Merge remote-tracking branch 'origin/released'

Conflicts:
	apps/xwininfo/configure.ac
	apps/xwininfo/xwininfo.c
	libX11/configure.ac
	libX11/specs/libX11/AppC.xml
	libX11/specs/libX11/AppD.xml
	libX11/specs/libX11/CH03.xml
	libX11/specs/libX11/CH04.xml
	libX11/specs/libX11/CH05.xml
	libX11/specs/libX11/CH06.xml
	libX11/specs/libX11/CH07.xml
	libX11/specs/libX11/CH08.xml
	libX11/specs/libX11/CH09.xml
	libX11/specs/libX11/CH11.xml
	libX11/specs/libX11/CH12.xml
	libX11/specs/libX11/CH13.xml
	libX11/specs/libX11/CH14.xml
	libX11/specs/libX11/CH15.xml
	libX11/specs/libX11/CH16.xml
	libxcb/configure.ac
	libxcb/src/c_client.py
	libxcb/src/xcb_auth.c
	libxcb/src/xcb_util.c
	mesalib/common.py
	mesalib/configs/linux-dri
	mesalib/docs/GL3.txt
	mesalib/docs/download.html
	mesalib/docs/install.html
	mesalib/include/GL/internal/dri_interface.h
	mesalib/scons/custom.py
	mesalib/scons/gallium.py
	mesalib/src/gallium/auxiliary/util/u_math.h
	mesalib/src/gallium/auxiliary/util/u_vbuf_mgr.c
	mesalib/src/glsl/ast_function.cpp
	mesalib/src/glsl/ast_to_hir.cpp
	mesalib/src/glsl/glcpp/glcpp-parse.y
	mesalib/src/glsl/glsl_parser_extras.cpp
	mesalib/src/glsl/glsl_parser_extras.h
	mesalib/src/glsl/ir.cpp
	mesalib/src/glsl/ir.h
	mesalib/src/glsl/ir_clone.cpp
	mesalib/src/glsl/ir_print_visitor.cpp
	mesalib/src/glsl/ir_validate.cpp
	mesalib/src/glsl/linker.cpp
	mesalib/src/glsl/main.cpp
	mesalib/src/glsl/opt_tree_grafting.cpp
	mesalib/src/mapi/glapi/gen/Makefile
	mesalib/src/mapi/glapi/gen/gl_XML.py
	mesalib/src/mapi/glapi/gen/gl_table.py
	mesalib/src/mapi/glapi/glapi_mapi_tmp.h
	mesalib/src/mapi/glapi/glapi_sparc.S
	mesalib/src/mapi/glapi/glapi_x86-64.S
	mesalib/src/mapi/glapi/glapi_x86.S
	mesalib/src/mapi/glapi/glapitemp.h
	mesalib/src/mapi/glapi/glprocs.h
	mesalib/src/mapi/mapi/u_thread.c
	mesalib/src/mapi/mapi/u_thread.h
	mesalib/src/mesa/SConscript
	mesalib/src/mesa/drivers/common/driverfuncs.c
	mesalib/src/mesa/drivers/windows/gldirect/dglcontext.c
	mesalib/src/mesa/main/api_arrayelt.c
	mesalib/src/mesa/main/bufferobj.c
	mesalib/src/mesa/main/compiler.h
	mesalib/src/mesa/main/dd.h
	mesalib/src/mesa/main/dlist.c
	mesalib/src/mesa/main/enums.c
	mesalib/src/mesa/main/es_generator.py
	mesalib/src/mesa/main/fbobject.c
	mesalib/src/mesa/main/imports.h
	mesalib/src/mesa/main/mtypes.h
	mesalib/src/mesa/main/pbo.c
	mesalib/src/mesa/main/remap_helper.h
	mesalib/src/mesa/main/shared.c
	mesalib/src/mesa/main/texgetimage.c
	mesalib/src/mesa/main/teximage.c
	mesalib/src/mesa/main/uniforms.c
	mesalib/src/mesa/program/ir_to_mesa.cpp
	mesalib/src/mesa/program/prog_optimize.c
	mesalib/src/mesa/program/register_allocate.c
	mesalib/src/mesa/program/register_allocate.h
	mesalib/src/mesa/sources.mak
	mesalib/src/mesa/state_tracker/st_cb_bufferobjects.c
	mesalib/src/mesa/state_tracker/st_cb_texture.c
	mesalib/src/mesa/state_tracker/st_texture.h
	mesalib/src/mesa/swrast/s_context.c
	mesalib/src/mesa/tnl/t_draw.c
	mesalib/src/mesa/vbo/vbo_exec_array.c
	mesalib/src/mesa/vbo/vbo_save_api.c
	pixman/pixman/pixman-arm-common.h
	pixman/pixman/pixman-fast-path.c
	pixman/pixman/pixman-inlines.h
	pixman/test/Makefile.am
	pixman/test/utils.c
	pixman/test/utils.h
	xorg-server/configure.ac
	xorg-server/fb/fbblt.c
	xorg-server/hw/xquartz/darwin.c
	xorg-server/xkeyboard-config/symbols/de

4 files changed, 2364 insertions, 2412 deletions

diff --git a/mesalib/src/mesa/swrast/s_aatritemp.h b/mesalib/src/mesa/swrast/s_aatritemp.h
index 4136df3a7..77b3ae6ec 100644
--- a/mesalib/src/mesa/swrast/s_aatritemp.h
+++ b/mesalib/src/mesa/swrast/s_aatritemp.h
@@ -1,331 +1,343 @@
-/*

- * Mesa 3-D graphics library

- * Version:  7.0.3

- *

- * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.

- *

- * Permission is hereby granted, free of charge, to any person obtaining a

- * copy of this software and associated documentation files (the "Software"),

- * 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.

- */

-

-

-/*

- * Antialiased Triangle Rasterizer Template

- *

- * This file is #include'd to generate custom AA triangle rasterizers.

- * NOTE: this code hasn't been optimized yet.  That'll come after it

- * works correctly.

- *

- * The following macros may be defined to indicate what auxillary information

- * must be copmuted across the triangle:

- *    DO_Z         - if defined, compute Z values

- *    DO_ATTRIBS   - if defined, compute texcoords, varying, etc.

- */

-

-/*void triangle( struct gl_context *ctx, GLuint v0, GLuint v1, GLuint v2, GLuint pv )*/

-{

-   const SWcontext *swrast = SWRAST_CONTEXT(ctx);

-   const GLfloat *p0 = v0->attrib[FRAG_ATTRIB_WPOS];

-   const GLfloat *p1 = v1->attrib[FRAG_ATTRIB_WPOS];

-   const GLfloat *p2 = v2->attrib[FRAG_ATTRIB_WPOS];

-   const SWvertex *vMin, *vMid, *vMax;

-   GLint iyMin, iyMax;

-   GLfloat yMin, yMax;

-   GLboolean ltor;

-   GLfloat majDx, majDy;  /* major (i.e. long) edge dx and dy */

-   

-   SWspan span;

-   

-#ifdef DO_Z

-   GLfloat zPlane[4];

-#endif

-   GLfloat rPlane[4], gPlane[4], bPlane[4], aPlane[4];

-#if defined(DO_ATTRIBS)

-   GLfloat attrPlane[FRAG_ATTRIB_MAX][4][4];

-   GLfloat wPlane[4];  /* win[3] */

-#endif

-   GLfloat bf = SWRAST_CONTEXT(ctx)->_BackfaceCullSign;

-   

-   (void) swrast;

-

-   INIT_SPAN(span, GL_POLYGON);

-   span.arrayMask = SPAN_COVERAGE;

-

-   /* determine bottom to top order of vertices */

-   {

-      GLfloat y0 = v0->attrib[FRAG_ATTRIB_WPOS][1];

-      GLfloat y1 = v1->attrib[FRAG_ATTRIB_WPOS][1];

-      GLfloat y2 = v2->attrib[FRAG_ATTRIB_WPOS][1];

-      if (y0 <= y1) {

-	 if (y1 <= y2) {

-	    vMin = v0;   vMid = v1;   vMax = v2;   /* y0<=y1<=y2 */

-	 }

-	 else if (y2 <= y0) {

-	    vMin = v2;   vMid = v0;   vMax = v1;   /* y2<=y0<=y1 */

-	 }

-	 else {

-	    vMin = v0;   vMid = v2;   vMax = v1;  bf = -bf; /* y0<=y2<=y1 */

-	 }

-      }

-      else {

-	 if (y0 <= y2) {

-	    vMin = v1;   vMid = v0;   vMax = v2;  bf = -bf; /* y1<=y0<=y2 */

-	 }

-	 else if (y2 <= y1) {

-	    vMin = v2;   vMid = v1;   vMax = v0;  bf = -bf; /* y2<=y1<=y0 */

-	 }

-	 else {

-	    vMin = v1;   vMid = v2;   vMax = v0;   /* y1<=y2<=y0 */

-	 }

-      }

-   }

-

-   majDx = vMax->attrib[FRAG_ATTRIB_WPOS][0] - vMin->attrib[FRAG_ATTRIB_WPOS][0];

-   majDy = vMax->attrib[FRAG_ATTRIB_WPOS][1] - vMin->attrib[FRAG_ATTRIB_WPOS][1];

-

-   /* front/back-face determination and cullling */

-   {

-      const GLfloat botDx = vMid->attrib[FRAG_ATTRIB_WPOS][0] - vMin->attrib[FRAG_ATTRIB_WPOS][0];

-      const GLfloat botDy = vMid->attrib[FRAG_ATTRIB_WPOS][1] - vMin->attrib[FRAG_ATTRIB_WPOS][1];

-      const GLfloat area = majDx * botDy - botDx * majDy;

-      /* Do backface culling */

-      if (area * bf < 0 || area == 0 || IS_INF_OR_NAN(area))

-	 return;

-      ltor = (GLboolean) (area < 0.0F);

-

-      span.facing = area * swrast->_BackfaceSign > 0.0F;

-   }

-

-   /* Plane equation setup:

-    * We evaluate plane equations at window (x,y) coordinates in order

-    * to compute color, Z, fog, texcoords, etc.  This isn't terribly

-    * efficient but it's easy and reliable.

-    */

-#ifdef DO_Z

-   compute_plane(p0, p1, p2, p0[2], p1[2], p2[2], zPlane);

-   span.arrayMask |= SPAN_Z;

-#endif

-   if (ctx->Light.ShadeModel == GL_SMOOTH) {

-      compute_plane(p0, p1, p2, v0->color[RCOMP], v1->color[RCOMP], v2->color[RCOMP], rPlane);

-      compute_plane(p0, p1, p2, v0->color[GCOMP], v1->color[GCOMP], v2->color[GCOMP], gPlane);

-      compute_plane(p0, p1, p2, v0->color[BCOMP], v1->color[BCOMP], v2->color[BCOMP], bPlane);

-      compute_plane(p0, p1, p2, v0->color[ACOMP], v1->color[ACOMP], v2->color[ACOMP], aPlane);

-   }

-   else {

-      constant_plane(v2->color[RCOMP], rPlane);

-      constant_plane(v2->color[GCOMP], gPlane);

-      constant_plane(v2->color[BCOMP], bPlane);

-      constant_plane(v2->color[ACOMP], aPlane);

-   }

-   span.arrayMask |= SPAN_RGBA;

-#if defined(DO_ATTRIBS)

-   {

-      const GLfloat invW0 = v0->attrib[FRAG_ATTRIB_WPOS][3];

-      const GLfloat invW1 = v1->attrib[FRAG_ATTRIB_WPOS][3];

-      const GLfloat invW2 = v2->attrib[FRAG_ATTRIB_WPOS][3];

-      compute_plane(p0, p1, p2, invW0, invW1, invW2, wPlane);

-      span.attrStepX[FRAG_ATTRIB_WPOS][3] = plane_dx(wPlane);

-      span.attrStepY[FRAG_ATTRIB_WPOS][3] = plane_dy(wPlane);

-      ATTRIB_LOOP_BEGIN

-         GLuint c;

-         if (swrast->_InterpMode[attr] == GL_FLAT) {

-            for (c = 0; c < 4; c++) {

-               constant_plane(v2->attrib[attr][c] * invW2, attrPlane[attr][c]);

-            }

-         }

-         else {

-            for (c = 0; c < 4; c++) {

-               const GLfloat a0 = v0->attrib[attr][c] * invW0;

-               const GLfloat a1 = v1->attrib[attr][c] * invW1;

-               const GLfloat a2 = v2->attrib[attr][c] * invW2;

-               compute_plane(p0, p1, p2, a0, a1, a2, attrPlane[attr][c]);

-            }

-         }

-         for (c = 0; c < 4; c++) {

-            span.attrStepX[attr][c] = plane_dx(attrPlane[attr][c]);

-            span.attrStepY[attr][c] = plane_dy(attrPlane[attr][c]);

-         }

-      ATTRIB_LOOP_END

-   }

-#endif

-

-   /* Begin bottom-to-top scan over the triangle.

-    * The long edge will either be on the left or right side of the

-    * triangle.  We always scan from the long edge toward the shorter

-    * edges, stopping when we find that coverage = 0.  If the long edge

-    * is on the left we scan left-to-right.  Else, we scan right-to-left.

