From 52011cfedaa930d61d8f60b283a2051093727582 Mon Sep 17 00:00:00 2001
From: marha <marha@users.sourceforge.net>
Date: Sun, 22 May 2011 13:28:34 +0000
Subject: xserver xkeyboard-config libX11 randrproto pixman glproto mesa git
 update 22 May 2011

---
 mesalib/src/mesa/swrast/s_span.c       | 3025 +++++++-------
 mesalib/src/mesa/swrast/s_texcombine.c | 1490 +++----
 mesalib/src/mesa/swrast/s_texfilter.c  | 7025 +++++++++++++++++---------------
 3 files changed, 5977 insertions(+), 5563 deletions(-)

(limited to 'mesalib/src/mesa/swrast')

diff --git a/mesalib/src/mesa/swrast/s_span.c b/mesalib/src/mesa/swrast/s_span.c
index f0524e061..7f88b6dd4 100644
--- a/mesalib/src/mesa/swrast/s_span.c
+++ b/mesalib/src/mesa/swrast/s_span.c
@@ -1,1509 +1,1516 @@
-/*
- * Mesa 3-D graphics library
- * Version:  7.5
- *
- * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
- * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included
- * in all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
- * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-
-
-/**
- * \file swrast/s_span.c
- * \brief Span processing functions used by all rasterization functions.
- * This is where all the per-fragment tests are performed
- * \author Brian Paul
- */
-
-#include "main/glheader.h"
-#include "main/colormac.h"
-#include "main/macros.h"
-#include "main/imports.h"
-#include "main/image.h"
-
-#include "s_atifragshader.h"
-#include "s_alpha.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_fragprog.h"
-#include "s_span.h"
-#include "s_stencil.h"
-#include "s_texcombine.h"
-
-
-/**
- * Set default fragment attributes for the span using the
- * current raster values.  Used prior to glDraw/CopyPixels
- * and glBitmap.
- */
-void
-_swrast_span_default_attribs(struct gl_context *ctx, SWspan *span)
-{
-   GLchan r, g, b, a;
-   /* Z*/
-   {
-      const GLfloat depthMax = ctx->DrawBuffer->_DepthMaxF;
-      if (ctx->DrawBuffer->Visual.depthBits <= 16)
-         span->z = FloatToFixed(ctx->Current.RasterPos[2] * depthMax + 0.5F);
-      else {
-         GLfloat tmpf = ctx->Current.RasterPos[2] * depthMax; 
-         tmpf = MIN2(tmpf, depthMax);
-         span->z = (GLint)tmpf;
-      }
-      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;
-
-   /* primary color, or color index */
-   UNCLAMPED_FLOAT_TO_CHAN(r, ctx->Current.RasterColor[0]);
-   UNCLAMPED_FLOAT_TO_CHAN(g, ctx->Current.RasterColor[1]);
-   UNCLAMPED_FLOAT_TO_CHAN(b, ctx->Current.RasterColor[2]);
-   UNCLAMPED_FLOAT_TO_CHAN(a, ctx->Current.RasterColor[3]);
-#if CHAN_TYPE == GL_FLOAT
-   span->red = r;
-   span->green = g;
-   span->blue = b;
-   span->alpha = a;
-#else
-   span->red   = IntToFixed(r);
-   span->green = IntToFixed(g);
-   span->blue  = IntToFixed(b);
-   span->alpha = IntToFixed(a);
-#endif
-   span->redStep = 0;
-   span->greenStep = 0;
-   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);
-
-   /* Secondary color */
-   if (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);
-      }
-   }
-}
-
-
-/**
- * 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]!
- */
-static INLINE void
-interpolate_active_attribs(struct gl_context *ctx, SWspan *span, GLbitfield attrMask)
-{
-   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] + span->leftClip * dv0dx;
-         GLfloat v1 = span->attrStart[attr][1] + span->leftClip * dv1dx;
-         GLfloat v2 = span->attrStart[attr][2] + span->leftClip * dv2dx;
-         GLfloat v3 = span->attrStart[attr][3] + span->leftClip * dv3dx;
-         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);
-      }
-   ATTRIB_LOOP_END
-}
-
-
-/**
- * Interpolate primary colors to fill in the span->array->rgba8 (or rgb16)
- * color array.
- */
-static INLINE void
-interpolate_int_colors(struct gl_context *ctx, SWspan *span)
-{
-   const GLuint n = span->end;
-   GLuint i;
-
-#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->rgba8;
-         if (span->interpMask & SPAN_FLAT) {
-            GLubyte color[4];
-            color[RCOMP] = FixedToInt(span->red);
-            color[GCOMP] = FixedToInt(span->green);
-            color[BCOMP] = FixedToInt(span->blue);
-            color[ACOMP] = FixedToInt(span->alpha);
-            for (i = 0; i < n; i++) {
-               COPY_4UBV(rgba[i], color);
-            }
-         }
-         else {
-            GLfixed r = span->red;
-            GLfixed g = span->green;
-            GLfixed b = span->blue;
-            GLfixed a = span->alpha;
-            GLint dr = span->redStep;
-            GLint dg = span->greenStep;
-            GLint db = span->blueStep;
-            GLint da = span->alphaStep;
-            for (i = 0; i < n; i++) {
-               rgba[i][RCOMP] = FixedToChan(r);
-               rgba[i][GCOMP] = FixedToChan(g);
-               rgba[i][BCOMP] = FixedToChan(b);
-               rgba[i][ACOMP] = FixedToChan(a);
-               r += dr;
-               g += dg;
-               b += db;
-               a += da;
-            }
-         }
-      }
-      break;
-   case GL_UNSIGNED_SHORT:
-      {
-         GLushort (*rgba)[4] = span->array->rgba16;
-         if (span->interpMask & SPAN_FLAT) {
-            GLushort color[4];
-            color[RCOMP] = FixedToInt(span->red);
-            color[GCOMP] = FixedToInt(span->green);
-            color[BCOMP] = FixedToInt(span->blue);
-            color[ACOMP] = FixedToInt(span->alpha);
-            for (i = 0; i < n; i++) {
-               COPY_4V(rgba[i], color);
-            }
-         }
-         else {
-            GLushort (*rgba)[4] = span->array->rgba16;
-            GLfixed r, g, b, a;
-            GLint dr, dg, db, da;
-            r = span->red;
-            g = span->green;
-            b = span->blue;
-            a = span->alpha;
-            dr = span->redStep;
-            dg = span->greenStep;
-            db = span->blueStep;
-            da = span->alphaStep;
-            for (i = 0; i < n; i++) {
-               rgba[i][RCOMP] = FixedToChan(r);
-               rgba[i][GCOMP] = FixedToChan(g);
-               rgba[i][BCOMP] = FixedToChan(b);
-               rgba[i][ACOMP] = FixedToChan(a);
-               r += dr;
-               g += dg;
-               b += db;
-               a += da;
-            }
-         }
-      }
-      break;
-#endif
-   case GL_FLOAT:
-      interpolate_active_attribs(ctx, span, FRAG_BIT_COL0);
-      break;
-   default:
-      _mesa_problem(ctx, "bad datatype 0x%x in interpolate_int_colors",
-                    span->array->ChanType);
-   }
-   span->arrayMask |= SPAN_RGBA;
-}
-
-
-/**
- * Populate the FRAG_ATTRIB_COL0 array.
- */
-static INLINE void
-interpolate_float_colors(SWspan *span)
-{
-   GLfloat (*col0)[4] = span->array->attribs[FRAG_ATTRIB_COL0];
-   const GLuint n = span->end;
-   GLuint i;
-
-   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]);
-      }
-   }
-   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);
-         }
-      }
-      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;
-         }
-      }
-   }
-
-   span->arrayAttribs |= FRAG_BIT_COL0;
-   span->array->ChanType = GL_FLOAT;
-}
-
-
-
-/**
- * Fill in the span.zArray array from the span->z, zStep values.
- */
-void
-_swrast_span_interpolate_z( const struct gl_context *ctx, SWspan *span )
-{
-   const GLuint n = span->end;
-   GLuint i;
-
-   ASSERT(!(span->arrayMask & SPAN_Z));
-
-   if (ctx->DrawBuffer->Visual.depthBits <= 16) {
-      GLfixed zval = span->z;
-      GLuint *z = span->array->z; 
-      for (i = 0; i < n; i++) {
-         z[i] = FixedToInt(zval);
-         zval += span->zStep;
-      }
-   }
-   else {
-      /* Deep Z buffer, no fixed->int shift */
-      GLuint zval = span->z;
-      GLuint *z = span->array->z;
-      for (i = 0; i < n; i++) {
-         z[i] = zval;
-         zval += span->zStep;
-      }
-   }
-   span->interpMask &= ~SPAN_Z;
-   span->arrayMask |= SPAN_Z;
-}
-
-
-/**
- * Compute mipmap LOD from partial derivatives.
- * This the ideal solution, as given in the OpenGL spec.
- */
-GLfloat
-_swrast_compute_lambda(GLfloat dsdx, GLfloat dsdy, GLfloat dtdx, GLfloat dtdy,
-                       GLfloat dqdx, GLfloat dqdy, GLfloat texW, GLfloat texH,
-                       GLfloat s, GLfloat t, GLfloat q, GLfloat invQ)
-{
-   GLfloat dudx = texW * ((s + dsdx) / (q + dqdx) - s * invQ);
-   GLfloat dvdx = texH * ((t + dtdx) / (q + dqdx) - t * invQ);
-   GLfloat dudy = texW * ((s + dsdy) / (q + dqdy) - s * invQ);
-   GLfloat dvdy = texH * ((t + dtdy) / (q + dqdy) - t * invQ);
-   GLfloat x = SQRTF(dudx * dudx + dvdx * dvdx);
-   GLfloat y = SQRTF(dudy * dudy + dvdy * dvdy);
-   GLfloat rho = MAX2(x, y);
-   GLfloat lambda = LOG2(rho);
-   return lambda;
-}
-
-
-/**
- * Compute mipmap LOD from partial derivatives.
- * This is a faster approximation than above function.
- */
-#if 0
-GLfloat
-_swrast_compute_lambda(GLfloat dsdx, GLfloat dsdy, GLfloat dtdx, GLfloat dtdy,
-                     GLfloat dqdx, GLfloat dqdy, GLfloat texW, GLfloat texH,
-                     GLfloat s, GLfloat t, GLfloat q, GLfloat invQ)
-{
-   GLfloat dsdx2 = (s + dsdx) / (q + dqdx) - s * invQ;
-   GLfloat dtdx2 = (t + dtdx) / (q + dqdx) - t * invQ;
-   GLfloat dsdy2 = (s + dsdy) / (q + dqdy) - s * invQ;
-   GLfloat dtdy2 = (t + dtdy) / (q + dqdy) - t * invQ;
-   GLfloat maxU, maxV, rho, lambda;
-   dsdx2 = FABSF(dsdx2);
-   dsdy2 = FABSF(dsdy2);
-   dtdx2 = FABSF(dtdx2);
-   dtdy2 = FABSF(dtdy2);
-   maxU = MAX2(dsdx2, dsdy2) * texW;
-   maxV = MAX2(dtdx2, dtdy2) * texH;
-   rho = MAX2(maxU, maxV);
-   lambda = LOG2(rho);
-   return lambda;
-}
-#endif
-
-
-/**
- * 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).
- */
-static void
-interpolate_texcoords(struct gl_context *ctx, SWspan *span)
-{
-   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] + span->leftClip * dsdx;
-         GLfloat t = span->attrStart[attr][1] + span->leftClip * dtdx;
-         GLfloat r = span->attrStart[attr][2] + span->leftClip * drdx;
-         GLfloat q = span->attrStart[attr][3] + span->leftClip * dqdx;
-
-         if (obj) {
-            const struct gl_texture_image *img = obj->Image[0][obj->BaseLevel];
-            needLambda = (obj->Sampler.MinFilter != obj->Sampler.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] + span->leftClip * dwdx;
-               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 {
-            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] + span->leftClip * dwdx;
-               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 */
-               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;
-               }
-            }
-         } /* lambda */
-      } /* if */
-   } /* for */
-}
-
-
-/**
- * Fill in the arrays->attribs[FRAG_ATTRIB_WPOS] array.
- */
-static INLINE void
-interpolate_wpos(struct gl_context *ctx, SWspan *span)
-{
-   GLfloat (*wpos)[4] = span->array->attribs[FRAG_ATTRIB_WPOS];
-   GLuint i;
-   const GLfloat zScale = 1.0F / ctx->DrawBuffer->_DepthMaxF;
-   GLfloat w, dw;
-
-   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;
-      }
-   }
-
-   dw = span->attrStepX[FRAG_ATTRIB_WPOS][3];
-   w = span->attrStart[FRAG_ATTRIB_WPOS][3] + span->leftClip * dw;
-   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 INLINE void
-stipple_polygon_span(struct gl_context *ctx, SWspan *span)
-{
-   GLubyte *mask = span->array->mask;
-
-   ASSERT(ctx->Polygon.StippleFlag);
-
-   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;
-         }
-      }
-   }
-   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;
-}
-
-
-/**
- * Clip a pixel span to the current buffer/window boundaries:
- * DrawBuffer->_Xmin, _Xmax, _Ymin, _Ymax.  This will accomplish
- * window clipping and scissoring.
- * Return:   GL_TRUE   some pixels still visible
- *           GL_FALSE  nothing visible
- */
-static INLINE GLuint
-clip_span( struct gl_context *ctx, SWspan *span )
-{
-   const GLint xmin = ctx->DrawBuffer->_Xmin;
-   const GLint xmax = ctx->DrawBuffer->_Xmax;
-   const GLint ymin = ctx->DrawBuffer->_Ymin;
-   const GLint ymax = ctx->DrawBuffer->_Ymax;
-
-   span->leftClip = 0;
-
-   if (span->arrayMask & SPAN_XY) {
-      /* arrays of x/y pixel coords */
-      const GLint *x = span->array->x;
-      const GLint *y = span->array->y;
-      const GLint n = span->end;
-      GLubyte *mask = span->array->mask;
-      GLint i;
-      if (span->arrayMask & SPAN_MASK) {
-         /* note: using & intead of && to reduce branches */
-         for (i = 0; i < n; i++) {
-            mask[i] &= (x[i] >= xmin) & (x[i] < xmax)
-                     & (y[i] >= ymin) & (y[i] < ymax);
-         }
-      }
-      else {
-         /* note: using & intead of && to reduce branches */
-         for (i = 0; i < n; i++) {
-            mask[i] = (x[i] >= xmin) & (x[i] < xmax)
-                    & (y[i] >= ymin) & (y[i] < ymax);
-         }
-      }
-      return GL_TRUE;  /* some pixels visible */
-   }
-   else {
-      /* horizontal span of pixels */
-      const GLint x = span->x;
-      const GLint y = span->y;
-      GLint n = span->end;
-
-      /* Trivial rejection tests */
-      if (y < ymin || y >= ymax || x + n <= xmin || x >= xmax) {
-         span->end = 0;
-         return GL_FALSE;  /* all pixels clipped */
-      }
-
-      /* Clip to right */
-      if (x + n > xmax) {
-         ASSERT(x < xmax);
-         n = span->end = xmax - x;
-      }
-
-      /* Clip to the left */
-      if (x < xmin) {
-         const GLint leftClip = xmin - x;
-         GLuint i;
-
-         ASSERT(leftClip > 0);
-         ASSERT(x + n > xmin);
-
-         /* Clip 'leftClip' pixels from the left side.
-          * The span->leftClip field will be applied when we interpolate
-          * fragment attributes.
-          * For arrays of values, shift them left.
-          */
-         for (i = 0; i < FRAG_ATTRIB_MAX; i++) {
-            if (span->interpMask & (1 << i)) {
-               GLuint j;
-               for (j = 0; j < 4; j++) {
-                  span->attrStart[i][j] += leftClip * span->attrStepX[i][j];
-               }
-            }
-         }
-
-         span->red += leftClip * span->redStep;
-         span->green += leftClip * span->greenStep;
-         span->blue += leftClip * span->blueStep;
-         span->alpha += leftClip * span->alphaStep;
-         span->index += leftClip * span->indexStep;
-         span->z += leftClip * span->zStep;
-         span->intTex[0] += leftClip * span->intTexStep[0];
-         span->intTex[1] += leftClip * span->intTexStep[1];
-
-#define SHIFT_ARRAY(ARRAY, SHIFT, LEN) \
-         memcpy(ARRAY, ARRAY + (SHIFT), (LEN) * sizeof(ARRAY[0]))
-
-         for (i = 0; i < FRAG_ATTRIB_MAX; i++) {
-            if (span->arrayAttribs & (1 << i)) {
-               /* shift array elements left by 'leftClip' */
-               SHIFT_ARRAY(span->array->attribs[i], leftClip, n - leftClip);
-            }
-         }
-
-         SHIFT_ARRAY(span->array->mask, leftClip, n - leftClip);
-         SHIFT_ARRAY(span->array->rgba8, leftClip, n - leftClip);
-         SHIFT_ARRAY(span->array->rgba16, leftClip, n - leftClip);
-         SHIFT_ARRAY(span->array->x, leftClip, n - leftClip);
-         SHIFT_ARRAY(span->array->y, leftClip, n - leftClip);
-         SHIFT_ARRAY(span->array->z, leftClip, n - leftClip);
-         SHIFT_ARRAY(span->array->index, leftClip, n - leftClip);
-         for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++) {
-            SHIFT_ARRAY(span->array->lambda[i], leftClip, n - leftClip);
-         }
-         SHIFT_ARRAY(span->array->coverage, leftClip, n - leftClip);
-
-#undef SHIFT_ARRAY
-
-         span->leftClip = leftClip;
-         span->x = xmin;
-         span->end -= leftClip;
-         span->writeAll = GL_FALSE;
-      }
-
-      ASSERT(span->x >= xmin);
-      ASSERT(span->x + span->end <= xmax);
-      ASSERT(span->y >= ymin);
-      ASSERT(span->y < ymax);
-
-      return GL_TRUE;  /* some pixels visible */
-   }
-}
-
-
-/**
- * Add specular colors to primary colors.
- * Only called during fixed-function operation.
- * Result is float color array (FRAG_ATTRIB_COL0).
- */
-static INLINE void
-add_specular(struct gl_context *ctx, SWspan *span)
-{
-   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);
-   (void) swrast; /* silence warning */
-
-   if (span->array->ChanType == GL_FLOAT) {
-      if ((span->arrayAttribs & FRAG_BIT_COL0) == 0) {
-         interpolate_active_attribs(ctx, span, FRAG_BIT_COL0);
-      }
-   }
-   else {
-      /* need float colors */
-      if ((span->arrayAttribs & FRAG_BIT_COL0) == 0) {
-         interpolate_float_colors(span);
-      }
-   }
-
-   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];
-      }
-   }
-
-   span->array->ChanType = GL_FLOAT;
-}
-
-
-/**
- * Apply antialiasing coverage value to alpha values.
- */
-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->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);
-         ASSERT(coverage[i] >= 0.0);
-         ASSERT(coverage[i] <= 1.0);
-      }
-   }
-   else if (span->array->ChanType == GL_UNSIGNED_SHORT) {
-      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->attribs[FRAG_ATTRIB_COL0];
-      for (i = 0; i < span->end; i++) {
-         rgba[i][ACOMP] = rgba[i][ACOMP] * coverage[i];
-         /* clamp later */
-      }
-   }
-}
-
-
-/**
- * Clamp span's float colors to [0,1]
- */
-static INLINE void
-clamp_colors(SWspan *span)
-{
-   GLfloat (*rgba)[4] = span->array->attribs[FRAG_ATTRIB_COL0];
-   GLuint i;
-   ASSERT(span->array->ChanType == GL_FLOAT);
-   for (i = 0; i < span->end; i++) {
-      rgba[i][RCOMP] = CLAMP(rgba[i][RCOMP], 0.0F, 1.0F);
-      rgba[i][GCOMP] = CLAMP(rgba[i][GCOMP], 0.0F, 1.0F);
-      rgba[i][BCOMP] = CLAMP(rgba[i][BCOMP], 0.0F, 1.0F);
-      rgba[i][ACOMP] = CLAMP(rgba[i][ACOMP], 0.0F, 1.0F);
-   }
-}
-
-
-/**
- * Convert the span's color arrays to the given type.
- * The only way 'output' can be greater than zero 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 INLINE void
-convert_color_type(SWspan *span, GLenum newType, GLuint output)
-{
-   GLvoid *src, *dst;
-
-   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->rgba8;
-   }
-   else {
-      ASSERT(span->array->ChanType == GL_UNSIGNED_SHORT);
-      src = span->array->rgba16;
-   }
-
-   if (newType == GL_UNSIGNED_BYTE) {
-      dst = span->array->rgba8;
-   }
-   else if (newType == GL_UNSIGNED_SHORT) {
-      dst = span->array->rgba16;
-   }
-   else {
-      dst = span->array->attribs[FRAG_ATTRIB_COL0];
-   }
-
-   _mesa_convert_colors(span->array->ChanType, src,
-                        newType, dst,
-                        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(struct gl_context *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);
-      }
-      else {
-         span->array->rgba = (void *) span->array->attribs[FRAG_ATTRIB_COL0];
-      }
-
-      if (span->primitive != GL_POINT ||
-	  (span->interpMask & SPAN_RGBA) ||
-	  ctx->Point.PointSprite) {
-         /* for single-pixel 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._EnabledCoordUnits) {
-      /* 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);
-   }
-}
-
-
-
-/**
- * Apply all the per-fragment operations to a span.
- * This now includes texturing (_swrast_write_texture_span() is history).
- * This function may modify any of the array values in the span.
- * span->interpMask and span->arrayMask may be changed but will be restored
- * to their original values before returning.
- */
-void
-_swrast_write_rgba_span( struct gl_context *ctx, SWspan *span)
-{
-   const SWcontext *swrast = SWRAST_CONTEXT(ctx);
-   const GLuint *colorMask = (GLuint *) ctx->Color.ColorMask;
-   const GLbitfield origInterpMask = span->interpMask;
-   const GLbitfield origArrayMask = span->arrayMask;
-   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._EnabledCoordUnits;
-   struct gl_framebuffer *fb = ctx->DrawBuffer;
-
-   /*
-   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);
-
-   /* Fragment write masks */
-   if (span->arrayMask & SPAN_MASK) {
-      /* mask was initialized by caller, probably glBitmap */
-      span->writeAll = GL_FALSE;
-   }
-   else {
-      memset(span->array->mask, 1, span->end);
-      span->writeAll = GL_TRUE;
-   }
-
-   /* Clip to window/scissor box */
-   if (!clip_span(ctx, span)) {
-      return;
-   }
-
-   ASSERT(span->end <= MAX_WIDTH);
-
-   /* Depth bounds test */
-   if (ctx->Depth.BoundsTest && fb->Visual.depthBits > 0) {
-      if (!_swrast_depth_bounds_test(ctx, span)) {
-         return;
-      }
-   }
-
-#ifdef DEBUG
-   /* Make sure all fragments are within window bounds */
-   if (span->arrayMask & SPAN_XY) {
-      /* array of pixel locations */
-      GLuint i;
-      for (i = 0; i < span->end; i++) {
-         if (span->array->mask[i]) {
-            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);
-         }
-      }
-   }
-#endif
-
-   /* Polygon Stippling */
-   if (ctx->Polygon.StippleFlag && span->primitive == GL_POLYGON) {
-      stipple_polygon_span(ctx, span);
-   }
-
-   /* This is the normal place to compute the fragment color/Z
-    * from texturing or shading.
-    */
-   if (shaderOrTexture && !swrast->_DeferredTexture) {
-      shade_texture_span(ctx, span);
-   }
-
-   /* Do the alpha test */
-   if (ctx->Color.AlphaEnabled) {
-      if (!_swrast_alpha_test(ctx, span)) {
-         /* all fragments failed test */
-         goto end;
-      }
-   }
-
-   /* Stencil and Z testing */
-   if (ctx->Stencil._Enabled || ctx->Depth.Test) {
-      if (!(span->arrayMask & SPAN_Z))
-         _swrast_span_interpolate_z(ctx, span);
-
-      if (ctx->Transform.DepthClamp)
-	 _swrast_depth_clamp_span(ctx, span);
-
-      if (ctx->Stencil._Enabled) {
-         /* Combined Z/stencil tests */
-         if (!_swrast_stencil_and_ztest_span(ctx, span)) {
-            /* all fragments failed test */
-            goto end;
-         }
-      }
-      else if (fb->Visual.depthBits > 0) {
-         /* Just regular depth testing */
-         ASSERT(ctx->Depth.Test);
-         ASSERT(span->arrayMask & SPAN_Z);
-         if (!_swrast_depth_test_span(ctx, span)) {
-            /* all fragments failed test */
-            goto end;
-         }
-      }
-   }
-
-   if (ctx->Query.CurrentOcclusionObject) {
-      /* update count of 'passed' fragments */
-      struct gl_query_object *q = ctx->Query.CurrentOcclusionObject;
-      GLuint i;
-      for (i = 0; i < span->end; i++)
-         q->Result += span->array->mask[i];
-   }
-
-   /* We had to wait until now to check for glColorMask(0,0,0,0) because of
-    * the occlusion test.
-    */
-   if (fb->_NumColorDrawBuffers == 1 && colorMask[0] == 0x0) {
-      /* no colors to write */
-      goto end;
-   }
-
-   /* If we were able to defer fragment color computation to now, there's
-    * a good chance that many fragments will have already been killed by
-    * Z/stencil testing.
-    */
-   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 (span->primitive == GL_BITMAP || !swrast->SpecularVertexAdd) {
-      /* Add primary and specular (diffuse + specular) colors */
-      if (!shader) {
-         if (ctx->Fog.ColorSumEnabled ||
-             (ctx->Light.Enabled &&
-              ctx->Light.Model.ColorControl == GL_SEPARATE_SPECULAR_COLOR)) {
-            add_specular(ctx, span);
-         }
-      }
-   }
-
-   /* Fog */
-   if (swrast->_FogEnabled) {
-      _swrast_fog_rgba_span(ctx, span);
-   }
-
-   /* Antialias coverage application */
-   if (span->arrayMask & SPAN_COVERAGE) {
-      apply_aa_coverage(span);
-   }
-
-   /* Clamp color/alpha values over the range [0.0, 1.0] before storage */
-   if (ctx->Color.ClampFragmentColor == GL_TRUE &&
-       span->array->ChanType == GL_FLOAT) {
-      clamp_colors(span);
-   }
-
-   /*
-    * Write to renderbuffers.
-    * Depending on glDrawBuffer() state and the which color outputs are
-    * written by the fragment shader, we may either replicate one color to
-    * all renderbuffers or write a different color to each renderbuffer.
-    * multiFragOutputs=TRUE for the later case.
-    */
-   {
-      const GLuint numBuffers = fb->_NumColorDrawBuffers;
-      const struct gl_fragment_program *fp = ctx->FragmentProgram._Current;
-      const GLboolean multiFragOutputs = 
-         (fp && fp->Base.OutputsWritten >= (1 << FRAG_RESULT_DATA0));
-      GLuint buf;
-
-      for (buf = 0; buf < numBuffers; buf++) {
-         struct gl_renderbuffer *rb = fb->_ColorDrawBuffers[buf];
-
-         /* color[fragOutput] will be written to buffer[buf] */
-
-         if (rb) {
-            GLchan rgbaSave[MAX_WIDTH][4];
-            const GLuint fragOutput = multiFragOutputs ? buf : 0;
-
-            /* set span->array->rgba to colors for render buffer's datatype */
-            if (rb->DataType != span->array->ChanType || fragOutput > 0) {
-               convert_color_type(span, rb->DataType, fragOutput);
-            }
-            else {
-               if (rb->DataType == GL_UNSIGNED_BYTE) {
-                  span->array->rgba = span->array->rgba8;
-               }
-               else if (rb->DataType == GL_UNSIGNED_SHORT) {
-                  span->array->rgba = (void *) span->array->rgba16;
-               }
-               else {
-                  span->array->rgba = (void *)
-                     span->array->attribs[FRAG_ATTRIB_COL0];
-               }
-            }
-
-            if (!multiFragOutputs && numBuffers > 1) {
-               /* save colors for second, third renderbuffer writes */
-               memcpy(rgbaSave, span->array->rgba,
-                      4 * span->end * sizeof(GLchan));
-            }
-
-            ASSERT(rb->_BaseFormat == GL_RGBA || rb->_BaseFormat == GL_RGB ||
-		   rb->_BaseFormat == GL_ALPHA);
-
-            if (ctx->Color._LogicOpEnabled) {
-               _swrast_logicop_rgba_span(ctx, rb, span);
-            }
-            else if ((ctx->Color.BlendEnabled >> buf) & 1) {
-               _swrast_blend_span(ctx, rb, span);
-            }
-
-            if (colorMask[buf] != 0xffffffff) {
-               _swrast_mask_rgba_span(ctx, rb, span, buf);
-            }
-
-            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 (!multiFragOutputs && numBuffers > 1) {
-               /* restore original span values */
-               memcpy(span->array->rgba, rgbaSave,
-                      4 * span->end * sizeof(GLchan));
-            }
-
-         } /* if rb */
-      } /* for buf */
-   }
-
-end:
-   /* restore these values before returning */
-   span->interpMask = origInterpMask;
-   span->arrayMask = origArrayMask;
-   span->arrayAttribs = origArrayAttribs;
-   span->array->ChanType = origChanType;
-   span->array->rgba = origRgba;
-}
-
-
-/**
- * Read RGBA pixels from a renderbuffer.  Clipping will be done to prevent
- * reading ouside the buffer's boundaries.
- * \param dstType  datatype for returned colors
- * \param rgba  the returned colors
- */
-void
-_swrast_read_rgba_span( struct gl_context *ctx, struct gl_renderbuffer *rb,
-                        GLuint n, GLint x, GLint y, GLenum dstType,
-                        GLvoid *rgba)
-{
-   const GLint bufWidth = (GLint) rb->Width;
-   const GLint bufHeight = (GLint) rb->Height;
-
-   if (y < 0 || y >= bufHeight || x + (GLint) n < 0 || x >= bufWidth) {
-      /* completely above, below, or right */
-      /* XXX maybe leave rgba values undefined? */
-      memset(rgba, 0, 4 * n * sizeof(GLchan));
-   }
-   else {
-      GLint skip, length;
-      if (x < 0) {
-         /* left edge clipping */
-         skip = -x;
-         length = (GLint) n - skip;
-         if (length < 0) {
-            /* completely left of window */
-            return;
-         }
-         if (length > bufWidth) {
-            length = bufWidth;
-         }
-      }
-      else if ((GLint) (x + n) > bufWidth) {
-         /* right edge clipping */
-         skip = 0;
-         length = bufWidth - x;
-         if (length < 0) {
-            /* completely to right of window */
-            return;
-         }
-      }
-      else {
-         /* no clipping */
-         skip = 0;
-         length = (GLint) n;
-      }
-
-      ASSERT(rb);
-      ASSERT(rb->GetRow);
-      ASSERT(rb->_BaseFormat == GL_RGBA ||
-	     rb->_BaseFormat == GL_RGB ||
-	     rb->_BaseFormat == GL_RG ||
-	     rb->_BaseFormat == GL_RED ||
-	     rb->_BaseFormat == GL_LUMINANCE ||
-	     rb->_BaseFormat == GL_INTENSITY ||
-	     rb->_BaseFormat == GL_LUMINANCE_ALPHA ||
-	     rb->_BaseFormat == GL_ALPHA);
-
-      if (rb->DataType == dstType) {
-         rb->GetRow(ctx, rb, length, x + skip, y,
-                    (GLubyte *) rgba + skip * RGBA_PIXEL_SIZE(rb->DataType));
-      }
-      else {
-         GLuint temp[MAX_WIDTH * 4];
-         rb->GetRow(ctx, rb, length, x + skip, y, temp);
-         _mesa_convert_colors(rb->DataType, temp,
-                   dstType, (GLubyte *) rgba + skip * RGBA_PIXEL_SIZE(dstType),
-                   length, NULL);
-      }
-   }
-}
-
-
-/**
- * Wrapper for gl_renderbuffer::GetValues() which does clipping to avoid
- * reading values outside the buffer bounds.
- * We can use this for reading any format/type of renderbuffer.
- * \param valueSize is the size in bytes of each value (pixel) put into the
- *                  values array.
- */
-void
-_swrast_get_values(struct gl_context *ctx, struct gl_renderbuffer *rb,
-                   GLuint count, const GLint x[], const GLint y[],
-                   void *values, GLuint valueSize)
-{
-   GLuint i, inCount = 0, inStart = 0;
-
-   for (i = 0; i < count; i++) {
-      if (x[i] >= 0 && y[i] >= 0 &&
-	  x[i] < (GLint) rb->Width && y[i] < (GLint) rb->Height) {
-         /* inside */
-         if (inCount == 0)
-            inStart = i;
-         inCount++;
-      }
-      else {
-         if (inCount > 0) {
-            /* read [inStart, inStart + inCount) */
-            rb->GetValues(ctx, rb, inCount, x + inStart, y + inStart,
-                          (GLubyte *) values + inStart * valueSize);
-            inCount = 0;
-         }
-      }
-   }
-   if (inCount > 0) {
-      /* read last values */
-      rb->GetValues(ctx, rb, inCount, x + inStart, y + inStart,
-                    (GLubyte *) values + inStart * valueSize);
-   }
-}
-
-
-/**
- * Wrapper for gl_renderbuffer::PutRow() which does clipping.
- * \param valueSize  size of each value (pixel) in bytes
- */
-void
-_swrast_put_row(struct gl_context *ctx, struct gl_renderbuffer *rb,
-                GLuint count, GLint x, GLint y,
-                const GLvoid *values, GLuint valueSize)
-{
-   GLint skip = 0;
-
-   if (y < 0 || y >= (GLint) rb->Height)
-      return; /* above or below */
-
-   if (x + (GLint) count <= 0 || x >= (GLint) rb->Width)
-      return; /* entirely left or right */
-
-   if ((GLint) (x + count) > (GLint) rb->Width) {
-      /* right clip */
-      GLint clip = x + count - rb->Width;
-      count -= clip;
-   }
-
-   if (x < 0) {
-      /* left clip */
-      skip = -x;
-      x = 0;
-      count -= skip;
-   }
-
-   rb->PutRow(ctx, rb, count, x, y,
-              (const GLubyte *) values + skip * valueSize, NULL);
-}
-
-
-/**
- * Wrapper for gl_renderbuffer::GetRow() which does clipping.
- * \param valueSize  size of each value (pixel) in bytes
- */
-void
-_swrast_get_row(struct gl_context *ctx, struct gl_renderbuffer *rb,
-                GLuint count, GLint x, GLint y,
-                GLvoid *values, GLuint valueSize)
-{
-   GLint skip = 0;
-
-   if (y < 0 || y >= (GLint) rb->Height)
-      return; /* above or below */
-
-   if (x + (GLint) count <= 0 || x >= (GLint) rb->Width)
-      return; /* entirely left or right */
-
-   if (x + count > rb->Width) {
-      /* right clip */
-      GLint clip = x + count - rb->Width;
-      count -= clip;
-   }
-
-   if (x < 0) {
-      /* left clip */
-      skip = -x;
-      x = 0;
-      count -= skip;
-   }
-
-   rb->GetRow(ctx, rb, count, x, y, (GLubyte *) values + skip * valueSize);
-}
-
-
-/**
- * Get RGBA pixels from the given renderbuffer.
- * Used by blending, logicop and masking functions.
- * \return pointer to the colors we read.
- */
-void *
-_swrast_get_dest_rgba(struct gl_context *ctx, struct gl_renderbuffer *rb,
-                      SWspan *span)
-{
-   const GLuint pixelSize = RGBA_PIXEL_SIZE(span->array->ChanType);
-   void *rbPixels;
-
-   /* Point rbPixels to a temporary space */
-   rbPixels = span->array->attribs[FRAG_ATTRIB_MAX - 1];
-
-   /* Get destination values from renderbuffer */
-   if (span->arrayMask & SPAN_XY) {
-      _swrast_get_values(ctx, rb, span->end, span->array->x, span->array->y,
-                         rbPixels, pixelSize);
-   }
-   else {
-      _swrast_get_row(ctx, rb, span->end, span->x, span->y,
-                      rbPixels, pixelSize);
-   }
-
-   return rbPixels;
-}
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.5
+ *
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+/**
+ * \file swrast/s_span.c
+ * \brief Span processing functions used by all rasterization functions.
+ * This is where all the per-fragment tests are performed
+ * \author Brian Paul
+ */
+
+#include "main/glheader.h"
+#include "main/colormac.h"
+#include "main/macros.h"
+#include "main/imports.h"
+#include "main/image.h"
+
+#include "s_atifragshader.h"
+#include "s_alpha.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_fragprog.h"
+#include "s_span.h"
+#include "s_stencil.h"
+#include "s_texcombine.h"
+
+
+/**
+ * Set default fragment attributes for the span using the
+ * current raster values.  Used prior to glDraw/CopyPixels
+ * and glBitmap.
+ */
+void
+_swrast_span_default_attribs(struct gl_context *ctx, SWspan *span)
+{
+   GLchan r, g, b, a;
+   /* Z*/
+   {
+      const GLfloat depthMax = ctx->DrawBuffer->_DepthMaxF;
+      if (ctx->DrawBuffer->Visual.depthBits <= 16)
+         span->z = FloatToFixed(ctx->Current.RasterPos[2] * depthMax + 0.5F);
+      else {
+         GLfloat tmpf = ctx->Current.RasterPos[2] * depthMax; 
+         tmpf = MIN2(tmpf, depthMax);
+         span->z = (GLint)tmpf;
+      }
+      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;
+
+   /* primary color, or color index */
+   UNCLAMPED_FLOAT_TO_CHAN(r, ctx->Current.RasterColor[0]);
+   UNCLAMPED_FLOAT_TO_CHAN(g, ctx->Current.RasterColor[1]);
+   UNCLAMPED_FLOAT_TO_CHAN(b, ctx->Current.RasterColor[2]);
+   UNCLAMPED_FLOAT_TO_CHAN(a, ctx->Current.RasterColor[3]);
+#if CHAN_TYPE == GL_FLOAT
+   span->red = r;
+   span->green = g;
+   span->blue = b;
+   span->alpha = a;
+#else
+   span->red   = IntToFixed(r);
+   span->green = IntToFixed(g);
+   span->blue  = IntToFixed(b);
+   span->alpha = IntToFixed(a);
+#endif
+   span->redStep = 0;
+   span->greenStep = 0;
+   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);
+
+   /* Secondary color */
+   if (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);
+      }
+   }
+}
+
+
+/**
+ * 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]!
+ */
+static INLINE void
+interpolate_active_attribs(struct gl_context *ctx, SWspan *span, GLbitfield attrMask)
+{
+   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] + span->leftClip * dv0dx;
+         GLfloat v1 = span->attrStart[attr][1] + span->leftClip * dv1dx;
+         GLfloat v2 = span->attrStart[attr][2] + span->leftClip * dv2dx;
+         GLfloat v3 = span->attrStart[attr][3] + span->leftClip * dv3dx;
+         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);
+      }
+   ATTRIB_LOOP_END
+}
+
+
+/**
+ * Interpolate primary colors to fill in the span->array->rgba8 (or rgb16)
+ * color array.
+ */
+static INLINE void
+interpolate_int_colors(struct gl_context *ctx, SWspan *span)
+{
+   const GLuint n = span->end;
+   GLuint i;
+
+#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->rgba8;
+         if (span->interpMask & SPAN_FLAT) {
+            GLubyte color[4];
+            color[RCOMP] = FixedToInt(span->red);
+            color[GCOMP] = FixedToInt(span->green);
+            color[BCOMP] = FixedToInt(span->blue);
+            color[ACOMP] = FixedToInt(span->alpha);
+            for (i = 0; i < n; i++) {
+               COPY_4UBV(rgba[i], color);
+            }
+         }
+         else {
+            GLfixed r = span->red;
+            GLfixed g = span->green;
+            GLfixed b = span->blue;
+            GLfixed a = span->alpha;
+            GLint dr = span->redStep;
+            GLint dg = span->greenStep;
+            GLint db = span->blueStep;
+            GLint da = span->alphaStep;
+            for (i = 0; i < n; i++) {
+               rgba[i][RCOMP] = FixedToChan(r);
+               rgba[i][GCOMP] = FixedToChan(g);
+               rgba[i][BCOMP] = FixedToChan(b);
+               rgba[i][ACOMP] = FixedToChan(a);
+               r += dr;
+               g += dg;
+               b += db;
+               a += da;
+            }
+         }
+      }
+      break;
+   case GL_UNSIGNED_SHORT:
+      {
+         GLushort (*rgba)[4] = span->array->rgba16;
+         if (span->interpMask & SPAN_FLAT) {
+            GLushort color[4];
+            color[RCOMP] = FixedToInt(span->red);
+            color[GCOMP] = FixedToInt(span->green);
+            color[BCOMP] = FixedToInt(span->blue);
+            color[ACOMP] = FixedToInt(span->alpha);
+            for (i = 0; i < n; i++) {
+               COPY_4V(rgba[i], color);
+            }
+         }
+         else {
+            GLushort (*rgba)[4] = span->array->rgba16;
+            GLfixed r, g, b, a;
+            GLint dr, dg, db, da;
+            r = span->red;
+            g = span->green;
+            b = span->blue;
+            a = span->alpha;
+            dr = span->redStep;
+            dg = span->greenStep;
+            db = span->blueStep;
+            da = span->alphaStep;
+            for (i = 0; i < n; i++) {
+               rgba[i][RCOMP] = FixedToChan(r);
+               rgba[i][GCOMP] = FixedToChan(g);
+               rgba[i][BCOMP] = FixedToChan(b);
+               rgba[i][ACOMP] = FixedToChan(a);
+               r += dr;
+               g += dg;
+               b += db;
+               a += da;
+            }
+         }
+      }
+      break;
+#endif
+   case GL_FLOAT:
+      interpolate_active_attribs(ctx, span, FRAG_BIT_COL0);
+      break;
+   default:
+      _mesa_problem(ctx, "bad datatype 0x%x in interpolate_int_colors",
+                    span->array->ChanType);
+   }
+   span->arrayMask |= SPAN_RGBA;
+}
+
+
+/**
+ * Populate the FRAG_ATTRIB_COL0 array.
+ */
+static INLINE void
+interpolate_float_colors(SWspan *span)
+{
+   GLfloat (*col0)[4] = span->array->attribs[FRAG_ATTRIB_COL0];
+   const GLuint n = span->end;
+   GLuint i;
+
+   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]);
+      }
+   }
+   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);
+         }
+      }
+      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;
+         }
+      }
+   }
+
+   span->arrayAttribs |= FRAG_BIT_COL0;
+   span->array->ChanType = GL_FLOAT;
+}
+
+
+
+/**
+ * Fill in the span.zArray array from the span->z, zStep values.
+ */
+void
+_swrast_span_interpolate_z( const struct gl_context *ctx, SWspan *span )
+{
+   const GLuint n = span->end;
+   GLuint i;
+
+   ASSERT(!(span->arrayMask & SPAN_Z));
+
+   if (ctx->DrawBuffer->Visual.depthBits <= 16) {
+      GLfixed zval = span->z;
+      GLuint *z = span->array->z; 
+      for (i = 0; i < n; i++) {
+         z[i] = FixedToInt(zval);
+         zval += span->zStep;
+      }
+   }
+   else {
+      /* Deep Z buffer, no fixed->int shift */
+      GLuint zval = span->z;
+      GLuint *z = span->array->z;
+      for (i = 0; i < n; i++) {
+         z[i] = zval;
+         zval += span->zStep;
+      }
+   }
+   span->interpMask &= ~SPAN_Z;
+   span->arrayMask |= SPAN_Z;
+}
+
+
+/**
+ * Compute mipmap LOD from partial derivatives.
+ * This the ideal solution, as given in the OpenGL spec.
+ */
+GLfloat
+_swrast_compute_lambda(GLfloat dsdx, GLfloat dsdy, GLfloat dtdx, GLfloat dtdy,
+                       GLfloat dqdx, GLfloat dqdy, GLfloat texW, GLfloat texH,
+                       GLfloat s, GLfloat t, GLfloat q, GLfloat invQ)
+{
+   GLfloat dudx = texW * ((s + dsdx) / (q + dqdx) - s * invQ);
+   GLfloat dvdx = texH * ((t + dtdx) / (q + dqdx) - t * invQ);
+   GLfloat dudy = texW * ((s + dsdy) / (q + dqdy) - s * invQ);
+   GLfloat dvdy = texH * ((t + dtdy) / (q + dqdy) - t * invQ);
+   GLfloat x = SQRTF(dudx * dudx + dvdx * dvdx);
+   GLfloat y = SQRTF(dudy * dudy + dvdy * dvdy);
+   GLfloat rho = MAX2(x, y);
+   GLfloat lambda = LOG2(rho);
+   return lambda;
+}
+
+
+/**
+ * Compute mipmap LOD from partial derivatives.
+ * This is a faster approximation than above function.
+ */
+#if 0
+GLfloat
+_swrast_compute_lambda(GLfloat dsdx, GLfloat dsdy, GLfloat dtdx, GLfloat dtdy,
+                     GLfloat dqdx, GLfloat dqdy, GLfloat texW, GLfloat texH,
+                     GLfloat s, GLfloat t, GLfloat q, GLfloat invQ)
+{
+   GLfloat dsdx2 = (s + dsdx) / (q + dqdx) - s * invQ;
+   GLfloat dtdx2 = (t + dtdx) / (q + dqdx) - t * invQ;
+   GLfloat dsdy2 = (s + dsdy) / (q + dqdy) - s * invQ;
+   GLfloat dtdy2 = (t + dtdy) / (q + dqdy) - t * invQ;
+   GLfloat maxU, maxV, rho, lambda;
+   dsdx2 = FABSF(dsdx2);
+   dsdy2 = FABSF(dsdy2);
+   dtdx2 = FABSF(dtdx2);
+   dtdy2 = FABSF(dtdy2);
+   maxU = MAX2(dsdx2, dsdy2) * texW;
+   maxV = MAX2(dtdx2, dtdy2) * texH;
+   rho = MAX2(maxU, maxV);
+   lambda = LOG2(rho);
+   return lambda;
+}
+#endif
+
+
+/**
+ * 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).
+ */
+static void
+interpolate_texcoords(struct gl_context *ctx, SWspan *span)
+{
+   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] + span->leftClip * dsdx;
+         GLfloat t = span->attrStart[attr][1] + span->leftClip * dtdx;
+         GLfloat r = span->attrStart[attr][2] + span->leftClip * drdx;
+         GLfloat q = span->attrStart[attr][3] + span->leftClip * dqdx;
+
+         if (obj) {
+            const struct gl_texture_image *img = obj->Image[0][obj->BaseLevel];
+            needLambda = (obj->Sampler.MinFilter != obj->Sampler.MagFilter)
+               || ctx->FragmentProgram._Current;
+            /* LOD is calculated directly in the ansiotropic filter, we can
+             * skip the normal lambda function as the result is ignored.
+             */
+            if (obj->Sampler.MaxAnisotropy > 1.0 &&
+                obj->Sampler.MinFilter == GL_LINEAR_MIPMAP_LINEAR) {
+               needLambda = GL_FALSE;
+            }
+            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] + span->leftClip * dwdx;
+               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 {
+            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] + span->leftClip * dwdx;
+               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 */
+               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;
+               }
+            }
+         } /* lambda */
+      } /* if */
+   } /* for */
+}
+
+
+/**
+ * Fill in the arrays->attribs[FRAG_ATTRIB_WPOS] array.
+ */
+static INLINE void
+interpolate_wpos(struct gl_context *ctx, SWspan *span)
+{
+   GLfloat (*wpos)[4] = span->array->attribs[FRAG_ATTRIB_WPOS];
+   GLuint i;
+   const GLfloat zScale = 1.0F / ctx->DrawBuffer->_DepthMaxF;
+   GLfloat w, dw;
+
+   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;
+      }
+   }
+
+   dw = span->attrStepX[FRAG_ATTRIB_WPOS][3];
+   w = span->attrStart[FRAG_ATTRIB_WPOS][3] + span->leftClip * dw;
+   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 INLINE void
+stipple_polygon_span(struct gl_context *ctx, SWspan *span)
+{
+   GLubyte *mask = span->array->mask;
+
+   ASSERT(ctx->Polygon.StippleFlag);
+
+   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;
+         }
+      }
+   }
+   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;
+}
+
+
+/**
+ * Clip a pixel span to the current buffer/window boundaries:
+ * DrawBuffer->_Xmin, _Xmax, _Ymin, _Ymax.  This will accomplish
+ * window clipping and scissoring.
