fix GL_DOT3_RGBA texture combiner mode in generated fragment programs (bug #11030)
[mesa.git] / src / mesa / swrast / s_context.c
index 94b57e4bbfd5432d088cc3a89da8668a64c6cc89..791850cb5028fa7c55d2917f96473a23796e6149 100644 (file)
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.1
+ * Version:  7.1
  *
- * Copyright (C) 1999-2004  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
 #include "context.h"
 #include "colormac.h"
 #include "mtypes.h"
-#include "program.h"
-#include "texobj.h"
-#include "nvfragprog.h"
-
+#include "prog_statevars.h"
+#include "teximage.h"
 #include "swrast.h"
 #include "s_blend.h"
 #include "s_context.h"
 #include "s_points.h"
 #include "s_span.h"
 #include "s_triangle.h"
-#include "s_texture.h"
+#include "s_texfilter.h"
 
 
-/*
+/**
  * Recompute the value of swrast->_RasterMask, etc. according to
- * the current context.
+ * the current context.  The _RasterMask field can be easily tested by
+ * drivers to determine certain basic GL state (does the primitive need
+ * stenciling, logic-op, fog, etc?).
  */
 static void
 _swrast_update_rasterflags( GLcontext *ctx )
 {
-   GLuint RasterMask = 0;
-
-   if (ctx->Color.AlphaEnabled)           RasterMask |= ALPHATEST_BIT;
-   if (ctx->Color.BlendEnabled)           RasterMask |= BLEND_BIT;
-   if (ctx->Depth.Test)                   RasterMask |= DEPTH_BIT;
-   if (ctx->Fog.Enabled)                  RasterMask |= FOG_BIT;
-   if (ctx->Scissor.Enabled)              RasterMask |= CLIP_BIT;
-   if (ctx->Stencil.Enabled)              RasterMask |= STENCIL_BIT;
+   SWcontext *swrast = SWRAST_CONTEXT(ctx);
+   GLbitfield rasterMask = 0;
+
+   if (ctx->Color.AlphaEnabled)           rasterMask |= ALPHATEST_BIT;
+   if (ctx->Color.BlendEnabled)           rasterMask |= BLEND_BIT;
+   if (ctx->Depth.Test)                   rasterMask |= DEPTH_BIT;
+   if (swrast->_FogEnabled)               rasterMask |= FOG_BIT;
+   if (ctx->Scissor.Enabled)              rasterMask |= CLIP_BIT;
+   if (ctx->Stencil.Enabled)              rasterMask |= STENCIL_BIT;
    if (ctx->Visual.rgbMode) {
       const GLuint colorMask = *((GLuint *) &ctx->Color.ColorMask);
-      if (colorMask != 0xffffffff)        RasterMask |= MASKING_BIT;
-      if (ctx->Color._LogicOpEnabled)     RasterMask |= LOGIC_OP_BIT;
-      if (ctx->Texture._EnabledUnits)     RasterMask |= TEXTURE_BIT;
+      if (colorMask != 0xffffffff)        rasterMask |= MASKING_BIT;
+      if (ctx->Color._LogicOpEnabled)     rasterMask |= LOGIC_OP_BIT;
+      if (ctx->Texture._EnabledUnits)     rasterMask |= TEXTURE_BIT;
    }
    else {
-      if (ctx->Color.IndexMask != 0xffffffff) RasterMask |= MASKING_BIT;
-      if (ctx->Color.IndexLogicOpEnabled)     RasterMask |= LOGIC_OP_BIT;
+      if (ctx->Color.IndexMask != 0xffffffff) rasterMask |= MASKING_BIT;
+      if (ctx->Color.IndexLogicOpEnabled)     rasterMask |= LOGIC_OP_BIT;
    }
 
-   if (ctx->DrawBuffer->UseSoftwareAlphaBuffers
-       && ctx->Color.ColorMask[ACOMP]
-       && ctx->Color.DrawBuffer != GL_NONE)
-      RasterMask |= ALPHABUF_BIT;
-
    if (   ctx->Viewport.X < 0
        || ctx->Viewport.X + ctx->Viewport.Width > (GLint) ctx->DrawBuffer->Width
        || ctx->Viewport.Y < 0
        || ctx->Viewport.Y + ctx->Viewport.Height > (GLint) ctx->DrawBuffer->Height) {
-      RasterMask |= CLIP_BIT;
+      rasterMask |= CLIP_BIT;
    }
 
-   if (ctx->Depth.OcclusionTest || ctx->Occlusion.Active)
-      RasterMask |= OCCLUSION_BIT;
+   if (ctx->Query.CurrentOcclusionObject)
+      rasterMask |= OCCLUSION_BIT;
 
