mesa: Prefix main includes with dir to avoid conflicts.
[mesa.git] / src / mesa / swrast / s_span.c
index 973ef3e3e4dab27bda97ae71853391b3a998d1f2..457dc4a6ce22925a19c9dd9436adb779522669a3 100644 (file)
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5
+ * Version:  7.1
  *
- * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
  * \author Brian Paul
  */
 
-#include "glheader.h"
-#include "colormac.h"
-#include "context.h"
-#include "macros.h"
-#include "imports.h"
+#include "main/glheader.h"
+#include "main/colormac.h"
+#include "main/context.h"
+#include "main/macros.h"
+#include "main/imports.h"
 #include "image.h"
 
 #include "s_atifragshader.h"
 #include "s_alpha.h"
-#include "s_arbshader.h"
 #include "s_blend.h"
 #include "s_context.h"
 #include "s_depth.h"
 #include "s_fog.h"
 #include "s_logic.h"
 #include "s_masking.h"
-#include "s_nvfragprog.h"
+#include "s_fragprog.h"
 #include "s_span.h"
 #include "s_stencil.h"
 #include "s_texcombine.h"
 
 
 /**
- * Init span's Z interpolation values to the RasterPos Z.
- * Used during setup for glDraw/CopyPixels.
+ * Set default fragment attributes for the span using the
+ * current raster values.  Used prior to glDraw/CopyPixels
+ * and glBitmap.
  */
 void
-_swrast_span_default_z( GLcontext *ctx, SWspan *span )
+_swrast_span_default_attribs(GLcontext *ctx, SWspan *span)
 {
-   const GLfloat depthMax = ctx->DrawBuffer->_DepthMaxF;
-   if (ctx->DrawBuffer->Visual.depthBits <= 16)
-      span->z = FloatToFixed(ctx->Current.RasterPos[2] * depthMax + 0.5F);
-   else
-      span->z = (GLint) (ctx->Current.RasterPos[2] * depthMax + 0.5F);
-   span->zStep = 0;
-   span->interpMask |= SPAN_Z;
-}
-
-
-/**
- * Init span's fog interpolation values to the RasterPos fog.
- * Used during setup for glDraw/CopyPixels.
- */
-void
-_swrast_span_default_fog( GLcontext *ctx, SWspan *span )
-{
-   span->fog = _swrast_z_to_fogfactor(ctx, ctx->Current.RasterDistance);
-   span->fogStep = span->dfogdx = span->dfogdy = 0.0F;
-   span->interpMask |= SPAN_FOG;
-}
+   /* Z*/
+   {
+      const GLfloat depthMax = ctx->DrawBuffer->_DepthMaxF;
+      if (ctx->DrawBuffer->Visual.depthBits <= 16)
+         span->z = FloatToFixed(ctx->Current.RasterPos[2] * depthMax + 0.5F);
+      else
+         span->z = (GLint) (ctx->Current.RasterPos[2] * depthMax + 0.5F);
+      span->zStep = 0;
+      span->interpMask |= SPAN_Z;
+   }
 
+   /* W (for perspective correction) */
+   span->attrStart[FRAG_ATTRIB_WPOS][3] = 1.0;
+   span->attrStepX[FRAG_ATTRIB_WPOS][3] = 0.0;
+   span->attrStepY[FRAG_ATTRIB_WPOS][3] = 0.0;
 
-/**
- * Init span's rgba or index interpolation values to the RasterPos color.
- * Used during setup for glDraw/CopyPixels.
- */
-void
-_swrast_span_default_color( GLcontext *ctx, SWspan *span )
-{
+   /* primary color, or color index */
    if (ctx->Visual.rgbMode) {
       GLchan r, g, b, a;
       UNCLAMPED_FLOAT_TO_CHAN(r, ctx->Current.RasterColor[0]);
@@ -111,62 +98,136 @@ _swrast_span_default_color( GLcontext *ctx, SWspan *span )
       span->blueStep = 0;
       span->alphaStep = 0;
       span->interpMask |= SPAN_RGBA;
+
+      COPY_4V(span->attrStart[FRAG_ATTRIB_COL0], ctx->Current.RasterColor);
+      ASSIGN_4V(span->attrStepX[FRAG_ATTRIB_COL0], 0.0, 0.0, 0.0, 0.0);
+      ASSIGN_4V(span->attrStepY[FRAG_ATTRIB_COL0], 0.0, 0.0, 0.0, 0.0);
    }
    else {
       span->index = FloatToFixed(ctx->Current.RasterIndex);
       span->indexStep = 0;
       span->interpMask |= SPAN_INDEX;
    }
+
+   /* Secondary color */
+   if (ctx->Visual.rgbMode && (ctx->Light.Enabled || ctx->Fog.ColorSumEnabled))
+   {
+      COPY_4V(span->attrStart[FRAG_ATTRIB_COL1], ctx->Current.RasterSecondaryColor);
+      ASSIGN_4V(span->attrStepX[FRAG_ATTRIB_COL1], 0.0, 0.0, 0.0, 0.0);
+      ASSIGN_4V(span->attrStepY[FRAG_ATTRIB_COL1], 0.0, 0.0, 0.0, 0.0);
+   }
+
+   /* fog */
+   {
+      const SWcontext *swrast = SWRAST_CONTEXT(ctx);
+      GLfloat fogVal; /* a coord or a blend factor */
+      if (swrast->_PreferPixelFog) {
+         /* fog blend factors will be computed from fog coordinates per pixel */
+         fogVal = ctx->Current.RasterDistance;
+      }
+      else {
+         /* fog blend factor should be computed from fogcoord now */
+         fogVal = _swrast_z_to_fogfactor(ctx, ctx->Current.RasterDistance);
+      }
+      span->attrStart[FRAG_ATTRIB_FOGC][0] = fogVal;
+      span->attrStepX[FRAG_ATTRIB_FOGC][0] = 0.0;
+      span->attrStepY[FRAG_ATTRIB_FOGC][0] = 0.0;
+   }
+
+   /* texcoords */
+   {
+      GLuint i;
+      for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
+         const GLuint attr = FRAG_ATTRIB_TEX0 + i;
+         const GLfloat *tc = ctx->Current.RasterTexCoords[i];
+         if (ctx->FragmentProgram._Current || ctx->ATIFragmentShader._Enabled) {
+            COPY_4V(span->attrStart[attr], tc);
+         }
+         else if (tc[3] > 0.0F) {
+            /* use (s/q, t/q, r/q, 1) */
+            span->attrStart[attr][0] = tc[0] / tc[3];
+            span->attrStart[attr][1] = tc[1] / tc[3];
+            span->attrStart[attr][2] = tc[2] / tc[3];
+            span->attrStart[attr][3] = 1.0;
+         }
+         else {
+            ASSIGN_4V(span->attrStart[attr], 0.0F, 0.0F, 0.0F, 1.0F);
+         }
+         ASSIGN_4V(span->attrStepX[attr], 0.0F, 0.0F, 0.0F, 0.0F);
+         ASSIGN_4V(span->attrStepY[attr], 0.0F, 0.0F, 0.0F, 0.0F);
+      }
+   }
 }
 
 
 /**
- * Init span's texcoord interpolation values to the RasterPos texcoords.
- * Used during setup for glDraw/CopyPixels.
+ * Interpolate the active attributes (and'd with attrMask) to
+ * fill in span->array->attribs[].
+ * Perspective correction will be done.  The point/line/triangle function
+ * should have computed attrStart/Step values for FRAG_ATTRIB_WPOS[3]!
  */
-void
-_swrast_span_default_texcoords( GLcontext *ctx, SWspan *span )
+static INLINE void
+interpolate_active_attribs(GLcontext *ctx, SWspan *span, GLbitfield attrMask)
 {
-   GLuint i;
-   for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
-      const GLfloat *tc = ctx->Current.RasterTexCoords[i];
-      if (ctx->FragmentProgram._Enabled || ctx->ATIFragmentShader._Enabled) {
-         COPY_4V(span->tex[i], tc);
-      }
-      else if (tc[3] > 0.0F) {
-         /* use (s/q, t/q, r/q, 1) */
-         span->tex[i][0] = tc[0] / tc[3];
-         span->tex[i][1] = tc[1] / tc[3];
-         span->tex[i][2] = tc[2] / tc[3];
-         span->tex[i][3] = 1.0;
-      }
-      else {
-         ASSIGN_4V(span->tex[i], 0.0F, 0.0F, 0.0F, 1.0F);
+   const SWcontext *swrast = SWRAST_CONTEXT(ctx);
+
+   /*
+    * Don't overwrite existing array values, such as colors that may have
+    * been produced by glDraw/CopyPixels.
+    */
+   attrMask &= ~span->arrayAttribs;
+
+   ATTRIB_LOOP_BEGIN
+      if (attrMask & (1 << attr)) {
+         const GLfloat dwdx = span->attrStepX[FRAG_ATTRIB_WPOS][3];
+         GLfloat w = span->attrStart[FRAG_ATTRIB_WPOS][3];
+         const GLfloat dv0dx = span->attrStepX[attr][0];
+         const GLfloat dv1dx = span->attrStepX[attr][1];
+         const GLfloat dv2dx = span->attrStepX[attr][2];
+         const GLfloat dv3dx = span->attrStepX[attr][3];
+         GLfloat v0 = span->attrStart[attr][0];
+         GLfloat v1 = span->attrStart[attr][1];
+         GLfloat v2 = span->attrStart[attr][2];
+         GLfloat v3 = span->attrStart[attr][3];
+         GLuint k;
+         for (k = 0; k < span->end; k++) {
+            const GLfloat invW = 1.0f / w;
+            span->array->attribs[attr][k][0] = v0 * invW;
+            span->array->attribs[attr][k][1] = v1 * invW;
+            span->array->attribs[attr][k][2] = v2 * invW;
+            span->array->attribs[attr][k][3] = v3 * invW;
+            v0 += dv0dx;
+            v1 += dv1dx;
+            v2 += dv2dx;
+            v3 += dv3dx;
+            w += dwdx;
+         }
+         ASSERT((span->arrayAttribs & (1 << attr)) == 0);
+         span->arrayAttribs |= (1 << attr);
       }
-      ASSIGN_4V(span->texStepX[i], 0.0F, 0.0F, 0.0F, 0.0F);
-      ASSIGN_4V(span->texStepY[i], 0.0F, 0.0F, 0.0F, 0.0F);
-   }
-   span->interpMask |= SPAN_TEXTURE;
+   ATTRIB_LOOP_END
 }
 