-    */

-   yMin = vMin->attrib[FRAG_ATTRIB_WPOS][1];

-   yMax = vMax->attrib[FRAG_ATTRIB_WPOS][1];

-   iyMin = (GLint) yMin;

-   iyMax = (GLint) yMax + 1;

-

-   if (ltor) {

-      /* scan left to right */

-      const GLfloat *pMin = vMin->attrib[FRAG_ATTRIB_WPOS];

-      const GLfloat *pMid = vMid->attrib[FRAG_ATTRIB_WPOS];

-      const GLfloat *pMax = vMax->attrib[FRAG_ATTRIB_WPOS];

-      const GLfloat dxdy = majDx / majDy;

-      const GLfloat xAdj = dxdy < 0.0F ? -dxdy : 0.0F;

-      GLfloat x = pMin[0] - (yMin - iyMin) * dxdy;

-      GLint iy;

-      for (iy = iyMin; iy < iyMax; iy++, x += dxdy) {

-         GLint ix, startX = (GLint) (x - xAdj);

-         GLuint count;

-         GLfloat coverage = 0.0F;

-

-         /* skip over fragments with zero coverage */

-         while (startX < MAX_WIDTH) {

-            coverage = compute_coveragef(pMin, pMid, pMax, startX, iy);

-            if (coverage > 0.0F)

-               break;

-            startX++;

-         }

-

-         /* enter interior of triangle */

-         ix = startX;

-

-#if defined(DO_ATTRIBS)

-         /* compute attributes at left-most fragment */

-         span.attrStart[FRAG_ATTRIB_WPOS][3] = solve_plane(ix + 0.5F, iy + 0.5F, wPlane);

-         ATTRIB_LOOP_BEGIN

-            GLuint c;

-            for (c = 0; c < 4; c++) {

-               span.attrStart[attr][c] = solve_plane(ix + 0.5F, iy + 0.5F, attrPlane[attr][c]);

-            }

-         ATTRIB_LOOP_END

-#endif

-

-         count = 0;

-         while (coverage > 0.0F) {

-            /* (cx,cy) = center of fragment */

-            const GLfloat cx = ix + 0.5F, cy = iy + 0.5F;

-            SWspanarrays *array = span.array;

-            array->coverage[count] = coverage;

-#ifdef DO_Z

-            array->z[count] = (GLuint) solve_plane(cx, cy, zPlane);

-#endif

-            array->rgba[count][RCOMP] = solve_plane_chan(cx, cy, rPlane);

-            array->rgba[count][GCOMP] = solve_plane_chan(cx, cy, gPlane);

-            array->rgba[count][BCOMP] = solve_plane_chan(cx, cy, bPlane);

-            array->rgba[count][ACOMP] = solve_plane_chan(cx, cy, aPlane);

-            ix++;

-            count++;

-            coverage = compute_coveragef(pMin, pMid, pMax, ix, iy);

-         }

-         

-         if (ix <= startX)

-            continue;

-         

-         span.x = startX;

-         span.y = iy;

-         span.end = (GLuint) ix - (GLuint) startX;

-         _swrast_write_rgba_span(ctx, &span);

-      }

-   }

-   else {

-      /* scan right to left */

-      const GLfloat *pMin = vMin->attrib[FRAG_ATTRIB_WPOS];

-      const GLfloat *pMid = vMid->attrib[FRAG_ATTRIB_WPOS];

-      const GLfloat *pMax = vMax->attrib[FRAG_ATTRIB_WPOS];

-      const GLfloat dxdy = majDx / majDy;

-      const GLfloat xAdj = dxdy > 0 ? dxdy : 0.0F;

-      GLfloat x = pMin[0] - (yMin - iyMin) * dxdy;

-      GLint iy;

-      for (iy = iyMin; iy < iyMax; iy++, x += dxdy) {

-         GLint ix, left, startX = (GLint) (x + xAdj);

-         GLuint count, n;

-         GLfloat coverage = 0.0F;

-         

-         /* make sure we're not past the window edge */

-         if (startX >= ctx->DrawBuffer->_Xmax) {

-            startX = ctx->DrawBuffer->_Xmax - 1;

-         }

-

-         /* skip fragments with zero coverage */

-         while (startX > 0) {

-            coverage = compute_coveragef(pMin, pMax, pMid, startX, iy);

-            if (coverage > 0.0F)

-               break;

-            startX--;

-         }

-         

-         /* enter interior of triangle */

-         ix = startX;

-         count = 0;

-         while (coverage > 0.0F) {

-            /* (cx,cy) = center of fragment */

-            const GLfloat cx = ix + 0.5F, cy = iy + 0.5F;

-            SWspanarrays *array = span.array;

-            ASSERT(ix >= 0);

-            array->coverage[ix] = coverage;

-#ifdef DO_Z

-            array->z[ix] = (GLuint) solve_plane(cx, cy, zPlane);

-#endif

-            array->rgba[ix][RCOMP] = solve_plane_chan(cx, cy, rPlane);

-            array->rgba[ix][GCOMP] = solve_plane_chan(cx, cy, gPlane);

-            array->rgba[ix][BCOMP] = solve_plane_chan(cx, cy, bPlane);

-            array->rgba[ix][ACOMP] = solve_plane_chan(cx, cy, aPlane);

-            ix--;

-            count++;

-            coverage = compute_coveragef(pMin, pMax, pMid, ix, iy);

-         }

-         

-#if defined(DO_ATTRIBS)

-         /* compute attributes at left-most fragment */

-         span.attrStart[FRAG_ATTRIB_WPOS][3] = solve_plane(ix + 1.5F, iy + 0.5F, wPlane);

-         ATTRIB_LOOP_BEGIN

-            GLuint c;

-            for (c = 0; c < 4; c++) {

-               span.attrStart[attr][c] = solve_plane(ix + 1.5F, iy + 0.5F, attrPlane[attr][c]);

-            }

-         ATTRIB_LOOP_END

-#endif

-

-         if (startX <= ix)

-            continue;

-

-         n = (GLuint) startX - (GLuint) ix;

-

-         left = ix + 1;

-

-         /* shift all values to the left */

-         /* XXX this is temporary */

-         {

-            SWspanarrays *array = span.array;

-            GLint j;

-            for (j = 0; j < (GLint) n; j++) {

-               array->coverage[j] = array->coverage[j + left];

-               COPY_CHAN4(array->rgba[j], array->rgba[j + left]);

-#ifdef DO_Z

-               array->z[j] = array->z[j + left];

-#endif

-            }

-         }

-

-         span.x = left;

-         span.y = iy;

-         span.end = n;

-         _swrast_write_rgba_span(ctx, &span);

-      }

-   }

-}

-

-

-#undef DO_Z

-#undef DO_ATTRIBS

-#undef DO_OCCLUSION_TEST

+/*
+ * Mesa 3-D graphics library
+ * Version:  7.0.3
+ *
+ * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * 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.
+ */
+
+
+/*
+ * Antialiased Triangle Rasterizer Template
+ *
+ * This file is #include'd to generate custom AA triangle rasterizers.
+ * NOTE: this code hasn't been optimized yet.  That'll come after it
+ * works correctly.
+ *
+ * The following macros may be defined to indicate what auxillary information
+ * must be copmuted across the triangle:
+ *    DO_Z         - if defined, compute Z values
+ *    DO_ATTRIBS   - if defined, compute texcoords, varying, etc.
+ */
+
+/*void triangle( struct gl_context *ctx, GLuint v0, GLuint v1, GLuint v2, GLuint pv )*/
+{
+   const SWcontext *swrast = SWRAST_CONTEXT(ctx);
+   const GLfloat *p0 = v0->attrib[FRAG_ATTRIB_WPOS];
+   const GLfloat *p1 = v1->attrib[FRAG_ATTRIB_WPOS];
+   const GLfloat *p2 = v2->attrib[FRAG_ATTRIB_WPOS];
+   const SWvertex *vMin, *vMid, *vMax;
+   GLint iyMin, iyMax;
+   GLfloat yMin, yMax;
+   GLboolean ltor;
+   GLfloat majDx, majDy;  /* major (i.e. long) edge dx and dy */
+   
+   SWspan span;
+   
+#ifdef DO_Z
+   GLfloat zPlane[4];
+#endif
+   GLfloat rPlane[4], gPlane[4], bPlane[4], aPlane[4];
+#if defined(DO_ATTRIBS)
+   GLfloat attrPlane[FRAG_ATTRIB_MAX][4][4];
+   GLfloat wPlane[4];  /* win[3] */
+#endif
+   GLfloat bf = SWRAST_CONTEXT(ctx)->_BackfaceCullSign;
+   
+   (void) swrast;
+
+   INIT_SPAN(span, GL_POLYGON);
+   span.arrayMask = SPAN_COVERAGE;
+
+   /* determine bottom to top order of vertices */
+   {
+      GLfloat y0 = v0->attrib[FRAG_ATTRIB_WPOS][1];
+      GLfloat y1 = v1->attrib[FRAG_ATTRIB_WPOS][1];
+      GLfloat y2 = v2->attrib[FRAG_ATTRIB_WPOS][1];
+      if (y0 <= y1) {
+	 if (y1 <= y2) {
+	    vMin = v0;   vMid = v1;   vMax = v2;   /* y0<=y1<=y2 */
+	 }
+	 else if (y2 <= y0) {
+	    vMin = v2;   vMid = v0;   vMax = v1;   /* y2<=y0<=y1 */
+	 }
+	 else {
+	    vMin = v0;   vMid = v2;   vMax = v1;  bf = -bf; /* y0<=y2<=y1 */
+	 }
+      }
+      else {
+	 if (y0 <= y2) {
+	    vMin = v1;   vMid = v0;   vMax = v2;  bf = -bf; /* y1<=y0<=y2 */
+	 }
+	 else if (y2 <= y1) {
+	    vMin = v2;   vMid = v1;   vMax = v0;  bf = -bf; /* y2<=y1<=y0 */
+	 }
+	 else {
+	    vMin = v1;   vMid = v2;   vMax = v0;   /* y1<=y2<=y0 */
+	 }
+      }
+   }
+
+   majDx = vMax->attrib[FRAG_ATTRIB_WPOS][0] - vMin->attrib[FRAG_ATTRIB_WPOS][0];
+   majDy = vMax->attrib[FRAG_ATTRIB_WPOS][1] - vMin->attrib[FRAG_ATTRIB_WPOS][1];
+
+   /* front/back-face determination and cullling */
+   {
+      const GLfloat botDx = vMid->attrib[FRAG_ATTRIB_WPOS][0] - vMin->attrib[FRAG_ATTRIB_WPOS][0];
+      const GLfloat botDy = vMid->attrib[FRAG_ATTRIB_WPOS][1] - vMin->attrib[FRAG_ATTRIB_WPOS][1];
+      const GLfloat area = majDx * botDy - botDx * majDy;
+      /* Do backface culling */
+      if (area * bf < 0 || area == 0 || IS_INF_OR_NAN(area))
+	 return;
+      ltor = (GLboolean) (area < 0.0F);
+
+      span.facing = area * swrast->_BackfaceSign > 0.0F;
+   }
+
+   /* Plane equation setup:
+    * We evaluate plane equations at window (x,y) coordinates in order
+    * to compute color, Z, fog, texcoords, etc.  This isn't terribly
+    * efficient but it's easy and reliable.
+    */
+#ifdef DO_Z
+   compute_plane(p0, p1, p2, p0[2], p1[2], p2[2], zPlane);
+   span.arrayMask |= SPAN_Z;
+#endif
+   if (ctx->Light.ShadeModel == GL_SMOOTH) {
+      compute_plane(p0, p1, p2, v0->color[RCOMP], v1->color[RCOMP], v2->color[RCOMP], rPlane);
+      compute_plane(p0, p1, p2, v0->color[GCOMP], v1->color[GCOMP], v2->color[GCOMP], gPlane);
+      compute_plane(p0, p1, p2, v0->color[BCOMP], v1->color[BCOMP], v2->color[BCOMP], bPlane);
+      compute_plane(p0, p1, p2, v0->color[ACOMP], v1->color[ACOMP], v2->color[ACOMP], aPlane);
+   }
+   else {
+      constant_plane(v2->color[RCOMP], rPlane);
+      constant_plane(v2->color[GCOMP], gPlane);
+      constant_plane(v2->color[BCOMP], bPlane);
+      constant_plane(v2->color[ACOMP], aPlane);
+   }
+   span.arrayMask |= SPAN_RGBA;
+#if defined(DO_ATTRIBS)
+   {
+      const GLfloat invW0 = v0->attrib[FRAG_ATTRIB_WPOS][3];
+      const GLfloat invW1 = v1->attrib[FRAG_ATTRIB_WPOS][3];
+      const GLfloat invW2 = v2->attrib[FRAG_ATTRIB_WPOS][3];
+      compute_plane(p0, p1, p2, invW0, invW1, invW2, wPlane);
+      span.attrStepX[FRAG_ATTRIB_WPOS][3] = plane_dx(wPlane);
+      span.attrStepY[FRAG_ATTRIB_WPOS][3] = plane_dy(wPlane);
+      ATTRIB_LOOP_BEGIN
+         GLuint c;
+         if (swrast->_InterpMode[attr] == GL_FLAT) {
+            for (c = 0; c < 4; c++) {
+               constant_plane(v2->attrib[attr][c] * invW2, attrPlane[attr][c]);
+            }
+         }
+         else {
+            for (c = 0; c < 4; c++) {
+               const GLfloat a0 = v0->attrib[attr][c] * invW0;
+               const GLfloat a1 = v1->attrib[attr][c] * invW1;
+               const GLfloat a2 = v2->attrib[attr][c] * invW2;
+               compute_plane(p0, p1, p2, a0, a1, a2, attrPlane[attr][c]);
+            }
+         }
+         for (c = 0; c < 4; c++) {
+            span.attrStepX[attr][c] = plane_dx(attrPlane[attr][c]);
+            span.attrStepY[attr][c] = plane_dy(attrPlane[attr][c]);
+         }
+      ATTRIB_LOOP_END
+   }
+#endif
+
+   /* Begin bottom-to-top scan over the triangle.
+    * The long edge will either be on the left or right side of the
+    * triangle.  We always scan from the long edge toward the shorter
+    * edges, stopping when we find that coverage = 0.  If the long edge
+    * is on the left we scan left-to-right.  Else, we scan right-to-left.
+    */
+   yMin = vMin->attrib[FRAG_ATTRIB_WPOS][1];
+   yMax = vMax->attrib[FRAG_ATTRIB_WPOS][1];
+   iyMin = (GLint) yMin;
+   iyMax = (GLint) yMax + 1;
+
+   if (ltor) {
+      /* scan left to right */
+      const GLfloat *pMin = vMin->attrib[FRAG_ATTRIB_WPOS];
+      const GLfloat *pMid = vMid->attrib[FRAG_ATTRIB_WPOS];
+      const GLfloat *pMax = vMax->attrib[FRAG_ATTRIB_WPOS];
+      const GLfloat dxdy = majDx / majDy;
+      const GLfloat xAdj = dxdy < 0.0F ? -dxdy : 0.0F;
+      GLint iy;
+#ifdef _OPENMP
+#pragma omp parallel for schedule(dynamic) private(iy) firstprivate(span)
+#endif
+      for (iy = iyMin; iy < iyMax; iy++) {
+         GLfloat x = pMin[0] - (yMin - iy) * dxdy;
+         GLint ix, startX = (GLint) (x - xAdj);
+         GLuint count;
+         GLfloat coverage = 0.0F;
+
+#ifdef _OPENMP
+         /* each thread needs to use a different (global) SpanArrays variable */
+         span.array = SWRAST_CONTEXT(ctx)->SpanArrays + omp_get_thread_num();
+#endif
+         /* skip over fragments with zero coverage */
+         while (startX < MAX_WIDTH) {
+            coverage = compute_coveragef(pMin, pMid, pMax, startX, iy);
+            if (coverage > 0.0F)
+               break;
+            startX++;
+         }
+
+         /* enter interior of triangle */
+         ix = startX;
+
+#if defined(DO_ATTRIBS)
+         /* compute attributes at left-most fragment */
+         span.attrStart[FRAG_ATTRIB_WPOS][3] = solve_plane(ix + 0.5F, iy + 0.5F, wPlane);
+         ATTRIB_LOOP_BEGIN
+            GLuint c;
+            for (c = 0; c < 4; c++) {
+               span.attrStart[attr][c] = solve_plane(ix + 0.5F, iy + 0.5F, attrPlane[attr][c]);
+            }
+         ATTRIB_LOOP_END
+#endif
+
+         count = 0;
+         while (coverage > 0.0F) {
+            /* (cx,cy) = center of fragment */
+            const GLfloat cx = ix + 0.5F, cy = iy + 0.5F;
+            SWspanarrays *array = span.array;
+            array->coverage[count] = coverage;
+#ifdef DO_Z
+            array->z[count] = (GLuint) solve_plane(cx, cy, zPlane);
+#endif
+            array->rgba[count][RCOMP] = solve_plane_chan(cx, cy, rPlane);
+            array->rgba[count][GCOMP] = solve_plane_chan(cx, cy, gPlane);
+            array->rgba[count][BCOMP] = solve_plane_chan(cx, cy, bPlane);
+            array->rgba[count][ACOMP] = solve_plane_chan(cx, cy, aPlane);
+            ix++;
+            count++;
+            coverage = compute_coveragef(pMin, pMid, pMax, ix, iy);
+         }
+         
+         if (ix > startX) {
+            span.x = startX;
+            span.y = iy;
+            span.end = (GLuint) ix - (GLuint) startX;
+            _swrast_write_rgba_span(ctx, &span);
+         }
+      }
+   }
+   else {
+      /* scan right to left */
+      const GLfloat *pMin = vMin->attrib[FRAG_ATTRIB_WPOS];
+      const GLfloat *pMid = vMid->attrib[FRAG_ATTRIB_WPOS];
+      const GLfloat *pMax = vMax->attrib[FRAG_ATTRIB_WPOS];
+      const GLfloat dxdy = majDx / majDy;
+      const GLfloat xAdj = dxdy > 0 ? dxdy : 0.0F;
+      GLint iy;
+#ifdef _OPENMP
+#pragma omp parallel for schedule(dynamic) private(iy) firstprivate(span)
+#endif
+      for (iy = iyMin; iy < iyMax; iy++) {
+         GLfloat x = pMin[0] - (yMin - iy) * dxdy;
+         GLint ix, left, startX = (GLint) (x + xAdj);
+         GLuint count, n;
+         GLfloat coverage = 0.0F;
+         
+#ifdef _OPENMP
+         /* each thread needs to use a different (global) SpanArrays variable */
+         span.array = SWRAST_CONTEXT(ctx)->SpanArrays + omp_get_thread_num();
+#endif
+         /* make sure we're not past the window edge */
+         if (startX >= ctx->DrawBuffer->_Xmax) {
+            startX = ctx->DrawBuffer->_Xmax - 1;
+         }
+
+         /* skip fragments with zero coverage */
+         while (startX > 0) {
+            coverage = compute_coveragef(pMin, pMax, pMid, startX, iy);
+            if (coverage > 0.0F)
+               break;
+            startX--;
+         }
+         
+         /* enter interior of triangle */
+         ix = startX;
+         count = 0;
+         while (coverage > 0.0F) {
+            /* (cx,cy) = center of fragment */
+            const GLfloat cx = ix + 0.5F, cy = iy + 0.5F;
+            SWspanarrays *array = span.array;
+            ASSERT(ix >= 0);
+            array->coverage[ix] = coverage;
+#ifdef DO_Z
+            array->z[ix] = (GLuint) solve_plane(cx, cy, zPlane);
+#endif
+            array->rgba[ix][RCOMP] = solve_plane_chan(cx, cy, rPlane);
+            array->rgba[ix][GCOMP] = solve_plane_chan(cx, cy, gPlane);
+            array->rgba[ix][BCOMP] = solve_plane_chan(cx, cy, bPlane);
+            array->rgba[ix][ACOMP] = solve_plane_chan(cx, cy, aPlane);
+            ix--;
+            count++;
+            coverage = compute_coveragef(pMin, pMax, pMid, ix, iy);
+         }
+         
+#if defined(DO_ATTRIBS)
+         /* compute attributes at left-most fragment */
+         span.attrStart[FRAG_ATTRIB_WPOS][3] = solve_plane(ix + 1.5F, iy + 0.5F, wPlane);
+         ATTRIB_LOOP_BEGIN
+            GLuint c;
+            for (c = 0; c < 4; c++) {
+               span.attrStart[attr][c] = solve_plane(ix + 1.5F, iy + 0.5F, attrPlane[attr][c]);
+            }
+         ATTRIB_LOOP_END
+#endif
+
+         if (startX > ix) {
+            n = (GLuint) startX - (GLuint) ix;
+
+            left = ix + 1;
+
+            /* shift all values to the left */
+            /* XXX this is temporary */
+            {
+               SWspanarrays *array = span.array;
+               GLint j;
+               for (j = 0; j < (GLint) n; j++) {
+                  array->coverage[j] = array->coverage[j + left];
+                  COPY_CHAN4(array->rgba[j], array->rgba[j + left]);
+#ifdef DO_Z
+                  array->z[j] = array->z[j + left];
+#endif
+               }
+            }
+
+            span.x = left;
+            span.y = iy;
+            span.end = n;
+            _swrast_write_rgba_span(ctx, &span);
+         }
+      }
+   }
+}
+
+
+#undef DO_Z
+#undef DO_ATTRIBS
+#undef DO_OCCLUSION_TEST
diff --git a/mesalib/src/mesa/swrast/s_context.c b/mesalib/src/mesa/swrast/s_context.c
index d5f49923c..fbecbc55e 100644
--- a/mesalib/src/mesa/swrast/s_context.c
+++ b/mesalib/src/mesa/swrast/s_context.c
@@ -417,84 +417,6 @@ _swrast_validate_blend_func(struct gl_context *ctx, GLuint n, const GLubyte mask
    swrast->BlendFunc( ctx, n, mask, src, dst, chanType );