 
    /* If we're not drawing to exactly one color buffer set the
     * MULTI_DRAW_BIT flag.  Also set it if we're drawing to no
     * buffers or the RGBA or CI mask disables all writes.
     */
-   if (_mesa_bitcount(ctx->Color._DrawDestMask) != 1) {
+   if (ctx->DrawBuffer->_NumColorDrawBuffers[0] != 1) {
       /* more than one color buffer designated for writing (or zero buffers) */
-      RasterMask |= MULTI_DRAW_BIT;
+      rasterMask |= MULTI_DRAW_BIT;
    }
    else if (ctx->Visual.rgbMode && *((GLuint *) ctx->Color.ColorMask) == 0) {
-      RasterMask |= MULTI_DRAW_BIT; /* all RGBA channels disabled */
+      rasterMask |= MULTI_DRAW_BIT; /* all RGBA channels disabled */
    }
    else if (!ctx->Visual.rgbMode && ctx->Color.IndexMask==0) {
-      RasterMask |= MULTI_DRAW_BIT; /* all color index bits disabled */
+      rasterMask |= MULTI_DRAW_BIT; /* all color index bits disabled */
+   }
+
+   if (ctx->FragmentProgram._Current) {
+      rasterMask |= FRAGPROG_BIT;
    }
 
-   if (ctx->FragmentProgram.Enabled) {
-      RasterMask |= FRAGPROG_BIT;
+   if (ctx->ATIFragmentShader._Enabled) {
+      rasterMask |= ATIFRAGSHADER_BIT;
    }
 
-   SWRAST_CONTEXT(ctx)->_RasterMask = RasterMask;
+#if CHAN_TYPE == GL_FLOAT
+   if (ctx->Color.ClampFragmentColor == GL_TRUE) {
+      rasterMask |= CLAMPING_BIT;
+   }
+#endif
+
+   SWRAST_CONTEXT(ctx)->_RasterMask = rasterMask;
 }
 
 
+/**
+ * Examine polycon culls tate to compute the _BackfaceSign field.
+ * _BackfaceSign will be 0 if no culling, -1 if culling back-faces,
+ * and 1 if culling front-faces.  The Polygon FrontFace state also
+ * factors in.
+ */
 static void
 _swrast_update_polygon( GLcontext *ctx )
 {
-   GLfloat backface_sign = 1;
+   GLfloat backface_sign;
 
    if (ctx->Polygon.CullFlag) {
-      backface_sign = 1;
-      switch(ctx->Polygon.CullFaceMode) {
+      backface_sign = 1.0;
+      switch (ctx->Polygon.CullFaceMode) {
       case GL_BACK:
-        if(ctx->Polygon.FrontFace==GL_CCW)
-           backface_sign = -1;
+        if (ctx->Polygon.FrontFace == GL_CCW)
+           backface_sign = -1.0;
         break;
       case GL_FRONT:
-        if(ctx->Polygon.FrontFace!=GL_CCW)
-           backface_sign = -1;
+        if (ctx->Polygon.FrontFace != GL_CCW)
+           backface_sign = -1.0;
         break;
-      default:
       case GL_FRONT_AND_BACK:
-        backface_sign = 0;
+         /* fallthrough */
+      default:
+        backface_sign = 0.0;
         break;
       }
    }
    else {
-      backface_sign = 0;
+      backface_sign = 0.0;
    }
 
    SWRAST_CONTEXT(ctx)->_BackfaceSign = backface_sign;
 }
 
 
+/**
+ * Update the _PreferPixelFog field to indicate if we need to compute
+ * fog blend factors (from the fog coords) per-fragment.
+ */
 static void
 _swrast_update_fog_hint( GLcontext *ctx )
 {
    SWcontext *swrast = SWRAST_CONTEXT(ctx);
    swrast->_PreferPixelFog = (!swrast->AllowVertexFog ||
-                              ctx->FragmentProgram.Enabled ||
+                              ctx->FragmentProgram._Current ||
                              (ctx->Hint.Fog == GL_NICEST &&
                               swrast->AllowPixelFog));
 }
 
 
 
-/*
+/**
  * Update the swrast->_AnyTextureCombine flag.
  */
 static void
@@ -171,41 +188,77 @@ _swrast_update_texture_env( GLcontext *ctx )
 }
 