 
 /**
- * Interpolate primary colors to fill in the span->array->color array.
+ * Interpolate primary colors to fill in the span->array->rgba8 (or rgb16)
+ * color array.
  */
-static void
-interpolate_colors(SWspan *span)
+static INLINE void
+interpolate_int_colors(GLcontext *ctx, SWspan *span)
 {
    const GLuint n = span->end;
    GLuint i;
 
-   ASSERT((span->interpMask & SPAN_RGBA)  &&
-          !(span->arrayMask & SPAN_RGBA));
+#if CHAN_BITS != 32
+   ASSERT(!(span->arrayMask & SPAN_RGBA));
+#endif
 
    switch (span->array->ChanType) {
 #if CHAN_BITS != 32
    case GL_UNSIGNED_BYTE:
       {
-         GLubyte (*rgba)[4] = span->array->color.sz1.rgba;
+         GLubyte (*rgba)[4] = span->array->rgba8;
          if (span->interpMask & SPAN_FLAT) {
             GLubyte color[4];
             color[RCOMP] = FixedToInt(span->red);
@@ -201,7 +262,7 @@ interpolate_colors(SWspan *span)
       break;
    case GL_UNSIGNED_SHORT:
       {
-         GLushort (*rgba)[4] = span->array->color.sz2.rgba;
+         GLushort (*rgba)[4] = span->array->rgba16;
          if (span->interpMask & SPAN_FLAT) {
             GLushort color[4];
             color[RCOMP] = FixedToInt(span->red);
@@ -213,7 +274,7 @@ interpolate_colors(SWspan *span)
             }
          }
          else {
-            GLushort (*rgba)[4] = span->array->color.sz2.rgba;
+            GLushort (*rgba)[4] = span->array->rgba16;
             GLfixed r, g, b, a;
             GLint dr, dg, db, da;
             r = span->red;
@@ -239,158 +300,78 @@ interpolate_colors(SWspan *span)
       break;
 #endif
    case GL_FLOAT:
-      {
-         GLfloat (*rgba)[4] = span->array->color.sz4.rgba;
-         GLfloat r, g, b, a, dr, dg, db, da;
-         r = span->red;
-         g = span->green;
-         b = span->blue;
-         a = span->alpha;
-         if (span->interpMask & SPAN_FLAT) {
-            dr = dg = db = da = 0.0;
-         }
-         else {
-            dr = span->redStep;
-            dg = span->greenStep;
-            db = span->blueStep;
-            da = span->alphaStep;
-         }
-         for (i = 0; i < n; i++) {
-            rgba[i][RCOMP] = r;
-            rgba[i][GCOMP] = g;
-            rgba[i][BCOMP] = b;
-            rgba[i][ACOMP] = a;
-            r += dr;
-            g += dg;
-            b += db;
-            a += da;
-         }
-      }
+      interpolate_active_attribs(ctx, span, FRAG_BIT_COL0);
       break;
    default:
-      _mesa_problem(NULL, "bad datatype in interpolate_colors");
+      _mesa_problem(NULL, "bad datatype in interpolate_int_colors");
    }
    span->arrayMask |= SPAN_RGBA;
 }
 
 
 /**
- * Interpolate specular/secondary colors.
+ * Populate the FRAG_ATTRIB_COL0 array.
  */
-static void
-interpolate_specular(SWspan *span)
+static INLINE void
+interpolate_float_colors(SWspan *span)
 {
+   GLfloat (*col0)[4] = span->array->attribs[FRAG_ATTRIB_COL0];
    const GLuint n = span->end;
    GLuint i;
 
-   switch (span->array->ChanType) {
-#if CHAN_BITS != 32
-   case GL_UNSIGNED_BYTE:
-      {
-         GLubyte (*spec)[4] = span->array->color.sz1.spec;
-         if (span->interpMask & SPAN_FLAT) {
-            GLubyte color[4];
-            color[RCOMP] = FixedToInt(span->specRed);
-            color[GCOMP] = FixedToInt(span->specGreen);
-            color[BCOMP] = FixedToInt(span->specBlue);
-            color[ACOMP] = 0;
-            for (i = 0; i < n; i++) {
-               COPY_4UBV(spec[i], color);
-            }
-         }
-         else {
-            GLfixed r = span->specRed;
-            GLfixed g = span->specGreen;
-            GLfixed b = span->specBlue;
-            GLint dr = span->specRedStep;
-            GLint dg = span->specGreenStep;
-            GLint db = span->specBlueStep;
-            for (i = 0; i < n; i++) {
-               spec[i][RCOMP] = CLAMP(FixedToChan(r), 0, 255);
-               spec[i][GCOMP] = CLAMP(FixedToChan(g), 0, 255);
-               spec[i][BCOMP] = CLAMP(FixedToChan(b), 0, 255);
-               spec[i][ACOMP] = 0;
-               r += dr;
-               g += dg;
-               b += db;
-            }
-         }
+   assert(!(span->arrayAttribs & FRAG_BIT_COL0));
+
+   if (span->arrayMask & SPAN_RGBA) {
+      /* convert array of int colors */
+      for (i = 0; i < n; i++) {
+         col0[i][0] = UBYTE_TO_FLOAT(span->array->rgba8[i][0]);
+         col0[i][1] = UBYTE_TO_FLOAT(span->array->rgba8[i][1]);
+         col0[i][2] = UBYTE_TO_FLOAT(span->array->rgba8[i][2]);
+         col0[i][3] = UBYTE_TO_FLOAT(span->array->rgba8[i][3]);
       }
-      break;
-   case GL_UNSIGNED_SHORT:
-      {
-         GLushort (*spec)[4] = span->array->color.sz2.spec;
-         if (span->interpMask & SPAN_FLAT) {
-            GLushort color[4];
-            color[RCOMP] = FixedToInt(span->specRed);
-            color[GCOMP] = FixedToInt(span->specGreen);
-            color[BCOMP] = FixedToInt(span->specBlue);
-            color[ACOMP] = 0;
-            for (i = 0; i < n; i++) {
-               COPY_4V(spec[i], color);
-            }
-         }
-         else {
-            GLfixed r = FloatToFixed(span->specRed);
-            GLfixed g = FloatToFixed(span->specGreen);
-            GLfixed b = FloatToFixed(span->specBlue);
-            GLint dr = FloatToFixed(span->specRedStep);
-            GLint dg = FloatToFixed(span->specGreenStep);
-            GLint db = FloatToFixed(span->specBlueStep);
-            for (i = 0; i < n; i++) {
-               spec[i][RCOMP] = FixedToInt(r);
-               spec[i][GCOMP] = FixedToInt(g);
-               spec[i][BCOMP] = FixedToInt(b);
-               spec[i][ACOMP] = 0;
-               r += dr;
-               g += dg;
-               b += db;
-            }
+   }
+   else {
+      /* interpolate red/green/blue/alpha to get float colors */
+      ASSERT(span->interpMask & SPAN_RGBA);
+      if (span->interpMask & SPAN_FLAT) {
+         GLfloat r = FixedToFloat(span->red);
+         GLfloat g = FixedToFloat(span->green);
+         GLfloat b = FixedToFloat(span->blue);
+         GLfloat a = FixedToFloat(span->alpha);
+         for (i = 0; i < n; i++) {
+            ASSIGN_4V(col0[i], r, g, b, a);
          }
       }
-      break;
-#endif
-   case GL_FLOAT:
-      {
-         GLfloat (*spec)[4] = span->array->color.sz4.spec;
-         if (span->interpMask & SPAN_FLAT) {
-            GLfloat color[4];
-            color[RCOMP] = span->specRed;
-            color[GCOMP] = span->specGreen;
-            color[BCOMP] = span->specBlue;
-            color[ACOMP] = 0.0F;
-            for (i = 0; i < n; i++) {
-               COPY_4V(spec[i], color);
-            }
-         }
-         else {
-            GLfloat r = span->specRed;
-            GLfloat g = span->specGreen;
-            GLfloat b = span->specBlue;
-            GLfloat dr = span->specRedStep;
-            GLfloat dg = span->specGreenStep;
-            GLfloat db = span->specBlueStep;
-            for (i = 0; i < n; i++) {
-               spec[i][RCOMP] = r;
-               spec[i][GCOMP] = g;
-               spec[i][BCOMP] = b;
-               spec[i][ACOMP] = 0.0F;
-               r += dr;
-               g += dg;
-               b += db;
-            }
+      else {
+         GLfloat r = FixedToFloat(span->red);
+         GLfloat g = FixedToFloat(span->green);
+         GLfloat b = FixedToFloat(span->blue);
+         GLfloat a = FixedToFloat(span->alpha);
+         GLfloat dr = FixedToFloat(span->redStep);
+         GLfloat dg = FixedToFloat(span->greenStep);
+         GLfloat db = FixedToFloat(span->blueStep);
+         GLfloat da = FixedToFloat(span->alphaStep);
+         for (i = 0; i < n; i++) {
+            col0[i][0] = r;
+            col0[i][1] = g;
+            col0[i][2] = b;
+            col0[i][3] = a;
+            r += dr;
+            g += dg;
+            b += db;
+            a += da;
          }
       }
-      break;
-   default:
-      _mesa_problem(NULL, "bad datatype in interpolate_specular");
    }
-   span->arrayMask |= SPAN_SPEC;
+
+   span->arrayAttribs |= FRAG_BIT_COL0;
+   span->array->ChanType = GL_FLOAT;
 }
 
