT_TWOSIDE     0x4 /* twoside tris */
+#define R200_TCL_FALLBACK_MATERIAL          0x8 /* material in vb */
+#define R200_TCL_FALLBACK_TEXGEN_0          0x10 /* texgen, unit 0 */
+#define R200_TCL_FALLBACK_TEXGEN_1          0x20 /* texgen, unit 1 */
+#define R200_TCL_FALLBACK_TEXGEN_2          0x40 /* texgen, unit 2 */
+#define R200_TCL_FALLBACK_TCL_DISABLE       0x80 /* user disable */
+#define R200_TCL_FALLBACK_BITMAP            0x100 /* draw bitmap with points */
+
+#define R200_MAX_TCL_VERTSIZE (4*4) /* using maos now... */
+
+#define TCL_FALLBACK( ctx, bit, mode )	r200TclFallback( ctx, bit, mode )
+
+
+#endif
+#endif

src/mesa/drivers/dri/r200/r200_tex.c

@@ -0,0 +1,988 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#include "glheader.h"
+#include "imports.h"
+#include "colormac.h"
+#include "context.h"
+#include "enums.h"
+#include "image.h"
+#include "mmath.h"
+#include "simple_list.h"
+#include "texformat.h"
+#include "texstore.h"
+#include "texutil.h"
+
+#include "r200_context.h"
+#include "r200_state.h"
+#include "r200_ioctl.h"
+#include "r200_swtcl.h"
+#include "r200_tex.h"
+
+
+/* =============================================================
+ * Utility functions:
+ */
+
+static void r200SetTexWrap( r200TexObjPtr t, GLenum swrap, GLenum twrap, GLenum rwrap )
+{
+   t->pp_txfilter &= ~(R200_CLAMP_S_MASK | R200_CLAMP_T_MASK);
+
+   switch ( swrap ) {
+   case GL_REPEAT:
+      t->pp_txfilter |= R200_CLAMP_S_WRAP;
+      break;
+   case GL_CLAMP:
+      t->pp_txfilter |= R200_CLAMP_S_CLAMP_LAST;
+      break;
+   case GL_CLAMP_TO_EDGE:
+      t->pp_txfilter |= R200_CLAMP_S_CLAMP_LAST;
+      break;
+   case GL_CLAMP_TO_BORDER:
+      t->pp_txfilter |= R200_CLAMP_S_CLAMP_BORDER;
+      break;
+   case GL_MIRRORED_REPEAT_ARB:
+      t->pp_txfilter |= R200_CLAMP_S_MIRROR;
+      break;
+   case GL_MIRROR_CLAMP_ATI:
+      t->pp_txfilter |= R200_CLAMP_S_MIRROR_CLAMP_BORDER;
+      break;
+   case GL_MIRROR_CLAMP_TO_EDGE_ATI:
+      t->pp_txfilter |= R200_CLAMP_S_MIRROR_CLAMP_LAST;
+      break;
+   default:
+      _mesa_problem(NULL, "bad S wrap mode in r200SetTexWrap");
+   }
+
+   switch ( twrap ) {
+   case GL_REPEAT:
+      t->pp_txfilter |= R200_CLAMP_T_WRAP;
+      break;
+   case GL_CLAMP:
+      t->pp_txfilter |= R200_CLAMP_T_CLAMP_LAST;
+      break;
+   case GL_CLAMP_TO_EDGE:
+      t->pp_txfilter |= R200_CLAMP_T_CLAMP_LAST;
+      break;
+   case GL_CLAMP_TO_BORDER:
+      t->pp_txfilter |= R200_CLAMP_T_CLAMP_BORDER;
+      break;
+   case GL_MIRRORED_REPEAT_ARB:
+      t->pp_txfilter |= R200_CLAMP_T_MIRROR;
+      break;
+   case GL_MIRROR_CLAMP_ATI:
+      t->pp_txfilter |= R200_CLAMP_T_MIRROR_CLAMP_BORDER;
+      break;
+   case GL_MIRROR_CLAMP_TO_EDGE_ATI:
+      t->pp_txfilter |= R200_CLAMP_T_MIRROR_CLAMP_LAST;
+      break;
+   default:
+      _mesa_problem(NULL, "bad S wrap mode in r200SetTexWrap");
+   }
+
+   t->pp_txformat_x &= ~R200_CLAMP_Q_MASK;
+
+   switch ( rwrap ) {
+   case GL_REPEAT:
+      t->pp_txformat_x |= R200_CLAMP_Q_WRAP;
+      break;
+   case GL_CLAMP:
+      t->pp_txformat_x |= R200_CLAMP_Q_CLAMP_LAST;
+      break;
+   case GL_CLAMP_TO_EDGE:
+      t->pp_txformat_x |= R200_CLAMP_Q_CLAMP_LAST;
+      break;
+   case GL_CLAMP_TO_BORDER:
+      t->pp_txformat_x |= R200_CLAMP_Q_CLAMP_BORDER;
+      break;
+   case GL_MIRRORED_REPEAT_ARB:
+      t->pp_txformat_x |= R200_CLAMP_Q_MIRROR;
+      break;
+   case GL_MIRROR_CLAMP_ATI:
+      t->pp_txformat_x |= R200_CLAMP_Q_MIRROR_CLAMP_BORDER;
+      break;
+   case GL_MIRROR_CLAMP_TO_EDGE_ATI:
+      t->pp_txformat_x |= R200_CLAMP_Q_MIRROR_CLAMP_LAST;
+      break;
+   default:
+      _mesa_problem(NULL, "bad R wrap mode in r200SetTexWrap");
+   }
+}
+
+static void r200SetTexMaxAnisotropy( r200TexObjPtr t, GLfloat max )
+{
+   t->pp_txfilter &= ~R200_MAX_ANISO_MASK;
+
+   if ( max == 1.0 ) {
+      t->pp_txfilter |= R200_MAX_ANISO_1_TO_1;
+   } else if ( max <= 2.0 ) {
+      t->pp_txfilter |= R200_MAX_ANISO_2_TO_1;
+   } else if ( max <= 4.0 ) {
+      t->pp_txfilter |= R200_MAX_ANISO_4_TO_1;
+   } else if ( max <= 8.0 ) {
+      t->pp_txfilter |= R200_MAX_ANISO_8_TO_1;
+   } else {
+      t->pp_txfilter |= R200_MAX_ANISO_16_TO_1;
+   }
+}
+
+static void r200SetTexFilter( r200TexObjPtr t, GLenum minf, GLenum magf )
+{
+   GLuint anisotropy = (t->pp_txfilter & R200_MAX_ANISO_MASK);
+
+   t->pp_txfilter &= ~(R200_MIN_FILTER_MASK | R200_MAG_FILTER_MASK);
+   t->pp_txformat_x &= ~R200_VOLUME_FILTER_MASK;
+
+   if ( anisotropy == R200_MAX_ANISO_1_TO_1 ) {
+      switch ( minf ) {
+      case GL_NEAREST:
+	 t->pp_txfilter |= R200_MIN_FILTER_NEAREST;
+	 break;
+      case GL_LINEAR:
+	 t->pp_txfilter |= R200_MIN_FILTER_LINEAR;
+	 break;
+      case GL_NEAREST_MIPMAP_NEAREST:
+	 t->pp_txfilter |= R200_MIN_FILTER_NEAREST_MIP_NEAREST;
+	 break;
+      case GL_NEAREST_MIPMAP_LINEAR:
+	 t->pp_txfilter |= R200_MIN_FILTER_LINEAR_MIP_NEAREST;
+	 break;
+      case GL_LINEAR_MIPMAP_NEAREST:
+	 t->pp_txfilter |= R200_MIN_FILTER_NEAREST_MIP_LINEAR;
+	 break;
+      case GL_LINEAR_MIPMAP_LINEAR:
+	 t->pp_txfilter |= R200_MIN_FILTER_LINEAR_MIP_LINEAR;
+	 break;
+      }
+   } else {
+      switch ( minf ) {
+      case GL_NEAREST:
+	 t->pp_txfilter |= R200_MIN_FILTER_ANISO_NEAREST;
+	 break;
+      case GL_LINEAR:
+	 t->pp_txfilter |= R200_MIN_FILTER_ANISO_LINEAR;
+	 break;
+      case GL_NEAREST_MIPMAP_NEAREST:
+      case GL_LINEAR_MIPMAP_NEAREST:
+	 t->pp_txfilter |= R200_MIN_FILTER_ANISO_NEAREST_MIP_NEAREST;
+	 break;
+      case GL_NEAREST_MIPMAP_LINEAR:
+      case GL_LINEAR_MIPMAP_LINEAR:
+	 t->pp_txfilter |= R200_MIN_FILTER_ANISO_NEAREST_MIP_LINEAR;
+	 break;
+      }
+   }
+
+   /* Note we don't have 3D mipmaps so only use the mag filter setting
+    * to set the 3D texture filter mode.
+    */
+   switch ( magf ) {
+   case GL_NEAREST:
+      t->pp_txfilter |= R200_MAG_FILTER_NEAREST;
+      t->pp_txformat_x |= R200_VOLUME_FILTER_NEAREST;
+      break;
+   case GL_LINEAR:
+      t->pp_txfilter |= R200_MAG_FILTER_LINEAR;
+      t->pp_txformat_x |= R200_VOLUME_FILTER_LINEAR;
+      break;
+   }
+}
+
+static void r200SetTexBorderColor( r200TexObjPtr t, GLubyte c[4] )
+{
+   t->pp_border_color = r200PackColor( 4, c[0], c[1], c[2], c[3] );
+}
+
+
+static r200TexObjPtr r200AllocTexObj( struct gl_texture_object *texObj )
+{
+   r200TexObjPtr t;
+
+   t = CALLOC_STRUCT( r200_tex_obj );
+   if (!t)
+      return NULL;
+
+   if ( R200_DEBUG & DEBUG_TEXTURE ) {
+      fprintf( stderr, "%s( %p, %p )\n", __FUNCTION__, texObj, t );
+   }
+
+   t->tObj = texObj;
+   make_empty_list( t );
+
+   /* Initialize non-image-dependent parts of the state:
+    */
+   r200SetTexWrap( t, texObj->WrapS, texObj->WrapT, texObj->WrapR );
+   r200SetTexMaxAnisotropy( t, texObj->MaxAnisotropy );
+   r200SetTexFilter( t, texObj->MinFilter, texObj->MagFilter );
+   r200SetTexBorderColor( t, texObj->_BorderChan );
+   return t;
+}
+
+
+static const struct gl_texture_format *
+r200ChooseTextureFormat( GLcontext *ctx, GLint internalFormat,
+                           GLenum format, GLenum type )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   const GLboolean do32bpt = ( rmesa->r200Screen->cpp == 4 );
+
+   switch ( internalFormat ) {
+   case 4:
+   case GL_RGBA:
+      if ( format == GL_BGRA ) {
+	 if ( type == GL_UNSIGNED_INT_8_8_8_8_REV ) {
+	    return &_mesa_texformat_argb8888;
+	 }
+         else if ( type == GL_UNSIGNED_SHORT_4_4_4_4_REV ) {
+            return &_mesa_texformat_argb4444;
+	 }
+         else if ( type == GL_UNSIGNED_SHORT_1_5_5_5_REV ) {
+	    return &_mesa_texformat_argb1555;
+	 }
+      }
+      return do32bpt ? &_mesa_texformat_rgba8888 : &_mesa_texformat_argb4444;
+
+   case 3:
+   case GL_RGB:
+      if ( format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5 ) {
+	 return &_mesa_texformat_rgb565;
+      }
+      return do32bpt ? &_mesa_texformat_rgba8888 : &_mesa_texformat_rgb565;
+
+   case GL_RGBA8:
+   case GL_RGB10_A2:
+   case GL_RGBA12:
+   case GL_RGBA16:
+      return do32bpt ? &_mesa_texformat_rgba8888 : &_mesa_texformat_argb4444;
+
+   case GL_RGBA4:
+   case GL_RGBA2:
+      return &_mesa_texformat_argb4444;
+
+   case GL_RGB5_A1:
+      return &_mesa_texformat_argb1555;
+
+   case GL_RGB8:
+   case GL_RGB10:
+   case GL_RGB12:
+   case GL_RGB16:
+      return do32bpt ? &_mesa_texformat_rgba8888 : &_mesa_texformat_rgb565;
+
+   case GL_RGB5:
+   case GL_RGB4:
+   case GL_R3_G3_B2:
+      return &_mesa_texformat_rgb565;
+
+   case GL_ALPHA:
+   case GL_ALPHA4:
+   case GL_ALPHA8:
+   case GL_ALPHA12:
+   case GL_ALPHA16:
+      return &_mesa_texformat_al88;
+
+   case 1:
+   case GL_LUMINANCE:
+   case GL_LUMINANCE4:
+   case GL_LUMINANCE8:
+   case GL_LUMINANCE12:
+   case GL_LUMINANCE16:
+      return &_mesa_texformat_al88;
+
+   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:
+      return &_mesa_texformat_al88;
+
+   case GL_INTENSITY:
+   case GL_INTENSITY4:
+   case GL_INTENSITY8:
+   case GL_INTENSITY12:
+   case GL_INTENSITY16:
+      /* At the moment, glean & conform both fail using the i8 internal
+       * format.
+       */
+      return &_mesa_texformat_al88;
+/*       return &_mesa_texformat_i8; */
+
+   case GL_YCBCR_MESA:
+      if (type == GL_UNSIGNED_SHORT_8_8_APPLE ||
+	  type == GL_UNSIGNED_BYTE)
+         return &_mesa_texformat_ycbcr;
+      else
+         return &_mesa_texformat_ycbcr_rev;
+
+   default:
+      _mesa_problem(ctx, "unexpected texture format in r200ChoosTexFormat");
+      return NULL;
+   }
+
+   return NULL; /* never get here */
+}
+
+
+static GLboolean
+r200ValidateClientStorage( GLcontext *ctx, GLenum target,
+			   GLint internalFormat,
+			   GLint srcWidth, GLint srcHeight, 
+                           GLenum format, GLenum type,  const void *pixels,
+			   const struct gl_pixelstore_attrib *packing,
+			   struct gl_texture_object *texObj,
+			   struct gl_texture_image *texImage)
+
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   int texelBytes;
+
+   if (0)
+      fprintf(stderr, "intformat %s format %s type %s\n",
+	      _mesa_lookup_enum_by_nr( internalFormat ),
+	      _mesa_lookup_enum_by_nr( format ),
+	      _mesa_lookup_enum_by_nr( type ));
+
+   if (!ctx->Unpack.ClientStorage)
+      return 0;
+
+   if (ctx->_ImageTransferState ||
+       texImage->IsCompressed ||
+       texObj->GenerateMipmap)
+      return 0;
+
+
+   /* This list is incomplete, may be different on ppc???
+    */
+   switch ( internalFormat ) {
+   case GL_RGBA:
+      if ( format == GL_BGRA && type == GL_UNSIGNED_INT_8_8_8_8_REV ) {
+	 texImage->TexFormat = &_mesa_texformat_argb8888;
+	 texelBytes = 4;
+      }
+      else
+	 return 0;
+      break;
+
+   case GL_YCBCR_MESA:
+      if ( format == GL_YCBCR_MESA && 
+	   type == GL_UNSIGNED_SHORT_8_8_REV_APPLE ) {
+	 texImage->TexFormat = &_mesa_texformat_ycbcr_rev;
+	 texelBytes = 2;
+      }
+      else if ( format == GL_YCBCR_MESA && 
+		(type == GL_UNSIGNED_SHORT_8_8_APPLE || 
+		 type == GL_UNSIGNED_BYTE)) {
+	 texImage->TexFormat = &_mesa_texformat_ycbcr;
+	 texelBytes = 2;
+      }
+      else
+	 return 0;
+      break;
+      
+	 
+   default:
+      return 0;
+   }
+
+   /* Could deal with these packing issues, but currently don't:
+    */
+   if (packing->SkipPixels || 
+       packing->SkipRows || 
+       packing->SwapBytes ||
+       packing->LsbFirst) {
+      return 0;
+   }
+
+   {      
+      GLint srcRowStride = _mesa_image_row_stride(packing, srcWidth,
+						  format, type);
+
+      
+      if (0)
+	 fprintf(stderr, "%s: srcRowStride %d/%x\n", 
+		 __FUNCTION__, srcRowStride, srcRowStride);
+
+      /* Could check this later in upload, pitch restrictions could be
+       * relaxed, but would need to store the image pitch somewhere,
+       * as packing details might change before image is uploaded:
+       */
+      if (!r200IsAgpMemory( rmesa, pixels, srcHeight * srcRowStride ) ||
+	  (srcRowStride & 63))
+	 return 0;
+
+
+      /* Have validated that _mesa_transfer_teximage would be a straight
+       * memcpy at this point.  NOTE: future calls to TexSubImage will
+       * overwrite the client data.  This is explicitly mentioned in the
+       * extension spec.
+       */
+      texImage->Data = (void *)pixels;
+      texImage->IsClientData = GL_TRUE;
+      texImage->RowStride = srcRowStride / texelBytes;
+      return 1;
+   }
+}
+
+
+static void r200TexImage1D( GLcontext *ctx, GLenum target, GLint level,
+                              GLint internalFormat,
+                              GLint width, 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 )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   r200TexObjPtr t = (r200TexObjPtr) texObj->DriverData;
+
+   if ( t ) {
+      r200SwapOutTexObj( rmesa, t );
+   }
+   else {
+      t = r200AllocTexObj( texObj );
+      if (!t) {
+         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage1D");
+         return;
+      }
+      texObj->DriverData = t;
+   }
+
+   /* Note, this will call r200ChooseTextureFormat */
+   _mesa_store_teximage1d(ctx, target, level, internalFormat,
+                          width, border, format, type, pixels,
+                          &ctx->Unpack, texObj, texImage);
+
+   t->dirty_images[0] |= (1 << level);
+}
+
+
+static void r200TexSubImage1D( GLcontext *ctx, GLenum target, GLint level,
+                                 GLint xoffset,
+                                 GLsizei width,
+                                 GLenum format, GLenum type,
+                                 const GLvoid *pixels,
+                                 const struct gl_pixelstore_attrib *packing,
+                                 struct gl_texture_object *texObj,
+                                 struct gl_texture_image *texImage )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   r200TexObjPtr t = (r200TexObjPtr)texObj->DriverData;
+
+   assert( t ); /* this _should_ be true */
+   if ( t ) {
+      r200SwapOutTexObj( rmesa, t );
+      t->dirty_images[0] |= (1 << level);
+   }
+   else {
+      t = r200AllocTexObj(texObj);
+      if (!t) {