 
-/*
+/**
+ * Determine if we can defer texturing/shading until after Z/stencil
+ * testing.  This potentially allows us to skip texturing/shading for
+ * lots of fragments.
+ */
+static void
+_swrast_update_deferred_texture(GLcontext *ctx)
+{
+   SWcontext *swrast = SWRAST_CONTEXT(ctx);
+   if (ctx->Color.AlphaEnabled) {
+      /* alpha test depends on post-texture/shader colors */
+      swrast->_DeferredTexture = GL_FALSE;
+   }
+   else {
+      const struct gl_fragment_program *fprog
+         = ctx->FragmentProgram._Current;
+      if (fprog && (fprog->Base.OutputsWritten & (1 << FRAG_RESULT_DEPR))) {
+         /* Z comes from fragment program/shader */
+         swrast->_DeferredTexture = GL_FALSE;
+      }
+      else if (ctx->Query.CurrentOcclusionObject) {
+         /* occlusion query depends on shader discard/kill results */
+         swrast->_DeferredTexture = GL_FALSE;
+      }
+      else {
+         swrast->_DeferredTexture = GL_TRUE;
+      }
+   }
+}
+
+
+/**
  * Update swrast->_FogColor and swrast->_FogEnable values.
  */
 static void
 _swrast_update_fog_state( GLcontext *ctx )
 {
    SWcontext *swrast = SWRAST_CONTEXT(ctx);
+   const struct gl_fragment_program *fp = ctx->FragmentProgram._Current;
 
-   /* convert fog color to GLchan values */
-   CLAMPED_FLOAT_TO_CHAN(swrast->_FogColor[RCOMP], ctx->Fog.Color[RCOMP]);
-   CLAMPED_FLOAT_TO_CHAN(swrast->_FogColor[GCOMP], ctx->Fog.Color[GCOMP]);
-   CLAMPED_FLOAT_TO_CHAN(swrast->_FogColor[BCOMP], ctx->Fog.Color[BCOMP]);
-
-   /* determine if fog is needed */
+   /* determine if fog is needed, and if so, which fog mode */
    swrast->_FogEnabled = GL_FALSE;
-   if (ctx->Fog.Enabled) {
-      swrast->_FogEnabled = GL_TRUE;
-   }
-   else if (ctx->FragmentProgram.Enabled &&
-        ctx->FragmentProgram.Current->Base.Target == GL_FRAGMENT_PROGRAM_ARB) {
-      const struct fragment_program *p;
-      p = (struct fragment_program *) ctx->FragmentProgram.Current;
-      if (p->FogOption != GL_NONE) {
+   if (fp && fp->Base.Target == GL_FRAGMENT_PROGRAM_ARB) {
+      if (fp->FogOption != GL_NONE) {
          swrast->_FogEnabled = GL_TRUE;
+         swrast->_FogMode = fp->FogOption;
       }
    }
+   else if (ctx->Fog.Enabled) {
+      swrast->_FogEnabled = GL_TRUE;
+      swrast->_FogMode = ctx->Fog.Mode;
+   }
 }
 
 
+/**
+ * Update state for running fragment programs.  Basically, load the
+ * program parameters with current state values.
+ */
 static void
-_swrast_update_fragment_program( GLcontext *ctx )
+_swrast_update_fragment_program(GLcontext *ctx, GLbitfield newState)
 {
-   if (ctx->FragmentProgram.Enabled) {
-      struct fragment_program *program = ctx->FragmentProgram.Current;
-      _mesa_load_state_parameters(ctx, program->Parameters);
+   const struct gl_fragment_program *fp = ctx->FragmentProgram._Current;
+   if (fp) {
+#if 0
+      /* XXX Need a way to trigger the initial loading of parameters
+       * even when there's no recent state changes.
+       */
+      if (fp->Base.Parameters->StateFlags & newState)
+#endif
+         _mesa_load_state_parameters(ctx, fp->Base.Parameters);
    }
 }
 
@@ -255,7 +308,8 @@ _swrast_update_fragment_program( GLcontext *ctx )
 
 
 
-/* Stub for swrast->Triangle to select a true triangle function
+/**
+ * Stub for swrast->Triangle to select a true triangle function
  * after a state change.
  */
 static void
@@ -268,10 +322,11 @@ _swrast_validate_triangle( GLcontext *ctx,
 
    _swrast_validate_derived( ctx );
    swrast->choose_triangle( ctx );
+   ASSERT(swrast->Triangle);
 
    if (ctx->Texture._EnabledUnits == 0
        && NEED_SECONDARY_COLOR(ctx)
-       && !ctx->FragmentProgram.Enabled) {
+       && !ctx->FragmentProgram._Current) {
       /* separate specular color, but no texture */
       swrast->SpecTriangle = swrast->Triangle;
       swrast->Triangle = _swrast_add_spec_terms_triangle;
@@ -280,6 +335,10 @@ _swrast_validate_triangle( GLcontext *ctx,
    swrast->Triangle( ctx, v0, v1, v2 );
 }
 
+/**
+ * Called via swrast->Line.  Examine current GL state and choose a software
+ * line routine.  Then call it.
+ */
 static void
 _swrast_validate_line( GLcontext *ctx, const SWvertex *v0, const SWvertex *v1 )
 {
@@ -287,18 +346,22 @@ _swrast_validate_line( GLcontext *ctx, const SWvertex *v0, const SWvertex *v1 )
 