+ * Return:   GL_TRUE   some pixels still visible
+ *           GL_FALSE  nothing visible
+ */
+static INLINE GLuint
+clip_span( struct gl_context *ctx, SWspan *span )
+{
+   const GLint xmin = ctx->DrawBuffer->_Xmin;
+   const GLint xmax = ctx->DrawBuffer->_Xmax;
+   const GLint ymin = ctx->DrawBuffer->_Ymin;
+   const GLint ymax = ctx->DrawBuffer->_Ymax;
+
+   span->leftClip = 0;
+
+   if (span->arrayMask & SPAN_XY) {
+      /* arrays of x/y pixel coords */
+      const GLint *x = span->array->x;
+      const GLint *y = span->array->y;
+      const GLint n = span->end;
+      GLubyte *mask = span->array->mask;
+      GLint i;
+      if (span->arrayMask & SPAN_MASK) {
+         /* note: using & intead of && to reduce branches */
+         for (i = 0; i < n; i++) {
+            mask[i] &= (x[i] >= xmin) & (x[i] < xmax)
+                     & (y[i] >= ymin) & (y[i] < ymax);
+         }
+      }
+      else {
+         /* note: using & intead of && to reduce branches */
+         for (i = 0; i < n; i++) {
+            mask[i] = (x[i] >= xmin) & (x[i] < xmax)
+                    & (y[i] >= ymin) & (y[i] < ymax);
+         }
+      }
+      return GL_TRUE;  /* some pixels visible */
+   }
+   else {
+      /* horizontal span of pixels */
+      const GLint x = span->x;
+      const GLint y = span->y;
+      GLint n = span->end;
+
+      /* Trivial rejection tests */
+      if (y < ymin || y >= ymax || x + n <= xmin || x >= xmax) {
+         span->end = 0;
+         return GL_FALSE;  /* all pixels clipped */
+      }
+
+      /* Clip to right */
+      if (x + n > xmax) {
+         ASSERT(x < xmax);
+         n = span->end = xmax - x;
+      }
+
+      /* Clip to the left */
+      if (x < xmin) {
+         const GLint leftClip = xmin - x;
+         GLuint i;
+
+         ASSERT(leftClip > 0);
+         ASSERT(x + n > xmin);
+
+         /* Clip 'leftClip' pixels from the left side.
+          * The span->leftClip field will be applied when we interpolate
+          * fragment attributes.
+          * For arrays of values, shift them left.
+          */
+         for (i = 0; i < FRAG_ATTRIB_MAX; i++) {
+            if (span->interpMask & (1 << i)) {
+               GLuint j;
+               for (j = 0; j < 4; j++) {
+                  span->attrStart[i][j] += leftClip * span->attrStepX[i][j];
+               }
+            }
+         }
+
+         span->red += leftClip * span->redStep;
+         span->green += leftClip * span->greenStep;
+         span->blue += leftClip * span->blueStep;
+         span->alpha += leftClip * span->alphaStep;
+         span->index += leftClip * span->indexStep;
+         span->z += leftClip * span->zStep;
+         span->intTex[0] += leftClip * span->intTexStep[0];
+         span->intTex[1] += leftClip * span->intTexStep[1];
+
+#define SHIFT_ARRAY(ARRAY, SHIFT, LEN) \
+         memcpy(ARRAY, ARRAY + (SHIFT), (LEN) * sizeof(ARRAY[0]))
+
+         for (i = 0; i < FRAG_ATTRIB_MAX; i++) {
+            if (span->arrayAttribs & (1 << i)) {
+               /* shift array elements left by 'leftClip' */
+               SHIFT_ARRAY(span->array->attribs[i], leftClip, n - leftClip);
+            }
+         }
+
+         SHIFT_ARRAY(span->array->mask, leftClip, n - leftClip);
+         SHIFT_ARRAY(span->array->rgba8, leftClip, n - leftClip);
+         SHIFT_ARRAY(span->array->rgba16, leftClip, n - leftClip);
+         SHIFT_ARRAY(span->array->x, leftClip, n - leftClip);
+         SHIFT_ARRAY(span->array->y, leftClip, n - leftClip);
+         SHIFT_ARRAY(span->array->z, leftClip, n - leftClip);
+         SHIFT_ARRAY(span->array->index, leftClip, n - leftClip);
+         for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++) {
+            SHIFT_ARRAY(span->array->lambda[i], leftClip, n - leftClip);
+         }
+         SHIFT_ARRAY(span->array->coverage, leftClip, n - leftClip);
+
+#undef SHIFT_ARRAY
+
+         span->leftClip = leftClip;
+         span->x = xmin;
+         span->end -= leftClip;
+         span->writeAll = GL_FALSE;
+      }
+
+      ASSERT(span->x >= xmin);
+      ASSERT(span->x + span->end <= xmax);
+      ASSERT(span->y >= ymin);
+      ASSERT(span->y < ymax);
+
+      return GL_TRUE;  /* some pixels visible */
+   }
+}
+
+
+/**
+ * Add specular colors to primary colors.
+ * Only called during fixed-function operation.
+ * Result is float color array (FRAG_ATTRIB_COL0).
+ */
+static INLINE void
+add_specular(struct gl_context *ctx, SWspan *span)
+{
+   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);
+   (void) swrast; /* silence warning */
+
+   if (span->array->ChanType == GL_FLOAT) {
+      if ((span->arrayAttribs & FRAG_BIT_COL0) == 0) {
+         interpolate_active_attribs(ctx, span, FRAG_BIT_COL0);
+      }
+   }
+   else {
+      /* need float colors */
+      if ((span->arrayAttribs & FRAG_BIT_COL0) == 0) {
+         interpolate_float_colors(span);
+      }
+   }
+
+   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];
+      }
+   }
+
+   span->array->ChanType = GL_FLOAT;
+}
+
+
+/**
+ * Apply antialiasing coverage value to alpha values.
+ */
+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->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);
+         ASSERT(coverage[i] >= 0.0);
+         ASSERT(coverage[i] <= 1.0);
+      }
+   }
+   else if (span->array->ChanType == GL_UNSIGNED_SHORT) {
+      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->attribs[FRAG_ATTRIB_COL0];
+      for (i = 0; i < span->end; i++) {
+         rgba[i][ACOMP] = rgba[i][ACOMP] * coverage[i];
+         /* clamp later */
+      }
+   }
+}
+
+
+/**
+ * Clamp span's float colors to [0,1]
+ */
+static INLINE void
+clamp_colors(SWspan *span)
+{
+   GLfloat (*rgba)[4] = span->array->attribs[FRAG_ATTRIB_COL0];
+   GLuint i;
+   ASSERT(span->array->ChanType == GL_FLOAT);
+   for (i = 0; i < span->end; i++) {
+      rgba[i][RCOMP] = CLAMP(rgba[i][RCOMP], 0.0F, 1.0F);
+      rgba[i][GCOMP] = CLAMP(rgba[i][GCOMP], 0.0F, 1.0F);
+      rgba[i][BCOMP] = CLAMP(rgba[i][BCOMP], 0.0F, 1.0F);
+      rgba[i][ACOMP] = CLAMP(rgba[i][ACOMP], 0.0F, 1.0F);
+   }
+}
+
+
+/**
+ * Convert the span's color arrays to the given type.
+ * The only way 'output' can be greater than zero 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 INLINE void
+convert_color_type(SWspan *span, GLenum newType, GLuint output)
+{
+   GLvoid *src, *dst;
+
+   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->rgba8;
+   }
+   else {
+      ASSERT(span->array->ChanType == GL_UNSIGNED_SHORT);
+      src = span->array->rgba16;
+   }
+
+   if (newType == GL_UNSIGNED_BYTE) {
+      dst = span->array->rgba8;
+   }
+   else if (newType == GL_UNSIGNED_SHORT) {
+      dst = span->array->rgba16;
+   }
+   else {
+      dst = span->array->attribs[FRAG_ATTRIB_COL0];
+   }
+
+   _mesa_convert_colors(span->array->ChanType, src,
+                        newType, dst,
+                        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(struct gl_context *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);
+      }
+      else {
+         span->array->rgba = (void *) span->array->attribs[FRAG_ATTRIB_COL0];
+      }
+
+      if (span->primitive != GL_POINT ||
+	  (span->interpMask & SPAN_RGBA) ||
+	  ctx->Point.PointSprite) {
+         /* for single-pixel 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._EnabledCoordUnits) {
+      /* 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);
+   }
+}
+
+
+
+/**
+ * Apply all the per-fragment operations to a span.
+ * This now includes texturing (_swrast_write_texture_span() is history).
+ * This function may modify any of the array values in the span.
+ * span->interpMask and span->arrayMask may be changed but will be restored
+ * to their original values before returning.
+ */
+void
+_swrast_write_rgba_span( struct gl_context *ctx, SWspan *span)
+{
+   const SWcontext *swrast = SWRAST_CONTEXT(ctx);
+   const GLuint *colorMask = (GLuint *) ctx->Color.ColorMask;
+   const GLbitfield origInterpMask = span->interpMask;
+   const GLbitfield origArrayMask = span->arrayMask;
+   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._EnabledCoordUnits;
+   struct gl_framebuffer *fb = ctx->DrawBuffer;
+
+   /*
+   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);
+
+   /* Fragment write masks */
+   if (span->arrayMask & SPAN_MASK) {
+      /* mask was initialized by caller, probably glBitmap */
+      span->writeAll = GL_FALSE;
+   }
+   else {
+      memset(span->array->mask, 1, span->end);
+      span->writeAll = GL_TRUE;
+   }
+
+   /* Clip to window/scissor box */
+   if (!clip_span(ctx, span)) {
+      return;
+   }
+
+   ASSERT(span->end <= MAX_WIDTH);
+
+   /* Depth bounds test */
+   if (ctx->Depth.BoundsTest && fb->Visual.depthBits > 0) {
+      if (!_swrast_depth_bounds_test(ctx, span)) {
+         return;
+      }
+   }
+
+#ifdef DEBUG
+   /* Make sure all fragments are within window bounds */
+   if (span->arrayMask & SPAN_XY) {
+      /* array of pixel locations */
+      GLuint i;
+      for (i = 0; i < span->end; i++) {
+         if (span->array->mask[i]) {
+            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);
+         }
+      }
+   }
+#endif
+
+   /* Polygon Stippling */
+   if (ctx->Polygon.StippleFlag && span->primitive == GL_POLYGON) {
+      stipple_polygon_span(ctx, span);
+   }
+
+   /* This is the normal place to compute the fragment color/Z
+    * from texturing or shading.
+    */
+   if (shaderOrTexture && !swrast->_DeferredTexture) {
+      shade_texture_span(ctx, span);
+   }
+
+   /* Do the alpha test */
+   if (ctx->Color.AlphaEnabled) {
+      if (!_swrast_alpha_test(ctx, span)) {
+         /* all fragments failed test */
+         goto end;
+      }
+   }
+
+   /* Stencil and Z testing */
+   if (ctx->Stencil._Enabled || ctx->Depth.Test) {
+      if (!(span->arrayMask & SPAN_Z))
+         _swrast_span_interpolate_z(ctx, span);
+
+      if (ctx->Transform.DepthClamp)
+	 _swrast_depth_clamp_span(ctx, span);
+
+      if (ctx->Stencil._Enabled) {
+         /* Combined Z/stencil tests */
+         if (!_swrast_stencil_and_ztest_span(ctx, span)) {
+            /* all fragments failed test */
+            goto end;
+         }
+      }
+      else if (fb->Visual.depthBits > 0) {
+         /* Just regular depth testing */
+         ASSERT(ctx->Depth.Test);
+         ASSERT(span->arrayMask & SPAN_Z);
+         if (!_swrast_depth_test_span(ctx, span)) {
+            /* all fragments failed test */
+            goto end;
+         }
+      }
+   }
+
+   if (ctx->Query.CurrentOcclusionObject) {
+      /* update count of 'passed' fragments */
+      struct gl_query_object *q = ctx->Query.CurrentOcclusionObject;
+      GLuint i;
+      for (i = 0; i < span->end; i++)
+         q->Result += span->array->mask[i];
+   }
+
+   /* We had to wait until now to check for glColorMask(0,0,0,0) because of
+    * the occlusion test.
+    */
+   if (fb->_NumColorDrawBuffers == 1 && colorMask[0] == 0x0) {
+      /* no colors to write */
+      goto end;
+   }
+
+   /* If we were able to defer fragment color computation to now, there's
+    * a good chance that many fragments will have already been killed by
+    * Z/stencil testing.
+    */
+   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 (span->primitive == GL_BITMAP || !swrast->SpecularVertexAdd) {
+      /* Add primary and specular (diffuse + specular) colors */
+      if (!shader) {
+         if (ctx->Fog.ColorSumEnabled ||
+             (ctx->Light.Enabled &&
+              ctx->Light.Model.ColorControl == GL_SEPARATE_SPECULAR_COLOR)) {
+            add_specular(ctx, span);
+         }
+      }
+   }
+
+   /* Fog */
+   if (swrast->_FogEnabled) {
+      _swrast_fog_rgba_span(ctx, span);
+   }
+
+   /* Antialias coverage application */
+   if (span->arrayMask & SPAN_COVERAGE) {
+      apply_aa_coverage(span);
+   }
+
+   /* Clamp color/alpha values over the range [0.0, 1.0] before storage */
+   if (ctx->Color.ClampFragmentColor == GL_TRUE &&
+       span->array->ChanType == GL_FLOAT) {
+      clamp_colors(span);
+   }
+
+   /*
+    * Write to renderbuffers.
+    * Depending on glDrawBuffer() state and the which color outputs are
+    * written by the fragment shader, we may either replicate one color to
+    * all renderbuffers or write a different color to each renderbuffer.
+    * multiFragOutputs=TRUE for the later case.
+    */
+   {
+      const GLuint numBuffers = fb->_NumColorDrawBuffers;
+      const struct gl_fragment_program *fp = ctx->FragmentProgram._Current;
+      const GLboolean multiFragOutputs = 
+         (fp && fp->Base.OutputsWritten >= (1 << FRAG_RESULT_DATA0));
+      GLuint buf;
+
+      for (buf = 0; buf < numBuffers; buf++) {
+         struct gl_renderbuffer *rb = fb->_ColorDrawBuffers[buf];
+
+         /* color[fragOutput] will be written to buffer[buf] */
+
+         if (rb) {
+            GLchan rgbaSave[MAX_WIDTH][4];
+            const GLuint fragOutput = multiFragOutputs ? buf : 0;
+
+            /* set span->array->rgba to colors for render buffer's datatype */
+            if (rb->DataType != span->array->ChanType || fragOutput > 0) {
+               convert_color_type(span, rb->DataType, fragOutput);
+            }
+            else {
+               if (rb->DataType == GL_UNSIGNED_BYTE) {
+                  span->array->rgba = span->array->rgba8;
+               }
+               else if (rb->DataType == GL_UNSIGNED_SHORT) {
+                  span->array->rgba = (void *) span->array->rgba16;
+               }
+               else {
+                  span->array->rgba = (void *)
+                     span->array->attribs[FRAG_ATTRIB_COL0];
+               }
+            }
+
+            if (!multiFragOutputs && numBuffers > 1) {
+               /* save colors for second, third renderbuffer writes */
+               memcpy(rgbaSave, span->array->rgba,
+                      4 * span->end * sizeof(GLchan));
+            }
+
+            ASSERT(rb->_BaseFormat == GL_RGBA || rb->_BaseFormat == GL_RGB ||
+		   rb->_BaseFormat == GL_ALPHA);
+
+            if (ctx->Color._LogicOpEnabled) {
+               _swrast_logicop_rgba_span(ctx, rb, span);
+            }
+            else if ((ctx->Color.BlendEnabled >> buf) & 1) {
+               _swrast_blend_span(ctx, rb, span);
+            }
+
+            if (colorMask[buf] != 0xffffffff) {
+               _swrast_mask_rgba_span(ctx, rb, span, buf);
+            }
+
+            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 (!multiFragOutputs && numBuffers > 1) {
+               /* restore original span values */
+               memcpy(span->array->rgba, rgbaSave,
+                      4 * span->end * sizeof(GLchan));
+            }
+
+         } /* if rb */
+      } /* for buf */
+   }
+
+end:
+   /* restore these values before returning */
+   span->interpMask = origInterpMask;
+   span->arrayMask = origArrayMask;
+   span->arrayAttribs = origArrayAttribs;
+   span->array->ChanType = origChanType;
+   span->array->rgba = origRgba;
+}
+
+
+/**
+ * Read RGBA pixels from a renderbuffer.  Clipping will be done to prevent
+ * reading ouside the buffer's boundaries.
+ * \param dstType  datatype for returned colors
+ * \param rgba  the returned colors
+ */
+void
+_swrast_read_rgba_span( struct gl_context *ctx, struct gl_renderbuffer *rb,
+                        GLuint n, GLint x, GLint y, GLenum dstType,
+                        GLvoid *rgba)
+{
+   const GLint bufWidth = (GLint) rb->Width;
+   const GLint bufHeight = (GLint) rb->Height;
+
+   if (y < 0 || y >= bufHeight || x + (GLint) n < 0 || x >= bufWidth) {
+      /* completely above, below, or right */
+      /* XXX maybe leave rgba values undefined? */
+      memset(rgba, 0, 4 * n * sizeof(GLchan));
+   }
+   else {
+      GLint skip, length;
+      if (x < 0) {
+         /* left edge clipping */
+         skip = -x;
+         length = (GLint) n - skip;
+         if (length < 0) {
+            /* completely left of window */
+            return;
+         }
+         if (length > bufWidth) {
+            length = bufWidth;
+         }
+      }
+      else if ((GLint) (x + n) > bufWidth) {
+         /* right edge clipping */
+         skip = 0;
+         length = bufWidth - x;
+         if (length < 0) {
+            /* completely to right of window */
+            return;
+         }
+      }
+      else {
+         /* no clipping */
+         skip = 0;
+         length = (GLint) n;
+      }
+
+      ASSERT(rb);
+      ASSERT(rb->GetRow);
+      ASSERT(rb->_BaseFormat == GL_RGBA ||
+	     rb->_BaseFormat == GL_RGB ||
+	     rb->_BaseFormat == GL_RG ||
+	     rb->_BaseFormat == GL_RED ||
+	     rb->_BaseFormat == GL_LUMINANCE ||
+	     rb->_BaseFormat == GL_INTENSITY ||
+	     rb->_BaseFormat == GL_LUMINANCE_ALPHA ||
+	     rb->_BaseFormat == GL_ALPHA);
+
+      if (rb->DataType == dstType) {
+         rb->GetRow(ctx, rb, length, x + skip, y,
+                    (GLubyte *) rgba + skip * RGBA_PIXEL_SIZE(rb->DataType));
+      }
+      else {
+         GLuint temp[MAX_WIDTH * 4];
+         rb->GetRow(ctx, rb, length, x + skip, y, temp);
+         _mesa_convert_colors(rb->DataType, temp,
+                   dstType, (GLubyte *) rgba + skip * RGBA_PIXEL_SIZE(dstType),
+                   length, NULL);
+      }
+   }
+}
+
+
+/**
+ * Wrapper for gl_renderbuffer::GetValues() which does clipping to avoid
+ * reading values outside the buffer bounds.
+ * We can use this for reading any format/type of renderbuffer.
+ * \param valueSize is the size in bytes of each value (pixel) put into the
+ *                  values array.
+ */
+void
+_swrast_get_values(struct gl_context *ctx, struct gl_renderbuffer *rb,
+                   GLuint count, const GLint x[], const GLint y[],
+                   void *values, GLuint valueSize)
+{
+   GLuint i, inCount = 0, inStart = 0;
+
+   for (i = 0; i < count; i++) {
+      if (x[i] >= 0 && y[i] >= 0 &&
+	  x[i] < (GLint) rb->Width && y[i] < (GLint) rb->Height) {
+         /* inside */
+         if (inCount == 0)
+            inStart = i;
+         inCount++;
+      }
+      else {
+         if (inCount > 0) {
+            /* read [inStart, inStart + inCount) */
+            rb->GetValues(ctx, rb, inCount, x + inStart, y + inStart,
+                          (GLubyte *) values + inStart * valueSize);
+            inCount = 0;
+         }
+      }
+   }
+   if (inCount > 0) {
+      /* read last values */
+      rb->GetValues(ctx, rb, inCount, x + inStart, y + inStart,
+                    (GLubyte *) values + inStart * valueSize);
+   }
+}
+
+
+/**
+ * Wrapper for gl_renderbuffer::PutRow() which does clipping.
+ * \param valueSize  size of each value (pixel) in bytes
+ */
+void
+_swrast_put_row(struct gl_context *ctx, struct gl_renderbuffer *rb,
+                GLuint count, GLint x, GLint y,
+                const GLvoid *values, GLuint valueSize)
+{
+   GLint skip = 0;
+
+   if (y < 0 || y >= (GLint) rb->Height)
+      return; /* above or below */
+
+   if (x + (GLint) count <= 0 || x >= (GLint) rb->Width)
+      return; /* entirely left or right */
+
+   if ((GLint) (x + count) > (GLint) rb->Width) {
+      /* right clip */
+      GLint clip = x + count - rb->Width;
+      count -= clip;
+   }
+
+   if (x < 0) {
+      /* left clip */
+      skip = -x;
+      x = 0;
+      count -= skip;
+   }
+
+   rb->PutRow(ctx, rb, count, x, y,
+              (const GLubyte *) values + skip * valueSize, NULL);
+}
+
+
+/**
+ * Wrapper for gl_renderbuffer::GetRow() which does clipping.
+ * \param valueSize  size of each value (pixel) in bytes
+ */
+void
+_swrast_get_row(struct gl_context *ctx, struct gl_renderbuffer *rb,
+                GLuint count, GLint x, GLint y,
+                GLvoid *values, GLuint valueSize)
+{
+   GLint skip = 0;
+
+   if (y < 0 || y >= (GLint) rb->Height)
+      return; /* above or below */
+
+   if (x + (GLint) count <= 0 || x >= (GLint) rb->Width)
+      return; /* entirely left or right */
+
+   if (x + count > rb->Width) {
+      /* right clip */
+      GLint clip = x + count - rb->Width;
+      count -= clip;
+   }
+
+   if (x < 0) {
+      /* left clip */
+      skip = -x;
+      x = 0;
+      count -= skip;
+   }
+
+   rb->GetRow(ctx, rb, count, x, y, (GLubyte *) values + skip * valueSize);
+}
+
+
+/**
+ * Get RGBA pixels from the given renderbuffer.
+ * Used by blending, logicop and masking functions.
+ * \return pointer to the colors we read.
+ */
+void *
+_swrast_get_dest_rgba(struct gl_context *ctx, struct gl_renderbuffer *rb,
+                      SWspan *span)
+{
+   const GLuint pixelSize = RGBA_PIXEL_SIZE(span->array->ChanType);
+   void *rbPixels;
+
+   /* Point rbPixels to a temporary space */
+   rbPixels = span->array->attribs[FRAG_ATTRIB_MAX - 1];
+
+   /* Get destination values from renderbuffer */
+   if (span->arrayMask & SPAN_XY) {
+      _swrast_get_values(ctx, rb, span->end, span->array->x, span->array->y,
+                         rbPixels, pixelSize);
+   }
+   else {
+      _swrast_get_row(ctx, rb, span->end, span->x, span->y,
+                      rbPixels, pixelSize);
+   }
+
+   return rbPixels;
+}
diff --git a/mesalib/src/mesa/swrast/s_texcombine.c b/mesalib/src/mesa/swrast/s_texcombine.c
index 7f49b6ffa..53ef2f890 100644
--- a/mesalib/src/mesa/swrast/s_texcombine.c
+++ b/mesalib/src/mesa/swrast/s_texcombine.c
@@ -1,739 +1,751 @@
-/*
- * Mesa 3-D graphics library
- * Version:  7.5
- *
- * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
- * Copyright (C) 2009  VMware, Inc.   All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included
- * in all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
- * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-
-
-#include "main/glheader.h"
-#include "main/context.h"
-#include "main/colormac.h"
-#include "main/imports.h"
-#include "main/pixeltransfer.h"
-#include "program/prog_instruction.h"
-
-#include "s_context.h"
-#include "s_texcombine.h"
-
-
-/**
- * Pointer to array of float[4]
- * This type makes the code below more concise and avoids a lot of casting.
- */
-typedef float (*float4_array)[4];
-
-
-/**
- * Return array of texels for given unit.
- */
-static INLINE float4_array
-get_texel_array(SWcontext *swrast, GLuint unit)
-{
-   return (float4_array) (swrast->TexelBuffer + unit * MAX_WIDTH * 4);
-}
-
-
-
-/**
- * Do texture application for:
- *  GL_EXT_texture_env_combine
- *  GL_ARB_texture_env_combine
- *  GL_EXT_texture_env_dot3
- *  GL_ARB_texture_env_dot3
- *  GL_ATI_texture_env_combine3
- *  GL_NV_texture_env_combine4
- *  conventional GL texture env modes
- *
- * \param ctx          rendering context
- * \param unit         the texture combiner unit
- * \param n            number of fragments to process (span width)
- * \param primary_rgba incoming fragment color array
- * \param texelBuffer  pointer to texel colors for all texture units
- * 
- * \param rgba         incoming/result fragment colors
- */
-static void
-texture_combine( struct gl_context *ctx, GLuint unit, GLuint n,
-                 const float4_array primary_rgba,
-                 const GLfloat *texelBuffer,
-                 GLchan (*rgbaChan)[4] )
-{
-   SWcontext *swrast = SWRAST_CONTEXT(ctx);
-   const struct gl_texture_unit *textureUnit = &(ctx->Texture.Unit[unit]);
-   const struct gl_tex_env_combine_state *combine = textureUnit->_CurrentCombine;
-   float4_array argRGB[MAX_COMBINER_TERMS];
-   float4_array argA[MAX_COMBINER_TERMS];
-   const GLfloat scaleRGB = (GLfloat) (1 << combine->ScaleShiftRGB);
-   const GLfloat scaleA = (GLfloat) (1 << combine->ScaleShiftA);
-   const GLuint numArgsRGB = combine->_NumArgsRGB;
-   const GLuint numArgsA = combine->_NumArgsA;
-   float4_array ccolor[4], rgba;
-   GLuint i, term;
-
-   /* alloc temp pixel buffers */
-   rgba = (float4_array) malloc(4 * n * sizeof(GLfloat));
-   if (!rgba) {
-      _mesa_error(ctx, GL_OUT_OF_MEMORY, "texture_combine");
-      return;
-   }
-
-   for (i = 0; i < numArgsRGB || i < numArgsA; i++) {
-      ccolor[i] = (float4_array) malloc(4 * n * sizeof(GLfloat));
-      if (!ccolor[i]) {
-         while (i) {
-            free(ccolor[i]);
-            i--;
-         }
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "texture_combine");
-         return;
-      }
-   }
-
-   for (i = 0; i < n; i++) {
-      rgba[i][RCOMP] = CHAN_TO_FLOAT(rgbaChan[i][RCOMP]);
-      rgba[i][GCOMP] = CHAN_TO_FLOAT(rgbaChan[i][GCOMP]);
-      rgba[i][BCOMP] = CHAN_TO_FLOAT(rgbaChan[i][BCOMP]);
-      rgba[i][ACOMP] = CHAN_TO_FLOAT(rgbaChan[i][ACOMP]);
-   }
-
-   /*
-   printf("modeRGB 0x%x  modeA 0x%x  srcRGB1 0x%x  srcA1 0x%x  srcRGB2 0x%x  srcA2 0x%x\n",
-          combine->ModeRGB,
-          combine->ModeA,
-          combine->SourceRGB[0],
-          combine->SourceA[0],
-          combine->SourceRGB[1],
-          combine->SourceA[1]);
-   */
-
-   /*
-    * Do operand setup for up to 4 operands.  Loop over the terms.
-    */
-   for (term = 0; term < numArgsRGB; term++) {
-      const GLenum srcRGB = combine->SourceRGB[term];
-      const GLenum operandRGB = combine->OperandRGB[term];
-
-      switch (srcRGB) {
-         case GL_TEXTURE:
-            argRGB[term] = get_texel_array(swrast, unit);
-            break;
-         case GL_PRIMARY_COLOR:
-            argRGB[term] = primary_rgba;
-            break;
-         case GL_PREVIOUS:
-            argRGB[term] = rgba;
-            break;
-         case GL_CONSTANT:
-            {
-               float4_array c = ccolor[term];
-               GLfloat red   = textureUnit->EnvColor[0];
-               GLfloat green = textureUnit->EnvColor[1];
-               GLfloat blue  = textureUnit->EnvColor[2];
-               GLfloat alpha = textureUnit->EnvColor[3];
-               for (i = 0; i < n; i++) {
-                  ASSIGN_4V(c[i], red, green, blue, alpha);
-               }
-               argRGB[term] = ccolor[term];
-            }
-            break;
-	 /* GL_ATI_texture_env_combine3 allows GL_ZERO & GL_ONE as sources.
-	  */
-	 case GL_ZERO:
-            {
-               float4_array c = ccolor[term];
-               for (i = 0; i < n; i++) {
-                  ASSIGN_4V(c[i], 0.0F, 0.0F, 0.0F, 0.0F);
-               }
-               argRGB[term] = ccolor[term];
-            }
-            break;
-	 case GL_ONE:
-            {
-               float4_array c = ccolor[term];
-               for (i = 0; i < n; i++) {
-                  ASSIGN_4V(c[i], 1.0F, 1.0F, 1.0F, 1.0F);
-               }
-               argRGB[term] = ccolor[term];
-            }
-            break;
-         default:
-            /* ARB_texture_env_crossbar source */
-            {
-               const GLuint srcUnit = srcRGB - GL_TEXTURE0;
-               ASSERT(srcUnit < ctx->Const.MaxTextureUnits);
-               if (!ctx->Texture.Unit[srcUnit]._ReallyEnabled)
-                  goto end;
-               argRGB[term] = get_texel_array(swrast, srcUnit);
-            }
-      }
-
-      if (operandRGB != GL_SRC_COLOR) {
-         float4_array src = argRGB[term];
-         float4_array dst = ccolor[term];
-
-         /* point to new arg[term] storage */
-         argRGB[term] = ccolor[term];
-
-         switch (operandRGB) {
-         case GL_ONE_MINUS_SRC_COLOR:
-            for (i = 0; i < n; i++) {
-               dst[i][RCOMP] = 1.0F - src[i][RCOMP];
-               dst[i][GCOMP] = 1.0F - src[i][GCOMP];
-               dst[i][BCOMP] = 1.0F - src[i][BCOMP];
-            }
-            break;
-         case GL_SRC_ALPHA:
-            for (i = 0; i < n; i++) {
-               dst[i][RCOMP] =
-               dst[i][GCOMP] =
-               dst[i][BCOMP] = src[i][ACOMP];
-            }
-            break;
-         case GL_ONE_MINUS_SRC_ALPHA:
-            for (i = 0; i < n; i++) {
-               dst[i][RCOMP] =
-               dst[i][GCOMP] =
-               dst[i][BCOMP] = 1.0F - src[i][ACOMP];
-            }
-            break;
-         default:
-            _mesa_problem(ctx, "Bad operandRGB");
-         }
-      }
-   }
-
-   /*
-    * Set up the argA[term] pointers
-    */
-   for (term = 0; term < numArgsA; term++) {
-      const GLenum srcA = combine->SourceA[term];
-      const GLenum operandA = combine->OperandA[term];
-
-      switch (srcA) {
-         case GL_TEXTURE:
-            argA[term] = get_texel_array(swrast, unit);
-            break;
-         case GL_PRIMARY_COLOR:
-            argA[term] = primary_rgba;
-            break;
-         case GL_PREVIOUS:
-            argA[term] = rgba;
-            break;
-         case GL_CONSTANT:
-            {
-               float4_array c = ccolor[term];
-               GLfloat alpha = textureUnit->EnvColor[3];
-               for (i = 0; i < n; i++)
-                  c[i][ACOMP] = alpha;
-               argA[term] = ccolor[term];
-            }
-            break;
-	 /* GL_ATI_texture_env_combine3 allows GL_ZERO & GL_ONE as sources.
-	  */
-	 case GL_ZERO:
-            {
-               float4_array c = ccolor[term];
-               for (i = 0; i < n; i++)
-                  c[i][ACOMP] = 0.0F;
-               argA[term] = ccolor[term];
-            }
-            break;
-	 case GL_ONE:
-            {
-               float4_array c = ccolor[term];
-               for (i = 0; i < n; i++)
-                  c[i][ACOMP] = 1.0F;
-               argA[term] = ccolor[term];
-            }
-            break;
-         default:
-            /* ARB_texture_env_crossbar source */
-            {
-               const GLuint srcUnit = srcA - GL_TEXTURE0;
-               ASSERT(srcUnit < ctx->Const.MaxTextureUnits);
-               if (!ctx->Texture.Unit[srcUnit]._ReallyEnabled)
-                  goto end;
-               argA[term] = get_texel_array(swrast, srcUnit);
-            }
-      }
-
-      if (operandA == GL_ONE_MINUS_SRC_ALPHA) {
-         float4_array src = argA[term];
-         float4_array dst = ccolor[term];
-         argA[term] = ccolor[term];
-         for (i = 0; i < n; i++) {
-            dst[i][ACOMP] = 1.0F - src[i][ACOMP];
-         }
-      }
-   }
-
-   /* RGB channel combine */
-   {
-      float4_array arg0 = argRGB[0];
-      float4_array arg1 = argRGB[1];
-      float4_array arg2 = argRGB[2];
-      float4_array arg3 = argRGB[3];
-
-      switch (combine->ModeRGB) {
-      case GL_REPLACE:
-         for (i = 0; i < n; i++) {
-            rgba[i][RCOMP] = arg0[i][RCOMP] * scaleRGB;
-            rgba[i][GCOMP] = arg0[i][GCOMP] * scaleRGB;
-            rgba[i][BCOMP] = arg0[i][BCOMP] * scaleRGB;
-         }
-         break;
-      case GL_MODULATE:
-         for (i = 0; i < n; i++) {
-            rgba[i][RCOMP] = arg0[i][RCOMP] * arg1[i][RCOMP] * scaleRGB;
-            rgba[i][GCOMP] = arg0[i][GCOMP] * arg1[i][GCOMP] * scaleRGB;
-            rgba[i][BCOMP] = arg0[i][BCOMP] * arg1[i][BCOMP] * scaleRGB;
-         }
-         break;
-      case GL_ADD:
-         if (textureUnit->EnvMode == GL_COMBINE4_NV) {
-            /* (a * b) + (c * d) */
-            for (i = 0; i < n; i++) {
-               rgba[i][RCOMP] = (arg0[i][RCOMP] * arg1[i][RCOMP] +
-                                 arg2[i][RCOMP] * arg3[i][RCOMP]) * scaleRGB;
-               rgba[i][GCOMP] = (arg0[i][GCOMP] * arg1[i][GCOMP] +
-                                 arg2[i][GCOMP] * arg3[i][GCOMP]) * scaleRGB;
-               rgba[i][BCOMP] = (arg0[i][BCOMP] * arg1[i][BCOMP] +
-                                 arg2[i][BCOMP] * arg3[i][BCOMP]) * scaleRGB;
-            }
-         }
-         else {
-            /* 2-term addition */
-            for (i = 0; i < n; i++) {
-               rgba[i][RCOMP] = (arg0[i][RCOMP] + arg1[i][RCOMP]) * scaleRGB;
-               rgba[i][GCOMP] = (arg0[i][GCOMP] + arg1[i][GCOMP]) * scaleRGB;
-               rgba[i][BCOMP] = (arg0[i][BCOMP] + arg1[i][BCOMP]) * scaleRGB;
-            }
-         }
-         break;
-      case GL_ADD_SIGNED:
-         if (textureUnit->EnvMode == GL_COMBINE4_NV) {
-            /* (a * b) + (c * d) - 0.5 */
-            for (i = 0; i < n; i++) {
-               rgba[i][RCOMP] = (arg0[i][RCOMP] * arg1[i][RCOMP] +
-                                 arg2[i][RCOMP] * arg3[i][RCOMP] - 0.5F) * scaleRGB;
-               rgba[i][GCOMP] = (arg0[i][GCOMP] * arg1[i][GCOMP] +
-                                 arg2[i][GCOMP] * arg3[i][GCOMP] - 0.5F) * scaleRGB;
-               rgba[i][BCOMP] = (arg0[i][BCOMP] * arg1[i][BCOMP] +
-                                 arg2[i][BCOMP] * arg3[i][BCOMP] - 0.5F) * scaleRGB;
-            }
-         }
-         else {
-            for (i = 0; i < n; i++) {
-               rgba[i][RCOMP] = (arg0[i][RCOMP] + arg1[i][RCOMP] - 0.5F) * scaleRGB;
-               rgba[i][GCOMP] = (arg0[i][GCOMP] + arg1[i][GCOMP] - 0.5F) * scaleRGB;
-               rgba[i][BCOMP] = (arg0[i][BCOMP] + arg1[i][BCOMP] - 0.5F) * scaleRGB;
-            }
-         }
-         break;
-      case GL_INTERPOLATE:
-         for (i = 0; i < n; i++) {
-            rgba[i][RCOMP] = (arg0[i][RCOMP] * arg2[i][RCOMP] +
-                          arg1[i][RCOMP] * (1.0F - arg2[i][RCOMP])) * scaleRGB;
-            rgba[i][GCOMP] = (arg0[i][GCOMP] * arg2[i][GCOMP] +
-                          arg1[i][GCOMP] * (1.0F - arg2[i][GCOMP])) * scaleRGB;
-            rgba[i][BCOMP] = (arg0[i][BCOMP] * arg2[i][BCOMP] +
-                          arg1[i][BCOMP] * (1.0F - arg2[i][BCOMP])) * scaleRGB;
-         }
-         break;
-      case GL_SUBTRACT:
-         for (i = 0; i < n; i++) {
-            rgba[i][RCOMP] = (arg0[i][RCOMP] - arg1[i][RCOMP]) * scaleRGB;
-            rgba[i][GCOMP] = (arg0[i][GCOMP] - arg1[i][GCOMP]) * scaleRGB;
-            rgba[i][BCOMP] = (arg0[i][BCOMP] - arg1[i][BCOMP]) * scaleRGB;
-         }
-         break;
-      case GL_DOT3_RGB_EXT:
-      case GL_DOT3_RGBA_EXT:
-         /* Do not scale the result by 1 2 or 4 */
-         for (i = 0; i < n; i++) {
-            GLfloat dot = ((arg0[i][RCOMP] - 0.5F) * (arg1[i][RCOMP] - 0.5F) +
-                           (arg0[i][GCOMP] - 0.5F) * (arg1[i][GCOMP] - 0.5F) +
-                           (arg0[i][BCOMP] - 0.5F) * (arg1[i][BCOMP] - 0.5F))
-               * 4.0F;
-            dot = CLAMP(dot, 0.0F, 1.0F);
-            rgba[i][RCOMP] = rgba[i][GCOMP] = rgba[i][BCOMP] = dot;
-         }
-         break;
-      case GL_DOT3_RGB:
-      case GL_DOT3_RGBA:
-         /* DO scale the result by 1 2 or 4 */
-         for (i = 0; i < n; i++) {
-            GLfloat dot = ((arg0[i][RCOMP] - 0.5F) * (arg1[i][RCOMP] - 0.5F) +
-                           (arg0[i][GCOMP] - 0.5F) * (arg1[i][GCOMP] - 0.5F) +
-                           (arg0[i][BCOMP] - 0.5F) * (arg1[i][BCOMP] - 0.5F))
-               * 4.0F * scaleRGB;
-            dot = CLAMP(dot, 0.0F, 1.0F);
-            rgba[i][RCOMP] = rgba[i][GCOMP] = rgba[i][BCOMP] = dot;
-         }
-         break;
-      case GL_MODULATE_ADD_ATI:
-         for (i = 0; i < n; i++) {
-            rgba[i][RCOMP] = ((arg0[i][RCOMP] * arg2[i][RCOMP]) +
-                              arg1[i][RCOMP]) * scaleRGB;
-            rgba[i][GCOMP] = ((arg0[i][GCOMP] * arg2[i][GCOMP]) +
-                              arg1[i][GCOMP]) * scaleRGB;
-            rgba[i][BCOMP] = ((arg0[i][BCOMP] * arg2[i][BCOMP]) +
-                              arg1[i][BCOMP]) * scaleRGB;
-	 }
-         break;
-      case GL_MODULATE_SIGNED_ADD_ATI:
-         for (i = 0; i < n; i++) {
-            rgba[i][RCOMP] = ((arg0[i][RCOMP] * arg2[i][RCOMP]) +
-                              arg1[i][RCOMP] - 0.5F) * scaleRGB;
-            rgba[i][GCOMP] = ((arg0[i][GCOMP] * arg2[i][GCOMP]) +
-                              arg1[i][GCOMP] - 0.5F) * scaleRGB;
-            rgba[i][BCOMP] = ((arg0[i][BCOMP] * arg2[i][BCOMP]) +
-                              arg1[i][BCOMP] - 0.5F) * scaleRGB;
-	 }
-         break;
-      case GL_MODULATE_SUBTRACT_ATI:
-         for (i = 0; i < n; i++) {
-            rgba[i][RCOMP] = ((arg0[i][RCOMP] * arg2[i][RCOMP]) -
-                              arg1[i][RCOMP]) * scaleRGB;
-            rgba[i][GCOMP] = ((arg0[i][GCOMP] * arg2[i][GCOMP]) -
-                              arg1[i][GCOMP]) * scaleRGB;
-            rgba[i][BCOMP] = ((arg0[i][BCOMP] * arg2[i][BCOMP]) -
-                              arg1[i][BCOMP]) * scaleRGB;
-	 }
-         break;
-      case GL_BUMP_ENVMAP_ATI:
-         /* this produces a fixed rgba color, and the coord calc is done elsewhere */
-         for (i = 0; i < n; i++) {
-            /* rgba result is 0,0,0,1 */
-            rgba[i][RCOMP] = 0.0;
-            rgba[i][GCOMP] = 0.0;
-            rgba[i][BCOMP] = 0.0;
-            rgba[i][ACOMP] = 1.0;
-	 }
-         goto end; /* no alpha processing */
-      default:
-         _mesa_problem(ctx, "invalid combine mode");
-      }
-   }
-
-   /* Alpha channel combine */
-   {
-      float4_array arg0 = argA[0];
-      float4_array arg1 = argA[1];
-      float4_array arg2 = argA[2];
-      float4_array arg3 = argA[3];
-
-      switch (combine->ModeA) {
-      case GL_REPLACE:
-         for (i = 0; i < n; i++) {
-            rgba[i][ACOMP] = arg0[i][ACOMP] * scaleA;
-         }
-         break;
-      case GL_MODULATE:
-         for (i = 0; i < n; i++) {
-            rgba[i][ACOMP] = arg0[i][ACOMP] * arg1[i][ACOMP] * scaleA;
-         }
-         break;
-      case GL_ADD:
-         if (textureUnit->EnvMode == GL_COMBINE4_NV) {
-            /* (a * b) + (c * d) */
-            for (i = 0; i < n; i++) {
-               rgba[i][ACOMP] = (arg0[i][ACOMP] * arg1[i][ACOMP] +
-                                 arg2[i][ACOMP] * arg3[i][ACOMP]) * scaleA;
-            }
-         }
-         else {
-            /* two-term add */
-            for (i = 0; i < n; i++) {
-               rgba[i][ACOMP] = (arg0[i][ACOMP] + arg1[i][ACOMP]) * scaleA;
-            }
-         }
-         break;
-      case GL_ADD_SIGNED:
-         if (textureUnit->EnvMode == GL_COMBINE4_NV) {
-            /* (a * b) + (c * d) - 0.5 */
-            for (i = 0; i < n; i++) {
-               rgba[i][ACOMP] = (arg0[i][ACOMP] * arg1[i][ACOMP] +
-                                 arg2[i][ACOMP] * arg3[i][ACOMP] -
-                                 0.5F) * scaleA;
-            }
-         }
-         else {
-            /* a + b - 0.5 */
-            for (i = 0; i < n; i++) {
-               rgba[i][ACOMP] = (arg0[i][ACOMP] + arg1[i][ACOMP] - 0.5F) * scaleA;
-            }
-         }
-         break;
-      case GL_INTERPOLATE:
-         for (i = 0; i < n; i++) {
-            rgba[i][ACOMP] = (arg0[i][ACOMP] * arg2[i][ACOMP] +
-                              arg1[i][ACOMP] * (1.0F - arg2[i][ACOMP]))
-               * scaleA;
-         }
-         break;
-      case GL_SUBTRACT:
-         for (i = 0; i < n; i++) {
-            rgba[i][ACOMP] = (arg0[i][ACOMP] - arg1[i][ACOMP]) * scaleA;
-         }
-         break;
-      case GL_MODULATE_ADD_ATI:
-         for (i = 0; i < n; i++) {
-            rgba[i][ACOMP] = ((arg0[i][ACOMP] * arg2[i][ACOMP])
-                              + arg1[i][ACOMP]) * scaleA;
-         }
-         break;
-      case GL_MODULATE_SIGNED_ADD_ATI:
-         for (i = 0; i < n; i++) {
-            rgba[i][ACOMP] = ((arg0[i][ACOMP] * arg2[i][ACOMP]) +
-                              arg1[i][ACOMP] - 0.5F) * scaleA;
-         }
-         break;
-      case GL_MODULATE_SUBTRACT_ATI:
-         for (i = 0; i < n; i++) {
-            rgba[i][ACOMP] = ((arg0[i][ACOMP] * arg2[i][ACOMP])
-                              - arg1[i][ACOMP]) * scaleA;
-         }
-         break;
-      default:
-         _mesa_problem(ctx, "invalid combine mode");
-      }
-   }
-
-   /* Fix the alpha component for GL_DOT3_RGBA_EXT/ARB combining.
-    * This is kind of a kludge.  It would have been better if the spec
-    * were written such that the GL_COMBINE_ALPHA value could be set to
-    * GL_DOT3.
-    */
-   if (combine->ModeRGB == GL_DOT3_RGBA_EXT ||
-       combine->ModeRGB == GL_DOT3_RGBA) {
-      for (i = 0; i < n; i++) {
-	 rgba[i][ACOMP] = rgba[i][RCOMP];
-      }
-   }
-
-   for (i = 0; i < n; i++) {
-      UNCLAMPED_FLOAT_TO_CHAN(rgbaChan[i][RCOMP], rgba[i][RCOMP]);
-      UNCLAMPED_FLOAT_TO_CHAN(rgbaChan[i][GCOMP], rgba[i][GCOMP]);
-      UNCLAMPED_FLOAT_TO_CHAN(rgbaChan[i][BCOMP], rgba[i][BCOMP]);
-      UNCLAMPED_FLOAT_TO_CHAN(rgbaChan[i][ACOMP], rgba[i][ACOMP]);
-   }
-
-end:
-   for (i = 0; i < numArgsRGB || i < numArgsA; i++) {
-      free(ccolor[i]);
-   }
-   free(rgba);
-}
-
-
-/**
- * Apply X/Y/Z/W/0/1 swizzle to an array of colors/texels.
- * See GL_EXT_texture_swizzle.
- */
-static void
-swizzle_texels(GLuint swizzle, GLuint count, float4_array texels)
-{
-   const GLuint swzR = GET_SWZ(swizzle, 0);
-   const GLuint swzG = GET_SWZ(swizzle, 1);
-   const GLuint swzB = GET_SWZ(swizzle, 2);
-   const GLuint swzA = GET_SWZ(swizzle, 3);
-   GLfloat vector[6];
-   GLuint i;
-
-   vector[SWIZZLE_ZERO] = 0;
-   vector[SWIZZLE_ONE] = 1.0F;
-
-   for (i = 0; i < count; i++) {
-      vector[SWIZZLE_X] = texels[i][0];
-      vector[SWIZZLE_Y] = texels[i][1];
-      vector[SWIZZLE_Z] = texels[i][2];
-      vector[SWIZZLE_W] = texels[i][3];
-      texels[i][RCOMP] = vector[swzR];
-      texels[i][GCOMP] = vector[swzG];
-      texels[i][BCOMP] = vector[swzB];
-      texels[i][ACOMP] = vector[swzA];
-   }
-}
-
-
-/**
- * Apply texture mapping to a span of fragments.