+         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage1D");
+         return;
+      }
+      texObj->DriverData = t;
+   }
+
+   _mesa_store_texsubimage1d(ctx, target, level, xoffset, width,
+			     format, type, pixels, packing, texObj,
+			     texImage);
+
+   t->dirty_images[0] |= (1 << level);
+}
+
+
+static void r200TexImage2D( 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 )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   r200TexObjPtr t = (r200TexObjPtr)texObj->DriverData;
+   GLuint face;
+
+   /* which cube face or ordinary 2D image */
+   switch (target) {
+   case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
+   case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
+   case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
+   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
+   case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
+   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
+      face = (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
+      ASSERT(face < 6);
+      break;
+   default:
+      face = 0;
+   }
+
+   if ( t ) {
+      r200SwapOutTexObj( rmesa, t );
+   }
+   else {
+      t = r200AllocTexObj( texObj );
+      if (!t) {
+         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage2D");
+         return;
+      }
+      texObj->DriverData = t;
+   }
+
+   texImage->IsClientData = GL_FALSE;
+
+   if (r200ValidateClientStorage( ctx, target, 
+				  internalFormat, 
+				  width, height, 
+				  format, type, pixels, 
+				  packing, texObj, texImage)) {
+      if (R200_DEBUG & DEBUG_TEXTURE)
+	 fprintf(stderr, "%s: Using client storage\n", __FUNCTION__); 
+   }
+   else {
+      if (R200_DEBUG & DEBUG_TEXTURE)
+	 fprintf(stderr, "%s: Using normal storage\n", __FUNCTION__); 
+
+      /* Normal path: copy (to cached memory) and eventually upload
+       * via another copy to agp memory and then a blit...  Could
+       * eliminate one copy by going straight to (permanent) agp.
+       *
+       * Note, this will call r200ChooseTextureFormat.
+       */
+      _mesa_store_teximage2d(ctx, target, level, internalFormat,
+			     width, height, border, format, type, pixels,
+			     &ctx->Unpack, texObj, texImage);
+      
+      t->dirty_images[face] |= (1 << level);
+   }
+}
+
+
+static void r200TexSubImage2D( 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 )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   r200TexObjPtr t = (r200TexObjPtr) texObj->DriverData;
+   GLuint face;
+
+/*     fprintf(stderr, "%s\n", __FUNCTION__); */
+
+   /* which cube face or ordinary 2D image */
+   switch (target) {
+   case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
+   case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
+   case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
+   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
+   case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
+   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
+      face = (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X;
+      ASSERT(face < 6);
+      break;
+   default:
+      face = 0;
+   }
+
+   assert( t ); /* this _should_ be true */
+   if ( t ) {
+      r200SwapOutTexObj( rmesa, t );
+   }
+   else {
+      t = r200AllocTexObj(texObj);
+      if (!t) {
+         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage2D");
+         return;
+      }
+      texObj->DriverData = t;
+   }
+
+   _mesa_store_texsubimage2d(ctx, target, level, xoffset, yoffset, width,
+			     height, format, type, pixels, packing, texObj,
+			     texImage);
+
+   t->dirty_images[face] |= (1 << level);
+}
+
+
+static void r200TexImage3D( GLcontext *ctx, GLenum target, GLint level,
+                            GLint internalFormat,
+                            GLint width, GLint height, GLint depth,
+                            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 )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   r200TexObjPtr t = (r200TexObjPtr)texObj->DriverData;
+
+   if ( t ) {
+      r200SwapOutTexObj( rmesa, t );
+   }
+   else {
+      t = r200AllocTexObj( texObj );
+      if (!t) {
+         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage3D");
+         return;
+      }
+      texObj->DriverData = t;
+   }
+
+   texImage->IsClientData = GL_FALSE;
+
+#if 0
+   if (r200ValidateClientStorage( ctx, target, 
+				  internalFormat, 
+				  width, height, 
+				  format, type, pixels, 
+				  packing, texObj, texImage)) {
+      if (R200_DEBUG & DEBUG_TEXTURE)
+	 fprintf(stderr, "%s: Using client storage\n", __FUNCTION__); 
+   }
+   else
+#endif
+   {
+      if (R200_DEBUG & DEBUG_TEXTURE)
+	 fprintf(stderr, "%s: Using normal storage\n", __FUNCTION__); 
+
+      /* Normal path: copy (to cached memory) and eventually upload
+       * via another copy to agp memory and then a blit...  Could
+       * eliminate one copy by going straight to (permanent) agp.
+       *
+       * Note, this will call r200ChooseTextureFormat.
+       */
+      _mesa_store_teximage3d(ctx, target, level, internalFormat,
+			     width, height, depth, border,
+                             format, type, pixels,
+			     &ctx->Unpack, texObj, texImage);
+      
+      t->dirty_images[0] |= (1 << level);
+   }
+}
+
+
+static void
+r200TexSubImage3D( GLcontext *ctx, GLenum target, GLint level,
+                   GLint xoffset, GLint yoffset, GLint zoffset,
+                   GLsizei width, GLsizei height, GLsizei depth,
+                   GLenum format, GLenum type,
+                   const GLvoid *pixels,
+                   const struct gl_pixelstore_attrib *packing,
+                   struct gl_texture_object *texObj,
+                   struct gl_texture_image *texImage )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   r200TexObjPtr t = (r200TexObjPtr) texObj->DriverData;
+
+/*     fprintf(stderr, "%s\n", __FUNCTION__); */
+
+   assert( t ); /* this _should_ be true */
+   if ( t ) {
+      r200SwapOutTexObj( rmesa, t );
+   }
+   else {
+      t = r200AllocTexObj(texObj);
+      if (!t) {
+         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage3D");
+         return;
+      }
+      texObj->DriverData = t;
+   }
+
+   _mesa_store_texsubimage3d(ctx, target, level, xoffset, yoffset, zoffset,
+                             width, height, depth,
+                             format, type, pixels, packing, texObj, texImage);
+
+   t->dirty_images[0] |= (1 << level);
+}
+
+
+
+static void r200TexEnv( GLcontext *ctx, GLenum target,
+			  GLenum pname, const GLfloat *param )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   GLuint unit = ctx->Texture.CurrentUnit;
+   struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
+
+   if ( R200_DEBUG & DEBUG_STATE ) {
+      fprintf( stderr, "%s( %s )\n",
+	       __FUNCTION__, _mesa_lookup_enum_by_nr( pname ) );
+   }
+
+   /* This is incorrect: Need to maintain this data for each of
+    * GL_TEXTURE_{123}D, GL_TEXTURE_RECTANGLE_NV, etc, and switch
+    * between them according to _ReallyEnabled.
+    */
+   switch ( pname ) {
+   case GL_TEXTURE_ENV_COLOR: {
+      GLubyte c[4];
+      GLuint envColor;
+      UNCLAMPED_FLOAT_TO_RGBA_CHAN( c, texUnit->EnvColor );
+      envColor = r200PackColor( 4, c[0], c[1], c[2], c[3] );
+      if ( rmesa->hw.tf.cmd[TF_TFACTOR_0 + unit] != envColor ) {
+	 R200_STATECHANGE( rmesa, tf );
+	 rmesa->hw.tf.cmd[TF_TFACTOR_0 + unit] = envColor;
+      }
+      break;
+   }
+
+   case GL_TEXTURE_LOD_BIAS_EXT: {
+      GLfloat bias;
+      GLuint b;
+      const int fixed_one = 0x8000000;
+
+      /* The R200's LOD bias is a signed 2's complement value with a
+       * range of -16.0 <= bias < 16.0. 
+       *
+       * NOTE: Add a small bias to the bias for conform mipsel.c test.
+       */
+      bias = *param + .01;
+      bias = CLAMP( bias, -16.0, 16.0 );
+      b = (int)(bias * fixed_one) & R200_LOD_BIAS_MASK;
+      
+      if ( (rmesa->hw.tex[unit].cmd[TEX_PP_TXFORMAT_X] & R200_LOD_BIAS_MASK) != b ) {
+	 R200_STATECHANGE( rmesa, tex[unit] );
+	 rmesa->hw.tex[unit].cmd[TEX_PP_TXFORMAT_X] &= ~R200_LOD_BIAS_MASK;
+	 rmesa->hw.tex[unit].cmd[TEX_PP_TXFORMAT_X] |= b;
+      }
+      break;
+   }
+
+   default:
+      return;
+   }
+}
+
+static void r200TexParameter( GLcontext *ctx, GLenum target,
+				struct gl_texture_object *texObj,
+				GLenum pname, const GLfloat *params )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   r200TexObjPtr t = (r200TexObjPtr) texObj->DriverData;
+
+   if ( R200_DEBUG & (DEBUG_STATE|DEBUG_TEXTURE) ) {
+      fprintf( stderr, "%s( %s )\n", __FUNCTION__,
+	       _mesa_lookup_enum_by_nr( pname ) );
+   }
+
+   if (!t)
+      return;
+
+   switch ( pname ) {
+   case GL_TEXTURE_MIN_FILTER:
+   case GL_TEXTURE_MAG_FILTER:
+   case GL_TEXTURE_MAX_ANISOTROPY_EXT:
+      r200SetTexMaxAnisotropy( t, texObj->MaxAnisotropy );
+      r200SetTexFilter( t, texObj->MinFilter, texObj->MagFilter );
+      break;
+
+   case GL_TEXTURE_WRAP_S:
+   case GL_TEXTURE_WRAP_T:
+   case GL_TEXTURE_WRAP_R:
+      r200SetTexWrap( t, texObj->WrapS, texObj->WrapT, texObj->WrapR );
+      break;
+
+   case GL_TEXTURE_BORDER_COLOR:
+      r200SetTexBorderColor( t, texObj->_BorderChan );
+      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 R200.  Since there's no LOD clamping,
+       * we just have to rely on loading the right subset of mipmap levels
+       * to simulate a clamped LOD.
+       */
+      r200SwapOutTexObj( rmesa, t );
+      break;
+
+   default:
+      return;
+   }
+
+   /* Mark this texobj as dirty (one bit per tex unit)
+    */
+   t->dirty_state = TEX_ALL;
+}
+
+
+
+static void r200BindTexture( GLcontext *ctx, GLenum target,
+			       struct gl_texture_object *texObj )
+{
+   r200TexObjPtr t = (r200TexObjPtr) texObj->DriverData;
+   GLuint unit = ctx->Texture.CurrentUnit;
+
+   if ( R200_DEBUG & (DEBUG_STATE|DEBUG_TEXTURE) ) {
+      fprintf( stderr, "%s( %p ) unit=%d\n", __FUNCTION__, texObj, unit );
+   }
+
+   if ( target == GL_TEXTURE_2D || target == GL_TEXTURE_1D ) {
+      if ( !t ) {
+	 t = r200AllocTexObj( texObj );
+	 texObj->DriverData = t;
+      }
+   }
+}
+
+static void r200DeleteTexture( GLcontext *ctx,
+				 struct gl_texture_object *texObj )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   r200TexObjPtr t = (r200TexObjPtr) texObj->DriverData;
+
+   if ( R200_DEBUG & (DEBUG_STATE|DEBUG_TEXTURE) ) {
+      fprintf( stderr, "%s( %p )\n", __FUNCTION__, texObj );
+   }
+
+   if ( t ) {
+      if ( rmesa ) {
+         R200_FIREVERTICES( rmesa );
+      }
+      r200DestroyTexObj( rmesa, t );
+      texObj->DriverData = NULL;
+   }
+}
+
+static GLboolean r200IsTextureResident( GLcontext *ctx,
+					  struct gl_texture_object *texObj )
+{
+   r200TexObjPtr t = (r200TexObjPtr) texObj->DriverData;
+
+   return ( t && t->memBlock );
+}
+
+
+static void r200InitTextureObjects( GLcontext *ctx )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   struct gl_texture_object *texObj;
+   GLuint tmp = ctx->Texture.CurrentUnit;
+
+   ctx->Texture.CurrentUnit = 0;
+
+   texObj = ctx->Texture.Unit[0].Current1D;
+   r200BindTexture( ctx, GL_TEXTURE_1D, texObj );
+   move_to_tail( &rmesa->texture.swapped,
+		 (r200TexObjPtr)texObj->DriverData );
+
+   texObj = ctx->Texture.Unit[0].Current2D;
+   r200BindTexture( ctx, GL_TEXTURE_2D, texObj );
+   move_to_tail( &rmesa->texture.swapped,
+		 (r200TexObjPtr)texObj->DriverData );
+
+   ctx->Texture.CurrentUnit = 1;
+
+   texObj = ctx->Texture.Unit[1].Current1D;
+   r200BindTexture( ctx, GL_TEXTURE_1D, texObj );
+   move_to_tail( &rmesa->texture.swapped,
+		 (r200TexObjPtr)texObj->DriverData );
+
+   texObj = ctx->Texture.Unit[1].Current2D;
+   r200BindTexture( ctx, GL_TEXTURE_2D, texObj );
+   move_to_tail( &rmesa->texture.swapped,
+		 (r200TexObjPtr)texObj->DriverData );
+
+   ctx->Texture.CurrentUnit = tmp;
+}
+
+/* Need:  
+ *  - Same GEN_MODE for all active bits
+ *  - Same EyePlane/ObjPlane for all active bits when using Eye/Obj
+ *  - STRQ presumably all supported (matrix means incoming R values
+ *    can end up in STQ, this has implications for vertex support,
+ *    presumably ok if maos is used, though?)
+ *  
+ * Basically impossible to do this on the fly - just collect some
+ * basic info & do the checks from ValidateState().
+ */
+static void r200TexGen( GLcontext *ctx,
+			  GLenum coord,
+			  GLenum pname,
+			  const GLfloat *params )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   GLuint unit = ctx->Texture.CurrentUnit;
+   rmesa->recheck_texgen[unit] = GL_TRUE;
+}
+
+
+void r200InitTextureFuncs( GLcontext *ctx )
+{
+   ctx->Driver.ChooseTextureFormat	= r200ChooseTextureFormat;
+   ctx->Driver.TexImage1D		= r200TexImage1D;
+   ctx->Driver.TexImage2D		= r200TexImage2D;
+#if ENABLE_HW_3D_TEXTURE
+   ctx->Driver.TexImage3D		= r200TexImage3D;
+#else
+   ctx->Driver.TexImage3D		= _mesa_store_teximage3d;
+#endif
+   ctx->Driver.TexSubImage1D		= r200TexSubImage1D;
+   ctx->Driver.TexSubImage2D		= r200TexSubImage2D;
+#if ENABLE_HW_3D_TEXTURE
+   ctx->Driver.TexSubImage3D		= r200TexSubImage3D;
+#else
+   ctx->Driver.TexSubImage3D		= _mesa_store_texsubimage3d;
+#endif
+   ctx->Driver.CopyTexImage1D		= _swrast_copy_teximage1d;
+   ctx->Driver.CopyTexImage2D		= _swrast_copy_teximage2d;
+   ctx->Driver.CopyTexSubImage1D	= _swrast_copy_texsubimage1d;
+   ctx->Driver.CopyTexSubImage2D	= _swrast_copy_texsubimage2d;
+   ctx->Driver.CopyTexSubImage3D 	= _swrast_copy_texsubimage3d;
+   ctx->Driver.TestProxyTexImage	= _mesa_test_proxy_teximage;
+
+   ctx->Driver.BindTexture		= r200BindTexture;
+   ctx->Driver.CreateTexture		= NULL; /* FIXME: Is this used??? */
+   ctx->Driver.DeleteTexture		= r200DeleteTexture;
+   ctx->Driver.IsTextureResident	= r200IsTextureResident;
+   ctx->Driver.PrioritizeTexture	= NULL;
+   ctx->Driver.ActiveTexture		= NULL;
+   ctx->Driver.UpdateTexturePalette	= NULL;
+
+   ctx->Driver.TexEnv			= r200TexEnv;
+   ctx->Driver.TexParameter		= r200TexParameter;
+   ctx->Driver.TexGen                   = r200TexGen;
+
+   r200InitTextureObjects( ctx );
+}