    _swrast_validate_derived( ctx );
    swrast->choose_line( ctx );
+   ASSERT(swrast->Line);
 
    if (ctx->Texture._EnabledUnits == 0
        && NEED_SECONDARY_COLOR(ctx)
-       && !ctx->FragmentProgram.Enabled) {
+       && !ctx->FragmentProgram._Current) {
       swrast->SpecLine = swrast->Line;
       swrast->Line = _swrast_add_spec_terms_line;
    }
 
-
    swrast->Line( ctx, v0, v1 );
 }
 
+/**
+ * Called via swrast->Point.  Examine current GL state and choose a software
+ * point routine.  Then call it.
+ */
 static void
 _swrast_validate_point( GLcontext *ctx, const SWvertex *v0 )
 {
@@ -309,7 +372,7 @@ _swrast_validate_point( GLcontext *ctx, const SWvertex *v0 )
 
    if (ctx->Texture._EnabledUnits == 0
        && NEED_SECONDARY_COLOR(ctx)
-       && !ctx->FragmentProgram.Enabled) {
+       && !ctx->FragmentProgram._Current) {
       swrast->SpecPoint = swrast->Point;
       swrast->Point = _swrast_add_spec_terms_point;
    }
@@ -318,59 +381,110 @@ _swrast_validate_point( GLcontext *ctx, const SWvertex *v0 )
 }
 
 
+/**
+ * Called via swrast->BlendFunc.  Examine GL state to choose a blending
+ * function, then call it.
+ */
 static void _ASMAPI
-_swrast_validate_blend_func( GLcontext *ctx, GLuint n,
-                            const GLubyte mask[],
-                            GLchan src[][4],
-                            CONST GLchan dst[][4] )
+_swrast_validate_blend_func(GLcontext *ctx, GLuint n, const GLubyte mask[],
+                            GLvoid *src, const GLvoid *dst,
+                            GLenum chanType )
 {
    SWcontext *swrast = SWRAST_CONTEXT(ctx);
 
-   _swrast_validate_derived( ctx );
-   _swrast_choose_blend_func( ctx );
+   _swrast_validate_derived( ctx ); /* why is this needed? */
+   _swrast_choose_blend_func( ctx, chanType );
 
-   swrast->BlendFunc( ctx, n, mask, src, dst );
+   swrast->BlendFunc( ctx, n, mask, src, dst, chanType );
 }
 
 
+/**
+ * Make sure we have texture image data for all the textures we may need
+ * for subsequent rendering.
+ */
 static void
-_swrast_validate_texture_sample( GLcontext *ctx, GLuint texUnit,
-                                const struct gl_texture_object *tObj,
-                                GLuint n, const GLfloat texcoords[][4],
-                                const GLfloat lambda[], GLchan rgba[][4] )
+_swrast_validate_texture_images(GLcontext *ctx)
 {
    SWcontext *swrast = SWRAST_CONTEXT(ctx);
-
-   _swrast_validate_derived( ctx );
-
-   /* Compute min/mag filter threshold */
-   if (tObj->MinFilter != tObj->MagFilter) {
-      if (tObj->MagFilter == GL_LINEAR
-          && (tObj->MinFilter == GL_NEAREST_MIPMAP_NEAREST ||
-              tObj->MinFilter == GL_NEAREST_MIPMAP_LINEAR)) {
-         swrast->_MinMagThresh[texUnit] = 0.5F;
-      }
-      else {
-         swrast->_MinMagThresh[texUnit] = 0.0F;
+   GLuint u;
+
+   if (!swrast->ValidateTextureImage || !ctx->Texture._EnabledUnits) {
+      /* no textures enabled, or no way to validate images! */
+      return;
+   }
+
+   for (u = 0; u < ctx->Const.MaxTextureImageUnits; u++) {
+      if (ctx->Texture.Unit[u]._ReallyEnabled) {
+         struct gl_texture_object *texObj = ctx->Texture.Unit[u]._Current;
+         ASSERT(texObj);
+         if (texObj) {
+            GLuint numFaces = (texObj->Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1;
+            GLuint face;
+            for (face = 0; face < numFaces; face++) {
+               GLint lvl;
+               for (lvl = texObj->BaseLevel; lvl <= texObj->_MaxLevel; lvl++) {
+                  struct gl_texture_image *texImg = texObj->Image[face][lvl];
+                  if (texImg && !texImg->Data) {
+                     swrast->ValidateTextureImage(ctx, texObj, face, lvl);
+                     ASSERT(texObj->Image[face][lvl]->Data);
+                  }
+               }
+            }
+         }
       }
    }
+}
 
-   swrast->TextureSample[texUnit] =
-      _swrast_choose_texture_sample_func( ctx, tObj );
 