- */
-void
-_swrast_texture_span( struct gl_context *ctx, SWspan *span )
-{
-   SWcontext *swrast = SWRAST_CONTEXT(ctx);
-   float4_array primary_rgba;
-   GLuint unit;
-
-   primary_rgba = (float4_array) malloc(span->end * 4 * sizeof(GLfloat));
-
-   if (!primary_rgba) {
-      _mesa_error(ctx, GL_OUT_OF_MEMORY, "texture_span");
-      return;
-   }
-
-   ASSERT(span->end <= MAX_WIDTH);
-
-   /*
-    * Save copy of the incoming fragment colors (the GL_PRIMARY_COLOR)
-    */
-   if (swrast->_TextureCombinePrimary) {
-      GLuint i;
-      for (i = 0; i < span->end; i++) {
-         primary_rgba[i][RCOMP] = CHAN_TO_FLOAT(span->array->rgba[i][RCOMP]);
-         primary_rgba[i][GCOMP] = CHAN_TO_FLOAT(span->array->rgba[i][GCOMP]);
-         primary_rgba[i][BCOMP] = CHAN_TO_FLOAT(span->array->rgba[i][BCOMP]);
-         primary_rgba[i][ACOMP] = CHAN_TO_FLOAT(span->array->rgba[i][ACOMP]);
-      }
-   }
-
-   /* First must sample all bump maps */
-   for (unit = 0; unit < ctx->Const.MaxTextureUnits; unit++) {
-      const struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-
-      if (texUnit->_ReallyEnabled &&
-         texUnit->_CurrentCombine->ModeRGB == GL_BUMP_ENVMAP_ATI) {
-         const GLfloat (*texcoords)[4] = (const GLfloat (*)[4])
-            span->array->attribs[FRAG_ATTRIB_TEX0 + unit];
-         float4_array targetcoords =
-            span->array->attribs[FRAG_ATTRIB_TEX0 +
-               ctx->Texture.Unit[unit].BumpTarget - GL_TEXTURE0];
-
-         const struct gl_texture_object *curObj = texUnit->_Current;
-         GLfloat *lambda = span->array->lambda[unit];
-         float4_array texels = get_texel_array(swrast, unit);
-         GLuint i;
-         GLfloat rotMatrix00 = ctx->Texture.Unit[unit].RotMatrix[0];
-         GLfloat rotMatrix01 = ctx->Texture.Unit[unit].RotMatrix[1];
-         GLfloat rotMatrix10 = ctx->Texture.Unit[unit].RotMatrix[2];
-         GLfloat rotMatrix11 = ctx->Texture.Unit[unit].RotMatrix[3];
-
-         /* adjust texture lod (lambda) */
-         if (span->arrayMask & SPAN_LAMBDA) {
-            if (texUnit->LodBias + curObj->Sampler.LodBias != 0.0F) {
-               /* apply LOD bias, but don't clamp yet */
-               const GLfloat bias = CLAMP(texUnit->LodBias + curObj->Sampler.LodBias,
-                                          -ctx->Const.MaxTextureLodBias,
-                                          ctx->Const.MaxTextureLodBias);
-               GLuint i;
-               for (i = 0; i < span->end; i++) {
-                  lambda[i] += bias;
-               }
-            }
-
-            if (curObj->Sampler.MinLod != -1000.0 ||
-                curObj->Sampler.MaxLod != 1000.0) {
-               /* apply LOD clamping to lambda */
-               const GLfloat min = curObj->Sampler.MinLod;
-               const GLfloat max = curObj->Sampler.MaxLod;
-               GLuint i;
-               for (i = 0; i < span->end; i++) {
-                  GLfloat l = lambda[i];
-                  lambda[i] = CLAMP(l, min, max);
-               }
-            }
-         }
-
-         /* Sample the texture (span->end = number of fragments) */
-         swrast->TextureSample[unit]( ctx, texUnit->_Current, span->end,
-                                      texcoords, lambda, texels );
-
-         /* manipulate the span values of the bump target
-            not sure this can work correctly even ignoring
-            the problem that channel is unsigned */
-         for (i = 0; i < span->end; i++) {
-            targetcoords[i][0] += (texels[i][0] * rotMatrix00 + texels[i][1] *
-                                  rotMatrix01) / targetcoords[i][3];
-            targetcoords[i][1] += (texels[i][0] * rotMatrix10 + texels[i][1] *
-                                  rotMatrix11) / targetcoords[i][3];
-         }
-      }
-   }
-
-   /*
-    * Must do all texture sampling before combining in order to
-    * accomodate GL_ARB_texture_env_crossbar.
-    */
-   for (unit = 0; unit < ctx->Const.MaxTextureUnits; unit++) {
-      const struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
-      if (texUnit->_ReallyEnabled &&
-          texUnit->_CurrentCombine->ModeRGB != GL_BUMP_ENVMAP_ATI) {
-         const GLfloat (*texcoords)[4] = (const GLfloat (*)[4])
-            span->array->attribs[FRAG_ATTRIB_TEX0 + unit];
-         const struct gl_texture_object *curObj = texUnit->_Current;
-         GLfloat *lambda = span->array->lambda[unit];
-         float4_array texels = get_texel_array(swrast, unit);
-
-         /* adjust texture lod (lambda) */
-         if (span->arrayMask & SPAN_LAMBDA) {
-            if (texUnit->LodBias + curObj->Sampler.LodBias != 0.0F) {
-               /* apply LOD bias, but don't clamp yet */
-               const GLfloat bias = CLAMP(texUnit->LodBias + curObj->Sampler.LodBias,
-                                          -ctx->Const.MaxTextureLodBias,
-                                          ctx->Const.MaxTextureLodBias);
-               GLuint i;
-               for (i = 0; i < span->end; i++) {
-                  lambda[i] += bias;
-               }
-            }
-
-            if (curObj->Sampler.MinLod != -1000.0 ||
-                curObj->Sampler.MaxLod != 1000.0) {
-               /* apply LOD clamping to lambda */
-               const GLfloat min = curObj->Sampler.MinLod;
-               const GLfloat max = curObj->Sampler.MaxLod;
-               GLuint i;
-               for (i = 0; i < span->end; i++) {
-                  GLfloat l = lambda[i];
-                  lambda[i] = CLAMP(l, min, max);
-               }
-            }
-         }
-
-         /* Sample the texture (span->end = number of fragments) */
-         swrast->TextureSample[unit]( ctx, texUnit->_Current, span->end,
-                                      texcoords, lambda, texels );
-
-         /* GL_EXT_texture_swizzle */
-         if (curObj->_Swizzle != SWIZZLE_NOOP) {
-            swizzle_texels(curObj->_Swizzle, span->end, texels);
-         }
-      }
-   }
-
-   /*
-    * OK, now apply the texture (aka texture combine/blend).
-    * We modify the span->color.rgba values.
-    */
-   for (unit = 0; unit < ctx->Const.MaxTextureUnits; unit++) {
-      if (ctx->Texture.Unit[unit]._ReallyEnabled) {
-         texture_combine( ctx, unit, span->end,
-                          primary_rgba,
-                          swrast->TexelBuffer,
-                          span->array->rgba );
-      }
-   }
-
-   free(primary_rgba);
-}
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.5
+ *
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+#include "main/glheader.h"
+#include "main/context.h"
+#include "main/colormac.h"
+#include "main/imports.h"
+#include "main/pixeltransfer.h"
+#include "program/prog_instruction.h"
+
+#include "s_context.h"
+#include "s_texcombine.h"
+
+
+/**
+ * Pointer to array of float[4]
+ * This type makes the code below more concise and avoids a lot of casting.
+ */
+typedef float (*float4_array)[4];
+
+
+/**
+ * Return array of texels for given unit.
+ */
+static INLINE float4_array
+get_texel_array(SWcontext *swrast, GLuint unit)
+{
+   return (float4_array) (swrast->TexelBuffer + unit * MAX_WIDTH * 4);
+}
+
+
+
+/**
+ * Do texture application for:
+ *  GL_EXT_texture_env_combine
+ *  GL_ARB_texture_env_combine
+ *  GL_EXT_texture_env_dot3
+ *  GL_ARB_texture_env_dot3
+ *  GL_ATI_texture_env_combine3
+ *  GL_NV_texture_env_combine4
+ *  conventional GL texture env modes
+ *
+ * \param ctx          rendering context
+ * \param unit         the texture combiner unit
+ * \param n            number of fragments to process (span width)
+ * \param primary_rgba incoming fragment color array
+ * \param texelBuffer  pointer to texel colors for all texture units
+ * 
+ * \param rgba         incoming/result fragment colors
+ */
+static void
+texture_combine( struct gl_context *ctx, GLuint unit, GLuint n,
+                 const float4_array primary_rgba,
+                 const GLfloat *texelBuffer,
+                 GLchan (*rgbaChan)[4] )
+{
+   SWcontext *swrast = SWRAST_CONTEXT(ctx);
+   const struct gl_texture_unit *textureUnit = &(ctx->Texture.Unit[unit]);
+   const struct gl_tex_env_combine_state *combine = textureUnit->_CurrentCombine;
+   float4_array argRGB[MAX_COMBINER_TERMS];
+   float4_array argA[MAX_COMBINER_TERMS];
+   const GLfloat scaleRGB = (GLfloat) (1 << combine->ScaleShiftRGB);
+   const GLfloat scaleA = (GLfloat) (1 << combine->ScaleShiftA);
+   const GLuint numArgsRGB = combine->_NumArgsRGB;
+   const GLuint numArgsA = combine->_NumArgsA;
+   float4_array ccolor[4], rgba;
+   GLuint i, term;
+
+   /* alloc temp pixel buffers */
+   rgba = (float4_array) malloc(4 * n * sizeof(GLfloat));
+   if (!rgba) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "texture_combine");
+      return;
+   }
+
+   for (i = 0; i < numArgsRGB || i < numArgsA; i++) {
+      ccolor[i] = (float4_array) malloc(4 * n * sizeof(GLfloat));
+      if (!ccolor[i]) {
+         while (i) {
+            free(ccolor[i]);
+            i--;
+         }
+         _mesa_error(ctx, GL_OUT_OF_MEMORY, "texture_combine");
+         return;
+      }
+   }
+
+   for (i = 0; i < n; i++) {
+      rgba[i][RCOMP] = CHAN_TO_FLOAT(rgbaChan[i][RCOMP]);
+      rgba[i][GCOMP] = CHAN_TO_FLOAT(rgbaChan[i][GCOMP]);
+      rgba[i][BCOMP] = CHAN_TO_FLOAT(rgbaChan[i][BCOMP]);
+      rgba[i][ACOMP] = CHAN_TO_FLOAT(rgbaChan[i][ACOMP]);
+   }
+
+   /*
+   printf("modeRGB 0x%x  modeA 0x%x  srcRGB1 0x%x  srcA1 0x%x  srcRGB2 0x%x  srcA2 0x%x\n",
+          combine->ModeRGB,
+          combine->ModeA,
+          combine->SourceRGB[0],
+          combine->SourceA[0],
+          combine->SourceRGB[1],
+          combine->SourceA[1]);
+   */
+
+   /*
+    * Do operand setup for up to 4 operands.  Loop over the terms.
+    */
+   for (term = 0; term < numArgsRGB; term++) {
+      const GLenum srcRGB = combine->SourceRGB[term];
+      const GLenum operandRGB = combine->OperandRGB[term];
+
+      switch (srcRGB) {
+         case GL_TEXTURE:
+            argRGB[term] = get_texel_array(swrast, unit);
+            break;
+         case GL_PRIMARY_COLOR:
+            argRGB[term] = primary_rgba;
+            break;
+         case GL_PREVIOUS:
+            argRGB[term] = rgba;
+            break;
+         case GL_CONSTANT:
+            {
+               float4_array c = ccolor[term];
+               GLfloat red   = textureUnit->EnvColor[0];
+               GLfloat green = textureUnit->EnvColor[1];
+               GLfloat blue  = textureUnit->EnvColor[2];
+               GLfloat alpha = textureUnit->EnvColor[3];
+               for (i = 0; i < n; i++) {
+                  ASSIGN_4V(c[i], red, green, blue, alpha);
+               }
+               argRGB[term] = ccolor[term];
+            }
+            break;
+	 /* GL_ATI_texture_env_combine3 allows GL_ZERO & GL_ONE as sources.
+	  */
+	 case GL_ZERO:
+            {
+               float4_array c = ccolor[term];
+               for (i = 0; i < n; i++) {
+                  ASSIGN_4V(c[i], 0.0F, 0.0F, 0.0F, 0.0F);
+               }
+               argRGB[term] = ccolor[term];
+            }
+            break;
+	 case GL_ONE:
+            {
+               float4_array c = ccolor[term];
+               for (i = 0; i < n; i++) {
+                  ASSIGN_4V(c[i], 1.0F, 1.0F, 1.0F, 1.0F);
+               }
+               argRGB[term] = ccolor[term];
+            }
+            break;
+         default:
+            /* ARB_texture_env_crossbar source */
+            {
+               const GLuint srcUnit = srcRGB - GL_TEXTURE0;
+               ASSERT(srcUnit < ctx->Const.MaxTextureUnits);
+               if (!ctx->Texture.Unit[srcUnit]._ReallyEnabled)
+                  goto end;
+               argRGB[term] = get_texel_array(swrast, srcUnit);
+            }
+      }
+
+      if (operandRGB != GL_SRC_COLOR) {
+         float4_array src = argRGB[term];
+         float4_array dst = ccolor[term];
+
+         /* point to new arg[term] storage */
+         argRGB[term] = ccolor[term];
+
+         switch (operandRGB) {
+         case GL_ONE_MINUS_SRC_COLOR:
+            for (i = 0; i < n; i++) {
+               dst[i][RCOMP] = 1.0F - src[i][RCOMP];
+               dst[i][GCOMP] = 1.0F - src[i][GCOMP];
+               dst[i][BCOMP] = 1.0F - src[i][BCOMP];
+            }
+            break;
+         case GL_SRC_ALPHA:
+            for (i = 0; i < n; i++) {
+               dst[i][RCOMP] =
+               dst[i][GCOMP] =
+               dst[i][BCOMP] = src[i][ACOMP];
+            }
+            break;
+         case GL_ONE_MINUS_SRC_ALPHA:
+            for (i = 0; i < n; i++) {
+               dst[i][RCOMP] =
+               dst[i][GCOMP] =
+               dst[i][BCOMP] = 1.0F - src[i][ACOMP];
+            }
+            break;
+         default:
+            _mesa_problem(ctx, "Bad operandRGB");
+         }
+      }
+   }
+
+   /*
+    * Set up the argA[term] pointers
+    */
+   for (term = 0; term < numArgsA; term++) {
+      const GLenum srcA = combine->SourceA[term];
+      const GLenum operandA = combine->OperandA[term];
+
+      switch (srcA) {
+         case GL_TEXTURE:
+            argA[term] = get_texel_array(swrast, unit);
+            break;
+         case GL_PRIMARY_COLOR:
+            argA[term] = primary_rgba;
+            break;
+         case GL_PREVIOUS:
+            argA[term] = rgba;
+            break;
+         case GL_CONSTANT:
+            {
+               float4_array c = ccolor[term];
+               GLfloat alpha = textureUnit->EnvColor[3];
+               for (i = 0; i < n; i++)
+                  c[i][ACOMP] = alpha;
+               argA[term] = ccolor[term];
+            }
+            break;
+	 /* GL_ATI_texture_env_combine3 allows GL_ZERO & GL_ONE as sources.
+	  */
+	 case GL_ZERO:
+            {
+               float4_array c = ccolor[term];
+               for (i = 0; i < n; i++)
+                  c[i][ACOMP] = 0.0F;
+               argA[term] = ccolor[term];
+            }
+            break;
+	 case GL_ONE:
+            {
+               float4_array c = ccolor[term];
+               for (i = 0; i < n; i++)
+                  c[i][ACOMP] = 1.0F;
+               argA[term] = ccolor[term];
+            }
+            break;
+         default:
+            /* ARB_texture_env_crossbar source */
+            {
+               const GLuint srcUnit = srcA - GL_TEXTURE0;
+               ASSERT(srcUnit < ctx->Const.MaxTextureUnits);
+               if (!ctx->Texture.Unit[srcUnit]._ReallyEnabled)
+                  goto end;
+               argA[term] = get_texel_array(swrast, srcUnit);
+            }
+      }
+
+      if (operandA == GL_ONE_MINUS_SRC_ALPHA) {
+         float4_array src = argA[term];
+         float4_array dst = ccolor[term];
+         argA[term] = ccolor[term];
+         for (i = 0; i < n; i++) {
+            dst[i][ACOMP] = 1.0F - src[i][ACOMP];
+         }
+      }
+   }
+
+   /* RGB channel combine */
+   {
+      float4_array arg0 = argRGB[0];
+      float4_array arg1 = argRGB[1];
+      float4_array arg2 = argRGB[2];
+      float4_array arg3 = argRGB[3];
+
+      switch (combine->ModeRGB) {
+      case GL_REPLACE:
+         for (i = 0; i < n; i++) {
+            rgba[i][RCOMP] = arg0[i][RCOMP] * scaleRGB;
+            rgba[i][GCOMP] = arg0[i][GCOMP] * scaleRGB;
+            rgba[i][BCOMP] = arg0[i][BCOMP] * scaleRGB;
+         }
+         break;
+      case GL_MODULATE:
+         for (i = 0; i < n; i++) {
+            rgba[i][RCOMP] = arg0[i][RCOMP] * arg1[i][RCOMP] * scaleRGB;
+            rgba[i][GCOMP] = arg0[i][GCOMP] * arg1[i][GCOMP] * scaleRGB;
+            rgba[i][BCOMP] = arg0[i][BCOMP] * arg1[i][BCOMP] * scaleRGB;
+         }
+         break;
+      case GL_ADD:
+         if (textureUnit->EnvMode == GL_COMBINE4_NV) {
+            /* (a * b) + (c * d) */
+            for (i = 0; i < n; i++) {
+               rgba[i][RCOMP] = (arg0[i][RCOMP] * arg1[i][RCOMP] +
+                                 arg2[i][RCOMP] * arg3[i][RCOMP]) * scaleRGB;
+               rgba[i][GCOMP] = (arg0[i][GCOMP] * arg1[i][GCOMP] +
+                                 arg2[i][GCOMP] * arg3[i][GCOMP]) * scaleRGB;
+               rgba[i][BCOMP] = (arg0[i][BCOMP] * arg1[i][BCOMP] +
+                                 arg2[i][BCOMP] * arg3[i][BCOMP]) * scaleRGB;
+            }
+         }
+         else {
+            /* 2-term addition */
+            for (i = 0; i < n; i++) {
+               rgba[i][RCOMP] = (arg0[i][RCOMP] + arg1[i][RCOMP]) * scaleRGB;
+               rgba[i][GCOMP] = (arg0[i][GCOMP] + arg1[i][GCOMP]) * scaleRGB;
+               rgba[i][BCOMP] = (arg0[i][BCOMP] + arg1[i][BCOMP]) * scaleRGB;
+            }
+         }
+         break;
+      case GL_ADD_SIGNED:
+         if (textureUnit->EnvMode == GL_COMBINE4_NV) {
+            /* (a * b) + (c * d) - 0.5 */
+            for (i = 0; i < n; i++) {
+               rgba[i][RCOMP] = (arg0[i][RCOMP] * arg1[i][RCOMP] +
+                                 arg2[i][RCOMP] * arg3[i][RCOMP] - 0.5F) * scaleRGB;
+               rgba[i][GCOMP] = (arg0[i][GCOMP] * arg1[i][GCOMP] +
+                                 arg2[i][GCOMP] * arg3[i][GCOMP] - 0.5F) * scaleRGB;
+               rgba[i][BCOMP] = (arg0[i][BCOMP] * arg1[i][BCOMP] +
+                                 arg2[i][BCOMP] * arg3[i][BCOMP] - 0.5F) * scaleRGB;
+            }
+         }
+         else {
+            for (i = 0; i < n; i++) {
+               rgba[i][RCOMP] = (arg0[i][RCOMP] + arg1[i][RCOMP] - 0.5F) * scaleRGB;
+               rgba[i][GCOMP] = (arg0[i][GCOMP] + arg1[i][GCOMP] - 0.5F) * scaleRGB;
+               rgba[i][BCOMP] = (arg0[i][BCOMP] + arg1[i][BCOMP] - 0.5F) * scaleRGB;
+            }
+         }
+         break;
+      case GL_INTERPOLATE:
+         for (i = 0; i < n; i++) {
+            rgba[i][RCOMP] = (arg0[i][RCOMP] * arg2[i][RCOMP] +
+                          arg1[i][RCOMP] * (1.0F - arg2[i][RCOMP])) * scaleRGB;
+            rgba[i][GCOMP] = (arg0[i][GCOMP] * arg2[i][GCOMP] +
+                          arg1[i][GCOMP] * (1.0F - arg2[i][GCOMP])) * scaleRGB;
+            rgba[i][BCOMP] = (arg0[i][BCOMP] * arg2[i][BCOMP] +
+                          arg1[i][BCOMP] * (1.0F - arg2[i][BCOMP])) * scaleRGB;
+         }
+         break;
+      case GL_SUBTRACT:
+         for (i = 0; i < n; i++) {
+            rgba[i][RCOMP] = (arg0[i][RCOMP] - arg1[i][RCOMP]) * scaleRGB;
+            rgba[i][GCOMP] = (arg0[i][GCOMP] - arg1[i][GCOMP]) * scaleRGB;
+            rgba[i][BCOMP] = (arg0[i][BCOMP] - arg1[i][BCOMP]) * scaleRGB;
+         }
+         break;
+      case GL_DOT3_RGB_EXT:
+      case GL_DOT3_RGBA_EXT:
+         /* Do not scale the result by 1 2 or 4 */
+         for (i = 0; i < n; i++) {
+            GLfloat dot = ((arg0[i][RCOMP] - 0.5F) * (arg1[i][RCOMP] - 0.5F) +
+                           (arg0[i][GCOMP] - 0.5F) * (arg1[i][GCOMP] - 0.5F) +
+                           (arg0[i][BCOMP] - 0.5F) * (arg1[i][BCOMP] - 0.5F))
+               * 4.0F;
+            dot = CLAMP(dot, 0.0F, 1.0F);
+            rgba[i][RCOMP] = rgba[i][GCOMP] = rgba[i][BCOMP] = dot;
+         }
+         break;
+      case GL_DOT3_RGB:
+      case GL_DOT3_RGBA:
+         /* DO scale the result by 1 2 or 4 */
+         for (i = 0; i < n; i++) {
+            GLfloat dot = ((arg0[i][RCOMP] - 0.5F) * (arg1[i][RCOMP] - 0.5F) +
+                           (arg0[i][GCOMP] - 0.5F) * (arg1[i][GCOMP] - 0.5F) +
+                           (arg0[i][BCOMP] - 0.5F) * (arg1[i][BCOMP] - 0.5F))
+               * 4.0F * scaleRGB;
+            dot = CLAMP(dot, 0.0F, 1.0F);
+            rgba[i][RCOMP] = rgba[i][GCOMP] = rgba[i][BCOMP] = dot;
+         }
+         break;
+      case GL_MODULATE_ADD_ATI:
+         for (i = 0; i < n; i++) {
+            rgba[i][RCOMP] = ((arg0[i][RCOMP] * arg2[i][RCOMP]) +
+                              arg1[i][RCOMP]) * scaleRGB;
+            rgba[i][GCOMP] = ((arg0[i][GCOMP] * arg2[i][GCOMP]) +
+                              arg1[i][GCOMP]) * scaleRGB;
+            rgba[i][BCOMP] = ((arg0[i][BCOMP] * arg2[i][BCOMP]) +
+                              arg1[i][BCOMP]) * scaleRGB;
+	 }
+         break;
+      case GL_MODULATE_SIGNED_ADD_ATI:
+         for (i = 0; i < n; i++) {
+            rgba[i][RCOMP] = ((arg0[i][RCOMP] * arg2[i][RCOMP]) +
+                              arg1[i][RCOMP] - 0.5F) * scaleRGB;
+            rgba[i][GCOMP] = ((arg0[i][GCOMP] * arg2[i][GCOMP]) +
+                              arg1[i][GCOMP] - 0.5F) * scaleRGB;
+            rgba[i][BCOMP] = ((arg0[i][BCOMP] * arg2[i][BCOMP]) +
+                              arg1[i][BCOMP] - 0.5F) * scaleRGB;
+	 }
+         break;
+      case GL_MODULATE_SUBTRACT_ATI:
+         for (i = 0; i < n; i++) {
+            rgba[i][RCOMP] = ((arg0[i][RCOMP] * arg2[i][RCOMP]) -
+                              arg1[i][RCOMP]) * scaleRGB;
+            rgba[i][GCOMP] = ((arg0[i][GCOMP] * arg2[i][GCOMP]) -
+                              arg1[i][GCOMP]) * scaleRGB;
+            rgba[i][BCOMP] = ((arg0[i][BCOMP] * arg2[i][BCOMP]) -
+                              arg1[i][BCOMP]) * scaleRGB;
+	 }
+         break;
+      case GL_BUMP_ENVMAP_ATI:
+         /* this produces a fixed rgba color, and the coord calc is done elsewhere */
+         for (i = 0; i < n; i++) {
+            /* rgba result is 0,0,0,1 */
+            rgba[i][RCOMP] = 0.0;
+            rgba[i][GCOMP] = 0.0;
+            rgba[i][BCOMP] = 0.0;
+            rgba[i][ACOMP] = 1.0;
+	 }
+         goto end; /* no alpha processing */
+      default:
+         _mesa_problem(ctx, "invalid combine mode");
+      }
+   }
+
+   /* Alpha channel combine */
+   {
+      float4_array arg0 = argA[0];
+      float4_array arg1 = argA[1];
+      float4_array arg2 = argA[2];
+      float4_array arg3 = argA[3];
+
+      switch (combine->ModeA) {
+      case GL_REPLACE:
+         for (i = 0; i < n; i++) {
+            rgba[i][ACOMP] = arg0[i][ACOMP] * scaleA;
+         }
+         break;
+      case GL_MODULATE:
+         for (i = 0; i < n; i++) {
+            rgba[i][ACOMP] = arg0[i][ACOMP] * arg1[i][ACOMP] * scaleA;
+         }
+         break;
+      case GL_ADD:
+         if (textureUnit->EnvMode == GL_COMBINE4_NV) {
+            /* (a * b) + (c * d) */
+            for (i = 0; i < n; i++) {
+               rgba[i][ACOMP] = (arg0[i][ACOMP] * arg1[i][ACOMP] +
+                                 arg2[i][ACOMP] * arg3[i][ACOMP]) * scaleA;
+            }
+         }
+         else {
+            /* two-term add */
+            for (i = 0; i < n; i++) {
+               rgba[i][ACOMP] = (arg0[i][ACOMP] + arg1[i][ACOMP]) * scaleA;
+            }
+         }
+         break;
+      case GL_ADD_SIGNED:
+         if (textureUnit->EnvMode == GL_COMBINE4_NV) {
+            /* (a * b) + (c * d) - 0.5 */
+            for (i = 0; i < n; i++) {
+               rgba[i][ACOMP] = (arg0[i][ACOMP] * arg1[i][ACOMP] +
+                                 arg2[i][ACOMP] * arg3[i][ACOMP] -
+                                 0.5F) * scaleA;
+            }
+         }
+         else {
+            /* a + b - 0.5 */
+            for (i = 0; i < n; i++) {
+               rgba[i][ACOMP] = (arg0[i][ACOMP] + arg1[i][ACOMP] - 0.5F) * scaleA;
+            }
+         }
+         break;
+      case GL_INTERPOLATE:
+         for (i = 0; i < n; i++) {
+            rgba[i][ACOMP] = (arg0[i][ACOMP] * arg2[i][ACOMP] +
+                              arg1[i][ACOMP] * (1.0F - arg2[i][ACOMP]))
+               * scaleA;
+         }
+         break;
+      case GL_SUBTRACT:
+         for (i = 0; i < n; i++) {
+            rgba[i][ACOMP] = (arg0[i][ACOMP] - arg1[i][ACOMP]) * scaleA;
+         }
+         break;
+      case GL_MODULATE_ADD_ATI:
+         for (i = 0; i < n; i++) {
+            rgba[i][ACOMP] = ((arg0[i][ACOMP] * arg2[i][ACOMP])
+                              + arg1[i][ACOMP]) * scaleA;
+         }
+         break;
+      case GL_MODULATE_SIGNED_ADD_ATI:
+         for (i = 0; i < n; i++) {
+            rgba[i][ACOMP] = ((arg0[i][ACOMP] * arg2[i][ACOMP]) +
+                              arg1[i][ACOMP] - 0.5F) * scaleA;
+         }
+         break;
+      case GL_MODULATE_SUBTRACT_ATI:
+         for (i = 0; i < n; i++) {
+            rgba[i][ACOMP] = ((arg0[i][ACOMP] * arg2[i][ACOMP])
+                              - arg1[i][ACOMP]) * scaleA;
+         }
+         break;
+      default:
+         _mesa_problem(ctx, "invalid combine mode");
+      }
+   }
+
+   /* Fix the alpha component for GL_DOT3_RGBA_EXT/ARB combining.
+    * This is kind of a kludge.  It would have been better if the spec
+    * were written such that the GL_COMBINE_ALPHA value could be set to
+    * GL_DOT3.
+    */
+   if (combine->ModeRGB == GL_DOT3_RGBA_EXT ||
+       combine->ModeRGB == GL_DOT3_RGBA) {
+      for (i = 0; i < n; i++) {
+	 rgba[i][ACOMP] = rgba[i][RCOMP];
+      }
+   }
+
+   for (i = 0; i < n; i++) {
+      UNCLAMPED_FLOAT_TO_CHAN(rgbaChan[i][RCOMP], rgba[i][RCOMP]);
+      UNCLAMPED_FLOAT_TO_CHAN(rgbaChan[i][GCOMP], rgba[i][GCOMP]);
+      UNCLAMPED_FLOAT_TO_CHAN(rgbaChan[i][BCOMP], rgba[i][BCOMP]);
+      UNCLAMPED_FLOAT_TO_CHAN(rgbaChan[i][ACOMP], rgba[i][ACOMP]);
+   }
+
+end:
+   for (i = 0; i < numArgsRGB || i < numArgsA; i++) {
+      free(ccolor[i]);
+   }
+   free(rgba);
+}
+
+
+/**
+ * Apply X/Y/Z/W/0/1 swizzle to an array of colors/texels.
+ * See GL_EXT_texture_swizzle.
+ */
+static void
+swizzle_texels(GLuint swizzle, GLuint count, float4_array texels)
+{
+   const GLuint swzR = GET_SWZ(swizzle, 0);
+   const GLuint swzG = GET_SWZ(swizzle, 1);
+   const GLuint swzB = GET_SWZ(swizzle, 2);
+   const GLuint swzA = GET_SWZ(swizzle, 3);
+   GLfloat vector[6];
+   GLuint i;
+
+   vector[SWIZZLE_ZERO] = 0;
+   vector[SWIZZLE_ONE] = 1.0F;
+
+   for (i = 0; i < count; i++) {
+      vector[SWIZZLE_X] = texels[i][0];
+      vector[SWIZZLE_Y] = texels[i][1];
+      vector[SWIZZLE_Z] = texels[i][2];
+      vector[SWIZZLE_W] = texels[i][3];
+      texels[i][RCOMP] = vector[swzR];
+      texels[i][GCOMP] = vector[swzG];
+      texels[i][BCOMP] = vector[swzB];
+      texels[i][ACOMP] = vector[swzA];
+   }
+}
+
+
+/**
+ * Apply texture mapping to a span of fragments.
+ */
+void
+_swrast_texture_span( struct gl_context *ctx, SWspan *span )
+{
+   SWcontext *swrast = SWRAST_CONTEXT(ctx);
+   float4_array primary_rgba;
+   GLuint unit;
+
+   primary_rgba = (float4_array) malloc(span->end * 4 * sizeof(GLfloat));
+
+   if (!primary_rgba) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "texture_span");
+      return;
+   }
+
+   ASSERT(span->end <= MAX_WIDTH);
+
+   /*
+    * Save copy of the incoming fragment colors (the GL_PRIMARY_COLOR)
+    */
+   if (swrast->_TextureCombinePrimary) {
+      GLuint i;
+      for (i = 0; i < span->end; i++) {
+         primary_rgba[i][RCOMP] = CHAN_TO_FLOAT(span->array->rgba[i][RCOMP]);
+         primary_rgba[i][GCOMP] = CHAN_TO_FLOAT(span->array->rgba[i][GCOMP]);
+         primary_rgba[i][BCOMP] = CHAN_TO_FLOAT(span->array->rgba[i][BCOMP]);
+         primary_rgba[i][ACOMP] = CHAN_TO_FLOAT(span->array->rgba[i][ACOMP]);
+      }
+   }
+
+   /* First must sample all bump maps */
+   for (unit = 0; unit < ctx->Const.MaxTextureUnits; unit++) {
+      const struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
+
+      if (texUnit->_ReallyEnabled &&
+         texUnit->_CurrentCombine->ModeRGB == GL_BUMP_ENVMAP_ATI) {
+         const GLfloat (*texcoords)[4] = (const GLfloat (*)[4])
+            span->array->attribs[FRAG_ATTRIB_TEX0 + unit];
+         float4_array targetcoords =
+            span->array->attribs[FRAG_ATTRIB_TEX0 +
+               ctx->Texture.Unit[unit].BumpTarget - GL_TEXTURE0];
+
+         const struct gl_texture_object *curObj = texUnit->_Current;
+         GLfloat *lambda = span->array->lambda[unit];
+         float4_array texels = get_texel_array(swrast, unit);
+         GLuint i;
+         GLfloat rotMatrix00 = ctx->Texture.Unit[unit].RotMatrix[0];
+         GLfloat rotMatrix01 = ctx->Texture.Unit[unit].RotMatrix[1];
+         GLfloat rotMatrix10 = ctx->Texture.Unit[unit].RotMatrix[2];
+         GLfloat rotMatrix11 = ctx->Texture.Unit[unit].RotMatrix[3];
+
+         /* adjust texture lod (lambda) */
+         if (span->arrayMask & SPAN_LAMBDA) {
+            if (texUnit->LodBias + curObj->Sampler.LodBias != 0.0F) {
+               /* apply LOD bias, but don't clamp yet */
+               const GLfloat bias = CLAMP(texUnit->LodBias + curObj->Sampler.LodBias,
+                                          -ctx->Const.MaxTextureLodBias,
+                                          ctx->Const.MaxTextureLodBias);
+               GLuint i;
+               for (i = 0; i < span->end; i++) {
+                  lambda[i] += bias;
+               }
+            }
+
+            if (curObj->Sampler.MinLod != -1000.0 ||
+                curObj->Sampler.MaxLod != 1000.0) {
+               /* apply LOD clamping to lambda */
+               const GLfloat min = curObj->Sampler.MinLod;
+               const GLfloat max = curObj->Sampler.MaxLod;
+               GLuint i;
+               for (i = 0; i < span->end; i++) {
+                  GLfloat l = lambda[i];
+                  lambda[i] = CLAMP(l, min, max);
+               }
+            }
+         }
+
+         /* Sample the texture (span->end = number of fragments) */
+         swrast->TextureSample[unit]( ctx, texUnit->_Current, span->end,
+                                      texcoords, lambda, texels );
+
+         /* manipulate the span values of the bump target
+            not sure this can work correctly even ignoring
+            the problem that channel is unsigned */
+         for (i = 0; i < span->end; i++) {
+            targetcoords[i][0] += (texels[i][0] * rotMatrix00 + texels[i][1] *
+                                  rotMatrix01) / targetcoords[i][3];
+            targetcoords[i][1] += (texels[i][0] * rotMatrix10 + texels[i][1] *
+                                  rotMatrix11) / targetcoords[i][3];
+         }
+      }
+   }
+
+   /*
+    * Must do all texture sampling before combining in order to
+    * accomodate GL_ARB_texture_env_crossbar.
+    */
+   for (unit = 0; unit < ctx->Const.MaxTextureUnits; unit++) {
+      const struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
+      if (texUnit->_ReallyEnabled &&
+          texUnit->_CurrentCombine->ModeRGB != GL_BUMP_ENVMAP_ATI) {
+         const GLfloat (*texcoords)[4] = (const GLfloat (*)[4])
+            span->array->attribs[FRAG_ATTRIB_TEX0 + unit];
+         const struct gl_texture_object *curObj = texUnit->_Current;
+         GLfloat *lambda = span->array->lambda[unit];
+         float4_array texels = get_texel_array(swrast, unit);
+
+         /* adjust texture lod (lambda) */
+         if (span->arrayMask & SPAN_LAMBDA) {
+            if (texUnit->LodBias + curObj->Sampler.LodBias != 0.0F) {
+               /* apply LOD bias, but don't clamp yet */
+               const GLfloat bias = CLAMP(texUnit->LodBias + curObj->Sampler.LodBias,
+                                          -ctx->Const.MaxTextureLodBias,
+                                          ctx->Const.MaxTextureLodBias);
+               GLuint i;
+               for (i = 0; i < span->end; i++) {
+                  lambda[i] += bias;
+               }
+            }
+
+            if (curObj->Sampler.MinLod != -1000.0 ||
+                curObj->Sampler.MaxLod != 1000.0) {
+               /* apply LOD clamping to lambda */
+               const GLfloat min = curObj->Sampler.MinLod;
+               const GLfloat max = curObj->Sampler.MaxLod;
+               GLuint i;
+               for (i = 0; i < span->end; i++) {
+                  GLfloat l = lambda[i];
+                  lambda[i] = CLAMP(l, min, max);
+               }
+            }
+         }
+         else if (curObj->Sampler.MaxAnisotropy > 1.0 &&
+                  curObj->Sampler.MinFilter == GL_LINEAR_MIPMAP_LINEAR) {
+            /* sample_lambda_2d_aniso is beeing used as texture_sample_func,
+             * it requires the current SWspan *span as an additional parameter.
+             * In order to keep the same function signature, the unused lambda
+             * parameter will be modified to actually contain the SWspan pointer.
+             * This is a Hack. To make it right, the texture_sample_func
+             * signature and all implementing functions need to be modified.
+             */
+            /* "hide" SWspan struct; cast to (GLfloat *) to suppress warning */
+            lambda = (GLfloat *)span;
+         }
+
+         /* Sample the texture (span->end = number of fragments) */
+         swrast->TextureSample[unit]( ctx, texUnit->_Current, span->end,
+                                      texcoords, lambda, texels );
+
+         /* GL_EXT_texture_swizzle */
+         if (curObj->_Swizzle != SWIZZLE_NOOP) {
+            swizzle_texels(curObj->_Swizzle, span->end, texels);
+         }
+      }
+   }
+
+   /*
+    * OK, now apply the texture (aka texture combine/blend).
+    * We modify the span->color.rgba values.
+    */
+   for (unit = 0; unit < ctx->Const.MaxTextureUnits; unit++) {
+      if (ctx->Texture.Unit[unit]._ReallyEnabled) {
+         texture_combine( ctx, unit, span->end,
+                          primary_rgba,
+                          swrast->TexelBuffer,
+                          span->array->rgba );
+      }
+   }
+
+   free(primary_rgba);
+}
diff --git a/mesalib/src/mesa/swrast/s_texfilter.c b/mesalib/src/mesa/swrast/s_texfilter.c
index 106f8b75f..237e5d28a 100644
--- a/mesalib/src/mesa/swrast/s_texfilter.c
+++ b/mesalib/src/mesa/swrast/s_texfilter.c
@@ -1,3315 +1,3710 @@
-/*
- * Mesa 3-D graphics library
- * Version:  7.3
- *
- * Copyright (C) 1999-2008  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"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included
- * in all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
- * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-
-
-#include "main/glheader.h"
-#include "main/context.h"
-#include "main/colormac.h"
-#include "main/imports.h"
-
-#include "s_context.h"
-#include "s_texfilter.h"
-
-
-/*
- * Note, the FRAC macro has to work perfectly.  Otherwise you'll sometimes
- * see 1-pixel bands of improperly weighted linear-filtered textures.
- * The tests/texwrap.c demo is a good test.
- * Also note, FRAC(x) doesn't truly return the fractional part of x for x < 0.
- * Instead, if x < 0 then FRAC(x) = 1 - true_frac(x).
- */
-#define FRAC(f)  ((f) - IFLOOR(f))
-
-
-
-/**
- * Linear interpolation macro
- */
-#define LERP(T, A, B)  ( (A) + (T) * ((B) - (A)) )
-
-
-/**
- * Do 2D/biliner interpolation of float values.
- * v00, v10, v01 and v11 are typically four texture samples in a square/box.
- * a and b are the horizontal and vertical interpolants.
- * It's important that this function is inlined when compiled with
- * optimization!  If we find that's not true on some systems, convert
- * to a macro.
- */
-static INLINE GLfloat
-lerp_2d(GLfloat a, GLfloat b,
-        GLfloat v00, GLfloat v10, GLfloat v01, GLfloat v11)
-{
-   const GLfloat temp0 = LERP(a, v00, v10);
-   const GLfloat temp1 = LERP(a, v01, v11);
-   return LERP(b, temp0, temp1);
-}
-
-
-/**
- * Do 3D/trilinear interpolation of float values.
- * \sa lerp_2d
- */
-static INLINE GLfloat
-lerp_3d(GLfloat a, GLfloat b, GLfloat c,
-        GLfloat v000, GLfloat v100, GLfloat v010, GLfloat v110,
-        GLfloat v001, GLfloat v101, GLfloat v011, GLfloat v111)
-{
-   const GLfloat temp00 = LERP(a, v000, v100);
-   const GLfloat temp10 = LERP(a, v010, v110);
-   const GLfloat temp01 = LERP(a, v001, v101);
-   const GLfloat temp11 = LERP(a, v011, v111);
-   const GLfloat temp0 = LERP(b, temp00, temp10);
-   const GLfloat temp1 = LERP(b, temp01, temp11);
-   return LERP(c, temp0, temp1);
-}
-
-
-/**
- * Do linear interpolation of colors.
- */
-static INLINE void
-lerp_rgba(GLfloat result[4], GLfloat t, const GLfloat a[4], const GLfloat b[4])
-{
-   result[0] = LERP(t, a[0], b[0]);
-   result[1] = LERP(t, a[1], b[1]);
-   result[2] = LERP(t, a[2], b[2]);
-   result[3] = LERP(t, a[3], b[3]);
-}
-
-
-/**
- * Do bilinear interpolation of colors.
- */
-static INLINE void
-lerp_rgba_2d(GLfloat result[4], GLfloat a, GLfloat b,
-             const GLfloat t00[4], const GLfloat t10[4],
-             const GLfloat t01[4], const GLfloat t11[4])
-{
-   result[0] = lerp_2d(a, b, t00[0], t10[0], t01[0], t11[0]);
-   result[1] = lerp_2d(a, b, t00[1], t10[1], t01[1], t11[1]);
-   result[2] = lerp_2d(a, b, t00[2], t10[2], t01[2], t11[2]);
-   result[3] = lerp_2d(a, b, t00[3], t10[3], t01[3], t11[3]);
-}
-
-
-/**
- * Do trilinear interpolation of colors.
- */
-static INLINE void
-lerp_rgba_3d(GLfloat result[4], GLfloat a, GLfloat b, GLfloat c,
-             const GLfloat t000[4], const GLfloat t100[4],
-             const GLfloat t010[4], const GLfloat t110[4],
-             const GLfloat t001[4], const GLfloat t101[4],
-             const GLfloat t011[4], const GLfloat t111[4])
-{
-   GLuint k;
-   /* compiler should unroll these short loops */
-   for (k = 0; k < 4; k++) {
-      result[k] = lerp_3d(a, b, c, t000[k], t100[k], t010[k], t110[k],
-                                   t001[k], t101[k], t011[k], t111[k]);
-   }
-}
-
-
-/**
- * Used for GL_REPEAT wrap mode.  Using A % B doesn't produce the
- * right results for A<0.  Casting to A to be unsigned only works if B
- * is a power of two.  Adding a bias to A (which is a multiple of B)
- * avoids the problems with A < 0 (for reasonable A) without using a
- * conditional.
- */
-#define REMAINDER(A, B) (((A) + (B) * 1024) % (B))
-
-
-/**
- * Used to compute texel locations for linear sampling.
- * Input:
- *    wrapMode = GL_REPEAT, GL_CLAMP, GL_CLAMP_TO_EDGE, GL_CLAMP_TO_BORDER
- *    s = texcoord in [0,1]
- *    size = width (or height or depth) of texture
- * Output:
- *    i0, i1 = returns two nearest texel indexes
- *    weight = returns blend factor between texels
- */
-static INLINE void
-linear_texel_locations(GLenum wrapMode,
-                       const struct gl_texture_image *img,
-                       GLint size, GLfloat s,
-                       GLint *i0, GLint *i1, GLfloat *weight)
-{
-   GLfloat u;
-   switch (wrapMode) {
-   case GL_REPEAT:
-      u = s * size - 0.5F;
-      if (img->_IsPowerOfTwo) {
-         *i0 = IFLOOR(u) & (size - 1);
-         *i1 = (*i0 + 1) & (size - 1);
-      }
-      else {
-         *i0 = REMAINDER(IFLOOR(u), size);
-         *i1 = REMAINDER(*i0 + 1, size);
-      }
-      break;
-   case GL_CLAMP_TO_EDGE:
-      if (s <= 0.0F)
-         u = 0.0F;
-      else if (s >= 1.0F)
-         u = (GLfloat) size;
-      else
-         u = s * size;
-      u -= 0.5F;
-      *i0 = IFLOOR(u);
-      *i1 = *i0 + 1;
-      if (*i0 < 0)
-         *i0 = 0;
-      if (*i1 >= (GLint) size)
-         *i1 = size - 1;
-      break;
-   case GL_CLAMP_TO_BORDER:
-      {
-         const GLfloat min = -1.0F / (2.0F * size);
-         const GLfloat max = 1.0F - min;
-         if (s <= min)
-            u = min * size;
-         else if (s >= max)
-            u = max * size;
-         else
-            u = s * size;
-         u -= 0.5F;
-         *i0 = IFLOOR(u);
-         *i1 = *i0 + 1;
-      }
-      break;
-   case GL_MIRRORED_REPEAT:
-      {
-         const GLint flr = IFLOOR(s);
-         if (flr & 1)
-            u = 1.0F - (s - (GLfloat) flr);
-         else
-            u = s - (GLfloat) flr;
-         u = (u * size) - 0.5F;
-         *i0 = IFLOOR(u);
-         *i1 = *i0 + 1;
-         if (*i0 < 0)
-            *i0 = 0;
-         if (*i1 >= (GLint) size)
-            *i1 = size - 1;
-      }
-      break;
-   case GL_MIRROR_CLAMP_EXT:
-      u = FABSF(s);
-      if (u >= 1.0F)
-         u = (GLfloat) size;
-      else
-         u *= size;
-      u -= 0.5F;
-      *i0 = IFLOOR(u);
-      *i1 = *i0 + 1;
-      break;
-   case GL_MIRROR_CLAMP_TO_EDGE_EXT:
-      u = FABSF(s);
-      if (u >= 1.0F)
-         u = (GLfloat) size;
-      else
-         u *= size;
-      u -= 0.5F;
-      *i0 = IFLOOR(u);
-      *i1 = *i0 + 1;
-      if (*i0 < 0)
-         *i0 = 0;
-      if (*i1 >= (GLint) size)
-         *i1 = size - 1;
-      break;
-   case GL_MIRROR_CLAMP_TO_BORDER_EXT:
-      {
-         const GLfloat min = -1.0F / (2.0F * size);
-         const GLfloat max = 1.0F - min;
-         u = FABSF(s);
-         if (u <= min)
-            u = min * size;
-         else if (u >= max)
-            u = max * size;
-         else
-            u *= size;
-         u -= 0.5F;
-         *i0 = IFLOOR(u);
-         *i1 = *i0 + 1;
-      }
-      break;
-   case GL_CLAMP:
-      if (s <= 0.0F)
-         u = 0.0F;
-      else if (s >= 1.0F)
-         u = (GLfloat) size;
-      else
-         u = s * size;
-      u -= 0.5F;
-      *i0 = IFLOOR(u);
-      *i1 = *i0 + 1;
-      break;
-   default:
-      _mesa_problem(NULL, "Bad wrap mode");
-      u = 0.0F;
-   }
-   *weight = FRAC(u);
-}
-
-
-/**
- * Used to compute texel location for nearest sampling.