src/mesa/drivers/dri/r200/r200_tex.h

@@ -0,0 +1,55 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#ifndef __R200_TEX_H__
+#define __R200_TEX_H__
+
+#ifdef GLX_DIRECT_RENDERING
+
+extern void r200UpdateTextureState( GLcontext *ctx );
+
+extern int r200UploadTexImages( r200ContextPtr rmesa, r200TexObjPtr t, GLuint face );
+
+extern void r200AgeTextures( r200ContextPtr rmesa, int heap );
+extern void r200DestroyTexObj( r200ContextPtr rmesa, r200TexObjPtr t );
+extern void r200SwapOutTexObj( r200ContextPtr rmesa, r200TexObjPtr t );
+
+extern void r200PrintLocalLRU( r200ContextPtr rmesa, int heap );
+extern void r200PrintGlobalLRU( r200ContextPtr rmesa, int heap );
+extern void r200UpdateTexLRU( r200ContextPtr rmesa, r200TexObjPtr t );
+
+extern void r200InitTextureFuncs( GLcontext *ctx );
+
+#endif
+#endif /* __R200_TEX_H__ */

src/mesa/drivers/dri/r200/r200_texmem.c

@@ -0,0 +1,811 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright (C) Tungsten Graphics 2002.  All Rights Reserved.  
+The Weather Channel, Inc. 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 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 ATI, VA LINUX SYSTEMS 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <martin@valinux.com>
+ *   Gareth Hughes <gareth@valinux.com>
+ *
+ */
+
+#include "glheader.h"
+#include "imports.h"
+#include "context.h"
+#include "colormac.h"
+#include "mmath.h"
+#include "macros.h"
+#include "simple_list.h"
+#include "radeon_reg.h" /* gets definition for usleep */
+#include "r200_context.h"
+#include "r200_state.h"
+#include "r200_ioctl.h"
+#include "r200_swtcl.h"
+#include "r200_tex.h"
+
+#include <unistd.h>  /* for usleep() */
+#include <errno.h>
+
+/* Destroy hardware state associated with texture `t'.
+ */
+void r200DestroyTexObj( r200ContextPtr rmesa, r200TexObjPtr t )
+{
+   if ( !t )
+      return;
+
+   if ( R200_DEBUG & DEBUG_TEXTURE ) {
+      fprintf( stderr, "%s( %p, %p )\n", __FUNCTION__, t, t->tObj );
+   }
+
+   if ( t->memBlock ) {
+      mmFreeMem( t->memBlock );
+      t->memBlock = NULL;
+   }
+
+   if ( t->tObj )
+      t->tObj->DriverData = NULL;
+
+   if ( rmesa ) {
+      if ( t == rmesa->state.texture.unit[0].texobj ) {
+         rmesa->state.texture.unit[0].texobj = NULL;
+	 remove_from_list( &rmesa->hw.tex[0] );
+	 make_empty_list( &rmesa->hw.tex[0] );
+	 remove_from_list( &rmesa->hw.cube[0] );
+	 make_empty_list( &rmesa->hw.cube[0] );
+      }
+
+      if ( t == rmesa->state.texture.unit[1].texobj ) {
+         rmesa->state.texture.unit[1].texobj = NULL;
+	 remove_from_list( &rmesa->hw.tex[1] );
+	 make_empty_list( &rmesa->hw.tex[1] );
+	 remove_from_list( &rmesa->hw.cube[1] );
+	 make_empty_list( &rmesa->hw.cube[1] );
+      }
+   }
+
+   remove_from_list( t );
+   FREE( t );
+}
+
+
+/* Keep track of swapped out texture objects.
+ */
+void r200SwapOutTexObj( r200ContextPtr rmesa, r200TexObjPtr t )
+{
+   GLuint face;
+
+   if ( R200_DEBUG & DEBUG_TEXTURE ) {
+      fprintf( stderr, "%s( %p, %p )\n", __FUNCTION__, t, t->tObj );
+   }
+
+   if (t->memBlock) {
+      mmFreeMem( t->memBlock );
+      t->memBlock = NULL;
+   }
+
+   if (!t->tObj) {
+      remove_from_list( t );
+      FREE( t );
+   }
+   else {
+      for (face = 0; face < 6; face++)
+	 t->dirty_images[face] = ~0;
+      move_to_tail( &rmesa->texture.swapped, t );
+   }
+}
+
+/* Print out debugging information about texture LRU.
+ */
+void r200PrintLocalLRU( r200ContextPtr rmesa, int heap )
+{
+   r200TexObjPtr t;
+   int sz = 1 << (rmesa->r200Screen->logTexGranularity[heap]);
+
+   fprintf( stderr, "\nLocal LRU, heap %d:\n", heap );
+
+   foreach ( t, &rmesa->texture.objects[heap] ) {
+      if (!t->memBlock)
+	 continue;
+      if (!t->tObj) {
+	 fprintf( stderr, "Placeholder %d at 0x%x sz 0x%x\n",
+		  t->memBlock->ofs / sz,
+		  t->memBlock->ofs,
+		  t->memBlock->size );
+      } else {
+	 fprintf( stderr, "Texture at 0x%x sz 0x%x\n",
+		  t->memBlock->ofs,
+		  t->memBlock->size );
+      }
+   }
+
+   fprintf( stderr, "\n" );
+}
+
+void r200PrintGlobalLRU( r200ContextPtr rmesa, int heap )
+{
+   radeon_tex_region_t *list = rmesa->sarea->texList[heap];
+   int i, j;
+
+   fprintf( stderr, "\nGlobal LRU, heap %d list %p:\n", heap, list );
+
+   for ( i = 0, j = RADEON_NR_TEX_REGIONS ; i < RADEON_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 == RADEON_NR_TEX_REGIONS ) break;
+   }
+
+   if ( j != RADEON_NR_TEX_REGIONS ) {
+      fprintf( stderr, "Loop detected in global LRU\n" );
+      for ( i = 0 ; i < RADEON_NR_TEX_REGIONS ; i++ ) {
+	 fprintf( stderr, "list[%d] age %d next %d prev %d\n",
+		  i, list[i].age, list[i].next, list[i].prev );
+      }
+   }
+
+   fprintf( stderr, "\n" );
+}
+
+/* Reset the global texture LRU.
+ */
+static void r200ResetGlobalLRU( r200ContextPtr rmesa, int heap )
+{
+   radeon_tex_region_t *list = rmesa->sarea->texList[heap];
+   int sz = 1 << rmesa->r200Screen->logTexGranularity[heap];
+   int i;
+
+   /*
+    * (Re)initialize the global circular LRU list.  The last element in
+    * the array (RADEON_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 <= rmesa->r200Screen->texSize[heap] ; i++ ) {
+      list[i].prev = i-1;
+      list[i].next = i+1;
+      list[i].age = 0;
+   }
+
+   i--;
+   list[0].prev = RADEON_NR_TEX_REGIONS;
+   list[i].prev = i-1;
+   list[i].next = RADEON_NR_TEX_REGIONS;
+   list[RADEON_NR_TEX_REGIONS].prev = i;
+   list[RADEON_NR_TEX_REGIONS].next = 0;
+   rmesa->sarea->texAge[heap] = 0;
+}
+
+/* Update the local and glock texture LRUs.
+ */
+void r200UpdateTexLRU(r200ContextPtr rmesa, r200TexObjPtr t )
+{
+   int heap = t->heap;
+   radeon_tex_region_t *list = rmesa->sarea->texList[heap];
+   int sz = rmesa->r200Screen->logTexGranularity[heap];
+   int i, start, end;
+
+   rmesa->texture.age[heap] = ++rmesa->sarea->texAge[heap];
+
+   if ( !t->memBlock ) 
+      return;
+
+   start = t->memBlock->ofs >> sz;
+   end = (t->memBlock->ofs + t->memBlock->size-1) >> sz;
+
+   /* Update our local LRU */
+   move_to_head( &rmesa->texture.objects[heap], t );
+
+   /* Update the global LRU */
+   for ( i = start ; i <= end ; i++ ) {
+      list[i].in_use = 1;
+      list[i].age = rmesa->texture.age[heap];
+
+      /* remove_from_list(i) */
+      list[(GLuint)list[i].next].prev = list[i].prev;
+      list[(GLuint)list[i].prev].next = list[i].next;
+
+      /* insert_at_head(list, i) */
+      list[i].prev = RADEON_NR_TEX_REGIONS;
+      list[i].next = list[RADEON_NR_TEX_REGIONS].next;
+      list[(GLuint)list[RADEON_NR_TEX_REGIONS].next].prev = i;
+      list[RADEON_NR_TEX_REGIONS].next = i;
+   }
+
+   if ( 0 ) {
+      r200PrintGlobalLRU( rmesa, t->heap );
+      r200PrintLocalLRU( rmesa, t->heap );
+   }
+}
+
+/* Update our notion of what textures have been changed since we last
+ * held the lock.  This pertains to both our local textures and the
+ * textures belonging to other clients.  Keep track of other client's
+ * textures by pushing a placeholder texture onto the LRU list -- these
+ * are denoted by (tObj == NULL).
+ */
+static void r200TexturesGone( r200ContextPtr rmesa, int heap,
+				int offset, int size, int in_use )
+{
+   r200TexObjPtr t, tmp;
+
+   foreach_s ( t, tmp, &rmesa->texture.objects[heap] ) {
+      if ( !t->memBlock ||
+	   t->memBlock->ofs >= offset + size ||
+	   t->memBlock->ofs + t->memBlock->size <= offset )
+	 continue;
+
+      /* It overlaps - kick it out.  Need to hold onto the currently
+       * bound objects, however.
+       */
+      r200SwapOutTexObj( rmesa, t );
+   }
+
+   if ( in_use ) {
+      t = (r200TexObjPtr) CALLOC( sizeof(*t) );
+      if ( !t ) return;
+
+      t->memBlock = mmAllocMem( rmesa->texture.heap[heap], size, 0, offset );
+      if ( !t->memBlock ) {
+	 fprintf( stderr, "Couldn't alloc placeholder sz %x ofs %x\n",
+		  (int)size, (int)offset );
+	 mmDumpMemInfo( rmesa->texture.heap[heap] );
+	 return;
+      }
+      insert_at_head( &rmesa->texture.objects[heap], t );
+   }
+}
+
+/* Update our client's shared texture state.  If another client has
+ * modified a region in which we have textures, then we need to figure
+ * out which of our textures has been removed, and update our global
+ * LRU.
+ */
+void r200AgeTextures( r200ContextPtr rmesa, int heap )
+{
+   RADEONSAREAPrivPtr sarea = rmesa->sarea;
+
+   if ( R200_DEBUG & DEBUG_TEXTURE ) {
+      fprintf(stderr, "%s %d\n", __FUNCTION__, heap);
+   }
+
+   if ( sarea->texAge[heap] != rmesa->texture.age[heap] ) {
+      int sz = 1 << rmesa->r200Screen->logTexGranularity[heap];
+      int nr = 0;
+      int idx;
+
+      for ( idx = sarea->texList[heap][RADEON_NR_TEX_REGIONS].prev ;
+	    idx != RADEON_NR_TEX_REGIONS && nr < RADEON_NR_TEX_REGIONS ;
+	    idx = sarea->texList[heap][idx].prev, nr++ )
+      {
+	 /* If switching texturing schemes, then the SAREA might not
+	  * have been properly cleared, so we need to reset the
+	  * global texture LRU.
+	  */
+	 if ( idx * sz > rmesa->r200Screen->texSize[heap] ) {
+	    nr = RADEON_NR_TEX_REGIONS;
+	    break;
+	 }
+
+	 if ( sarea->texList[heap][idx].age > rmesa->texture.age[heap] ) {
+	    r200TexturesGone( rmesa, heap, idx * sz, sz,
+				sarea->texList[heap][idx].in_use );
+	 }
+      }
+
+      if ( nr == RADEON_NR_TEX_REGIONS ) {
+	 r200TexturesGone( rmesa, heap, 0,
+			     rmesa->r200Screen->texSize[heap], 0 );
+	 r200ResetGlobalLRU( rmesa, heap );
+      }
+
+      rmesa->texture.age[heap] = sarea->texAge[heap];
+   }
+}
+
+
+/* ------------------------------------------------------------
+ * Texture image conversions
+ */
+
+
+static void r200UploadAGPClientSubImage( r200ContextPtr rmesa,
+					 r200TexObjPtr t, 
+					 struct gl_texture_image *texImage,
+					 GLint hwlevel,
+					 GLint x, GLint y, 
+					 GLint width, GLint height )
+{
+   const struct gl_texture_format *texFormat = texImage->TexFormat;
+   GLuint srcPitch, dstPitch;
+   int blit_format;
+   int srcOffset;
+
+   /*
+    * XXX it appears that we always upload the full image, not a subimage.
+    * I.e. x==0, y==0, width=texWidth, height=texWidth.  If this is ever
+    * changed, the src pitch will have to change.
+    */
+   switch ( texFormat->TexelBytes ) {
+   case 1:
+      blit_format = R200_CP_COLOR_FORMAT_CI8;
+      srcPitch = t->image[0][0].width * texFormat->TexelBytes;
+      dstPitch = t->image[0][0].width * texFormat->TexelBytes;
+      break;
+   case 2:
+      blit_format = R200_CP_COLOR_FORMAT_RGB565;
+      srcPitch = t->image[0][0].width * texFormat->TexelBytes;
+      dstPitch = t->image[0][0].width * texFormat->TexelBytes;
+      break;
+   case 4:
+      blit_format = R200_CP_COLOR_FORMAT_ARGB8888;
+      srcPitch = t->image[0][0].width * texFormat->TexelBytes;
+      dstPitch = t->image[0][0].width * texFormat->TexelBytes;
+      break;
+   default:
+      return;
+   }
+
+   t->image[0][hwlevel].data = texImage->Data;
+   srcOffset = r200AgpOffsetFromVirtual( rmesa, texImage->Data );
+
+   assert( srcOffset != ~0 );
+
+   /* Don't currently need to cope with small pitches?
+    */
+   width = texImage->Width;
+   height = texImage->Height;
+
+   r200EmitWait( rmesa, RADEON_WAIT_3D );
+
+   r200EmitBlit( rmesa, blit_format, 
+		 srcPitch,  
+		 srcOffset,   
+		 dstPitch,
+		 t->bufAddr,
+		 x, 
+		 y, 
+		 t->image[0][hwlevel].x + x,
+		 t->image[0][hwlevel].y + y, 
+		 width,
+		 height );
+
+   r200EmitWait( rmesa, RADEON_WAIT_2D );
+}
+
+static void r200UploadRectSubImage( r200ContextPtr rmesa,
+				    r200TexObjPtr t, 
+				    struct gl_texture_image *texImage,
+				    GLint x, GLint y, 
+				    GLint width, GLint height )
+{
+   const struct gl_texture_format *texFormat = texImage->TexFormat;
+   int blit_format, dstPitch, done;
+
+   switch ( texFormat->TexelBytes ) {
+   case 1:
+      blit_format = R200_CP_COLOR_FORMAT_CI8;
+      break;
+   case 2:
+      blit_format = R200_CP_COLOR_FORMAT_RGB565;
+      break;
+   case 4:
+      blit_format = R200_CP_COLOR_FORMAT_ARGB8888;
+      break;
+   default:
+      return;
+   }
+
+   t->image[0][0].data = texImage->Data;
+
+   /* Currently don't need to cope with small pitches.
+    */
+   width = texImage->Width;
+   height = texImage->Height;
+   dstPitch = t->pp_txpitch + 32;
+
+   if (rmesa->prefer_agp_client_texturing && texImage->IsClientData) {
+      /* In this case, could also use agp texturing.  This is
+       * currently disabled, but has been tested & works.
+       */
+      t->pp_txoffset = r200AgpOffsetFromVirtual( rmesa, texImage->Data );
+      t->pp_txpitch = texImage->RowStride * texFormat->TexelBytes - 32;
+
+      if (R200_DEBUG & DEBUG_TEXTURE)
+	 fprintf(stderr, 
+		 "Using agp texturing for rectangular client texture\n");
+
+      /* Release FB memory allocated for this image:
+       */
+      if ( t->memBlock ) {
+	 mmFreeMem( t->memBlock );
+	 t->memBlock = NULL;
+      }
+
+   }
+   else if (texImage->IsClientData) {
+      /* Data already in agp memory, with usable pitch.
+       */
+      GLuint srcPitch;
+      srcPitch = texImage->RowStride * texFormat->TexelBytes;
+      r200EmitBlit( rmesa, 
+		    blit_format, 
+		    srcPitch,
+		    r200AgpOffsetFromVirtual( rmesa, texImage->Data ),   
+		    dstPitch, t->bufAddr,
+		    0, 0, 
+		    0, 0, 
+		    width, height );
+   }
+   else {
+      /* Data not in agp memory, or bad pitch.
+       */
+      for (done = 0; done < height ; ) {
+	 struct r200_dma_region region;
+	 int lines = MIN2( height - done, RADEON_BUFFER_SIZE / dstPitch );
+	 int src_pitch;
+	 char *tex;
+
+         src_pitch = texImage->RowStride * texFormat->TexelBytes;
+
+	 tex = (char *)texImage->Data + done * src_pitch;
+
+	 memset(&region, 0, sizeof(region));
+	 r200AllocDmaRegion( rmesa, &region, lines * dstPitch, 64 );
+
+	 /* Copy texdata to dma:
+	  */
+	 if (0)
+	    fprintf(stderr, "%s: src_pitch %d dst_pitch %d\n",
+		    __FUNCTION__, src_pitch, dstPitch);
+
+	 if (src_pitch == dstPitch) {
+	    memcpy( region.address, tex, lines * src_pitch );
+	 } 
+	 else {
+	    char *buf = region.address;
+	    int i;
+	    for (i = 0 ; i < lines ; i++) {
+	       memcpy( buf, tex, src_pitch );
+	       buf += dstPitch;
+	       tex += src_pitch;
+	    }
+	 }
+
+	 r200EmitWait( rmesa, RADEON_WAIT_3D );
+
+	 /* Blit to framebuffer
+	  */
+	 r200EmitBlit( rmesa, 
+		       blit_format, 
+		       dstPitch, GET_START( &region ),   
+		       dstPitch, t->bufAddr,
+		       0, 0, 
+		       0, done, 
+		       width, lines );
+	 
+	 r200EmitWait( rmesa, RADEON_WAIT_2D );
+
+	 r200ReleaseDmaRegion( rmesa, &region, __FUNCTION__ );
+	 done += lines;
+      }
+   }
+}
+
+
+/* Upload the texture image associated with texture `t' at level `level'
+ * at the address relative to `start'.
+ */
+static void r200UploadSubImage( r200ContextPtr rmesa,
+				r200TexObjPtr t, 
+				GLint hwlevel,
+				GLint x, GLint y, GLint width, GLint height,
+                                GLuint face )
+{
+   struct gl_texture_image *texImage = NULL;
+   const struct gl_texture_format *texFormat;
+   GLint texelsPerDword = 0;
+   GLuint format, pitch, offset;
+   GLint imageWidth, imageHeight;
+   GLint ret;
+   drmRadeonTexture tex;
+   drmRadeonTexImage tmp;
+   int level = hwlevel + t->firstLevel;
+
+   if ( R200_DEBUG & DEBUG_TEXTURE ) {
+      fprintf( stderr, "%s level %d %dx%d\n", __FUNCTION__,
+	       level, width, height);
+   }
+
+   ASSERT(face < 6);
+
+   /* Ensure we have a valid texture to upload */
+   if ( ( hwlevel < 0 ) || ( hwlevel >= RADEON_MAX_TEXTURE_LEVELS ) ) {
+      _mesa_problem(NULL, "bad texture level in r200UploadSubimage");
+      return;
+   }
+
+   switch (face) {
+   case 0:
+      texImage = t->tObj->Image[level];
+      break;
+   case 1:
+      texImage = t->tObj->NegX[level];
+      break;
+   case 2:
+      texImage = t->tObj->PosY[level];
+      break;
+   case 3:
+      texImage = t->tObj->NegY[level];
+      break;
+   case 4:
+      texImage = t->tObj->PosZ[level];
+      break;
+   case 5:
+      texImage = t->tObj->NegZ[level];
+      break;
+   }
+
+   if ( !texImage ) {
+      if ( R200_DEBUG & DEBUG_TEXTURE )
+	 fprintf( stderr, "%s: texImage %d is NULL!\n", __FUNCTION__, level );
+      return;
+   }
+   if ( !texImage->Data ) {
+      if ( R200_DEBUG & DEBUG_TEXTURE )
+	 fprintf( stderr, "%s: image data is NULL!\n", __FUNCTION__ );
+      return;
+   }
+
+
+   if (t->tObj->Target == GL_TEXTURE_RECTANGLE_NV) {
+      assert(level == 0);
+      assert(hwlevel == 0);
+      if ( R200_DEBUG & DEBUG_TEXTURE )
+	 fprintf( stderr, "%s: image data is rectangular\n", __FUNCTION__);
+      r200UploadRectSubImage( rmesa, t, texImage, x, y, width, height );
+      return;
+   }
+   else if (texImage->IsClientData) {
+      if ( R200_DEBUG & DEBUG_TEXTURE )
+	 fprintf( stderr, "%s: image data is in agp client storage\n",
+		  __FUNCTION__);
+      r200UploadAGPClientSubImage( rmesa, t, texImage, hwlevel,
+				   x, y, width, height );
+      return;
+   }
+   else if ( R200_DEBUG & DEBUG_TEXTURE )
+      fprintf( stderr, "%s: image data is in normal memory\n",
+	       __FUNCTION__);
+      
+
+   texFormat = texImage->TexFormat;
+
+   switch ( texFormat->TexelBytes ) {
+   case 1:
+      texelsPerDword = 4;
+      break;
+   case 2:
+      texelsPerDword = 2;
+      break;
+   case 4:
+      texelsPerDword = 1;
+      break;
+   }
+
+   format = t->pp_txformat & R200_TXFORMAT_FORMAT_MASK;
+
+   imageWidth = texImage->Width;
+   imageHeight = texImage->Height;
+
+   offset = t->bufAddr;
+
+   if (texFormat->TexelBytes == 0)
+      pitch = (t->image[face][0].width * 1) / 64;
+   else
+      pitch = (t->image[face][0].width * texFormat->TexelBytes) / 64;
+
+
+   if ( R200_DEBUG & (DEBUG_TEXTURE|DEBUG_IOCTL) )
+   {
+      GLint imageX = 0;
+      GLint imageY = 0;
+      GLint blitX = t->image[face][hwlevel].x;
+      GLint blitY = t->image[face][hwlevel].y;
+      GLint blitWidth = t->image[face][hwlevel].width;
+      GLint blitHeight = t->image[face][hwlevel].height;
+      fprintf( stderr, "   upload image: %d,%d at %d,%d\n",
+	       imageWidth, imageHeight, imageX, imageY );
+      fprintf( stderr, "   upload  blit: %d,%d at %d,%d\n",
+	       blitWidth, blitHeight, blitX, blitY );
+      fprintf( stderr, "       blit ofs: 0x%07x pitch: 0x%x "
+	       "level: %d/%d format: %x\n",
+	       (GLuint)offset, (GLuint)pitch, hwlevel, level, format );
+   }
+
+   t->image[face][hwlevel].data = texImage->Data;
+
+   /* Init the DRM_RADEON_TEXTURE command / drmRadeonTexture struct.
+    * NOTE: we're always use a 1KB-wide blit and I8 texture format.
+    * We used to use 1, 2 and 4-byte texels and used to use the texture
+    * width to dictate the blit width - but that won't work for compressed
+    * textures. (Brian)
+    */
+   tex.offset = offset;
+   tex.pitch = BLIT_WIDTH_BYTES / 64;
+   tex.format = R200_TXFORMAT_I8; /* any 1-byte texel format */
+   if (texImage->TexFormat->TexelBytes) {
+      tex.width = imageWidth * texImage->TexFormat->TexelBytes; /* in bytes */
+      tex.height = imageHeight;
+   }
+   else {
+      tex.width = imageWidth; /* compressed */
+      tex.height = imageHeight;
+      if (tex.height < 4)
+         tex.height = 4;
+   }
+   tex.image = &tmp;
+
+   /* copy (x,y,width,height,data) */
+   memcpy( &tmp, &t->image[face][hwlevel], sizeof(drmRadeonTexImage) );
+
+   LOCK_HARDWARE( rmesa );
+   do {
+      ret = drmCommandWriteRead( rmesa->dri.fd, DRM_RADEON_TEXTURE,
+                                 &tex, sizeof(drmRadeonTexture) );
+      if (ret) {
+	 if (R200_DEBUG & DEBUG_IOCTL)
+	    fprintf(stderr, "DRM_RADEON_TEXTURE:  again!\n");
+	 usleep(1);
+      }
+   } while ( ret && errno == EAGAIN );
+
+   UNLOCK_HARDWARE( rmesa );
+
+   if ( ret ) {
+      fprintf( stderr, "DRM_R200_TEXTURE: return = %d\n", ret );
+      fprintf( stderr, "   offset=0x%08x pitch=0x%x format=%d\n",
+	       offset, pitch, format );
+      fprintf( stderr, "   image width=%d height=%d\n",
+	       imageWidth, imageHeight );
+      fprintf( stderr, "    blit width=%d height=%d data=%p\n",
+	       t->image[face][hwlevel].width, t->image[face][hwlevel].height,
+	       t->image[face][hwlevel].data );
+      exit( 1 );
+   }
+}
+
+
+
+/* Upload the texture images associated with texture `t'.  This might
+ * require removing our own and/or other client's texture objects to
+ * make room for these images.
+ */
+int r200UploadTexImages( r200ContextPtr rmesa, r200TexObjPtr t, GLuint face )
+{
+   const int numLevels = t->lastLevel - t->firstLevel + 1;
+   int heap;
+   r200TexObjPtr t0 = rmesa->state.texture.unit[0].texobj;
+   r200TexObjPtr t1 = rmesa->state.texture.unit[1].texobj;
+
+   if ( R200_DEBUG & (DEBUG_TEXTURE|DEBUG_IOCTL) ) {
+      fprintf( stderr, "%s( %p, %p ) sz=%d lvls=%d-%d\n", __FUNCTION__,
+	       rmesa->glCtx, t->tObj, t->totalSize,
+	       t->firstLevel, t->lastLevel );
+   }
+
+   if ( !t || t->totalSize == 0 )
+      return 0;
+
+   if (R200_DEBUG & DEBUG_SYNC) {
+      fprintf(stderr, "\nSyncing\n\n");
+      R200_FIREVERTICES( rmesa );
+      r200Finish( rmesa->glCtx );
+   }
+
+   LOCK_HARDWARE( rmesa );
+
+   /* Choose the heap appropriately */
+   heap = t->heap = RADEON_CARD_HEAP;
+
+   /* Do we need to eject LRU texture objects? */
+   if ( !t->memBlock ) {
+      /* Allocate a memory block on a 1k boundary (1<<10 == 1024) */
+      t->memBlock = mmAllocMem( rmesa->texture.heap[heap],
+				t->totalSize, 10, 0 );
+
+
+      /* Kick out textures until the requested texture fits */
+      while ( !t->memBlock ) {
+	 if ( rmesa->texture.objects[heap].prev == t0 ||
+	      rmesa->texture.objects[heap].prev == t1 ) {
+	    fprintf( stderr,
+		     "r200UploadTexImages: ran into bound texture\n" );
+	    UNLOCK_HARDWARE( rmesa );
+	    return -1;
+	 }
+	 if ( rmesa->texture.objects[heap].prev ==
+	      &rmesa->texture.objects[heap] ) {
+	    if ( rmesa->r200Screen->IsPCI ) {
+	       fprintf( stderr, "r200UploadTexImages: upload texture "
+			"failure on local texture heaps, sz=%d\n",
+			t->totalSize );
+	       UNLOCK_HARDWARE( rmesa );
+	       return -1;
+	    } else {
+	       fprintf( stderr, "r200UploadTexImages: upload texture "
+			"failure on both local and AGP texture heaps, "
+			"sz=%d\n",
+			t->totalSize );
+	       UNLOCK_HARDWARE( rmesa );
+	       return -1;
+	    }
+	 }
+
+	 r200SwapOutTexObj( rmesa, rmesa->texture.objects[heap].prev );
+
+	 t->memBlock = mmAllocMem( rmesa->texture.heap[heap],
+				   t->totalSize, 12, 0 );
+      }
+
+      /* Set the base offset of the texture image */
+      t->bufAddr = rmesa->r200Screen->texOffset[heap] + t->memBlock->ofs;
+      t->pp_txoffset = t->bufAddr;
+
+      /* Mark this texobj as dirty on all units:
+       */
+      t->dirty_state = TEX_ALL;
+   }
+
+   /* Let the world know we've used this memory recently */
+   r200UpdateTexLRU( rmesa, t );
+   UNLOCK_HARDWARE( rmesa );
+
+   /* Upload any images that are new */
+   if (t->dirty_images[face]) {
+      int hwlevel;
+      for ( hwlevel = 0 ; hwlevel < numLevels ; hwlevel++ ) {
+         if ( t->dirty_images[face] & (1 << (hwlevel+t->firstLevel)) ) {
+            r200UploadSubImage( rmesa, t, hwlevel, 
+                                0, 0,
+                                t->image[face][hwlevel].width, 
+                                t->image[face][hwlevel].height, face );
+         }
+      }
+      t->dirty_images[face] = 0;
+   }
+
+
+   if (R200_DEBUG & DEBUG_SYNC) {
+      fprintf(stderr, "\nSyncing\n\n");
+      r200Finish( rmesa->glCtx );
+   }
+
+   return 0;
+}

