softpipe: Split changing fields of quad_header into input, inout and output parts.

15 files changed, 183 insertions, 173 deletions

diff --git a/src/gallium/drivers/softpipe/sp_fs_exec.c b/src/gallium/drivers/softpipe/sp_fs_exec.c
index d0456731be..701ee4c72f 100644
--- a/src/gallium/drivers/softpipe/sp_fs_exec.c
+++ b/src/gallium/drivers/softpipe/sp_fs_exec.c
@@ -106,7 +106,7 @@ exec_run( const struct sp_fragment_shader *base,
 
    /* Compute X, Y, Z, W vals for this quad */
    sp_setup_pos_vector(quad->posCoef, 
-		       (float)quad->x0, (float)quad->y0, 
+		       (float)quad->input.x0, (float)quad->input.y0, 
 		       &machine->QuadPos);
    
    return tgsi_exec_machine_run( machine );
diff --git a/src/gallium/drivers/softpipe/sp_fs_sse.c b/src/gallium/drivers/softpipe/sp_fs_sse.c
index 35653a8e48..496ed43df2 100644
--- a/src/gallium/drivers/softpipe/sp_fs_sse.c
+++ b/src/gallium/drivers/softpipe/sp_fs_sse.c
@@ -88,7 +88,7 @@ fs_sse_run( const struct sp_fragment_shader *base,
 
    /* Compute X, Y, Z, W vals for this quad -- place in temp[0] for now */
    sp_setup_pos_vector(quad->posCoef, 
-		       (float)quad->x0, (float)quad->y0, 
+		       (float)quad->input.x0, (float)quad->input.y0, 
 		       machine->Temps);
 
    /* init kill mask */
diff --git a/src/gallium/drivers/softpipe/sp_headers.h b/src/gallium/drivers/softpipe/sp_headers.h
index ae2ee210fc..4a42cb3c19 100644
--- a/src/gallium/drivers/softpipe/sp_headers.h
+++ b/src/gallium/drivers/softpipe/sp_headers.h
@@ -59,20 +59,31 @@
  * Encodes everything we need to know about a 2x2 pixel block.  Uses
  * "Channel-Serial" or "SoA" layout.  
  */
-struct quad_header {
+struct quad_header_input
+{
    int x0;
    int y0;
-   unsigned mask:4;
+   float coverage[QUAD_SIZE];    /** fragment coverage for antialiasing */
    unsigned facing:1;   /**< Front (0) or back (1) facing? */
    unsigned prim:2;     /**< PRIM_POINT, LINE, TRI */
+};
+
+struct quad_header_inout
+{
+   unsigned mask:4;
+};
 
-   struct {
-      /** colors in SOA format (rrrr, gggg, bbbb, aaaa) */
-      float color[PIPE_MAX_COLOR_BUFS][NUM_CHANNELS][QUAD_SIZE];
-      float depth[QUAD_SIZE];
-   } outputs;
+struct quad_header_output
+{
+   /** colors in SOA format (rrrr, gggg, bbbb, aaaa) */
+   float color[PIPE_MAX_COLOR_BUFS][NUM_CHANNELS][QUAD_SIZE];
+   float depth[QUAD_SIZE];
+};
 
-   float coverage[QUAD_SIZE];    /** fragment coverage for antialiasing */
+struct quad_header {
+   struct quad_header_input input;
+   struct quad_header_inout inout;
+   struct quad_header_output output;
 
    const struct tgsi_interp_coef *coef;
    const struct tgsi_interp_coef *posCoef;
@@ -80,5 +91,5 @@ struct quad_header {
    unsigned nr_attrs;
 };
 
-
 #endif /* SP_HEADERS_H */
+
diff --git a/src/gallium/drivers/softpipe/sp_quad_alpha_test.c b/src/gallium/drivers/softpipe/sp_quad_alpha_test.c
index 7d3580fb4f..5bebd141e9 100644
--- a/src/gallium/drivers/softpipe/sp_quad_alpha_test.c
+++ b/src/gallium/drivers/softpipe/sp_quad_alpha_test.c
@@ -17,11 +17,10 @@ alpha_test_quad(struct quad_stage *qs, struct quad_header *quad)
    const float ref = softpipe->depth_stencil->alpha.ref;
    unsigned passMask = 0x0, j;
    const uint cbuf = 0; /* only output[0].alpha is tested */
-   const float *aaaa = quad->outputs.color[cbuf][3];
+   const float *aaaa = quad->output.color[cbuf][3];
 
    switch (softpipe->depth_stencil->alpha.func) {
    case PIPE_FUNC_NEVER:
-      quad->mask = 0x0;
       break;
    case PIPE_FUNC_LESS:
       /*
@@ -76,9 +75,9 @@ alpha_test_quad(struct quad_stage *qs, struct quad_header *quad)
       assert(0);
    }
 