-   swrast->TextureSample[texUnit]( ctx, texUnit, tObj, n, texcoords,
-                                   lambda, rgba );
+/**
+ * Free the texture image data attached to all currently enabled
+ * textures.  Meant to be called by device drivers when transitioning
+ * from software to hardware rendering.
+ */
+void
+_swrast_eject_texture_images(GLcontext *ctx)
+{
+   GLuint u;
+
+   if (!ctx->Texture._EnabledUnits) {
+      /* no textures enabled */
+      return;
+   }
+
+   for (u = 0; u < ctx->Const.MaxTextureImageUnits; u++) {
+      if (ctx->Texture.Unit[u]._ReallyEnabled) {
+         struct gl_texture_object *texObj = ctx->Texture.Unit[u]._Current;
+         ASSERT(texObj);
+         if (texObj) {
+            GLuint numFaces = (texObj->Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1;
+            GLuint face;
+            for (face = 0; face < numFaces; face++) {
+               GLint lvl;
+               for (lvl = texObj->BaseLevel; lvl <= texObj->_MaxLevel; lvl++) {
+                  struct gl_texture_image *texImg = texObj->Image[face][lvl];
+                  if (texImg && texImg->Data) {
+                     _mesa_free_texmemory(texImg->Data);
+                     texImg->Data = NULL;
+                  }
+               }
+            }
+         }
+      }
+   }
 }
 
 
+
 static void
-_swrast_sleep( GLcontext *ctx, GLuint new_state )
+_swrast_sleep( GLcontext *ctx, GLbitfield new_state )
 {
+   (void) ctx; (void) new_state;
 }
 
 
 static void
-_swrast_invalidate_state( GLcontext *ctx, GLuint new_state )
+_swrast_invalidate_state( GLcontext *ctx, GLbitfield new_state )
 {
    SWcontext *swrast = SWRAST_CONTEXT(ctx);
    GLuint i;
@@ -386,39 +500,135 @@ _swrast_invalidate_state( GLcontext *ctx, GLuint new_state )
       new_state = ~0;
    }
 
-   if (new_state & swrast->invalidate_triangle)
+   if (new_state & swrast->InvalidateTriangleMask)
       swrast->Triangle = _swrast_validate_triangle;
 
-   if (new_state & swrast->invalidate_line)
+   if (new_state & swrast->InvalidateLineMask)
       swrast->Line = _swrast_validate_line;
 
-   if (new_state & swrast->invalidate_point)
+   if (new_state & swrast->InvalidatePointMask)
       swrast->Point = _swrast_validate_point;
 
    if (new_state & _SWRAST_NEW_BLEND_FUNC)
       swrast->BlendFunc = _swrast_validate_blend_func;
 
    if (new_state & _SWRAST_NEW_TEXTURE_SAMPLE_FUNC)
-      for (i = 0 ; i < ctx->Const.MaxTextureUnits ; i++)
-        swrast->TextureSample[i] = _swrast_validate_texture_sample;
+      for (i = 0 ; i < ctx->Const.MaxTextureImageUnits ; i++)
+        swrast->TextureSample[i] = NULL;
+}
 