src/mesa/drivers/dri/r200/r200_vtxfmt.h

@@ -0,0 +1,128 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#ifndef __R200_VTXFMT_H__
+#define __R200_VTXFMT_H__
+
+#ifdef GLX_DIRECT_RENDERING
+
+#include "r200_context.h"
+
+
+
+extern struct r200_vb vb;
+
+
+extern void r200VtxfmtUpdate( GLcontext *ctx );
+extern void r200VtxfmtInit( GLcontext *ctx );
+extern void r200VtxfmtInvalidate( GLcontext *ctx );
+extern void r200VtxfmtDestroy( GLcontext *ctx );
+extern void r200VtxfmtInitChoosers( GLvertexformat *vfmt );
+
+extern void r200VtxfmtMakeCurrent( GLcontext *ctx );
+extern void r200VtxfmtUnbindContext( GLcontext *ctx );
+
+extern void r200_copy_to_current( GLcontext *ctx );
+
+#define DFN( FUNC, CACHE)				\
+do {							\
+   char *start = (char *)&FUNC;				\
+   char *end = (char *)&FUNC##_end;			\
+   insert_at_head( &CACHE, dfn );			\
+   dfn->key[0] = key[0];					\
+   dfn->key[1] = key[1];					\
+   dfn->code = ALIGN_MALLOC( end - start, 16 );		\
+   memcpy (dfn->code, start, end - start);		\
+}							\
+while ( 0 )
+
+#define FIXUP( CODE, OFFSET, CHECKVAL, NEWVAL )	\
+do {						\
+   int *icode = (int *)(CODE+OFFSET);		\
+   assert (*icode == CHECKVAL);			\
+   *icode = (int)NEWVAL;			\
+} while (0)
+
+
+/* Useful for figuring out the offsets:
+ */
+#define FIXUP2( CODE, OFFSET, CHECKVAL, NEWVAL )		\
+do {								\
+   while (*(int *)(CODE+OFFSET) != CHECKVAL) OFFSET++;		\
+   /*fprintf(stderr, "%s/%d CVAL %x OFFSET %d VAL %x\n", __FUNCTION__,*/ \
+   /*	   __LINE__, CHECKVAL, OFFSET, (int)(NEWVAL));*/		\
+   *(int *)(CODE+OFFSET) = (int)(NEWVAL);				\
+   OFFSET += 4;							\
+} while (0)
+
+/* 
+ */
+void r200InitCodegen( struct dfn_generators *gen );
+void r200InitX86Codegen( struct dfn_generators *gen );
+void r200InitSSECodegen( struct dfn_generators *gen );
+
+
+
+/* Defined in r200_vtxfmt_x86.c
+ */
+struct dynfn *r200_makeX86Vertex2f( GLcontext *, const int * );
+struct dynfn *r200_makeX86Vertex2fv( GLcontext *, const int * );
+struct dynfn *r200_makeX86Vertex3f( GLcontext *, const int * );
+struct dynfn *r200_makeX86Vertex3fv( GLcontext *, const int * );
+struct dynfn *r200_makeX86Color4ub( GLcontext *, const int * );
+struct dynfn *r200_makeX86Color4ubv( GLcontext *, const int * );
+struct dynfn *r200_makeX86Color3ub( GLcontext *, const int * );
+struct dynfn *r200_makeX86Color3ubv( GLcontext *, const int * );
+struct dynfn *r200_makeX86Color4f( GLcontext *, const int * );
+struct dynfn *r200_makeX86Color4fv( GLcontext *, const int * );
+struct dynfn *r200_makeX86Color3f( GLcontext *, const int * );
+struct dynfn *r200_makeX86Color3fv( GLcontext *, const int * );
+struct dynfn *r200_makeX86SecondaryColor3ubEXT( GLcontext *, const int * );
+struct dynfn *r200_makeX86SecondaryColor3ubvEXT( GLcontext *, const int * );
+struct dynfn *r200_makeX86SecondaryColor3fEXT( GLcontext *, const int * );
+struct dynfn *r200_makeX86SecondaryColor3fvEXT( GLcontext *, const int * );
+struct dynfn *r200_makeX86Normal3f( GLcontext *, const int * );
+struct dynfn *r200_makeX86Normal3fv( GLcontext *, const int * );
+struct dynfn *r200_makeX86TexCoord2f( GLcontext *, const int * );
+struct dynfn *r200_makeX86TexCoord2fv( GLcontext *, const int * );
+struct dynfn *r200_makeX86TexCoord1f( GLcontext *, const int * );
+struct dynfn *r200_makeX86TexCoord1fv( GLcontext *, const int * );
+struct dynfn *r200_makeX86MultiTexCoord2fARB( GLcontext *, const int * );
+struct dynfn *r200_makeX86MultiTexCoord2fvARB( GLcontext *, const int * );
+struct dynfn *r200_makeX86MultiTexCoord1fARB( GLcontext *, const int * );
+struct dynfn *r200_makeX86MultiTexCoord1fvARB( GLcontext *, const int * );
+
+
+#endif
+#endif