 
+
 /* Fill in the span.color.index array from the interpolation values */
-static void
+static INLINE void
 interpolate_indexes(GLcontext *ctx, SWspan *span)
 {
    GLfixed index = span->index;
@@ -399,8 +380,8 @@ interpolate_indexes(GLcontext *ctx, SWspan *span)
    GLuint *indexes = span->array->index;
    GLuint i;
    (void) ctx;
-   ASSERT((span->interpMask & SPAN_INDEX)  &&
-         !(span->arrayMask & SPAN_INDEX));
+
+   ASSERT(!(span->arrayMask & SPAN_INDEX));
 
    if ((span->interpMask & SPAN_FLAT) || (indexStep == 0)) {
       /* constant color */
@@ -421,35 +402,16 @@ interpolate_indexes(GLcontext *ctx, SWspan *span)
 }
 
 
-/* Fill in the span.array.fog values from the interpolation values */
-static void
-interpolate_fog(const GLcontext *ctx, SWspan *span)
-{
-   GLfloat *fog = span->array->fog;
-   const GLfloat fogStep = span->fogStep;
-   GLfloat fogCoord = span->fog;
-   const GLuint haveW = (span->interpMask & SPAN_W);
-   const GLfloat wStep = haveW ? span->dwdx : 0.0F;
-   GLfloat w = haveW ? span->w : 1.0F;
-   GLuint i;
-   for (i = 0; i < span->end; i++) {
-      fog[i] = fogCoord / w;
-      fogCoord += fogStep;
-      w += wStep;
-   }
-   span->arrayMask |= SPAN_FOG;
-}
-
-
-/* Fill in the span.zArray array from the interpolation values */
+/**
+ * Fill in the span.zArray array from the span->z, zStep values.
+ */
 void
 _swrast_span_interpolate_z( const GLcontext *ctx, SWspan *span )
 {
    const GLuint n = span->end;
    GLuint i;
 
-   ASSERT((span->interpMask & SPAN_Z)  &&
-         !(span->arrayMask & SPAN_Z));
+   ASSERT(!(span->arrayMask & SPAN_Z));
 
    if (ctx->DrawBuffer->Visual.depthBits <= 16) {
       GLfixed zval = span->z;
@@ -473,7 +435,8 @@ _swrast_span_interpolate_z( const GLcontext *ctx, SWspan *span )
 }
 
 
-/*
+/**
+ * Compute mipmap LOD from partial derivatives.
  * This the ideal solution, as given in the OpenGL spec.
  */
 #if 0
@@ -495,8 +458,9 @@ compute_lambda(GLfloat dsdx, GLfloat dsdy, GLfloat dtdx, GLfloat dtdy,
 #endif
 
 
-/*
- * This is a faster approximation
+/**
+ * Compute mipmap LOD from partial derivatives.
+ * This is a faster approximation than above function.
  */
 GLfloat
 _swrast_compute_lambda(GLfloat dsdx, GLfloat dsdy, GLfloat dtdx, GLfloat dtdy,
@@ -521,351 +485,223 @@ _swrast_compute_lambda(GLfloat dsdx, GLfloat dsdy, GLfloat dtdx, GLfloat dtdy,
 
 
 /**
- * Fill in the span.texcoords array from the interpolation values.
+ * Fill in the span.array->attrib[FRAG_ATTRIB_TEXn] arrays from the
+ * using the attrStart/Step values.
+ *
+ * This function only used during fixed-function fragment processing.
+ *
  * Note: in the places where we divide by Q (or mult by invQ) we're
  * really doing two things: perspective correction and texcoord
  * projection.  Remember, for texcoord (s,t,r,q) we need to index
  * texels with (s/q, t/q, r/q).
- * If we're using a fragment program, we never do the division
- * for texcoord projection.  That's done by the TXP instruction
- * or user-written code.
  */
 static void
 interpolate_texcoords(GLcontext *ctx, SWspan *span)
 {
-   ASSERT(span->interpMask & SPAN_TEXTURE);
-   ASSERT(!(span->arrayMask & SPAN_TEXTURE));
-
-   if (ctx->Texture._EnabledCoordUnits > 1) {
-      /* multitexture */
-      GLuint u;
-      span->arrayMask |= SPAN_TEXTURE;
-      /* XXX CoordUnits vs. ImageUnits */
-      for (u = 0; u < ctx->Const.MaxTextureUnits; u++) {
-         if (ctx->Texture._EnabledCoordUnits & (1 << u)) {
-            const struct gl_texture_object *obj =ctx->Texture.Unit[u]._Current;
-            GLfloat texW, texH;
-            GLboolean needLambda;
-            if (obj) {
-               const struct gl_texture_image *img = obj->Image[0][obj->BaseLevel];
-               needLambda = (obj->MinFilter != obj->MagFilter)
-                  || ctx->FragmentProgram._Enabled;
-               texW = img->WidthScale;
-               texH = img->HeightScale;
-            }
-            else {
-               /* using a fragment program */
-               texW = 1.0;
-               texH = 1.0;
-               needLambda = GL_FALSE;
-            }
-            if (needLambda) {
-               GLfloat (*texcoord)[4] = span->array->texcoords[u];
-               GLfloat *lambda = span->array->lambda[u];
-               const GLfloat dsdx = span->texStepX[u][0];
-               const GLfloat dsdy = span->texStepY[u][0];
-               const GLfloat dtdx = span->texStepX[u][1];
-               const GLfloat dtdy = span->texStepY[u][1];
-               const GLfloat drdx = span->texStepX[u][2];
-               const GLfloat dqdx = span->texStepX[u][3];
-               const GLfloat dqdy = span->texStepY[u][3];
-               GLfloat s = span->tex[u][0];
-               GLfloat t = span->tex[u][1];
-               GLfloat r = span->tex[u][2];
-               GLfloat q = span->tex[u][3];
-               GLuint i;
-               if (ctx->FragmentProgram._Enabled || ctx->ATIFragmentShader._Enabled ||
-                   ctx->ShaderObjects._FragmentShaderPresent) {
-                  /* do perspective correction but don't divide s, t, r by q */
-                  const GLfloat dwdx = span->dwdx;
-                  GLfloat w = span->w;
-                  for (i = 0; i < span->end; i++) {
-                     const GLfloat invW = 1.0F / w;
-                     texcoord[i][0] = s * invW;
-                     texcoord[i][1] = t * invW;
-                     texcoord[i][2] = r * invW;
-                     texcoord[i][3] = q * invW;
-                     lambda[i] = _swrast_compute_lambda(dsdx, dsdy, dtdx, dtdy,
-                                                        dqdx, dqdy, texW, texH,
-                                                        s, t, q, invW);
-                     s += dsdx;
-                     t += dtdx;
-                     r += drdx;
-                     q += dqdx;
-                     w += dwdx;
-                  }
+   const GLuint maxUnit
+      = (ctx->Texture._EnabledCoordUnits > 1) ? ctx->Const.MaxTextureUnits : 1;
+   GLuint u;
+
+   /* XXX CoordUnits vs. ImageUnits */
+   for (u = 0; u < maxUnit; u++) {
+      if (ctx->Texture._EnabledCoordUnits & (1 << u)) {
+         const GLuint attr = FRAG_ATTRIB_TEX0 + u;
+         const struct gl_texture_object *obj = ctx->Texture.Unit[u]._Current;
+         GLfloat texW, texH;
+         GLboolean needLambda;
+         GLfloat (*texcoord)[4] = span->array->attribs[attr];
+         GLfloat *lambda = span->array->lambda[u];
+         const GLfloat dsdx = span->attrStepX[attr][0];
+         const GLfloat dsdy = span->attrStepY[attr][0];
+         const GLfloat dtdx = span->attrStepX[attr][1];
+         const GLfloat dtdy = span->attrStepY[attr][1];
+         const GLfloat drdx = span->attrStepX[attr][2];
+         const GLfloat dqdx = span->attrStepX[attr][3];
+         const GLfloat dqdy = span->attrStepY[attr][3];
+         GLfloat s = span->attrStart[attr][0];
+         GLfloat t = span->attrStart[attr][1];
+         GLfloat r = span->attrStart[attr][2];
+         GLfloat q = span->attrStart[attr][3];
+
+         if (obj) {
+            const struct gl_texture_image *img = obj->Image[0][obj->BaseLevel];
+            needLambda = (obj->MinFilter != obj->MagFilter)
+               || ctx->FragmentProgram._Current;
+            texW = img->WidthScale;
+            texH = img->HeightScale;
+         }
+         else {
+            /* using a fragment program */
+            texW = 1.0;
+            texH = 1.0;
+            needLambda = GL_FALSE;
+         }
 