 }

 

-

-/**

- * Make sure we have texture image data for all the textures we may need

- * for subsequent rendering.

- */

-static void

-_swrast_validate_texture_images(struct gl_context *ctx)

-{

-   SWcontext *swrast = SWRAST_CONTEXT(ctx);

-   GLuint u;

-

-   if (!swrast->ValidateTextureImage || !ctx->Texture._EnabledUnits) {

-      /* no textures enabled, or no way to validate images! */

-      return;

-   }

-

-   for (u = 0; u < ctx->Const.MaxTextureImageUnits; u++) {

-      if (ctx->Texture.Unit[u]._ReallyEnabled) {

-         struct gl_texture_object *texObj = ctx->Texture.Unit[u]._Current;

-         ASSERT(texObj);

-         if (texObj) {

-            GLuint numFaces = (texObj->Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1;

-            GLuint face;

-            for (face = 0; face < numFaces; face++) {

-               GLint lvl;

-               for (lvl = texObj->BaseLevel; lvl <= texObj->_MaxLevel; lvl++) {

-                  struct gl_texture_image *texImg = texObj->Image[face][lvl];

-                  if (texImg && !texImg->Data) {

-                     swrast->ValidateTextureImage(ctx, texObj, face, lvl);

-                     ASSERT(texObj->Image[face][lvl]->Data);

-                  }

-               }

-            }

-         }

-      }

-   }

-}

-

-

-/**

- * Free the texture image data attached to all currently enabled

- * textures.  Meant to be called by device drivers when transitioning

- * from software to hardware rendering.

- */

-void

-_swrast_eject_texture_images(struct gl_context *ctx)

-{

-   GLuint u;

-

-   if (!ctx->Texture._EnabledUnits) {

-      /* no textures enabled */

-      return;

-   }

-

-   for (u = 0; u < ctx->Const.MaxTextureImageUnits; u++) {

-      if (ctx->Texture.Unit[u]._ReallyEnabled) {

-         struct gl_texture_object *texObj = ctx->Texture.Unit[u]._Current;

-         ASSERT(texObj);

-         if (texObj) {

-            GLuint numFaces = (texObj->Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1;

-            GLuint face;

-            for (face = 0; face < numFaces; face++) {

-               GLint lvl;

-               for (lvl = texObj->BaseLevel; lvl <= texObj->_MaxLevel; lvl++) {

-                  struct gl_texture_image *texImg = texObj->Image[face][lvl];

-                  if (texImg && texImg->Data) {

-                     _mesa_free_texmemory(texImg->Data);

-                     texImg->Data = NULL;

-                  }

-               }

-            }

-         }

-      }

-   }

-}

-

-

-

 static void

 _swrast_sleep( struct gl_context *ctx, GLbitfield new_state )

 {

@@ -640,7 +562,6 @@ _swrast_validate_derived( struct gl_context *ctx )
 

       if (swrast->NewState & (_NEW_TEXTURE | _NEW_PROGRAM)) {

          _swrast_update_texture_samplers( ctx );

-         _swrast_validate_texture_images(ctx);

       }

 

       if (swrast->NewState & (_NEW_COLOR | _NEW_PROGRAM))

@@ -772,6 +693,11 @@ _swrast_CreateContext( struct gl_context *ctx )
 {

    GLuint i;

    SWcontext *swrast = (SWcontext *)CALLOC(sizeof(SWcontext));

+#ifdef _OPENMP

+   const GLint maxThreads = omp_get_max_threads();

+#else

+   const GLint maxThreads = 1;

+#endif

 

    if (SWRAST_DEBUG) {

       _mesa_debug(ctx, "_swrast_CreateContext\n");

@@ -806,19 +732,25 @@ _swrast_CreateContext( struct gl_context *ctx )
    for (i = 0; i < MAX_TEXTURE_IMAGE_UNITS; i++)

       swrast->TextureSample[i] = NULL;

 

-   swrast->SpanArrays = MALLOC_STRUCT(sw_span_arrays);

+   /* SpanArrays is global and shared by all SWspan instances. However, when

+    * using multiple threads, it is necessary to have one SpanArrays instance

+    * per thread.

+    */

+   swrast->SpanArrays = (SWspanarrays *) MALLOC(maxThreads * sizeof(SWspanarrays));

    if (!swrast->SpanArrays) {

       FREE(swrast);

       return GL_FALSE;

    }

-   swrast->SpanArrays->ChanType = CHAN_TYPE;

+   for(i = 0; i < maxThreads; i++) {

+      swrast->SpanArrays[i].ChanType = CHAN_TYPE;

 #if CHAN_TYPE == GL_UNSIGNED_BYTE

-   swrast->SpanArrays->rgba = swrast->SpanArrays->rgba8;

+      swrast->SpanArrays[i].rgba = swrast->SpanArrays[i].rgba8;

 #elif CHAN_TYPE == GL_UNSIGNED_SHORT

-   swrast->SpanArrays->rgba = swrast->SpanArrays->rgba16;

+      swrast->SpanArrays[i].rgba = swrast->SpanArrays[i].rgba16;

 #else

-   swrast->SpanArrays->rgba = swrast->SpanArrays->attribs[FRAG_ATTRIB_COL0];

+      swrast->SpanArrays[i].rgba = swrast->SpanArrays[i].attribs[FRAG_ATTRIB_COL0];

 #endif

+   }

 

    /* init point span buffer */

    swrast->PointSpan.primitive = GL_POINT;

@@ -826,7 +758,10 @@ _swrast_CreateContext( struct gl_context *ctx )
    swrast->PointSpan.facing = 0;

    swrast->PointSpan.array = swrast->SpanArrays;

 

-   swrast->TexelBuffer = (GLfloat *) MALLOC(ctx->Const.MaxTextureImageUnits *

+   /* TexelBuffer is also global and normally shared by all SWspan instances;

+    * when running with multiple threads, create one per thread.

+    */

+   swrast->TexelBuffer = (GLfloat *) MALLOC(ctx->Const.MaxTextureImageUnits * maxThreads *

                                            MAX_WIDTH * 4 * sizeof(GLfloat));

    if (!swrast->TexelBuffer) {

       FREE(swrast->SpanArrays);

diff --git a/mesalib/src/mesa/swrast/s_stencil.c b/mesalib/src/mesa/swrast/s_stencil.c
index 999fe3c3f..fa5093a34 100644
--- a/mesalib/src/mesa/swrast/s_stencil.c
+++ b/mesalib/src/mesa/swrast/s_stencil.c
@@ -1,1245 +1,1246 @@
-/*

- * Mesa 3-D graphics library

- * Version:  7.1

- *

- * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.

- *

- * Permission is hereby granted, free of charge, to any person obtaining a

- * copy of this software and associated documentation files (the "Software"),

- * 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/imports.h"

-

-#include "s_context.h"

-#include "s_depth.h"

-#include "s_stencil.h"

-#include "s_span.h"

-

-

-

-/* Stencil Logic:

-

-IF stencil test fails THEN

-   Apply fail-op to stencil value

-   Don't write the pixel (RGBA,Z)

-ELSE

-   IF doing depth test && depth test fails THEN

-      Apply zfail-op to stencil value

-      Write RGBA and Z to appropriate buffers

-   ELSE

-      Apply zpass-op to stencil value

-ENDIF

-

-*/

-

-

-/**

- * Apply the given stencil operator to the array of stencil values.

- * Don't touch stencil[i] if mask[i] is zero.