- */
-static INLINE GLint
-nearest_texel_location(GLenum wrapMode,
-                       const struct gl_texture_image *img,
-                       GLint size, GLfloat s)
-{
-   GLint i;
-
-   switch (wrapMode) {
-   case GL_REPEAT:
-      /* s limited to [0,1) */
-      /* i limited to [0,size-1] */
-      i = IFLOOR(s * size);
-      if (img->_IsPowerOfTwo)
-         i &= (size - 1);
-      else
-         i = REMAINDER(i, size);
-      return i;
-   case GL_CLAMP_TO_EDGE:
-      {
-         /* s limited to [min,max] */
-         /* i limited to [0, size-1] */
-         const GLfloat min = 1.0F / (2.0F * size);
-         const GLfloat max = 1.0F - min;
-         if (s < min)
-            i = 0;
-         else if (s > max)
-            i = size - 1;
-         else
-            i = IFLOOR(s * size);
-      }
-      return i;
-   case GL_CLAMP_TO_BORDER:
-      {
-         /* s limited to [min,max] */
-         /* i limited to [-1, size] */
-         const GLfloat min = -1.0F / (2.0F * size);
-         const GLfloat max = 1.0F - min;
-         if (s <= min)
-            i = -1;
-         else if (s >= max)
-            i = size;
-         else
-            i = IFLOOR(s * size);
-      }
-      return i;
-   case GL_MIRRORED_REPEAT:
-      {
-         const GLfloat min = 1.0F / (2.0F * size);
-         const GLfloat max = 1.0F - min;
-         const GLint flr = IFLOOR(s);
-         GLfloat u;
-         if (flr & 1)
-            u = 1.0F - (s - (GLfloat) flr);
-         else
-            u = s - (GLfloat) flr;
-         if (u < min)
-            i = 0;
-         else if (u > max)
-            i = size - 1;
-         else
-            i = IFLOOR(u * size);
-      }
-      return i;
-   case GL_MIRROR_CLAMP_EXT:
-      {
-         /* s limited to [0,1] */
-         /* i limited to [0,size-1] */
-         const GLfloat u = FABSF(s);
-         if (u <= 0.0F)
-            i = 0;
-         else if (u >= 1.0F)
-            i = size - 1;
-         else
-            i = IFLOOR(u * size);
-      }
-      return i;
-   case GL_MIRROR_CLAMP_TO_EDGE_EXT:
-      {
-         /* s limited to [min,max] */
-         /* i limited to [0, size-1] */
-         const GLfloat min = 1.0F / (2.0F * size);
-         const GLfloat max = 1.0F - min;
-         const GLfloat u = FABSF(s);
-         if (u < min)
-            i = 0;
-         else if (u > max)
-            i = size - 1;
-         else
-            i = IFLOOR(u * size);
-      }
-      return i;
-   case GL_MIRROR_CLAMP_TO_BORDER_EXT:
-      {
-         /* s limited to [min,max] */
-         /* i limited to [0, size-1] */
-         const GLfloat min = -1.0F / (2.0F * size);
-         const GLfloat max = 1.0F - min;
-         const GLfloat u = FABSF(s);
-         if (u < min)
-            i = -1;
-         else if (u > max)
-            i = size;
-         else
-            i = IFLOOR(u * size);
-      }
-      return i;
-   case GL_CLAMP:
-      /* s limited to [0,1] */
-      /* i limited to [0,size-1] */
-      if (s <= 0.0F)
-         i = 0;
-      else if (s >= 1.0F)
-         i = size - 1;
-      else
-         i = IFLOOR(s * size);
-      return i;
-   default:
-      _mesa_problem(NULL, "Bad wrap mode");
-      return 0;
-   }
-}
-
-
-/* Power of two image sizes only */
-static INLINE void
-linear_repeat_texel_location(GLuint size, GLfloat s,
-                             GLint *i0, GLint *i1, GLfloat *weight)
-{
-   GLfloat u = s * size - 0.5F;
-   *i0 = IFLOOR(u) & (size - 1);
-   *i1 = (*i0 + 1) & (size - 1);
-   *weight = FRAC(u);
-}
-
-
-/**
- * Do clamp/wrap for a texture rectangle coord, GL_NEAREST filter mode.
- */
-static INLINE GLint
-clamp_rect_coord_nearest(GLenum wrapMode, GLfloat coord, GLint max)
-{
-   switch (wrapMode) {
-   case GL_CLAMP:
-      return IFLOOR( CLAMP(coord, 0.0F, max - 1) );
-   case GL_CLAMP_TO_EDGE:
-      return IFLOOR( CLAMP(coord, 0.5F, max - 0.5F) );
-   case GL_CLAMP_TO_BORDER:
-      return IFLOOR( CLAMP(coord, -0.5F, max + 0.5F) );
-   default:
-      _mesa_problem(NULL, "bad wrapMode in clamp_rect_coord_nearest");
-      return 0;
-   }
-}
-
-
-/**
- * As above, but GL_LINEAR filtering.
- */
-static INLINE void
-clamp_rect_coord_linear(GLenum wrapMode, GLfloat coord, GLint max,
-                        GLint *i0out, GLint *i1out, GLfloat *weight)
-{
-   GLfloat fcol;
-   GLint i0, i1;
-   switch (wrapMode) {
-   case GL_CLAMP:
-      /* Not exactly what the spec says, but it matches NVIDIA output */
-      fcol = CLAMP(coord - 0.5F, 0.0F, max - 1);
-      i0 = IFLOOR(fcol);
-      i1 = i0 + 1;
-      break;
-   case GL_CLAMP_TO_EDGE:
-      fcol = CLAMP(coord, 0.5F, max - 0.5F);
-      fcol -= 0.5F;
-      i0 = IFLOOR(fcol);
-      i1 = i0 + 1;
-      if (i1 > max - 1)
-         i1 = max - 1;
-      break;
-   case GL_CLAMP_TO_BORDER:
-      fcol = CLAMP(coord, -0.5F, max + 0.5F);
-      fcol -= 0.5F;
-      i0 = IFLOOR(fcol);
-      i1 = i0 + 1;
-      break;
-   default:
-      _mesa_problem(NULL, "bad wrapMode in clamp_rect_coord_linear");
-      i0 = i1 = 0;
-      fcol = 0.0F;
-   }
-   *i0out = i0;
-   *i1out = i1;
-   *weight = FRAC(fcol);
-}
-
-
-/**
- * Compute slice/image to use for 1D or 2D array texture.
- */
-static INLINE GLint
-tex_array_slice(GLfloat coord, GLsizei size)
-{
-   GLint slice = IFLOOR(coord + 0.5f);
-   slice = CLAMP(slice, 0, size - 1);
-   return slice;
-}
-
-
-/**
- * Compute nearest integer texcoords for given texobj and coordinate.
- * NOTE: only used for depth texture sampling.
- */
-static INLINE void
-nearest_texcoord(const struct gl_texture_object *texObj,
-                 GLuint level,
-                 const GLfloat texcoord[4],
-                 GLint *i, GLint *j, GLint *k)
-{
-   const struct gl_texture_image *img = texObj->Image[0][level];
-   const GLint width = img->Width;
-   const GLint height = img->Height;
-   const GLint depth = img->Depth;
-
-   switch (texObj->Target) {
-   case GL_TEXTURE_RECTANGLE_ARB:
-      *i = clamp_rect_coord_nearest(texObj->Sampler.WrapS, texcoord[0], width);
-      *j = clamp_rect_coord_nearest(texObj->Sampler.WrapT, texcoord[1], height);
-      *k = 0;
-      break;
-   case GL_TEXTURE_1D:
-      *i = nearest_texel_location(texObj->Sampler.WrapS, img, width, texcoord[0]);
-      *j = 0;
-      *k = 0;
-      break;
-   case GL_TEXTURE_2D:
-      *i = nearest_texel_location(texObj->Sampler.WrapS, img, width, texcoord[0]);
-      *j = nearest_texel_location(texObj->Sampler.WrapT, img, height, texcoord[1]);
-      *k = 0;
-      break;
-   case GL_TEXTURE_1D_ARRAY_EXT:
-      *i = nearest_texel_location(texObj->Sampler.WrapS, img, width, texcoord[0]);
-      *j = tex_array_slice(texcoord[1], height);
-      *k = 0;
-      break;
-   case GL_TEXTURE_2D_ARRAY_EXT:
-      *i = nearest_texel_location(texObj->Sampler.WrapS, img, width, texcoord[0]);
-      *j = nearest_texel_location(texObj->Sampler.WrapT, img, height, texcoord[1]);
-      *k = tex_array_slice(texcoord[2], depth);
-      break;
-   default:
-      *i = *j = *k = 0;
-   }
-}
-
-
-/**
- * Compute linear integer texcoords for given texobj and coordinate.
- * NOTE: only used for depth texture sampling.
- */
-static INLINE void
-linear_texcoord(const struct gl_texture_object *texObj,
-                GLuint level,
-                const GLfloat texcoord[4],
-                GLint *i0, GLint *i1, GLint *j0, GLint *j1, GLint *slice,
-                GLfloat *wi, GLfloat *wj)
-{
-   const struct gl_texture_image *img = texObj->Image[0][level];
-   const GLint width = img->Width;
-   const GLint height = img->Height;
-   const GLint depth = img->Depth;
-
-   switch (texObj->Target) {
-   case GL_TEXTURE_RECTANGLE_ARB:
-      clamp_rect_coord_linear(texObj->Sampler.WrapS, texcoord[0],
-                              width, i0, i1, wi);
-      clamp_rect_coord_linear(texObj->Sampler.WrapT, texcoord[1],
-                              height, j0, j1, wj);
-      *slice = 0;
-      break;
-
-   case GL_TEXTURE_1D:
-   case GL_TEXTURE_2D:
-      linear_texel_locations(texObj->Sampler.WrapS, img, width,
-                             texcoord[0], i0, i1, wi);
-      linear_texel_locations(texObj->Sampler.WrapT, img, height,
-                             texcoord[1], j0, j1, wj);
-      *slice = 0;
-      break;
-
-   case GL_TEXTURE_1D_ARRAY_EXT:
-      linear_texel_locations(texObj->Sampler.WrapS, img, width,
-                             texcoord[0], i0, i1, wi);
-      *j0 = tex_array_slice(texcoord[1], height);
-      *j1 = *j0;
-      *slice = 0;
-      break;
-
-   case GL_TEXTURE_2D_ARRAY_EXT:
-      linear_texel_locations(texObj->Sampler.WrapS, img, width,
-                             texcoord[0], i0, i1, wi);
-      linear_texel_locations(texObj->Sampler.WrapT, img, height,
-                             texcoord[1], j0, j1, wj);
-      *slice = tex_array_slice(texcoord[2], depth);
-      break;
-
-   default:
-      *slice = 0;
-   }
-}
-
-
-
-/**
- * For linear interpolation between mipmap levels N and N+1, this function
- * computes N.
- */
-static INLINE GLint
-linear_mipmap_level(const struct gl_texture_object *tObj, GLfloat lambda)
-{
-   if (lambda < 0.0F)
-      return tObj->BaseLevel;
-   else if (lambda > tObj->_MaxLambda)
-      return (GLint) (tObj->BaseLevel + tObj->_MaxLambda);
-   else
-      return (GLint) (tObj->BaseLevel + lambda);
-}
-
-
-/**
- * Compute the nearest mipmap level to take texels from.
- */
-static INLINE GLint
-nearest_mipmap_level(const struct gl_texture_object *tObj, GLfloat lambda)
-{
-   GLfloat l;
-   GLint level;
-   if (lambda <= 0.5F)
-      l = 0.0F;
-   else if (lambda > tObj->_MaxLambda + 0.4999F)
-      l = tObj->_MaxLambda + 0.4999F;
-   else
-      l = lambda;
-   level = (GLint) (tObj->BaseLevel + l + 0.5F);
-   if (level > tObj->_MaxLevel)
-      level = tObj->_MaxLevel;
-   return level;
-}
-
-
-
-/*
- * Bitflags for texture border color sampling.
- */
-#define I0BIT   1
-#define I1BIT   2
-#define J0BIT   4
-#define J1BIT   8
-#define K0BIT  16
-#define K1BIT  32
-
-
-
-/**
- * The lambda[] array values are always monotonic.  Either the whole span
- * will be minified, magnified, or split between the two.  This function
- * determines the subranges in [0, n-1] that are to be minified or magnified.
- */
-static INLINE void
-compute_min_mag_ranges(const struct gl_texture_object *tObj,
-                       GLuint n, const GLfloat lambda[],
-                       GLuint *minStart, GLuint *minEnd,
-                       GLuint *magStart, GLuint *magEnd)
-{
-   GLfloat minMagThresh;
-
-   /* we shouldn't be here if minfilter == magfilter */
-   ASSERT(tObj->Sampler.MinFilter != tObj->Sampler.MagFilter);
-
-   /* This bit comes from the OpenGL spec: */
-   if (tObj->Sampler.MagFilter == GL_LINEAR
-       && (tObj->Sampler.MinFilter == GL_NEAREST_MIPMAP_NEAREST ||
-           tObj->Sampler.MinFilter == GL_NEAREST_MIPMAP_LINEAR)) {
-      minMagThresh = 0.5F;
-   }
-   else {
-      minMagThresh = 0.0F;
-   }
-
-#if 0
-   /* DEBUG CODE: Verify that lambda[] is monotonic.
-    * We can't really use this because the inaccuracy in the LOG2 function
-    * causes this test to fail, yet the resulting texturing is correct.
-    */
-   if (n > 1) {
-      GLuint i;
-      printf("lambda delta = %g\n", lambda[0] - lambda[n-1]);
-      if (lambda[0] >= lambda[n-1]) { /* decreasing */
-         for (i = 0; i < n - 1; i++) {
-            ASSERT((GLint) (lambda[i] * 10) >= (GLint) (lambda[i+1] * 10));
-         }
-      }
-      else { /* increasing */
-         for (i = 0; i < n - 1; i++) {
-            ASSERT((GLint) (lambda[i] * 10) <= (GLint) (lambda[i+1] * 10));
-         }
-      }
-   }
-#endif /* DEBUG */
-
-   if (lambda[0] <= minMagThresh && (n <= 1 || lambda[n-1] <= minMagThresh)) {
-      /* magnification for whole span */
-      *magStart = 0;
-      *magEnd = n;
-      *minStart = *minEnd = 0;
-   }
-   else if (lambda[0] > minMagThresh && (n <=1 || lambda[n-1] > minMagThresh)) {
-      /* minification for whole span */
-      *minStart = 0;
-      *minEnd = n;
-      *magStart = *magEnd = 0;
-   }
-   else {
-      /* a mix of minification and magnification */
-      GLuint i;
-      if (lambda[0] > minMagThresh) {
-         /* start with minification */
-         for (i = 1; i < n; i++) {
-            if (lambda[i] <= minMagThresh)
-               break;
-         }
-         *minStart = 0;
-         *minEnd = i;
-         *magStart = i;
-         *magEnd = n;
-      }
-      else {
-         /* start with magnification */
-         for (i = 1; i < n; i++) {
-            if (lambda[i] > minMagThresh)
-               break;
-         }
-         *magStart = 0;
-         *magEnd = i;
-         *minStart = i;
-         *minEnd = n;
-      }
-   }
-
-#if 0
-   /* Verify the min/mag Start/End values
-    * We don't use this either (see above)
-    */
-   {
-      GLint i;
-      for (i = 0; i < n; i++) {
-         if (lambda[i] > minMagThresh) {
-            /* minification */
-            ASSERT(i >= *minStart);
-            ASSERT(i < *minEnd);
-         }
-         else {
-            /* magnification */
-            ASSERT(i >= *magStart);
-            ASSERT(i < *magEnd);
-         }
-      }
-   }
-#endif
-}
-
-
-/**
- * When we sample the border color, it must be interpreted according to
- * the base texture format.  Ex: if the texture base format it GL_ALPHA,
- * we return (0,0,0,BorderAlpha).
- */
-static INLINE void
-get_border_color(const struct gl_texture_object *tObj,
-                 const struct gl_texture_image *img,
-                 GLfloat rgba[4])
-{
-   switch (img->_BaseFormat) {
-   case GL_RGB:
-      rgba[0] = tObj->Sampler.BorderColor.f[0];
-      rgba[1] = tObj->Sampler.BorderColor.f[1];
-      rgba[2] = tObj->Sampler.BorderColor.f[2];
-      rgba[3] = 1.0F;
-      break;
-   case GL_ALPHA:
-      rgba[0] = rgba[1] = rgba[2] = 0.0;
-      rgba[3] = tObj->Sampler.BorderColor.f[3];
-      break;
-   case GL_LUMINANCE:
-      rgba[0] = rgba[1] = rgba[2] = tObj->Sampler.BorderColor.f[0];
-      rgba[3] = 1.0;
-      break;
-   case GL_LUMINANCE_ALPHA:
-      rgba[0] = rgba[1] = rgba[2] = tObj->Sampler.BorderColor.f[0];
-      rgba[3] = tObj->Sampler.BorderColor.f[3];
-      break;
-   case GL_INTENSITY:
-      rgba[0] = rgba[1] = rgba[2] = rgba[3] = tObj->Sampler.BorderColor.f[0];
-      break;
-   default:
-      COPY_4V(rgba, tObj->Sampler.BorderColor.f);
-   }
-}
-
-
-/**********************************************************************/
-/*                    1-D Texture Sampling Functions                  */
-/**********************************************************************/
-
-/**
- * Return the texture sample for coordinate (s) using GL_NEAREST filter.
- */
-static INLINE void
-sample_1d_nearest(struct gl_context *ctx,
-                  const struct gl_texture_object *tObj,
-                  const struct gl_texture_image *img,
-                  const GLfloat texcoord[4], GLfloat rgba[4])
-{
-   const GLint width = img->Width2;  /* without border, power of two */
-   GLint i;
-   i = nearest_texel_location(tObj->Sampler.WrapS, img, width, texcoord[0]);
-   /* skip over the border, if any */
-   i += img->Border;
-   if (i < 0 || i >= (GLint) img->Width) {
-      /* Need this test for GL_CLAMP_TO_BORDER mode */
-      get_border_color(tObj, img, rgba);
-   }
-   else {
-      img->FetchTexelf(img, i, 0, 0, rgba);
-   }
-}
-
-
-/**
- * Return the texture sample for coordinate (s) using GL_LINEAR filter.
- */
-static INLINE void
-sample_1d_linear(struct gl_context *ctx,
-                 const struct gl_texture_object *tObj,
-                 const struct gl_texture_image *img,
-                 const GLfloat texcoord[4], GLfloat rgba[4])
-{
-   const GLint width = img->Width2;
-   GLint i0, i1;
-   GLbitfield useBorderColor = 0x0;
-   GLfloat a;
-   GLfloat t0[4], t1[4];  /* texels */
-
-   linear_texel_locations(tObj->Sampler.WrapS, img, width, texcoord[0], &i0, &i1, &a);
-
-   if (img->Border) {
-      i0 += img->Border;
-      i1 += img->Border;
-   }
-   else {
-      if (i0 < 0 || i0 >= width)   useBorderColor |= I0BIT;
-      if (i1 < 0 || i1 >= width)   useBorderColor |= I1BIT;
-   }
-
-   /* fetch texel colors */
-   if (useBorderColor & I0BIT) {
-      get_border_color(tObj, img, t0);
-   }
-   else {
-      img->FetchTexelf(img, i0, 0, 0, t0);
-   }
-   if (useBorderColor & I1BIT) {
-      get_border_color(tObj, img, t1);
-   }
-   else {
-      img->FetchTexelf(img, i1, 0, 0, t1);
-   }
-
-   lerp_rgba(rgba, a, t0, t1);
-}
-
-
-static void
-sample_1d_nearest_mipmap_nearest(struct gl_context *ctx,
-                                 const struct gl_texture_object *tObj,
-                                 GLuint n, const GLfloat texcoord[][4],
-                                 const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   ASSERT(lambda != NULL);
-   for (i = 0; i < n; i++) {
-      GLint level = nearest_mipmap_level(tObj, lambda[i]);
-      sample_1d_nearest(ctx, tObj, tObj->Image[0][level], texcoord[i], rgba[i]);
-   }
-}
-
-
-static void
-sample_1d_linear_mipmap_nearest(struct gl_context *ctx,
-                                const struct gl_texture_object *tObj,
-                                GLuint n, const GLfloat texcoord[][4],
-                                const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   ASSERT(lambda != NULL);
-   for (i = 0; i < n; i++) {
-      GLint level = nearest_mipmap_level(tObj, lambda[i]);
-      sample_1d_linear(ctx, tObj, tObj->Image[0][level], texcoord[i], rgba[i]);
-   }
-}
-
-
-static void
-sample_1d_nearest_mipmap_linear(struct gl_context *ctx,
-                                const struct gl_texture_object *tObj,
-                                GLuint n, const GLfloat texcoord[][4],
-                                const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   ASSERT(lambda != NULL);
-   for (i = 0; i < n; i++) {
-      GLint level = linear_mipmap_level(tObj, lambda[i]);
-      if (level >= tObj->_MaxLevel) {
-         sample_1d_nearest(ctx, tObj, tObj->Image[0][tObj->_MaxLevel],
-                           texcoord[i], rgba[i]);
-      }
-      else {
-         GLfloat t0[4], t1[4];
-         const GLfloat f = FRAC(lambda[i]);
-         sample_1d_nearest(ctx, tObj, tObj->Image[0][level  ], texcoord[i], t0);
-         sample_1d_nearest(ctx, tObj, tObj->Image[0][level+1], texcoord[i], t1);
-         lerp_rgba(rgba[i], f, t0, t1);
-      }
-   }
-}
-
-
-static void
-sample_1d_linear_mipmap_linear(struct gl_context *ctx,
-                               const struct gl_texture_object *tObj,
-                               GLuint n, const GLfloat texcoord[][4],
-                               const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   ASSERT(lambda != NULL);
-   for (i = 0; i < n; i++) {
-      GLint level = linear_mipmap_level(tObj, lambda[i]);
-      if (level >= tObj->_MaxLevel) {
-         sample_1d_linear(ctx, tObj, tObj->Image[0][tObj->_MaxLevel],
-                          texcoord[i], rgba[i]);
-      }
-      else {
-         GLfloat t0[4], t1[4];
-         const GLfloat f = FRAC(lambda[i]);
-         sample_1d_linear(ctx, tObj, tObj->Image[0][level  ], texcoord[i], t0);
-         sample_1d_linear(ctx, tObj, tObj->Image[0][level+1], texcoord[i], t1);
-         lerp_rgba(rgba[i], f, t0, t1);
-      }
-   }
-}
-
-
-/** Sample 1D texture, nearest filtering for both min/magnification */
-static void
-sample_nearest_1d( struct gl_context *ctx,
-                   const struct gl_texture_object *tObj, GLuint n,
-                   const GLfloat texcoords[][4], const GLfloat lambda[],
-                   GLfloat rgba[][4] )
-{
-   GLuint i;
-   struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
-   (void) lambda;
-   for (i = 0; i < n; i++) {
-      sample_1d_nearest(ctx, tObj, image, texcoords[i], rgba[i]);
-   }
-}
-
-
-/** Sample 1D texture, linear filtering for both min/magnification */
-static void
-sample_linear_1d( struct gl_context *ctx,
-                  const struct gl_texture_object *tObj, GLuint n,
-                  const GLfloat texcoords[][4], const GLfloat lambda[],
-                  GLfloat rgba[][4] )
-{
-   GLuint i;
-   struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
-   (void) lambda;
-   for (i = 0; i < n; i++) {
-      sample_1d_linear(ctx, tObj, image, texcoords[i], rgba[i]);
-   }
-}
-
-
-/** Sample 1D texture, using lambda to choose between min/magnification */
-static void
-sample_lambda_1d( struct gl_context *ctx,
-                  const struct gl_texture_object *tObj, GLuint n,
-                  const GLfloat texcoords[][4],
-                  const GLfloat lambda[], GLfloat rgba[][4] )
-{
-   GLuint minStart, minEnd;  /* texels with minification */
-   GLuint magStart, magEnd;  /* texels with magnification */
-   GLuint i;
-
-   ASSERT(lambda != NULL);
-   compute_min_mag_ranges(tObj, n, lambda,
-                          &minStart, &minEnd, &magStart, &magEnd);
-
-   if (minStart < minEnd) {
-      /* do the minified texels */
-      const GLuint m = minEnd - minStart;
-      switch (tObj->Sampler.MinFilter) {
-      case GL_NEAREST:
-         for (i = minStart; i < minEnd; i++)
-            sample_1d_nearest(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
-                              texcoords[i], rgba[i]);
-         break;
-      case GL_LINEAR:
-         for (i = minStart; i < minEnd; i++)
-            sample_1d_linear(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
-                             texcoords[i], rgba[i]);
-         break;
-      case GL_NEAREST_MIPMAP_NEAREST:
-         sample_1d_nearest_mipmap_nearest(ctx, tObj, m, texcoords + minStart,
-                                          lambda + minStart, rgba + minStart);
-         break;
-      case GL_LINEAR_MIPMAP_NEAREST:
-         sample_1d_linear_mipmap_nearest(ctx, tObj, m, texcoords + minStart,
-                                         lambda + minStart, rgba + minStart);
-         break;
-      case GL_NEAREST_MIPMAP_LINEAR:
-         sample_1d_nearest_mipmap_linear(ctx, tObj, m, texcoords + minStart,
-                                         lambda + minStart, rgba + minStart);
-         break;
-      case GL_LINEAR_MIPMAP_LINEAR:
-         sample_1d_linear_mipmap_linear(ctx, tObj, m, texcoords + minStart,
-                                        lambda + minStart, rgba + minStart);
-         break;
-      default:
-         _mesa_problem(ctx, "Bad min filter in sample_1d_texture");
-         return;
-      }
-   }
-
-   if (magStart < magEnd) {
-      /* do the magnified texels */
-      switch (tObj->Sampler.MagFilter) {
-      case GL_NEAREST:
-         for (i = magStart; i < magEnd; i++)
-            sample_1d_nearest(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
-                              texcoords[i], rgba[i]);
-         break;
-      case GL_LINEAR:
-         for (i = magStart; i < magEnd; i++)
-            sample_1d_linear(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
-                             texcoords[i], rgba[i]);
-         break;
-      default:
-         _mesa_problem(ctx, "Bad mag filter in sample_1d_texture");
-         return;
-      }
-   }
-}
-
-
-/**********************************************************************/
-/*                    2-D Texture Sampling Functions                  */
-/**********************************************************************/
-
-
-/**
- * Return the texture sample for coordinate (s,t) using GL_NEAREST filter.
- */
-static INLINE void
-sample_2d_nearest(struct gl_context *ctx,
-                  const struct gl_texture_object *tObj,
-                  const struct gl_texture_image *img,
-                  const GLfloat texcoord[4],
-                  GLfloat rgba[])
-{
-   const GLint width = img->Width2;    /* without border, power of two */
-   const GLint height = img->Height2;  /* without border, power of two */
-   GLint i, j;
-   (void) ctx;
-
-   i = nearest_texel_location(tObj->Sampler.WrapS, img, width, texcoord[0]);
-   j = nearest_texel_location(tObj->Sampler.WrapT, img, height, texcoord[1]);
-
-   /* skip over the border, if any */
-   i += img->Border;
-   j += img->Border;
-
-   if (i < 0 || i >= (GLint) img->Width || j < 0 || j >= (GLint) img->Height) {
-      /* Need this test for GL_CLAMP_TO_BORDER mode */
-      get_border_color(tObj, img, rgba);
-   }
-   else {
-      img->FetchTexelf(img, i, j, 0, rgba);
-   }
-}
-
-
-/**
- * Return the texture sample for coordinate (s,t) using GL_LINEAR filter.
- * New sampling code contributed by Lynn Quam <quam@ai.sri.com>.
- */
-static INLINE void
-sample_2d_linear(struct gl_context *ctx,
-                 const struct gl_texture_object *tObj,
-                 const struct gl_texture_image *img,
-                 const GLfloat texcoord[4],
-                 GLfloat rgba[])
-{
-   const GLint width = img->Width2;
-   const GLint height = img->Height2;
-   GLint i0, j0, i1, j1;
-   GLbitfield useBorderColor = 0x0;
-   GLfloat a, b;
-   GLfloat t00[4], t10[4], t01[4], t11[4]; /* sampled texel colors */
-
-   linear_texel_locations(tObj->Sampler.WrapS, img, width, texcoord[0],  &i0, &i1, &a);
-   linear_texel_locations(tObj->Sampler.WrapT, img, height, texcoord[1], &j0, &j1, &b);
-
-   if (img->Border) {
-      i0 += img->Border;
-      i1 += img->Border;
-      j0 += img->Border;
-      j1 += img->Border;
-   }
-   else {
-      if (i0 < 0 || i0 >= width)   useBorderColor |= I0BIT;
-      if (i1 < 0 || i1 >= width)   useBorderColor |= I1BIT;
-      if (j0 < 0 || j0 >= height)  useBorderColor |= J0BIT;
-      if (j1 < 0 || j1 >= height)  useBorderColor |= J1BIT;
-   }
-
-   /* fetch four texel colors */
-   if (useBorderColor & (I0BIT | J0BIT)) {
-      get_border_color(tObj, img, t00);
-   }
-   else {
-      img->FetchTexelf(img, i0, j0, 0, t00);
-   }
-   if (useBorderColor & (I1BIT | J0BIT)) {
-      get_border_color(tObj, img, t10);
-   }
-   else {
-      img->FetchTexelf(img, i1, j0, 0, t10);
-   }
-   if (useBorderColor & (I0BIT | J1BIT)) {
-      get_border_color(tObj, img, t01);
-   }
-   else {
-      img->FetchTexelf(img, i0, j1, 0, t01);
-   }
-   if (useBorderColor & (I1BIT | J1BIT)) {
-      get_border_color(tObj, img, t11);
-   }
-   else {
-      img->FetchTexelf(img, i1, j1, 0, t11);
-   }
-
-   lerp_rgba_2d(rgba, a, b, t00, t10, t01, t11);
-}
-
-
-/**
- * As above, but we know WRAP_S == REPEAT and WRAP_T == REPEAT.
- * We don't have to worry about the texture border.
- */
-static INLINE void
-sample_2d_linear_repeat(struct gl_context *ctx,
-                        const struct gl_texture_object *tObj,
-                        const struct gl_texture_image *img,
-                        const GLfloat texcoord[4],
-                        GLfloat rgba[])
-{
-   const GLint width = img->Width2;
-   const GLint height = img->Height2;
-   GLint i0, j0, i1, j1;
-   GLfloat wi, wj;
-   GLfloat t00[4], t10[4], t01[4], t11[4]; /* sampled texel colors */
-
-   (void) ctx;
-
-   ASSERT(tObj->Sampler.WrapS == GL_REPEAT);
-   ASSERT(tObj->Sampler.WrapT == GL_REPEAT);
-   ASSERT(img->Border == 0);
-   ASSERT(img->_BaseFormat != GL_COLOR_INDEX);
-   ASSERT(img->_IsPowerOfTwo);
-
-   linear_repeat_texel_location(width,  texcoord[0], &i0, &i1, &wi);
-   linear_repeat_texel_location(height, texcoord[1], &j0, &j1, &wj);
-
-   img->FetchTexelf(img, i0, j0, 0, t00);
-   img->FetchTexelf(img, i1, j0, 0, t10);
-   img->FetchTexelf(img, i0, j1, 0, t01);
-   img->FetchTexelf(img, i1, j1, 0, t11);
-
-   lerp_rgba_2d(rgba, wi, wj, t00, t10, t01, t11);
-}
-
-
-static void
-sample_2d_nearest_mipmap_nearest(struct gl_context *ctx,
-                                 const struct gl_texture_object *tObj,
-                                 GLuint n, const GLfloat texcoord[][4],
-                                 const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   for (i = 0; i < n; i++) {
-      GLint level = nearest_mipmap_level(tObj, lambda[i]);
-      sample_2d_nearest(ctx, tObj, tObj->Image[0][level], texcoord[i], rgba[i]);
-   }
-}
-
-
-static void
-sample_2d_linear_mipmap_nearest(struct gl_context *ctx,
-                                const struct gl_texture_object *tObj,
-                                GLuint n, const GLfloat texcoord[][4],
-                                const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   ASSERT(lambda != NULL);
-   for (i = 0; i < n; i++) {
-      GLint level = nearest_mipmap_level(tObj, lambda[i]);
-      sample_2d_linear(ctx, tObj, tObj->Image[0][level], texcoord[i], rgba[i]);
-   }
-}
-
-
-static void
-sample_2d_nearest_mipmap_linear(struct gl_context *ctx,
-                                const struct gl_texture_object *tObj,
-                                GLuint n, const GLfloat texcoord[][4],
-                                const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   ASSERT(lambda != NULL);
-   for (i = 0; i < n; i++) {
-      GLint level = linear_mipmap_level(tObj, lambda[i]);
-      if (level >= tObj->_MaxLevel) {
-         sample_2d_nearest(ctx, tObj, tObj->Image[0][tObj->_MaxLevel],
-                           texcoord[i], rgba[i]);
-      }
-      else {
-         GLfloat t0[4], t1[4];  /* texels */
-         const GLfloat f = FRAC(lambda[i]);
-         sample_2d_nearest(ctx, tObj, tObj->Image[0][level  ], texcoord[i], t0);
-         sample_2d_nearest(ctx, tObj, tObj->Image[0][level+1], texcoord[i], t1);
-         lerp_rgba(rgba[i], f, t0, t1);
-      }
-   }
-}
-
-
-static void
-sample_2d_linear_mipmap_linear( struct gl_context *ctx,
-                                const struct gl_texture_object *tObj,
-                                GLuint n, const GLfloat texcoord[][4],
-                                const GLfloat lambda[], GLfloat rgba[][4] )
-{
-   GLuint i;
-   ASSERT(lambda != NULL);
-   for (i = 0; i < n; i++) {
-      GLint level = linear_mipmap_level(tObj, lambda[i]);
-      if (level >= tObj->_MaxLevel) {
-         sample_2d_linear(ctx, tObj, tObj->Image[0][tObj->_MaxLevel],
-                          texcoord[i], rgba[i]);
-      }
-      else {
-         GLfloat t0[4], t1[4];  /* texels */
-         const GLfloat f = FRAC(lambda[i]);
-         sample_2d_linear(ctx, tObj, tObj->Image[0][level  ], texcoord[i], t0);
-         sample_2d_linear(ctx, tObj, tObj->Image[0][level+1], texcoord[i], t1);
-         lerp_rgba(rgba[i], f, t0, t1);
-      }
-   }
-}
-
-
-static void
-sample_2d_linear_mipmap_linear_repeat(struct gl_context *ctx,
-                                      const struct gl_texture_object *tObj,
-                                      GLuint n, const GLfloat texcoord[][4],
-                                      const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   ASSERT(lambda != NULL);
-   ASSERT(tObj->Sampler.WrapS == GL_REPEAT);
-   ASSERT(tObj->Sampler.WrapT == GL_REPEAT);
-   for (i = 0; i < n; i++) {
-      GLint level = linear_mipmap_level(tObj, lambda[i]);
-      if (level >= tObj->_MaxLevel) {
-         sample_2d_linear_repeat(ctx, tObj, tObj->Image[0][tObj->_MaxLevel],
-                                 texcoord[i], rgba[i]);
-      }
-      else {
-         GLfloat t0[4], t1[4];  /* texels */
-         const GLfloat f = FRAC(lambda[i]);
-         sample_2d_linear_repeat(ctx, tObj, tObj->Image[0][level  ],
-                                 texcoord[i], t0);
-         sample_2d_linear_repeat(ctx, tObj, tObj->Image[0][level+1],
-                                 texcoord[i], t1);
-         lerp_rgba(rgba[i], f, t0, t1);
-      }
-   }
-}
-
-
-/** Sample 2D texture, nearest filtering for both min/magnification */
-static void
-sample_nearest_2d(struct gl_context *ctx,
-                  const struct gl_texture_object *tObj, GLuint n,
-                  const GLfloat texcoords[][4],
-                  const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
-   (void) lambda;
-   for (i = 0; i < n; i++) {
-      sample_2d_nearest(ctx, tObj, image, texcoords[i], rgba[i]);
-   }
-}
-
-
-/** Sample 2D texture, linear filtering for both min/magnification */
-static void
-sample_linear_2d(struct gl_context *ctx,
-                 const struct gl_texture_object *tObj, GLuint n,
-                 const GLfloat texcoords[][4],
-                 const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
-   (void) lambda;
-   if (tObj->Sampler.WrapS == GL_REPEAT &&
-       tObj->Sampler.WrapT == GL_REPEAT &&
-       image->_IsPowerOfTwo &&
-       image->Border == 0) {
-      for (i = 0; i < n; i++) {
-         sample_2d_linear_repeat(ctx, tObj, image, texcoords[i], rgba[i]);
-      }
-   }
-   else {
-      for (i = 0; i < n; i++) {
-         sample_2d_linear(ctx, tObj, image, texcoords[i], rgba[i]);
-      }
-   }
-}
-
-
-/**
- * Optimized 2-D texture sampling:
- *    S and T wrap mode == GL_REPEAT
- *    GL_NEAREST min/mag filter
- *    No border, 
- *    RowStride == Width,
- *    Format = GL_RGB
- */
-static void
-opt_sample_rgb_2d(struct gl_context *ctx,
-                  const struct gl_texture_object *tObj,
-                  GLuint n, const GLfloat texcoords[][4],
-                  const GLfloat lambda[], GLfloat rgba[][4])
-{
-   const struct gl_texture_image *img = tObj->Image[0][tObj->BaseLevel];
-   const GLfloat width = (GLfloat) img->Width;
-   const GLfloat height = (GLfloat) img->Height;
-   const GLint colMask = img->Width - 1;
-   const GLint rowMask = img->Height - 1;
-   const GLint shift = img->WidthLog2;
-   GLuint k;
-   (void) ctx;
-   (void) lambda;
-   ASSERT(tObj->Sampler.WrapS==GL_REPEAT);
-   ASSERT(tObj->Sampler.WrapT==GL_REPEAT);
-   ASSERT(img->Border==0);
-   ASSERT(img->TexFormat == MESA_FORMAT_RGB888);
-   ASSERT(img->_IsPowerOfTwo);
-
-   for (k=0; k<n; k++) {
-      GLint i = IFLOOR(texcoords[k][0] * width) & colMask;
-      GLint j = IFLOOR(texcoords[k][1] * height) & rowMask;
-      GLint pos = (j << shift) | i;
-      GLubyte *texel = ((GLubyte *) img->Data) + 3*pos;
-      rgba[k][RCOMP] = UBYTE_TO_FLOAT(texel[2]);
-      rgba[k][GCOMP] = UBYTE_TO_FLOAT(texel[1]);
-      rgba[k][BCOMP] = UBYTE_TO_FLOAT(texel[0]);
-      rgba[k][ACOMP] = 1.0F;
-   }
-}
-
-
-/**
- * Optimized 2-D texture sampling:
- *    S and T wrap mode == GL_REPEAT
- *    GL_NEAREST min/mag filter
- *    No border
- *    RowStride == Width,
- *    Format = GL_RGBA
- */
-static void
-opt_sample_rgba_2d(struct gl_context *ctx,
-                   const struct gl_texture_object *tObj,
-                   GLuint n, const GLfloat texcoords[][4],
-                   const GLfloat lambda[], GLfloat rgba[][4])
-{
-   const struct gl_texture_image *img = tObj->Image[0][tObj->BaseLevel];
-   const GLfloat width = (GLfloat) img->Width;
-   const GLfloat height = (GLfloat) img->Height;
-   const GLint colMask = img->Width - 1;
-   const GLint rowMask = img->Height - 1;
-   const GLint shift = img->WidthLog2;
-   GLuint i;
-   (void) ctx;
-   (void) lambda;
-   ASSERT(tObj->Sampler.WrapS==GL_REPEAT);
-   ASSERT(tObj->Sampler.WrapT==GL_REPEAT);
-   ASSERT(img->Border==0);
-   ASSERT(img->TexFormat == MESA_FORMAT_RGBA8888);
-   ASSERT(img->_IsPowerOfTwo);
-
-   for (i = 0; i < n; i++) {
-      const GLint col = IFLOOR(texcoords[i][0] * width) & colMask;
-      const GLint row = IFLOOR(texcoords[i][1] * height) & rowMask;
-      const GLint pos = (row << shift) | col;
-      const GLuint texel = *((GLuint *) img->Data + pos);
-      rgba[i][RCOMP] = UBYTE_TO_FLOAT( (texel >> 24)        );
-      rgba[i][GCOMP] = UBYTE_TO_FLOAT( (texel >> 16) & 0xff );
-      rgba[i][BCOMP] = UBYTE_TO_FLOAT( (texel >>  8) & 0xff );
-      rgba[i][ACOMP] = UBYTE_TO_FLOAT( (texel      ) & 0xff );
-   }
-}
-
-
-/** Sample 2D texture, using lambda to choose between min/magnification */
-static void
-sample_lambda_2d(struct gl_context *ctx,
-                 const struct gl_texture_object *tObj,
-                 GLuint n, const GLfloat texcoords[][4],
-                 const GLfloat lambda[], GLfloat rgba[][4])
-{
-   const struct gl_texture_image *tImg = tObj->Image[0][tObj->BaseLevel];
-   GLuint minStart, minEnd;  /* texels with minification */
-   GLuint magStart, magEnd;  /* texels with magnification */
-
-   const GLboolean repeatNoBorderPOT = (tObj->Sampler.WrapS == GL_REPEAT)
-      && (tObj->Sampler.WrapT == GL_REPEAT)
-      && (tImg->Border == 0 && (tImg->Width == tImg->RowStride))
-      && (tImg->_BaseFormat != GL_COLOR_INDEX)
-      && tImg->_IsPowerOfTwo;
-
-   ASSERT(lambda != NULL);
-   compute_min_mag_ranges(tObj, n, lambda,
-                          &minStart, &minEnd, &magStart, &magEnd);
-
-   if (minStart < minEnd) {
-      /* do the minified texels */
-      const GLuint m = minEnd - minStart;
-      switch (tObj->Sampler.MinFilter) {
-      case GL_NEAREST:
-         if (repeatNoBorderPOT) {
-            switch (tImg->TexFormat) {
-            case MESA_FORMAT_RGB888:
-               opt_sample_rgb_2d(ctx, tObj, m, texcoords + minStart,
-                                 NULL, rgba + minStart);
-               break;
-            case MESA_FORMAT_RGBA8888:
-	       opt_sample_rgba_2d(ctx, tObj, m, texcoords + minStart,
-                                  NULL, rgba + minStart);
-               break;
-            default:
-               sample_nearest_2d(ctx, tObj, m, texcoords + minStart,
-                                 NULL, rgba + minStart );
-            }
-         }
-         else {
-            sample_nearest_2d(ctx, tObj, m, texcoords + minStart,
-                              NULL, rgba + minStart);
-         }
-         break;
-      case GL_LINEAR:
-	 sample_linear_2d(ctx, tObj, m, texcoords + minStart,
-			  NULL, rgba + minStart);
-         break;
-      case GL_NEAREST_MIPMAP_NEAREST:
-         sample_2d_nearest_mipmap_nearest(ctx, tObj, m,
-                                          texcoords + minStart,
-                                          lambda + minStart, rgba + minStart);
-         break;
-      case GL_LINEAR_MIPMAP_NEAREST:
-         sample_2d_linear_mipmap_nearest(ctx, tObj, m, texcoords + minStart,
-                                         lambda + minStart, rgba + minStart);
-         break;
-      case GL_NEAREST_MIPMAP_LINEAR:
-         sample_2d_nearest_mipmap_linear(ctx, tObj, m, texcoords + minStart,
-                                         lambda + minStart, rgba + minStart);
-         break;
-      case GL_LINEAR_MIPMAP_LINEAR:
-         if (repeatNoBorderPOT)
-            sample_2d_linear_mipmap_linear_repeat(ctx, tObj, m,
-                  texcoords + minStart, lambda + minStart, rgba + minStart);
-         else
-            sample_2d_linear_mipmap_linear(ctx, tObj, m, texcoords + minStart,
-                                        lambda + minStart, rgba + minStart);
-         break;
-      default:
-         _mesa_problem(ctx, "Bad min filter in sample_2d_texture");
-         return;
-      }
-   }
-
-   if (magStart < magEnd) {
-      /* do the magnified texels */
-      const GLuint m = magEnd - magStart;
-
-      switch (tObj->Sampler.MagFilter) {
-      case GL_NEAREST:
-         if (repeatNoBorderPOT) {
-            switch (tImg->TexFormat) {
-            case MESA_FORMAT_RGB888:
-               opt_sample_rgb_2d(ctx, tObj, m, texcoords + magStart,
-                                 NULL, rgba + magStart);
-               break;
-            case MESA_FORMAT_RGBA8888:
-	       opt_sample_rgba_2d(ctx, tObj, m, texcoords + magStart,
-                                  NULL, rgba + magStart);
-               break;
-            default:
-               sample_nearest_2d(ctx, tObj, m, texcoords + magStart,
-                                 NULL, rgba + magStart );
-            }
-         }
-         else {
-            sample_nearest_2d(ctx, tObj, m, texcoords + magStart,
-                              NULL, rgba + magStart);
-         }
-         break;
-      case GL_LINEAR:
-	 sample_linear_2d(ctx, tObj, m, texcoords + magStart,
-			  NULL, rgba + magStart);
-         break;
-      default:
-         _mesa_problem(ctx, "Bad mag filter in sample_lambda_2d");
-      }
-   }
-}
-
-
-
-/**********************************************************************/
-/*                    3-D Texture Sampling Functions                  */
-/**********************************************************************/
-
-/**
- * Return the texture sample for coordinate (s,t,r) using GL_NEAREST filter.
- */
-static INLINE void
-sample_3d_nearest(struct gl_context *ctx,
-                  const struct gl_texture_object *tObj,
-                  const struct gl_texture_image *img,
-                  const GLfloat texcoord[4],
-                  GLfloat rgba[4])
-{
-   const GLint width = img->Width2;     /* without border, power of two */
-   const GLint height = img->Height2;   /* without border, power of two */
-   const GLint depth = img->Depth2;     /* without border, power of two */
-   GLint i, j, k;
-   (void) ctx;
-
-   i = nearest_texel_location(tObj->Sampler.WrapS, img, width, texcoord[0]);
-   j = nearest_texel_location(tObj->Sampler.WrapT, img, height, texcoord[1]);
-   k = nearest_texel_location(tObj->Sampler.WrapR, img, depth, texcoord[2]);
-
-   if (i < 0 || i >= (GLint) img->Width ||
-       j < 0 || j >= (GLint) img->Height ||
-       k < 0 || k >= (GLint) img->Depth) {
-      /* Need this test for GL_CLAMP_TO_BORDER mode */
-      get_border_color(tObj, img, rgba);
-   }
-   else {
-      img->FetchTexelf(img, i, j, k, rgba);
-   }
-}
-
-
-/**
- * Return the texture sample for coordinate (s,t,r) using GL_LINEAR filter.