-   if (ctx->Visual.rgbMode) {
-      ASSERT(swrast->Driver.WriteRGBASpan);
-      ASSERT(swrast->Driver.WriteRGBSpan);
-      ASSERT(swrast->Driver.WriteMonoRGBASpan);
-      ASSERT(swrast->Driver.WriteRGBAPixels);
-      ASSERT(swrast->Driver.WriteMonoRGBAPixels);
-      ASSERT(swrast->Driver.ReadRGBASpan);
-      ASSERT(swrast->Driver.ReadRGBAPixels);
+
+void
+_swrast_update_texture_samplers(GLcontext *ctx)
+{
+   SWcontext *swrast = SWRAST_CONTEXT(ctx);
+   GLuint u;
+
+   for (u = 0; u < ctx->Const.MaxTextureImageUnits; u++) {
+      const struct gl_texture_object *tObj = ctx->Texture.Unit[u]._Current;
+      /* Note: If tObj is NULL, the sample function will be a simple
+       * function that just returns opaque black (0,0,0,1).
+       */
+      swrast->TextureSample[u] = _swrast_choose_texture_sample_func(ctx, tObj);
+   }
+}
+
+
+/**
+ * Update swrast->_ActiveAttribs, swrast->_NumActiveAttribs,
+ * swrast->_ActiveAtttribMask.
+ */
+static void
+_swrast_update_active_attribs(GLcontext *ctx)
+{
+   SWcontext *swrast = SWRAST_CONTEXT(ctx);
+   GLuint attribsMask;
+
+   /*
+    * Compute _ActiveAttribsMask = which fragment attributes are needed.
+    */
+   if (ctx->FragmentProgram._Current) {
+      /* fragment program/shader */
+      attribsMask = ctx->FragmentProgram._Current->Base.InputsRead;
+      attribsMask &= ~FRAG_BIT_WPOS; /* WPOS is always handled specially */
+   }
+   else if (ctx->ATIFragmentShader._Enabled) {
+      attribsMask = ~0;  /* XXX fix me */
    }
    else {
-      ASSERT(swrast->Driver.WriteCI32Span);
-      ASSERT(swrast->Driver.WriteCI8Span);
-      ASSERT(swrast->Driver.WriteMonoCISpan);
-      ASSERT(swrast->Driver.WriteCI32Pixels);
-      ASSERT(swrast->Driver.WriteMonoCIPixels);
-      ASSERT(swrast->Driver.ReadCI32Span);
-      ASSERT(swrast->Driver.ReadCI32Pixels);
+      /* fixed function */
+      attribsMask = 0x0;
+
+#if CHAN_TYPE == GL_FLOAT
+      attribsMask |= FRAG_BIT_COL0;
+#endif
+
+      if (ctx->Fog.ColorSumEnabled ||
+          (ctx->Light.Enabled &&
+           ctx->Light.Model.ColorControl == GL_SEPARATE_SPECULAR_COLOR)) {
+         attribsMask |= FRAG_BIT_COL1;
+      }
+
+      if (swrast->_FogEnabled)
+         attribsMask |= FRAG_BIT_FOGC;
+
+      attribsMask |= (ctx->Texture._EnabledUnits << FRAG_ATTRIB_TEX0);
+   }
+
+   swrast->_ActiveAttribMask = attribsMask;
+
+   /* Update _ActiveAttribs[] list */
+   {
+      GLuint i, num = 0;
+      for (i = 0; i < FRAG_ATTRIB_MAX; i++) {
+         if (attribsMask & (1 << i)) {
+            swrast->_ActiveAttribs[num++] = i;
+            /* how should this attribute be interpolated? */
+            if (i == FRAG_ATTRIB_COL0 || i == FRAG_ATTRIB_COL1)
+               swrast->_InterpMode[i] = ctx->Light.ShadeModel;
+            else
+               swrast->_InterpMode[i] = GL_SMOOTH;
+         }
+      }
+      swrast->_NumActiveAttribs = num;
+   }
+}
+
+
+/**
+ * Update the swrast->_ColorOutputsMask which indicates which color
+ * renderbuffers (aka rendertargets) are being written to by the current
+ * fragment program.
+ * We also take glDrawBuffers() into account to skip outputs that are
+ * set to GL_NONE.
+ */
+static void
+_swrast_update_color_outputs(GLcontext *ctx)
+{
+   SWcontext *swrast = SWRAST_CONTEXT(ctx);
+   const struct gl_framebuffer *fb = ctx->DrawBuffer;
+
+   swrast->_ColorOutputsMask = 0;
+   swrast->_NumColorOutputs = 0;
+
+   if (ctx->FragmentProgram._Current) {
+      const GLbitfield outputsWritten
+         = ctx->FragmentProgram._Current->Base.OutputsWritten;
+      GLuint output;
+      for (output = 0; output < ctx->Const.MaxDrawBuffers; output++) {
+         if ((outputsWritten & (1 << (FRAG_RESULT_DATA0 + output)))
+             && (fb->_NumColorDrawBuffers[output] > 0)) {
+            swrast->_ColorOutputsMask |= (1 << output);
+            swrast->_NumColorOutputs = output + 1;
+         }
+      }
+   }
+   if (swrast->_ColorOutputsMask == 0x0) {
+      /* no fragment program, or frag prog didn't write to gl_FragData[] */
+      if (fb->_NumColorDrawBuffers[0] > 0) {
+         swrast->_ColorOutputsMask = 0x1;
+         swrast->_NumColorOutputs = 1;
+      }
    }
 }
 
@@ -429,9 +639,6 @@ _swrast_validate_derived( GLcontext *ctx )
    SWcontext *swrast = SWRAST_CONTEXT(ctx);
 
    if (swrast->NewState) {
-      if (swrast->NewState & _SWRAST_NEW_RASTERMASK)
-        _swrast_update_rasterflags( ctx );
-
       if (swrast->NewState & _NEW_POLYGON)
         _swrast_update_polygon( ctx );
 
@@ -441,11 +648,42 @@ _swrast_validate_derived( GLcontext *ctx )
       if (swrast->NewState & _SWRAST_NEW_TEXTURE_ENV_MODE)
         _swrast_update_texture_env( ctx );
 
-      if (swrast->NewState & _NEW_FOG)
+      if (swrast->NewState & (_NEW_FOG | _NEW_PROGRAM))
          _swrast_update_fog_state( ctx );
 