+         if (needLambda) {
+            GLuint i;
+            if (ctx->FragmentProgram._Current
+                || ctx->ATIFragmentShader._Enabled) {
+               /* do perspective correction but don't divide s, t, r by q */
+               const GLfloat dwdx = span->attrStepX[FRAG_ATTRIB_WPOS][3];
+               GLfloat w = span->attrStart[FRAG_ATTRIB_WPOS][3];
+               for (i = 0; i < span->end; i++) {
+                  const GLfloat invW = 1.0F / w;
+                  texcoord[i][0] = s * invW;
+                  texcoord[i][1] = t * invW;
+                  texcoord[i][2] = r * invW;
+                  texcoord[i][3] = q * invW;
+                  lambda[i] = _swrast_compute_lambda(dsdx, dsdy, dtdx, dtdy,
+                                                     dqdx, dqdy, texW, texH,
+                                                     s, t, q, invW);
+                  s += dsdx;
+                  t += dtdx;
+                  r += drdx;
+                  q += dqdx;
+                  w += dwdx;
                }
-               else {
-                  for (i = 0; i < span->end; i++) {
-                     const GLfloat invQ = (q == 0.0F) ? 1.0F : (1.0F / q);
-                     texcoord[i][0] = s * invQ;
-                     texcoord[i][1] = t * invQ;
-                     texcoord[i][2] = r * invQ;
-                     texcoord[i][3] = q;
-                     lambda[i] = _swrast_compute_lambda(dsdx, dsdy, dtdx, dtdy,
-                                                        dqdx, dqdy, texW, texH,
-                                                        s, t, q, invQ);
-                     s += dsdx;
-                     t += dtdx;
-                     r += drdx;
-                     q += dqdx;
-                  }
-               }
-               span->arrayMask |= SPAN_LAMBDA;
             }
             else {
-               GLfloat (*texcoord)[4] = span->array->texcoords[u];
-               GLfloat *lambda = span->array->lambda[u];
-               const GLfloat dsdx = span->texStepX[u][0];
-               const GLfloat dtdx = span->texStepX[u][1];
-               const GLfloat drdx = span->texStepX[u][2];
-               const GLfloat dqdx = span->texStepX[u][3];
-               GLfloat s = span->tex[u][0];
-               GLfloat t = span->tex[u][1];
-               GLfloat r = span->tex[u][2];
-               GLfloat q = span->tex[u][3];
-               GLuint i;
-               if (ctx->FragmentProgram._Enabled || ctx->ATIFragmentShader._Enabled ||
-                   ctx->ShaderObjects._FragmentShaderPresent) {
-                  /* do perspective correction but don't divide s, t, r by q */
-                  const GLfloat dwdx = span->dwdx;
-                  GLfloat w = span->w;
-                  for (i = 0; i < span->end; i++) {
-                     const GLfloat invW = 1.0F / w;
-                     texcoord[i][0] = s * invW;
-                     texcoord[i][1] = t * invW;
-                     texcoord[i][2] = r * invW;
-                     texcoord[i][3] = q * invW;
-                     lambda[i] = 0.0;
-                     s += dsdx;
-                     t += dtdx;
-                     r += drdx;
-                     q += dqdx;
-                     w += dwdx;
-                  }
-               }
-               else if (dqdx == 0.0F) {
-                  /* Ortho projection or polygon's parallel to window X axis */
+               for (i = 0; i < span->end; i++) {
                   const GLfloat invQ = (q == 0.0F) ? 1.0F : (1.0F / q);
-                  for (i = 0; i < span->end; i++) {
-                     texcoord[i][0] = s * invQ;
-                     texcoord[i][1] = t * invQ;
-                     texcoord[i][2] = r * invQ;
-                     texcoord[i][3] = q;
-                     lambda[i] = 0.0;
-                     s += dsdx;
-                     t += dtdx;
-                     r += drdx;
-                  }
-               }
-               else {
-                  for (i = 0; i < span->end; i++) {
-                     const GLfloat invQ = (q == 0.0F) ? 1.0F : (1.0F / q);
-                     texcoord[i][0] = s * invQ;
-                     texcoord[i][1] = t * invQ;
-                     texcoord[i][2] = r * invQ;
-                     texcoord[i][3] = q;
-                     lambda[i] = 0.0;
-                     s += dsdx;
-                     t += dtdx;
-                     r += drdx;
-                     q += dqdx;
-                  }
+                  texcoord[i][0] = s * invQ;
+                  texcoord[i][1] = t * invQ;
+                  texcoord[i][2] = r * invQ;
+                  texcoord[i][3] = q;
+                  lambda[i] = _swrast_compute_lambda(dsdx, dsdy, dtdx, dtdy,
+                                                     dqdx, dqdy, texW, texH,
+                                                     s, t, q, invQ);
+                  s += dsdx;
+                  t += dtdx;
+                  r += drdx;
+                  q += dqdx;
                }
-            } /* lambda */
-         } /* if */
-      } /* for */
-   }
-   else {
-      /* single texture */
-      const struct gl_texture_object *obj = ctx->Texture.Unit[0]._Current;
-      GLfloat texW, texH;
-      GLboolean needLambda;
-      if (obj) {
-         const struct gl_texture_image *img = obj->Image[0][obj->BaseLevel];
-         needLambda = (obj->MinFilter != obj->MagFilter)
-            || ctx->FragmentProgram._Enabled;
-         texW = (GLfloat) img->WidthScale;
-         texH = (GLfloat) img->HeightScale;
-      }
-      else {
-         needLambda = GL_FALSE;
-         texW = texH = 1.0;
-      }
-      span->arrayMask |= SPAN_TEXTURE;
-      if (needLambda) {
-         /* just texture unit 0, with lambda */
-         GLfloat (*texcoord)[4] = span->array->texcoords[0];
-         GLfloat *lambda = span->array->lambda[0];
-         const GLfloat dsdx = span->texStepX[0][0];
-         const GLfloat dsdy = span->texStepY[0][0];
-         const GLfloat dtdx = span->texStepX[0][1];
-         const GLfloat dtdy = span->texStepY[0][1];
-         const GLfloat drdx = span->texStepX[0][2];
-         const GLfloat dqdx = span->texStepX[0][3];
-         const GLfloat dqdy = span->texStepY[0][3];
-         GLfloat s = span->tex[0][0];
-         GLfloat t = span->tex[0][1];
-         GLfloat r = span->tex[0][2];
-         GLfloat q = span->tex[0][3];
-         GLuint i;
-         if (ctx->FragmentProgram._Enabled || ctx->ATIFragmentShader._Enabled ||
-             ctx->ShaderObjects._FragmentShaderPresent) {
-            /* do perspective correction but don't divide s, t, r by q */
-            const GLfloat dwdx = span->dwdx;
-            GLfloat w = span->w;
-            for (i = 0; i < span->end; i++) {
-               const GLfloat invW = 1.0F / w;
-               texcoord[i][0] = s * invW;
-               texcoord[i][1] = t * invW;
-               texcoord[i][2] = r * invW;
-               texcoord[i][3] = q * invW;
-               lambda[i] = _swrast_compute_lambda(dsdx, dsdy, dtdx, dtdy,
-                                                  dqdx, dqdy, texW, texH,
-                                                  s, t, q, invW);
-               s += dsdx;
-               t += dtdx;
-               r += drdx;
-               q += dqdx;
-               w += dwdx;
             }
+            span->arrayMask |= SPAN_LAMBDA;
          }
          else {
-            /* tex.c */
-            for (i = 0; i < span->end; i++) {
-               const GLfloat invQ = (q == 0.0F) ? 1.0F : (1.0F / q);
-               lambda[i] = _swrast_compute_lambda(dsdx, dsdy, dtdx, dtdy,
-                                                dqdx, dqdy, texW, texH,
-                                                s, t, q, invQ);
-               texcoord[i][0] = s * invQ;
-               texcoord[i][1] = t * invQ;
-               texcoord[i][2] = r * invQ;
-               texcoord[i][3] = q;
-               s += dsdx;
-               t += dtdx;
-               r += drdx;
-               q += dqdx;
-            }
-         }
-         span->arrayMask |= SPAN_LAMBDA;
-      }
-      else {
-         /* just texture 0, without lambda */
-         GLfloat (*texcoord)[4] = span->array->texcoords[0];
-         const GLfloat dsdx = span->texStepX[0][0];
-         const GLfloat dtdx = span->texStepX[0][1];
-         const GLfloat drdx = span->texStepX[0][2];
-         const GLfloat dqdx = span->texStepX[0][3];
-         GLfloat s = span->tex[0][0];
-         GLfloat t = span->tex[0][1];
-         GLfloat r = span->tex[0][2];
-         GLfloat q = span->tex[0][3];
-         GLuint i;
-         if (ctx->FragmentProgram._Enabled || ctx->ATIFragmentShader._Enabled ||
-             ctx->ShaderObjects._FragmentShaderPresent) {
-            /* do perspective correction but don't divide s, t, r by q */
-            const GLfloat dwdx = span->dwdx;
-            GLfloat w = span->w;
-            for (i = 0; i < span->end; i++) {
-               const GLfloat invW = 1.0F / w;
-               texcoord[i][0] = s * invW;
-               texcoord[i][1] = t * invW;
-               texcoord[i][2] = r * invW;
-               texcoord[i][3] = q * invW;
-               s += dsdx;
-               t += dtdx;
-               r += drdx;
-               q += dqdx;
-               w += dwdx;
-            }
-         }
-         else if (dqdx == 0.0F) {
-            /* Ortho projection or polygon's parallel to window X axis */
-            const GLfloat invQ = (q == 0.0F) ? 1.0F : (1.0F / q);
-            for (i = 0; i < span->end; i++) {
-               texcoord[i][0] = s * invQ;
-               texcoord[i][1] = t * invQ;
-               texcoord[i][2] = r * invQ;
-               texcoord[i][3] = q;
-               s += dsdx;
-               t += dtdx;
-               r += drdx;
+            GLuint i;
+            if (ctx->FragmentProgram._Current ||
+                ctx->ATIFragmentShader._Enabled) {
+               /* do perspective correction but don't divide s, t, r by q */
+               const GLfloat dwdx = span->attrStepX[FRAG_ATTRIB_WPOS][3];
+               GLfloat w = span->attrStart[FRAG_ATTRIB_WPOS][3];
+               for (i = 0; i < span->end; i++) {
+                  const GLfloat invW = 1.0F / w;
+                  texcoord[i][0] = s * invW;
+                  texcoord[i][1] = t * invW;
+                  texcoord[i][2] = r * invW;
+                  texcoord[i][3] = q * invW;
+                  lambda[i] = 0.0;
+                  s += dsdx;
+                  t += dtdx;
+                  r += drdx;
+                  q += dqdx;
+                  w += dwdx;
+               }
             }
-         }
-         else {
-            for (i = 0; i < span->end; i++) {
+            else if (dqdx == 0.0F) {
+               /* Ortho projection or polygon's parallel to window X axis */
                const GLfloat invQ = (q == 0.0F) ? 1.0F : (1.0F / q);
-               texcoord[i][0] = s * invQ;
-               texcoord[i][1] = t * invQ;
-               texcoord[i][2] = r * invQ;
-               texcoord[i][3] = q;
-               s += dsdx;
-               t += dtdx;
-               r += drdx;
-               q += dqdx;
+               for (i = 0; i < span->end; i++) {
+                  texcoord[i][0] = s * invQ;
+                  texcoord[i][1] = t * invQ;
+                  texcoord[i][2] = r * invQ;
+                  texcoord[i][3] = q;
+                  lambda[i] = 0.0;
+                  s += dsdx;
+                  t += dtdx;
+                  r += drdx;
+               }
             }
-         }
-      }
-   }
+            else {
+               for (i = 0; i < span->end; i++) {
+                  const GLfloat invQ = (q == 0.0F) ? 1.0F : (1.0F / q);
+                  texcoord[i][0] = s * invQ;
+                  texcoord[i][1] = t * invQ;
+                  texcoord[i][2] = r * invQ;
+                  texcoord[i][3] = q;
+                  lambda[i] = 0.0;
+                  s += dsdx;
+                  t += dtdx;
+                  r += drdx;
+                  q += dqdx;
+               }
+            }
+         } /* lambda */
+      } /* if */
+   } /* for */
 }
 