- */
-static void
-sample_3d_linear(struct gl_context *ctx,
-                 const struct gl_texture_object *tObj,
-                 const struct gl_texture_image *img,
-                 const GLfloat texcoord[4],
-                 GLfloat rgba[4])
-{
-   const GLint width = img->Width2;
-   const GLint height = img->Height2;
-   const GLint depth = img->Depth2;
-   GLint i0, j0, k0, i1, j1, k1;
-   GLbitfield useBorderColor = 0x0;
-   GLfloat a, b, c;
-   GLfloat t000[4], t010[4], t001[4], t011[4];
-   GLfloat t100[4], t110[4], t101[4], t111[4];
-
-   linear_texel_locations(tObj->Sampler.WrapS, img, width, texcoord[0],  &i0, &i1, &a);
-   linear_texel_locations(tObj->Sampler.WrapT, img, height, texcoord[1], &j0, &j1, &b);
-   linear_texel_locations(tObj->Sampler.WrapR, img, depth, texcoord[2],  &k0, &k1, &c);
-
-   if (img->Border) {
-      i0 += img->Border;
-      i1 += img->Border;
-      j0 += img->Border;
-      j1 += img->Border;
-      k0 += img->Border;
-      k1 += img->Border;
-   }
-   else {
-      /* check if sampling texture border color */
-      if (i0 < 0 || i0 >= width)   useBorderColor |= I0BIT;
-      if (i1 < 0 || i1 >= width)   useBorderColor |= I1BIT;
-      if (j0 < 0 || j0 >= height)  useBorderColor |= J0BIT;
-      if (j1 < 0 || j1 >= height)  useBorderColor |= J1BIT;
-      if (k0 < 0 || k0 >= depth)   useBorderColor |= K0BIT;
-      if (k1 < 0 || k1 >= depth)   useBorderColor |= K1BIT;
-   }
-
-   /* Fetch texels */
-   if (useBorderColor & (I0BIT | J0BIT | K0BIT)) {
-      get_border_color(tObj, img, t000);
-   }
-   else {
-      img->FetchTexelf(img, i0, j0, k0, t000);
-   }
-   if (useBorderColor & (I1BIT | J0BIT | K0BIT)) {
-      get_border_color(tObj, img, t100);
-   }
-   else {
-      img->FetchTexelf(img, i1, j0, k0, t100);
-   }
-   if (useBorderColor & (I0BIT | J1BIT | K0BIT)) {
-      get_border_color(tObj, img, t010);
-   }
-   else {
-      img->FetchTexelf(img, i0, j1, k0, t010);
-   }
-   if (useBorderColor & (I1BIT | J1BIT | K0BIT)) {
-      get_border_color(tObj, img, t110);
-   }
-   else {
-      img->FetchTexelf(img, i1, j1, k0, t110);
-   }
-
-   if (useBorderColor & (I0BIT | J0BIT | K1BIT)) {
-      get_border_color(tObj, img, t001);
-   }
-   else {
-      img->FetchTexelf(img, i0, j0, k1, t001);
-   }
-   if (useBorderColor & (I1BIT | J0BIT | K1BIT)) {
-      get_border_color(tObj, img, t101);
-   }
-   else {
-      img->FetchTexelf(img, i1, j0, k1, t101);
-   }
-   if (useBorderColor & (I0BIT | J1BIT | K1BIT)) {
-      get_border_color(tObj, img, t011);
-   }
-   else {
-      img->FetchTexelf(img, i0, j1, k1, t011);
-   }
-   if (useBorderColor & (I1BIT | J1BIT | K1BIT)) {
-      get_border_color(tObj, img, t111);
-   }
-   else {
-      img->FetchTexelf(img, i1, j1, k1, t111);
-   }
-
-   /* trilinear interpolation of samples */
-   lerp_rgba_3d(rgba, a, b, c, t000, t100, t010, t110, t001, t101, t011, t111);
-}
-
-
-static void
-sample_3d_nearest_mipmap_nearest(struct gl_context *ctx,
-                                 const struct gl_texture_object *tObj,
-                                 GLuint n, const GLfloat texcoord[][4],
-                                 const GLfloat lambda[], GLfloat rgba[][4] )
-{
-   GLuint i;
-   for (i = 0; i < n; i++) {
-      GLint level = nearest_mipmap_level(tObj, lambda[i]);
-      sample_3d_nearest(ctx, tObj, tObj->Image[0][level], texcoord[i], rgba[i]);
-   }
-}
-
-
-static void
-sample_3d_linear_mipmap_nearest(struct gl_context *ctx,
-                                const struct gl_texture_object *tObj,
-                                GLuint n, const GLfloat texcoord[][4],
-                                const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   ASSERT(lambda != NULL);
-   for (i = 0; i < n; i++) {
-      GLint level = nearest_mipmap_level(tObj, lambda[i]);
-      sample_3d_linear(ctx, tObj, tObj->Image[0][level], texcoord[i], rgba[i]);
-   }
-}
-
-
-static void
-sample_3d_nearest_mipmap_linear(struct gl_context *ctx,
-                                const struct gl_texture_object *tObj,
-                                GLuint n, const GLfloat texcoord[][4],
-                                const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   ASSERT(lambda != NULL);
-   for (i = 0; i < n; i++) {
-      GLint level = linear_mipmap_level(tObj, lambda[i]);
-      if (level >= tObj->_MaxLevel) {
-         sample_3d_nearest(ctx, tObj, tObj->Image[0][tObj->_MaxLevel],
-                           texcoord[i], rgba[i]);
-      }
-      else {
-         GLfloat t0[4], t1[4];  /* texels */
-         const GLfloat f = FRAC(lambda[i]);
-         sample_3d_nearest(ctx, tObj, tObj->Image[0][level  ], texcoord[i], t0);
-         sample_3d_nearest(ctx, tObj, tObj->Image[0][level+1], texcoord[i], t1);
-         lerp_rgba(rgba[i], f, t0, t1);
-      }
-   }
-}
-
-
-static void
-sample_3d_linear_mipmap_linear(struct gl_context *ctx,
-                               const struct gl_texture_object *tObj,
-                               GLuint n, const GLfloat texcoord[][4],
-                               const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   ASSERT(lambda != NULL);
-   for (i = 0; i < n; i++) {
-      GLint level = linear_mipmap_level(tObj, lambda[i]);
-      if (level >= tObj->_MaxLevel) {
-         sample_3d_linear(ctx, tObj, tObj->Image[0][tObj->_MaxLevel],
-                          texcoord[i], rgba[i]);
-      }
-      else {
-         GLfloat t0[4], t1[4];  /* texels */
-         const GLfloat f = FRAC(lambda[i]);
-         sample_3d_linear(ctx, tObj, tObj->Image[0][level  ], texcoord[i], t0);
-         sample_3d_linear(ctx, tObj, tObj->Image[0][level+1], texcoord[i], t1);
-         lerp_rgba(rgba[i], f, t0, t1);
-      }
-   }
-}
-
-
-/** Sample 3D texture, nearest filtering for both min/magnification */
-static void
-sample_nearest_3d(struct gl_context *ctx,
-                  const struct gl_texture_object *tObj, GLuint n,
-                  const GLfloat texcoords[][4], const GLfloat lambda[],
-                  GLfloat rgba[][4])
-{
-   GLuint i;
-   struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
-   (void) lambda;
-   for (i = 0; i < n; i++) {
-      sample_3d_nearest(ctx, tObj, image, texcoords[i], rgba[i]);
-   }
-}
-
-
-/** Sample 3D texture, linear filtering for both min/magnification */
-static void
-sample_linear_3d(struct gl_context *ctx,
-                 const struct gl_texture_object *tObj, GLuint n,
-                 const GLfloat texcoords[][4],
-		 const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
-   (void) lambda;
-   for (i = 0; i < n; i++) {
-      sample_3d_linear(ctx, tObj, image, texcoords[i], rgba[i]);
-   }
-}
-
-
-/** Sample 3D texture, using lambda to choose between min/magnification */
-static void
-sample_lambda_3d(struct gl_context *ctx,
-                 const struct gl_texture_object *tObj, GLuint n,
-                 const GLfloat texcoords[][4], const GLfloat lambda[],
-                 GLfloat rgba[][4])
-{
-   GLuint minStart, minEnd;  /* texels with minification */
-   GLuint magStart, magEnd;  /* texels with magnification */
-   GLuint i;
-
-   ASSERT(lambda != NULL);
-   compute_min_mag_ranges(tObj, n, lambda,
-                          &minStart, &minEnd, &magStart, &magEnd);
-
-   if (minStart < minEnd) {
-      /* do the minified texels */
-      GLuint m = minEnd - minStart;
-      switch (tObj->Sampler.MinFilter) {
-      case GL_NEAREST:
-         for (i = minStart; i < minEnd; i++)
-            sample_3d_nearest(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
-                              texcoords[i], rgba[i]);
-         break;
-      case GL_LINEAR:
-         for (i = minStart; i < minEnd; i++)
-            sample_3d_linear(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
-                             texcoords[i], rgba[i]);
-         break;
-      case GL_NEAREST_MIPMAP_NEAREST:
-         sample_3d_nearest_mipmap_nearest(ctx, tObj, m, texcoords + minStart,
-                                          lambda + minStart, rgba + minStart);
-         break;
-      case GL_LINEAR_MIPMAP_NEAREST:
-         sample_3d_linear_mipmap_nearest(ctx, tObj, m, texcoords + minStart,
-                                         lambda + minStart, rgba + minStart);
-         break;
-      case GL_NEAREST_MIPMAP_LINEAR:
-         sample_3d_nearest_mipmap_linear(ctx, tObj, m, texcoords + minStart,
-                                         lambda + minStart, rgba + minStart);
-         break;
-      case GL_LINEAR_MIPMAP_LINEAR:
-         sample_3d_linear_mipmap_linear(ctx, tObj, m, texcoords + minStart,
-                                        lambda + minStart, rgba + minStart);
-         break;
-      default:
-         _mesa_problem(ctx, "Bad min filter in sample_3d_texture");
-         return;
-      }
-   }
-
-   if (magStart < magEnd) {
-      /* do the magnified texels */
-      switch (tObj->Sampler.MagFilter) {
-      case GL_NEAREST:
-         for (i = magStart; i < magEnd; i++)
-            sample_3d_nearest(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
-                              texcoords[i], rgba[i]);
-         break;
-      case GL_LINEAR:
-         for (i = magStart; i < magEnd; i++)
-            sample_3d_linear(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
-                             texcoords[i], rgba[i]);
-         break;
-      default:
-         _mesa_problem(ctx, "Bad mag filter in sample_3d_texture");
-         return;
-      }
-   }
-}
-
-
-/**********************************************************************/
-/*                Texture Cube Map Sampling Functions                 */
-/**********************************************************************/
-
-/**
- * Choose one of six sides of a texture cube map given the texture
- * coord (rx,ry,rz).  Return pointer to corresponding array of texture
- * images.
- */
-static const struct gl_texture_image **
-choose_cube_face(const struct gl_texture_object *texObj,
-                 const GLfloat texcoord[4], GLfloat newCoord[4])
-{
-   /*
-      major axis
-      direction     target                             sc     tc    ma
-      ----------    -------------------------------    ---    ---   ---
-       +rx          TEXTURE_CUBE_MAP_POSITIVE_X_EXT    -rz    -ry   rx
-       -rx          TEXTURE_CUBE_MAP_NEGATIVE_X_EXT    +rz    -ry   rx
-       +ry          TEXTURE_CUBE_MAP_POSITIVE_Y_EXT    +rx    +rz   ry
-       -ry          TEXTURE_CUBE_MAP_NEGATIVE_Y_EXT    +rx    -rz   ry
-       +rz          TEXTURE_CUBE_MAP_POSITIVE_Z_EXT    +rx    -ry   rz
-       -rz          TEXTURE_CUBE_MAP_NEGATIVE_Z_EXT    -rx    -ry   rz
-   */
-   const GLfloat rx = texcoord[0];
-   const GLfloat ry = texcoord[1];
-   const GLfloat rz = texcoord[2];
-   const GLfloat arx = FABSF(rx), ary = FABSF(ry), arz = FABSF(rz);
-   GLuint face;
-   GLfloat sc, tc, ma;
-
-   if (arx >= ary && arx >= arz) {
-      if (rx >= 0.0F) {
-         face = FACE_POS_X;
-         sc = -rz;
-         tc = -ry;
-         ma = arx;
-      }
-      else {
-         face = FACE_NEG_X;
-         sc = rz;
-         tc = -ry;
-         ma = arx;
-      }
-   }
-   else if (ary >= arx && ary >= arz) {
-      if (ry >= 0.0F) {
-         face = FACE_POS_Y;
-         sc = rx;
-         tc = rz;
-         ma = ary;
-      }
-      else {
-         face = FACE_NEG_Y;
-         sc = rx;
-         tc = -rz;
-         ma = ary;
-      }
-   }
-   else {
-      if (rz > 0.0F) {
-         face = FACE_POS_Z;
-         sc = rx;
-         tc = -ry;
-         ma = arz;
-      }
-      else {
-         face = FACE_NEG_Z;
-         sc = -rx;
-         tc = -ry;
-         ma = arz;
-      }
-   }
-
-   { 
-      const float ima = 1.0F / ma;
-      newCoord[0] = ( sc * ima + 1.0F ) * 0.5F;
-      newCoord[1] = ( tc * ima + 1.0F ) * 0.5F;
-   }
-
-   return (const struct gl_texture_image **) texObj->Image[face];
-}
-
-
-static void
-sample_nearest_cube(struct gl_context *ctx,
-		    const struct gl_texture_object *tObj, GLuint n,
-                    const GLfloat texcoords[][4], const GLfloat lambda[],
-                    GLfloat rgba[][4])
-{
-   GLuint i;
-   (void) lambda;
-   for (i = 0; i < n; i++) {
-      const struct gl_texture_image **images;
-      GLfloat newCoord[4];
-      images = choose_cube_face(tObj, texcoords[i], newCoord);
-      sample_2d_nearest(ctx, tObj, images[tObj->BaseLevel],
-                        newCoord, rgba[i]);
-   }
-}
-
-
-static void
-sample_linear_cube(struct gl_context *ctx,
-		   const struct gl_texture_object *tObj, GLuint n,
-                   const GLfloat texcoords[][4],
-		   const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   (void) lambda;
-   for (i = 0; i < n; i++) {
-      const struct gl_texture_image **images;
-      GLfloat newCoord[4];
-      images = choose_cube_face(tObj, texcoords[i], newCoord);
-      sample_2d_linear(ctx, tObj, images[tObj->BaseLevel],
-                       newCoord, rgba[i]);
-   }
-}
-
-
-static void
-sample_cube_nearest_mipmap_nearest(struct gl_context *ctx,
-                                   const struct gl_texture_object *tObj,
-                                   GLuint n, const GLfloat texcoord[][4],
-                                   const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   ASSERT(lambda != NULL);
-   for (i = 0; i < n; i++) {
-      const struct gl_texture_image **images;
-      GLfloat newCoord[4];
-      GLint level;
-      images = choose_cube_face(tObj, texcoord[i], newCoord);
-
-      /* XXX we actually need to recompute lambda here based on the newCoords.
-       * But we would need the texcoords of adjacent fragments to compute that
-       * properly, and we don't have those here.
-       * For now, do an approximation:  subtracting 1 from the chosen mipmap
-       * level seems to work in some test cases.
-       * The same adjustment is done in the next few functions.
-      */
-      level = nearest_mipmap_level(tObj, lambda[i]);
-      level = MAX2(level - 1, 0);
-
-      sample_2d_nearest(ctx, tObj, images[level], newCoord, rgba[i]);
-   }
-}
-
-
-static void
-sample_cube_linear_mipmap_nearest(struct gl_context *ctx,
-                                  const struct gl_texture_object *tObj,
-                                  GLuint n, const GLfloat texcoord[][4],
-                                  const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   ASSERT(lambda != NULL);
-   for (i = 0; i < n; i++) {
-      const struct gl_texture_image **images;
-      GLfloat newCoord[4];
-      GLint level = nearest_mipmap_level(tObj, lambda[i]);
-      level = MAX2(level - 1, 0); /* see comment above */
-      images = choose_cube_face(tObj, texcoord[i], newCoord);
-      sample_2d_linear(ctx, tObj, images[level], newCoord, rgba[i]);
-   }
-}
-
-
-static void
-sample_cube_nearest_mipmap_linear(struct gl_context *ctx,
-                                  const struct gl_texture_object *tObj,
-                                  GLuint n, const GLfloat texcoord[][4],
-                                  const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   ASSERT(lambda != NULL);
-   for (i = 0; i < n; i++) {
-      const struct gl_texture_image **images;
-      GLfloat newCoord[4];
-      GLint level = linear_mipmap_level(tObj, lambda[i]);
-      level = MAX2(level - 1, 0); /* see comment above */
-      images = choose_cube_face(tObj, texcoord[i], newCoord);
-      if (level >= tObj->_MaxLevel) {
-         sample_2d_nearest(ctx, tObj, images[tObj->_MaxLevel],
-                           newCoord, rgba[i]);
-      }
-      else {
-         GLfloat t0[4], t1[4];  /* texels */
-         const GLfloat f = FRAC(lambda[i]);
-         sample_2d_nearest(ctx, tObj, images[level  ], newCoord, t0);
-         sample_2d_nearest(ctx, tObj, images[level+1], newCoord, t1);
-         lerp_rgba(rgba[i], f, t0, t1);
-      }
-   }
-}
-
-
-static void
-sample_cube_linear_mipmap_linear(struct gl_context *ctx,
-                                 const struct gl_texture_object *tObj,
-                                 GLuint n, const GLfloat texcoord[][4],
-                                 const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   ASSERT(lambda != NULL);
-   for (i = 0; i < n; i++) {
-      const struct gl_texture_image **images;
-      GLfloat newCoord[4];
-      GLint level = linear_mipmap_level(tObj, lambda[i]);
-      level = MAX2(level - 1, 0); /* see comment above */
-      images = choose_cube_face(tObj, texcoord[i], newCoord);
-      if (level >= tObj->_MaxLevel) {
-         sample_2d_linear(ctx, tObj, images[tObj->_MaxLevel],
-                          newCoord, rgba[i]);
-      }
-      else {
-         GLfloat t0[4], t1[4];
-         const GLfloat f = FRAC(lambda[i]);
-         sample_2d_linear(ctx, tObj, images[level  ], newCoord, t0);
-         sample_2d_linear(ctx, tObj, images[level+1], newCoord, t1);
-         lerp_rgba(rgba[i], f, t0, t1);
-      }
-   }
-}
-
-
-/** Sample cube texture, using lambda to choose between min/magnification */
-static void
-sample_lambda_cube(struct gl_context *ctx,
-		   const struct gl_texture_object *tObj, GLuint n,
-		   const GLfloat texcoords[][4], const GLfloat lambda[],
-		   GLfloat rgba[][4])
-{
-   GLuint minStart, minEnd;  /* texels with minification */
-   GLuint magStart, magEnd;  /* texels with magnification */
-
-   ASSERT(lambda != NULL);
-   compute_min_mag_ranges(tObj, n, lambda,
-                          &minStart, &minEnd, &magStart, &magEnd);
-
-   if (minStart < minEnd) {
-      /* do the minified texels */
-      const GLuint m = minEnd - minStart;
-      switch (tObj->Sampler.MinFilter) {
-      case GL_NEAREST:
-         sample_nearest_cube(ctx, tObj, m, texcoords + minStart,
-                             lambda + minStart, rgba + minStart);
-         break;
-      case GL_LINEAR:
-         sample_linear_cube(ctx, tObj, m, texcoords + minStart,
-                            lambda + minStart, rgba + minStart);
-         break;
-      case GL_NEAREST_MIPMAP_NEAREST:
-         sample_cube_nearest_mipmap_nearest(ctx, tObj, m,
-                                            texcoords + minStart,
-                                           lambda + minStart, rgba + minStart);
-         break;
-      case GL_LINEAR_MIPMAP_NEAREST:
-         sample_cube_linear_mipmap_nearest(ctx, tObj, m,
-                                           texcoords + minStart,
-                                           lambda + minStart, rgba + minStart);
-         break;
-      case GL_NEAREST_MIPMAP_LINEAR:
-         sample_cube_nearest_mipmap_linear(ctx, tObj, m,
-                                           texcoords + minStart,
-                                           lambda + minStart, rgba + minStart);
-         break;
-      case GL_LINEAR_MIPMAP_LINEAR:
-         sample_cube_linear_mipmap_linear(ctx, tObj, m,
-                                          texcoords + minStart,
-                                          lambda + minStart, rgba + minStart);
-         break;
-      default:
-         _mesa_problem(ctx, "Bad min filter in sample_lambda_cube");
-      }
-   }
-
-   if (magStart < magEnd) {
-      /* do the magnified texels */
-      const GLuint m = magEnd - magStart;
-      switch (tObj->Sampler.MagFilter) {
-      case GL_NEAREST:
-         sample_nearest_cube(ctx, tObj, m, texcoords + magStart,
-                             lambda + magStart, rgba + magStart);
-         break;
-      case GL_LINEAR:
-         sample_linear_cube(ctx, tObj, m, texcoords + magStart,
-                            lambda + magStart, rgba + magStart);
-         break;
-      default:
-         _mesa_problem(ctx, "Bad mag filter in sample_lambda_cube");
-      }
-   }
-}
-
-
-/**********************************************************************/
-/*               Texture Rectangle Sampling Functions                 */
-/**********************************************************************/
-
-
-static void
-sample_nearest_rect(struct gl_context *ctx,
-		    const struct gl_texture_object *tObj, GLuint n,
-                    const GLfloat texcoords[][4], const GLfloat lambda[],
-                    GLfloat rgba[][4])
-{
-   const struct gl_texture_image *img = tObj->Image[0][0];
-   const GLint width = img->Width;
-   const GLint height = img->Height;
-   GLuint i;
-
-   (void) ctx;
-   (void) lambda;
-
-   ASSERT(tObj->Sampler.WrapS == GL_CLAMP ||
-          tObj->Sampler.WrapS == GL_CLAMP_TO_EDGE ||
-          tObj->Sampler.WrapS == GL_CLAMP_TO_BORDER);
-   ASSERT(tObj->Sampler.WrapT == GL_CLAMP ||
-          tObj->Sampler.WrapT == GL_CLAMP_TO_EDGE ||
-          tObj->Sampler.WrapT == GL_CLAMP_TO_BORDER);
-   ASSERT(img->_BaseFormat != GL_COLOR_INDEX);
-
-   for (i = 0; i < n; i++) {
-      GLint row, col;
-      col = clamp_rect_coord_nearest(tObj->Sampler.WrapS, texcoords[i][0], width);
-      row = clamp_rect_coord_nearest(tObj->Sampler.WrapT, texcoords[i][1], height);
-      if (col < 0 || col >= width || row < 0 || row >= height)
-         get_border_color(tObj, img, rgba[i]);
-      else
-         img->FetchTexelf(img, col, row, 0, rgba[i]);
-   }
-}
-
-
-static void
-sample_linear_rect(struct gl_context *ctx,
-		   const struct gl_texture_object *tObj, GLuint n,
-                   const GLfloat texcoords[][4],
-		   const GLfloat lambda[], GLfloat rgba[][4])
-{
-   const struct gl_texture_image *img = tObj->Image[0][0];
-   const GLint width = img->Width;
-   const GLint height = img->Height;
-   GLuint i;
-
-   (void) ctx;
-   (void) lambda;
-
-   ASSERT(tObj->Sampler.WrapS == GL_CLAMP ||
-          tObj->Sampler.WrapS == GL_CLAMP_TO_EDGE ||
-          tObj->Sampler.WrapS == GL_CLAMP_TO_BORDER);
-   ASSERT(tObj->Sampler.WrapT == GL_CLAMP ||
-          tObj->Sampler.WrapT == GL_CLAMP_TO_EDGE ||
-          tObj->Sampler.WrapT == GL_CLAMP_TO_BORDER);
-   ASSERT(img->_BaseFormat != GL_COLOR_INDEX);
-
-   for (i = 0; i < n; i++) {
-      GLint i0, j0, i1, j1;
-      GLfloat t00[4], t01[4], t10[4], t11[4];
-      GLfloat a, b;
-      GLbitfield useBorderColor = 0x0;
-
-      clamp_rect_coord_linear(tObj->Sampler.WrapS, texcoords[i][0], width,
-                              &i0, &i1, &a);
-      clamp_rect_coord_linear(tObj->Sampler.WrapT, texcoords[i][1], height,
-                              &j0, &j1, &b);
-
-      /* compute integer rows/columns */
-      if (i0 < 0 || i0 >= width)   useBorderColor |= I0BIT;
-      if (i1 < 0 || i1 >= width)   useBorderColor |= I1BIT;
-      if (j0 < 0 || j0 >= height)  useBorderColor |= J0BIT;
-      if (j1 < 0 || j1 >= height)  useBorderColor |= J1BIT;
-
-      /* get four texel samples */
-      if (useBorderColor & (I0BIT | J0BIT))
-         get_border_color(tObj, img, t00);
-      else
-         img->FetchTexelf(img, i0, j0, 0, t00);
-
-      if (useBorderColor & (I1BIT | J0BIT))
-         get_border_color(tObj, img, t10);
-      else
-         img->FetchTexelf(img, i1, j0, 0, t10);
-
-      if (useBorderColor & (I0BIT | J1BIT))
-         get_border_color(tObj, img, t01);
-      else
-         img->FetchTexelf(img, i0, j1, 0, t01);
-
-      if (useBorderColor & (I1BIT | J1BIT))
-         get_border_color(tObj, img, t11);
-      else
-         img->FetchTexelf(img, i1, j1, 0, t11);
-
-      lerp_rgba_2d(rgba[i], a, b, t00, t10, t01, t11);
-   }
-}
-
-
-/** Sample Rect texture, using lambda to choose between min/magnification */
-static void
-sample_lambda_rect(struct gl_context *ctx,
-		   const struct gl_texture_object *tObj, GLuint n,
-		   const GLfloat texcoords[][4], const GLfloat lambda[],
-		   GLfloat rgba[][4])
-{
-   GLuint minStart, minEnd, magStart, magEnd;
-
-   /* We only need lambda to decide between minification and magnification.
-    * There is no mipmapping with rectangular textures.
-    */
-   compute_min_mag_ranges(tObj, n, lambda,
-                          &minStart, &minEnd, &magStart, &magEnd);
-
-   if (minStart < minEnd) {
-      if (tObj->Sampler.MinFilter == GL_NEAREST) {
-         sample_nearest_rect(ctx, tObj, minEnd - minStart,
-                             texcoords + minStart, NULL, rgba + minStart);
-      }
-      else {
-         sample_linear_rect(ctx, tObj, minEnd - minStart,
-                            texcoords + minStart, NULL, rgba + minStart);
-      }
-   }
-   if (magStart < magEnd) {
-      if (tObj->Sampler.MagFilter == GL_NEAREST) {
-         sample_nearest_rect(ctx, tObj, magEnd - magStart,
-                             texcoords + magStart, NULL, rgba + magStart);
-      }
-      else {
-         sample_linear_rect(ctx, tObj, magEnd - magStart,
-                            texcoords + magStart, NULL, rgba + magStart);
-      }
-   }
-}
-
-
-/**********************************************************************/
-/*                2D Texture Array Sampling Functions                 */
-/**********************************************************************/
-
-/**
- * Return the texture sample for coordinate (s,t,r) using GL_NEAREST filter.
- */
-static void
-sample_2d_array_nearest(struct gl_context *ctx,
-                        const struct gl_texture_object *tObj,
-                        const struct gl_texture_image *img,
-                        const GLfloat texcoord[4],
-                        GLfloat rgba[4])
-{
-   const GLint width = img->Width2;     /* without border, power of two */
-   const GLint height = img->Height2;   /* without border, power of two */
-   const GLint depth = img->Depth;
-   GLint i, j;
-   GLint array;
-   (void) ctx;
-
-   i = nearest_texel_location(tObj->Sampler.WrapS, img, width, texcoord[0]);
-   j = nearest_texel_location(tObj->Sampler.WrapT, img, height, texcoord[1]);
-   array = tex_array_slice(texcoord[2], depth);
-
-   if (i < 0 || i >= (GLint) img->Width ||
-       j < 0 || j >= (GLint) img->Height ||
-       array < 0 || array >= (GLint) img->Depth) {
-      /* Need this test for GL_CLAMP_TO_BORDER mode */
-      get_border_color(tObj, img, rgba);
-   }
-   else {
-      img->FetchTexelf(img, i, j, array, rgba);
-   }
-}
-
-
-/**
- * Return the texture sample for coordinate (s,t,r) using GL_LINEAR filter.
- */
-static void
-sample_2d_array_linear(struct gl_context *ctx,
-                       const struct gl_texture_object *tObj,
-                       const struct gl_texture_image *img,
-                       const GLfloat texcoord[4],
-                       GLfloat rgba[4])
-{
-   const GLint width = img->Width2;
-   const GLint height = img->Height2;
-   const GLint depth = img->Depth;
-   GLint i0, j0, i1, j1;
-   GLint array;
-   GLbitfield useBorderColor = 0x0;
-   GLfloat a, b;
-   GLfloat t00[4], t01[4], t10[4], t11[4];
-
-   linear_texel_locations(tObj->Sampler.WrapS, img, width,  texcoord[0], &i0, &i1, &a);
-   linear_texel_locations(tObj->Sampler.WrapT, img, height, texcoord[1], &j0, &j1, &b);
-   array = tex_array_slice(texcoord[2], depth);
-
-   if (array < 0 || array >= depth) {
-      COPY_4V(rgba, tObj->Sampler.BorderColor.f);
-   }
-   else {
-      if (img->Border) {
-	 i0 += img->Border;
-	 i1 += img->Border;
-	 j0 += img->Border;
-	 j1 += img->Border;
-      }
-      else {
-	 /* check if sampling texture border color */
-	 if (i0 < 0 || i0 >= width)   useBorderColor |= I0BIT;
-	 if (i1 < 0 || i1 >= width)   useBorderColor |= I1BIT;
-	 if (j0 < 0 || j0 >= height)  useBorderColor |= J0BIT;
-	 if (j1 < 0 || j1 >= height)  useBorderColor |= J1BIT;
-      }
-
-      /* Fetch texels */
-      if (useBorderColor & (I0BIT | J0BIT)) {
-         get_border_color(tObj, img, t00);
-      }
-      else {
-	 img->FetchTexelf(img, i0, j0, array, t00);
-      }
-      if (useBorderColor & (I1BIT | J0BIT)) {
-         get_border_color(tObj, img, t10);
-      }
-      else {
-	 img->FetchTexelf(img, i1, j0, array, t10);
-      }
-      if (useBorderColor & (I0BIT | J1BIT)) {
-         get_border_color(tObj, img, t01);
-      }
-      else {
-	 img->FetchTexelf(img, i0, j1, array, t01);
-      }
-      if (useBorderColor & (I1BIT | J1BIT)) {
-         get_border_color(tObj, img, t11);
-      }
-      else {
-	 img->FetchTexelf(img, i1, j1, array, t11);
-      }
-      
-      /* trilinear interpolation of samples */
-      lerp_rgba_2d(rgba, a, b, t00, t10, t01, t11);
-   }
-}
-
-
-static void
-sample_2d_array_nearest_mipmap_nearest(struct gl_context *ctx,
-                                       const struct gl_texture_object *tObj,
-                                       GLuint n, const GLfloat texcoord[][4],
-                                       const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   for (i = 0; i < n; i++) {
-      GLint level = nearest_mipmap_level(tObj, lambda[i]);
-      sample_2d_array_nearest(ctx, tObj, tObj->Image[0][level], texcoord[i],
-                              rgba[i]);
-   }
-}
-
-
-static void
-sample_2d_array_linear_mipmap_nearest(struct gl_context *ctx,
-                                      const struct gl_texture_object *tObj,
-                                      GLuint n, const GLfloat texcoord[][4],
-                                      const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   ASSERT(lambda != NULL);
-   for (i = 0; i < n; i++) {
-      GLint level = nearest_mipmap_level(tObj, lambda[i]);
-      sample_2d_array_linear(ctx, tObj, tObj->Image[0][level],
-                             texcoord[i], rgba[i]);
-   }
-}
-
-
-static void
-sample_2d_array_nearest_mipmap_linear(struct gl_context *ctx,
-                                      const struct gl_texture_object *tObj,
-                                      GLuint n, const GLfloat texcoord[][4],
-                                      const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   ASSERT(lambda != NULL);
-   for (i = 0; i < n; i++) {
-      GLint level = linear_mipmap_level(tObj, lambda[i]);
-      if (level >= tObj->_MaxLevel) {
-         sample_2d_array_nearest(ctx, tObj, tObj->Image[0][tObj->_MaxLevel],
-                                 texcoord[i], rgba[i]);
-      }
-      else {
-         GLfloat t0[4], t1[4];  /* texels */
-         const GLfloat f = FRAC(lambda[i]);
-         sample_2d_array_nearest(ctx, tObj, tObj->Image[0][level  ],
-                                 texcoord[i], t0);
-         sample_2d_array_nearest(ctx, tObj, tObj->Image[0][level+1],
-                                 texcoord[i], t1);
-         lerp_rgba(rgba[i], f, t0, t1);
-      }
-   }
-}
-
-
-static void
-sample_2d_array_linear_mipmap_linear(struct gl_context *ctx,
-                                     const struct gl_texture_object *tObj,
-                                     GLuint n, const GLfloat texcoord[][4],
-                                     const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   ASSERT(lambda != NULL);
-   for (i = 0; i < n; i++) {
-      GLint level = linear_mipmap_level(tObj, lambda[i]);
-      if (level >= tObj->_MaxLevel) {
-         sample_2d_array_linear(ctx, tObj, tObj->Image[0][tObj->_MaxLevel],
-                          texcoord[i], rgba[i]);
-      }
-      else {
-         GLfloat t0[4], t1[4];  /* texels */
-         const GLfloat f = FRAC(lambda[i]);
-         sample_2d_array_linear(ctx, tObj, tObj->Image[0][level  ],
-                                texcoord[i], t0);
-         sample_2d_array_linear(ctx, tObj, tObj->Image[0][level+1],
-                                texcoord[i], t1);
-         lerp_rgba(rgba[i], f, t0, t1);
-      }
-   }
-}
-
-
-/** Sample 2D Array texture, nearest filtering for both min/magnification */
-static void
-sample_nearest_2d_array(struct gl_context *ctx,
-                        const struct gl_texture_object *tObj, GLuint n,
-                        const GLfloat texcoords[][4], const GLfloat lambda[],
-                        GLfloat rgba[][4])
-{
-   GLuint i;
-   struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
-   (void) lambda;
-   for (i = 0; i < n; i++) {
-      sample_2d_array_nearest(ctx, tObj, image, texcoords[i], rgba[i]);
-   }
-}
-
-
-
-/** Sample 2D Array texture, linear filtering for both min/magnification */
-static void
-sample_linear_2d_array(struct gl_context *ctx,
-                       const struct gl_texture_object *tObj, GLuint n,
-                       const GLfloat texcoords[][4],
-                       const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
-   (void) lambda;
-   for (i = 0; i < n; i++) {
-      sample_2d_array_linear(ctx, tObj, image, texcoords[i], rgba[i]);
-   }
-}
-
-
-/** Sample 2D Array texture, using lambda to choose between min/magnification */
-static void
-sample_lambda_2d_array(struct gl_context *ctx,
-                       const struct gl_texture_object *tObj, GLuint n,
-                       const GLfloat texcoords[][4], const GLfloat lambda[],
-                       GLfloat rgba[][4])
-{
-   GLuint minStart, minEnd;  /* texels with minification */
-   GLuint magStart, magEnd;  /* texels with magnification */
-   GLuint i;
-
-   ASSERT(lambda != NULL);
-   compute_min_mag_ranges(tObj, n, lambda,
-                          &minStart, &minEnd, &magStart, &magEnd);
-
-   if (minStart < minEnd) {
-      /* do the minified texels */
-      GLuint m = minEnd - minStart;
-      switch (tObj->Sampler.MinFilter) {
-      case GL_NEAREST:
-         for (i = minStart; i < minEnd; i++)
-            sample_2d_array_nearest(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
-                                    texcoords[i], rgba[i]);
-         break;
-      case GL_LINEAR:
-         for (i = minStart; i < minEnd; i++)
-            sample_2d_array_linear(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
-                                   texcoords[i], rgba[i]);
-         break;
-      case GL_NEAREST_MIPMAP_NEAREST:
-         sample_2d_array_nearest_mipmap_nearest(ctx, tObj, m,
-                                                texcoords + minStart,
-                                                lambda + minStart,
-                                                rgba + minStart);
-         break;
-      case GL_LINEAR_MIPMAP_NEAREST:
-         sample_2d_array_linear_mipmap_nearest(ctx, tObj, m, 
-                                               texcoords + minStart,
-                                               lambda + minStart,
-                                               rgba + minStart);
-         break;
-      case GL_NEAREST_MIPMAP_LINEAR:
-         sample_2d_array_nearest_mipmap_linear(ctx, tObj, m,
-                                               texcoords + minStart,
-                                               lambda + minStart,
-                                               rgba + minStart);
-         break;
-      case GL_LINEAR_MIPMAP_LINEAR:
-         sample_2d_array_linear_mipmap_linear(ctx, tObj, m, 
-                                              texcoords + minStart,
-                                              lambda + minStart, 
-                                              rgba + minStart);
-         break;
-      default:
-         _mesa_problem(ctx, "Bad min filter in sample_2d_array_texture");
-         return;
-      }
-   }
-
-   if (magStart < magEnd) {
-      /* do the magnified texels */
-      switch (tObj->Sampler.MagFilter) {
-      case GL_NEAREST:
-         for (i = magStart; i < magEnd; i++)
-            sample_2d_array_nearest(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
-                              texcoords[i], rgba[i]);
-         break;
-      case GL_LINEAR:
-         for (i = magStart; i < magEnd; i++)
-            sample_2d_array_linear(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
-                                   texcoords[i], rgba[i]);
-         break;
-      default:
-         _mesa_problem(ctx, "Bad mag filter in sample_2d_array_texture");
-         return;
-      }
-   }
-}
-
-
-
-
-/**********************************************************************/
-/*                1D Texture Array Sampling Functions                 */
-/**********************************************************************/
-
-/**
- * Return the texture sample for coordinate (s,t,r) using GL_NEAREST filter.
- */
-static void
-sample_1d_array_nearest(struct gl_context *ctx,
-                        const struct gl_texture_object *tObj,
-                        const struct gl_texture_image *img,
-                        const GLfloat texcoord[4],
-                        GLfloat rgba[4])
-{
-   const GLint width = img->Width2;     /* without border, power of two */
-   const GLint height = img->Height;
-   GLint i;
-   GLint array;
-   (void) ctx;
-
-   i = nearest_texel_location(tObj->Sampler.WrapS, img, width, texcoord[0]);
-   array = tex_array_slice(texcoord[1], height);
-
-   if (i < 0 || i >= (GLint) img->Width ||
-       array < 0 || array >= (GLint) img->Height) {
-      /* Need this test for GL_CLAMP_TO_BORDER mode */
-      get_border_color(tObj, img, rgba);
-   }
-   else {
-      img->FetchTexelf(img, i, array, 0, rgba);
-   }
-}
-
-
-/**
- * Return the texture sample for coordinate (s,t,r) using GL_LINEAR filter.
- */
-static void
-sample_1d_array_linear(struct gl_context *ctx,
-                       const struct gl_texture_object *tObj,
-                       const struct gl_texture_image *img,
-                       const GLfloat texcoord[4],
-                       GLfloat rgba[4])
-{
-   const GLint width = img->Width2;
-   const GLint height = img->Height;
-   GLint i0, i1;
-   GLint array;
-   GLbitfield useBorderColor = 0x0;
-   GLfloat a;
-   GLfloat t0[4], t1[4];
-
-   linear_texel_locations(tObj->Sampler.WrapS, img, width, texcoord[0], &i0, &i1, &a);
-   array = tex_array_slice(texcoord[1], height);
-
-   if (img->Border) {
-      i0 += img->Border;
-      i1 += img->Border;
-   }
-   else {
-      /* check if sampling texture border color */
-      if (i0 < 0 || i0 >= width)   useBorderColor |= I0BIT;
-      if (i1 < 0 || i1 >= width)   useBorderColor |= I1BIT;
-   }
-
-   if (array < 0 || array >= height)   useBorderColor |= K0BIT;
-
-   /* Fetch texels */
-   if (useBorderColor & (I0BIT | K0BIT)) {
-      get_border_color(tObj, img, t0);
-   }
-   else {
-      img->FetchTexelf(img, i0, array, 0, t0);
-   }
-   if (useBorderColor & (I1BIT | K0BIT)) {
-      get_border_color(tObj, img, t1);
-   }
-   else {
-      img->FetchTexelf(img, i1, array, 0, t1);
-   }
-
-   /* bilinear interpolation of samples */
-   lerp_rgba(rgba, a, t0, t1);
-}
-
-
-static void
-sample_1d_array_nearest_mipmap_nearest(struct gl_context *ctx,
-                                       const struct gl_texture_object *tObj,
-                                       GLuint n, const GLfloat texcoord[][4],
-                                       const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   for (i = 0; i < n; i++) {
-      GLint level = nearest_mipmap_level(tObj, lambda[i]);
-      sample_1d_array_nearest(ctx, tObj, tObj->Image[0][level], texcoord[i],
-                              rgba[i]);
-   }
-}
-
-
-static void
-sample_1d_array_linear_mipmap_nearest(struct gl_context *ctx,
-                                      const struct gl_texture_object *tObj,
-                                      GLuint n, const GLfloat texcoord[][4],
-                                      const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   ASSERT(lambda != NULL);
-   for (i = 0; i < n; i++) {
-      GLint level = nearest_mipmap_level(tObj, lambda[i]);
-      sample_1d_array_linear(ctx, tObj, tObj->Image[0][level],
-                             texcoord[i], rgba[i]);
-   }
-}
-
-
-static void
-sample_1d_array_nearest_mipmap_linear(struct gl_context *ctx,
-                                      const struct gl_texture_object *tObj,
-                                      GLuint n, const GLfloat texcoord[][4],
-                                      const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   ASSERT(lambda != NULL);
-   for (i = 0; i < n; i++) {
-      GLint level = linear_mipmap_level(tObj, lambda[i]);
-      if (level >= tObj->_MaxLevel) {
-         sample_1d_array_nearest(ctx, tObj, tObj->Image[0][tObj->_MaxLevel],
-                                 texcoord[i], rgba[i]);
-      }
-      else {
-         GLfloat t0[4], t1[4];  /* texels */
-         const GLfloat f = FRAC(lambda[i]);
-         sample_1d_array_nearest(ctx, tObj, tObj->Image[0][level  ], texcoord[i], t0);
-         sample_1d_array_nearest(ctx, tObj, tObj->Image[0][level+1], texcoord[i], t1);
-         lerp_rgba(rgba[i], f, t0, t1);
-      }
-   }
-}
-
-
-static void
-sample_1d_array_linear_mipmap_linear(struct gl_context *ctx,
-                                     const struct gl_texture_object *tObj,
-                                     GLuint n, const GLfloat texcoord[][4],
-                                     const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   ASSERT(lambda != NULL);
-   for (i = 0; i < n; i++) {
-      GLint level = linear_mipmap_level(tObj, lambda[i]);
-      if (level >= tObj->_MaxLevel) {
-         sample_1d_array_linear(ctx, tObj, tObj->Image[0][tObj->_MaxLevel],
-                          texcoord[i], rgba[i]);
-      }
-      else {
-         GLfloat t0[4], t1[4];  /* texels */
-         const GLfloat f = FRAC(lambda[i]);
-         sample_1d_array_linear(ctx, tObj, tObj->Image[0][level  ], texcoord[i], t0);
-         sample_1d_array_linear(ctx, tObj, tObj->Image[0][level+1], texcoord[i], t1);
-         lerp_rgba(rgba[i], f, t0, t1);
-      }
-   }
-}
-
-
-/** Sample 1D Array texture, nearest filtering for both min/magnification */
-static void
-sample_nearest_1d_array(struct gl_context *ctx,
-                        const struct gl_texture_object *tObj, GLuint n,
-                        const GLfloat texcoords[][4], const GLfloat lambda[],
-                        GLfloat rgba[][4])
-{
-   GLuint i;
-   struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
-   (void) lambda;
-   for (i = 0; i < n; i++) {
-      sample_1d_array_nearest(ctx, tObj, image, texcoords[i], rgba[i]);
-   }
-}
-
-
-/** Sample 1D Array texture, linear filtering for both min/magnification */
-static void
-sample_linear_1d_array(struct gl_context *ctx,
-                       const struct gl_texture_object *tObj, GLuint n,
-                       const GLfloat texcoords[][4],
-                       const GLfloat lambda[], GLfloat rgba[][4])
-{
-   GLuint i;
-   struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
-   (void) lambda;
-   for (i = 0; i < n; i++) {
-      sample_1d_array_linear(ctx, tObj, image, texcoords[i], rgba[i]);
-   }
-}
-
-
-/** Sample 1D Array texture, using lambda to choose between min/magnification */
-static void
-sample_lambda_1d_array(struct gl_context *ctx,
-                       const struct gl_texture_object *tObj, GLuint n,
-                       const GLfloat texcoords[][4], const GLfloat lambda[],
-                       GLfloat rgba[][4])
-{
-   GLuint minStart, minEnd;  /* texels with minification */
-   GLuint magStart, magEnd;  /* texels with magnification */
-   GLuint i;
-
-   ASSERT(lambda != NULL);
-   compute_min_mag_ranges(tObj, n, lambda,
-                          &minStart, &minEnd, &magStart, &magEnd);
-
-   if (minStart < minEnd) {
-      /* do the minified texels */
-      GLuint m = minEnd - minStart;
-      switch (tObj->Sampler.MinFilter) {
-      case GL_NEAREST:
-         for (i = minStart; i < minEnd; i++)
-            sample_1d_array_nearest(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
-                                    texcoords[i], rgba[i]);
-         break;
-      case GL_LINEAR:
-         for (i = minStart; i < minEnd; i++)
-            sample_1d_array_linear(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
-                                   texcoords[i], rgba[i]);
-         break;
-      case GL_NEAREST_MIPMAP_NEAREST:
-         sample_1d_array_nearest_mipmap_nearest(ctx, tObj, m, texcoords + minStart,
-                                                lambda + minStart, rgba + minStart);
-         break;
-      case GL_LINEAR_MIPMAP_NEAREST:
-         sample_1d_array_linear_mipmap_nearest(ctx, tObj, m, 
-                                               texcoords + minStart,
-                                               lambda + minStart,
-                                               rgba + minStart);
-         break;
-      case GL_NEAREST_MIPMAP_LINEAR:
-         sample_1d_array_nearest_mipmap_linear(ctx, tObj, m, texcoords + minStart,
-                                               lambda + minStart, rgba + minStart);
-         break;
-      case GL_LINEAR_MIPMAP_LINEAR:
-         sample_1d_array_linear_mipmap_linear(ctx, tObj, m, 
-                                              texcoords + minStart,
-                                              lambda + minStart, 
-                                              rgba + minStart);
-         break;
-      default:
-         _mesa_problem(ctx, "Bad min filter in sample_1d_array_texture");
-         return;
-      }
-   }
-
-   if (magStart < magEnd) {
-      /* do the magnified texels */
-      switch (tObj->Sampler.MagFilter) {
-      case GL_NEAREST:
-         for (i = magStart; i < magEnd; i++)
-            sample_1d_array_nearest(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
-                              texcoords[i], rgba[i]);
-         break;
-      case GL_LINEAR:
-         for (i = magStart; i < magEnd; i++)
-            sample_1d_array_linear(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
-                                   texcoords[i], rgba[i]);
-         break;
-      default:
-         _mesa_problem(ctx, "Bad mag filter in sample_1d_array_texture");
-         return;
-      }
-   }
-}
-
-
-/**
- * Compare texcoord against depth sample.  Return 1.0 or the ambient value.
- */
-static INLINE GLfloat
-shadow_compare(GLenum function, GLfloat coord, GLfloat depthSample,
-               GLfloat ambient)
-{
-   switch (function) {
-   case GL_LEQUAL:
-      return (coord <= depthSample) ? 1.0F : ambient;
-   case GL_GEQUAL:
-      return (coord >= depthSample) ? 1.0F : ambient;
-   case GL_LESS:
-      return (coord < depthSample) ? 1.0F : ambient;
-   case GL_GREATER:
-      return (coord > depthSample) ? 1.0F : ambient;
-   case GL_EQUAL:
-      return (coord == depthSample) ? 1.0F : ambient;
-   case GL_NOTEQUAL:
-      return (coord != depthSample) ? 1.0F : ambient;
-   case GL_ALWAYS:
-      return 1.0F;
-   case GL_NEVER:
-      return ambient;
-   case GL_NONE:
-      return depthSample;
-   default:
-      _mesa_problem(NULL, "Bad compare func in shadow_compare");
-      return ambient;
-   }
-}
-
-
-/**
- * Compare texcoord against four depth samples.
- */
-static INLINE GLfloat
-shadow_compare4(GLenum function, GLfloat coord,
-                GLfloat depth00, GLfloat depth01,
-                GLfloat depth10, GLfloat depth11,
-                GLfloat ambient, GLfloat wi, GLfloat wj)
-{
-   const GLfloat d = (1.0F - (GLfloat) ambient) * 0.25F;
-   GLfloat luminance = 1.0F;
-
-   switch (function) {
-   case GL_LEQUAL:
-      if (coord > depth00)  luminance -= d;
-      if (coord > depth01)  luminance -= d;
-      if (coord > depth10)  luminance -= d;
-      if (coord > depth11)  luminance -= d;
-      return luminance;
-   case GL_GEQUAL:
-      if (coord < depth00)  luminance -= d;
-      if (coord < depth01)  luminance -= d;
-      if (coord < depth10)  luminance -= d;
-      if (coord < depth11)  luminance -= d;
-      return luminance;
-   case GL_LESS:
-      if (coord >= depth00)  luminance -= d;
-      if (coord >= depth01)  luminance -= d;
-      if (coord >= depth10)  luminance -= d;
-      if (coord >= depth11)  luminance -= d;
-      return luminance;
-   case GL_GREATER:
-      if (coord <= depth00)  luminance -= d;
-      if (coord <= depth01)  luminance -= d;
-      if (coord <= depth10)  luminance -= d;
-      if (coord <= depth11)  luminance -= d;
-      return luminance;
-   case GL_EQUAL:
-      if (coord != depth00)  luminance -= d;
-      if (coord != depth01)  luminance -= d;
-      if (coord != depth10)  luminance -= d;
-      if (coord != depth11)  luminance -= d;
-      return luminance;
-   case GL_NOTEQUAL:
-      if (coord == depth00)  luminance -= d;
-      if (coord == depth01)  luminance -= d;
-      if (coord == depth10)  luminance -= d;
-      if (coord == depth11)  luminance -= d;
-      return luminance;
-   case GL_ALWAYS:
-      return 1.0F;
-   case GL_NEVER:
-      return ambient;
-   case GL_NONE:
-      /* ordinary bilinear filtering */
-      return lerp_2d(wi, wj, depth00, depth10, depth01, depth11);
-   default:
-      _mesa_problem(NULL, "Bad compare func in sample_compare4");
-      return ambient;
-   }
-}
-
-
-/**
- * Choose the mipmap level to use when sampling from a depth texture.
- */
-static int
-choose_depth_texture_level(const struct gl_texture_object *tObj, GLfloat lambda)
-{
-   GLint level;
-
-   if (tObj->Sampler.MinFilter == GL_NEAREST || tObj->Sampler.MinFilter == GL_LINEAR) {
-      /* no mipmapping - use base level */
-      level = tObj->BaseLevel;
-   }
-   else {
-      /* choose mipmap level */
-      lambda = CLAMP(lambda, tObj->Sampler.MinLod, tObj->Sampler.MaxLod);
-      level = (GLint) lambda;
-      level = CLAMP(level, tObj->BaseLevel, tObj->_MaxLevel);
-   }
-
-   return level;
-}
-
-
-/**
- * Sample a shadow/depth texture.  This function is incomplete.  It doesn't
- * check for minification vs. magnification, etc.