- * Input:  n - size of stencil array

- *         oper - the stencil buffer operator

- *         face - 0 or 1 for front or back face operation

- *         stencil - array of stencil values

- *         mask - array [n] of flag:  1=apply operator, 0=don't apply operator

- * Output:  stencil - modified values

- */

-static void

-apply_stencil_op( const struct gl_context *ctx, GLenum oper, GLuint face,

-                  GLuint n, GLstencil stencil[], const GLubyte mask[] )

-{

-   const GLstencil ref = ctx->Stencil.Ref[face];

-   const GLstencil wrtmask = ctx->Stencil.WriteMask[face];

-   const GLstencil invmask = (GLstencil) (~wrtmask);

-   const GLstencil stencilMax = (1 << ctx->DrawBuffer->Visual.stencilBits) - 1;

-   GLuint i;

-

-   switch (oper) {

-      case GL_KEEP:

-         /* do nothing */

-         break;

-      case GL_ZERO:

-	 if (invmask==0) {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-		  stencil[i] = 0;

-	       }

-	    }

-	 }

-	 else {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-		  stencil[i] = (GLstencil) (stencil[i] & invmask);

-	       }

-	    }

-	 }

-	 break;

-      case GL_REPLACE:

-	 if (invmask==0) {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-                  stencil[i] = ref;

-	       }

-	    }

-	 }

-	 else {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-		  GLstencil s = stencil[i];

-		  stencil[i] = (GLstencil) ((invmask & s ) | (wrtmask & ref));

-	       }

-	    }

-	 }

-	 break;

-      case GL_INCR:

-	 if (invmask==0) {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-		  GLstencil s = stencil[i];

-		  if (s < stencilMax) {

-		     stencil[i] = (GLstencil) (s+1);

-		  }

-	       }

-	    }

-	 }

-	 else {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-		  /* VERIFY logic of adding 1 to a write-masked value */

-		  GLstencil s = stencil[i];

-		  if (s < stencilMax) {

-		     stencil[i] = (GLstencil) ((invmask & s) | (wrtmask & (s+1)));

-		  }

-	       }

-	    }

-	 }

-	 break;

-      case GL_DECR:

-	 if (invmask==0) {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-		  GLstencil s = stencil[i];

-		  if (s>0) {

-		     stencil[i] = (GLstencil) (s-1);

-		  }

-	       }

-	    }

-	 }

-	 else {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-		  /* VERIFY logic of subtracting 1 to a write-masked value */

-		  GLstencil s = stencil[i];

-		  if (s>0) {

-		     stencil[i] = (GLstencil) ((invmask & s) | (wrtmask & (s-1)));

-		  }

-	       }

-	    }

-	 }

-	 break;

-      case GL_INCR_WRAP_EXT:

-	 if (invmask==0) {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-                  stencil[i]++;

-	       }

-	    }

-	 }

-	 else {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-                  GLstencil s = stencil[i];

-                  stencil[i] = (GLstencil) ((invmask & s) | (wrtmask & (s+1)));

-	       }

-	    }

-	 }

-	 break;

-      case GL_DECR_WRAP_EXT:

-	 if (invmask==0) {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-		  stencil[i]--;

-	       }

-	    }

-	 }

-	 else {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-                  GLstencil s = stencil[i];

-                  stencil[i] = (GLstencil) ((invmask & s) | (wrtmask & (s-1)));

-	       }

-	    }

-	 }

-	 break;

-      case GL_INVERT:

-	 if (invmask==0) {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-		  GLstencil s = stencil[i];

-		  stencil[i] = (GLstencil) ~s;

-	       }

-	    }

-	 }

-	 else {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-		  GLstencil s = stencil[i];

-		  stencil[i] = (GLstencil) ((invmask & s) | (wrtmask & ~s));

-	       }

-	    }

-	 }

-	 break;

-      default:

-         _mesa_problem(ctx, "Bad stencil op in apply_stencil_op");

-   }

-}

-

-

-

-

-/**

- * Apply stencil test to an array of stencil values (before depth buffering).

- * Input:  face - 0 or 1 for front or back-face polygons

- *         n - number of pixels in the array

- *         stencil - array of [n] stencil values

- *         mask - array [n] of flag:  0=skip the pixel, 1=stencil the pixel

- * Output:  mask - pixels which fail the stencil test will have their

- *                 mask flag set to 0.

- *          stencil - updated stencil values (where the test passed)

- * Return:  GL_FALSE = all pixels failed, GL_TRUE = zero or more pixels passed.

- */

-static GLboolean

-do_stencil_test( struct gl_context *ctx, GLuint face, GLuint n, GLstencil stencil[],

-                 GLubyte mask[] )

-{

-   GLubyte fail[MAX_WIDTH];

-   GLboolean allfail = GL_FALSE;

-   GLuint i;

-   const GLuint valueMask = ctx->Stencil.ValueMask[face];

-   const GLstencil r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);

-   GLstencil s;

-

-   ASSERT(n <= MAX_WIDTH);

-

-   /*

-    * Perform stencil test.  The results of this operation are stored

-    * in the fail[] array:

-    *   IF fail[i] is non-zero THEN

-    *       the stencil fail operator is to be applied

-    *   ELSE

-    *       the stencil fail operator is not to be applied

-    *   ENDIF

-    */

-   switch (ctx->Stencil.Function[face]) {

-      case GL_NEVER:

-         /* never pass; always fail */

-         for (i=0;i<n;i++) {

-	    if (mask[i]) {

-	       mask[i] = 0;

-	       fail[i] = 1;

-	    }

-	    else {

-	       fail[i] = 0;

-	    }

-	 }

-	 allfail = GL_TRUE;

-	 break;

-      case GL_LESS:

-	 for (i=0;i<n;i++) {

-	    if (mask[i]) {

-	       s = (GLstencil) (stencil[i] & valueMask);

-	       if (r < s) {

-		  /* passed */

-		  fail[i] = 0;

-	       }

-	       else {

-		  fail[i] = 1;

-		  mask[i] = 0;

-	       }

-	    }

-	    else {

-	       fail[i] = 0;

-	    }

-	 }

-	 break;

-      case GL_LEQUAL:

-	 for (i=0;i<n;i++) {

-	    if (mask[i]) {

-	       s = (GLstencil) (stencil[i] & valueMask);

-	       if (r <= s) {

-		  /* pass */

-		  fail[i] = 0;

-	       }

-	       else {

-		  fail[i] = 1;

-		  mask[i] = 0;

-	       }

-	    }

-	    else {

-	       fail[i] = 0;

-	    }

-	 }

-	 break;

-      case GL_GREATER:

-	 for (i=0;i<n;i++) {

-	    if (mask[i]) {

-	       s = (GLstencil) (stencil[i] & valueMask);

-	       if (r > s) {

-		  /* passed */

-		  fail[i] = 0;

-	       }

-	       else {

-		  fail[i] = 1;

-		  mask[i] = 0;

-	       }

-	    }

-	    else {

-	       fail[i] = 0;

-	    }

-	 }

-	 break;

-      case GL_GEQUAL:

-	 for (i=0;i<n;i++) {

-	    if (mask[i]) {

-	       s = (GLstencil) (stencil[i] & valueMask);

-	       if (r >= s) {

-		  /* passed */

-		  fail[i] = 0;

-	       }

-	       else {

-		  fail[i] = 1;

-		  mask[i] = 0;

-	       }

-	    }

-	    else {

-	       fail[i] = 0;

-	    }

-	 }

-	 break;

-      case GL_EQUAL:

-	 for (i=0;i<n;i++) {

-	    if (mask[i]) {

-	       s = (GLstencil) (stencil[i] & valueMask);

-	       if (r == s) {

-		  /* passed */

-		  fail[i] = 0;

-	       }

-	       else {

-		  fail[i] = 1;

-		  mask[i] = 0;

-	       }

-	    }

-	    else {

-	       fail[i] = 0;

-	    }

-	 }

-	 break;

-      case GL_NOTEQUAL:

-	 for (i=0;i<n;i++) {

-	    if (mask[i]) {

-	       s = (GLstencil) (stencil[i] & valueMask);

-	       if (r != s) {

-		  /* passed */

-		  fail[i] = 0;

-	       }

-	       else {

-		  fail[i] = 1;

-		  mask[i] = 0;

-	       }

-	    }

-	    else {

-	       fail[i] = 0;

-	    }

-	 }

-	 break;

-      case GL_ALWAYS:

-	 /* always pass */

-	 for (i=0;i<n;i++) {

-	    fail[i] = 0;

-	 }

-	 break;

-      default:

-         _mesa_problem(ctx, "Bad stencil func in gl_stencil_span");

-         return 0;

-   }

-

-   if (ctx->Stencil.FailFunc[face] != GL_KEEP) {

-      apply_stencil_op( ctx, ctx->Stencil.FailFunc[face], face, n, stencil, fail );

-   }

-

-   return !allfail;

-}

-

-

-/**

- * Compute the zpass/zfail masks by comparing the pre- and post-depth test

- * masks.

- */

-static INLINE void

-compute_pass_fail_masks(GLuint n, const GLubyte origMask[],

-                        const GLubyte newMask[],

-                        GLubyte passMask[], GLubyte failMask[])

-{

-   GLuint i;

-   for (i = 0; i < n; i++) {

-      ASSERT(newMask[i] == 0 || newMask[i] == 1);

-      passMask[i] = origMask[i] & newMask[i];

-      failMask[i] = origMask[i] & (newMask[i] ^ 1);

-   }

-}

-

-

-/**

- * Apply stencil and depth testing to the span of pixels.

- * Both software and hardware stencil buffers are acceptable.

- * Input:  n - number of pixels in the span

- *         x, y - location of leftmost pixel in span

- *         z - array [n] of z values

- *         mask - array [n] of flags  (1=test this pixel, 0=skip the pixel)

- * Output:  mask - array [n] of flags (1=stencil and depth test passed)

- * Return: GL_FALSE - all fragments failed the testing

- *         GL_TRUE - one or more fragments passed the testing

- *

- */

-static GLboolean

-stencil_and_ztest_span(struct gl_context *ctx, SWspan *span, GLuint face)

-{

-   struct gl_framebuffer *fb = ctx->DrawBuffer;

-   struct gl_renderbuffer *rb = fb->_StencilBuffer;

-   GLstencil stencilRow[MAX_WIDTH];

-   GLstencil *stencil;

-   const GLuint n = span->end;

-   const GLint x = span->x;

-   const GLint y = span->y;

-   GLubyte *mask = span->array->mask;

-

-   ASSERT((span->arrayMask & SPAN_XY) == 0);

-   ASSERT(ctx->Stencil.Enabled);

-   ASSERT(n <= MAX_WIDTH);

-#ifdef DEBUG

-   if (ctx->Depth.Test) {

-      ASSERT(span->arrayMask & SPAN_Z);

-   }

-#endif

-

-   stencil = (GLstencil *) rb->GetPointer(ctx, rb, x, y);

-   if (!stencil) {

-      rb->GetRow(ctx, rb, n, x, y, stencilRow);

-      stencil = stencilRow;

-   }

-

-   /*

-    * Apply the stencil test to the fragments.

-    * failMask[i] is 1 if the stencil test failed.

-    */

-   if (do_stencil_test( ctx, face, n, stencil, mask ) == GL_FALSE) {

-      /* all fragments failed the stencil test, we're done. */

-      span->writeAll = GL_FALSE;

-      if (!rb->GetPointer(ctx, rb, 0, 0)) {

-         /* put updated stencil values into buffer */

-         rb->PutRow(ctx, rb, n, x, y, stencil, NULL);

-      }

-      return GL_FALSE;

-   }

-

-   /*

-    * Some fragments passed the stencil test, apply depth test to them

-    * and apply Zpass and Zfail stencil ops.

-    */

-   if (ctx->Depth.Test == GL_FALSE) {

-      /*

-       * No depth buffer, just apply zpass stencil function to active pixels.

-       */

-      apply_stencil_op( ctx, ctx->Stencil.ZPassFunc[face], face, n, stencil, mask );

-   }

-   else {

-      /*

-       * Perform depth buffering, then apply zpass or zfail stencil function.

-       */

-      GLubyte passMask[MAX_WIDTH], failMask[MAX_WIDTH], origMask[MAX_WIDTH];

-

-      /* save the current mask bits */

-      memcpy(origMask, mask, n * sizeof(GLubyte));

-

-      /* apply the depth test */

-      _swrast_depth_test_span(ctx, span);

-

-      compute_pass_fail_masks(n, origMask, mask, passMask, failMask);

-

-      /* apply the pass and fail operations */

-      if (ctx->Stencil.ZFailFunc[face] != GL_KEEP) {

-         apply_stencil_op( ctx, ctx->Stencil.ZFailFunc[face], face,

-                           n, stencil, failMask );

-      }

-      if (ctx->Stencil.ZPassFunc[face] != GL_KEEP) {

-         apply_stencil_op( ctx, ctx->Stencil.ZPassFunc[face], face,

-                           n, stencil, passMask );

-      }

-   }

-

-   /*

-    * Write updated stencil values back into hardware stencil buffer.

-    */

-   if (!rb->GetPointer(ctx, rb, 0, 0)) {

-      rb->PutRow(ctx, rb, n, x, y, stencil, NULL);

-   }

-   

-   span->writeAll = GL_FALSE;

-   

-   return GL_TRUE;  /* one or more fragments passed both tests */

-}

-

-

-

-/*

- * Return the address of a stencil buffer value given the window coords:

- */

-#define STENCIL_ADDRESS(X, Y)  (stencilStart + (Y) * stride + (X))

-

-

-

-/**

- * Apply the given stencil operator for each pixel in the array whose

- * mask flag is set.

- * \note  This is for software stencil buffers only.

- * Input:  n - number of pixels in the span

- *         x, y - array of [n] pixels

- *         operator - the stencil buffer operator

- *         mask - array [n] of flag:  1=apply operator, 0=don't apply operator

- */

-static void

-apply_stencil_op_to_pixels( struct gl_context *ctx,

-                            GLuint n, const GLint x[], const GLint y[],

-                            GLenum oper, GLuint face, const GLubyte mask[] )

-{

-   struct gl_framebuffer *fb = ctx->DrawBuffer;

-   struct gl_renderbuffer *rb = fb->_StencilBuffer;

-   const GLstencil stencilMax = (1 << fb->Visual.stencilBits) - 1;

-   const GLstencil ref = ctx->Stencil.Ref[face];

-   const GLstencil wrtmask = ctx->Stencil.WriteMask[face];

-   const GLstencil invmask = (GLstencil) (~wrtmask);

-   GLuint i;

-   GLstencil *stencilStart = (GLubyte *) rb->Data;

-   const GLuint stride = rb->Width;

-

-   ASSERT(rb->GetPointer(ctx, rb, 0, 0));

-   ASSERT(sizeof(GLstencil) == 1);

-

-   switch (oper) {

-      case GL_KEEP:

-         /* do nothing */

-         break;

-      case GL_ZERO:

-	 if (invmask==0) {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );

-                  *sptr = 0;

-	       }

-	    }

-	 }

-	 else {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );

-		  *sptr = (GLstencil) (invmask & *sptr);

-	       }

-	    }

-	 }

-	 break;

-      case GL_REPLACE:

-	 if (invmask==0) {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );

-                  *sptr = ref;

-	       }

-	    }

-	 }

-	 else {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );

-		  *sptr = (GLstencil) ((invmask & *sptr ) | (wrtmask & ref));

-	       }

-	    }

-	 }

-	 break;

-      case GL_INCR:

-	 if (invmask==0) {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );

-		  if (*sptr < stencilMax) {

-		     *sptr = (GLstencil) (*sptr + 1);

-		  }

-	       }

-	    }

-	 }

-	 else {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );

-		  if (*sptr < stencilMax) {

-		     *sptr = (GLstencil) ((invmask & *sptr) | (wrtmask & (*sptr+1)));

-		  }

-	       }

-	    }

-	 }

-	 break;

-      case GL_DECR:

-	 if (invmask==0) {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );

-		  if (*sptr>0) {

-		     *sptr = (GLstencil) (*sptr - 1);

-		  }

-	       }

-	    }

-	 }

-	 else {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );

-		  if (*sptr>0) {

-		     *sptr = (GLstencil) ((invmask & *sptr) | (wrtmask & (*sptr-1)));

-		  }

-	       }

-	    }

-	 }

-	 break;

-      case GL_INCR_WRAP_EXT:

-	 if (invmask==0) {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );

-                  *sptr = (GLstencil) (*sptr + 1);

-	       }

-	    }

-	 }

-	 else {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );

-                  *sptr = (GLstencil) ((invmask & *sptr) | (wrtmask & (*sptr+1)));

-	       }

-	    }

-	 }

-	 break;

-      case GL_DECR_WRAP_EXT:

-	 if (invmask==0) {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );

-                  *sptr = (GLstencil) (*sptr - 1);

-	       }

-	    }

-	 }

-	 else {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );

-                  *sptr = (GLstencil) ((invmask & *sptr) | (wrtmask & (*sptr-1)));

-	       }

-	    }

-	 }

-	 break;

-      case GL_INVERT:

-	 if (invmask==0) {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );

-                  *sptr = (GLstencil) (~*sptr);

-	       }

-	    }

-	 }

-	 else {

-	    for (i=0;i<n;i++) {

-	       if (mask[i]) {

-                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );

-                  *sptr = (GLstencil) ((invmask & *sptr) | (wrtmask & ~*sptr));

-	       }

-	    }

-	 }

-	 break;

-      default:

-         _mesa_problem(ctx, "Bad stencilop in apply_stencil_op_to_pixels");

-   }

-}

-

-

-

-/**

- * Apply stencil test to an array of pixels before depth buffering.