-      if (swrast->NewState & _NEW_PROGRAM)
-        _swrast_update_fragment_program( ctx );
+      if (swrast->NewState & (_NEW_MODELVIEW |
+                              _NEW_PROJECTION |
+                              _NEW_TEXTURE_MATRIX |
+                              _NEW_FOG |
+                              _NEW_LIGHT |
+                              _NEW_LINE |
+                              _NEW_TEXTURE |
+                              _NEW_TRANSFORM |
+                              _NEW_POINT |
+                              _NEW_VIEWPORT |
+                              _NEW_PROGRAM))
+        _swrast_update_fragment_program( ctx, swrast->NewState );
+
+      if (swrast->NewState & (_NEW_TEXTURE | _NEW_PROGRAM)) {
+         _swrast_update_texture_samplers( ctx );
+         _swrast_validate_texture_images(ctx);
+      }
+
+      if (swrast->NewState & (_NEW_COLOR | _NEW_PROGRAM))
+         _swrast_update_deferred_texture(ctx);
+
+      if (swrast->NewState & _SWRAST_NEW_RASTERMASK)
+        _swrast_update_rasterflags( ctx );
+
+      if (swrast->NewState & (_NEW_DEPTH |
+                              _NEW_FOG |
+                              _NEW_LIGHT |
+                              _NEW_PROGRAM |
+                              _NEW_TEXTURE))
+         _swrast_update_active_attribs(ctx);
+
+      if (swrast->NewState & (_NEW_PROGRAM | _NEW_BUFFERS))
+         _swrast_update_color_outputs(ctx);
 
       swrast->NewState = 0;
       swrast->StateChanges = 0;
@@ -508,7 +746,7 @@ _swrast_Point( GLcontext *ctx, const SWvertex *v0 )
 }
 
 void
-_swrast_InvalidateState( GLcontext *ctx, GLuint new_state )
+_swrast_InvalidateState( GLcontext *ctx, GLbitfield new_state )
 {
    if (SWRAST_DEBUG) {
       _mesa_debug(ctx, "_swrast_InvalidateState\n");
@@ -565,9 +803,9 @@ _swrast_CreateContext( GLcontext *ctx )
    swrast->choose_line = _swrast_choose_line;
    swrast->choose_triangle = _swrast_choose_triangle;
 
-   swrast->invalidate_point = _SWRAST_NEW_POINT;
-   swrast->invalidate_line = _SWRAST_NEW_LINE;
-   swrast->invalidate_triangle = _SWRAST_NEW_TRIANGLE;
+   swrast->InvalidatePointMask = _SWRAST_NEW_POINT;
+   swrast->InvalidateLineMask = _SWRAST_NEW_LINE;
+   swrast->InvalidateTriangleMask = _SWRAST_NEW_TRIANGLE;
 
    swrast->Point = _swrast_validate_point;
    swrast->Line = _swrast_validate_line;
@@ -578,35 +816,34 @@ _swrast_CreateContext( GLcontext *ctx )
    swrast->AllowVertexFog = GL_TRUE;
    swrast->AllowPixelFog = GL_TRUE;
 
-   if (ctx->Visual.doubleBufferMode)
-      swrast->CurrentBufferBit = DD_BACK_LEFT_BIT;
-   else
-      swrast->CurrentBufferBit = DD_FRONT_LEFT_BIT;
-
    /* Optimized Accum buffer */
-   swrast->_IntegerAccumMode = GL_TRUE;
+   swrast->_IntegerAccumMode = GL_FALSE;
    swrast->_IntegerAccumScaler = 0.0;
 
    for (i = 0; i < MAX_TEXTURE_IMAGE_UNITS; i++)
-      swrast->TextureSample[i] = _swrast_validate_texture_sample;
+      swrast->TextureSample[i] = NULL;
 
-   swrast->SpanArrays = MALLOC_STRUCT(span_arrays);
+   swrast->SpanArrays = MALLOC_STRUCT(sw_span_arrays);
    if (!swrast->SpanArrays) {
       FREE(swrast);
       return GL_FALSE;
    }
+   swrast->SpanArrays->ChanType = CHAN_TYPE;
+#if CHAN_TYPE == GL_UNSIGNED_BYTE
+   swrast->SpanArrays->rgba = swrast->SpanArrays->rgba8;
+#elif CHAN_TYPE == GL_UNSIGNED_SHORT
+   swrast->SpanArrays->rgba = swrast->SpanArrays->rgba16;
+#else
+   swrast->SpanArrays->rgba = swrast->SpanArrays->attribs[FRAG_ATTRIB_COL0];
+#endif
 
    /* init point span buffer */
    swrast->PointSpan.primitive = GL_POINT;
-   swrast->PointSpan.start = 0;
    swrast->PointSpan.end = 0;
    swrast->PointSpan.facing = 0;
    swrast->PointSpan.array = swrast->SpanArrays;
 