- */
-static void
-sample_depth_texture( struct gl_context *ctx,
-                      const struct gl_texture_object *tObj, GLuint n,
-                      const GLfloat texcoords[][4], const GLfloat lambda[],
-                      GLfloat texel[][4] )
-{
-   const GLint level = choose_depth_texture_level(tObj, lambda[0]);
-   const struct gl_texture_image *img = tObj->Image[0][level];
-   const GLint width = img->Width;
-   const GLint height = img->Height;
-   const GLint depth = img->Depth;
-   const GLuint compare_coord = (tObj->Target == GL_TEXTURE_2D_ARRAY_EXT)
-       ? 3 : 2;
-   GLfloat ambient;
-   GLenum function;
-   GLfloat result;
-
-   ASSERT(img->_BaseFormat == GL_DEPTH_COMPONENT ||
-          img->_BaseFormat == GL_DEPTH_STENCIL_EXT);
-
-   ASSERT(tObj->Target == GL_TEXTURE_1D ||
-          tObj->Target == GL_TEXTURE_2D ||
-          tObj->Target == GL_TEXTURE_RECTANGLE_NV ||
-          tObj->Target == GL_TEXTURE_1D_ARRAY_EXT ||
-          tObj->Target == GL_TEXTURE_2D_ARRAY_EXT);
-
-   ambient = tObj->Sampler.CompareFailValue;
-
-   /* XXXX if tObj->Sampler.MinFilter != tObj->Sampler.MagFilter, we're ignoring lambda */
-
-   function = (tObj->Sampler.CompareMode == GL_COMPARE_R_TO_TEXTURE_ARB) ?
-      tObj->Sampler.CompareFunc : GL_NONE;
-
-   if (tObj->Sampler.MagFilter == GL_NEAREST) {
-      GLuint i;
-      for (i = 0; i < n; i++) {
-         GLfloat depthSample, depthRef;
-         GLint col, row, slice;
-
-         nearest_texcoord(tObj, level, texcoords[i], &col, &row, &slice);
-
-         if (col >= 0 && row >= 0 && col < width && row < height && 
-             slice >= 0 && slice < depth) {
-            img->FetchTexelf(img, col, row, slice, &depthSample);
-         }
-         else {
-            depthSample = tObj->Sampler.BorderColor.f[0];
-         }
-
-         depthRef = CLAMP(texcoords[i][compare_coord], 0.0F, 1.0F);
-
-         result = shadow_compare(function, depthRef, depthSample, ambient);
-
-         switch (tObj->Sampler.DepthMode) {
-         case GL_LUMINANCE:
-            ASSIGN_4V(texel[i], result, result, result, 1.0F);
-            break;
-         case GL_INTENSITY:
-            ASSIGN_4V(texel[i], result, result, result, result);
-            break;
-         case GL_ALPHA:
-            ASSIGN_4V(texel[i], 0.0F, 0.0F, 0.0F, result);
-            break;
-         case GL_RED:
-            ASSIGN_4V(texel[i], result, 0.0F, 0.0F, 1.0F);
-            break;
-         default:
-            _mesa_problem(ctx, "Bad depth texture mode");
-         }
-      }
-   }
-   else {
-      GLuint i;
-      ASSERT(tObj->Sampler.MagFilter == GL_LINEAR);
-      for (i = 0; i < n; i++) {
-         GLfloat depth00, depth01, depth10, depth11, depthRef;
-         GLint i0, i1, j0, j1;
-         GLint slice;
-         GLfloat wi, wj;
-         GLuint useBorderTexel;
-
-         linear_texcoord(tObj, level, texcoords[i], &i0, &i1, &j0, &j1, &slice,
-                         &wi, &wj);
-
-         useBorderTexel = 0;
-         if (img->Border) {
-            i0 += img->Border;
-            i1 += img->Border;
-            if (tObj->Target != GL_TEXTURE_1D_ARRAY_EXT) {
-               j0 += img->Border;
-               j1 += img->Border;
-            }
-         }
-         else {
-            if (i0 < 0 || i0 >= (GLint) width)   useBorderTexel |= I0BIT;
-            if (i1 < 0 || i1 >= (GLint) width)   useBorderTexel |= I1BIT;
-            if (j0 < 0 || j0 >= (GLint) height)  useBorderTexel |= J0BIT;
-            if (j1 < 0 || j1 >= (GLint) height)  useBorderTexel |= J1BIT;
-         }
-
-         if (slice < 0 || slice >= (GLint) depth) {
-            depth00 = tObj->Sampler.BorderColor.f[0];
-            depth01 = tObj->Sampler.BorderColor.f[0];
-            depth10 = tObj->Sampler.BorderColor.f[0];
-            depth11 = tObj->Sampler.BorderColor.f[0];
-         }
-         else {
-            /* get four depth samples from the texture */
-            if (useBorderTexel & (I0BIT | J0BIT)) {
-               depth00 = tObj->Sampler.BorderColor.f[0];
-            }
-            else {
-               img->FetchTexelf(img, i0, j0, slice, &depth00);
-            }
-            if (useBorderTexel & (I1BIT | J0BIT)) {
-               depth10 = tObj->Sampler.BorderColor.f[0];
-            }
-            else {
-               img->FetchTexelf(img, i1, j0, slice, &depth10);
-            }
-
-            if (tObj->Target != GL_TEXTURE_1D_ARRAY_EXT) {
-               if (useBorderTexel & (I0BIT | J1BIT)) {
-                  depth01 = tObj->Sampler.BorderColor.f[0];
-               }
-               else {
-                  img->FetchTexelf(img, i0, j1, slice, &depth01);
-               }
-               if (useBorderTexel & (I1BIT | J1BIT)) {
-                  depth11 = tObj->Sampler.BorderColor.f[0];
-               }
-               else {
-                  img->FetchTexelf(img, i1, j1, slice, &depth11);
-               }
-            }
-            else {
-               depth01 = depth00;
-               depth11 = depth10;
-            }
-         }
-
-         depthRef = CLAMP(texcoords[i][compare_coord], 0.0F, 1.0F);
-
-         result = shadow_compare4(function, depthRef,
-                                  depth00, depth01, depth10, depth11,
-                                  ambient, wi, wj);
-
-         switch (tObj->Sampler.DepthMode) {
-         case GL_LUMINANCE:
-            ASSIGN_4V(texel[i], result, result, result, 1.0F);
-            break;
-         case GL_INTENSITY:
-            ASSIGN_4V(texel[i], result, result, result, result);
-            break;
-         case GL_ALPHA:
-            ASSIGN_4V(texel[i], 0.0F, 0.0F, 0.0F, result);
-            break;
-         default:
-            _mesa_problem(ctx, "Bad depth texture mode");
-         }
-
-      }  /* for */
-   }  /* if filter */
-}
-
-
-/**
- * We use this function when a texture object is in an "incomplete" state.
- * When a fragment program attempts to sample an incomplete texture we
- * return black (see issue 23 in GL_ARB_fragment_program spec).
- * Note: fragment programs don't observe the texture enable/disable flags.
- */
-static void
-null_sample_func( struct gl_context *ctx,
-		  const struct gl_texture_object *tObj, GLuint n,
-		  const GLfloat texcoords[][4], const GLfloat lambda[],
-		  GLfloat rgba[][4])
-{
-   GLuint i;
-   (void) ctx;
-   (void) tObj;
-   (void) texcoords;
-   (void) lambda;
-   for (i = 0; i < n; i++) {
-      rgba[i][RCOMP] = 0;
-      rgba[i][GCOMP] = 0;
-      rgba[i][BCOMP] = 0;
-      rgba[i][ACOMP] = 1.0;
-   }
-}
-
-
-/**
- * Choose the texture sampling function for the given texture object.
- */
-texture_sample_func
-_swrast_choose_texture_sample_func( struct gl_context *ctx,
-				    const struct gl_texture_object *t )
-{
-   if (!t || !t->_Complete) {
-      return &null_sample_func;
-   }
-   else {
-      const GLboolean needLambda =
-         (GLboolean) (t->Sampler.MinFilter != t->Sampler.MagFilter);
-      const GLenum format = t->Image[0][t->BaseLevel]->_BaseFormat;
-
-      switch (t->Target) {
-      case GL_TEXTURE_1D:
-         if (format == GL_DEPTH_COMPONENT || format == GL_DEPTH_STENCIL_EXT) {
-            return &sample_depth_texture;
-         }
-         else if (needLambda) {
-            return &sample_lambda_1d;
-         }
-         else if (t->Sampler.MinFilter == GL_LINEAR) {
-            return &sample_linear_1d;
-         }
-         else {
-            ASSERT(t->Sampler.MinFilter == GL_NEAREST);
-            return &sample_nearest_1d;
-         }
-      case GL_TEXTURE_2D:
-         if (format == GL_DEPTH_COMPONENT || format == GL_DEPTH_STENCIL_EXT) {
-            return &sample_depth_texture;
-         }
-         else if (needLambda) {
-            return &sample_lambda_2d;
-         }
-         else if (t->Sampler.MinFilter == GL_LINEAR) {
-            return &sample_linear_2d;
-         }
-         else {
-            /* check for a few optimized cases */
-            const struct gl_texture_image *img = t->Image[0][t->BaseLevel];
-            ASSERT(t->Sampler.MinFilter == GL_NEAREST);
-            if (t->Sampler.WrapS == GL_REPEAT &&
-                t->Sampler.WrapT == GL_REPEAT &&
-                img->_IsPowerOfTwo &&
-                img->Border == 0 &&
-                img->TexFormat == MESA_FORMAT_RGB888) {
-               return &opt_sample_rgb_2d;
-            }
-            else if (t->Sampler.WrapS == GL_REPEAT &&
-                     t->Sampler.WrapT == GL_REPEAT &&
-                     img->_IsPowerOfTwo &&
-                     img->Border == 0 &&
-                     img->TexFormat == MESA_FORMAT_RGBA8888) {
-               return &opt_sample_rgba_2d;
-            }
-            else {
-               return &sample_nearest_2d;
-            }
-         }
-      case GL_TEXTURE_3D:
-         if (needLambda) {
-            return &sample_lambda_3d;
-         }
-         else if (t->Sampler.MinFilter == GL_LINEAR) {
-            return &sample_linear_3d;
-         }
-         else {
-            ASSERT(t->Sampler.MinFilter == GL_NEAREST);
-            return &sample_nearest_3d;
-         }
-      case GL_TEXTURE_CUBE_MAP:
-         if (needLambda) {
-            return &sample_lambda_cube;
-         }
-         else if (t->Sampler.MinFilter == GL_LINEAR) {
-            return &sample_linear_cube;
-         }
-         else {
-            ASSERT(t->Sampler.MinFilter == GL_NEAREST);
-            return &sample_nearest_cube;
-         }
-      case GL_TEXTURE_RECTANGLE_NV:
-         if (format == GL_DEPTH_COMPONENT || format == GL_DEPTH_STENCIL_EXT) {
-            return &sample_depth_texture;
-         }
-         else if (needLambda) {
-            return &sample_lambda_rect;
-         }
-         else if (t->Sampler.MinFilter == GL_LINEAR) {
-            return &sample_linear_rect;
-         }
-         else {
-            ASSERT(t->Sampler.MinFilter == GL_NEAREST);
-            return &sample_nearest_rect;
-         }
-      case GL_TEXTURE_1D_ARRAY_EXT:
-         if (needLambda) {
-            return &sample_lambda_1d_array;
-         }
-         else if (t->Sampler.MinFilter == GL_LINEAR) {
-            return &sample_linear_1d_array;
-         }
-         else {
-            ASSERT(t->Sampler.MinFilter == GL_NEAREST);
-            return &sample_nearest_1d_array;
-         }
-      case GL_TEXTURE_2D_ARRAY_EXT:
-         if (needLambda) {
-            return &sample_lambda_2d_array;
-         }
-         else if (t->Sampler.MinFilter == GL_LINEAR) {
-            return &sample_linear_2d_array;
-         }
-         else {
-            ASSERT(t->Sampler.MinFilter == GL_NEAREST);
-            return &sample_nearest_2d_array;
-         }
-      default:
-         _mesa_problem(ctx,
-                       "invalid target in _swrast_choose_texture_sample_func");
-         return &null_sample_func;
-      }
-   }
-}
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.3
+ *
+ * Copyright (C) 1999-2008  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"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+#include "main/glheader.h"
+#include "main/context.h"
+#include "main/colormac.h"
+#include "main/imports.h"
+
+#include "s_context.h"
+#include "s_texfilter.h"
+
+
+/*
+ * Note, the FRAC macro has to work perfectly.  Otherwise you'll sometimes
+ * see 1-pixel bands of improperly weighted linear-filtered textures.
+ * The tests/texwrap.c demo is a good test.
+ * Also note, FRAC(x) doesn't truly return the fractional part of x for x < 0.
+ * Instead, if x < 0 then FRAC(x) = 1 - true_frac(x).
+ */
+#define FRAC(f)  ((f) - IFLOOR(f))
+
+
+
+/**
+ * Linear interpolation macro
+ */
+#define LERP(T, A, B)  ( (A) + (T) * ((B) - (A)) )
+
+
+/**
+ * Do 2D/biliner interpolation of float values.
+ * v00, v10, v01 and v11 are typically four texture samples in a square/box.
+ * a and b are the horizontal and vertical interpolants.
+ * It's important that this function is inlined when compiled with
+ * optimization!  If we find that's not true on some systems, convert
+ * to a macro.
+ */
+static INLINE GLfloat
+lerp_2d(GLfloat a, GLfloat b,
+        GLfloat v00, GLfloat v10, GLfloat v01, GLfloat v11)
+{
+   const GLfloat temp0 = LERP(a, v00, v10);
+   const GLfloat temp1 = LERP(a, v01, v11);
+   return LERP(b, temp0, temp1);
+}
+
+
+/**
+ * Do 3D/trilinear interpolation of float values.
+ * \sa lerp_2d
+ */
+static INLINE GLfloat
+lerp_3d(GLfloat a, GLfloat b, GLfloat c,
+        GLfloat v000, GLfloat v100, GLfloat v010, GLfloat v110,
+        GLfloat v001, GLfloat v101, GLfloat v011, GLfloat v111)
+{
+   const GLfloat temp00 = LERP(a, v000, v100);
+   const GLfloat temp10 = LERP(a, v010, v110);
+   const GLfloat temp01 = LERP(a, v001, v101);
+   const GLfloat temp11 = LERP(a, v011, v111);
+   const GLfloat temp0 = LERP(b, temp00, temp10);
+   const GLfloat temp1 = LERP(b, temp01, temp11);
+   return LERP(c, temp0, temp1);
+}
+
+
+/**
+ * Do linear interpolation of colors.
+ */
+static INLINE void
+lerp_rgba(GLfloat result[4], GLfloat t, const GLfloat a[4], const GLfloat b[4])
+{
+   result[0] = LERP(t, a[0], b[0]);
+   result[1] = LERP(t, a[1], b[1]);
+   result[2] = LERP(t, a[2], b[2]);
+   result[3] = LERP(t, a[3], b[3]);
+}
+
+
+/**
+ * Do bilinear interpolation of colors.
+ */
+static INLINE void
+lerp_rgba_2d(GLfloat result[4], GLfloat a, GLfloat b,
+             const GLfloat t00[4], const GLfloat t10[4],
+             const GLfloat t01[4], const GLfloat t11[4])
+{
+   result[0] = lerp_2d(a, b, t00[0], t10[0], t01[0], t11[0]);
+   result[1] = lerp_2d(a, b, t00[1], t10[1], t01[1], t11[1]);
+   result[2] = lerp_2d(a, b, t00[2], t10[2], t01[2], t11[2]);
+   result[3] = lerp_2d(a, b, t00[3], t10[3], t01[3], t11[3]);
+}
+
+
+/**
+ * Do trilinear interpolation of colors.
+ */
+static INLINE void
+lerp_rgba_3d(GLfloat result[4], GLfloat a, GLfloat b, GLfloat c,
+             const GLfloat t000[4], const GLfloat t100[4],
+             const GLfloat t010[4], const GLfloat t110[4],
+             const GLfloat t001[4], const GLfloat t101[4],
+             const GLfloat t011[4], const GLfloat t111[4])
+{
+   GLuint k;
+   /* compiler should unroll these short loops */
+   for (k = 0; k < 4; k++) {
+      result[k] = lerp_3d(a, b, c, t000[k], t100[k], t010[k], t110[k],
+                                   t001[k], t101[k], t011[k], t111[k]);
+   }
+}
+
+
+/**
+ * Used for GL_REPEAT wrap mode.  Using A % B doesn't produce the
+ * right results for A<0.  Casting to A to be unsigned only works if B
+ * is a power of two.  Adding a bias to A (which is a multiple of B)
+ * avoids the problems with A < 0 (for reasonable A) without using a
+ * conditional.
+ */
+#define REMAINDER(A, B) (((A) + (B) * 1024) % (B))
+
+
+/**
+ * Used to compute texel locations for linear sampling.
+ * Input:
+ *    wrapMode = GL_REPEAT, GL_CLAMP, GL_CLAMP_TO_EDGE, GL_CLAMP_TO_BORDER
+ *    s = texcoord in [0,1]
+ *    size = width (or height or depth) of texture
+ * Output:
+ *    i0, i1 = returns two nearest texel indexes
+ *    weight = returns blend factor between texels
+ */
+static INLINE void
+linear_texel_locations(GLenum wrapMode,
+                       const struct gl_texture_image *img,
+                       GLint size, GLfloat s,
+                       GLint *i0, GLint *i1, GLfloat *weight)
+{
+   GLfloat u;
+   switch (wrapMode) {
+   case GL_REPEAT:
+      u = s * size - 0.5F;
+      if (img->_IsPowerOfTwo) {
+         *i0 = IFLOOR(u) & (size - 1);
+         *i1 = (*i0 + 1) & (size - 1);
+      }
+      else {
+         *i0 = REMAINDER(IFLOOR(u), size);
+         *i1 = REMAINDER(*i0 + 1, size);
+      }
+      break;
+   case GL_CLAMP_TO_EDGE:
+      if (s <= 0.0F)
+         u = 0.0F;
+      else if (s >= 1.0F)
+         u = (GLfloat) size;
+      else
+         u = s * size;
+      u -= 0.5F;
+      *i0 = IFLOOR(u);
+      *i1 = *i0 + 1;
+      if (*i0 < 0)
+         *i0 = 0;
+      if (*i1 >= (GLint) size)
+         *i1 = size - 1;
+      break;
+   case GL_CLAMP_TO_BORDER:
+      {
+         const GLfloat min = -1.0F / (2.0F * size);
+         const GLfloat max = 1.0F - min;
+         if (s <= min)
+            u = min * size;
+         else if (s >= max)
+            u = max * size;
+         else
+            u = s * size;
+         u -= 0.5F;
+         *i0 = IFLOOR(u);
+         *i1 = *i0 + 1;
+      }
+      break;
+   case GL_MIRRORED_REPEAT:
+      {
+         const GLint flr = IFLOOR(s);
+         if (flr & 1)
+            u = 1.0F - (s - (GLfloat) flr);
+         else
+            u = s - (GLfloat) flr;
+         u = (u * size) - 0.5F;
+         *i0 = IFLOOR(u);
+         *i1 = *i0 + 1;
+         if (*i0 < 0)
+            *i0 = 0;
+         if (*i1 >= (GLint) size)
+            *i1 = size - 1;
+      }
+      break;
+   case GL_MIRROR_CLAMP_EXT:
+      u = FABSF(s);
+      if (u >= 1.0F)
+         u = (GLfloat) size;
+      else
+         u *= size;
+      u -= 0.5F;
+      *i0 = IFLOOR(u);
+      *i1 = *i0 + 1;
+      break;
+   case GL_MIRROR_CLAMP_TO_EDGE_EXT:
+      u = FABSF(s);
+      if (u >= 1.0F)
+         u = (GLfloat) size;
+      else
+         u *= size;
+      u -= 0.5F;
+      *i0 = IFLOOR(u);
+      *i1 = *i0 + 1;
+      if (*i0 < 0)
+         *i0 = 0;
+      if (*i1 >= (GLint) size)
+         *i1 = size - 1;
+      break;
+   case GL_MIRROR_CLAMP_TO_BORDER_EXT:
+      {
+         const GLfloat min = -1.0F / (2.0F * size);
+         const GLfloat max = 1.0F - min;
+         u = FABSF(s);
+         if (u <= min)
+            u = min * size;
+         else if (u >= max)
+            u = max * size;
+         else
+            u *= size;
+         u -= 0.5F;
+         *i0 = IFLOOR(u);
+         *i1 = *i0 + 1;
+      }
+      break;
+   case GL_CLAMP:
+      if (s <= 0.0F)
+         u = 0.0F;
+      else if (s >= 1.0F)
+         u = (GLfloat) size;
+      else
+         u = s * size;
+      u -= 0.5F;
+      *i0 = IFLOOR(u);
+      *i1 = *i0 + 1;
+      break;
+   default:
+      _mesa_problem(NULL, "Bad wrap mode");
+      u = 0.0F;
+   }
+   *weight = FRAC(u);
+}
+
+
+/**
+ * Used to compute texel location for nearest sampling.
+ */
+static INLINE GLint
+nearest_texel_location(GLenum wrapMode,
+                       const struct gl_texture_image *img,
+                       GLint size, GLfloat s)
+{
+   GLint i;
+
+   switch (wrapMode) {
+   case GL_REPEAT:
+      /* s limited to [0,1) */
+      /* i limited to [0,size-1] */
+      i = IFLOOR(s * size);
+      if (img->_IsPowerOfTwo)
+         i &= (size - 1);
+      else
+         i = REMAINDER(i, size);
+      return i;
+   case GL_CLAMP_TO_EDGE:
+      {
+         /* s limited to [min,max] */
+         /* i limited to [0, size-1] */
+         const GLfloat min = 1.0F / (2.0F * size);
+         const GLfloat max = 1.0F - min;
+         if (s < min)
+            i = 0;
+         else if (s > max)
+            i = size - 1;
+         else
+            i = IFLOOR(s * size);
+      }
+      return i;
+   case GL_CLAMP_TO_BORDER:
+      {
+         /* s limited to [min,max] */
+         /* i limited to [-1, size] */
+         const GLfloat min = -1.0F / (2.0F * size);
+         const GLfloat max = 1.0F - min;
+         if (s <= min)
+            i = -1;
+         else if (s >= max)
+            i = size;
+         else
+            i = IFLOOR(s * size);
+      }
+      return i;
+   case GL_MIRRORED_REPEAT:
+      {
+         const GLfloat min = 1.0F / (2.0F * size);
+         const GLfloat max = 1.0F - min;
+         const GLint flr = IFLOOR(s);
+         GLfloat u;
+         if (flr & 1)
+            u = 1.0F - (s - (GLfloat) flr);
+         else
+            u = s - (GLfloat) flr;
+         if (u < min)
+            i = 0;
+         else if (u > max)
+            i = size - 1;
+         else
+            i = IFLOOR(u * size);
+      }
+      return i;
+   case GL_MIRROR_CLAMP_EXT:
+      {
+         /* s limited to [0,1] */
+         /* i limited to [0,size-1] */
+         const GLfloat u = FABSF(s);
+         if (u <= 0.0F)
+            i = 0;
+         else if (u >= 1.0F)
+            i = size - 1;
+         else
+            i = IFLOOR(u * size);
+      }
+      return i;
+   case GL_MIRROR_CLAMP_TO_EDGE_EXT:
+      {
+         /* s limited to [min,max] */
+         /* i limited to [0, size-1] */
+         const GLfloat min = 1.0F / (2.0F * size);
+         const GLfloat max = 1.0F - min;
+         const GLfloat u = FABSF(s);
+         if (u < min)
+            i = 0;
+         else if (u > max)
+            i = size - 1;
+         else
+            i = IFLOOR(u * size);
+      }
+      return i;
+   case GL_MIRROR_CLAMP_TO_BORDER_EXT:
+      {
+         /* s limited to [min,max] */
+         /* i limited to [0, size-1] */
+         const GLfloat min = -1.0F / (2.0F * size);
+         const GLfloat max = 1.0F - min;
+         const GLfloat u = FABSF(s);
+         if (u < min)
+            i = -1;
+         else if (u > max)
+            i = size;
+         else
+            i = IFLOOR(u * size);
+      }
+      return i;
+   case GL_CLAMP:
+      /* s limited to [0,1] */
+      /* i limited to [0,size-1] */
+      if (s <= 0.0F)
+         i = 0;
+      else if (s >= 1.0F)
+         i = size - 1;
+      else
+         i = IFLOOR(s * size);
+      return i;
+   default:
+      _mesa_problem(NULL, "Bad wrap mode");
+      return 0;
+   }
+}
+
+
+/* Power of two image sizes only */
+static INLINE void
+linear_repeat_texel_location(GLuint size, GLfloat s,
+                             GLint *i0, GLint *i1, GLfloat *weight)
+{
+   GLfloat u = s * size - 0.5F;
+   *i0 = IFLOOR(u) & (size - 1);
+   *i1 = (*i0 + 1) & (size - 1);
+   *weight = FRAC(u);
+}
+
+
+/**
+ * Do clamp/wrap for a texture rectangle coord, GL_NEAREST filter mode.
+ */
+static INLINE GLint
+clamp_rect_coord_nearest(GLenum wrapMode, GLfloat coord, GLint max)
+{
+   switch (wrapMode) {
+   case GL_CLAMP:
+      return IFLOOR( CLAMP(coord, 0.0F, max - 1) );
+   case GL_CLAMP_TO_EDGE:
+      return IFLOOR( CLAMP(coord, 0.5F, max - 0.5F) );
+   case GL_CLAMP_TO_BORDER:
+      return IFLOOR( CLAMP(coord, -0.5F, max + 0.5F) );
+   default:
+      _mesa_problem(NULL, "bad wrapMode in clamp_rect_coord_nearest");
+      return 0;
+   }
+}
+
+
+/**
+ * As above, but GL_LINEAR filtering.
+ */
+static INLINE void
+clamp_rect_coord_linear(GLenum wrapMode, GLfloat coord, GLint max,
+                        GLint *i0out, GLint *i1out, GLfloat *weight)
+{
+   GLfloat fcol;
+   GLint i0, i1;
+   switch (wrapMode) {
+   case GL_CLAMP:
+      /* Not exactly what the spec says, but it matches NVIDIA output */
+      fcol = CLAMP(coord - 0.5F, 0.0F, max - 1);
+      i0 = IFLOOR(fcol);
+      i1 = i0 + 1;
+      break;
+   case GL_CLAMP_TO_EDGE:
+      fcol = CLAMP(coord, 0.5F, max - 0.5F);
+      fcol -= 0.5F;
+      i0 = IFLOOR(fcol);
+      i1 = i0 + 1;
+      if (i1 > max - 1)
+         i1 = max - 1;
+      break;
+   case GL_CLAMP_TO_BORDER:
+      fcol = CLAMP(coord, -0.5F, max + 0.5F);
+      fcol -= 0.5F;
+      i0 = IFLOOR(fcol);
+      i1 = i0 + 1;
+      break;
+   default:
+      _mesa_problem(NULL, "bad wrapMode in clamp_rect_coord_linear");
+      i0 = i1 = 0;
+      fcol = 0.0F;
+   }
+   *i0out = i0;
+   *i1out = i1;
+   *weight = FRAC(fcol);
+}
+
+
+/**
+ * Compute slice/image to use for 1D or 2D array texture.
+ */
+static INLINE GLint
+tex_array_slice(GLfloat coord, GLsizei size)
+{
+   GLint slice = IFLOOR(coord + 0.5f);
+   slice = CLAMP(slice, 0, size - 1);
+   return slice;
+}
+
+
+/**
+ * Compute nearest integer texcoords for given texobj and coordinate.
+ * NOTE: only used for depth texture sampling.
+ */
+static INLINE void
+nearest_texcoord(const struct gl_texture_object *texObj,
+                 GLuint level,
+                 const GLfloat texcoord[4],
+                 GLint *i, GLint *j, GLint *k)
+{
+   const struct gl_texture_image *img = texObj->Image[0][level];
+   const GLint width = img->Width;
+   const GLint height = img->Height;
+   const GLint depth = img->Depth;
+
+   switch (texObj->Target) {
+   case GL_TEXTURE_RECTANGLE_ARB:
+      *i = clamp_rect_coord_nearest(texObj->Sampler.WrapS, texcoord[0], width);
+      *j = clamp_rect_coord_nearest(texObj->Sampler.WrapT, texcoord[1], height);
+      *k = 0;
+      break;
+   case GL_TEXTURE_1D:
+      *i = nearest_texel_location(texObj->Sampler.WrapS, img, width, texcoord[0]);
+      *j = 0;
+      *k = 0;
+      break;
+   case GL_TEXTURE_2D:
+      *i = nearest_texel_location(texObj->Sampler.WrapS, img, width, texcoord[0]);
+      *j = nearest_texel_location(texObj->Sampler.WrapT, img, height, texcoord[1]);
+      *k = 0;
+      break;
+   case GL_TEXTURE_1D_ARRAY_EXT:
+      *i = nearest_texel_location(texObj->Sampler.WrapS, img, width, texcoord[0]);
+      *j = tex_array_slice(texcoord[1], height);
+      *k = 0;
+      break;
+   case GL_TEXTURE_2D_ARRAY_EXT:
+      *i = nearest_texel_location(texObj->Sampler.WrapS, img, width, texcoord[0]);
+      *j = nearest_texel_location(texObj->Sampler.WrapT, img, height, texcoord[1]);
+      *k = tex_array_slice(texcoord[2], depth);
+      break;
+   default:
+      *i = *j = *k = 0;
+   }
+}
+
+
+/**
+ * Compute linear integer texcoords for given texobj and coordinate.
+ * NOTE: only used for depth texture sampling.
+ */
+static INLINE void
+linear_texcoord(const struct gl_texture_object *texObj,
+                GLuint level,
+                const GLfloat texcoord[4],
+                GLint *i0, GLint *i1, GLint *j0, GLint *j1, GLint *slice,
+                GLfloat *wi, GLfloat *wj)
+{
+   const struct gl_texture_image *img = texObj->Image[0][level];
+   const GLint width = img->Width;
+   const GLint height = img->Height;
+   const GLint depth = img->Depth;
+
+   switch (texObj->Target) {
+   case GL_TEXTURE_RECTANGLE_ARB:
+      clamp_rect_coord_linear(texObj->Sampler.WrapS, texcoord[0],
+                              width, i0, i1, wi);
+      clamp_rect_coord_linear(texObj->Sampler.WrapT, texcoord[1],
+                              height, j0, j1, wj);
+      *slice = 0;
+      break;
+
+   case GL_TEXTURE_1D:
+   case GL_TEXTURE_2D:
+      linear_texel_locations(texObj->Sampler.WrapS, img, width,
+                             texcoord[0], i0, i1, wi);
+      linear_texel_locations(texObj->Sampler.WrapT, img, height,
+                             texcoord[1], j0, j1, wj);
+      *slice = 0;
+      break;
+
+   case GL_TEXTURE_1D_ARRAY_EXT:
+      linear_texel_locations(texObj->Sampler.WrapS, img, width,
+                             texcoord[0], i0, i1, wi);
+      *j0 = tex_array_slice(texcoord[1], height);
+      *j1 = *j0;
+      *slice = 0;
+      break;
+
+   case GL_TEXTURE_2D_ARRAY_EXT:
+      linear_texel_locations(texObj->Sampler.WrapS, img, width,
+                             texcoord[0], i0, i1, wi);
+      linear_texel_locations(texObj->Sampler.WrapT, img, height,
+                             texcoord[1], j0, j1, wj);
+      *slice = tex_array_slice(texcoord[2], depth);
+      break;
+
+   default:
+      *slice = 0;
+   }
+}
+
+
+
+/**
+ * For linear interpolation between mipmap levels N and N+1, this function
+ * computes N.
+ */
+static INLINE GLint
+linear_mipmap_level(const struct gl_texture_object *tObj, GLfloat lambda)
+{
+   if (lambda < 0.0F)
+      return tObj->BaseLevel;
+   else if (lambda > tObj->_MaxLambda)
+      return (GLint) (tObj->BaseLevel + tObj->_MaxLambda);
+   else
+      return (GLint) (tObj->BaseLevel + lambda);
+}
+
+
+/**
+ * Compute the nearest mipmap level to take texels from.
+ */
+static INLINE GLint
+nearest_mipmap_level(const struct gl_texture_object *tObj, GLfloat lambda)
+{
+   GLfloat l;
+   GLint level;
+   if (lambda <= 0.5F)
+      l = 0.0F;
+   else if (lambda > tObj->_MaxLambda + 0.4999F)
+      l = tObj->_MaxLambda + 0.4999F;
+   else
+      l = lambda;
+   level = (GLint) (tObj->BaseLevel + l + 0.5F);
+   if (level > tObj->_MaxLevel)
+      level = tObj->_MaxLevel;
+   return level;
+}
+
+
+
+/*
+ * Bitflags for texture border color sampling.
+ */
+#define I0BIT   1
+#define I1BIT   2
+#define J0BIT   4
+#define J1BIT   8
+#define K0BIT  16
+#define K1BIT  32
+
+
+
+/**
+ * The lambda[] array values are always monotonic.  Either the whole span
+ * will be minified, magnified, or split between the two.  This function
+ * determines the subranges in [0, n-1] that are to be minified or magnified.
+ */
+static INLINE void
+compute_min_mag_ranges(const struct gl_texture_object *tObj,
+                       GLuint n, const GLfloat lambda[],
+                       GLuint *minStart, GLuint *minEnd,
+                       GLuint *magStart, GLuint *magEnd)
+{
+   GLfloat minMagThresh;
+
+   /* we shouldn't be here if minfilter == magfilter */
+   ASSERT(tObj->Sampler.MinFilter != tObj->Sampler.MagFilter);
+
+   /* This bit comes from the OpenGL spec: */
+   if (tObj->Sampler.MagFilter == GL_LINEAR
+       && (tObj->Sampler.MinFilter == GL_NEAREST_MIPMAP_NEAREST ||
+           tObj->Sampler.MinFilter == GL_NEAREST_MIPMAP_LINEAR)) {
+      minMagThresh = 0.5F;
+   }
+   else {
+      minMagThresh = 0.0F;
+   }
+
+#if 0
+   /* DEBUG CODE: Verify that lambda[] is monotonic.
+    * We can't really use this because the inaccuracy in the LOG2 function
+    * causes this test to fail, yet the resulting texturing is correct.
+    */
+   if (n > 1) {
+      GLuint i;
+      printf("lambda delta = %g\n", lambda[0] - lambda[n-1]);
+      if (lambda[0] >= lambda[n-1]) { /* decreasing */
+         for (i = 0; i < n - 1; i++) {
+            ASSERT((GLint) (lambda[i] * 10) >= (GLint) (lambda[i+1] * 10));
+         }
+      }
+      else { /* increasing */
+         for (i = 0; i < n - 1; i++) {
+            ASSERT((GLint) (lambda[i] * 10) <= (GLint) (lambda[i+1] * 10));
+         }
+      }
+   }
+#endif /* DEBUG */
+
+   if (lambda[0] <= minMagThresh && (n <= 1 || lambda[n-1] <= minMagThresh)) {
+      /* magnification for whole span */
+      *magStart = 0;
+      *magEnd = n;
+      *minStart = *minEnd = 0;
+   }
+   else if (lambda[0] > minMagThresh && (n <=1 || lambda[n-1] > minMagThresh)) {
+      /* minification for whole span */
+      *minStart = 0;
+      *minEnd = n;
+      *magStart = *magEnd = 0;
+   }
+   else {
+      /* a mix of minification and magnification */
+      GLuint i;
+      if (lambda[0] > minMagThresh) {
+         /* start with minification */
+         for (i = 1; i < n; i++) {
+            if (lambda[i] <= minMagThresh)
+               break;
+         }
+         *minStart = 0;
+         *minEnd = i;
+         *magStart = i;
+         *magEnd = n;
+      }
+      else {
+         /* start with magnification */
+         for (i = 1; i < n; i++) {
+            if (lambda[i] > minMagThresh)
+               break;
+         }
+         *magStart = 0;
+         *magEnd = i;
+         *minStart = i;
+         *minEnd = n;
+      }
+   }
+
+#if 0
+   /* Verify the min/mag Start/End values
+    * We don't use this either (see above)
+    */
+   {
+      GLint i;
+      for (i = 0; i < n; i++) {
+         if (lambda[i] > minMagThresh) {
+            /* minification */
+            ASSERT(i >= *minStart);
+            ASSERT(i < *minEnd);
+         }
+         else {
+            /* magnification */
+            ASSERT(i >= *magStart);
+            ASSERT(i < *magEnd);
+         }
+      }
+   }
+#endif
+}
+
+
+/**
+ * When we sample the border color, it must be interpreted according to
+ * the base texture format.  Ex: if the texture base format it GL_ALPHA,
+ * we return (0,0,0,BorderAlpha).
+ */
+static INLINE void
+get_border_color(const struct gl_texture_object *tObj,
+                 const struct gl_texture_image *img,
+                 GLfloat rgba[4])
+{
+   switch (img->_BaseFormat) {
+   case GL_RGB:
+      rgba[0] = tObj->Sampler.BorderColor.f[0];
+      rgba[1] = tObj->Sampler.BorderColor.f[1];
+      rgba[2] = tObj->Sampler.BorderColor.f[2];
+      rgba[3] = 1.0F;
+      break;
+   case GL_ALPHA:
+      rgba[0] = rgba[1] = rgba[2] = 0.0;
+      rgba[3] = tObj->Sampler.BorderColor.f[3];
+      break;
+   case GL_LUMINANCE:
+      rgba[0] = rgba[1] = rgba[2] = tObj->Sampler.BorderColor.f[0];
+      rgba[3] = 1.0;
+      break;
+   case GL_LUMINANCE_ALPHA:
+      rgba[0] = rgba[1] = rgba[2] = tObj->Sampler.BorderColor.f[0];
+      rgba[3] = tObj->Sampler.BorderColor.f[3];
+      break;
+   case GL_INTENSITY:
+      rgba[0] = rgba[1] = rgba[2] = rgba[3] = tObj->Sampler.BorderColor.f[0];
+      break;
+   default:
+      COPY_4V(rgba, tObj->Sampler.BorderColor.f);
+   }
+}
+
+
+/**********************************************************************/
+/*                    1-D Texture Sampling Functions                  */
+/**********************************************************************/
+
+/**
+ * Return the texture sample for coordinate (s) using GL_NEAREST filter.
+ */
+static INLINE void
+sample_1d_nearest(struct gl_context *ctx,
+                  const struct gl_texture_object *tObj,
+                  const struct gl_texture_image *img,
+                  const GLfloat texcoord[4], GLfloat rgba[4])
+{
+   const GLint width = img->Width2;  /* without border, power of two */
+   GLint i;
+   i = nearest_texel_location(tObj->Sampler.WrapS, img, width, texcoord[0]);
+   /* skip over the border, if any */
+   i += img->Border;
+   if (i < 0 || i >= (GLint) img->Width) {
+      /* Need this test for GL_CLAMP_TO_BORDER mode */
+      get_border_color(tObj, img, rgba);
+   }
+   else {
+      img->FetchTexelf(img, i, 0, 0, rgba);
+   }
+}
+
+
+/**
+ * Return the texture sample for coordinate (s) using GL_LINEAR filter.
+ */
+static INLINE void
+sample_1d_linear(struct gl_context *ctx,
+                 const struct gl_texture_object *tObj,
+                 const struct gl_texture_image *img,
+                 const GLfloat texcoord[4], GLfloat rgba[4])
+{
+   const GLint width = img->Width2;
+   GLint i0, i1;
+   GLbitfield useBorderColor = 0x0;
+   GLfloat a;
+   GLfloat t0[4], t1[4];  /* texels */
+
+   linear_texel_locations(tObj->Sampler.WrapS, img, width, texcoord[0], &i0, &i1, &a);
+
+   if (img->Border) {
+      i0 += img->Border;
+      i1 += img->Border;
+   }
+   else {
+      if (i0 < 0 || i0 >= width)   useBorderColor |= I0BIT;
+      if (i1 < 0 || i1 >= width)   useBorderColor |= I1BIT;
+   }
+
+   /* fetch texel colors */
+   if (useBorderColor & I0BIT) {
+      get_border_color(tObj, img, t0);
+   }
+   else {
+      img->FetchTexelf(img, i0, 0, 0, t0);
+   }
+   if (useBorderColor & I1BIT) {
+      get_border_color(tObj, img, t1);
+   }
+   else {
+      img->FetchTexelf(img, i1, 0, 0, t1);
+   }
+
+   lerp_rgba(rgba, a, t0, t1);
+}
+
+
+static void
+sample_1d_nearest_mipmap_nearest(struct gl_context *ctx,
+                                 const struct gl_texture_object *tObj,
+                                 GLuint n, const GLfloat texcoord[][4],
+                                 const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   ASSERT(lambda != NULL);
+   for (i = 0; i < n; i++) {
+      GLint level = nearest_mipmap_level(tObj, lambda[i]);
+      sample_1d_nearest(ctx, tObj, tObj->Image[0][level], texcoord[i], rgba[i]);
+   }
+}
+
+
+static void
+sample_1d_linear_mipmap_nearest(struct gl_context *ctx,
+                                const struct gl_texture_object *tObj,
+                                GLuint n, const GLfloat texcoord[][4],
+                                const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   ASSERT(lambda != NULL);
+   for (i = 0; i < n; i++) {
+      GLint level = nearest_mipmap_level(tObj, lambda[i]);
+      sample_1d_linear(ctx, tObj, tObj->Image[0][level], texcoord[i], rgba[i]);
+   }
+}
+
+
+static void
+sample_1d_nearest_mipmap_linear(struct gl_context *ctx,
+                                const struct gl_texture_object *tObj,
+                                GLuint n, const GLfloat texcoord[][4],
+                                const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   ASSERT(lambda != NULL);
+   for (i = 0; i < n; i++) {
+      GLint level = linear_mipmap_level(tObj, lambda[i]);
+      if (level >= tObj->_MaxLevel) {
+         sample_1d_nearest(ctx, tObj, tObj->Image[0][tObj->_MaxLevel],
+                           texcoord[i], rgba[i]);
+      }
+      else {
+         GLfloat t0[4], t1[4];
+         const GLfloat f = FRAC(lambda[i]);
+         sample_1d_nearest(ctx, tObj, tObj->Image[0][level  ], texcoord[i], t0);
+         sample_1d_nearest(ctx, tObj, tObj->Image[0][level+1], texcoord[i], t1);
+         lerp_rgba(rgba[i], f, t0, t1);
+      }
+   }
+}
+
+
+static void
+sample_1d_linear_mipmap_linear(struct gl_context *ctx,
+                               const struct gl_texture_object *tObj,
+                               GLuint n, const GLfloat texcoord[][4],
+                               const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   ASSERT(lambda != NULL);
+   for (i = 0; i < n; i++) {
+      GLint level = linear_mipmap_level(tObj, lambda[i]);
+      if (level >= tObj->_MaxLevel) {
+         sample_1d_linear(ctx, tObj, tObj->Image[0][tObj->_MaxLevel],
+                          texcoord[i], rgba[i]);
+      }
+      else {
+         GLfloat t0[4], t1[4];
+         const GLfloat f = FRAC(lambda[i]);
+         sample_1d_linear(ctx, tObj, tObj->Image[0][level  ], texcoord[i], t0);
+         sample_1d_linear(ctx, tObj, tObj->Image[0][level+1], texcoord[i], t1);
+         lerp_rgba(rgba[i], f, t0, t1);
+      }
+   }
+}
+
+
+/** Sample 1D texture, nearest filtering for both min/magnification */
+static void
+sample_nearest_1d( struct gl_context *ctx,
+                   const struct gl_texture_object *tObj, GLuint n,
+                   const GLfloat texcoords[][4], const GLfloat lambda[],
+                   GLfloat rgba[][4] )
+{
+   GLuint i;
+   struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
+   (void) lambda;
+   for (i = 0; i < n; i++) {
+      sample_1d_nearest(ctx, tObj, image, texcoords[i], rgba[i]);
+   }
+}
+
+
+/** Sample 1D texture, linear filtering for both min/magnification */
+static void
+sample_linear_1d( struct gl_context *ctx,
+                  const struct gl_texture_object *tObj, GLuint n,
+                  const GLfloat texcoords[][4], const GLfloat lambda[],
+                  GLfloat rgba[][4] )
+{
+   GLuint i;
+   struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
+   (void) lambda;
+   for (i = 0; i < n; i++) {
+      sample_1d_linear(ctx, tObj, image, texcoords[i], rgba[i]);
+   }
+}
+
+
+/** Sample 1D texture, using lambda to choose between min/magnification */
+static void
+sample_lambda_1d( struct gl_context *ctx,
+                  const struct gl_texture_object *tObj, GLuint n,
+                  const GLfloat texcoords[][4],
+                  const GLfloat lambda[], GLfloat rgba[][4] )
+{
+   GLuint minStart, minEnd;  /* texels with minification */
+   GLuint magStart, magEnd;  /* texels with magnification */
+   GLuint i;
+
+   ASSERT(lambda != NULL);
+   compute_min_mag_ranges(tObj, n, lambda,
+                          &minStart, &minEnd, &magStart, &magEnd);
+
+   if (minStart < minEnd) {
+      /* do the minified texels */
+      const GLuint m = minEnd - minStart;
+      switch (tObj->Sampler.MinFilter) {
+      case GL_NEAREST:
+         for (i = minStart; i < minEnd; i++)
+            sample_1d_nearest(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
+                              texcoords[i], rgba[i]);
+         break;
+      case GL_LINEAR:
+         for (i = minStart; i < minEnd; i++)
+            sample_1d_linear(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
+                             texcoords[i], rgba[i]);
+         break;
+      case GL_NEAREST_MIPMAP_NEAREST:
+         sample_1d_nearest_mipmap_nearest(ctx, tObj, m, texcoords + minStart,
+                                          lambda + minStart, rgba + minStart);
+         break;
+      case GL_LINEAR_MIPMAP_NEAREST:
+         sample_1d_linear_mipmap_nearest(ctx, tObj, m, texcoords + minStart,
+                                         lambda + minStart, rgba + minStart);
+         break;
+      case GL_NEAREST_MIPMAP_LINEAR:
+         sample_1d_nearest_mipmap_linear(ctx, tObj, m, texcoords + minStart,
+                                         lambda + minStart, rgba + minStart);
+         break;
+      case GL_LINEAR_MIPMAP_LINEAR:
+         sample_1d_linear_mipmap_linear(ctx, tObj, m, texcoords + minStart,
+                                        lambda + minStart, rgba + minStart);
+         break;
+      default:
+         _mesa_problem(ctx, "Bad min filter in sample_1d_texture");
+         return;
+      }
+   }
+
+   if (magStart < magEnd) {
+      /* do the magnified texels */
+      switch (tObj->Sampler.MagFilter) {
+      case GL_NEAREST:
+         for (i = magStart; i < magEnd; i++)
+            sample_1d_nearest(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
+                              texcoords[i], rgba[i]);
+         break;
+      case GL_LINEAR:
+         for (i = magStart; i < magEnd; i++)
+            sample_1d_linear(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
+                             texcoords[i], rgba[i]);
+         break;
+      default:
+         _mesa_problem(ctx, "Bad mag filter in sample_1d_texture");
+         return;
+      }
+   }
+}
+
+
+/**********************************************************************/
+/*                    2-D Texture Sampling Functions                  */
+/**********************************************************************/
+
+
+/**
+ * Return the texture sample for coordinate (s,t) using GL_NEAREST filter.
+ */
+static INLINE void
+sample_2d_nearest(struct gl_context *ctx,
+                  const struct gl_texture_object *tObj,
+                  const struct gl_texture_image *img,
+                  const GLfloat texcoord[4],
+                  GLfloat rgba[])
+{
+   const GLint width = img->Width2;    /* without border, power of two */
+   const GLint height = img->Height2;  /* without border, power of two */
+   GLint i, j;
+   (void) ctx;
+
+   i = nearest_texel_location(tObj->Sampler.WrapS, img, width, texcoord[0]);
+   j = nearest_texel_location(tObj->Sampler.WrapT, img, height, texcoord[1]);
+
+   /* skip over the border, if any */
+   i += img->Border;
+   j += img->Border;
+
+   if (i < 0 || i >= (GLint) img->Width || j < 0 || j >= (GLint) img->Height) {
+      /* Need this test for GL_CLAMP_TO_BORDER mode */
+      get_border_color(tObj, img, rgba);
+   }
+   else {
+      img->FetchTexelf(img, i, j, 0, rgba);
+   }
+}
+
+
+/**
+ * Return the texture sample for coordinate (s,t) using GL_LINEAR filter.
+ * New sampling code contributed by Lynn Quam <quam@ai.sri.com>.