- *

- * \note Used for software stencil buffer only.

- * Input:  n - number of pixels in the span

- *         x, y - array of [n] pixels to stencil

- *         mask - array [n] of flag:  0=skip the pixel, 1=stencil the pixel

- * Output:  mask - pixels which fail the stencil test will have their

- *                 mask flag set to 0.

- * \return  GL_FALSE = all pixels failed, GL_TRUE = zero or more pixels passed.

- */

-static GLboolean

-stencil_test_pixels( struct gl_context *ctx, GLuint face, GLuint n,

-                     const GLint x[], const GLint y[], GLubyte mask[] )

-{

-   const struct gl_framebuffer *fb = ctx->DrawBuffer;

-   struct gl_renderbuffer *rb = fb->_StencilBuffer;

-   GLubyte fail[MAX_WIDTH];

-   GLstencil r, s;

-   GLuint i;

-   GLboolean allfail = GL_FALSE;

-   const GLuint valueMask = ctx->Stencil.ValueMask[face];

-   const GLstencil *stencilStart = (GLstencil *) rb->Data;

-   const GLuint stride = rb->Width;

-

-   ASSERT(rb->GetPointer(ctx, rb, 0, 0));

-   ASSERT(sizeof(GLstencil) == 1);

-

-   /*

-    * Perform stencil test.  The results of this operation are stored

-    * in the fail[] array:

-    *   IF fail[i] is non-zero THEN

-    *       the stencil fail operator is to be applied

-    *   ELSE

-    *       the stencil fail operator is not to be applied

-    *   ENDIF

-    */

-

-   switch (ctx->Stencil.Function[face]) {

-      case GL_NEVER:

-         /* always fail */

-         for (i=0;i<n;i++) {

-	    if (mask[i]) {

-	       mask[i] = 0;

-	       fail[i] = 1;

-	    }

-	    else {

-	       fail[i] = 0;

-	    }

-	 }

-	 allfail = GL_TRUE;

-	 break;

-      case GL_LESS:

-	 r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);

-	 for (i=0;i<n;i++) {

-	    if (mask[i]) {

-               const GLstencil *sptr = STENCIL_ADDRESS(x[i],y[i]);

-	       s = (GLstencil) (*sptr & valueMask);

-	       if (r < s) {

-		  /* passed */

-		  fail[i] = 0;

-	       }

-	       else {

-		  fail[i] = 1;

-		  mask[i] = 0;

-	       }

-	    }

-	    else {

-	       fail[i] = 0;

-	    }

-	 }

-	 break;

-      case GL_LEQUAL:

-	 r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);

-	 for (i=0;i<n;i++) {

-	    if (mask[i]) {

-               const GLstencil *sptr = STENCIL_ADDRESS(x[i],y[i]);

-	       s = (GLstencil) (*sptr & valueMask);

-	       if (r <= s) {

-		  /* pass */

-		  fail[i] = 0;

-	       }

-	       else {

-		  fail[i] = 1;

-		  mask[i] = 0;

-	       }

-	    }

-	    else {

-	       fail[i] = 0;

-	    }

-	 }

-	 break;

-      case GL_GREATER:

-	 r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);

-	 for (i=0;i<n;i++) {

-	    if (mask[i]) {

-               const GLstencil *sptr = STENCIL_ADDRESS(x[i],y[i]);

-	       s = (GLstencil) (*sptr & valueMask);

-	       if (r > s) {

-		  /* passed */

-		  fail[i] = 0;

-	       }

-	       else {

-		  fail[i] = 1;

-		  mask[i] = 0;

-	       }

-	    }

-	    else {

-	       fail[i] = 0;

-	    }

-	 }

-	 break;

-      case GL_GEQUAL:

-	 r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);

-	 for (i=0;i<n;i++) {

-	    if (mask[i]) {

-               const GLstencil *sptr = STENCIL_ADDRESS(x[i],y[i]);

-	       s = (GLstencil) (*sptr & valueMask);

-	       if (r >= s) {

-		  /* passed */

-		  fail[i] = 0;

-	       }

-	       else {

-		  fail[i] = 1;

-		  mask[i] = 0;

-	       }

-	    }

-	    else {

-	       fail[i] = 0;

-	    }

-	 }

-	 break;

-      case GL_EQUAL:

-	 r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);

-	 for (i=0;i<n;i++) {

-	    if (mask[i]) {

-               const GLstencil *sptr = STENCIL_ADDRESS(x[i],y[i]);

-	       s = (GLstencil) (*sptr & valueMask);

-	       if (r == s) {

-		  /* passed */

-		  fail[i] = 0;

-	       }

-	       else {

-		  fail[i] = 1;

-		  mask[i] = 0;

-	       }

-	    }

-	    else {

-	       fail[i] = 0;

-	    }

-	 }

-	 break;

-      case GL_NOTEQUAL:

-	 r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);

-	 for (i=0;i<n;i++) {

-	    if (mask[i]) {

-               const GLstencil *sptr = STENCIL_ADDRESS(x[i],y[i]);

-	       s = (GLstencil) (*sptr & valueMask);

-	       if (r != s) {

-		  /* passed */

-		  fail[i] = 0;

-	       }

-	       else {

-		  fail[i] = 1;

-		  mask[i] = 0;

-	       }

-	    }

-	    else {

-	       fail[i] = 0;

-	    }

-	 }

-	 break;

-      case GL_ALWAYS:

-	 /* always pass */

-	 for (i=0;i<n;i++) {

-	    fail[i] = 0;

-	 }

-	 break;

-      default:

-         _mesa_problem(ctx, "Bad stencil func in gl_stencil_pixels");

-         return 0;

-   }

-

-   if (ctx->Stencil.FailFunc[face] != GL_KEEP) {

-      apply_stencil_op_to_pixels( ctx, n, x, y, ctx->Stencil.FailFunc[face],

-                                  face, fail );

-   }

-

-   return !allfail;

-}

-

-

-

-

-/**

- * Apply stencil and depth testing to an array of pixels.

- * This is used both for software and hardware stencil buffers.

- *

- * The comments in this function are a bit sparse but the code is

- * almost identical to stencil_and_ztest_span(), which is well

- * commented.

- *

- * Input:  n - number of pixels in the array

- *         x, y - array of [n] pixel positions

- *         z - array [n] of z values

- *         mask - array [n] of flags  (1=test this pixel, 0=skip the pixel)

- * Output: mask - array [n] of flags (1=stencil and depth test passed)

- * Return: GL_FALSE - all fragments failed the testing

- *         GL_TRUE - one or more fragments passed the testing

- */

-static GLboolean

-stencil_and_ztest_pixels( struct gl_context *ctx, SWspan *span, GLuint face )

-{

-   GLubyte passMask[MAX_WIDTH], failMask[MAX_WIDTH], origMask[MAX_WIDTH];

-   struct gl_framebuffer *fb = ctx->DrawBuffer;

-   struct gl_renderbuffer *rb = fb->_StencilBuffer;

-   const GLuint n = span->end;

-   const GLint *x = span->array->x;

-   const GLint *y = span->array->y;

-   GLubyte *mask = span->array->mask;

-

-   ASSERT(span->arrayMask & SPAN_XY);

-   ASSERT(ctx->Stencil.Enabled);

-   ASSERT(n <= MAX_WIDTH);

-

-   if (!rb->GetPointer(ctx, rb, 0, 0)) {

-      /* No direct access */

-      GLstencil stencil[MAX_WIDTH];

-

-      ASSERT(rb->DataType == GL_UNSIGNED_BYTE);

-      _swrast_get_values(ctx, rb, n, x, y, stencil, sizeof(GLubyte));

-

-      memcpy(origMask, mask, n * sizeof(GLubyte));          

-

-      (void) do_stencil_test(ctx, face, n, stencil, mask);

-

-      if (ctx->Depth.Test == GL_FALSE) {

-         apply_stencil_op(ctx, ctx->Stencil.ZPassFunc[face], face,

-                          n, stencil, mask);

-      }

-      else {

-         GLubyte tmpMask[MAX_WIDTH]; 

-         memcpy(tmpMask, mask, n * sizeof(GLubyte));

-

-         _swrast_depth_test_span(ctx, span);

-

-         compute_pass_fail_masks(n, tmpMask, mask, passMask, failMask);

-

-         if (ctx->Stencil.ZFailFunc[face] != GL_KEEP) {

-            apply_stencil_op(ctx, ctx->Stencil.ZFailFunc[face], face,

-                             n, stencil, failMask);

-         }

-         if (ctx->Stencil.ZPassFunc[face] != GL_KEEP) {

-            apply_stencil_op(ctx, ctx->Stencil.ZPassFunc[face], face,

-                             n, stencil, passMask);

-         }

-      }

-

-      /* Write updated stencil values into hardware stencil buffer */

-      rb->PutValues(ctx, rb, n, x, y, stencil, origMask);

-

-      return GL_TRUE;

-   }

-   else {

-      /* Direct access to stencil buffer */

-

-      if (stencil_test_pixels(ctx, face, n, x, y, mask) == GL_FALSE) {

-         /* all fragments failed the stencil test, we're done. */

-         return GL_FALSE;

-      }

-

-      if (ctx->Depth.Test==GL_FALSE) {

-         apply_stencil_op_to_pixels(ctx, n, x, y,

-                                    ctx->Stencil.ZPassFunc[face], face, mask);

-      }

-      else {

-         memcpy(origMask, mask, n * sizeof(GLubyte));

-

-         _swrast_depth_test_span(ctx, span);

-

-         compute_pass_fail_masks(n, origMask, mask, passMask, failMask);

-

-         if (ctx->Stencil.ZFailFunc[face] != GL_KEEP) {

-            apply_stencil_op_to_pixels(ctx, n, x, y,

-                                       ctx->Stencil.ZFailFunc[face],

-                                       face, failMask);

-         }

-         if (ctx->Stencil.ZPassFunc[face] != GL_KEEP) {

-            apply_stencil_op_to_pixels(ctx, n, x, y,

-                                       ctx->Stencil.ZPassFunc[face],

-                                       face, passMask);

-         }

-      }

-

-      return GL_TRUE;  /* one or more fragments passed both tests */

-   }

-}

-

-

-/**

- * /return GL_TRUE = one or more fragments passed,

- * GL_FALSE = all fragments failed.

- */

-GLboolean

-_swrast_stencil_and_ztest_span(struct gl_context *ctx, SWspan *span)

-{

-   const GLuint face = (span->facing == 0) ? 0 : ctx->Stencil._BackFace;

-

-   if (span->arrayMask & SPAN_XY)

-      return stencil_and_ztest_pixels(ctx, span, face);

-   else

-      return stencil_and_ztest_span(ctx, span, face);

-}

-

-

-#if 0

-GLuint

-clip_span(GLuint bufferWidth, GLuint bufferHeight,

-          GLint x, GLint y, GLuint *count)

-{

-   GLuint n = *count;

-   GLuint skipPixels = 0;

-

-   if (y < 0 || y >= bufferHeight || x + n <= 0 || x >= bufferWidth) {

-      /* totally out of bounds */

-      n = 0;

-   }

-   else {

-      /* left clip */

-      if (x < 0) {

-         skipPixels = -x;

-         x = 0;

-         n -= skipPixels;

-      }

-      /* right clip */

-      if (x + n > bufferWidth) {

-         GLint dx = x + n - bufferWidth;

-         n -= dx;

-      }

-   }

-

-   *count = n;

-

-   return skipPixels;

-}

-#endif

-

-

-/**

- * Return a span of stencil values from the stencil buffer.

- * Used for glRead/CopyPixels

- * Input:  n - how many pixels

- *         x,y - location of first pixel

- * Output:  stencil - the array of stencil values

- */

-void

-_swrast_read_stencil_span(struct gl_context *ctx, struct gl_renderbuffer *rb,

-                          GLint n, GLint x, GLint y, GLstencil stencil[])

-{

-   if (y < 0 || y >= (GLint) rb->Height ||

-       x + n <= 0 || x >= (GLint) rb->Width) {

-      /* span is completely outside framebuffer */

-      return; /* undefined values OK */

-   }

-

-   if (x < 0) {

-      GLint dx = -x;

-      x = 0;

-      n -= dx;

-      stencil += dx;

-   }

-   if (x + n > (GLint) rb->Width) {

-      GLint dx = x + n - rb->Width;

-      n -= dx;

-   }

-   if (n <= 0) {

-      return;

-   }

-

-   rb->GetRow(ctx, rb, n, x, y, stencil);

-}

-

-

-

-/**

- * Write a span of stencil values to the stencil buffer.  This function

- * applies the stencil write mask when needed.