src/mesa/drivers/dri/r200/r200_vtxfmt_c.c

@@ -0,0 +1,797 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#include "glheader.h"
+#include "imports.h"
+#include "mtypes.h"
+#include "colormac.h"
+#include "simple_list.h"
+#include "api_noop.h"
+#include "vtxfmt.h"
+
+#include "r200_vtxfmt.h"
+
+/* Fallback versions of all the entrypoints for situations where
+ * codegen isn't available.  This is still a lot faster than the
+ * vb/pipeline implementation in Mesa.
+ */
+static void r200_Vertex3f( GLfloat x, GLfloat y, GLfloat z )
+{
+   int i;
+
+   *vb.dmaptr++ = *(int *)&x;
+   *vb.dmaptr++ = *(int *)&y;
+   *vb.dmaptr++ = *(int *)&z;
+
+   for (i = 3; i < vb.vertex_size; i++)
+      *vb.dmaptr++ = vb.vertex[i].i;
+   
+   if (--vb.counter == 0)
+      vb.notify();
+}
+
+
+static void r200_Vertex3fv( const GLfloat *v )
+{
+   int i;
+
+   *vb.dmaptr++ = *(int *)&v[0];
+   *vb.dmaptr++ = *(int *)&v[1];
+   *vb.dmaptr++ = *(int *)&v[2];
+
+   for (i = 3; i < vb.vertex_size; i++)
+      *vb.dmaptr++ = vb.vertex[i].i;
+   
+   if (--vb.counter == 0)
+      vb.notify();
+}
+
+
+static void r200_Vertex2f( GLfloat x, GLfloat y )
+{
+   int i;
+   
+   *vb.dmaptr++ = *(int *)&x;
+   *vb.dmaptr++ = *(int *)&y;
+   *vb.dmaptr++ = 0;
+
+   for (i = 3; i < vb.vertex_size; i++)
+      *vb.dmaptr++ = vb.vertex[i].i;
+
+   if (--vb.counter == 0)
+      vb.notify();
+}
+
+
+static void r200_Vertex2fv( const GLfloat *v )
+{
+   int i;
+
+   *vb.dmaptr++ = *(int *)&v[0];
+   *vb.dmaptr++ = *(int *)&v[1];
+   *vb.dmaptr++ = 0;
+
+   for (i = 3; i < vb.vertex_size; i++)
+      *vb.dmaptr++ = vb.vertex[i].i;
+   
+   if (--vb.counter == 0)
+      vb.notify();
+}
+
+
+
+/* Color for ubyte (packed) color formats:
+ */
+static void r200_Color3ub_ub( GLubyte r, GLubyte g, GLubyte b )
+{
+   r200_color_t *dest = vb.colorptr;
+   dest->red	= r;
+   dest->green	= g;
+   dest->blue	= b;
+   dest->alpha	= 0xff;
+}
+
+static void r200_Color3ubv_ub( const GLubyte *v )
+{
+   r200_color_t *dest = vb.colorptr;
+   dest->red	= v[0];
+   dest->green	= v[1];
+   dest->blue	= v[2];
+   dest->alpha	= 0xff;
+}
+
+static void r200_Color4ub_ub( GLubyte r, GLubyte g, GLubyte b, GLubyte a )
+{
+   r200_color_t *dest = vb.colorptr;
+   dest->red	= r;
+   dest->green	= g;
+   dest->blue	= b;
+   dest->alpha	= a;
+}
+
+static void r200_Color4ubv_ub( const GLubyte *v )
+{
+   *(GLuint *)vb.colorptr = LE32_TO_CPU(*(GLuint *)v);
+}
+
+
+static void r200_Color3f_ub( GLfloat r, GLfloat g, GLfloat b )
+{
+   r200_color_t *dest = vb.colorptr;
+   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,   r );
+   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, g );
+   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  b );
+   dest->alpha = 255;
+}
+
+static void r200_Color3fv_ub( const GLfloat *v )
+{
+   r200_color_t *dest = vb.colorptr;
+   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,   v[0] );
+   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, v[1] );
+   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  v[2] );
+   dest->alpha = 255;
+}
+
+static void r200_Color4f_ub( GLfloat r, GLfloat g, GLfloat b, GLfloat a )
+{
+   r200_color_t *dest = vb.colorptr;
+   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,   r );
+   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, g );
+   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  b );
+   UNCLAMPED_FLOAT_TO_UBYTE( dest->alpha, a );
+}
+
+static void r200_Color4fv_ub( const GLfloat *v )
+{
+   r200_color_t *dest = vb.colorptr;
+   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,	  v[0] );
+   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, v[1] );
+   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  v[2] );
+   UNCLAMPED_FLOAT_TO_UBYTE( dest->alpha, v[3] );
+}
+
+
+/* Color for float color+alpha formats:
+ */
+static void r200_Color3ub_4f( GLubyte r, GLubyte g, GLubyte b )
+{
+   GLfloat *dest = vb.floatcolorptr;
+   dest[0] = UBYTE_TO_FLOAT(r);
+   dest[1] = UBYTE_TO_FLOAT(g);
+   dest[2] = UBYTE_TO_FLOAT(b);
+   dest[3] = 1.0;
+}
+
+static void r200_Color3ubv_4f( const GLubyte *v )
+{
+   GLfloat *dest = vb.floatcolorptr;
+   dest[0] = UBYTE_TO_FLOAT(v[0]);
+   dest[1] = UBYTE_TO_FLOAT(v[1]);
+   dest[2] = UBYTE_TO_FLOAT(v[2]);
+   dest[3] = 1.0;
+}
+
+static void r200_Color4ub_4f( GLubyte r, GLubyte g, GLubyte b, GLubyte a )
+{
+   GLfloat *dest = vb.floatcolorptr;
+   dest[0] = UBYTE_TO_FLOAT(r);
+   dest[1] = UBYTE_TO_FLOAT(g);
+   dest[2] = UBYTE_TO_FLOAT(b);
+   dest[3] = UBYTE_TO_FLOAT(a);
+}
+
+static void r200_Color4ubv_4f( const GLubyte *v )
+{
+   GLfloat *dest = vb.floatcolorptr;
+   dest[0] = UBYTE_TO_FLOAT(v[0]);
+   dest[1] = UBYTE_TO_FLOAT(v[1]);
+   dest[2] = UBYTE_TO_FLOAT(v[2]);
+   dest[3] = UBYTE_TO_FLOAT(v[3]);
+}
+
+
+static void r200_Color3f_4f( GLfloat r, GLfloat g, GLfloat b )
+{
+   GLfloat *dest = vb.floatcolorptr;
+   dest[0] = r;
+   dest[1] = g;
+   dest[2] = b;
+   dest[3] = 1.0;		
+}
+
+static void r200_Color3fv_4f( const GLfloat *v )
+{
+   GLfloat *dest = vb.floatcolorptr;
+   dest[0] = v[0];
+   dest[1] = v[1];
+   dest[2] = v[2];
+   dest[3] = 1.0;
+}
+
+static void r200_Color4f_4f( GLfloat r, GLfloat g, GLfloat b, GLfloat a )
+{
+   GLfloat *dest = vb.floatcolorptr;
+   dest[0] = r;
+   dest[1] = g;
+   dest[2] = b;
+   dest[3] = a;
+}
+
+static void r200_Color4fv_4f( const GLfloat *v )
+{
+   GLfloat *dest = vb.floatcolorptr;
+   dest[0] = v[0];
+   dest[1] = v[1];
+   dest[2] = v[2];
+   dest[3] = v[3];
+}
+
+
+/* Color for float color formats:
+ */
+static void r200_Color3ub_3f( GLubyte r, GLubyte g, GLubyte b )
+{
+   GLfloat *dest = vb.floatcolorptr;
+   dest[0] = UBYTE_TO_FLOAT(r);
+   dest[1] = UBYTE_TO_FLOAT(g);
+   dest[2] = UBYTE_TO_FLOAT(b);
+}
+
+static void r200_Color3ubv_3f( const GLubyte *v )
+{
+   GLfloat *dest = vb.floatcolorptr;
+   dest[0] = UBYTE_TO_FLOAT(v[0]);
+   dest[1] = UBYTE_TO_FLOAT(v[1]);
+   dest[2] = UBYTE_TO_FLOAT(v[2]);
+}
+
+static void r200_Color4ub_3f( GLubyte r, GLubyte g, GLubyte b, GLubyte a )
+{
+   GLfloat *dest = vb.floatcolorptr;
+   dest[0] = UBYTE_TO_FLOAT(r);
+   dest[1] = UBYTE_TO_FLOAT(g);
+   dest[2] = UBYTE_TO_FLOAT(b);
+   vb.context->Current.Attrib[VERT_ATTRIB_COLOR0][3] = UBYTE_TO_FLOAT(a);
+}
+
+static void r200_Color4ubv_3f( const GLubyte *v )
+{
+   GLfloat *dest = vb.floatcolorptr;
+   dest[0] = UBYTE_TO_FLOAT(v[0]);
+   dest[1] = UBYTE_TO_FLOAT(v[1]);
+   dest[2] = UBYTE_TO_FLOAT(v[2]);
+   vb.context->Current.Attrib[VERT_ATTRIB_COLOR0][3] = UBYTE_TO_FLOAT(v[3]);
+}
+
+
+static void r200_Color3f_3f( GLfloat r, GLfloat g, GLfloat b )
+{
+   GLfloat *dest = vb.floatcolorptr;
+   dest[0] = r;
+   dest[1] = g;
+   dest[2] = b;
+}
+
+static void r200_Color3fv_3f( const GLfloat *v )
+{
+   GLfloat *dest = vb.floatcolorptr;
+   dest[0] = v[0];
+   dest[1] = v[1];
+   dest[2] = v[2];
+}
+
+static void r200_Color4f_3f( GLfloat r, GLfloat g, GLfloat b, GLfloat a )
+{
+   GLfloat *dest = vb.floatcolorptr;
+   dest[0] = r;
+   dest[1] = g;
+   dest[2] = b;
+   vb.context->Current.Attrib[VERT_ATTRIB_COLOR0][3] = a;
+}
+
+static void r200_Color4fv_3f( const GLfloat *v )
+{
+   GLfloat *dest = vb.floatcolorptr;
+   dest[0] = v[0];
+   dest[1] = v[1];
+   dest[2] = v[2];
+   vb.context->Current.Attrib[VERT_ATTRIB_COLOR0][3] = v[3]; 
+}
+
+
+/* Secondary Color:
+ */
+static void r200_SecondaryColor3ubEXT_ub( GLubyte r, GLubyte g, GLubyte b )
+{
+   r200_color_t *dest = vb.specptr;
+   dest->red	= r;
+   dest->green	= g;
+   dest->blue	= b;
+   dest->alpha	= 0xff;
+}
+
+static void r200_SecondaryColor3ubvEXT_ub( const GLubyte *v )
+{
+   r200_color_t *dest = vb.specptr;
+   dest->red	= v[0];
+   dest->green	= v[1];
+   dest->blue	= v[2];
+   dest->alpha	= 0xff;
+}
+
+static void r200_SecondaryColor3fEXT_ub( GLfloat r, GLfloat g, GLfloat b )
+{
+   r200_color_t *dest = vb.specptr;
+   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,	  r );
+   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, g );
+   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  b );
+   dest->alpha = 255;
+}
+
+static void r200_SecondaryColor3fvEXT_ub( const GLfloat *v )
+{
+   r200_color_t *dest = vb.specptr;
+   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,	  v[0] );
+   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, v[1] );
+   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  v[2] );
+   dest->alpha = 255;
+}
+
+static void r200_SecondaryColor3ubEXT_3f( GLubyte r, GLubyte g, GLubyte b )
+{
+   GLfloat *dest = vb.floatspecptr;
+   dest[0] = UBYTE_TO_FLOAT(r);
+   dest[1] = UBYTE_TO_FLOAT(g);
+   dest[2] = UBYTE_TO_FLOAT(b);
+   dest[3] = 1.0;
+}
+
+static void r200_SecondaryColor3ubvEXT_3f( const GLubyte *v )
+{
+   GLfloat *dest = vb.floatspecptr;
+   dest[0] = UBYTE_TO_FLOAT(v[0]);
+   dest[1] = UBYTE_TO_FLOAT(v[1]);
+   dest[2] = UBYTE_TO_FLOAT(v[2]);
+   dest[3] = 1.0;
+}
+
+static void r200_SecondaryColor3fEXT_3f( GLfloat r, GLfloat g, GLfloat b )
+{
+   GLfloat *dest = vb.floatspecptr;
+   dest[0] = r;
+   dest[1] = g;
+   dest[2] = b;
+   dest[3] = 1.0;
+}
+
+static void r200_SecondaryColor3fvEXT_3f( const GLfloat *v )
+{
+   GLfloat *dest = vb.floatspecptr;
+   dest[0] = v[0];
+   dest[1] = v[1];
+   dest[2] = v[2];
+   dest[3] = 1.0;
+}
+
+
+
+/* Normal
+ */
+static void r200_Normal3f( GLfloat n0, GLfloat n1, GLfloat n2 )
+{
+   GLfloat *dest = vb.normalptr;
+   dest[0] = n0;
+   dest[1] = n1;
+   dest[2] = n2;
+}
+
+static void r200_Normal3fv( const GLfloat *v )
+{
+   GLfloat *dest = vb.normalptr;
+   dest[0] = v[0];
+   dest[1] = v[1];
+   dest[2] = v[2];
+}
+
+
+/* TexCoord
+ */
+static void r200_TexCoord1f( GLfloat s )
+{
+   GLfloat *dest = vb.texcoordptr[0];
+   dest[0] = s;
+   dest[1] = 0;
+}
+
+static void r200_TexCoord1fv( const GLfloat *v )
+{
+   GLfloat *dest = vb.texcoordptr[0];
+   dest[0] = v[0];
+   dest[1] = 0;
+}
+
+static void r200_TexCoord2f( GLfloat s, GLfloat t )
+{
+   GLfloat *dest = vb.texcoordptr[0];
+   dest[0] = s;
+   dest[1] = t;
+}
+
+static void r200_TexCoord2fv( const GLfloat *v )
+{
+   GLfloat *dest = vb.texcoordptr[0];
+   dest[0] = v[0];
+   dest[1] = v[1];
+}
+
+
+/* MultiTexcoord
+ */
+static void r200_MultiTexCoord1fARB( GLenum target, GLfloat s  )
+{
+   GLfloat *dest = vb.texcoordptr[(target - GL_TEXTURE0_ARB)&1];
+   dest[0] = s;
+   dest[1] = 0;
+}
+
+static void r200_MultiTexCoord1fvARB( GLenum target, const GLfloat *v )
+{
+   GLfloat *dest = vb.texcoordptr[(target - GL_TEXTURE0_ARB)&1];
+   dest[0] = v[0];
+   dest[1] = 0;
+}
+
+static void r200_MultiTexCoord2fARB( GLenum target, GLfloat s, GLfloat t )
+{
+   GLfloat *dest = vb.texcoordptr[(target - GL_TEXTURE0_ARB)&1];
+   dest[0] = s;
+   dest[1] = t;
+}
+
+static void r200_MultiTexCoord2fvARB( GLenum target, const GLfloat *v )
+{
+   GLfloat *dest = vb.texcoordptr[(target - GL_TEXTURE0_ARB)&1];
+   dest[0] = v[0];
+   dest[1] = v[1];
+}
+
+static struct dynfn *lookup( struct dynfn *l, const int *key )
+{
+   struct dynfn *f;
+
+   foreach( f, l ) {
+      if (f->key[0] == key[0] && f->key[1] == key[1]) 
+	 return f;
+   }
+
+   return 0;
+}
+
+/* Can't use the loopback template for this:
+ */
+
+#define CHOOSE(FN, FNTYPE, MASK0, MASK1, ARGS1, ARGS2 )			\
+static void choose_##FN ARGS1						\
+{									\
+   r200ContextPtr rmesa = R200_CONTEXT(vb.context);			\
+   int key[2];								\
+   struct dynfn *dfn;							\
+									\
+   key[0] = rmesa->vb.vtxfmt_0 & MASK0;				\
+   key[1] = rmesa->vb.vtxfmt_1 & MASK1;				\
+									\
+   dfn = lookup( &rmesa->vb.dfn_cache.FN, key );			\
+   if (dfn == 0)							\
+      dfn = rmesa->vb.codegen.FN( vb.context, key );			\
+   else if (R200_DEBUG & DEBUG_CODEGEN)					\
+      fprintf(stderr, "%s -- cached codegen\n", __FUNCTION__ );		\
+									\
+   if (dfn)								\
+      vb.context->Exec->FN = (FNTYPE)(dfn->code);			\
+   else {								\
+      if (R200_DEBUG & DEBUG_CODEGEN)					\
+	 fprintf(stderr, "%s -- generic version\n", __FUNCTION__ );	\
+      vb.context->Exec->FN = r200_##FN;					\
+   }									\
+									\
+   vb.context->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;		\
+   vb.context->Exec->FN ARGS2;						\
+}
+
+
+
+/* For the _3f case, only allow one color function to be hooked in at
+ * a time.  Eventually, use a similar mechanism to allow selecting the
+ * color component of the vertex format based on client behaviour.  
+ *
+ * Note:  Perform these actions even if there is a codegen or cached 
+ * codegen version of the chosen function.
+ */
+#define CHOOSE_COLOR(FN, FNTYPE, NR, MASK0, MASK1, ARGS1, ARGS2 )	\
+static void choose_##FN ARGS1						\
+{									\
+   GLcontext *ctx = vb.context;						\
+   r200ContextPtr rmesa = R200_CONTEXT(vb.context);			\
+   int key[2];								\
+   struct dynfn *dfn;							\
+									\
+   key[0] = rmesa->vb.vtxfmt_0 & MASK0;					\
+   key[1] = rmesa->vb.vtxfmt_1 & MASK1;					\
+									\
+   if (VTX_COLOR(rmesa->vb.vtxfmt_0,0) == R200_VTX_PK_RGBA) {		\
+      ctx->Exec->FN = r200_##FN##_ub;					\
+   }									\
+   else if (VTX_COLOR(rmesa->vb.vtxfmt_0,0) == R200_VTX_FP_RGB) {	\
+									\
+      if (rmesa->vb.installed_color_3f_sz != NR) {			\
+         rmesa->vb.installed_color_3f_sz = NR;				\
+         if (NR == 3) ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = 1.0;	\
+         if (ctx->Driver.NeedFlush & FLUSH_UPDATE_CURRENT) {		\
+            r200_copy_to_current( ctx );				\
+            _mesa_install_exec_vtxfmt( ctx, &rmesa->vb.vtxfmt );	\
+            ctx->Exec->FN ARGS2;					\
+            return;							\
+         }								\
+      }									\
+									\
+      ctx->Exec->FN = r200_##FN##_3f;					\
+   }									\
+   else {								\
+      ctx->Exec->FN = r200_##FN##_4f;					\
+   }									\
+									\
+									\
+   dfn = lookup( &rmesa->vb.dfn_cache.FN, key );			\
+   if (!dfn) dfn = rmesa->vb.codegen.FN( ctx, key );			\
+									\
+   if (dfn) {								\
+      if (R200_DEBUG & DEBUG_CODEGEN)					\
+         fprintf(stderr, "%s -- codegen version\n", __FUNCTION__ );	\
+      ctx->Exec->FN = (FNTYPE)dfn->code;				\
+   }									\
+   else if (R200_DEBUG & DEBUG_CODEGEN)					\
+         fprintf(stderr, "%s -- 'c' version\n", __FUNCTION__ );		\
+									\
+   ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;			\
+   ctx->Exec->FN ARGS2;							\
+}
+
+
+
+/* Right now there are both _ub and _3f versions of the secondary color
+ * functions.  Currently, we only set-up the hardware to use the _ub versions.
+ * The _3f versions are needed for the cases where secondary color isn't used
+ * in the vertex format, but it still needs to be stored in the context
+ * state vector.
+ */
+#define CHOOSE_SECONDARY_COLOR(FN, FNTYPE, MASK0, MASK1, ARGS1, ARGS2 )	\
+static void choose_##FN ARGS1						\
+{									\
+   r200ContextPtr rmesa = R200_CONTEXT(vb.context);			\
+   int key[2];								\
+   struct dynfn *dfn;							\
+									\
+   key[0] = rmesa->vb.vtxfmt_0 & MASK0;					\
+   key[1] = rmesa->vb.vtxfmt_1 & MASK1;					\
+									\
+   dfn = lookup( &rmesa->vb.dfn_cache.FN, key );			\
+   if (dfn == 0)							\
+      dfn = rmesa->vb.codegen.FN( vb.context, key );			\
+   else  if (R200_DEBUG & DEBUG_CODEGEN)				\
+      fprintf(stderr, "%s -- cached version\n", __FUNCTION__ );		\
+									\
+   if (dfn)								\
+      vb.context->Exec->FN = (FNTYPE)(dfn->code);			\
+   else {								\
+      if (R200_DEBUG & DEBUG_CODEGEN)					\
+         fprintf(stderr, "%s -- generic version\n", __FUNCTION__ );	\
+      vb.context->Exec->FN = (VTX_COLOR(rmesa->vb.vtxfmt_0,1) == R200_VTX_PK_RGBA) \
+	  ? r200_##FN##_ub : r200_##FN##_3f;				\
+   }									\
+									\
+   vb.context->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;		\
+   vb.context->Exec->FN ARGS2;						\
+}
+
+
+
+
+
+/* VTXFMT_0
+ */
+#define MASK_XYZW  (R200_VTX_W0|R200_VTX_Z0)
+#define MASK_NORM  (MASK_XYZW|R200_VTX_N0)
+#define MASK_COLOR (MASK_NORM |(R200_VTX_COLOR_MASK<<R200_VTX_COLOR_0_SHIFT))
+#define MASK_SPEC  (MASK_COLOR|(R200_VTX_COLOR_MASK<<R200_VTX_COLOR_1_SHIFT))
+
+/* VTXFMT_1
+ */
+#define MASK_ST0 (0x7 << R200_VTX_TEX0_COMP_CNT_SHIFT)
+
+
+
+typedef void (*p4f)( GLfloat, GLfloat, GLfloat, GLfloat );
+typedef void (*p3f)( GLfloat, GLfloat, GLfloat );
+typedef void (*p2f)( GLfloat, GLfloat );
+typedef void (*p1f)( GLfloat );
+typedef void (*pe2f)( GLenum, GLfloat, GLfloat );
+typedef void (*pe1f)( GLenum, GLfloat );
+typedef void (*p4ub)( GLubyte, GLubyte, GLubyte, GLubyte );
+typedef void (*p3ub)( GLubyte, GLubyte, GLubyte );
+typedef void (*pfv)( const GLfloat * );
+typedef void (*pefv)( GLenum, const GLfloat * );
+typedef void (*pubv)( const GLubyte * );
+
+
+CHOOSE(Normal3f, p3f, MASK_NORM, 0, 
+       (GLfloat a,GLfloat b,GLfloat c), (a,b,c))
+CHOOSE(Normal3fv, pfv, MASK_NORM, 0, 
+       (const GLfloat *v), (v))
+
+CHOOSE_COLOR(Color4ub, p4ub, 4, MASK_COLOR, 0,
+	(GLubyte a,GLubyte b, GLubyte c, GLubyte d), (a,b,c,d))
+CHOOSE_COLOR(Color4ubv, pubv, 4, MASK_COLOR, 0, 
+	(const GLubyte *v), (v))
+CHOOSE_COLOR(Color3ub, p3ub, 3, MASK_COLOR, 0, 
+	(GLubyte a,GLubyte b, GLubyte c), (a,b,c))
+CHOOSE_COLOR(Color3ubv, pubv, 3, MASK_COLOR, 0, 
+	(const GLubyte *v), (v))
+
+CHOOSE_COLOR(Color4f, p4f, 4, MASK_COLOR, 0, 
+	(GLfloat a,GLfloat b, GLfloat c, GLfloat d), (a,b,c,d))
+CHOOSE_COLOR(Color4fv, pfv, 4, MASK_COLOR, 0, 
+	(const GLfloat *v), (v))
+CHOOSE_COLOR(Color3f, p3f, 3, MASK_COLOR, 0,
+	(GLfloat a,GLfloat b, GLfloat c), (a,b,c))
+CHOOSE_COLOR(Color3fv, pfv, 3, MASK_COLOR, 0,
+	(const GLfloat *v), (v))
+
+
+CHOOSE_SECONDARY_COLOR(SecondaryColor3ubEXT, p3ub, MASK_SPEC, 0, 
+	(GLubyte a,GLubyte b, GLubyte c), (a,b,c))
+CHOOSE_SECONDARY_COLOR(SecondaryColor3ubvEXT, pubv, MASK_SPEC, 0, 
+	(const GLubyte *v), (v))
+CHOOSE_SECONDARY_COLOR(SecondaryColor3fEXT, p3f, MASK_SPEC, 0,
+	(GLfloat a,GLfloat b, GLfloat c), (a,b,c))
+CHOOSE_SECONDARY_COLOR(SecondaryColor3fvEXT, pfv, MASK_SPEC, 0,
+	(const GLfloat *v), (v))
+
+CHOOSE(TexCoord2f, p2f, ~0, MASK_ST0, 
+       (GLfloat a,GLfloat b), (a,b))
+CHOOSE(TexCoord2fv, pfv, ~0, MASK_ST0, 
+       (const GLfloat *v), (v))
+CHOOSE(TexCoord1f, p1f, ~0, MASK_ST0, 
+       (GLfloat a), (a))
+CHOOSE(TexCoord1fv, pfv, ~0, MASK_ST0, 
+       (const GLfloat *v), (v))
+
+CHOOSE(MultiTexCoord2fARB, pe2f, ~0, ~0,
+	 (GLenum u,GLfloat a,GLfloat b), (u,a,b))
+CHOOSE(MultiTexCoord2fvARB, pefv, ~0, ~0,
+	(GLenum u,const GLfloat *v), (u,v))
+CHOOSE(MultiTexCoord1fARB, pe1f, ~0, ~0,
+	 (GLenum u,GLfloat a), (u,a))
+CHOOSE(MultiTexCoord1fvARB, pefv, ~0, ~0,
+	(GLenum u,const GLfloat *v), (u,v))
+
+CHOOSE(Vertex3f, p3f, ~0, ~0, 
+       (GLfloat a,GLfloat b,GLfloat c), (a,b,c))
+CHOOSE(Vertex3fv, pfv, ~0, ~0, 
+       (const GLfloat *v), (v))
+CHOOSE(Vertex2f, p2f, ~0, ~0, 
+       (GLfloat a,GLfloat b), (a,b))
+CHOOSE(Vertex2fv, pfv, ~0, ~0, 
+       (const GLfloat *v), (v))
+
+
+
+
+
+void r200VtxfmtInitChoosers( GLvertexformat *vfmt )
+{
+   vfmt->Color3f = choose_Color3f;
+   vfmt->Color3fv = choose_Color3fv;
+   vfmt->Color3ub = choose_Color3ub;
+   vfmt->Color3ubv = choose_Color3ubv;
+   vfmt->Color4f = choose_Color4f;
+   vfmt->Color4fv = choose_Color4fv;
+   vfmt->Color4ub = choose_Color4ub;
+   vfmt->Color4ubv = choose_Color4ubv;
+   vfmt->SecondaryColor3fEXT = choose_SecondaryColor3fEXT;
+   vfmt->SecondaryColor3fvEXT = choose_SecondaryColor3fvEXT;
+   vfmt->SecondaryColor3ubEXT = choose_SecondaryColor3ubEXT;
+   vfmt->SecondaryColor3ubvEXT = choose_SecondaryColor3ubvEXT;
+   vfmt->MultiTexCoord1fARB = choose_MultiTexCoord1fARB;
+   vfmt->MultiTexCoord1fvARB = choose_MultiTexCoord1fvARB;
+   vfmt->MultiTexCoord2fARB = choose_MultiTexCoord2fARB;
+   vfmt->MultiTexCoord2fvARB = choose_MultiTexCoord2fvARB;
+   vfmt->Normal3f = choose_Normal3f;
+   vfmt->Normal3fv = choose_Normal3fv;
+   vfmt->TexCoord1f = choose_TexCoord1f;
+   vfmt->TexCoord1fv = choose_TexCoord1fv;
+   vfmt->TexCoord2f = choose_TexCoord2f;
+   vfmt->TexCoord2fv = choose_TexCoord2fv;
+   vfmt->Vertex2f = choose_Vertex2f;
+   vfmt->Vertex2fv = choose_Vertex2fv;
+   vfmt->Vertex3f = choose_Vertex3f;
+   vfmt->Vertex3fv = choose_Vertex3fv;
+}
+
+
+static struct dynfn *codegen_noop( GLcontext *ctx, const int *key )
+{
+   (void) ctx; (void) key;
+   return 0;
+}
+
+void r200InitCodegen( struct dfn_generators *gen )
+{
+   gen->Vertex3f = codegen_noop;
+   gen->Vertex3fv = codegen_noop;
+   gen->Color4ub = codegen_noop;
+   gen->Color4ubv = codegen_noop;
+   gen->Normal3f = codegen_noop;
+   gen->Normal3fv = codegen_noop;
+   gen->TexCoord2f = codegen_noop;
+   gen->TexCoord2fv = codegen_noop;
+   gen->MultiTexCoord2fARB = codegen_noop;
+   gen->MultiTexCoord2fvARB = codegen_noop;
+   gen->Vertex2f = codegen_noop;
+   gen->Vertex2fv = codegen_noop;
+   gen->Color3ub = codegen_noop;
+   gen->Color3ubv = codegen_noop;
+   gen->Color4f = codegen_noop;
+   gen->Color4fv = codegen_noop;
+   gen->Color3f = codegen_noop;
+   gen->Color3fv = codegen_noop;
+   gen->SecondaryColor3fEXT = codegen_noop;
+   gen->SecondaryColor3fvEXT = codegen_noop;
+   gen->SecondaryColor3ubEXT = codegen_noop;
+   gen->SecondaryColor3ubvEXT = codegen_noop;
+   gen->TexCoord1f = codegen_noop;
+   gen->TexCoord1fv = codegen_noop;
+   gen->MultiTexCoord1fARB = codegen_noop;
+   gen->MultiTexCoord1fvARB = codegen_noop;
+
+   if (!getenv("R200_NO_CODEGEN")) {
+#if defined(USE_X86_ASM)
+      r200InitX86Codegen( gen );
+#endif
+
+#if defined(USE_SSE_ASM)
+      r200InitSSECodegen( gen );
+#endif
+   }
+}