+ */
+static INLINE void
+sample_2d_linear(struct gl_context *ctx,
+                 const struct gl_texture_object *tObj,
+                 const struct gl_texture_image *img,
+                 const GLfloat texcoord[4],
+                 GLfloat rgba[])
+{
+   const GLint width = img->Width2;
+   const GLint height = img->Height2;
+   GLint i0, j0, i1, j1;
+   GLbitfield useBorderColor = 0x0;
+   GLfloat a, b;
+   GLfloat t00[4], t10[4], t01[4], t11[4]; /* sampled texel colors */
+
+   linear_texel_locations(tObj->Sampler.WrapS, img, width, texcoord[0],  &i0, &i1, &a);
+   linear_texel_locations(tObj->Sampler.WrapT, img, height, texcoord[1], &j0, &j1, &b);
+
+   if (img->Border) {
+      i0 += img->Border;
+      i1 += img->Border;
+      j0 += img->Border;
+      j1 += img->Border;
+   }
+   else {
+      if (i0 < 0 || i0 >= width)   useBorderColor |= I0BIT;
+      if (i1 < 0 || i1 >= width)   useBorderColor |= I1BIT;
+      if (j0 < 0 || j0 >= height)  useBorderColor |= J0BIT;
+      if (j1 < 0 || j1 >= height)  useBorderColor |= J1BIT;
+   }
+
+   /* fetch four texel colors */
+   if (useBorderColor & (I0BIT | J0BIT)) {
+      get_border_color(tObj, img, t00);
+   }
+   else {
+      img->FetchTexelf(img, i0, j0, 0, t00);
+   }
+   if (useBorderColor & (I1BIT | J0BIT)) {
+      get_border_color(tObj, img, t10);
+   }
+   else {
+      img->FetchTexelf(img, i1, j0, 0, t10);
+   }
+   if (useBorderColor & (I0BIT | J1BIT)) {
+      get_border_color(tObj, img, t01);
+   }
+   else {
+      img->FetchTexelf(img, i0, j1, 0, t01);
+   }
+   if (useBorderColor & (I1BIT | J1BIT)) {
+      get_border_color(tObj, img, t11);
+   }
+   else {
+      img->FetchTexelf(img, i1, j1, 0, t11);
+   }
+
+   lerp_rgba_2d(rgba, a, b, t00, t10, t01, t11);
+}
+
+
+/**
+ * As above, but we know WRAP_S == REPEAT and WRAP_T == REPEAT.
+ * We don't have to worry about the texture border.
+ */
+static INLINE void
+sample_2d_linear_repeat(struct gl_context *ctx,
+                        const struct gl_texture_object *tObj,
+                        const struct gl_texture_image *img,
+                        const GLfloat texcoord[4],
+                        GLfloat rgba[])
+{
+   const GLint width = img->Width2;
+   const GLint height = img->Height2;
+   GLint i0, j0, i1, j1;
+   GLfloat wi, wj;
+   GLfloat t00[4], t10[4], t01[4], t11[4]; /* sampled texel colors */
+
+   (void) ctx;
+
+   ASSERT(tObj->Sampler.WrapS == GL_REPEAT);
+   ASSERT(tObj->Sampler.WrapT == GL_REPEAT);
+   ASSERT(img->Border == 0);
+   ASSERT(img->_BaseFormat != GL_COLOR_INDEX);
+   ASSERT(img->_IsPowerOfTwo);
+
+   linear_repeat_texel_location(width,  texcoord[0], &i0, &i1, &wi);
+   linear_repeat_texel_location(height, texcoord[1], &j0, &j1, &wj);
+
+   img->FetchTexelf(img, i0, j0, 0, t00);
+   img->FetchTexelf(img, i1, j0, 0, t10);
+   img->FetchTexelf(img, i0, j1, 0, t01);
+   img->FetchTexelf(img, i1, j1, 0, t11);
+
+   lerp_rgba_2d(rgba, wi, wj, t00, t10, t01, t11);
+}
+
+
+static void
+sample_2d_nearest_mipmap_nearest(struct gl_context *ctx,
+                                 const struct gl_texture_object *tObj,
+                                 GLuint n, const GLfloat texcoord[][4],
+                                 const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   for (i = 0; i < n; i++) {
+      GLint level = nearest_mipmap_level(tObj, lambda[i]);
+      sample_2d_nearest(ctx, tObj, tObj->Image[0][level], texcoord[i], rgba[i]);
+   }
+}
+
+
+static void
+sample_2d_linear_mipmap_nearest(struct gl_context *ctx,
+                                const struct gl_texture_object *tObj,
+                                GLuint n, const GLfloat texcoord[][4],
+                                const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   ASSERT(lambda != NULL);
+   for (i = 0; i < n; i++) {
+      GLint level = nearest_mipmap_level(tObj, lambda[i]);
+      sample_2d_linear(ctx, tObj, tObj->Image[0][level], texcoord[i], rgba[i]);
+   }
+}
+
+
+static void
+sample_2d_nearest_mipmap_linear(struct gl_context *ctx,
+                                const struct gl_texture_object *tObj,
+                                GLuint n, const GLfloat texcoord[][4],
+                                const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   ASSERT(lambda != NULL);
+   for (i = 0; i < n; i++) {
+      GLint level = linear_mipmap_level(tObj, lambda[i]);
+      if (level >= tObj->_MaxLevel) {
+         sample_2d_nearest(ctx, tObj, tObj->Image[0][tObj->_MaxLevel],
+                           texcoord[i], rgba[i]);
+      }
+      else {
+         GLfloat t0[4], t1[4];  /* texels */
+         const GLfloat f = FRAC(lambda[i]);
+         sample_2d_nearest(ctx, tObj, tObj->Image[0][level  ], texcoord[i], t0);
+         sample_2d_nearest(ctx, tObj, tObj->Image[0][level+1], texcoord[i], t1);
+         lerp_rgba(rgba[i], f, t0, t1);
+      }
+   }
+}
+
+
+static void
+sample_2d_linear_mipmap_linear( struct gl_context *ctx,
+                                const struct gl_texture_object *tObj,
+                                GLuint n, const GLfloat texcoord[][4],
+                                const GLfloat lambda[], GLfloat rgba[][4] )
+{
+   GLuint i;
+   ASSERT(lambda != NULL);
+   for (i = 0; i < n; i++) {
+      GLint level = linear_mipmap_level(tObj, lambda[i]);
+      if (level >= tObj->_MaxLevel) {
+         sample_2d_linear(ctx, tObj, tObj->Image[0][tObj->_MaxLevel],
+                          texcoord[i], rgba[i]);
+      }
+      else {
+         GLfloat t0[4], t1[4];  /* texels */
+         const GLfloat f = FRAC(lambda[i]);
+         sample_2d_linear(ctx, tObj, tObj->Image[0][level  ], texcoord[i], t0);
+         sample_2d_linear(ctx, tObj, tObj->Image[0][level+1], texcoord[i], t1);
+         lerp_rgba(rgba[i], f, t0, t1);
+      }
+   }
+}
+
+
+static void
+sample_2d_linear_mipmap_linear_repeat(struct gl_context *ctx,
+                                      const struct gl_texture_object *tObj,
+                                      GLuint n, const GLfloat texcoord[][4],
+                                      const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   ASSERT(lambda != NULL);
+   ASSERT(tObj->Sampler.WrapS == GL_REPEAT);
+   ASSERT(tObj->Sampler.WrapT == GL_REPEAT);
+   for (i = 0; i < n; i++) {
+      GLint level = linear_mipmap_level(tObj, lambda[i]);
+      if (level >= tObj->_MaxLevel) {
+         sample_2d_linear_repeat(ctx, tObj, tObj->Image[0][tObj->_MaxLevel],
+                                 texcoord[i], rgba[i]);
+      }
+      else {
+         GLfloat t0[4], t1[4];  /* texels */
+         const GLfloat f = FRAC(lambda[i]);
+         sample_2d_linear_repeat(ctx, tObj, tObj->Image[0][level  ],
+                                 texcoord[i], t0);
+         sample_2d_linear_repeat(ctx, tObj, tObj->Image[0][level+1],
+                                 texcoord[i], t1);
+         lerp_rgba(rgba[i], f, t0, t1);
+      }
+   }
+}
+
+
+/** Sample 2D texture, nearest filtering for both min/magnification */
+static void
+sample_nearest_2d(struct gl_context *ctx,
+                  const struct gl_texture_object *tObj, GLuint n,
+                  const GLfloat texcoords[][4],
+                  const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
+   (void) lambda;
+   for (i = 0; i < n; i++) {
+      sample_2d_nearest(ctx, tObj, image, texcoords[i], rgba[i]);
+   }
+}
+
+
+/** Sample 2D texture, linear filtering for both min/magnification */
+static void
+sample_linear_2d(struct gl_context *ctx,
+                 const struct gl_texture_object *tObj, GLuint n,
+                 const GLfloat texcoords[][4],
+                 const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
+   (void) lambda;
+   if (tObj->Sampler.WrapS == GL_REPEAT &&
+       tObj->Sampler.WrapT == GL_REPEAT &&
+       image->_IsPowerOfTwo &&
+       image->Border == 0) {
+      for (i = 0; i < n; i++) {
+         sample_2d_linear_repeat(ctx, tObj, image, texcoords[i], rgba[i]);
+      }
+   }
+   else {
+      for (i = 0; i < n; i++) {
+         sample_2d_linear(ctx, tObj, image, texcoords[i], rgba[i]);
+      }
+   }
+}
+
+
+/**
+ * Optimized 2-D texture sampling:
+ *    S and T wrap mode == GL_REPEAT
+ *    GL_NEAREST min/mag filter
+ *    No border, 
+ *    RowStride == Width,
+ *    Format = GL_RGB
+ */
+static void
+opt_sample_rgb_2d(struct gl_context *ctx,
+                  const struct gl_texture_object *tObj,
+                  GLuint n, const GLfloat texcoords[][4],
+                  const GLfloat lambda[], GLfloat rgba[][4])
+{
+   const struct gl_texture_image *img = tObj->Image[0][tObj->BaseLevel];
+   const GLfloat width = (GLfloat) img->Width;
+   const GLfloat height = (GLfloat) img->Height;
+   const GLint colMask = img->Width - 1;
+   const GLint rowMask = img->Height - 1;
+   const GLint shift = img->WidthLog2;
+   GLuint k;
+   (void) ctx;
+   (void) lambda;
+   ASSERT(tObj->Sampler.WrapS==GL_REPEAT);
+   ASSERT(tObj->Sampler.WrapT==GL_REPEAT);
+   ASSERT(img->Border==0);
+   ASSERT(img->TexFormat == MESA_FORMAT_RGB888);
+   ASSERT(img->_IsPowerOfTwo);
+
+   for (k=0; k<n; k++) {
+      GLint i = IFLOOR(texcoords[k][0] * width) & colMask;
+      GLint j = IFLOOR(texcoords[k][1] * height) & rowMask;
+      GLint pos = (j << shift) | i;
+      GLubyte *texel = ((GLubyte *) img->Data) + 3*pos;
+      rgba[k][RCOMP] = UBYTE_TO_FLOAT(texel[2]);
+      rgba[k][GCOMP] = UBYTE_TO_FLOAT(texel[1]);
+      rgba[k][BCOMP] = UBYTE_TO_FLOAT(texel[0]);
+      rgba[k][ACOMP] = 1.0F;
+   }
+}
+
+
+/**
+ * Optimized 2-D texture sampling:
+ *    S and T wrap mode == GL_REPEAT
+ *    GL_NEAREST min/mag filter
+ *    No border
+ *    RowStride == Width,
+ *    Format = GL_RGBA
+ */
+static void
+opt_sample_rgba_2d(struct gl_context *ctx,
+                   const struct gl_texture_object *tObj,
+                   GLuint n, const GLfloat texcoords[][4],
+                   const GLfloat lambda[], GLfloat rgba[][4])
+{
+   const struct gl_texture_image *img = tObj->Image[0][tObj->BaseLevel];
+   const GLfloat width = (GLfloat) img->Width;
+   const GLfloat height = (GLfloat) img->Height;
+   const GLint colMask = img->Width - 1;
+   const GLint rowMask = img->Height - 1;
+   const GLint shift = img->WidthLog2;
+   GLuint i;
+   (void) ctx;
+   (void) lambda;
+   ASSERT(tObj->Sampler.WrapS==GL_REPEAT);
+   ASSERT(tObj->Sampler.WrapT==GL_REPEAT);
+   ASSERT(img->Border==0);
+   ASSERT(img->TexFormat == MESA_FORMAT_RGBA8888);
+   ASSERT(img->_IsPowerOfTwo);
+
+   for (i = 0; i < n; i++) {
+      const GLint col = IFLOOR(texcoords[i][0] * width) & colMask;
+      const GLint row = IFLOOR(texcoords[i][1] * height) & rowMask;
+      const GLint pos = (row << shift) | col;
+      const GLuint texel = *((GLuint *) img->Data + pos);
+      rgba[i][RCOMP] = UBYTE_TO_FLOAT( (texel >> 24)        );
+      rgba[i][GCOMP] = UBYTE_TO_FLOAT( (texel >> 16) & 0xff );
+      rgba[i][BCOMP] = UBYTE_TO_FLOAT( (texel >>  8) & 0xff );
+      rgba[i][ACOMP] = UBYTE_TO_FLOAT( (texel      ) & 0xff );
+   }
+}
+
+
+/** Sample 2D texture, using lambda to choose between min/magnification */
+static void
+sample_lambda_2d(struct gl_context *ctx,
+                 const struct gl_texture_object *tObj,
+                 GLuint n, const GLfloat texcoords[][4],
+                 const GLfloat lambda[], GLfloat rgba[][4])
+{
+   const struct gl_texture_image *tImg = tObj->Image[0][tObj->BaseLevel];
+   GLuint minStart, minEnd;  /* texels with minification */
+   GLuint magStart, magEnd;  /* texels with magnification */
+
+   const GLboolean repeatNoBorderPOT = (tObj->Sampler.WrapS == GL_REPEAT)
+      && (tObj->Sampler.WrapT == GL_REPEAT)
+      && (tImg->Border == 0 && (tImg->Width == tImg->RowStride))
+      && (tImg->_BaseFormat != GL_COLOR_INDEX)
+      && tImg->_IsPowerOfTwo;
+
+   ASSERT(lambda != NULL);
+   compute_min_mag_ranges(tObj, n, lambda,
+                          &minStart, &minEnd, &magStart, &magEnd);
+
+   if (minStart < minEnd) {
+      /* do the minified texels */
+      const GLuint m = minEnd - minStart;
+      switch (tObj->Sampler.MinFilter) {
+      case GL_NEAREST:
+         if (repeatNoBorderPOT) {
+            switch (tImg->TexFormat) {
+            case MESA_FORMAT_RGB888:
+               opt_sample_rgb_2d(ctx, tObj, m, texcoords + minStart,
+                                 NULL, rgba + minStart);
+               break;
+            case MESA_FORMAT_RGBA8888:
+	       opt_sample_rgba_2d(ctx, tObj, m, texcoords + minStart,
+                                  NULL, rgba + minStart);
+               break;
+            default:
+               sample_nearest_2d(ctx, tObj, m, texcoords + minStart,
+                                 NULL, rgba + minStart );
+            }
+         }
+         else {
+            sample_nearest_2d(ctx, tObj, m, texcoords + minStart,
+                              NULL, rgba + minStart);
+         }
+         break;
+      case GL_LINEAR:
+	 sample_linear_2d(ctx, tObj, m, texcoords + minStart,
+			  NULL, rgba + minStart);
+         break;
+      case GL_NEAREST_MIPMAP_NEAREST:
+         sample_2d_nearest_mipmap_nearest(ctx, tObj, m,
+                                          texcoords + minStart,
+                                          lambda + minStart, rgba + minStart);
+         break;
+      case GL_LINEAR_MIPMAP_NEAREST:
+         sample_2d_linear_mipmap_nearest(ctx, tObj, m, texcoords + minStart,
+                                         lambda + minStart, rgba + minStart);
+         break;
+      case GL_NEAREST_MIPMAP_LINEAR:
+         sample_2d_nearest_mipmap_linear(ctx, tObj, m, texcoords + minStart,
+                                         lambda + minStart, rgba + minStart);
+         break;
+      case GL_LINEAR_MIPMAP_LINEAR:
+         if (repeatNoBorderPOT)
+            sample_2d_linear_mipmap_linear_repeat(ctx, tObj, m,
+                  texcoords + minStart, lambda + minStart, rgba + minStart);
+         else
+            sample_2d_linear_mipmap_linear(ctx, tObj, m, texcoords + minStart,
+                                        lambda + minStart, rgba + minStart);
+         break;
+      default:
+         _mesa_problem(ctx, "Bad min filter in sample_2d_texture");
+         return;
+      }
+   }
+
+   if (magStart < magEnd) {
+      /* do the magnified texels */
+      const GLuint m = magEnd - magStart;
+
+      switch (tObj->Sampler.MagFilter) {
+      case GL_NEAREST:
+         if (repeatNoBorderPOT) {
+            switch (tImg->TexFormat) {
+            case MESA_FORMAT_RGB888:
+               opt_sample_rgb_2d(ctx, tObj, m, texcoords + magStart,
+                                 NULL, rgba + magStart);
+               break;
+            case MESA_FORMAT_RGBA8888:
+	       opt_sample_rgba_2d(ctx, tObj, m, texcoords + magStart,
+                                  NULL, rgba + magStart);
+               break;
+            default:
+               sample_nearest_2d(ctx, tObj, m, texcoords + magStart,
+                                 NULL, rgba + magStart );
+            }
+         }
+         else {
+            sample_nearest_2d(ctx, tObj, m, texcoords + magStart,
+                              NULL, rgba + magStart);
+         }
+         break;
+      case GL_LINEAR:
+	 sample_linear_2d(ctx, tObj, m, texcoords + magStart,
+			  NULL, rgba + magStart);
+         break;
+      default:
+         _mesa_problem(ctx, "Bad mag filter in sample_lambda_2d");
+      }
+   }
+}
+
+
+/* For anisotropic filtering */
+#define WEIGHT_LUT_SIZE 1024
+
+static GLfloat *weightLut = NULL;
+
+/**
+ * Creates the look-up table used to speed-up EWA sampling
+ */
+static void
+create_filter_table(void)
+{
+   GLuint i;
+   if (!weightLut) {
+      weightLut = (GLfloat *) malloc(WEIGHT_LUT_SIZE * sizeof(GLfloat));
+
+      for (i = 0; i < WEIGHT_LUT_SIZE; ++i) {
+         GLfloat alpha = 2;
+         GLfloat r2 = (GLfloat) i / (GLfloat) (WEIGHT_LUT_SIZE - 1);
+         GLfloat weight = (GLfloat) exp(-alpha * r2);
+         weightLut[i] = weight;
+      }
+   }
+}
+
+
+/**
+ * Elliptical weighted average (EWA) filter for producing high quality
+ * anisotropic filtered results.
+ * Based on the Higher Quality Elliptical Weighted Avarage Filter
+ * published by Paul S. Heckbert in his Master's Thesis
+ * "Fundamentals of Texture Mapping and Image Warping" (1989)
+ */
+static void
+sample_2d_ewa(struct gl_context *ctx,
+              const struct gl_texture_object *tObj,
+              const GLfloat texcoord[4],
+              const GLfloat dudx, const GLfloat dvdx,
+              const GLfloat dudy, const GLfloat dvdy, const GLint lod,
+              GLfloat rgba[])
+{
+   GLint level = lod > 0 ? lod : 0;
+   GLfloat scaling = 1.0 / (1 << level);
+   const struct gl_texture_image *img =	tObj->Image[0][level];
+   const struct gl_texture_image *mostDetailedImage =
+      tObj->Image[0][tObj->BaseLevel];
+   GLfloat tex_u=-0.5 + texcoord[0] * mostDetailedImage->WidthScale * scaling;
+   GLfloat tex_v=-0.5 + texcoord[1] * mostDetailedImage->HeightScale * scaling;
+
+   GLfloat ux = dudx * scaling;
+   GLfloat vx = dvdx * scaling;
+   GLfloat uy = dudy * scaling;
+   GLfloat vy = dvdy * scaling;
+
+   /* compute ellipse coefficients to bound the region: 
+    * A*x*x + B*x*y + C*y*y = F.
+    */
+   GLfloat A = vx*vx+vy*vy+1;
+   GLfloat B = -2*(ux*vx+uy*vy);
+   GLfloat C = ux*ux+uy*uy+1;
+   GLfloat F = A*C-B*B/4.0;
+
+   /* check if it is an ellipse */
+   /* ASSERT(F > 0.0); */
+
+   /* Compute the ellipse's (u,v) bounding box in texture space */
+   GLfloat d = -B*B+4.0*C*A;
+   GLfloat box_u = 2.0 / d * sqrt(d*C*F); /* box_u -> half of bbox with   */
+   GLfloat box_v = 2.0 / d * sqrt(A*d*F); /* box_v -> half of bbox height */
+
+   GLint u0 = floor(tex_u - box_u);
+   GLint u1 = ceil (tex_u + box_u);
+   GLint v0 = floor(tex_v - box_v);
+   GLint v1 = ceil (tex_v + box_v);
+
+   GLfloat num[4] = {0.0F, 0.0F, 0.0F, 0.0F};
+   GLfloat newCoord[2];
+   GLfloat den = 0.0F;
+   GLfloat ddq;
+   GLfloat U = u0 - tex_u;
+   GLint v;
+
+   /* Scale ellipse formula to directly index the Filter Lookup Table.
+    * i.e. scale so that F = WEIGHT_LUT_SIZE-1
+    */
+   double formScale = (double) (WEIGHT_LUT_SIZE - 1) / F;
+   A *= formScale;
+   B *= formScale;
+   C *= formScale;
+   /* F *= formScale; */ /* no need to scale F as we don't use it below here */
+
+   /* Heckbert MS thesis, p. 59; scan over the bounding box of the ellipse
+    * and incrementally update the value of Ax^2+Bxy*Cy^2; when this
+    * value, q, is less than F, we're inside the ellipse
+    */
+   ddq = 2 * A;
+   for (v = v0; v <= v1; ++v) {
+      GLfloat V = v - tex_v;
+      GLfloat dq = A * (2 * U + 1) + B * V;
+      GLfloat q = (C * V + B * U) * V + A * U * U;
+
+      GLint u;
+      for (u = u0; u <= u1; ++u) {
+         /* Note that the ellipse has been pre-scaled so F = WEIGHT_LUT_SIZE - 1 */
+         if (q < WEIGHT_LUT_SIZE) {
+            /* as a LUT is used, q must never be negative;
+             * should not happen, though
+             */
+            const GLint qClamped = q >= 0.0F ? q : 0;
+            GLfloat weight = weightLut[qClamped];
+
+            newCoord[0] = u / ((GLfloat) img->Width2);
+            newCoord[1] = v / ((GLfloat) img->Height2);
+
+            sample_2d_nearest(ctx, tObj, img, newCoord, rgba);
+            num[0] += weight * rgba[0];
+            num[1] += weight * rgba[1];
+            num[2] += weight * rgba[2];
+            num[3] += weight * rgba[3];
+
+            den += weight;
+         }
+         q += dq;
+         dq += ddq;
+      }
+   }
+
+   if (den <= 0.0F) {
+      /* Reaching this place would mean
+       * that no pixels intersected the ellipse.
+       * This should never happen because
+       * the filter we use always
+       * intersects at least one pixel.
+       */
+
+      /*rgba[0]=0;
+      rgba[1]=0;
+      rgba[2]=0;
+      rgba[3]=0;*/
+      /* not enough pixels in resampling, resort to direct interpolation */
+      sample_2d_linear(ctx, tObj, img, texcoord, rgba);
+      return;
+   }
+
+   rgba[0] = num[0] / den;
+   rgba[1] = num[1] / den;
+   rgba[2] = num[2] / den;
+   rgba[3] = num[3] / den;
+}
+
+
+/**
+ * Anisotropic filtering using footprint assembly as outlined in the
+ * EXT_texture_filter_anisotropic spec:
+ * http://www.opengl.org/registry/specs/EXT/texture_filter_anisotropic.txt
+ * Faster than EWA but has less quality (more aliasing effects)
+ */
+static void
+sample_2d_footprint(struct gl_context *ctx,
+                 const struct gl_texture_object *tObj,
+                 const GLfloat texcoord[4],
+                 const GLfloat dudx, const GLfloat dvdx,
+                 const GLfloat dudy, const GLfloat dvdy, const GLint lod,
+                 GLfloat rgba[])
+{
+   GLint level = lod > 0 ? lod : 0;
+   GLfloat scaling = 1.0F / (1 << level);
+   const struct gl_texture_image *img = tObj->Image[0][level];
+
+   GLfloat ux = dudx * scaling;
+   GLfloat vx = dvdx * scaling;
+   GLfloat uy = dudy * scaling;
+   GLfloat vy = dvdy * scaling;
+
+   GLfloat Px2 = ux * ux + vx * vx; /* squared length of dx */
+   GLfloat Py2 = uy * uy + vy * vy; /* squared length of dy */
+
+   GLint numSamples;
+   GLfloat ds;
+   GLfloat dt;
+
+   GLfloat num[4] = {0.0F, 0.0F, 0.0F, 0.0F};
+   GLfloat newCoord[2];
+   GLint s;
+
+   /*  Calculate the per anisotropic sample offsets in s,t space. */
+   if (Px2 > Py2) {
+      numSamples = ceil(SQRTF(Px2));
+      ds = ux / ((GLfloat) img->Width2);
+      dt = vx / ((GLfloat) img->Height2);
+   }
+   else {
+      numSamples = ceil(SQRTF(Py2));
+      ds = uy / ((GLfloat) img->Width2);
+      dt = vy / ((GLfloat) img->Height2);
+   }
+
+   for (s = 0; s<numSamples; s++) {
+      newCoord[0] = texcoord[0] + ds * ((GLfloat)(s+1) / (numSamples+1) -0.5);
+      newCoord[1] = texcoord[1] + dt * ((GLfloat)(s+1) / (numSamples+1) -0.5);
+
+      sample_2d_linear(ctx, tObj, img, newCoord, rgba);
+      num[0] += rgba[0];
+      num[1] += rgba[1];
+      num[2] += rgba[2];
+      num[3] += rgba[3];
+   }
+
+   rgba[0] = num[0] / numSamples;
+   rgba[1] = num[1] / numSamples;
+   rgba[2] = num[2] / numSamples;
+   rgba[3] = num[3] / numSamples;
+}
+
+
+/**
+ * Returns the index of the specified texture object in the
+ * gl_context texture unit array.
+ */
+static INLINE GLuint
+texture_unit_index(const struct gl_context *ctx,
+                   const struct gl_texture_object *tObj)
+{
+   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.Unit[u]._Current == tObj)
+         break; /* found */
+   }
+   if (u >= maxUnit)
+      u = 0; /* not found, use 1st one; should never happen */
+   
+   return u;
+}
+
+
+/**
+ * Sample 2D texture using an anisotropic filter.
+ * NOTE: the const GLfloat lambda_iso[] parameter does *NOT* contain
+ * the lambda float array but a "hidden" SWspan struct which is required
+ * by this function but is not available in the texture_sample_func signature.
+ * See _swrast_texture_span( struct gl_context *ctx, SWspan *span ) on how
+ * this function is called.
+ */
+static void
+sample_lambda_2d_aniso(struct gl_context *ctx,
+                       const struct gl_texture_object *tObj,
+                       GLuint n, const GLfloat texcoords[][4],
+                       const GLfloat lambda_iso[], GLfloat rgba[][4])
+{
+   const struct gl_texture_image *tImg = tObj->Image[0][tObj->BaseLevel];
+   const GLfloat maxEccentricity =
+      tObj->Sampler.MaxAnisotropy * tObj->Sampler.MaxAnisotropy;
+   
+   /* re-calculate the lambda values so that they are usable with anisotropic
+    * filtering
+    */
+   SWspan *span = (SWspan *)lambda_iso; /* access the "hidden" SWspan struct */
+
+   /* based on interpolate_texcoords(struct gl_context *ctx, SWspan *span)
+    * in swrast/s_span.c
+    */
+   
+   /* find the texture unit index by looking up the current texture object
+    * from the context list of available texture objects.
+    */
+   const GLuint u = texture_unit_index(ctx, tObj);
+   const GLuint attr = FRAG_ATTRIB_TEX0 + u;
+   GLfloat texW, texH;
+
+   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 dqdx = span->attrStepX[attr][3];
+   const GLfloat dqdy = span->attrStepY[attr][3];
+   GLfloat s = span->attrStart[attr][0] + span->leftClip * dsdx;
+   GLfloat t = span->attrStart[attr][1] + span->leftClip * dtdx;
+   GLfloat q = span->attrStart[attr][3] + span->leftClip * dqdx;
+
+   /* from swrast/s_texcombine.c _swrast_texture_span */
+   const struct gl_texture_unit *texUnit = &ctx->Texture.Unit[u];
+   const GLboolean adjustLOD =
+      (texUnit->LodBias + tObj->Sampler.LodBias != 0.0F)
+      || (tObj->Sampler.MinLod != -1000.0 || tObj->Sampler.MaxLod != 1000.0);
+
+   GLuint i;
+   
+   /* on first access create the lookup table containing the filter weights. */
+   if (!weightLut) {
+      create_filter_table();
+   }
+
+   texW = tImg->WidthScale;
+   texH = tImg->HeightScale;
+
+   for (i = 0; i < n; i++) {
+      const GLfloat invQ = (q == 0.0F) ? 1.0F : (1.0F / q);
+      
+      GLfloat dudx = texW * ((s + dsdx) / (q + dqdx) - s * invQ);
+      GLfloat dvdx = texH * ((t + dtdx) / (q + dqdx) - t * invQ);
+      GLfloat dudy = texW * ((s + dsdy) / (q + dqdy) - s * invQ);
+      GLfloat dvdy = texH * ((t + dtdy) / (q + dqdy) - t * invQ);
+      
+      /* note: instead of working with Px and Py, we will use the 
+       * squared length instead, to avoid sqrt.
+       */
+      GLfloat Px2 = dudx * dudx + dvdx * dvdx;
+      GLfloat Py2 = dudy * dudy + dvdy * dvdy;
+
+      GLfloat Pmax2;
+      GLfloat Pmin2;
+      GLfloat e;
+      GLfloat lod;
+
+      s += dsdx;
+      t += dtdx;
+      q += dqdx;
+      
+      if (Px2 < Py2) {
+         Pmax2 = Py2;
+         Pmin2 = Px2;
+      }
+      else {
+         Pmax2 = Px2;
+         Pmin2 = Py2;
+      }
+      
+      /* if the eccentricity of the ellipse is too big, scale up the shorter
+       * of the two vectors to limit the maximum amount of work per pixel
+       */
+      e = Pmax2 / Pmin2;
+      if (e > maxEccentricity) {
+         /* GLfloat s=e / maxEccentricity;
+            minor[0] *= s;
+            minor[1] *= s;
+            Pmin2 *= s; */
+         Pmin2 = Pmax2 / maxEccentricity;
+      }
+      
+      /* note: we need to have Pmin=sqrt(Pmin2) here, but we can avoid
+       * this since 0.5*log(x) = log(sqrt(x))
+       */
+      lod = 0.5 * LOG2(Pmin2);
+      
+      if (adjustLOD) {
+         /* from swrast/s_texcombine.c _swrast_texture_span */
+         if (texUnit->LodBias + tObj->Sampler.LodBias != 0.0F) {
+            /* apply LOD bias, but don't clamp yet */
+            const GLfloat bias =
+               CLAMP(texUnit->LodBias + tObj->Sampler.LodBias,
+                     -ctx->Const.MaxTextureLodBias,
+                     ctx->Const.MaxTextureLodBias);
+            lod += bias;
+
+            if (tObj->Sampler.MinLod != -1000.0 ||
+                tObj->Sampler.MaxLod != 1000.0) {
+               /* apply LOD clamping to lambda */
+               lod = CLAMP(lod, tObj->Sampler.MinLod, tObj->Sampler.MaxLod);
+            }
+         }
+      }
+      
+      /* If the ellipse covers the whole image, we can
+       * simply return the average of the whole image.
+       */
+      if (lod >= tObj->_MaxLevel) {
+         sample_2d_linear(ctx, tObj, tObj->Image[0][tObj->_MaxLevel],
+                          texcoords[i], rgba[i]);
+      }
+      else {
+         /* don't bother interpolating between multiple LODs; it doesn't
+          * seem to be worth the extra running time.
+          */
+         sample_2d_ewa(ctx, tObj, texcoords[i],
+                       dudx, dvdx, dudy, dvdy, floor(lod), rgba[i]);
+
+         /* unused: */
+         (void) sample_2d_footprint;
+         /*
+         sample_2d_footprint(ctx, tObj, texcoords[i],
+                             dudx, dvdx, dudy, dvdy, floor(lod), rgba[i]);
+         */
+      }
+   }
+}
+
+
+
+/**********************************************************************/
+/*                    3-D Texture Sampling Functions                  */
+/**********************************************************************/
+
+/**
+ * Return the texture sample for coordinate (s,t,r) using GL_NEAREST filter.
+ */
+static INLINE void
+sample_3d_nearest(struct gl_context *ctx,
+                  const struct gl_texture_object *tObj,
+                  const struct gl_texture_image *img,
+                  const GLfloat texcoord[4],
+                  GLfloat rgba[4])
+{
+   const GLint width = img->Width2;     /* without border, power of two */
+   const GLint height = img->Height2;   /* without border, power of two */
+   const GLint depth = img->Depth2;     /* without border, power of two */
+   GLint i, j, k;
+   (void) ctx;
+
+   i = nearest_texel_location(tObj->Sampler.WrapS, img, width, texcoord[0]);
+   j = nearest_texel_location(tObj->Sampler.WrapT, img, height, texcoord[1]);
+   k = nearest_texel_location(tObj->Sampler.WrapR, img, depth, texcoord[2]);
+
+   if (i < 0 || i >= (GLint) img->Width ||
+       j < 0 || j >= (GLint) img->Height ||
+       k < 0 || k >= (GLint) img->Depth) {
+      /* Need this test for GL_CLAMP_TO_BORDER mode */
+      get_border_color(tObj, img, rgba);
+   }
+   else {
+      img->FetchTexelf(img, i, j, k, rgba);
+   }
+}
+
+
+/**
+ * Return the texture sample for coordinate (s,t,r) using GL_LINEAR filter.
+ */
+static void
+sample_3d_linear(struct gl_context *ctx,
+                 const struct gl_texture_object *tObj,
+                 const struct gl_texture_image *img,
+                 const GLfloat texcoord[4],
+                 GLfloat rgba[4])
+{
+   const GLint width = img->Width2;
+   const GLint height = img->Height2;
+   const GLint depth = img->Depth2;
+   GLint i0, j0, k0, i1, j1, k1;
+   GLbitfield useBorderColor = 0x0;
+   GLfloat a, b, c;
+   GLfloat t000[4], t010[4], t001[4], t011[4];
+   GLfloat t100[4], t110[4], t101[4], t111[4];
+
+   linear_texel_locations(tObj->Sampler.WrapS, img, width, texcoord[0],  &i0, &i1, &a);
+   linear_texel_locations(tObj->Sampler.WrapT, img, height, texcoord[1], &j0, &j1, &b);
+   linear_texel_locations(tObj->Sampler.WrapR, img, depth, texcoord[2],  &k0, &k1, &c);
+
+   if (img->Border) {
+      i0 += img->Border;
+      i1 += img->Border;
+      j0 += img->Border;
+      j1 += img->Border;
+      k0 += img->Border;
+      k1 += img->Border;
+   }
+   else {
+      /* check if sampling texture border color */
+      if (i0 < 0 || i0 >= width)   useBorderColor |= I0BIT;
+      if (i1 < 0 || i1 >= width)   useBorderColor |= I1BIT;
+      if (j0 < 0 || j0 >= height)  useBorderColor |= J0BIT;
+      if (j1 < 0 || j1 >= height)  useBorderColor |= J1BIT;
+      if (k0 < 0 || k0 >= depth)   useBorderColor |= K0BIT;
+      if (k1 < 0 || k1 >= depth)   useBorderColor |= K1BIT;
+   }
+
+   /* Fetch texels */
+   if (useBorderColor & (I0BIT | J0BIT | K0BIT)) {
+      get_border_color(tObj, img, t000);
+   }
+   else {
+      img->FetchTexelf(img, i0, j0, k0, t000);
+   }
+   if (useBorderColor & (I1BIT | J0BIT | K0BIT)) {
+      get_border_color(tObj, img, t100);
+   }
+   else {
+      img->FetchTexelf(img, i1, j0, k0, t100);
+   }
+   if (useBorderColor & (I0BIT | J1BIT | K0BIT)) {
+      get_border_color(tObj, img, t010);
+   }
+   else {
+      img->FetchTexelf(img, i0, j1, k0, t010);
+   }
+   if (useBorderColor & (I1BIT | J1BIT | K0BIT)) {
+      get_border_color(tObj, img, t110);
+   }
+   else {
+      img->FetchTexelf(img, i1, j1, k0, t110);
+   }
+
+   if (useBorderColor & (I0BIT | J0BIT | K1BIT)) {
+      get_border_color(tObj, img, t001);
+   }
+   else {
+      img->FetchTexelf(img, i0, j0, k1, t001);
+   }
+   if (useBorderColor & (I1BIT | J0BIT | K1BIT)) {
+      get_border_color(tObj, img, t101);
+   }
+   else {
+      img->FetchTexelf(img, i1, j0, k1, t101);
+   }
+   if (useBorderColor & (I0BIT | J1BIT | K1BIT)) {
+      get_border_color(tObj, img, t011);
+   }
+   else {
+      img->FetchTexelf(img, i0, j1, k1, t011);
+   }
+   if (useBorderColor & (I1BIT | J1BIT | K1BIT)) {
+      get_border_color(tObj, img, t111);
+   }
+   else {
+      img->FetchTexelf(img, i1, j1, k1, t111);
+   }
+
+   /* trilinear interpolation of samples */
+   lerp_rgba_3d(rgba, a, b, c, t000, t100, t010, t110, t001, t101, t011, t111);
+}
+
+
+static void
+sample_3d_nearest_mipmap_nearest(struct gl_context *ctx,
+                                 const struct gl_texture_object *tObj,
+                                 GLuint n, const GLfloat texcoord[][4],
+                                 const GLfloat lambda[], GLfloat rgba[][4] )
+{
+   GLuint i;
+   for (i = 0; i < n; i++) {
+      GLint level = nearest_mipmap_level(tObj, lambda[i]);
+      sample_3d_nearest(ctx, tObj, tObj->Image[0][level], texcoord[i], rgba[i]);
+   }
+}
+
+
+static void
+sample_3d_linear_mipmap_nearest(struct gl_context *ctx,
+                                const struct gl_texture_object *tObj,
+                                GLuint n, const GLfloat texcoord[][4],
+                                const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   ASSERT(lambda != NULL);
+   for (i = 0; i < n; i++) {
+      GLint level = nearest_mipmap_level(tObj, lambda[i]);
+      sample_3d_linear(ctx, tObj, tObj->Image[0][level], texcoord[i], rgba[i]);
+   }
+}
+
+
+static void
+sample_3d_nearest_mipmap_linear(struct gl_context *ctx,
+                                const struct gl_texture_object *tObj,
+                                GLuint n, const GLfloat texcoord[][4],
+                                const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   ASSERT(lambda != NULL);
+   for (i = 0; i < n; i++) {
+      GLint level = linear_mipmap_level(tObj, lambda[i]);
+      if (level >= tObj->_MaxLevel) {
+         sample_3d_nearest(ctx, tObj, tObj->Image[0][tObj->_MaxLevel],
+                           texcoord[i], rgba[i]);
+      }
+      else {
+         GLfloat t0[4], t1[4];  /* texels */
+         const GLfloat f = FRAC(lambda[i]);
+         sample_3d_nearest(ctx, tObj, tObj->Image[0][level  ], texcoord[i], t0);
+         sample_3d_nearest(ctx, tObj, tObj->Image[0][level+1], texcoord[i], t1);
+         lerp_rgba(rgba[i], f, t0, t1);
+      }
+   }
+}
+
+
+static void
+sample_3d_linear_mipmap_linear(struct gl_context *ctx,
+                               const struct gl_texture_object *tObj,
+                               GLuint n, const GLfloat texcoord[][4],
+                               const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   ASSERT(lambda != NULL);
+   for (i = 0; i < n; i++) {
+      GLint level = linear_mipmap_level(tObj, lambda[i]);
+      if (level >= tObj->_MaxLevel) {
+         sample_3d_linear(ctx, tObj, tObj->Image[0][tObj->_MaxLevel],
+                          texcoord[i], rgba[i]);
+      }
+      else {
+         GLfloat t0[4], t1[4];  /* texels */
+         const GLfloat f = FRAC(lambda[i]);
+         sample_3d_linear(ctx, tObj, tObj->Image[0][level  ], texcoord[i], t0);
+         sample_3d_linear(ctx, tObj, tObj->Image[0][level+1], texcoord[i], t1);
+         lerp_rgba(rgba[i], f, t0, t1);
+      }
+   }
+}
+
+
+/** Sample 3D texture, nearest filtering for both min/magnification */
+static void
+sample_nearest_3d(struct gl_context *ctx,
+                  const struct gl_texture_object *tObj, GLuint n,
+                  const GLfloat texcoords[][4], const GLfloat lambda[],
+                  GLfloat rgba[][4])
+{
+   GLuint i;
+   struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
+   (void) lambda;
+   for (i = 0; i < n; i++) {
+      sample_3d_nearest(ctx, tObj, image, texcoords[i], rgba[i]);
+   }
+}
+
+
+/** Sample 3D texture, linear filtering for both min/magnification */
+static void
+sample_linear_3d(struct gl_context *ctx,
+                 const struct gl_texture_object *tObj, GLuint n,
+                 const GLfloat texcoords[][4],
+		 const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
+   (void) lambda;
+   for (i = 0; i < n; i++) {
+      sample_3d_linear(ctx, tObj, image, texcoords[i], rgba[i]);
+   }
+}
+
+
+/** Sample 3D texture, using lambda to choose between min/magnification */
+static void
+sample_lambda_3d(struct gl_context *ctx,
+                 const struct gl_texture_object *tObj, GLuint n,
+                 const GLfloat texcoords[][4], const GLfloat lambda[],
+                 GLfloat rgba[][4])
+{
+   GLuint minStart, minEnd;  /* texels with minification */
+   GLuint magStart, magEnd;  /* texels with magnification */
+   GLuint i;
+
+   ASSERT(lambda != NULL);
+   compute_min_mag_ranges(tObj, n, lambda,
+                          &minStart, &minEnd, &magStart, &magEnd);
+
+   if (minStart < minEnd) {
+      /* do the minified texels */
+      GLuint m = minEnd - minStart;
+      switch (tObj->Sampler.MinFilter) {
+      case GL_NEAREST:
+         for (i = minStart; i < minEnd; i++)
+            sample_3d_nearest(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
+                              texcoords[i], rgba[i]);
+         break;
+      case GL_LINEAR:
+         for (i = minStart; i < minEnd; i++)
+            sample_3d_linear(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
+                             texcoords[i], rgba[i]);
+         break;
+      case GL_NEAREST_MIPMAP_NEAREST:
+         sample_3d_nearest_mipmap_nearest(ctx, tObj, m, texcoords + minStart,
+                                          lambda + minStart, rgba + minStart);
+         break;
+      case GL_LINEAR_MIPMAP_NEAREST:
+         sample_3d_linear_mipmap_nearest(ctx, tObj, m, texcoords + minStart,
+                                         lambda + minStart, rgba + minStart);
+         break;
+      case GL_NEAREST_MIPMAP_LINEAR:
+         sample_3d_nearest_mipmap_linear(ctx, tObj, m, texcoords + minStart,
+                                         lambda + minStart, rgba + minStart);
+         break;
+      case GL_LINEAR_MIPMAP_LINEAR:
+         sample_3d_linear_mipmap_linear(ctx, tObj, m, texcoords + minStart,
+                                        lambda + minStart, rgba + minStart);
+         break;
+      default:
+         _mesa_problem(ctx, "Bad min filter in sample_3d_texture");
+         return;
+      }
+   }
+
+   if (magStart < magEnd) {
+      /* do the magnified texels */
+      switch (tObj->Sampler.MagFilter) {
+      case GL_NEAREST:
+         for (i = magStart; i < magEnd; i++)
+            sample_3d_nearest(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
+                              texcoords[i], rgba[i]);
+         break;
+      case GL_LINEAR:
+         for (i = magStart; i < magEnd; i++)
+            sample_3d_linear(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
+                             texcoords[i], rgba[i]);
+         break;
+      default:
+         _mesa_problem(ctx, "Bad mag filter in sample_3d_texture");
+         return;
+      }
+   }
+}
+
+
+/**********************************************************************/
+/*                Texture Cube Map Sampling Functions                 */
+/**********************************************************************/
+
+/**
+ * Choose one of six sides of a texture cube map given the texture
+ * coord (rx,ry,rz).  Return pointer to corresponding array of texture
+ * images.
+ */
+static const struct gl_texture_image **
+choose_cube_face(const struct gl_texture_object *texObj,
+                 const GLfloat texcoord[4], GLfloat newCoord[4])
+{
+   /*
+      major axis
+      direction     target                             sc     tc    ma
+      ----------    -------------------------------    ---    ---   ---
+       +rx          TEXTURE_CUBE_MAP_POSITIVE_X_EXT    -rz    -ry   rx
+       -rx          TEXTURE_CUBE_MAP_NEGATIVE_X_EXT    +rz    -ry   rx
+       +ry          TEXTURE_CUBE_MAP_POSITIVE_Y_EXT    +rx    +rz   ry
+       -ry          TEXTURE_CUBE_MAP_NEGATIVE_Y_EXT    +rx    -rz   ry
+       +rz          TEXTURE_CUBE_MAP_POSITIVE_Z_EXT    +rx    -ry   rz
+       -rz          TEXTURE_CUBE_MAP_NEGATIVE_Z_EXT    -rx    -ry   rz
+   */
+   const GLfloat rx = texcoord[0];
+   const GLfloat ry = texcoord[1];
+   const GLfloat rz = texcoord[2];
+   const GLfloat arx = FABSF(rx), ary = FABSF(ry), arz = FABSF(rz);
+   GLuint face;
+   GLfloat sc, tc, ma;
+
+   if (arx >= ary && arx >= arz) {
+      if (rx >= 0.0F) {
+         face = FACE_POS_X;
+         sc = -rz;
+         tc = -ry;
+         ma = arx;
+      }
+      else {
+         face = FACE_NEG_X;
+         sc = rz;
+         tc = -ry;
+         ma = arx;
+      }
+   }
+   else if (ary >= arx && ary >= arz) {
+      if (ry >= 0.0F) {
+         face = FACE_POS_Y;
+         sc = rx;
+         tc = rz;
+         ma = ary;
+      }
+      else {
+         face = FACE_NEG_Y;
+         sc = rx;
+         tc = -rz;
+         ma = ary;
+      }
+   }
+   else {
+      if (rz > 0.0F) {
+         face = FACE_POS_Z;
+         sc = rx;
+         tc = -ry;
+         ma = arz;
+      }
+      else {
+         face = FACE_NEG_Z;
+         sc = -rx;
+         tc = -ry;
+         ma = arz;
+      }
+   }
+
+   { 
+      const float ima = 1.0F / ma;
+      newCoord[0] = ( sc * ima + 1.0F ) * 0.5F;
+      newCoord[1] = ( tc * ima + 1.0F ) * 0.5F;
+   }
+
+   return (const struct gl_texture_image **) texObj->Image[face];
+}
+
+
+static void
+sample_nearest_cube(struct gl_context *ctx,
+		    const struct gl_texture_object *tObj, GLuint n,
+                    const GLfloat texcoords[][4], const GLfloat lambda[],
+                    GLfloat rgba[][4])
+{
+   GLuint i;
+   (void) lambda;
+   for (i = 0; i < n; i++) {
+      const struct gl_texture_image **images;
+      GLfloat newCoord[4];
+      images = choose_cube_face(tObj, texcoords[i], newCoord);
+      sample_2d_nearest(ctx, tObj, images[tObj->BaseLevel],
+                        newCoord, rgba[i]);
+   }
+}
+
+
+static void
+sample_linear_cube(struct gl_context *ctx,
+		   const struct gl_texture_object *tObj, GLuint n,
+                   const GLfloat texcoords[][4],
+		   const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   (void) lambda;
+   for (i = 0; i < n; i++) {
+      const struct gl_texture_image **images;
+      GLfloat newCoord[4];
+      images = choose_cube_face(tObj, texcoords[i], newCoord);
+      sample_2d_linear(ctx, tObj, images[tObj->BaseLevel],
+                       newCoord, rgba[i]);
+   }
+}
+
+
+static void
+sample_cube_nearest_mipmap_nearest(struct gl_context *ctx,
+                                   const struct gl_texture_object *tObj,
+                                   GLuint n, const GLfloat texcoord[][4],
+                                   const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   ASSERT(lambda != NULL);
+   for (i = 0; i < n; i++) {
+      const struct gl_texture_image **images;
+      GLfloat newCoord[4];
+      GLint level;
+      images = choose_cube_face(tObj, texcoord[i], newCoord);
+
+      /* XXX we actually need to recompute lambda here based on the newCoords.
+       * But we would need the texcoords of adjacent fragments to compute that
+       * properly, and we don't have those here.
+       * For now, do an approximation:  subtracting 1 from the chosen mipmap
+       * level seems to work in some test cases.
+       * The same adjustment is done in the next few functions.