 
 /**
- * Fill in the span.varying array from the interpolation values.
+ * Fill in the arrays->attribs[FRAG_ATTRIB_WPOS] array.
  */
-static void
-interpolate_varying(GLcontext *ctx, SWspan *span)
+static INLINE void
+interpolate_wpos(GLcontext *ctx, SWspan *span)
 {
-   GLuint i, j;
-
-   ASSERT(span->interpMask & SPAN_VARYING);
-   ASSERT(!(span->arrayMask & SPAN_VARYING));
-
-   span->arrayMask |= SPAN_VARYING;
-
-   for (i = 0; i < MAX_VARYING_VECTORS; i++) {
-      for (j = 0; j < VARYINGS_PER_VECTOR; j++) {
-         const GLfloat dvdx = span->varStepX[i][j];
-         GLfloat v = span->var[i][j];
-         const GLfloat dwdx = span->dwdx;
-         GLfloat w = span->w;
-         GLuint k;
+   GLfloat (*wpos)[4] = span->array->attribs[FRAG_ATTRIB_WPOS];
+   GLuint i;
+   const GLfloat zScale = 1.0 / ctx->DrawBuffer->_DepthMaxF;
+   GLfloat w, dw;
 
-         for (k = 0; k < span->end; k++) {
-            GLfloat invW = 1.0f / w;
-            span->array->varying[k][i][j] = v * invW;
-            v += dvdx;
-            w += dwdx;
-         }
+   if (span->arrayMask & SPAN_XY) {
+      for (i = 0; i < span->end; i++) {
+         wpos[i][0] = (GLfloat) span->array->x[i];
+         wpos[i][1] = (GLfloat) span->array->y[i];
+      }
+   }
+   else {
+      for (i = 0; i < span->end; i++) {
+         wpos[i][0] = (GLfloat) span->x + i;
+         wpos[i][1] = (GLfloat) span->y;
       }
    }
+
+   w = span->attrStart[FRAG_ATTRIB_WPOS][3];
+   dw = span->attrStepX[FRAG_ATTRIB_WPOS][3];
+   for (i = 0; i < span->end; i++) {
+      wpos[i][2] = (GLfloat) span->array->z[i] * zScale;
+      wpos[i][3] = w;
+      w += dw;
+   }
 }
 
 
 /**
  * Apply the current polygon stipple pattern to a span of pixels.
  */
-static void
-stipple_polygon_span( GLcontext *ctx, SWspan *span )
+static INLINE void
+stipple_polygon_span(GLcontext *ctx, SWspan *span)
 {
-   const GLuint highbit = 0x80000000;
-   const GLuint stipple = ctx->PolygonStipple[span->y % 32];
    GLubyte *mask = span->array->mask;
-   GLuint i, m;
 
    ASSERT(ctx->Polygon.StippleFlag);
-   ASSERT((span->arrayMask & SPAN_XY) == 0);
 
-   m = highbit >> (GLuint) (span->x % 32);
-
-   for (i = 0; i < span->end; i++) {
-      if ((m & stipple) == 0) {
-        mask[i] = 0;
+   if (span->arrayMask & SPAN_XY) {
+      /* arrays of x/y pixel coords */
+      GLuint i;
+      for (i = 0; i < span->end; i++) {
+         const GLint col = span->array->x[i] % 32;
+         const GLint row = span->array->y[i] % 32;
+         const GLuint stipple = ctx->PolygonStipple[row];
+         if (((1 << col) & stipple) == 0) {
+            mask[i] = 0;
+         }
       }
-      m = m >> 1;
-      if (m == 0) {
-         m = highbit;
+   }
+   else {
+      /* horizontal span of pixels */
+      const GLuint highBit = 1 << 31;
+      const GLuint stipple = ctx->PolygonStipple[span->y % 32];
+      GLuint i, m = highBit >> (GLuint) (span->x % 32);
+      for (i = 0; i < span->end; i++) {
+         if ((m & stipple) == 0) {
+            mask[i] = 0;
+         }
+         m = m >> 1;
+         if (m == 0) {
+            m = highBit;
+         }
       }
    }
    span->writeAll = GL_FALSE;
@@ -879,7 +715,7 @@ stipple_polygon_span( GLcontext *ctx, SWspan *span )
  * Return:   GL_TRUE   some pixels still visible
  *           GL_FALSE  nothing visible
  */
-static GLuint
+static INLINE GLuint
 clip_span( GLcontext *ctx, SWspan *span )
 {
    const GLint xmin = ctx->DrawBuffer->_Xmin;
@@ -958,7 +794,9 @@ _swrast_write_index_span( GLcontext *ctx, SWspan *span)
    ASSERT(span->primitive == GL_POINT  ||  span->primitive == GL_LINE ||
          span->primitive == GL_POLYGON  ||  span->primitive == GL_BITMAP);
    ASSERT((span->interpMask | span->arrayMask) & SPAN_INDEX);
+   /*
    ASSERT((span->interpMask & span->arrayMask) == 0);
+   */
 
    if (span->arrayMask & SPAN_MASK) {
       /* mask was initialized by caller, probably glBitmap */
@@ -1005,7 +843,7 @@ _swrast_write_index_span( GLcontext *ctx, SWspan *span)
 
    /* Stencil and Z testing */
    if (ctx->Depth.Test || ctx->Stencil.Enabled) {
-      if (span->interpMask & SPAN_Z)
+      if (!(span->arrayMask & SPAN_Z))
          _swrast_span_interpolate_z(ctx, span);
 
       if (ctx->Stencil.Enabled) {
@@ -1046,7 +884,7 @@ _swrast_write_index_span( GLcontext *ctx, SWspan *span)
        ctx->Color.IndexLogicOpEnabled ||
        ctx->Color.IndexMask != 0xffffffff ||
        (span->arrayMask & SPAN_COVERAGE)) {
-      if (span->interpMask & SPAN_INDEX) {
+      if (!(span->arrayMask & SPAN_INDEX) /*span->interpMask & SPAN_INDEX*/) {
          interpolate_indexes(ctx, span);
       }
    }
@@ -1095,7 +933,7 @@ _swrast_write_index_span( GLcontext *ctx, SWspan *span)
             _swrast_mask_ci_span(ctx, rb, span);
          }
 
-         if ((span->interpMask & SPAN_INDEX) && span->indexStep == 0) {
+         if (!(span->arrayMask & SPAN_INDEX) && span->indexStep == 0) {
             /* all fragments have same color index */
             GLubyte index8;
             GLushort index16;
@@ -1175,76 +1013,65 @@ _swrast_write_index_span( GLcontext *ctx, SWspan *span)
 