- * Used for glDraw/CopyPixels

- * Input:  n - how many pixels

- *         x, y - location of first pixel

- *         stencil - the array of stencil values

- */

-void

-_swrast_write_stencil_span(struct gl_context *ctx, GLint n, GLint x, GLint y,

-                           const GLstencil stencil[] )

-{

-   struct gl_framebuffer *fb = ctx->DrawBuffer;

-   struct gl_renderbuffer *rb = fb->_StencilBuffer;

-   const GLuint stencilMax = (1 << fb->Visual.stencilBits) - 1;

-   const GLuint stencilMask = ctx->Stencil.WriteMask[0];

-

-   if (y < 0 || y >= (GLint) rb->Height ||

-       x + n <= 0 || x >= (GLint) rb->Width) {

-      /* span is completely outside framebuffer */

-      return; /* undefined values OK */

-   }

-   if (x < 0) {

-      GLint dx = -x;

-      x = 0;

-      n -= dx;

-      stencil += dx;

-   }

-   if (x + n > (GLint) rb->Width) {

-      GLint dx = x + n - rb->Width;

-      n -= dx;

-   }

-   if (n <= 0) {

-      return;

-   }

-

-   if ((stencilMask & stencilMax) != stencilMax) {

-      /* need to apply writemask */

-      GLstencil destVals[MAX_WIDTH], newVals[MAX_WIDTH];

-      GLint i;

-      rb->GetRow(ctx, rb, n, x, y, destVals);

-      for (i = 0; i < n; i++) {

-         newVals[i]

-            = (stencil[i] & stencilMask) | (destVals[i] & ~stencilMask);

-      }

-      rb->PutRow(ctx, rb, n, x, y, newVals, NULL);

-   }

-   else {

-      rb->PutRow(ctx, rb, n, x, y, stencil, NULL);

-   }

-}

-

-

-

-/**

- * Clear the stencil buffer.

- */

-void

-_swrast_clear_stencil_buffer( struct gl_context *ctx, struct gl_renderbuffer *rb )

-{

-   const GLubyte stencilBits = ctx->DrawBuffer->Visual.stencilBits;

-   const GLuint mask = ctx->Stencil.WriteMask[0];

-   const GLuint invMask = ~mask;

-   const GLuint clearVal = (ctx->Stencil.Clear & mask);

-   const GLuint stencilMax = (1 << stencilBits) - 1;

-   GLint x, y, width, height;

-

-   if (!rb || mask == 0)

-      return;

-

-   ASSERT(rb->DataType == GL_UNSIGNED_BYTE ||

-          rb->DataType == GL_UNSIGNED_SHORT);

-

-   ASSERT(rb->_BaseFormat == GL_STENCIL_INDEX);

-

-   /* compute region to clear */

-   x = ctx->DrawBuffer->_Xmin;

-   y = ctx->DrawBuffer->_Ymin;

-   width  = ctx->DrawBuffer->_Xmax - ctx->DrawBuffer->_Xmin;

-   height = ctx->DrawBuffer->_Ymax - ctx->DrawBuffer->_Ymin;

-

-   if (rb->GetPointer(ctx, rb, 0, 0)) {

-      /* Direct buffer access */

-      if ((mask & stencilMax) != stencilMax) {

-         /* need to mask the clear */

-         if (rb->DataType == GL_UNSIGNED_BYTE) {

-            GLint i, j;

-            for (i = 0; i < height; i++) {

-               GLubyte *stencil = (GLubyte*) rb->GetPointer(ctx, rb, x, y + i);

-               for (j = 0; j < width; j++) {

-                  stencil[j] = (stencil[j] & invMask) | clearVal;

-               }

-            }

-         }

-         else {

-            GLint i, j;

-            for (i = 0; i < height; i++) {

-               GLushort *stencil = (GLushort*) rb->GetPointer(ctx, rb, x, y + i);

-               for (j = 0; j < width; j++) {

-                  stencil[j] = (stencil[j] & invMask) | clearVal;

-               }

-            }

-         }

-      }

-      else {

-         /* no bit masking */

-         if (width == (GLint) rb->Width && rb->DataType == GL_UNSIGNED_BYTE) {

-            /* optimized case */

-            /* Note: bottom-to-top raster assumed! */

-            GLubyte *stencil = (GLubyte *) rb->GetPointer(ctx, rb, x, y);

-            GLuint len = width * height * sizeof(GLubyte);

-            memset(stencil, clearVal, len);

-         }

-         else {

-            /* general case */

-            GLint i;

-            for (i = 0; i < height; i++) {

-               GLvoid *stencil = rb->GetPointer(ctx, rb, x, y + i);

-               if (rb->DataType == GL_UNSIGNED_BYTE) {

-                  memset(stencil, clearVal, width);

-               }

-               else {

-                  _mesa_memset16((short unsigned int*) stencil, clearVal, width);

-               }

-            }

-         }

-      }

-   }

-   else {

-      /* no direct access */

-      if ((mask & stencilMax) != stencilMax) {

-         /* need to mask the clear */

-         if (rb->DataType == GL_UNSIGNED_BYTE) {

-            GLint i, j;

-            for (i = 0; i < height; i++) {

-               GLubyte stencil[MAX_WIDTH];

-               rb->GetRow(ctx, rb, width, x, y + i, stencil);

-               for (j = 0; j < width; j++) {

-                  stencil[j] = (stencil[j] & invMask) | clearVal;

-               }

-               rb->PutRow(ctx, rb, width, x, y + i, stencil, NULL);

-            }

-         }

-         else {

-            GLint i, j;

-            for (i = 0; i < height; i++) {

-               GLushort stencil[MAX_WIDTH];

-               rb->GetRow(ctx, rb, width, x, y + i, stencil);

-               for (j = 0; j < width; j++) {

-                  stencil[j] = (stencil[j] & invMask) | clearVal;

-               }

-               rb->PutRow(ctx, rb, width, x, y + i, stencil, NULL);

-            }

-         }

-      }

-      else {

-         /* no bit masking */

-         const GLubyte clear8 = (GLubyte) clearVal;

-         const GLushort clear16 = (GLushort) clearVal;

-         const void *clear;

-         GLint i;

-         if (rb->DataType == GL_UNSIGNED_BYTE) {

-            clear = &clear8;

-         }

-         else {

-            clear = &clear16;

-         }

-         for (i = 0; i < height; i++) {

-            rb->PutMonoRow(ctx, rb, width, x, y + i, clear, NULL);