src/mesa/drivers/dri/r200/r200_vtxfmt_sse.c

@@ -0,0 +1,95 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#include "glheader.h"
+#include "imports.h"
+#include "simple_list.h" 
+#include "r200_vtxfmt.h"
+
+#if defined(USE_SSE_ASM)
+
+/* Build specialized versions of the immediate calls on the fly for
+ * the current state.  ???P4 SSE2 versions???
+ */
+
+
+static struct dynfn *makeSSENormal3fv( GLcontext *ctx, const int *key )
+{
+   /* Requires P4 (sse2?)
+    */
+   static unsigned char temp[] = {
+      0x8b, 0x44, 0x24, 0x04,          	/*  mov    0x4(%esp,1),%eax */
+      0xba, 0x78, 0x56, 0x34, 0x12,   	/*  mov    $0x12345678,%edx */
+      0xf3, 0x0f, 0x7e, 0x00,          	/*  movq   (%eax),%xmm0 */
+      0x66, 0x0f, 0x6e, 0x48, 0x08,    	/*  movd   0x8(%eax),%xmm1 */
+      0x66, 0x0f, 0xd6, 0x42, 0x0c,    	/*  movq   %xmm0,0xc(%edx) */
+      0x66, 0x0f, 0x7e, 0x4a, 0x14,    	/*  movd   %xmm1,0x14(%edx) */
+      0xc3,                   	        /*  ret     */
+   };
+
+
+   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   insert_at_head( &rmesa->vb.dfn_cache.Normal3fv, dfn );
+   dfn->key[0] = key[0];
+   dfn->key[1] = key[1];
+
+   dfn->code = ALIGN_MALLOC( sizeof(temp), 16 );
+   memcpy (dfn->code, temp, sizeof(temp));
+   FIXUP(dfn->code, 5, 0x0, (int)vb.normalptr); 
+   return dfn;
+}
+
+void r200InitSSECodegen( struct dfn_generators *gen )
+{
+   /* Need to: 
+    *    - check kernel sse support
+    *    - check p4/sse2
+    */
+   (void) makeSSENormal3fv;
+}
+
+
+#else 
+
+void r200InitSSECodegen( struct dfn_generators *gen )
+{
+   (void) gen;
+}
+
+#endif
+
+
+
+