+      */
+      level = nearest_mipmap_level(tObj, lambda[i]);
+      level = MAX2(level - 1, 0);
+
+      sample_2d_nearest(ctx, tObj, images[level], newCoord, rgba[i]);
+   }
+}
+
+
+static void
+sample_cube_linear_mipmap_nearest(struct gl_context *ctx,
+                                  const struct gl_texture_object *tObj,
+                                  GLuint n, const GLfloat texcoord[][4],
+                                  const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   ASSERT(lambda != NULL);
+   for (i = 0; i < n; i++) {
+      const struct gl_texture_image **images;
+      GLfloat newCoord[4];
+      GLint level = nearest_mipmap_level(tObj, lambda[i]);
+      level = MAX2(level - 1, 0); /* see comment above */
+      images = choose_cube_face(tObj, texcoord[i], newCoord);
+      sample_2d_linear(ctx, tObj, images[level], newCoord, rgba[i]);
+   }
+}
+
+
+static void
+sample_cube_nearest_mipmap_linear(struct gl_context *ctx,
+                                  const struct gl_texture_object *tObj,
+                                  GLuint n, const GLfloat texcoord[][4],
+                                  const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   ASSERT(lambda != NULL);
+   for (i = 0; i < n; i++) {
+      const struct gl_texture_image **images;
+      GLfloat newCoord[4];
+      GLint level = linear_mipmap_level(tObj, lambda[i]);
+      level = MAX2(level - 1, 0); /* see comment above */
+      images = choose_cube_face(tObj, texcoord[i], newCoord);
+      if (level >= tObj->_MaxLevel) {
+         sample_2d_nearest(ctx, tObj, images[tObj->_MaxLevel],
+                           newCoord, rgba[i]);
+      }
+      else {
+         GLfloat t0[4], t1[4];  /* texels */
+         const GLfloat f = FRAC(lambda[i]);
+         sample_2d_nearest(ctx, tObj, images[level  ], newCoord, t0);
+         sample_2d_nearest(ctx, tObj, images[level+1], newCoord, t1);
+         lerp_rgba(rgba[i], f, t0, t1);
+      }
+   }
+}
+
+
+static void
+sample_cube_linear_mipmap_linear(struct gl_context *ctx,
+                                 const struct gl_texture_object *tObj,
+                                 GLuint n, const GLfloat texcoord[][4],
+                                 const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   ASSERT(lambda != NULL);
+   for (i = 0; i < n; i++) {
+      const struct gl_texture_image **images;
+      GLfloat newCoord[4];
+      GLint level = linear_mipmap_level(tObj, lambda[i]);
+      level = MAX2(level - 1, 0); /* see comment above */
+      images = choose_cube_face(tObj, texcoord[i], newCoord);
+      if (level >= tObj->_MaxLevel) {
+         sample_2d_linear(ctx, tObj, images[tObj->_MaxLevel],
+                          newCoord, rgba[i]);
+      }
+      else {
+         GLfloat t0[4], t1[4];
+         const GLfloat f = FRAC(lambda[i]);
+         sample_2d_linear(ctx, tObj, images[level  ], newCoord, t0);
+         sample_2d_linear(ctx, tObj, images[level+1], newCoord, t1);
+         lerp_rgba(rgba[i], f, t0, t1);
+      }
+   }
+}
+
+
+/** Sample cube texture, using lambda to choose between min/magnification */
+static void
+sample_lambda_cube(struct gl_context *ctx,
+		   const struct gl_texture_object *tObj, GLuint n,
+		   const GLfloat texcoords[][4], const GLfloat lambda[],
+		   GLfloat rgba[][4])
+{
+   GLuint minStart, minEnd;  /* texels with minification */
+   GLuint magStart, magEnd;  /* texels with magnification */
+
+   ASSERT(lambda != NULL);
+   compute_min_mag_ranges(tObj, n, lambda,
+                          &minStart, &minEnd, &magStart, &magEnd);
+
+   if (minStart < minEnd) {
+      /* do the minified texels */
+      const GLuint m = minEnd - minStart;
+      switch (tObj->Sampler.MinFilter) {
+      case GL_NEAREST:
+         sample_nearest_cube(ctx, tObj, m, texcoords + minStart,
+                             lambda + minStart, rgba + minStart);
+         break;
+      case GL_LINEAR:
+         sample_linear_cube(ctx, tObj, m, texcoords + minStart,
+                            lambda + minStart, rgba + minStart);
+         break;
+      case GL_NEAREST_MIPMAP_NEAREST:
+         sample_cube_nearest_mipmap_nearest(ctx, tObj, m,
+                                            texcoords + minStart,
+                                           lambda + minStart, rgba + minStart);
+         break;
+      case GL_LINEAR_MIPMAP_NEAREST:
+         sample_cube_linear_mipmap_nearest(ctx, tObj, m,
+                                           texcoords + minStart,
+                                           lambda + minStart, rgba + minStart);
+         break;
+      case GL_NEAREST_MIPMAP_LINEAR:
+         sample_cube_nearest_mipmap_linear(ctx, tObj, m,
+                                           texcoords + minStart,
+                                           lambda + minStart, rgba + minStart);
+         break;
+      case GL_LINEAR_MIPMAP_LINEAR:
+         sample_cube_linear_mipmap_linear(ctx, tObj, m,
+                                          texcoords + minStart,
+                                          lambda + minStart, rgba + minStart);
+         break;
+      default:
+         _mesa_problem(ctx, "Bad min filter in sample_lambda_cube");
+      }
+   }
+
+   if (magStart < magEnd) {
+      /* do the magnified texels */
+      const GLuint m = magEnd - magStart;
+      switch (tObj->Sampler.MagFilter) {
+      case GL_NEAREST:
+         sample_nearest_cube(ctx, tObj, m, texcoords + magStart,
+                             lambda + magStart, rgba + magStart);
+         break;
+      case GL_LINEAR:
+         sample_linear_cube(ctx, tObj, m, texcoords + magStart,
+                            lambda + magStart, rgba + magStart);
+         break;
+      default:
+         _mesa_problem(ctx, "Bad mag filter in sample_lambda_cube");
+      }
+   }
+}
+
+
+/**********************************************************************/
+/*               Texture Rectangle Sampling Functions                 */
+/**********************************************************************/
+
+
+static void
+sample_nearest_rect(struct gl_context *ctx,
+		    const struct gl_texture_object *tObj, GLuint n,
+                    const GLfloat texcoords[][4], const GLfloat lambda[],
+                    GLfloat rgba[][4])
+{
+   const struct gl_texture_image *img = tObj->Image[0][0];
+   const GLint width = img->Width;
+   const GLint height = img->Height;
+   GLuint i;
+
+   (void) ctx;
+   (void) lambda;
+
+   ASSERT(tObj->Sampler.WrapS == GL_CLAMP ||
+          tObj->Sampler.WrapS == GL_CLAMP_TO_EDGE ||
+          tObj->Sampler.WrapS == GL_CLAMP_TO_BORDER);
+   ASSERT(tObj->Sampler.WrapT == GL_CLAMP ||
+          tObj->Sampler.WrapT == GL_CLAMP_TO_EDGE ||
+          tObj->Sampler.WrapT == GL_CLAMP_TO_BORDER);
+   ASSERT(img->_BaseFormat != GL_COLOR_INDEX);
+
+   for (i = 0; i < n; i++) {
+      GLint row, col;
+      col = clamp_rect_coord_nearest(tObj->Sampler.WrapS, texcoords[i][0], width);
+      row = clamp_rect_coord_nearest(tObj->Sampler.WrapT, texcoords[i][1], height);
+      if (col < 0 || col >= width || row < 0 || row >= height)
+         get_border_color(tObj, img, rgba[i]);
+      else
+         img->FetchTexelf(img, col, row, 0, rgba[i]);
+   }
+}
+
+
+static void
+sample_linear_rect(struct gl_context *ctx,
+		   const struct gl_texture_object *tObj, GLuint n,
+                   const GLfloat texcoords[][4],
+		   const GLfloat lambda[], GLfloat rgba[][4])
+{
+   const struct gl_texture_image *img = tObj->Image[0][0];
+   const GLint width = img->Width;
+   const GLint height = img->Height;
+   GLuint i;
+
+   (void) ctx;
+   (void) lambda;
+
+   ASSERT(tObj->Sampler.WrapS == GL_CLAMP ||
+          tObj->Sampler.WrapS == GL_CLAMP_TO_EDGE ||
+          tObj->Sampler.WrapS == GL_CLAMP_TO_BORDER);
+   ASSERT(tObj->Sampler.WrapT == GL_CLAMP ||
+          tObj->Sampler.WrapT == GL_CLAMP_TO_EDGE ||
+          tObj->Sampler.WrapT == GL_CLAMP_TO_BORDER);
+   ASSERT(img->_BaseFormat != GL_COLOR_INDEX);
+
+   for (i = 0; i < n; i++) {
+      GLint i0, j0, i1, j1;
+      GLfloat t00[4], t01[4], t10[4], t11[4];
+      GLfloat a, b;
+      GLbitfield useBorderColor = 0x0;
+
+      clamp_rect_coord_linear(tObj->Sampler.WrapS, texcoords[i][0], width,
+                              &i0, &i1, &a);
+      clamp_rect_coord_linear(tObj->Sampler.WrapT, texcoords[i][1], height,
+                              &j0, &j1, &b);
+
+      /* compute integer rows/columns */
+      if (i0 < 0 || i0 >= width)   useBorderColor |= I0BIT;
+      if (i1 < 0 || i1 >= width)   useBorderColor |= I1BIT;
+      if (j0 < 0 || j0 >= height)  useBorderColor |= J0BIT;
+      if (j1 < 0 || j1 >= height)  useBorderColor |= J1BIT;
+
+      /* get four texel samples */
+      if (useBorderColor & (I0BIT | J0BIT))
+         get_border_color(tObj, img, t00);
+      else
+         img->FetchTexelf(img, i0, j0, 0, t00);
+
+      if (useBorderColor & (I1BIT | J0BIT))
+         get_border_color(tObj, img, t10);
+      else
+         img->FetchTexelf(img, i1, j0, 0, t10);
+
+      if (useBorderColor & (I0BIT | J1BIT))
+         get_border_color(tObj, img, t01);
+      else
+         img->FetchTexelf(img, i0, j1, 0, t01);
+
+      if (useBorderColor & (I1BIT | J1BIT))
+         get_border_color(tObj, img, t11);
+      else
+         img->FetchTexelf(img, i1, j1, 0, t11);
+
+      lerp_rgba_2d(rgba[i], a, b, t00, t10, t01, t11);
+   }
+}
+
+
+/** Sample Rect texture, using lambda to choose between min/magnification */
+static void
+sample_lambda_rect(struct gl_context *ctx,
+		   const struct gl_texture_object *tObj, GLuint n,
+		   const GLfloat texcoords[][4], const GLfloat lambda[],
+		   GLfloat rgba[][4])
+{
+   GLuint minStart, minEnd, magStart, magEnd;
+
+   /* We only need lambda to decide between minification and magnification.
+    * There is no mipmapping with rectangular textures.
+    */
+   compute_min_mag_ranges(tObj, n, lambda,
+                          &minStart, &minEnd, &magStart, &magEnd);
+
+   if (minStart < minEnd) {
+      if (tObj->Sampler.MinFilter == GL_NEAREST) {
+         sample_nearest_rect(ctx, tObj, minEnd - minStart,
+                             texcoords + minStart, NULL, rgba + minStart);
+      }
+      else {
+         sample_linear_rect(ctx, tObj, minEnd - minStart,
+                            texcoords + minStart, NULL, rgba + minStart);
+      }
+   }
+   if (magStart < magEnd) {
+      if (tObj->Sampler.MagFilter == GL_NEAREST) {
+         sample_nearest_rect(ctx, tObj, magEnd - magStart,
+                             texcoords + magStart, NULL, rgba + magStart);
+      }
+      else {
+         sample_linear_rect(ctx, tObj, magEnd - magStart,
+                            texcoords + magStart, NULL, rgba + magStart);
+      }
+   }
+}
+
+
+/**********************************************************************/
+/*                2D Texture Array Sampling Functions                 */
+/**********************************************************************/
+
+/**
+ * Return the texture sample for coordinate (s,t,r) using GL_NEAREST filter.
+ */
+static void
+sample_2d_array_nearest(struct gl_context *ctx,
+                        const struct gl_texture_object *tObj,
+                        const struct gl_texture_image *img,
+                        const GLfloat texcoord[4],
+                        GLfloat rgba[4])
+{
+   const GLint width = img->Width2;     /* without border, power of two */
+   const GLint height = img->Height2;   /* without border, power of two */
+   const GLint depth = img->Depth;
+   GLint i, j;
+   GLint array;
+   (void) ctx;
+
+   i = nearest_texel_location(tObj->Sampler.WrapS, img, width, texcoord[0]);
+   j = nearest_texel_location(tObj->Sampler.WrapT, img, height, texcoord[1]);
+   array = tex_array_slice(texcoord[2], depth);
+
+   if (i < 0 || i >= (GLint) img->Width ||
+       j < 0 || j >= (GLint) img->Height ||
+       array < 0 || array >= (GLint) img->Depth) {
+      /* Need this test for GL_CLAMP_TO_BORDER mode */
+      get_border_color(tObj, img, rgba);
+   }
+   else {
+      img->FetchTexelf(img, i, j, array, rgba);
+   }
+}
+
+
+/**
+ * Return the texture sample for coordinate (s,t,r) using GL_LINEAR filter.
+ */
+static void
+sample_2d_array_linear(struct gl_context *ctx,
+                       const struct gl_texture_object *tObj,
+                       const struct gl_texture_image *img,
+                       const GLfloat texcoord[4],
+                       GLfloat rgba[4])
+{
+   const GLint width = img->Width2;
+   const GLint height = img->Height2;
+   const GLint depth = img->Depth;
+   GLint i0, j0, i1, j1;
+   GLint array;
+   GLbitfield useBorderColor = 0x0;
+   GLfloat a, b;
+   GLfloat t00[4], t01[4], t10[4], t11[4];
+
+   linear_texel_locations(tObj->Sampler.WrapS, img, width,  texcoord[0], &i0, &i1, &a);
+   linear_texel_locations(tObj->Sampler.WrapT, img, height, texcoord[1], &j0, &j1, &b);
+   array = tex_array_slice(texcoord[2], depth);
+
+   if (array < 0 || array >= depth) {
+      COPY_4V(rgba, tObj->Sampler.BorderColor.f);
+   }
+   else {
+      if (img->Border) {
+	 i0 += img->Border;
+	 i1 += img->Border;
+	 j0 += img->Border;
+	 j1 += img->Border;
+      }
+      else {
+	 /* check if sampling texture border color */
+	 if (i0 < 0 || i0 >= width)   useBorderColor |= I0BIT;
+	 if (i1 < 0 || i1 >= width)   useBorderColor |= I1BIT;
+	 if (j0 < 0 || j0 >= height)  useBorderColor |= J0BIT;
+	 if (j1 < 0 || j1 >= height)  useBorderColor |= J1BIT;
+      }
+
+      /* Fetch texels */
+      if (useBorderColor & (I0BIT | J0BIT)) {
+         get_border_color(tObj, img, t00);
+      }
+      else {
+	 img->FetchTexelf(img, i0, j0, array, t00);
+      }
+      if (useBorderColor & (I1BIT | J0BIT)) {
+         get_border_color(tObj, img, t10);
+      }
+      else {
+	 img->FetchTexelf(img, i1, j0, array, t10);
+      }
+      if (useBorderColor & (I0BIT | J1BIT)) {
+         get_border_color(tObj, img, t01);
+      }
+      else {
+	 img->FetchTexelf(img, i0, j1, array, t01);
+      }
+      if (useBorderColor & (I1BIT | J1BIT)) {
+         get_border_color(tObj, img, t11);
+      }
+      else {
+	 img->FetchTexelf(img, i1, j1, array, t11);
+      }
+      
+      /* trilinear interpolation of samples */
+      lerp_rgba_2d(rgba, a, b, t00, t10, t01, t11);
+   }
+}
+
+
+static void
+sample_2d_array_nearest_mipmap_nearest(struct gl_context *ctx,
+                                       const struct gl_texture_object *tObj,
+                                       GLuint n, const GLfloat texcoord[][4],
+                                       const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   for (i = 0; i < n; i++) {
+      GLint level = nearest_mipmap_level(tObj, lambda[i]);
+      sample_2d_array_nearest(ctx, tObj, tObj->Image[0][level], texcoord[i],
+                              rgba[i]);
+   }
+}
+
+
+static void
+sample_2d_array_linear_mipmap_nearest(struct gl_context *ctx,
+                                      const struct gl_texture_object *tObj,
+                                      GLuint n, const GLfloat texcoord[][4],
+                                      const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   ASSERT(lambda != NULL);
+   for (i = 0; i < n; i++) {
+      GLint level = nearest_mipmap_level(tObj, lambda[i]);
+      sample_2d_array_linear(ctx, tObj, tObj->Image[0][level],
+                             texcoord[i], rgba[i]);
+   }
+}
+
+
+static void
+sample_2d_array_nearest_mipmap_linear(struct gl_context *ctx,
+                                      const struct gl_texture_object *tObj,
+                                      GLuint n, const GLfloat texcoord[][4],
+                                      const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   ASSERT(lambda != NULL);
+   for (i = 0; i < n; i++) {
+      GLint level = linear_mipmap_level(tObj, lambda[i]);
+      if (level >= tObj->_MaxLevel) {
+         sample_2d_array_nearest(ctx, tObj, tObj->Image[0][tObj->_MaxLevel],
+                                 texcoord[i], rgba[i]);
+      }
+      else {
+         GLfloat t0[4], t1[4];  /* texels */
+         const GLfloat f = FRAC(lambda[i]);
+         sample_2d_array_nearest(ctx, tObj, tObj->Image[0][level  ],
+                                 texcoord[i], t0);
+         sample_2d_array_nearest(ctx, tObj, tObj->Image[0][level+1],
+                                 texcoord[i], t1);
+         lerp_rgba(rgba[i], f, t0, t1);
+      }
+   }
+}
+
+
+static void
+sample_2d_array_linear_mipmap_linear(struct gl_context *ctx,
+                                     const struct gl_texture_object *tObj,
+                                     GLuint n, const GLfloat texcoord[][4],
+                                     const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   ASSERT(lambda != NULL);
+   for (i = 0; i < n; i++) {
+      GLint level = linear_mipmap_level(tObj, lambda[i]);
+      if (level >= tObj->_MaxLevel) {
+         sample_2d_array_linear(ctx, tObj, tObj->Image[0][tObj->_MaxLevel],
+                          texcoord[i], rgba[i]);
+      }
+      else {
+         GLfloat t0[4], t1[4];  /* texels */
+         const GLfloat f = FRAC(lambda[i]);
+         sample_2d_array_linear(ctx, tObj, tObj->Image[0][level  ],
+                                texcoord[i], t0);
+         sample_2d_array_linear(ctx, tObj, tObj->Image[0][level+1],
+                                texcoord[i], t1);
+         lerp_rgba(rgba[i], f, t0, t1);
+      }
+   }
+}
+
+
+/** Sample 2D Array texture, nearest filtering for both min/magnification */
+static void
+sample_nearest_2d_array(struct gl_context *ctx,
+                        const struct gl_texture_object *tObj, GLuint n,
+                        const GLfloat texcoords[][4], const GLfloat lambda[],
+                        GLfloat rgba[][4])
+{
+   GLuint i;
+   struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
+   (void) lambda;
+   for (i = 0; i < n; i++) {
+      sample_2d_array_nearest(ctx, tObj, image, texcoords[i], rgba[i]);
+   }
+}
+
+
+
+/** Sample 2D Array texture, linear filtering for both min/magnification */
+static void
+sample_linear_2d_array(struct gl_context *ctx,
+                       const struct gl_texture_object *tObj, GLuint n,
+                       const GLfloat texcoords[][4],
+                       const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
+   (void) lambda;
+   for (i = 0; i < n; i++) {
+      sample_2d_array_linear(ctx, tObj, image, texcoords[i], rgba[i]);
+   }
+}
+
+
+/** Sample 2D Array texture, using lambda to choose between min/magnification */
+static void
+sample_lambda_2d_array(struct gl_context *ctx,
+                       const struct gl_texture_object *tObj, GLuint n,
+                       const GLfloat texcoords[][4], const GLfloat lambda[],
+                       GLfloat rgba[][4])
+{
+   GLuint minStart, minEnd;  /* texels with minification */
+   GLuint magStart, magEnd;  /* texels with magnification */
+   GLuint i;
+
+   ASSERT(lambda != NULL);
+   compute_min_mag_ranges(tObj, n, lambda,
+                          &minStart, &minEnd, &magStart, &magEnd);
+
+   if (minStart < minEnd) {
+      /* do the minified texels */
+      GLuint m = minEnd - minStart;
+      switch (tObj->Sampler.MinFilter) {
+      case GL_NEAREST:
+         for (i = minStart; i < minEnd; i++)
+            sample_2d_array_nearest(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
+                                    texcoords[i], rgba[i]);
+         break;
+      case GL_LINEAR:
+         for (i = minStart; i < minEnd; i++)
+            sample_2d_array_linear(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
+                                   texcoords[i], rgba[i]);
+         break;
+      case GL_NEAREST_MIPMAP_NEAREST:
+         sample_2d_array_nearest_mipmap_nearest(ctx, tObj, m,
+                                                texcoords + minStart,
+                                                lambda + minStart,
+                                                rgba + minStart);
+         break;
+      case GL_LINEAR_MIPMAP_NEAREST:
+         sample_2d_array_linear_mipmap_nearest(ctx, tObj, m, 
+                                               texcoords + minStart,
+                                               lambda + minStart,
+                                               rgba + minStart);
+         break;
+      case GL_NEAREST_MIPMAP_LINEAR:
+         sample_2d_array_nearest_mipmap_linear(ctx, tObj, m,
+                                               texcoords + minStart,
+                                               lambda + minStart,
+                                               rgba + minStart);
+         break;
+      case GL_LINEAR_MIPMAP_LINEAR:
+         sample_2d_array_linear_mipmap_linear(ctx, tObj, m, 
+                                              texcoords + minStart,
+                                              lambda + minStart, 
+                                              rgba + minStart);
+         break;
+      default:
+         _mesa_problem(ctx, "Bad min filter in sample_2d_array_texture");
+         return;
+      }
+   }
+
+   if (magStart < magEnd) {
+      /* do the magnified texels */
+      switch (tObj->Sampler.MagFilter) {
+      case GL_NEAREST:
+         for (i = magStart; i < magEnd; i++)
+            sample_2d_array_nearest(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
+                              texcoords[i], rgba[i]);
+         break;
+      case GL_LINEAR:
+         for (i = magStart; i < magEnd; i++)
+            sample_2d_array_linear(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
+                                   texcoords[i], rgba[i]);
+         break;
+      default:
+         _mesa_problem(ctx, "Bad mag filter in sample_2d_array_texture");
+         return;
+      }
+   }
+}
+
+
+
+
+/**********************************************************************/
+/*                1D Texture Array Sampling Functions                 */
+/**********************************************************************/
+
+/**
+ * Return the texture sample for coordinate (s,t,r) using GL_NEAREST filter.
+ */
+static void
+sample_1d_array_nearest(struct gl_context *ctx,
+                        const struct gl_texture_object *tObj,
+                        const struct gl_texture_image *img,
+                        const GLfloat texcoord[4],
+                        GLfloat rgba[4])
+{
+   const GLint width = img->Width2;     /* without border, power of two */
+   const GLint height = img->Height;
+   GLint i;
+   GLint array;
+   (void) ctx;
+
+   i = nearest_texel_location(tObj->Sampler.WrapS, img, width, texcoord[0]);
+   array = tex_array_slice(texcoord[1], height);
+
+   if (i < 0 || i >= (GLint) img->Width ||
+       array < 0 || array >= (GLint) img->Height) {
+      /* Need this test for GL_CLAMP_TO_BORDER mode */
+      get_border_color(tObj, img, rgba);
+   }
+   else {
+      img->FetchTexelf(img, i, array, 0, rgba);
+   }
+}
+
+
+/**
+ * Return the texture sample for coordinate (s,t,r) using GL_LINEAR filter.
+ */
+static void
+sample_1d_array_linear(struct gl_context *ctx,
+                       const struct gl_texture_object *tObj,
+                       const struct gl_texture_image *img,
+                       const GLfloat texcoord[4],
+                       GLfloat rgba[4])
+{
+   const GLint width = img->Width2;
+   const GLint height = img->Height;
+   GLint i0, i1;
+   GLint array;
+   GLbitfield useBorderColor = 0x0;
+   GLfloat a;
+   GLfloat t0[4], t1[4];
+
+   linear_texel_locations(tObj->Sampler.WrapS, img, width, texcoord[0], &i0, &i1, &a);
+   array = tex_array_slice(texcoord[1], height);
+
+   if (img->Border) {
+      i0 += img->Border;
+      i1 += img->Border;
+   }
+   else {
+      /* check if sampling texture border color */
+      if (i0 < 0 || i0 >= width)   useBorderColor |= I0BIT;
+      if (i1 < 0 || i1 >= width)   useBorderColor |= I1BIT;
+   }
+
+   if (array < 0 || array >= height)   useBorderColor |= K0BIT;
+
+   /* Fetch texels */
+   if (useBorderColor & (I0BIT | K0BIT)) {
+      get_border_color(tObj, img, t0);
+   }
+   else {
+      img->FetchTexelf(img, i0, array, 0, t0);
+   }
+   if (useBorderColor & (I1BIT | K0BIT)) {
+      get_border_color(tObj, img, t1);
+   }
+   else {
+      img->FetchTexelf(img, i1, array, 0, t1);
+   }
+
+   /* bilinear interpolation of samples */
+   lerp_rgba(rgba, a, t0, t1);
+}
+
+
+static void
+sample_1d_array_nearest_mipmap_nearest(struct gl_context *ctx,
+                                       const struct gl_texture_object *tObj,
+                                       GLuint n, const GLfloat texcoord[][4],
+                                       const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   for (i = 0; i < n; i++) {
+      GLint level = nearest_mipmap_level(tObj, lambda[i]);
+      sample_1d_array_nearest(ctx, tObj, tObj->Image[0][level], texcoord[i],
+                              rgba[i]);
+   }
+}
+
+
+static void
+sample_1d_array_linear_mipmap_nearest(struct gl_context *ctx,
+                                      const struct gl_texture_object *tObj,
+                                      GLuint n, const GLfloat texcoord[][4],
+                                      const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   ASSERT(lambda != NULL);
+   for (i = 0; i < n; i++) {
+      GLint level = nearest_mipmap_level(tObj, lambda[i]);
+      sample_1d_array_linear(ctx, tObj, tObj->Image[0][level],
+                             texcoord[i], rgba[i]);
+   }
+}
+
+
+static void
+sample_1d_array_nearest_mipmap_linear(struct gl_context *ctx,
+                                      const struct gl_texture_object *tObj,
+                                      GLuint n, const GLfloat texcoord[][4],
+                                      const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   ASSERT(lambda != NULL);
+   for (i = 0; i < n; i++) {
+      GLint level = linear_mipmap_level(tObj, lambda[i]);
+      if (level >= tObj->_MaxLevel) {
+         sample_1d_array_nearest(ctx, tObj, tObj->Image[0][tObj->_MaxLevel],
+                                 texcoord[i], rgba[i]);
+      }
+      else {
+         GLfloat t0[4], t1[4];  /* texels */
+         const GLfloat f = FRAC(lambda[i]);
+         sample_1d_array_nearest(ctx, tObj, tObj->Image[0][level  ], texcoord[i], t0);
+         sample_1d_array_nearest(ctx, tObj, tObj->Image[0][level+1], texcoord[i], t1);
+         lerp_rgba(rgba[i], f, t0, t1);
+      }
+   }
+}
+
+
+static void
+sample_1d_array_linear_mipmap_linear(struct gl_context *ctx,
+                                     const struct gl_texture_object *tObj,
+                                     GLuint n, const GLfloat texcoord[][4],
+                                     const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   ASSERT(lambda != NULL);
+   for (i = 0; i < n; i++) {
+      GLint level = linear_mipmap_level(tObj, lambda[i]);
+      if (level >= tObj->_MaxLevel) {
+         sample_1d_array_linear(ctx, tObj, tObj->Image[0][tObj->_MaxLevel],
+                          texcoord[i], rgba[i]);
+      }
+      else {
+         GLfloat t0[4], t1[4];  /* texels */
+         const GLfloat f = FRAC(lambda[i]);
+         sample_1d_array_linear(ctx, tObj, tObj->Image[0][level  ], texcoord[i], t0);
+         sample_1d_array_linear(ctx, tObj, tObj->Image[0][level+1], texcoord[i], t1);
+         lerp_rgba(rgba[i], f, t0, t1);
+      }
+   }
+}
+
+
+/** Sample 1D Array texture, nearest filtering for both min/magnification */
+static void
+sample_nearest_1d_array(struct gl_context *ctx,
+                        const struct gl_texture_object *tObj, GLuint n,
+                        const GLfloat texcoords[][4], const GLfloat lambda[],
+                        GLfloat rgba[][4])
+{
+   GLuint i;
+   struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
+   (void) lambda;
+   for (i = 0; i < n; i++) {
+      sample_1d_array_nearest(ctx, tObj, image, texcoords[i], rgba[i]);
+   }
+}
+
+
+/** Sample 1D Array texture, linear filtering for both min/magnification */
+static void
+sample_linear_1d_array(struct gl_context *ctx,
+                       const struct gl_texture_object *tObj, GLuint n,
+                       const GLfloat texcoords[][4],
+                       const GLfloat lambda[], GLfloat rgba[][4])
+{
+   GLuint i;
+   struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
+   (void) lambda;
+   for (i = 0; i < n; i++) {
+      sample_1d_array_linear(ctx, tObj, image, texcoords[i], rgba[i]);
+   }
+}
+
+
+/** Sample 1D Array texture, using lambda to choose between min/magnification */
+static void
+sample_lambda_1d_array(struct gl_context *ctx,
+                       const struct gl_texture_object *tObj, GLuint n,
+                       const GLfloat texcoords[][4], const GLfloat lambda[],
+                       GLfloat rgba[][4])
+{
+   GLuint minStart, minEnd;  /* texels with minification */
+   GLuint magStart, magEnd;  /* texels with magnification */
+   GLuint i;
+
+   ASSERT(lambda != NULL);
+   compute_min_mag_ranges(tObj, n, lambda,
+                          &minStart, &minEnd, &magStart, &magEnd);
+
+   if (minStart < minEnd) {
+      /* do the minified texels */
+      GLuint m = minEnd - minStart;
+      switch (tObj->Sampler.MinFilter) {
+      case GL_NEAREST:
+         for (i = minStart; i < minEnd; i++)
+            sample_1d_array_nearest(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
+                                    texcoords[i], rgba[i]);
+         break;
+      case GL_LINEAR:
+         for (i = minStart; i < minEnd; i++)
+            sample_1d_array_linear(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
+                                   texcoords[i], rgba[i]);
+         break;
+      case GL_NEAREST_MIPMAP_NEAREST:
+         sample_1d_array_nearest_mipmap_nearest(ctx, tObj, m, texcoords + minStart,
+                                                lambda + minStart, rgba + minStart);
+         break;
+      case GL_LINEAR_MIPMAP_NEAREST:
+         sample_1d_array_linear_mipmap_nearest(ctx, tObj, m, 
+                                               texcoords + minStart,
+                                               lambda + minStart,
+                                               rgba + minStart);
+         break;
+      case GL_NEAREST_MIPMAP_LINEAR:
+         sample_1d_array_nearest_mipmap_linear(ctx, tObj, m, texcoords + minStart,
+                                               lambda + minStart, rgba + minStart);
+         break;
+      case GL_LINEAR_MIPMAP_LINEAR:
+         sample_1d_array_linear_mipmap_linear(ctx, tObj, m, 
+                                              texcoords + minStart,
+                                              lambda + minStart, 
+                                              rgba + minStart);
+         break;
+      default:
+         _mesa_problem(ctx, "Bad min filter in sample_1d_array_texture");
+         return;
+      }
+   }
+
+   if (magStart < magEnd) {
+      /* do the magnified texels */
+      switch (tObj->Sampler.MagFilter) {
+      case GL_NEAREST:
+         for (i = magStart; i < magEnd; i++)
+            sample_1d_array_nearest(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
+                              texcoords[i], rgba[i]);
+         break;
+      case GL_LINEAR:
+         for (i = magStart; i < magEnd; i++)
+            sample_1d_array_linear(ctx, tObj, tObj->Image[0][tObj->BaseLevel],
+                                   texcoords[i], rgba[i]);
+         break;
+      default:
+         _mesa_problem(ctx, "Bad mag filter in sample_1d_array_texture");
+         return;
+      }
+   }
+}
+
+
+/**
+ * Compare texcoord against depth sample.  Return 1.0 or the ambient value.
+ */
+static INLINE GLfloat
+shadow_compare(GLenum function, GLfloat coord, GLfloat depthSample,
+               GLfloat ambient)
+{
+   switch (function) {
+   case GL_LEQUAL:
+      return (coord <= depthSample) ? 1.0F : ambient;
+   case GL_GEQUAL:
+      return (coord >= depthSample) ? 1.0F : ambient;
+   case GL_LESS:
+      return (coord < depthSample) ? 1.0F : ambient;
+   case GL_GREATER:
+      return (coord > depthSample) ? 1.0F : ambient;
+   case GL_EQUAL:
+      return (coord == depthSample) ? 1.0F : ambient;
+   case GL_NOTEQUAL:
+      return (coord != depthSample) ? 1.0F : ambient;
+   case GL_ALWAYS:
+      return 1.0F;
+   case GL_NEVER:
+      return ambient;
+   case GL_NONE:
+      return depthSample;
+   default:
+      _mesa_problem(NULL, "Bad compare func in shadow_compare");
+      return ambient;
+   }
+}
+
+
+/**
+ * Compare texcoord against four depth samples.
+ */
+static INLINE GLfloat
+shadow_compare4(GLenum function, GLfloat coord,
+                GLfloat depth00, GLfloat depth01,
+                GLfloat depth10, GLfloat depth11,
+                GLfloat ambient, GLfloat wi, GLfloat wj)
+{
+   const GLfloat d = (1.0F - (GLfloat) ambient) * 0.25F;
+   GLfloat luminance = 1.0F;
+
+   switch (function) {
+   case GL_LEQUAL:
+      if (coord > depth00)  luminance -= d;
+      if (coord > depth01)  luminance -= d;
+      if (coord > depth10)  luminance -= d;
+      if (coord > depth11)  luminance -= d;
+      return luminance;
+   case GL_GEQUAL:
+      if (coord < depth00)  luminance -= d;
+      if (coord < depth01)  luminance -= d;
+      if (coord < depth10)  luminance -= d;
+      if (coord < depth11)  luminance -= d;
+      return luminance;
+   case GL_LESS:
+      if (coord >= depth00)  luminance -= d;
+      if (coord >= depth01)  luminance -= d;
+      if (coord >= depth10)  luminance -= d;
+      if (coord >= depth11)  luminance -= d;
+      return luminance;
+   case GL_GREATER:
+      if (coord <= depth00)  luminance -= d;
+      if (coord <= depth01)  luminance -= d;
+      if (coord <= depth10)  luminance -= d;
+      if (coord <= depth11)  luminance -= d;
+      return luminance;
+   case GL_EQUAL:
+      if (coord != depth00)  luminance -= d;
+      if (coord != depth01)  luminance -= d;
+      if (coord != depth10)  luminance -= d;
+      if (coord != depth11)  luminance -= d;
+      return luminance;
+   case GL_NOTEQUAL:
+      if (coord == depth00)  luminance -= d;
+      if (coord == depth01)  luminance -= d;
+      if (coord == depth10)  luminance -= d;
+      if (coord == depth11)  luminance -= d;
+      return luminance;
+   case GL_ALWAYS:
+      return 1.0F;
+   case GL_NEVER:
+      return ambient;
+   case GL_NONE:
+      /* ordinary bilinear filtering */
+      return lerp_2d(wi, wj, depth00, depth10, depth01, depth11);
+   default:
+      _mesa_problem(NULL, "Bad compare func in sample_compare4");
+      return ambient;
+   }
+}
+
+
+/**
+ * Choose the mipmap level to use when sampling from a depth texture.
+ */
+static int
+choose_depth_texture_level(const struct gl_texture_object *tObj, GLfloat lambda)
+{
+   GLint level;
+
+   if (tObj->Sampler.MinFilter == GL_NEAREST || tObj->Sampler.MinFilter == GL_LINEAR) {
+      /* no mipmapping - use base level */
+      level = tObj->BaseLevel;
+   }
+   else {
+      /* choose mipmap level */
+      lambda = CLAMP(lambda, tObj->Sampler.MinLod, tObj->Sampler.MaxLod);
+      level = (GLint) lambda;
+      level = CLAMP(level, tObj->BaseLevel, tObj->_MaxLevel);
+   }
+
+   return level;
+}
+
+
+/**
+ * Sample a shadow/depth texture.  This function is incomplete.  It doesn't
+ * check for minification vs. magnification, etc.
+ */
+static void
+sample_depth_texture( struct gl_context *ctx,
+                      const struct gl_texture_object *tObj, GLuint n,
+                      const GLfloat texcoords[][4], const GLfloat lambda[],
+                      GLfloat texel[][4] )
+{
+   const GLint level = choose_depth_texture_level(tObj, lambda[0]);
+   const struct gl_texture_image *img = tObj->Image[0][level];
+   const GLint width = img->Width;
+   const GLint height = img->Height;
+   const GLint depth = img->Depth;
+   const GLuint compare_coord = (tObj->Target == GL_TEXTURE_2D_ARRAY_EXT)
+       ? 3 : 2;
+   GLfloat ambient;
+   GLenum function;
+   GLfloat result;
+
+   ASSERT(img->_BaseFormat == GL_DEPTH_COMPONENT ||
+          img->_BaseFormat == GL_DEPTH_STENCIL_EXT);
+
+   ASSERT(tObj->Target == GL_TEXTURE_1D ||
+          tObj->Target == GL_TEXTURE_2D ||
+          tObj->Target == GL_TEXTURE_RECTANGLE_NV ||
+          tObj->Target == GL_TEXTURE_1D_ARRAY_EXT ||
+          tObj->Target == GL_TEXTURE_2D_ARRAY_EXT);
+
+   ambient = tObj->Sampler.CompareFailValue;
+
+   /* XXXX if tObj->Sampler.MinFilter != tObj->Sampler.MagFilter, we're ignoring lambda */
+
+   function = (tObj->Sampler.CompareMode == GL_COMPARE_R_TO_TEXTURE_ARB) ?
+      tObj->Sampler.CompareFunc : GL_NONE;
+
+   if (tObj->Sampler.MagFilter == GL_NEAREST) {
+      GLuint i;
+      for (i = 0; i < n; i++) {
+         GLfloat depthSample, depthRef;
+         GLint col, row, slice;
+
+         nearest_texcoord(tObj, level, texcoords[i], &col, &row, &slice);
+
+         if (col >= 0 && row >= 0 && col < width && row < height && 
+             slice >= 0 && slice < depth) {
+            img->FetchTexelf(img, col, row, slice, &depthSample);
+         }
+         else {
+            depthSample = tObj->Sampler.BorderColor.f[0];
+         }
+
+         depthRef = CLAMP(texcoords[i][compare_coord], 0.0F, 1.0F);
+
+         result = shadow_compare(function, depthRef, depthSample, ambient);
+
+         switch (tObj->Sampler.DepthMode) {
+         case GL_LUMINANCE:
+            ASSIGN_4V(texel[i], result, result, result, 1.0F);
+            break;
+         case GL_INTENSITY:
+            ASSIGN_4V(texel[i], result, result, result, result);
+            break;
+         case GL_ALPHA:
+            ASSIGN_4V(texel[i], 0.0F, 0.0F, 0.0F, result);
+            break;
+         case GL_RED:
+            ASSIGN_4V(texel[i], result, 0.0F, 0.0F, 1.0F);
+            break;
+         default:
+            _mesa_problem(ctx, "Bad depth texture mode");
+         }
+      }
+   }
+   else {
+      GLuint i;
+      ASSERT(tObj->Sampler.MagFilter == GL_LINEAR);
+      for (i = 0; i < n; i++) {
+         GLfloat depth00, depth01, depth10, depth11, depthRef;
+         GLint i0, i1, j0, j1;
+         GLint slice;
+         GLfloat wi, wj;
+         GLuint useBorderTexel;
+
+         linear_texcoord(tObj, level, texcoords[i], &i0, &i1, &j0, &j1, &slice,
+                         &wi, &wj);
+
+         useBorderTexel = 0;
+         if (img->Border) {
+            i0 += img->Border;
+            i1 += img->Border;
+            if (tObj->Target != GL_TEXTURE_1D_ARRAY_EXT) {
+               j0 += img->Border;
+               j1 += img->Border;
+            }
+         }
+         else {
+            if (i0 < 0 || i0 >= (GLint) width)   useBorderTexel |= I0BIT;
+            if (i1 < 0 || i1 >= (GLint) width)   useBorderTexel |= I1BIT;
+            if (j0 < 0 || j0 >= (GLint) height)  useBorderTexel |= J0BIT;
+            if (j1 < 0 || j1 >= (GLint) height)  useBorderTexel |= J1BIT;
+         }
+
+         if (slice < 0 || slice >= (GLint) depth) {
+            depth00 = tObj->Sampler.BorderColor.f[0];
+            depth01 = tObj->Sampler.BorderColor.f[0];
+            depth10 = tObj->Sampler.BorderColor.f[0];
+            depth11 = tObj->Sampler.BorderColor.f[0];
+         }
+         else {
+            /* get four depth samples from the texture */
+            if (useBorderTexel & (I0BIT | J0BIT)) {
+               depth00 = tObj->Sampler.BorderColor.f[0];
+            }
+            else {
+               img->FetchTexelf(img, i0, j0, slice, &depth00);
+            }
+            if (useBorderTexel & (I1BIT | J0BIT)) {
+               depth10 = tObj->Sampler.BorderColor.f[0];
+            }
+            else {
+               img->FetchTexelf(img, i1, j0, slice, &depth10);
+            }
+
+            if (tObj->Target != GL_TEXTURE_1D_ARRAY_EXT) {
+               if (useBorderTexel & (I0BIT | J1BIT)) {
+                  depth01 = tObj->Sampler.BorderColor.f[0];
+               }
+               else {
+                  img->FetchTexelf(img, i0, j1, slice, &depth01);
+               }
+               if (useBorderTexel & (I1BIT | J1BIT)) {
+                  depth11 = tObj->Sampler.BorderColor.f[0];
+               }
+               else {
+                  img->FetchTexelf(img, i1, j1, slice, &depth11);
+               }
+            }
+            else {
+               depth01 = depth00;
+               depth11 = depth10;
+            }
+         }
+
+         depthRef = CLAMP(texcoords[i][compare_coord], 0.0F, 1.0F);
+
+         result = shadow_compare4(function, depthRef,
+                                  depth00, depth01, depth10, depth11,
+                                  ambient, wi, wj);
+
+         switch (tObj->Sampler.DepthMode) {
+         case GL_LUMINANCE:
+            ASSIGN_4V(texel[i], result, result, result, 1.0F);
+            break;
+         case GL_INTENSITY:
+            ASSIGN_4V(texel[i], result, result, result, result);
+            break;
+         case GL_ALPHA:
+            ASSIGN_4V(texel[i], 0.0F, 0.0F, 0.0F, result);
+            break;
+         default:
+            _mesa_problem(ctx, "Bad depth texture mode");
+         }
+
+      }  /* for */
+   }  /* if filter */
+}
+
+
+/**
+ * We use this function when a texture object is in an "incomplete" state.
+ * When a fragment program attempts to sample an incomplete texture we
+ * return black (see issue 23 in GL_ARB_fragment_program spec).
+ * Note: fragment programs don't observe the texture enable/disable flags.
+ */
+static void
+null_sample_func( struct gl_context *ctx,
+		  const struct gl_texture_object *tObj, GLuint n,
+		  const GLfloat texcoords[][4], const GLfloat lambda[],
+		  GLfloat rgba[][4])
+{
+   GLuint i;
+   (void) ctx;
+   (void) tObj;
+   (void) texcoords;
+   (void) lambda;
+   for (i = 0; i < n; i++) {
+      rgba[i][RCOMP] = 0;
+      rgba[i][GCOMP] = 0;
+      rgba[i][BCOMP] = 0;
+      rgba[i][ACOMP] = 1.0;
+   }
+}
+
+
+/**
+ * Choose the texture sampling function for the given texture object.
+ */
+texture_sample_func
+_swrast_choose_texture_sample_func( struct gl_context *ctx,
+				    const struct gl_texture_object *t )
+{
+   if (!t || !t->_Complete) {
+      return &null_sample_func;
+   }
+   else {
+      const GLboolean needLambda =
+         (GLboolean) (t->Sampler.MinFilter != t->Sampler.MagFilter);
+      const GLenum format = t->Image[0][t->BaseLevel]->_BaseFormat;
+
+      switch (t->Target) {
+      case GL_TEXTURE_1D:
+         if (format == GL_DEPTH_COMPONENT || format == GL_DEPTH_STENCIL_EXT) {
+            return &sample_depth_texture;
+         }
+         else if (needLambda) {
+            return &sample_lambda_1d;
+         }
+         else if (t->Sampler.MinFilter == GL_LINEAR) {
+            return &sample_linear_1d;
+         }
+         else {
+            ASSERT(t->Sampler.MinFilter == GL_NEAREST);
+            return &sample_nearest_1d;
+         }
+      case GL_TEXTURE_2D:
+         if (format == GL_DEPTH_COMPONENT || format == GL_DEPTH_STENCIL_EXT) {
+            return &sample_depth_texture;
+         }
+         else if (needLambda) {
+            /* Anisotropic filtering extension. Activated only if mipmaps are used */
+            if (t->Sampler.MaxAnisotropy > 1.0 &&
+                t->Sampler.MinFilter == GL_LINEAR_MIPMAP_LINEAR) {
+               return &sample_lambda_2d_aniso;
+            }
+            return &sample_lambda_2d;
+         }
+         else if (t->Sampler.MinFilter == GL_LINEAR) {
+            return &sample_linear_2d;
+         }
+         else {
+            /* check for a few optimized cases */
+            const struct gl_texture_image *img = t->Image[0][t->BaseLevel];
+            ASSERT(t->Sampler.MinFilter == GL_NEAREST);
+            if (t->Sampler.WrapS == GL_REPEAT &&
+                t->Sampler.WrapT == GL_REPEAT &&
+                img->_IsPowerOfTwo &&
+                img->Border == 0 &&
+                img->TexFormat == MESA_FORMAT_RGB888) {
+               return &opt_sample_rgb_2d;
+            }
+            else if (t->Sampler.WrapS == GL_REPEAT &&
+                     t->Sampler.WrapT == GL_REPEAT &&
+                     img->_IsPowerOfTwo &&
+                     img->Border == 0 &&
+                     img->TexFormat == MESA_FORMAT_RGBA8888) {
+               return &opt_sample_rgba_2d;
+            }
+            else {
+               return &sample_nearest_2d;
+            }
+         }
+      case GL_TEXTURE_3D:
+         if (needLambda) {
+            return &sample_lambda_3d;
+         }
+         else if (t->Sampler.MinFilter == GL_LINEAR) {
+            return &sample_linear_3d;
+         }
+         else {
+            ASSERT(t->Sampler.MinFilter == GL_NEAREST);
+            return &sample_nearest_3d;
+         }
+      case GL_TEXTURE_CUBE_MAP:
+         if (needLambda) {
+            return &sample_lambda_cube;
+         }
+         else if (t->Sampler.MinFilter == GL_LINEAR) {
+            return &sample_linear_cube;
+         }
+         else {
+            ASSERT(t->Sampler.MinFilter == GL_NEAREST);
+            return &sample_nearest_cube;
+         }
+      case GL_TEXTURE_RECTANGLE_NV:
+         if (format == GL_DEPTH_COMPONENT || format == GL_DEPTH_STENCIL_EXT) {
+            return &sample_depth_texture;
+         }
+         else if (needLambda) {
+            return &sample_lambda_rect;
+         }
+         else if (t->Sampler.MinFilter == GL_LINEAR) {
+            return &sample_linear_rect;
+         }
+         else {
+            ASSERT(t->Sampler.MinFilter == GL_NEAREST);
+            return &sample_nearest_rect;
+         }
+      case GL_TEXTURE_1D_ARRAY_EXT:
+         if (needLambda) {
+            return &sample_lambda_1d_array;
+         }
+         else if (t->Sampler.MinFilter == GL_LINEAR) {
+            return &sample_linear_1d_array;
+         }
+         else {
+            ASSERT(t->Sampler.MinFilter == GL_NEAREST);
+            return &sample_nearest_1d_array;
+         }
+      case GL_TEXTURE_2D_ARRAY_EXT:
+         if (needLambda) {
+            return &sample_lambda_2d_array;
+         }
+         else if (t->Sampler.MinFilter == GL_LINEAR) {
+            return &sample_linear_2d_array;
+         }
+         else {
+            ASSERT(t->Sampler.MinFilter == GL_NEAREST);
+            return &sample_nearest_2d_array;
+         }
+      default:
+         _mesa_problem(ctx,
+                       "invalid target in _swrast_choose_texture_sample_func");
+         return &null_sample_func;
+      }
+   }
+}
-- 
cgit v1.2.3