-   assert(ctx->Const.MaxTextureUnits > 0);
-   assert(ctx->Const.MaxTextureUnits <= MAX_TEXTURE_UNITS);
-
-   swrast->TexelBuffer = (GLchan *) MALLOC(ctx->Const.MaxTextureUnits *
+   swrast->TexelBuffer = (GLchan *) MALLOC(ctx->Const.MaxTextureImageUnits *
                                            MAX_WIDTH * 4 * sizeof(GLchan));
    if (!swrast->TexelBuffer) {
       FREE(swrast->SpanArrays);
@@ -650,10 +887,7 @@ _swrast_flush( GLcontext *ctx )
    /* flush any pending fragments from rendering points */
    if (swrast->PointSpan.end > 0) {
       if (ctx->Visual.rgbMode) {
-         if (ctx->Texture._EnabledCoordUnits)
-            _swrast_write_texture_span(ctx, &(swrast->PointSpan));
-         else
-            _swrast_write_rgba_span(ctx, &(swrast->PointSpan));
+         _swrast_write_rgba_span(ctx, &(swrast->PointSpan));
       }
       else {
          _swrast_write_index_span(ctx, &(swrast->PointSpan));
@@ -702,52 +936,34 @@ _swrast_print_vertex( GLcontext *ctx, const SWvertex *v )
 
    if (SWRAST_DEBUG_VERTICES) {
       _mesa_debug(ctx, "win %f %f %f %f\n",
-                  v->win[0], v->win[1], v->win[2], v->win[3]);
+                  v->attrib[FRAG_ATTRIB_WPOS][0],
+                  v->attrib[FRAG_ATTRIB_WPOS][1],
+                  v->attrib[FRAG_ATTRIB_WPOS][2],
+                  v->attrib[FRAG_ATTRIB_WPOS][3]);
 
-      for (i = 0 ; i < ctx->Const.MaxTextureUnits ; i++)
+      for (i = 0 ; i < ctx->Const.MaxTextureCoordUnits ; i++)
         if (ctx->Texture.Unit[i]._ReallyEnabled)
            _mesa_debug(ctx, "texcoord[%d] %f %f %f %f\n", i,
-                        v->texcoord[i][0], v->texcoord[i][1],
-                        v->texcoord[i][2], v->texcoord[i][3]);
+                        v->attrib[FRAG_ATTRIB_TEX0 + i][0],
+                        v->attrib[FRAG_ATTRIB_TEX0 + i][1],
+                        v->attrib[FRAG_ATTRIB_TEX0 + i][2],
+                        v->attrib[FRAG_ATTRIB_TEX0 + i][3]);
 
 #if CHAN_TYPE == GL_FLOAT
       _mesa_debug(ctx, "color %f %f %f %f\n",
                   v->color[0], v->color[1], v->color[2], v->color[3]);
-      _mesa_debug(ctx, "spec %f %f %f %f\n",
-                  v->specular[0], v->specular[1],
-                  v->specular[2], v->specular[3]);
 #else
       _mesa_debug(ctx, "color %d %d %d %d\n",
                   v->color[0], v->color[1], v->color[2], v->color[3]);
-      _mesa_debug(ctx, "spec %d %d %d %d\n",
-                  v->specular[0], v->specular[1],
-                  v->specular[2], v->specular[3]);
 #endif
-      _mesa_debug(ctx, "fog %f\n", v->fog);
-      _mesa_debug(ctx, "index %d\n", v->index);
+      _mesa_debug(ctx, "spec %g %g %g %g\n",
+                  v->attrib[FRAG_ATTRIB_COL1][0],
+                  v->attrib[FRAG_ATTRIB_COL1][1],
+                  v->attrib[FRAG_ATTRIB_COL1][2],
+                  v->attrib[FRAG_ATTRIB_COL1][3]);
+      _mesa_debug(ctx, "fog %f\n", v->attrib[FRAG_ATTRIB_FOGC][0]);
+      _mesa_debug(ctx, "index %d\n", v->attrib[FRAG_ATTRIB_CI][0]);
       _mesa_debug(ctx, "pointsize %f\n", v->pointSize);
       _mesa_debug(ctx, "\n");
    }
 }
-
-
-/**
- * Validate access to a PBO to be sure we're not going to read/write
- * out of buffer bounds.
- */
-GLvoid *
-_swrast_validate_pbo_access(const struct gl_pixelstore_attrib *pack,
-                            GLsizei width, GLsizei height, GLsizei depth,
-                            GLenum format, GLenum type, GLvoid *ptr)
-{
-   if (pack->BufferObj->Name == 0) {
-      /* no PBO */
-      return ptr;
-   }
-   else if (_mesa_validate_pbo_access(pack, width, height, depth, format,
-                                      type, ptr)) {
-      return ADD_POINTERS(pack->BufferObj->Data, ptr);
-   }
-   /* bad access! */
-   return NULL;
-}