src/mesa/drivers/dri/r200/r200_vtxfmt_x86.c

@@ -0,0 +1,462 @@
+/* $XFree86$ */
+/*
+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 <keith@tungstengraphics.com>
+ */
+
+#include "glheader.h"
+#include "imports.h"
+#include "mmath.h" 
+#include "simple_list.h" 
+#include "r200_vtxfmt.h"
+
+#if defined(USE_X86_ASM)
+
+#define EXTERN( FUNC )		\
+extern const char *FUNC;	\
+extern const char *FUNC##_end
+
+EXTERN ( _x86_Normal3fv );
+EXTERN ( _x86_Normal3f );
+EXTERN ( _x86_Vertex3fv_6 );
+EXTERN ( _x86_Vertex3fv_8 );
+EXTERN ( _x86_Vertex3fv );
+EXTERN ( _x86_Vertex3f_4 );
+EXTERN ( _x86_Vertex3f_6 );
+EXTERN ( _x86_Vertex3f );
+EXTERN ( _x86_Color4ubv_ub );
+EXTERN ( _x86_Color4ubv_4f );
+EXTERN ( _x86_Color4ub_ub );
+EXTERN ( _x86_Color3fv_3f );
+EXTERN ( _x86_Color3f_3f );
+EXTERN ( _x86_TexCoord2fv );
+EXTERN ( _x86_TexCoord2f );
+EXTERN ( _x86_MultiTexCoord2fvARB );
+EXTERN ( _x86_MultiTexCoord2fvARB_2 );
+EXTERN ( _x86_MultiTexCoord2fARB );
+EXTERN ( _x86_MultiTexCoord2fARB_2 );
+
+
+/* Build specialized versions of the immediate calls on the fly for
+ * the current state.  Generic x86 versions.
+ */
+
+struct dynfn *r200_makeX86Vertex3f( GLcontext *ctx, const int *key )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
+
+   if (R200_DEBUG & DEBUG_CODEGEN)
+      fprintf(stderr, "%s 0x%08x 0x%08x %d\n", __FUNCTION__, 
+	      key[0], key[1], vb.vertex_size );
+
+   switch (vb.vertex_size) {
+   case 4: {
+
+      DFN ( _x86_Vertex3f_4, rmesa->vb.dfn_cache.Vertex3f );
+      FIXUP(dfn->code, 2, 0x0, (int)&vb.dmaptr);
+      FIXUP(dfn->code, 25, 0x0, (int)&vb.vertex[3]);
+      FIXUP(dfn->code, 36, 0x0, (int)&vb.counter);
+      FIXUP(dfn->code, 46, 0x0, (int)&vb.dmaptr);
+      FIXUP(dfn->code, 51, 0x0, (int)&vb.counter);
+      FIXUP(dfn->code, 60, 0x0, (int)&vb.notify);
+      break;
+   }
+   case 6: {
+
+      DFN ( _x86_Vertex3f_6, rmesa->vb.dfn_cache.Vertex3f );
+      FIXUP(dfn->code, 3, 0x0, (int)&vb.dmaptr);
+      FIXUP(dfn->code, 28, 0x0, (int)&vb.vertex[3]);
+      FIXUP(dfn->code, 34, 0x0, (int)&vb.vertex[4]);
+      FIXUP(dfn->code, 40, 0x0, (int)&vb.vertex[5]);
+      FIXUP(dfn->code, 57, 0x0, (int)&vb.counter);
+      FIXUP(dfn->code, 63, 0x0, (int)&vb.dmaptr);
+      FIXUP(dfn->code, 70, 0x0, (int)&vb.counter);
+      FIXUP(dfn->code, 79, 0x0, (int)&vb.notify);
+      break;
+   }
+   default: {
+
+      DFN ( _x86_Vertex3f, rmesa->vb.dfn_cache.Vertex3f );
+      FIXUP(dfn->code, 3, 0x0, (int)&vb.vertex[3]);
+      FIXUP(dfn->code, 9, 0x0, (int)&vb.dmaptr);
+      FIXUP(dfn->code, 37, 0x0, vb.vertex_size-3);
+      FIXUP(dfn->code, 44, 0x0, (int)&vb.counter);
+      FIXUP(dfn->code, 50, 0x0, (int)&vb.dmaptr);
+      FIXUP(dfn->code, 56, 0x0, (int)&vb.counter);
+      FIXUP(dfn->code, 67, 0x0, (int)&vb.notify);
+   break;
+   }
+   }
+
+   return dfn;
+}
+
+
+
+struct dynfn *r200_makeX86Vertex3fv( GLcontext *ctx, const int *key )
+{
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
+
+   if (R200_DEBUG & DEBUG_CODEGEN)
+      fprintf(stderr, "%s 0x%08x 0x%08x %d\n", __FUNCTION__, 
+	      key[0], key[1], vb.vertex_size );
+
+   switch (vb.vertex_size) {
+   case 6: {
+
+      DFN ( _x86_Vertex3fv_6, rmesa->vb.dfn_cache.Vertex3fv );
+      FIXUP(dfn->code, 1, 0x00000000, (int)&vb.dmaptr);
+      FIXUP(dfn->code, 27, 0x0000001c, (int)&vb.vertex[3]);
+      FIXUP(dfn->code, 33, 0x00000020, (int)&vb.vertex[4]);
+      FIXUP(dfn->code, 45, 0x00000024, (int)&vb.vertex[5]);
+      FIXUP(dfn->code, 56, 0x00000000, (int)&vb.dmaptr);
+      FIXUP(dfn->code, 61, 0x00000004, (int)&vb.counter);
+      FIXUP(dfn->code, 67, 0x00000004, (int)&vb.counter);
+      FIXUP(dfn->code, 76, 0x00000008, (int)&vb.notify);
+      break;
+   }
+   
+
+   case 8: {
+
+      DFN ( _x86_Vertex3fv_8, rmesa->vb.dfn_cache.Vertex3fv );
+      FIXUP(dfn->code, 1, 0x00000000, (int)&vb.dmaptr);
+      FIXUP(dfn->code, 27, 0x0000001c, (int)&vb.vertex[3]);
+      FIXUP(dfn->code, 33, 0x00000020, (int)&vb.vertex[4]);
+      FIXUP(dfn->code, 45, 0x0000001c, (int)&vb.vertex[5]);
+      FIXUP(dfn->code, 51, 0x00000020, (int)&vb.vertex[6]);
+      FIXUP(dfn->code, 63, 0x00000024, (int)&vb.vertex[7]);
+      FIXUP(dfn->code, 74, 0x00000000, (int)&vb.dmaptr);
+      FIXUP(dfn->code, 79, 0x00000004, (int)&vb.counter);
+      FIXUP(dfn->code, 85, 0x00000004, (int)&vb.counter);
+      FIXUP(dfn->code, 94, 0x00000008, (int)&vb.notify);
+      break;
+   }
+   
+
+
+   default: {
+
+      DFN ( _x86_Vertex3fv, rmesa->vb.dfn_cache.Vertex3fv );
+      FIXUP(dfn->code, 8, 0x01010101, (int)&vb.dmaptr);
+      FIXUP(dfn->code, 32, 0x00000006, vb.vertex_size-3);
+      FIXUP(dfn->code, 37, 0x00000058, (int)&vb.vertex[3]);
+      FIXUP(dfn->code, 45, 0x01010101, (int)&vb.dmaptr);
+      FIXUP(dfn->code, 50, 0x02020202, (int)&vb.counter);
+      FIXUP(dfn->code, 58, 0x02020202, (int)&vb.counter);
+      FIXUP(dfn->code, 67, 0x0, (int)&vb.notify);
+   break;
+   }
+   }
+
+   return dfn;
+}
+
+struct dynfn *r200_makeX86Normal3fv( GLcontext *ctx, const int *key )
+{
+   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+   int i = 0;
+
+   if (R200_DEBUG & DEBUG_CODEGEN)
+      fprintf(stderr, "%s 0x%08x\n", __FUNCTION__, key[0] );
+
+   DFN ( _x86_Normal3fv, rmesa->vb.dfn_cache.Normal3fv );
+
+   FIXUP2(dfn->code, i, 0x0, (int)vb.normalptr); 
+   FIXUP2(dfn->code, i, 0x4, 4+(int)vb.normalptr); 
+   FIXUP2(dfn->code, i, 0x8, 8+(int)vb.normalptr); 
+   /*fprintf(stderr, "%s done\n", __FUNCTION__);*/
+   return dfn;
+}
+
+struct dynfn *r200_makeX86Normal3f( GLcontext *ctx, const int *key )
+{
+   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+
+   if (R200_DEBUG & DEBUG_CODEGEN)
+      fprintf(stderr, "%s 0x%08x\n", __FUNCTION__, key[0] );
+
+   DFN ( _x86_Normal3f, rmesa->vb.dfn_cache.Normal3f );
+   FIXUP(dfn->code, 1, 0x12345678, (int)vb.normalptr); 
+   return dfn;
+}
+
+struct dynfn *r200_makeX86Color4ubv( GLcontext *ctx, const int *key )
+{
+   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+
+
+   if (R200_DEBUG & DEBUG_CODEGEN)
+      fprintf(stderr, "%s 0x%08x\n", __FUNCTION__, key[0] );
+
+   if (VTX_COLOR(key[0],0) == R200_VTX_PK_RGBA) {
+      DFN ( _x86_Color4ubv_ub, rmesa->vb.dfn_cache.Color4ubv);
+      FIXUP(dfn->code, 5, 0x12345678, (int)vb.colorptr); 
+      return dfn;
+   } 
+   else {
+
+      DFN ( _x86_Color4ubv_4f, rmesa->vb.dfn_cache.Color4ubv);
+      FIXUP(dfn->code, 2, 0x00000000, (int)_mesa_ubyte_to_float_color_tab); 
+      FIXUP(dfn->code, 27, 0xdeadbeaf, (int)vb.floatcolorptr); 
+      FIXUP(dfn->code, 33, 0xdeadbeaf, (int)vb.floatcolorptr+4); 
+      FIXUP(dfn->code, 55, 0xdeadbeaf, (int)vb.floatcolorptr+8); 
+      FIXUP(dfn->code, 61, 0xdeadbeaf, (int)vb.floatcolorptr+12); 
+      return dfn;
+   }
+}
+
+struct dynfn *r200_makeX86Color4ub( GLcontext *ctx, const int *key )
+{
+   if (R200_DEBUG & DEBUG_CODEGEN)
+      fprintf(stderr, "%s 0x%08x\n", __FUNCTION__, key[0] );
+
+   if (VTX_COLOR(key[0],0) == R200_VTX_PK_RGBA) {
+      struct dynfn *dfn = MALLOC_STRUCT( dynfn );
+      r200ContextPtr rmesa = R200_CONTEXT(ctx);
+
+      DFN ( _x86_Color4ub_ub, rmesa->vb.dfn_cache.Color4ub );
+      FIXUP(dfn->code, 18, 0x0, (int)vb.colorptr); 
+      FIXUP(dfn->code, 24, 0x0, (int)vb.colorptr+1); 
+      FIXUP(dfn->code, 30, 0x0, (int)vb.colorptr+2); 
+      FIXUP(dfn->code, 36, 0x0, (int)vb.colorptr+3); 
+      return dfn;
+   }
+   else
+      return 0;
+}
+
+
+struct dynfn *r200_makeX86Color3fv( GLcontext *ctx, const int *key )
+{
+   if (VTX_COLOR(key[0],0) != R200_VTX_FP_RGB) 
+      return 0;
+   else
+   {
+      struct dynfn *dfn = MALLOC_STRUCT( dynfn );
+      r200ContextPtr rmesa = R200_CONTEXT(ctx);
+
+      if (R200_DEBUG & DEBUG_CODEGEN)
+	 fprintf(stderr, "%s 0x%08x\n", __FUNCTION__, key[0] );
+
+      DFN ( _x86_Color3fv_3f, rmesa->vb.dfn_cache.Color3fv );
+      FIXUP(dfn->code, 5, 0x0, (int)vb.floatcolorptr); 
+      return dfn;
+   }
+}
+
+struct dynfn *r200_makeX86Color3f( GLcontext *ctx, const int *key )
+{
+   if (VTX_COLOR(key[0],0) != R200_VTX_FP_RGB) 
+      return 0;
+   else
+   {
+      struct dynfn *dfn = MALLOC_STRUCT( dynfn );
+      r200ContextPtr rmesa = R200_CONTEXT(ctx);
+
+      if (R200_DEBUG & DEBUG_CODEGEN)
+	 fprintf(stderr, "%s 0x%08x\n", __FUNCTION__, key[0] );
+
+      DFN ( _x86_Color3f_3f, rmesa->vb.dfn_cache.Color3f );
+      FIXUP(dfn->code, 1, 0x12345678, (int)vb.floatcolorptr); 
+      return dfn;
+   }
+}
+
+
+
+struct dynfn *r200_makeX86TexCoord2fv( GLcontext *ctx, const int *key )
+{
+
+   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+
+   if (R200_DEBUG & DEBUG_CODEGEN)
+      fprintf(stderr, "%s 0x%08x 0x%08x\n", __FUNCTION__, key[0], key[1] );
+
+   DFN ( _x86_TexCoord2fv, rmesa->vb.dfn_cache.TexCoord2fv );
+   FIXUP(dfn->code, 5, 0x12345678, (int)vb.texcoordptr[0]); 
+   return dfn;
+}
+
+struct dynfn *r200_makeX86TexCoord2f( GLcontext *ctx, const int *key )
+{
+
+   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+
+   if (R200_DEBUG & DEBUG_CODEGEN)
+      fprintf(stderr, "%s 0x%08x 0x%08x\n", __FUNCTION__, key[0], key[1] );
+
+   DFN ( _x86_TexCoord2f, rmesa->vb.dfn_cache.TexCoord2f );
+   FIXUP(dfn->code, 1, 0x12345678, (int)vb.texcoordptr[0]); 
+   return dfn;
+}
+
+struct dynfn *r200_makeX86MultiTexCoord2fvARB( GLcontext *ctx, const int *key )
+{
+#if 0
+   static  char temp[] = {
+      0x8b, 0x44, 0x24, 0x04,          	/* mov    0x4(%esp,1),%eax */
+      0x8b, 0x4c, 0x24, 0x08,          	/* mov    0x8(%esp,1),%ecx */
+      0x2d, 0xc0, 0x84, 0x00, 0x00,    	/* sub    $0x84c0,%eax */
+      0x83, 0xe0, 0x01,             	/* and    $0x1,%eax */
+      0x8b, 0x11,                	/* mov    (%ecx),%edx */
+      0xc1, 0xe0, 0x03,             	/* shl    $0x3,%eax */
+      0x8b, 0x49, 0x04,             	/* mov    0x4(%ecx),%ecx */
+      0x89, 0x90, 0, 0, 0, 0,/* mov    %edx,DEST(%eax) */
+      0x89, 0x88, 0, 0, 0, 0,/* mov    %ecx,DEST+8(%eax) */
+      0xc3,                     	/* ret     */
+   };
+   static char temp2[] = {
+      0x8b, 0x44, 0x24, 0x04,          	/* mov    0x4(%esp,1),%eax */
+      0x8b, 0x4c, 0x24, 0x08,          	/* mov    0x8(%esp,1),%ecx */
+      0x2d, 0xc0, 0x84, 0x00, 0x00,    	/* sub    $0x84c0,%eax */
+      0x83, 0xe0, 0x01,             	/* and    $0x1,%eax */
+      0x8b, 0x14, 0x85, 0, 0, 0, 0, /* mov    DEST(,%eax,4),%edx */
+      0x8b, 0x01,                	/* mov    (%ecx),%eax */
+      0x89, 0x02,                	/* mov    %eax,(%edx) */
+      0x8b, 0x41, 0x04,             	/* mov    0x4(%ecx),%eax */
+      0x89, 0x42, 0x04,             	/* mov    %eax,0x4(%edx) */
+      0xc3,                     	/* ret     */
+   };
+#endif
+
+   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+
+   if (R200_DEBUG & DEBUG_CODEGEN)
+      fprintf(stderr, "%s 0x%08x 0x%08x\n", __FUNCTION__, key[0], key[1] );
+
+   if (vb.texcoordptr[1] == vb.texcoordptr[0]+4) {
+      DFN ( _x86_MultiTexCoord2fvARB, rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
+      FIXUP(dfn->code, 26, 0xdeadbeef, (int)vb.texcoordptr[0]);	
+      FIXUP(dfn->code, 32, 0xdeadbeef, (int)vb.texcoordptr[0]+4);
+   } else {
+      DFN ( _x86_MultiTexCoord2fvARB_2, rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
+      FIXUP(dfn->code, 19, 0x0, (int)vb.texcoordptr);
+   }
+   return dfn;
+}
+
+struct dynfn *r200_makeX86MultiTexCoord2fARB( GLcontext *ctx, 
+					      const int *key )
+{
+#if 0
+   static  char temp[] = {
+      0x8b, 0x44, 0x24, 0x04,          	/* mov    0x4(%esp,1),%eax */
+      0x8b, 0x54, 0x24, 0x08,          	/* mov    0x8(%esp,1),%edx */
+      0x2d, 0xc0, 0x84, 0x00, 0x00,    	/* sub    $0x84c0,%eax */
+      0x8b, 0x4c, 0x24, 0x0c,          	/* mov    0xc(%esp,1),%ecx */
+      0x83, 0xe0, 0x01,             	/* and    $0x1,%eax */
+      0xc1, 0xe0, 0x03,             	/* shl    $0x3,%eax */
+      0x89, 0x90, 0, 0, 0, 0,	/* mov    %edx,DEST(%eax) */
+      0x89, 0x88, 0, 0, 0, 0,	/* mov    %ecx,DEST+8(%eax) */
+      0xc3,                     	/* ret     */
+   };
+
+   static char temp2[] = {
+      0x8b, 0x44, 0x24, 0x04,          	/* mov    0x4(%esp,1),%eax */
+      0x8b, 0x54, 0x24, 0x08,          	/* mov    0x8(%esp,1),%edx */
+      0x2d, 0xc0, 0x84, 0x00, 0x00,    	/* sub    $0x84c0,%eax */
+      0x8b, 0x4c, 0x24, 0x0c,          	/* mov    0xc(%esp,1),%ecx */
+      0x83, 0xe0, 0x01,             	/* and    $0x1,%eax */
+      0x8b, 0x04, 0x85, 0, 0, 0, 0,     /* mov    DEST(,%eax,4),%eax */
+      0x89, 0x10,                	/* mov    %edx,(%eax) */
+      0x89, 0x48, 0x04,             	/* mov    %ecx,0x4(%eax) */
+      0xc3,                   	        /* ret     */
+   };
+#endif
+   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
+   r200ContextPtr rmesa = R200_CONTEXT(ctx);
+
+   if (R200_DEBUG & DEBUG_CODEGEN)
+      fprintf(stderr, "%s 0x%08x 0x%08x\n", __FUNCTION__, key[0], key[1] );
+
+   if (vb.texcoordptr[1] == vb.texcoordptr[0]+4) {
+      DFN ( _x86_MultiTexCoord2fARB, rmesa->vb.dfn_cache.MultiTexCoord2fARB );
+      FIXUP(dfn->code, 25, 0xdeadbeef, (int)vb.texcoordptr[0]); 
+      FIXUP(dfn->code, 31, 0xdeadbeef, (int)vb.texcoordptr[0]+4); 
+   }
+   else {
+      /* Note: this might get generated multiple times, even though the
+       * actual emitted code is the same.
+       */
+      DFN ( _x86_MultiTexCoord2fARB_2, rmesa->vb.dfn_cache.MultiTexCoord2fARB );
+      FIXUP(dfn->code, 23, 0x0, (int)vb.texcoordptr); 
+   }      
+   return dfn;
+}
+
+
+void r200InitX86Codegen( struct dfn_generators *gen )
+{
+   gen->Vertex3f = r200_makeX86Vertex3f;
+   gen->Vertex3fv = r200_makeX86Vertex3fv;
+   gen->Color4ub = r200_makeX86Color4ub; /* PKCOLOR only */
+   gen->Color4ubv = r200_makeX86Color4ubv; /* PKCOLOR only */
+   gen->Normal3f = r200_makeX86Normal3f;
+   gen->Normal3fv = r200_makeX86Normal3fv;
+   gen->TexCoord2f = r200_makeX86TexCoord2f;
+   gen->TexCoord2fv = r200_makeX86TexCoord2fv;
+   gen->MultiTexCoord2fARB = r200_makeX86MultiTexCoord2fARB;
+   gen->MultiTexCoord2fvARB = r200_makeX86MultiTexCoord2fvARB;
+   gen->Color3f = r200_makeX86Color3f;
+   gen->Color3fv = r200_makeX86Color3fv;
+
+   /* Not done:
+    */
+/*     gen->Vertex2f = r200_makeX86Vertex2f; */
+/*     gen->Vertex2fv = r200_makeX86Vertex2fv; */
+/*     gen->Color3ub = r200_makeX86Color3ub; */
+/*     gen->Color3ubv = r200_makeX86Color3ubv; */
+/*     gen->Color4f = r200_makeX86Color4f; */
+/*     gen->Color4fv = r200_makeX86Color4fv; */
+/*     gen->TexCoord1f = r200_makeX86TexCoord1f; */
+/*     gen->TexCoord1fv = r200_makeX86TexCoord1fv; */
+/*     gen->MultiTexCoord1fARB = r200_makeX86MultiTexCoord1fARB; */
+/*     gen->MultiTexCoord1fvARB = r200_makeX86MultiTexCoord1fvARB; */
+}
+
+
+#else 
+
+void r200InitX86Codegen( struct dfn_generators *gen )
+{
+   (void) gen;
+}
+
+#endif