-         }

-      }

-   }

-}

+/*
+ * Mesa 3-D graphics library
+ * Version:  7.1
+ *
+ * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * 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/imports.h"
+
+#include "s_context.h"
+#include "s_depth.h"
+#include "s_stencil.h"
+#include "s_span.h"
+
+
+
+/* Stencil Logic:
+
+IF stencil test fails THEN
+   Apply fail-op to stencil value
+   Don't write the pixel (RGBA,Z)
+ELSE
+   IF doing depth test && depth test fails THEN
+      Apply zfail-op to stencil value
+      Write RGBA and Z to appropriate buffers
+   ELSE
+      Apply zpass-op to stencil value
+ENDIF
+
+*/
+
+
+/**
+ * Apply the given stencil operator to the array of stencil values.
+ * Don't touch stencil[i] if mask[i] is zero.
+ * Input:  n - size of stencil array
+ *         oper - the stencil buffer operator
+ *         face - 0 or 1 for front or back face operation
+ *         stencil - array of stencil values
+ *         mask - array [n] of flag:  1=apply operator, 0=don't apply operator
+ * Output:  stencil - modified values
+ */
+static void
+apply_stencil_op( const struct gl_context *ctx, GLenum oper, GLuint face,
+                  GLuint n, GLstencil stencil[], const GLubyte mask[] )
+{
+   const GLstencil ref = ctx->Stencil.Ref[face];
+   const GLstencil wrtmask = ctx->Stencil.WriteMask[face];
+   const GLstencil invmask = (GLstencil) (~wrtmask);
+   const GLstencil stencilMax = (1 << ctx->DrawBuffer->Visual.stencilBits) - 1;
+   GLuint i;
+
+   switch (oper) {
+      case GL_KEEP:
+         /* do nothing */
+         break;
+      case GL_ZERO:
+	 if (invmask==0) {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+		  stencil[i] = 0;
+	       }
+	    }
+	 }
+	 else {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+		  stencil[i] = (GLstencil) (stencil[i] & invmask);
+	       }
+	    }
+	 }
+	 break;
+      case GL_REPLACE:
+	 if (invmask==0) {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+                  stencil[i] = ref;
+	       }
+	    }
+	 }
+	 else {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+		  GLstencil s = stencil[i];
+		  stencil[i] = (GLstencil) ((invmask & s ) | (wrtmask & ref));
+	       }
+	    }
+	 }
+	 break;
+      case GL_INCR:
+	 if (invmask==0) {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+		  GLstencil s = stencil[i];
+		  if (s < stencilMax) {
+		     stencil[i] = (GLstencil) (s+1);
+		  }
+	       }
+	    }
+	 }
+	 else {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+		  /* VERIFY logic of adding 1 to a write-masked value */
+		  GLstencil s = stencil[i];
+		  if (s < stencilMax) {
+		     stencil[i] = (GLstencil) ((invmask & s) | (wrtmask & (s+1)));
+		  }
+	       }
+	    }
+	 }
+	 break;
+      case GL_DECR:
+	 if (invmask==0) {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+		  GLstencil s = stencil[i];
+		  if (s>0) {
+		     stencil[i] = (GLstencil) (s-1);
+		  }
+	       }
+	    }
+	 }
+	 else {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+		  /* VERIFY logic of subtracting 1 to a write-masked value */
+		  GLstencil s = stencil[i];
+		  if (s>0) {
+		     stencil[i] = (GLstencil) ((invmask & s) | (wrtmask & (s-1)));
+		  }
+	       }
+	    }
+	 }
+	 break;
+      case GL_INCR_WRAP_EXT:
+	 if (invmask==0) {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+                  stencil[i]++;
+	       }
+	    }
+	 }
+	 else {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+                  GLstencil s = stencil[i];
+                  stencil[i] = (GLstencil) ((invmask & s) | (wrtmask & (s+1)));
+	       }
+	    }
+	 }
+	 break;
+      case GL_DECR_WRAP_EXT:
+	 if (invmask==0) {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+		  stencil[i]--;
+	       }
+	    }
+	 }
+	 else {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+                  GLstencil s = stencil[i];
+                  stencil[i] = (GLstencil) ((invmask & s) | (wrtmask & (s-1)));
+	       }
+	    }
+	 }
+	 break;
+      case GL_INVERT:
+	 if (invmask==0) {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+		  GLstencil s = stencil[i];
+		  stencil[i] = (GLstencil) ~s;
+	       }
+	    }
+	 }
+	 else {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+		  GLstencil s = stencil[i];
+		  stencil[i] = (GLstencil) ((invmask & s) | (wrtmask & ~s));
+	       }
+	    }
+	 }
+	 break;
+      default:
+         _mesa_problem(ctx, "Bad stencil op in apply_stencil_op");
+   }
+}
+
+
+
+
+/**
+ * Apply stencil test to an array of stencil values (before depth buffering).
+ * Input:  face - 0 or 1 for front or back-face polygons
+ *         n - number of pixels in the array
+ *         stencil - array of [n] stencil values
+ *         mask - array [n] of flag:  0=skip the pixel, 1=stencil the pixel
+ * Output:  mask - pixels which fail the stencil test will have their
+ *                 mask flag set to 0.
+ *          stencil - updated stencil values (where the test passed)
+ * Return:  GL_FALSE = all pixels failed, GL_TRUE = zero or more pixels passed.
+ */
+static GLboolean
+do_stencil_test( struct gl_context *ctx, GLuint face, GLuint n, GLstencil stencil[],
+                 GLubyte mask[] )
+{
+   GLubyte fail[MAX_WIDTH];
+   GLboolean allfail = GL_FALSE;
+   GLuint i;
+   const GLuint valueMask = ctx->Stencil.ValueMask[face];
+   const GLstencil r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);
+   GLstencil s;
+
+   ASSERT(n <= MAX_WIDTH);
+
+   /*
+    * Perform stencil test.  The results of this operation are stored
+    * in the fail[] array:
+    *   IF fail[i] is non-zero THEN
+    *       the stencil fail operator is to be applied
+    *   ELSE
+    *       the stencil fail operator is not to be applied
+    *   ENDIF
+    */
+   switch (ctx->Stencil.Function[face]) {
+      case GL_NEVER:
+         /* never pass; always fail */
+         for (i=0;i<n;i++) {
+	    if (mask[i]) {
+	       mask[i] = 0;
+	       fail[i] = 1;
+	    }
+	    else {
+	       fail[i] = 0;
+	    }
+	 }
+	 allfail = GL_TRUE;
+	 break;
+      case GL_LESS:
+	 for (i=0;i<n;i++) {
+	    if (mask[i]) {
+	       s = (GLstencil) (stencil[i] & valueMask);
+	       if (r < s) {
+		  /* passed */
+		  fail[i] = 0;
+	       }
+	       else {
+		  fail[i] = 1;
+		  mask[i] = 0;
+	       }
+	    }
+	    else {
+	       fail[i] = 0;
+	    }
+	 }
+	 break;
+      case GL_LEQUAL:
+	 for (i=0;i<n;i++) {
+	    if (mask[i]) {
+	       s = (GLstencil) (stencil[i] & valueMask);
+	       if (r <= s) {
+		  /* pass */
+		  fail[i] = 0;
+	       }
+	       else {
+		  fail[i] = 1;
+		  mask[i] = 0;
+	       }
+	    }
+	    else {
+	       fail[i] = 0;
+	    }
+	 }
+	 break;
+      case GL_GREATER:
+	 for (i=0;i<n;i++) {
+	    if (mask[i]) {
+	       s = (GLstencil) (stencil[i] & valueMask);
+	       if (r > s) {
+		  /* passed */
+		  fail[i] = 0;
+	       }
+	       else {
+		  fail[i] = 1;
+		  mask[i] = 0;
+	       }
+	    }
+	    else {
+	       fail[i] = 0;
+	    }
+	 }
+	 break;
+      case GL_GEQUAL:
+	 for (i=0;i<n;i++) {
+	    if (mask[i]) {
+	       s = (GLstencil) (stencil[i] & valueMask);
+	       if (r >= s) {
+		  /* passed */
+		  fail[i] = 0;
+	       }
+	       else {
+		  fail[i] = 1;
+		  mask[i] = 0;
+	       }
+	    }
+	    else {
+	       fail[i] = 0;
+	    }
+	 }
+	 break;
+      case GL_EQUAL:
+	 for (i=0;i<n;i++) {
+	    if (mask[i]) {
+	       s = (GLstencil) (stencil[i] & valueMask);
+	       if (r == s) {
+		  /* passed */
+		  fail[i] = 0;
+	       }
+	       else {
+		  fail[i] = 1;
+		  mask[i] = 0;
+	       }
+	    }
+	    else {
+	       fail[i] = 0;
+	    }
+	 }
+	 break;
+      case GL_NOTEQUAL:
+	 for (i=0;i<n;i++) {
+	    if (mask[i]) {
+	       s = (GLstencil) (stencil[i] & valueMask);
+	       if (r != s) {
+		  /* passed */
+		  fail[i] = 0;
+	       }
+	       else {
+		  fail[i] = 1;
+		  mask[i] = 0;
+	       }
+	    }
+	    else {
+	       fail[i] = 0;
+	    }
+	 }
+	 break;
+      case GL_ALWAYS:
+	 /* always pass */
+	 for (i=0;i<n;i++) {
+	    fail[i] = 0;
+	 }
+	 break;
+      default:
+         _mesa_problem(ctx, "Bad stencil func in gl_stencil_span");
+         return 0;
+   }
+
+   if (ctx->Stencil.FailFunc[face] != GL_KEEP) {
+      apply_stencil_op( ctx, ctx->Stencil.FailFunc[face], face, n, stencil, fail );
+   }
+
+   return !allfail;
+}
+
+
+/**
+ * Compute the zpass/zfail masks by comparing the pre- and post-depth test
+ * masks.
+ */
+static INLINE void
+compute_pass_fail_masks(GLuint n, const GLubyte origMask[],
+                        const GLubyte newMask[],
+                        GLubyte passMask[], GLubyte failMask[])
+{
+   GLuint i;
+   for (i = 0; i < n; i++) {
+      ASSERT(newMask[i] == 0 || newMask[i] == 1);
+      passMask[i] = origMask[i] & newMask[i];
+      failMask[i] = origMask[i] & (newMask[i] ^ 1);
+   }
+}
+
+
+/**
+ * Apply stencil and depth testing to the span of pixels.
+ * Both software and hardware stencil buffers are acceptable.
+ * Input:  n - number of pixels in the span
+ *         x, y - location of leftmost pixel in span
+ *         z - array [n] of z values
+ *         mask - array [n] of flags  (1=test this pixel, 0=skip the pixel)
+ * Output:  mask - array [n] of flags (1=stencil and depth test passed)
+ * Return: GL_FALSE - all fragments failed the testing
+ *         GL_TRUE - one or more fragments passed the testing
+ *
+ */
+static GLboolean
+stencil_and_ztest_span(struct gl_context *ctx, SWspan *span, GLuint face)
+{
+   struct gl_framebuffer *fb = ctx->DrawBuffer;
+   struct gl_renderbuffer *rb = fb->_StencilBuffer;
+   GLstencil stencilRow[MAX_WIDTH];
+   GLstencil *stencil;
+   const GLuint n = span->end;
+   const GLint x = span->x;
+   const GLint y = span->y;
+   GLubyte *mask = span->array->mask;
+
+   ASSERT((span->arrayMask & SPAN_XY) == 0);
+   ASSERT(ctx->Stencil.Enabled);
+   ASSERT(n <= MAX_WIDTH);
+#ifdef DEBUG
+   if (ctx->Depth.Test) {
+      ASSERT(span->arrayMask & SPAN_Z);
+   }
+#endif
+
+   stencil = (GLstencil *) rb->GetPointer(ctx, rb, x, y);
+   if (!stencil) {
+      rb->GetRow(ctx, rb, n, x, y, stencilRow);
+      stencil = stencilRow;
+   }
+
+   /*
+    * Apply the stencil test to the fragments.
+    * failMask[i] is 1 if the stencil test failed.
+    */
+   if (do_stencil_test( ctx, face, n, stencil, mask ) == GL_FALSE) {
+      /* all fragments failed the stencil test, we're done. */
+      span->writeAll = GL_FALSE;
+      if (!rb->GetPointer(ctx, rb, 0, 0)) {
+         /* put updated stencil values into buffer */
+         rb->PutRow(ctx, rb, n, x, y, stencil, NULL);
+      }
+      return GL_FALSE;
+   }
+
+   /*
+    * Some fragments passed the stencil test, apply depth test to them
+    * and apply Zpass and Zfail stencil ops.
+    */
+   if (ctx->Depth.Test == GL_FALSE ||
+       ctx->DrawBuffer->_DepthBuffer == NULL) {
+      /*
+       * No depth buffer, just apply zpass stencil function to active pixels.
+       */
+      apply_stencil_op( ctx, ctx->Stencil.ZPassFunc[face], face, n, stencil, mask );
+   }
+   else {
+      /*
+       * Perform depth buffering, then apply zpass or zfail stencil function.
+       */
+      GLubyte passMask[MAX_WIDTH], failMask[MAX_WIDTH], origMask[MAX_WIDTH];
+
+      /* save the current mask bits */
+      memcpy(origMask, mask, n * sizeof(GLubyte));
+
+      /* apply the depth test */
+      _swrast_depth_test_span(ctx, span);
+
+      compute_pass_fail_masks(n, origMask, mask, passMask, failMask);
+
+      /* apply the pass and fail operations */
+      if (ctx->Stencil.ZFailFunc[face] != GL_KEEP) {
+         apply_stencil_op( ctx, ctx->Stencil.ZFailFunc[face], face,
+                           n, stencil, failMask );
+      }
+      if (ctx->Stencil.ZPassFunc[face] != GL_KEEP) {
+         apply_stencil_op( ctx, ctx->Stencil.ZPassFunc[face], face,
+                           n, stencil, passMask );
+      }
+   }
+
+   /*
+    * Write updated stencil values back into hardware stencil buffer.
+    */
+   if (!rb->GetPointer(ctx, rb, 0, 0)) {
+      rb->PutRow(ctx, rb, n, x, y, stencil, NULL);
+   }
+   
+   span->writeAll = GL_FALSE;
+   
+   return GL_TRUE;  /* one or more fragments passed both tests */
+}
+
+
+
+/*
+ * Return the address of a stencil buffer value given the window coords:
+ */
+#define STENCIL_ADDRESS(X, Y)  (stencilStart + (Y) * stride + (X))
+
+
+
+/**
+ * Apply the given stencil operator for each pixel in the array whose
+ * mask flag is set.
+ * \note  This is for software stencil buffers only.
+ * Input:  n - number of pixels in the span
+ *         x, y - array of [n] pixels
+ *         operator - the stencil buffer operator
+ *         mask - array [n] of flag:  1=apply operator, 0=don't apply operator
+ */
+static void
+apply_stencil_op_to_pixels( struct gl_context *ctx,
+                            GLuint n, const GLint x[], const GLint y[],
+                            GLenum oper, GLuint face, const GLubyte mask[] )
+{
+   struct gl_framebuffer *fb = ctx->DrawBuffer;
+   struct gl_renderbuffer *rb = fb->_StencilBuffer;
+   const GLstencil stencilMax = (1 << fb->Visual.stencilBits) - 1;
+   const GLstencil ref = ctx->Stencil.Ref[face];
+   const GLstencil wrtmask = ctx->Stencil.WriteMask[face];
+   const GLstencil invmask = (GLstencil) (~wrtmask);
+   GLuint i;
+   GLstencil *stencilStart = (GLubyte *) rb->Data;
+   const GLuint stride = rb->Width;
+
+   ASSERT(rb->GetPointer(ctx, rb, 0, 0));
+   ASSERT(sizeof(GLstencil) == 1);
+
+   switch (oper) {
+      case GL_KEEP:
+         /* do nothing */
+         break;
+      case GL_ZERO:
+	 if (invmask==0) {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
+                  *sptr = 0;
+	       }
+	    }
+	 }
+	 else {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
+		  *sptr = (GLstencil) (invmask & *sptr);
+	       }
+	    }
+	 }
+	 break;
+      case GL_REPLACE:
+	 if (invmask==0) {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
+                  *sptr = ref;
+	       }
+	    }
+	 }
+	 else {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
+		  *sptr = (GLstencil) ((invmask & *sptr ) | (wrtmask & ref));
+	       }
+	    }
+	 }
+	 break;
+      case GL_INCR:
+	 if (invmask==0) {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
+		  if (*sptr < stencilMax) {
+		     *sptr = (GLstencil) (*sptr + 1);
+		  }
+	       }
+	    }
+	 }
+	 else {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
+		  if (*sptr < stencilMax) {
+		     *sptr = (GLstencil) ((invmask & *sptr) | (wrtmask & (*sptr+1)));
+		  }
+	       }
+	    }
+	 }
+	 break;
+      case GL_DECR:
+	 if (invmask==0) {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
+		  if (*sptr>0) {
+		     *sptr = (GLstencil) (*sptr - 1);
+		  }
+	       }
+	    }
+	 }
+	 else {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
+		  if (*sptr>0) {
+		     *sptr = (GLstencil) ((invmask & *sptr) | (wrtmask & (*sptr-1)));
+		  }
+	       }
+	    }
+	 }
+	 break;
+      case GL_INCR_WRAP_EXT:
+	 if (invmask==0) {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
+                  *sptr = (GLstencil) (*sptr + 1);
+	       }
+	    }
+	 }
+	 else {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
+                  *sptr = (GLstencil) ((invmask & *sptr) | (wrtmask & (*sptr+1)));
+	       }
+	    }
+	 }
+	 break;
+      case GL_DECR_WRAP_EXT:
+	 if (invmask==0) {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
+                  *sptr = (GLstencil) (*sptr - 1);
+	       }
+	    }
+	 }
+	 else {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
+                  *sptr = (GLstencil) ((invmask & *sptr) | (wrtmask & (*sptr-1)));
+	       }
+	    }
+	 }
+	 break;
+      case GL_INVERT:
+	 if (invmask==0) {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
+                  *sptr = (GLstencil) (~*sptr);
+	       }
+	    }
+	 }
+	 else {
+	    for (i=0;i<n;i++) {
+	       if (mask[i]) {
+                  GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
+                  *sptr = (GLstencil) ((invmask & *sptr) | (wrtmask & ~*sptr));
+	       }
+	    }
+	 }
+	 break;
+      default:
+         _mesa_problem(ctx, "Bad stencilop in apply_stencil_op_to_pixels");
+   }
+}
+
+
+
+/**
+ * Apply stencil test to an array of pixels before depth buffering.
+ *
+ * \note Used for software stencil buffer only.
+ * Input:  n - number of pixels in the span
+ *         x, y - array of [n] pixels to stencil
+ *         mask - array [n] of flag:  0=skip the pixel, 1=stencil the pixel