 
 /**
- * Add specular color to base color.  This is used only when
- * GL_LIGHT_MODEL_COLOR_CONTROL = GL_SEPARATE_SPECULAR_COLOR.
+ * Add specular colors to primary colors.
+ * Only called during fixed-function operation.
+ * Result is float color array (FRAG_ATTRIB_COL0).
  */
-static void
+static INLINE void
 add_specular(GLcontext *ctx, SWspan *span)
 {
-   switch (span->array->ChanType) {
-   case GL_UNSIGNED_BYTE:
-      {
-         GLubyte (*rgba)[4] = span->array->color.sz1.rgba;
-         GLubyte (*spec)[4] = span->array->color.sz1.spec;
-         GLuint i;
-         for (i = 0; i < span->end; i++) {
-            GLint r = rgba[i][RCOMP] + spec[i][RCOMP];
-            GLint g = rgba[i][GCOMP] + spec[i][GCOMP];
-            GLint b = rgba[i][BCOMP] + spec[i][BCOMP];
-            GLint a = rgba[i][ACOMP] + spec[i][ACOMP];
-            rgba[i][RCOMP] = MIN2(r, 255);
-            rgba[i][GCOMP] = MIN2(g, 255);
-            rgba[i][BCOMP] = MIN2(b, 255);
-            rgba[i][ACOMP] = MIN2(a, 255);
-         }
+   const SWcontext *swrast = SWRAST_CONTEXT(ctx);
+   const GLubyte *mask = span->array->mask;
+   GLfloat (*col0)[4] = span->array->attribs[FRAG_ATTRIB_COL0];
+   GLfloat (*col1)[4] = span->array->attribs[FRAG_ATTRIB_COL1];
+   GLuint i;
+
+   ASSERT(!ctx->FragmentProgram._Current);
+   ASSERT(span->arrayMask & SPAN_RGBA);
+   ASSERT(swrast->_ActiveAttribMask & FRAG_BIT_COL1);
+
+   if (span->array->ChanType == GL_FLOAT) {
+      if ((span->arrayAttribs & FRAG_BIT_COL0) == 0) {
+         interpolate_active_attribs(ctx, span, FRAG_BIT_COL0);
       }
-      break;
-   case GL_UNSIGNED_SHORT:
-      {
-         GLushort (*rgba)[4] = span->array->color.sz2.rgba;
-         GLushort (*spec)[4] = span->array->color.sz2.spec;
-         GLuint i;
-         for (i = 0; i < span->end; i++) {
-            GLint r = rgba[i][RCOMP] + spec[i][RCOMP];
-            GLint g = rgba[i][GCOMP] + spec[i][GCOMP];
-            GLint b = rgba[i][BCOMP] + spec[i][BCOMP];
-            GLint a = rgba[i][ACOMP] + spec[i][ACOMP];
-            rgba[i][RCOMP] = MIN2(r, 65535);
-            rgba[i][GCOMP] = MIN2(g, 65535);
-            rgba[i][BCOMP] = MIN2(b, 65535);
-            rgba[i][ACOMP] = MIN2(a, 65535);
-         }
+   }
+   else {
+      /* need float colors */
+      if ((span->arrayAttribs & FRAG_BIT_COL0) == 0) {
+         interpolate_float_colors(span);
       }
-      break;
-   case GL_FLOAT:
-      {
-         GLfloat (*rgba)[4] = span->array->color.sz4.rgba;
-         GLfloat (*spec)[4] = span->array->color.sz4.spec;
-         GLuint i;
-         for (i = 0; i < span->end; i++) {
-            rgba[i][RCOMP] += spec[i][RCOMP];
-            rgba[i][GCOMP] += spec[i][GCOMP];
-            rgba[i][BCOMP] += spec[i][BCOMP];
-            rgba[i][ACOMP] += spec[i][ACOMP];
-         }
+   }
+
+   if ((span->arrayAttribs & FRAG_BIT_COL1) == 0) {
+      /* XXX could avoid this and interpolate COL1 in the loop below */
+      interpolate_active_attribs(ctx, span, FRAG_BIT_COL1);
+   }
+
+   ASSERT(span->arrayAttribs & FRAG_BIT_COL0);
+   ASSERT(span->arrayAttribs & FRAG_BIT_COL1);
+
+   for (i = 0; i < span->end; i++) {
+      if (mask[i]) {
+         col0[i][0] += col1[i][0];
+         col0[i][1] += col1[i][1];
+         col0[i][2] += col1[i][2];
       }
-      break;
-   default:
-      _mesa_problem(ctx, "Invalid datatype in add_specular");
    }
+
+   span->array->ChanType = GL_FLOAT;
 }
 
 
 /**
  * Apply antialiasing coverage value to alpha values.
  */
-static void
+static INLINE void
 apply_aa_coverage(SWspan *span)
 {
    const GLfloat *coverage = span->array->coverage;
    GLuint i;
    if (span->array->ChanType == GL_UNSIGNED_BYTE) {
-      GLubyte (*rgba)[4] = span->array->color.sz1.rgba;
+      GLubyte (*rgba)[4] = span->array->rgba8;
       for (i = 0; i < span->end; i++) {
          const GLfloat a = rgba[i][ACOMP] * coverage[i];
          rgba[i][ACOMP] = (GLubyte) CLAMP(a, 0.0, 255.0);
@@ -1253,16 +1080,17 @@ apply_aa_coverage(SWspan *span)
       }
    }
    else if (span->array->ChanType == GL_UNSIGNED_SHORT) {
-      GLushort (*rgba)[4] = span->array->color.sz2.rgba;
+      GLushort (*rgba)[4] = span->array->rgba16;
       for (i = 0; i < span->end; i++) {
          const GLfloat a = rgba[i][ACOMP] * coverage[i];
          rgba[i][ACOMP] = (GLushort) CLAMP(a, 0.0, 65535.0);
       }
    }
    else {
-      GLfloat (*rgba)[4] = span->array->color.sz4.rgba;
+      GLfloat (*rgba)[4] = span->array->attribs[FRAG_ATTRIB_COL0];
       for (i = 0; i < span->end; i++) {
          rgba[i][ACOMP] = rgba[i][ACOMP] * coverage[i];
+         /* clamp later */
       }
    }
 }
@@ -1271,10 +1099,10 @@ apply_aa_coverage(SWspan *span)
 /**
  * Clamp span's float colors to [0,1]
  */
-static void
+static INLINE void
 clamp_colors(SWspan *span)
 {
-   GLfloat (*rgba)[4] = span->array->color.sz4.rgba;
+   GLfloat (*rgba)[4] = span->array->attribs[FRAG_ATTRIB_COL0];
    GLuint i;
    ASSERT(span->array->ChanType == GL_FLOAT);
    for (i = 0; i < span->end; i++) {
@@ -1288,28 +1116,35 @@ clamp_colors(SWspan *span)
 
 /**
  * Convert the span's color arrays to the given type.
+ * The only way 'output' can be greater than one is when we have a fragment
+ * program that writes to gl_FragData[1] or higher.
+ * \param output  which fragment program color output is being processed
  */
-static void
-convert_color_type(GLcontext *ctx, SWspan *span, GLenum newType)
+static INLINE void
+convert_color_type(SWspan *span, GLenum newType, GLuint output)
 {
    GLvoid *src, *dst;
-   if (span->array->ChanType == GL_UNSIGNED_BYTE) {
-      src = span->array->color.sz1.rgba;
+
+   if (output > 0 || span->array->ChanType == GL_FLOAT) {
+      src = span->array->attribs[FRAG_ATTRIB_COL0 + output];
+      span->array->ChanType = GL_FLOAT;
    }
    else if (span->array->ChanType == GL_UNSIGNED_BYTE) {
-      src = span->array->color.sz2.rgba;
+      src = span->array->rgba8;
    }
    else {
-      src = span->array->color.sz4.rgba;
+      ASSERT(span->array->ChanType == GL_UNSIGNED_SHORT);
+      src = span->array->rgba16;
    }
+
    if (newType == GL_UNSIGNED_BYTE) {
-      dst = span->array->color.sz1.rgba;
+      dst = span->array->rgba8;
    }
-   else if (newType == GL_UNSIGNED_BYTE) {
-      dst = span->array->color.sz2.rgba;
+   else if (newType == GL_UNSIGNED_SHORT) {
+      dst = span->array->rgba16;
    }
    else {
-      dst = span->array->color.sz4.rgba;
+      dst = span->array->attribs[FRAG_ATTRIB_COL0];
    }
 
    _mesa_convert_colors(span->array->ChanType, src,
@@ -1317,6 +1152,75 @@ convert_color_type(GLcontext *ctx, SWspan *span, GLenum newType)
                         span->end, span->array->mask);
 
    span->array->ChanType = newType;
+   span->array->rgba = dst;
+}
+
+
+
+/**
+ * Apply fragment shader, fragment program or normal texturing to span.
+ */
+static INLINE void
+shade_texture_span(GLcontext *ctx, SWspan *span)
+{
+   GLbitfield inputsRead;
+
+   /* Determine which fragment attributes are actually needed */
+   if (ctx->FragmentProgram._Current) {
+      inputsRead = ctx->FragmentProgram._Current->Base.InputsRead;
+   }
+   else {
+      /* XXX we could be a bit smarter about this */
+      inputsRead = ~0;
+   }
+
+   if (ctx->FragmentProgram._Current ||
+       ctx->ATIFragmentShader._Enabled) {
+      /* programmable shading */
+      if (span->primitive == GL_BITMAP && span->array->ChanType != GL_FLOAT) {
+         convert_color_type(span, GL_FLOAT, 0);
+      }
+      if (span->primitive != GL_POINT || ctx->Point.PointSprite) {
+         /* for points, we populated the arrays already */
+         interpolate_active_attribs(ctx, span, ~0);
+      }
+      span->array->ChanType = GL_FLOAT;
+
+      if (!(span->arrayMask & SPAN_Z))
+         _swrast_span_interpolate_z (ctx, span);
+
+#if 0
+      if (inputsRead & FRAG_BIT_WPOS)
+#else
+      /* XXX always interpolate wpos so that DDX/DDY work */
+#endif
+         interpolate_wpos(ctx, span);
+
+      /* Run fragment program/shader now */
+      if (ctx->FragmentProgram._Current) {
+         _swrast_exec_fragment_program(ctx, span);
+      }
+      else {
+         ASSERT(ctx->ATIFragmentShader._Enabled);
+         _swrast_exec_fragment_shader(ctx, span);
+      }
+   }
+   else if (ctx->Texture._EnabledUnits) {
+      /* conventional texturing */
+
+#if CHAN_BITS == 32
+      if ((span->arrayAttribs & FRAG_BIT_COL0) == 0) {
+         interpolate_int_colors(ctx, span);
+      }
+#else
+      if (!(span->arrayMask & SPAN_RGBA))
+         interpolate_int_colors(ctx, span);
+#endif
+      if ((span->arrayAttribs & FRAG_BITS_TEX_ANY) == 0x0)
+         interpolate_texcoords(ctx, span);
+
+      _swrast_texture_span(ctx, span);
+   }
 }
 