src/mesa/drivers/dri/r200/r200_vtxtmp_x86.S

@@ -0,0 +1,410 @@
+/* $XFree86: xc/lib/GL/mesa/src/drv/r200/r200_vtxtmp_x86.S,v 1.1 2002/10/30 12:51:53 alanh Exp $ */
+/**************************************************************************
+
+Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas.
+
+All Rights Reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a
+copy of this software and associated documentation files (the "Software"),
+to deal in the Software without restriction, including without limitation
+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
+ATI, TUNGSTEN GRAPHICS 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.
+
+**************************************************************************/
+
+#define GLOBL( x )	\
+.globl x;		\
+x:
+
+.data
+.align 4
+GLOBL( _x86_Normal3fv)
+	movl 4(%esp), %eax      /* load 'v' off stack */
+	movl (%eax), %ecx       /* load v[0] */
+	movl 4(%eax), %edx      /* load v[1] */
+	movl 8(%eax), %eax      /* load v[2] */
+	movl %ecx, 0      	/* store v[0] to current vertex */
+	movl %edx, 4      	/* store v[1] to current vertex */
+	movl %eax, 8      	/* store v[2] to current vertex */
+	ret
+GLOBL ( _x86_Normal3fv_end )
+
+/*
+	vertex 3f vertex size 4
+*/
+	
+GLOBL ( _x86_Vertex3f_4 )
+	movl	(0), %ecx
+	movl	4(%esp), %eax
+	movl	8(%esp), %edx
+	movl	%eax, (%ecx)
+	movl	%edx, 4(%ecx)
+	movl	12(%esp), %eax
+	movl	(0), %edx
+	movl	%eax, 8(%ecx)
+	movl	%edx, 12(%ecx)
+	movl	(0), %eax
+	addl	$16, %ecx
+	dec 	%eax
+	movl	%ecx, (0)
+	movl	%eax, (0)
+	je	.1 
+	ret
+.1:	jmp	*0
+	
+GLOBL ( _x86_Vertex3f_4_end )
+
+/*
+	vertex 3f vertex size 6
+*/
+GLOBL ( _x86_Vertex3f_6 )
+	push	%edi
+	movl	(0), %edi
+	movl	8(%esp), %eax
+	movl	12(%esp), %edx
+	movl	16(%esp), %ecx
+	movl	%eax, (%edi)
+	movl	%edx, 4(%edi)
+	movl	%ecx, 8(%edi)
+	movl	(0), %eax
+	movl	(0), %edx
+	movl	(0), %ecx
+	movl	%eax, 12(%edi)
+	movl	%edx, 16(%edi)
+	movl	%ecx, 20(%edi)
+	addl	$24, %edi
+	movl	(0), %eax
+	movl	%edi, (0)
+	dec 	%eax
+	pop 	%edi
+	movl	%eax, (0)
+	je	.2
+	ret
+.2:	jmp	*0
+GLOBL ( _x86_Vertex3f_6_end )
+/*
+	vertex 3f generic size
+*/
+GLOBL ( _x86_Vertex3f )
+	push	%edi
+	push	%esi
+	movl	$0, %esi
+	movl	(0), %edi
+	movl	12(%esp), %eax
+	movl	16(%esp), %edx
+	movl	20(%esp), %ecx
+	movl	%eax, (%edi)
+	movl	%edx, 4(%edi)
+	movl	%ecx, 8(%edi)
+	addl	$12, %edi
+	movl	$0, %ecx
+	repz
+	movsl %ds:(%esi), %es:(%edi)
+	movl	(0), %eax
+	movl	%edi, (0)
+	dec 	%eax
+	movl	%eax, (0)
+	pop 	%esi
+	pop 	%edi
+	je  	.3
+	ret
+.3:	jmp	*0
+
+GLOBL ( _x86_Vertex3f_end )
+
+/*
+	Vertex 3fv vertex size 6
+*/
+GLOBL ( _x86_Vertex3fv_6 )
+	movl	(0), %eax
+	movl	4(%esp), %ecx
+	movl	(%ecx), %edx
+	movl	%edx, (%eax)
+	movl	4(%ecx), %edx
+	movl	8(%ecx), %ecx
+	movl	%edx, 4(%eax)
+	movl	%ecx, 8(%eax)
+	movl	(28), %edx
+	movl	(32), %ecx
+	movl	%edx, 12(%eax)
+	movl	%ecx, 16(%eax)
+	movl	(36), %edx
+	movl	%edx, 20(%eax)
+	addl	$24, %eax
+	movl	%eax, 0
+	movl	4, %eax
+	dec 	%eax
+	movl	%eax, 4
+	je	.4
+	ret
+.4:	jmp    *8
+	
+GLOBL ( _x86_Vertex3fv_6_end )
+
+/*
+	Vertex 3fv vertex size 8
+*/
+GLOBL ( _x86_Vertex3fv_8 )
+	movl	(0), %eax
+	movl	4(%esp), %ecx
+	movl	(%ecx), %edx
+	movl	%edx ,(%eax)
+	movl	4(%ecx) ,%edx
+	movl	8(%ecx) ,%ecx
+	movl	%edx, 4(%eax)
+	movl	%ecx, 8(%eax)
+	movl	(28), %edx
+	movl	(32), %ecx
+	movl	%edx, 12(%eax)
+	movl	%ecx, 16(%eax)
+	movl	(28), %edx
+	movl	(32), %ecx
+	movl	%edx, 20(%eax)
+	movl	%ecx, 24(%eax)
+	movl	(36), %edx
+	movl	%edx, 28(%eax)
+	addl	$32, %eax
+	movl	%eax, (0)
+	movl	4, %eax
+	dec	%eax
+	movl    %eax, (4)
+	je	.5
+	ret
+.5:	jmp    *8
+	
+GLOBL ( _x86_Vertex3fv_8_end )
+
+/*
+	Vertex 3fv generic vertex size
+*/
+GLOBL ( _x86_Vertex3fv )
+	movl	4(%esp), %edx
+	push	%edi
+	push	%esi
+	movl	(0x1010101), %edi
+	movl	(%edx), %eax
+	movl	4(%edx), %ecx
+	movl	8(%edx), %esi
+	movl	%eax, (%edi)
+	movl	%ecx, 4(%edi)
+	movl	%esi, 8(%edi)
+	addl	$12, %edi
+	movl	$6, %ecx
+	movl	$0x58, %esi
+	repz
+	movsl %ds:(%esi), %es:(%edi)
+	movl	%edi, (0x1010101)
+	movl	(0x2020202), %eax
+	pop	%esi
+	pop	%edi
+	dec	%eax
+	movl	%eax, (0x2020202)
+	je	.6
+	ret
+.6:	jmp    *0
+GLOBL ( _x86_Vertex3fv_end )
+
+/*
+	Normal 3f
+*/
+GLOBL ( _x86_Normal3f )
+	movl	$0x12345678, %edx
+	movl	4(%esp), %eax
+	movl	%eax, (%edx)
+	movl	8(%esp), %eax
+	movl	%eax, 4(%edx)
+	movl	12(%esp), %eax
+	movl	%eax, 8(%edx)
+	ret
+GLOBL ( _x86_Normal3f_end )
+
+/*
+	Color 4ubv_ub
+*/
+GLOBL ( _x86_Color4ubv_ub )
+	movl 4(%esp), %eax
+	movl $0x12345678, %edx
+	movl (%eax), %eax
+	movl %eax, (%edx)
+	ret
+GLOBL ( _x86_Color4ubv_ub_end )
+
+/*
+	Color 4ubv 4f
+*/
+GLOBL ( _x86_Color4ubv_4f )
+	push	%ebx
+	movl	$0, %edx
+	xor	%eax, %eax
+	xor	%ecx, %ecx
+	movl	8(%esp), %ebx
+	movl	(%ebx), %ebx
+	mov	%bl, %al
+	mov	%bh, %cl
+	movl	(%edx,%eax,4),%eax
+	movl	(%edx,%ecx,4),%ecx
+	movl	%eax, (0xdeadbeaf)
+	movl	%ecx, (0xdeadbeaf)
+	xor	%eax, %eax
+	xor	%ecx, %ecx
+	shr	$16, %ebx
+	mov	%bl, %al
+	mov	%bh, %cl
+	movl	(%edx,%eax,4), %eax
+	movl	(%edx,%ecx,4), %ecx
+	movl	%eax, (0xdeadbeaf)
+	movl	%ecx, (0xdeadbeaf)
+	pop	%ebx
+	ret
+GLOBL ( _x86_Color4ubv_4f_end )
+
+/*
+
+	Color4ub_ub
+*/
+GLOBL( _x86_Color4ub_ub )
+	push	%ebx
+	movl	8(%esp), %eax
+	movl	12(%esp), %edx
+	movl	16(%esp), %ecx
+	movl	20(%esp), %ebx
+	mov	%al, (0)
+	mov	%dl, (0)
+	mov	%cl, (0)
+	mov	%bl, (0)
+	pop	%ebx
+	ret
+GLOBL( _x86_Color4ub_ub_end )
+
+/*
+	Color3fv_3f
+*/
+GLOBL( _x86_Color3fv_3f )
+	movl	4(%esp), %eax
+	movl	$0, %edx
+	movl	(%eax), %ecx
+	movl	%ecx, (%edx)
+	movl	4(%eax), %ecx
+	movl	%ecx, 4(%edx)
+	movl	8(%eax), %ecx
+	movl	%ecx, 8(%edx)
+	ret
+GLOBL( _x86_Color3fv_3f_end )
+
+/*
+	Color3f_3f
+*/
+GLOBL( _x86_Color3f_3f )
+	movl	$0x12345678, %edx
+	movl	4(%esp), %eax
+	movl	%eax, (%edx)
+	movl	8(%esp,1), %eax
+	movl	%eax, 4(%edx)
+	movl	12(%esp), %eax
+	movl	%eax, 8(%edx)
+	ret
+GLOBL( _x86_Color3f_3f_end )
+
+/*
+	TexCoord2fv
+*/
+
+GLOBL( _x86_TexCoord2fv )
+	movl	4(%esp), %eax
+	movl	$0x12345678, %edx
+	movl	(%eax), %ecx
+	movl	4(%eax), %eax
+	movl	%ecx, (%edx)
+	movl	%eax, 4(%edx)
+	ret
+
+GLOBL( _x86_TexCoord2fv_end )
+/*
+	TexCoord2f
+*/
+GLOBL( _x86_TexCoord2f )
+	movl	$0x12345678, %edx
+	movl	4(%esp), %eax
+	movl	8(%esp), %ecx
+	movl	%eax, (%edx)
+	movl	%ecx, 4(%edx)
+	ret
+GLOBL( _x86_TexCoord2f_end )
+
+/*
+	MultiTexCoord2fvARB st0/st1
+*/
+GLOBL( _x86_MultiTexCoord2fvARB )
+
+	movl	4(%esp), %eax
+	movl	8(%esp), %ecx
+	sub	$0x84c0, %eax
+	and	$1, %eax
+	movl	(%ecx), %edx
+	shl	$3, %eax
+	movl	4(%ecx), %ecx
+	movl	%edx, 0xdeadbeef(%eax)
+	movl	%ecx, 0xdeadbeef(%eax)
+	ret
+GLOBL( _x86_MultiTexCoord2fvARB_end )
+/*
+	MultiTexCoord2fvARB
+*/
+
+GLOBL( _x86_MultiTexCoord2fvARB_2 )
+	movl	4(%esp,1), %eax
+	movl	8(%esp,1), %ecx
+	sub	$0x84c0, %eax
+	and	$0x1, %eax
+	movl	0(,%eax,4), %edx
+	movl	(%ecx), %eax
+	movl	%eax, (%edx)
+	movl	4(%ecx), %eax
+	movl	%eax, 4(%edx)
+	ret
+
+GLOBL( _x86_MultiTexCoord2fvARB_2_end )
+
+/*
+	MultiTexCoord2fARB st0/st1
+*/
+GLOBL( _x86_MultiTexCoord2fARB )
+	movl	4(%esp), %eax
+	movl	8(%esp), %edx
+	sub	$0x84c0, %eax
+	movl	12(%esp), %ecx
+	and	$1, %eax
+	shl	$3, %eax
+	movl	%edx, 0xdeadbeef(%eax)
+	movl	%ecx, 0xdeadbeef(%eax)
+	ret
+GLOBL( _x86_MultiTexCoord2fARB_end )
+
+/*
+	MultiTexCoord2fARB
+*/
+GLOBL( _x86_MultiTexCoord2fARB_2 )
+	movl	4(%esp), %eax
+	movl	8(%esp), %edx
+	sub	$0x84c0, %eax
+	movl	12(%esp,1), %ecx
+	and	$1,%eax
+	movl	0(,%eax,4), %eax
+	movl	%edx, (%eax)
+	movl	%ecx, 4(%eax)
+	ret
+GLOBL( _x86_MultiTexCoord2fARB_2_end )