+ * Output:  mask - pixels which fail the stencil test will have their
+ *                 mask flag set to 0.
+ * \return  GL_FALSE = all pixels failed, GL_TRUE = zero or more pixels passed.
+ */
+static GLboolean
+stencil_test_pixels( struct gl_context *ctx, GLuint face, GLuint n,
+                     const GLint x[], const GLint y[], GLubyte mask[] )
+{
+   const struct gl_framebuffer *fb = ctx->DrawBuffer;
+   struct gl_renderbuffer *rb = fb->_StencilBuffer;
+   GLubyte fail[MAX_WIDTH];
+   GLstencil r, s;
+   GLuint i;
+   GLboolean allfail = GL_FALSE;
+   const GLuint valueMask = ctx->Stencil.ValueMask[face];
+   const GLstencil *stencilStart = (GLstencil *) rb->Data;
+   const GLuint stride = rb->Width;
+
+   ASSERT(rb->GetPointer(ctx, rb, 0, 0));
+   ASSERT(sizeof(GLstencil) == 1);
+
+   /*
+    * Perform stencil test.  The results of this operation are stored
+    * in the fail[] array:
+    *   IF fail[i] is non-zero THEN
+    *       the stencil fail operator is to be applied
+    *   ELSE
+    *       the stencil fail operator is not to be applied
+    *   ENDIF
+    */
+
+   switch (ctx->Stencil.Function[face]) {
+      case GL_NEVER:
+         /* always fail */
+         for (i=0;i<n;i++) {
+	    if (mask[i]) {
+	       mask[i] = 0;
+	       fail[i] = 1;
+	    }
+	    else {
+	       fail[i] = 0;
+	    }
+	 }
+	 allfail = GL_TRUE;
+	 break;
+      case GL_LESS:
+	 r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);
+	 for (i=0;i<n;i++) {
+	    if (mask[i]) {
+               const GLstencil *sptr = STENCIL_ADDRESS(x[i],y[i]);
+	       s = (GLstencil) (*sptr & valueMask);
+	       if (r < s) {
+		  /* passed */
+		  fail[i] = 0;
+	       }
+	       else {
+		  fail[i] = 1;
+		  mask[i] = 0;
+	       }
+	    }
+	    else {
+	       fail[i] = 0;
+	    }
+	 }
+	 break;
+      case GL_LEQUAL:
+	 r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);
+	 for (i=0;i<n;i++) {
+	    if (mask[i]) {
+               const GLstencil *sptr = STENCIL_ADDRESS(x[i],y[i]);
+	       s = (GLstencil) (*sptr & valueMask);
+	       if (r <= s) {
+		  /* pass */
+		  fail[i] = 0;
+	       }
+	       else {
+		  fail[i] = 1;
+		  mask[i] = 0;
+	       }
+	    }
+	    else {
+	       fail[i] = 0;
+	    }
+	 }
+	 break;
+      case GL_GREATER:
+	 r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);
+	 for (i=0;i<n;i++) {
+	    if (mask[i]) {
+               const GLstencil *sptr = STENCIL_ADDRESS(x[i],y[i]);
+	       s = (GLstencil) (*sptr & valueMask);
+	       if (r > s) {
+		  /* passed */
+		  fail[i] = 0;
+	       }
+	       else {
+		  fail[i] = 1;
+		  mask[i] = 0;
+	       }
+	    }
+	    else {
+	       fail[i] = 0;
+	    }
+	 }
+	 break;
+      case GL_GEQUAL:
+	 r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);
+	 for (i=0;i<n;i++) {
+	    if (mask[i]) {
+               const GLstencil *sptr = STENCIL_ADDRESS(x[i],y[i]);
+	       s = (GLstencil) (*sptr & valueMask);
+	       if (r >= s) {
+		  /* passed */
+		  fail[i] = 0;
+	       }
+	       else {
+		  fail[i] = 1;
+		  mask[i] = 0;
+	       }
+	    }
+	    else {
+	       fail[i] = 0;
+	    }
+	 }
+	 break;
+      case GL_EQUAL:
+	 r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);
+	 for (i=0;i<n;i++) {
+	    if (mask[i]) {
+               const GLstencil *sptr = STENCIL_ADDRESS(x[i],y[i]);
+	       s = (GLstencil) (*sptr & valueMask);
+	       if (r == s) {
+		  /* passed */
+		  fail[i] = 0;
+	       }
+	       else {
+		  fail[i] = 1;
+		  mask[i] = 0;
+	       }
+	    }
+	    else {
+	       fail[i] = 0;
+	    }
+	 }
+	 break;
+      case GL_NOTEQUAL:
+	 r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);
+	 for (i=0;i<n;i++) {
+	    if (mask[i]) {
+               const GLstencil *sptr = STENCIL_ADDRESS(x[i],y[i]);
+	       s = (GLstencil) (*sptr & valueMask);
+	       if (r != s) {
+		  /* passed */
+		  fail[i] = 0;
+	       }
+	       else {
+		  fail[i] = 1;
+		  mask[i] = 0;
+	       }
+	    }
+	    else {
+	       fail[i] = 0;
+	    }
+	 }
+	 break;
+      case GL_ALWAYS:
+	 /* always pass */
+	 for (i=0;i<n;i++) {
+	    fail[i] = 0;
+	 }
+	 break;
+      default:
+         _mesa_problem(ctx, "Bad stencil func in gl_stencil_pixels");
+         return 0;
+   }
+
+   if (ctx->Stencil.FailFunc[face] != GL_KEEP) {
+      apply_stencil_op_to_pixels( ctx, n, x, y, ctx->Stencil.FailFunc[face],
+                                  face, fail );
+   }
+
+   return !allfail;
+}
+
+
+
+
+/**
+ * Apply stencil and depth testing to an array of pixels.
+ * This is used both for software and hardware stencil buffers.
+ *
+ * The comments in this function are a bit sparse but the code is
+ * almost identical to stencil_and_ztest_span(), which is well
+ * commented.
+ *
+ * Input:  n - number of pixels in the array
+ *         x, y - array of [n] pixel positions
+ *         z - array [n] of z values
+ *         mask - array [n] of flags  (1=test this pixel, 0=skip the pixel)
+ * Output: mask - array [n] of flags (1=stencil and depth test passed)
+ * Return: GL_FALSE - all fragments failed the testing
+ *         GL_TRUE - one or more fragments passed the testing
+ */
+static GLboolean
+stencil_and_ztest_pixels( struct gl_context *ctx, SWspan *span, GLuint face )
+{
+   GLubyte passMask[MAX_WIDTH], failMask[MAX_WIDTH], origMask[MAX_WIDTH];
+   struct gl_framebuffer *fb = ctx->DrawBuffer;
+   struct gl_renderbuffer *rb = fb->_StencilBuffer;
+   const GLuint n = span->end;
+   const GLint *x = span->array->x;
+   const GLint *y = span->array->y;
+   GLubyte *mask = span->array->mask;
+
+   ASSERT(span->arrayMask & SPAN_XY);
+   ASSERT(ctx->Stencil.Enabled);
+   ASSERT(n <= MAX_WIDTH);
+
+   if (!rb->GetPointer(ctx, rb, 0, 0)) {
+      /* No direct access */
+      GLstencil stencil[MAX_WIDTH];
+
+      ASSERT(rb->DataType == GL_UNSIGNED_BYTE);
+      _swrast_get_values(ctx, rb, n, x, y, stencil, sizeof(GLubyte));
+
+      memcpy(origMask, mask, n * sizeof(GLubyte));          
+
+      (void) do_stencil_test(ctx, face, n, stencil, mask);
+
+      if (ctx->Depth.Test == GL_FALSE) {
+         apply_stencil_op(ctx, ctx->Stencil.ZPassFunc[face], face,
+                          n, stencil, mask);
+      }
+      else {
+         GLubyte tmpMask[MAX_WIDTH]; 
+         memcpy(tmpMask, mask, n * sizeof(GLubyte));
+
+         _swrast_depth_test_span(ctx, span);
+
+         compute_pass_fail_masks(n, tmpMask, mask, passMask, failMask);
+
+         if (ctx->Stencil.ZFailFunc[face] != GL_KEEP) {
+            apply_stencil_op(ctx, ctx->Stencil.ZFailFunc[face], face,
+                             n, stencil, failMask);
+         }
+         if (ctx->Stencil.ZPassFunc[face] != GL_KEEP) {
+            apply_stencil_op(ctx, ctx->Stencil.ZPassFunc[face], face,
+                             n, stencil, passMask);
+         }
+      }
+
+      /* Write updated stencil values into hardware stencil buffer */
+      rb->PutValues(ctx, rb, n, x, y, stencil, origMask);
+
+      return GL_TRUE;
+   }
+   else {
+      /* Direct access to stencil buffer */
+
+      if (stencil_test_pixels(ctx, face, n, x, y, mask) == GL_FALSE) {
+         /* all fragments failed the stencil test, we're done. */
+         return GL_FALSE;
+      }
+
+      if (ctx->Depth.Test==GL_FALSE) {
+         apply_stencil_op_to_pixels(ctx, n, x, y,
+                                    ctx->Stencil.ZPassFunc[face], face, mask);
+      }
+      else {
+         memcpy(origMask, mask, n * sizeof(GLubyte));
+
+         _swrast_depth_test_span(ctx, span);
+
+         compute_pass_fail_masks(n, origMask, mask, passMask, failMask);
+
+         if (ctx->Stencil.ZFailFunc[face] != GL_KEEP) {
+            apply_stencil_op_to_pixels(ctx, n, x, y,
+                                       ctx->Stencil.ZFailFunc[face],
+                                       face, failMask);
+         }
+         if (ctx->Stencil.ZPassFunc[face] != GL_KEEP) {
+            apply_stencil_op_to_pixels(ctx, n, x, y,
+                                       ctx->Stencil.ZPassFunc[face],
+                                       face, passMask);
+         }
+      }
+
+      return GL_TRUE;  /* one or more fragments passed both tests */
+   }
+}
+
+
+/**
+ * /return GL_TRUE = one or more fragments passed,
+ * GL_FALSE = all fragments failed.
+ */
+GLboolean
+_swrast_stencil_and_ztest_span(struct gl_context *ctx, SWspan *span)
+{
+   const GLuint face = (span->facing == 0) ? 0 : ctx->Stencil._BackFace;
+
+   if (span->arrayMask & SPAN_XY)
+      return stencil_and_ztest_pixels(ctx, span, face);
+   else
+      return stencil_and_ztest_span(ctx, span, face);
+}
+
+
+#if 0
+GLuint
+clip_span(GLuint bufferWidth, GLuint bufferHeight,
+          GLint x, GLint y, GLuint *count)
+{
+   GLuint n = *count;
+   GLuint skipPixels = 0;
+
+   if (y < 0 || y >= bufferHeight || x + n <= 0 || x >= bufferWidth) {
+      /* totally out of bounds */
+      n = 0;
+   }
+   else {
+      /* left clip */
+      if (x < 0) {
+         skipPixels = -x;
+         x = 0;
+         n -= skipPixels;
+      }
+      /* right clip */
+      if (x + n > bufferWidth) {
+         GLint dx = x + n - bufferWidth;
+         n -= dx;
+      }
+   }
+
+   *count = n;
+
+   return skipPixels;
+}
+#endif
+
+
+/**
+ * Return a span of stencil values from the stencil buffer.
+ * Used for glRead/CopyPixels
+ * Input:  n - how many pixels
+ *         x,y - location of first pixel
+ * Output:  stencil - the array of stencil values
+ */
+void
+_swrast_read_stencil_span(struct gl_context *ctx, struct gl_renderbuffer *rb,
+                          GLint n, GLint x, GLint y, GLstencil stencil[])
+{
+   if (y < 0 || y >= (GLint) rb->Height ||
+       x + n <= 0 || x >= (GLint) rb->Width) {
+      /* span is completely outside framebuffer */
+      return; /* undefined values OK */
+   }
+
+   if (x < 0) {
+      GLint dx = -x;
+      x = 0;
+      n -= dx;
+      stencil += dx;
+   }
+   if (x + n > (GLint) rb->Width) {
+      GLint dx = x + n - rb->Width;
+      n -= dx;
+   }
+   if (n <= 0) {
+      return;
+   }
+
+   rb->GetRow(ctx, rb, n, x, y, stencil);
+}
+
+
+
+/**
+ * Write a span of stencil values to the stencil buffer.  This function
+ * applies the stencil write mask when needed.
+ * Used for glDraw/CopyPixels
+ * Input:  n - how many pixels
+ *         x, y - location of first pixel
+ *         stencil - the array of stencil values
+ */
+void
+_swrast_write_stencil_span(struct gl_context *ctx, GLint n, GLint x, GLint y,
+                           const GLstencil stencil[] )
+{
+   struct gl_framebuffer *fb = ctx->DrawBuffer;
+   struct gl_renderbuffer *rb = fb->_StencilBuffer;
+   const GLuint stencilMax = (1 << fb->Visual.stencilBits) - 1;
+   const GLuint stencilMask = ctx->Stencil.WriteMask[0];
+
+   if (y < 0 || y >= (GLint) rb->Height ||
+       x + n <= 0 || x >= (GLint) rb->Width) {
+      /* span is completely outside framebuffer */
+      return; /* undefined values OK */
+   }
+   if (x < 0) {
+      GLint dx = -x;
+      x = 0;
+      n -= dx;
+      stencil += dx;
+   }
+   if (x + n > (GLint) rb->Width) {
+      GLint dx = x + n - rb->Width;
+      n -= dx;
+   }
+   if (n <= 0) {
+      return;
+   }
+
+   if ((stencilMask & stencilMax) != stencilMax) {
+      /* need to apply writemask */
+      GLstencil destVals[MAX_WIDTH], newVals[MAX_WIDTH];
+      GLint i;
+      rb->GetRow(ctx, rb, n, x, y, destVals);
+      for (i = 0; i < n; i++) {
+         newVals[i]
+            = (stencil[i] & stencilMask) | (destVals[i] & ~stencilMask);
+      }
+      rb->PutRow(ctx, rb, n, x, y, newVals, NULL);
+   }
+   else {
+      rb->PutRow(ctx, rb, n, x, y, stencil, NULL);
+   }
+}
+
+
+
+/**
+ * Clear the stencil buffer.
+ */
+void
+_swrast_clear_stencil_buffer( struct gl_context *ctx, struct gl_renderbuffer *rb )
+{
+   const GLubyte stencilBits = ctx->DrawBuffer->Visual.stencilBits;
+   const GLuint mask = ctx->Stencil.WriteMask[0];
+   const GLuint invMask = ~mask;
+   const GLuint clearVal = (ctx->Stencil.Clear & mask);
+   const GLuint stencilMax = (1 << stencilBits) - 1;
+   GLint x, y, width, height;
+
+   if (!rb || mask == 0)
+      return;
+
+   ASSERT(rb->DataType == GL_UNSIGNED_BYTE ||
+          rb->DataType == GL_UNSIGNED_SHORT);
+
+   ASSERT(rb->_BaseFormat == GL_STENCIL_INDEX);
+
+   /* compute region to clear */
+   x = ctx->DrawBuffer->_Xmin;
+   y = ctx->DrawBuffer->_Ymin;
+   width  = ctx->DrawBuffer->_Xmax - ctx->DrawBuffer->_Xmin;
+   height = ctx->DrawBuffer->_Ymax - ctx->DrawBuffer->_Ymin;
+
+   if (rb->GetPointer(ctx, rb, 0, 0)) {
+      /* Direct buffer access */
+      if ((mask & stencilMax) != stencilMax) {
+         /* need to mask the clear */
+         if (rb->DataType == GL_UNSIGNED_BYTE) {
+            GLint i, j;
+            for (i = 0; i < height; i++) {
+               GLubyte *stencil = (GLubyte*) rb->GetPointer(ctx, rb, x, y + i);
+               for (j = 0; j < width; j++) {
+                  stencil[j] = (stencil[j] & invMask) | clearVal;
+               }
+            }
+         }
+         else {
+            GLint i, j;
+            for (i = 0; i < height; i++) {
+               GLushort *stencil = (GLushort*) rb->GetPointer(ctx, rb, x, y + i);
+               for (j = 0; j < width; j++) {
+                  stencil[j] = (stencil[j] & invMask) | clearVal;
+               }
+            }
+         }
+      }
+      else {
+         /* no bit masking */
+         if (width == (GLint) rb->Width && rb->DataType == GL_UNSIGNED_BYTE) {
+            /* optimized case */
+            /* Note: bottom-to-top raster assumed! */
+            GLubyte *stencil = (GLubyte *) rb->GetPointer(ctx, rb, x, y);
+            GLuint len = width * height * sizeof(GLubyte);
+            memset(stencil, clearVal, len);
+         }
+         else {
+            /* general case */
+            GLint i;
+            for (i = 0; i < height; i++) {
+               GLvoid *stencil = rb->GetPointer(ctx, rb, x, y + i);
+               if (rb->DataType == GL_UNSIGNED_BYTE) {
+                  memset(stencil, clearVal, width);
+               }
+               else {
+                  _mesa_memset16((short unsigned int*) stencil, clearVal, width);
+               }
+            }
+         }
+      }
+   }
+   else {
+      /* no direct access */
+      if ((mask & stencilMax) != stencilMax) {
+         /* need to mask the clear */
+         if (rb->DataType == GL_UNSIGNED_BYTE) {
+            GLint i, j;
+            for (i = 0; i < height; i++) {
+               GLubyte stencil[MAX_WIDTH];
+               rb->GetRow(ctx, rb, width, x, y + i, stencil);
+               for (j = 0; j < width; j++) {
+                  stencil[j] = (stencil[j] & invMask) | clearVal;
+               }
+               rb->PutRow(ctx, rb, width, x, y + i, stencil, NULL);
+            }
+         }
+         else {
+            GLint i, j;
+            for (i = 0; i < height; i++) {
+               GLushort stencil[MAX_WIDTH];
+               rb->GetRow(ctx, rb, width, x, y + i, stencil);
+               for (j = 0; j < width; j++) {
+                  stencil[j] = (stencil[j] & invMask) | clearVal;
+               }
+               rb->PutRow(ctx, rb, width, x, y + i, stencil, NULL);
+            }
+         }
+      }
+      else {
+         /* no bit masking */
+         const GLubyte clear8 = (GLubyte) clearVal;
+         const GLushort clear16 = (GLushort) clearVal;
+         const void *clear;
+         GLint i;
+         if (rb->DataType == GL_UNSIGNED_BYTE) {
+            clear = &clear8;
+         }
+         else {
+            clear = &clear16;
+         }
+         for (i = 0; i < height; i++) {
+            rb->PutMonoRow(ctx, rb, width, x, y + i, clear, NULL);
+         }
+      }
+   }
+}
diff --git a/mesalib/src/mesa/swrast/s_texcombine.c b/mesalib/src/mesa/swrast/s_texcombine.c
index 53ef2f890..80b9dff3c 100644
--- a/mesalib/src/mesa/swrast/s_texcombine.c
+++ b/mesalib/src/mesa/swrast/s_texcombine.c
@@ -1,751 +1,755 @@
-/*

- * 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);

-}

+/*
+ * 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)
+{
+#ifdef _OPENMP
+   return (float4_array) (swrast->TexelBuffer + unit * MAX_WIDTH * 4 * omp_get_num_threads() + (MAX_WIDTH * 4 * omp_get_thread_num()));
+#else
+   return (float4_array) (swrast->TexelBuffer + unit * MAX_WIDTH * 4);
+#endif
+}
+
+
+
+/**
+ * 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);
+}