 
@@ -1331,28 +1235,31 @@ convert_color_type(GLcontext *ctx, SWspan *span, GLenum newType)
 void
 _swrast_write_rgba_span( GLcontext *ctx, SWspan *span)
 {
+   const SWcontext *swrast = SWRAST_CONTEXT(ctx);
    const GLuint colorMask = *((GLuint *) ctx->Color.ColorMask);
-   SWcontext *swrast = SWRAST_CONTEXT(ctx);
    const GLbitfield origInterpMask = span->interpMask;
    const GLbitfield origArrayMask = span->arrayMask;
-   const GLenum chanType = span->array->ChanType;
-   const GLboolean deferredTexture = !(ctx->Color.AlphaEnabled ||
-                                       ctx->FragmentProgram._Enabled ||
-                                       ctx->ShaderObjects._FragmentShaderPresent);
+   const GLbitfield origArrayAttribs = span->arrayAttribs;
+   const GLenum origChanType = span->array->ChanType;
+   void * const origRgba = span->array->rgba;
+   const GLboolean shader = (ctx->FragmentProgram._Current
+                             || ctx->ATIFragmentShader._Enabled);
+   const GLboolean shaderOrTexture = shader || ctx->Texture._EnabledUnits;
+   struct gl_framebuffer *fb = ctx->DrawBuffer;
+   GLuint output;
+
+   /*
+   printf("%s()  interp 0x%x  array 0x%x\n", __FUNCTION__,
+          span->interpMask, span->arrayMask);
+   */
 
    ASSERT(span->primitive == GL_POINT ||
           span->primitive == GL_LINE ||
          span->primitive == GL_POLYGON ||
           span->primitive == GL_BITMAP);
    ASSERT(span->end <= MAX_WIDTH);
-   ASSERT((span->interpMask & span->arrayMask) == 0);
-   ASSERT((span->interpMask & SPAN_RGBA) ^ (span->arrayMask & SPAN_RGBA));
-
-   /*
-   printf("%s()  interp 0x%x  array 0x%x\n", __FUNCTION__,
-          span->interpMask, span->arrayMask);
-   */
 
+   /* Fragment write masks */
    if (span->arrayMask & SPAN_MASK) {
       /* mask was initialized by caller, probably glBitmap */
       span->writeAll = GL_FALSE;
@@ -1375,10 +1282,10 @@ _swrast_write_rgba_span( GLcontext *ctx, SWspan *span)
       GLuint i;
       for (i = 0; i < span->end; i++) {
          if (span->array->mask[i]) {
-            assert(span->array->x[i] >= ctx->DrawBuffer->_Xmin);
-            assert(span->array->x[i] < ctx->DrawBuffer->_Xmax);
-            assert(span->array->y[i] >= ctx->DrawBuffer->_Ymin);
-            assert(span->array->y[i] < ctx->DrawBuffer->_Ymax);
+            assert(span->array->x[i] >= fb->_Xmin);
+            assert(span->array->x[i] < fb->_Xmax);
+            assert(span->array->y[i] >= fb->_Ymin);
+            assert(span->array->y[i] < fb->_Ymax);
          }
       }
    }
@@ -1389,90 +1296,32 @@ _swrast_write_rgba_span( GLcontext *ctx, SWspan *span)
       stipple_polygon_span(ctx, span);
    }
 
-   /* Interpolate texcoords? */
-   if (ctx->Texture._EnabledCoordUnits && (span->interpMask & SPAN_TEXTURE)) {
-      interpolate_texcoords(ctx, span);
-   }
-
-   if (ctx->ShaderObjects._FragmentShaderPresent) {
-      interpolate_varying(ctx, span);
-   }
-
-   /* This is the normal place to compute the resulting fragment color/Z.
-    * As an optimization, we try to defer this until after Z/stencil
-    * testing in order to try to avoid computing colors that we won't
-    * actually need.
+   /* This is the normal place to compute the fragment color/Z
+    * from texturing or shading.
     */
-   if (!deferredTexture) {
-      /* Now we need the rgba array, fill it in if needed */
-      if (span->interpMask & SPAN_RGBA)
-         interpolate_colors(span);
-
-      if (span->interpMask & SPAN_SPEC)
-         interpolate_specular(span);
-
-      if (span->interpMask & SPAN_FOG)
-         interpolate_fog(ctx, span);
-
-      /* XXX need this code below too? */
-
-      /* use float colors if running a fragment program or shader */
-      if (ctx->ShaderObjects._FragmentShaderPresent ||
-          ctx->FragmentProgram._Enabled ||
-          ctx->ATIFragmentShader._Enabled) {
-         const GLenum oldType = span->array->ChanType;
-         /* work with float colors */
-         if (oldType != GL_FLOAT) {
-            GLvoid *src = (oldType == GL_UNSIGNED_BYTE)
-               ? (GLvoid *) span->array->color.sz1.rgba
-               : (GLvoid *) span->array->color.sz2.rgba;
-            _mesa_convert_colors(oldType, src,
-                                 GL_FLOAT, span->array->color.sz4.rgba,
-                                 span->end, span->array->mask);
-            span->array->ChanType = GL_FLOAT;
-         }
-      }
+   if (shaderOrTexture && !swrast->_DeferredTexture) {
+      shade_texture_span(ctx, span);
+   }
 
-      /* Compute fragment colors with fragment program or texture lookups */
-#if FEATURE_ARB_fragment_shader
-      if (ctx->ShaderObjects._FragmentShaderPresent) {
-         if (span->interpMask & SPAN_Z)
-            _swrast_span_interpolate_z (ctx, span);
-         _swrast_exec_arbshader (ctx, span);
-      }
-      else
-#endif
-      if (ctx->FragmentProgram._Enabled) {
-         /* frag prog may need Z values */
-         if (span->interpMask & SPAN_Z)
-            _swrast_span_interpolate_z(ctx, span);
-         _swrast_exec_fragment_program( ctx, span );
-      }
-      else if (ctx->ATIFragmentShader._Enabled)
-         _swrast_exec_fragment_shader( ctx, span );
-      else if (ctx->Texture._EnabledUnits && (span->arrayMask & SPAN_TEXTURE))
-         _swrast_texture_span( ctx, span );
-
-      /* Do the alpha test */
-      if (ctx->Color.AlphaEnabled) {
-         if (!_swrast_alpha_test(ctx, span)) {
-            goto end;
-        }
+   /* Do the alpha test */
+   if (ctx->Color.AlphaEnabled) {
+      if (!_swrast_alpha_test(ctx, span)) {
+         goto end;
       }
    }
 
    /* Stencil and Z testing */
    if (ctx->Stencil.Enabled || ctx->Depth.Test) {
-      if (span->interpMask & SPAN_Z)
+      if (!(span->arrayMask & SPAN_Z))
          _swrast_span_interpolate_z(ctx, span);
 
-      if (ctx->Stencil.Enabled && ctx->DrawBuffer->Visual.stencilBits > 0) {
+      if (ctx->Stencil.Enabled && fb->Visual.stencilBits > 0) {
          /* Combined Z/stencil tests */
          if (!_swrast_stencil_and_ztest_span(ctx, span)) {
             goto end;
          }
       }
-      else if (ctx->DrawBuffer->Visual.depthBits > 0) {
+      else if (fb->Visual.depthBits > 0) {
          /* Just regular depth testing */
          ASSERT(ctx->Depth.Test);
          ASSERT(span->arrayMask & SPAN_Z);
@@ -1503,52 +1352,28 @@ _swrast_write_rgba_span( GLcontext *ctx, SWspan *span)
     * a good chance that many fragments will have already been killed by
     * Z/stencil testing.
     */
-   if (deferredTexture) {
-      /* Now we need the rgba array, fill it in if needed */
-      if (span->interpMask & SPAN_RGBA)
-         interpolate_colors(span);
-
-      if (span->interpMask & SPAN_SPEC)
-         interpolate_specular(span);
-
-      if (span->interpMask & SPAN_FOG)
-         interpolate_fog(ctx, span);
-
-#if FEATURE_ARB_fragment_shader
-      if (ctx->ShaderObjects._FragmentShaderPresent) {
-         if (span->interpMask & SPAN_Z)
-            _swrast_span_interpolate_z(ctx, span);
-         _swrast_exec_arbshader (ctx, span);
-      }
-      else
-#endif
-      if (ctx->FragmentProgram._Enabled)
-         _swrast_exec_fragment_program( ctx, span );
-      else if (ctx->ATIFragmentShader._Enabled)
-         _swrast_exec_fragment_shader( ctx, span );
-      else if (ctx->Texture._EnabledUnits && (span->arrayMask & SPAN_TEXTURE))
-         _swrast_texture_span( ctx, span );
+   if (shaderOrTexture && swrast->_DeferredTexture) {
+      shade_texture_span(ctx, span);
+   }
+
+#if CHAN_BITS == 32
+   if ((span->arrayAttribs & FRAG_BIT_COL0) == 0) {
+      interpolate_active_attribs(ctx, span, FRAG_BIT_COL0);
+   }
+#else
+   if ((span->arrayMask & SPAN_RGBA) == 0) {
+      interpolate_int_colors(ctx, span);
    }
+#endif
 
    ASSERT(span->arrayMask & SPAN_RGBA);
 
-   if (!ctx->FragmentProgram._Enabled &&
-       !ctx->ShaderObjects._FragmentShaderPresent) {
+   if (!shader) {
       /* Add base and specular colors */
       if (ctx->Fog.ColorSumEnabled ||
           (ctx->Light.Enabled &&
            ctx->Light.Model.ColorControl == GL_SEPARATE_SPECULAR_COLOR)) {
-         if (span->interpMask & SPAN_SPEC) {
-            interpolate_specular(span);
-         }
-         if (span->arrayMask & SPAN_SPEC) {
-            add_specular(ctx, span);
-         }
-         else {
-            /* We probably added the base/specular colors during the
-             * vertex stage!
-             */
-         }
+         add_specular(ctx, span);
       }
    }
 
@@ -1563,7 +1388,7 @@ _swrast_write_rgba_span( GLcontext *ctx, SWspan *span)
    }
 
    /* Clamp color/alpha values over the range [0.0, 1.0] before storage */
-   if (ctx->Color.ClampFragmentColor &&
+   if (ctx->Color.ClampFragmentColor == GL_TRUE &&
        span->array->ChanType == GL_FLOAT) {
       clamp_colors(span);
    }
@@ -1571,76 +1396,82 @@ _swrast_write_rgba_span( GLcontext *ctx, SWspan *span)
    /*
     * Write to renderbuffers
     */
-   {
-      struct gl_framebuffer *fb = ctx->DrawBuffer;
-      const GLuint output = 0; /* only frag progs can write to other outputs */
-      const GLuint numDrawBuffers = fb->_NumColorDrawBuffers[output];
-      GLchan rgbaSave[MAX_WIDTH][4];
-      GLuint buf;
-
-      if (numDrawBuffers > 0) {
-         if (fb->_ColorDrawBuffers[output][0]->DataType
-             != span->array->ChanType) {
-            convert_color_type(ctx, span,
-                               fb->_ColorDrawBuffers[output][0]->DataType);
-         }
-      }
-
-      if (numDrawBuffers > 1) {
-         /* save colors for second, third renderbuffer writes */
-         _mesa_memcpy(rgbaSave, span->array->rgba,
-                      4 * span->end * sizeof(GLchan));
-      }
-
-      for (buf = 0; buf < numDrawBuffers; buf++) {
-         struct gl_renderbuffer *rb = fb->_ColorDrawBuffers[output][buf];
-         ASSERT(rb->_BaseFormat == GL_RGBA || rb->_BaseFormat == GL_RGB);
-
-         if (ctx->Color._LogicOpEnabled) {
-            _swrast_logicop_rgba_span(ctx, rb, span);
-         }
-         else if (ctx->Color.BlendEnabled) {
-            _swrast_blend_span(ctx, rb, span);
-         }
-
-         if (colorMask != 0xffffffff) {
-            _swrast_mask_rgba_span(ctx, rb, span);
-         }
-
-         if (span->arrayMask & SPAN_XY) {
-            /* array of pixel coords */
-            ASSERT(rb->PutValues);
-            rb->PutValues(ctx, rb, span->end,
-                          span->array->x, span->array->y,
-                          span->array->rgba, span->array->mask);
-         }
-         else {
-            /* horizontal run of pixels */
-            ASSERT(rb->PutRow);
-            rb->PutRow(ctx, rb, span->end, span->x, span->y, span->array->rgba,
-                       span->writeAll ? NULL: span->array->mask);
-         }
-
-         if (buf + 1 < numDrawBuffers) {
-            /* restore original span values */
-            _mesa_memcpy(span->array->rgba, rgbaSave,
-                         4 * span->end * sizeof(GLchan));
-         }
-      } /* for buf */
-
-   }
+   /* Loop over color outputs (GL_ARB_draw_buffers) written by frag prog */
+   for (output = 0; output < swrast->_NumColorOutputs; output++) {
+      if (swrast->_ColorOutputsMask & (1 << output)) {
+        const GLuint numDrawBuffers = fb->_NumColorDrawBuffers[output];
+        GLchan rgbaSave[MAX_WIDTH][4];
+        GLuint buf;
+
+        ASSERT(numDrawBuffers > 0);
+
+        if (fb->_ColorDrawBuffers[output][0]->DataType
+            != span->array->ChanType || output > 0) {
+           convert_color_type(span,
+                              fb->_ColorDrawBuffers[output][0]->DataType,
+                              output);
+        }
+
+        if (numDrawBuffers > 1) {
+           /* save colors for second, third renderbuffer writes */
+           _mesa_memcpy(rgbaSave, span->array->rgba,
+                        4 * span->end * sizeof(GLchan));
+        }
+
+        /* Loop over renderbuffers (i.e. GL_FRONT_AND_BACK) */
+        for (buf = 0; buf < numDrawBuffers; buf++) {
+           struct gl_renderbuffer *rb = fb->_ColorDrawBuffers[output][buf];
+           ASSERT(rb->_BaseFormat == GL_RGBA || rb->_BaseFormat == GL_RGB);
+
+           if (ctx->Color._LogicOpEnabled) {
+              _swrast_logicop_rgba_span(ctx, rb, span);
+           }
+           else if (ctx->Color.BlendEnabled) {
+              _swrast_blend_span(ctx, rb, span);
+           }
+
+           if (colorMask != 0xffffffff) {
+              _swrast_mask_rgba_span(ctx, rb, span);
+           }
+
+           if (span->arrayMask & SPAN_XY) {
+              /* array of pixel coords */
+              ASSERT(rb->PutValues);
+              rb->PutValues(ctx, rb, span->end,
+                            span->array->x, span->array->y,
+                            span->array->rgba, span->array->mask);
+           }
+           else {
+              /* horizontal run of pixels */
+              ASSERT(rb->PutRow);
+              rb->PutRow(ctx, rb, span->end, span->x, span->y,
+                         span->array->rgba,
+                         span->writeAll ? NULL: span->array->mask);
+           }
+
+           if (buf + 1 < numDrawBuffers) {
+              /* restore original span values */
+              _mesa_memcpy(span->array->rgba, rgbaSave,
+                           4 * span->end * sizeof(GLchan));
+           }
+        } /* for buf */
+      } /* if output is written to */
+   } /* for output */
 
 end:
+   /* restore these values before returning */
    span->interpMask = origInterpMask;
    span->arrayMask = origArrayMask;
-   span->array->ChanType = chanType; /* restore */
+   span->arrayAttribs = origArrayAttribs;
+   span->array->ChanType = origChanType;
+   span->array->rgba = origRgba;
 }
 
 
 /**
- * Read RGBA pixels from frame buffer.  Clipping will be done to prevent
+ * Read RGBA pixels from a renderbuffer.  Clipping will be done to prevent
  * reading ouside the buffer's boundaries.
- * \param type  datatype for returned colors
+ * \param dstType  datatype for returned colors
  * \param rgba  the returned colors
  */
 void
@@ -1705,7 +1536,7 @@ _swrast_read_rgba_span( GLcontext *ctx, struct gl_renderbuffer *rb,
 
 
 /**
- * Read CI pixels from frame buffer.  Clipping will be done to prevent
+ * Read CI pixels from a renderbuffer.  Clipping will be done to prevent
  * reading ouside the buffer's boundaries.
  */
 void
@@ -1787,7 +1618,8 @@ _swrast_get_values(GLcontext *ctx, struct gl_renderbuffer *rb,
    GLuint i, inCount = 0, inStart = 0;
 
    for (i = 0; i < count; i++) {
-      if (x[i] >= 0 && y[i] >= 0 && x[i] < rb->Width && y[i] < rb->Height) {
+      if (x[i] >= 0 && y[i] >= 0 &&
+         x[i] < (GLint) rb->Width && y[i] < (GLint) rb->Height) {
          /* inside */
          if (inCount == 0)
             inStart = i;
@@ -1821,13 +1653,13 @@ _swrast_put_row(GLcontext *ctx, struct gl_renderbuffer *rb,
 {
    GLint skip = 0;
 
-   if (y < 0 || y >= rb->Height)
+   if (y < 0 || y >= (GLint) rb->Height)
       return; /* above or below */
 
-   if (x + (GLint) count <= 0 || x >= rb->Width)
+   if (x + (GLint) count <= 0 || x >= (GLint) rb->Width)
       return; /* entirely left or right */
 
-   if (x + count > rb->Width) {
+   if ((GLint) (x + count) > (GLint) rb->Width) {
       /* right clip */
       GLint clip = x + count - rb->Width;
       count -= clip;
@@ -1856,10 +1688,10 @@ _swrast_get_row(GLcontext *ctx, struct gl_renderbuffer *rb,
 {
    GLint skip = 0;
 
-   if (y < 0 || y >= rb->Height)
+   if (y < 0 || y >= (GLint) rb->Height)
       return; /* above or below */
 
-   if (x + (GLint) count <= 0 || x >= rb->Width)
+   if (x + (GLint) count <= 0 || x >= (GLint) rb->Width)
       return; /* entirely left or right */
 
    if (x + count > rb->Width) {
@@ -1894,18 +1726,9 @@ _swrast_get_dest_rgba(GLcontext *ctx, struct gl_renderbuffer *rb,
    void *rbPixels;
 
    /*
-    * Determine pixel size (in bytes).
     * Point rbPixels to a temporary space (use specular color arrays).
     */
-   if (span->array->ChanType == GL_UNSIGNED_BYTE) {
-      rbPixels = span->array->color.sz1.spec;
-   }
-   else if (span->array->ChanType == GL_UNSIGNED_SHORT) {
-      rbPixels = span->array->color.sz2.spec;
-   }
-   else {
-      rbPixels = span->array->color.sz4.spec;
-   }
+   rbPixels = span->array->attribs[FRAG_ATTRIB_COL1];
 
    /* Get destination values from renderbuffer */
    if (span->arrayMask & SPAN_XY) {