-   quad->mask &= passMask;
+   quad->inout.mask &= passMask;
 
-   if (quad->mask)
+   if (quad->inout.mask)
       qs->next->run(qs->next, quad);
 }
 
diff --git a/src/gallium/drivers/softpipe/sp_quad_blend.c b/src/gallium/drivers/softpipe/sp_quad_blend.c
index a834accb86..6f64c6e584 100644
--- a/src/gallium/drivers/softpipe/sp_quad_blend.c
+++ b/src/gallium/drivers/softpipe/sp_quad_blend.c
@@ -114,14 +114,14 @@ logicop_quad(struct quad_stage *qs, struct quad_header *quad)
       struct softpipe_cached_tile *
          tile = sp_get_cached_tile(softpipe,
                                    softpipe->cbuf_cache[cbuf],
-                                   quad->x0, quad->y0);
-      float (*quadColor)[4] = quad->outputs.color[cbuf];
+                                   quad->input.x0, quad->input.y0);
+      float (*quadColor)[4] = quad->output.color[cbuf];
       uint i, j;
 
       /* get/swizzle dest colors */
       for (j = 0; j < QUAD_SIZE; j++) {
-         int x = (quad->x0 & (TILE_SIZE-1)) + (j & 1);
-         int y = (quad->y0 & (TILE_SIZE-1)) + (j >> 1);
+         int x = (quad->input.x0 & (TILE_SIZE-1)) + (j & 1);
+         int y = (quad->input.y0 & (TILE_SIZE-1)) + (j >> 1);
          for (i = 0; i < 4; i++) {
             dest[i][j] = tile->data.color[y][x][i];
          }
@@ -244,14 +244,14 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad)
       struct softpipe_cached_tile *tile
          = sp_get_cached_tile(softpipe,
                               softpipe->cbuf_cache[cbuf],
-                              quad->x0, quad->y0);
-      float (*quadColor)[4] = quad->outputs.color[cbuf];
+                              quad->input.x0, quad->input.y0);
+      float (*quadColor)[4] = quad->output.color[cbuf];
       uint i, j;
 
       /* get/swizzle dest colors */
       for (j = 0; j < QUAD_SIZE; j++) {
-         int x = (quad->x0 & (TILE_SIZE-1)) + (j & 1);
-         int y = (quad->y0 & (TILE_SIZE-1)) + (j >> 1);
+         int x = (quad->input.x0 & (TILE_SIZE-1)) + (j & 1);
+         int y = (quad->input.y0 & (TILE_SIZE-1)) + (j >> 1);
          for (i = 0; i < 4; i++) {
             dest[i][j] = tile->data.color[y][x][i];
          }
diff --git a/src/gallium/drivers/softpipe/sp_quad_colormask.c b/src/gallium/drivers/softpipe/sp_quad_colormask.c
index f72f31db97..f32bdfab78 100644
--- a/src/gallium/drivers/softpipe/sp_quad_colormask.c
+++ b/src/gallium/drivers/softpipe/sp_quad_colormask.c
@@ -56,14 +56,14 @@ colormask_quad(struct quad_stage *qs, struct quad_header *quad)
       struct softpipe_cached_tile *tile
          = sp_get_cached_tile(softpipe,
                               softpipe->cbuf_cache[cbuf],
-                              quad->x0, quad->y0);
-      float (*quadColor)[4] = quad->outputs.color[cbuf];
+                              quad->input.x0, quad->input.y0);
+      float (*quadColor)[4] = quad->output.color[cbuf];
       uint i, j;
 
       /* get/swizzle dest colors */
       for (j = 0; j < QUAD_SIZE; j++) {
-         int x = (quad->x0 & (TILE_SIZE-1)) + (j & 1);
-         int y = (quad->y0 & (TILE_SIZE-1)) + (j >> 1);
+         int x = (quad->input.x0 & (TILE_SIZE-1)) + (j & 1);
+         int y = (quad->input.y0 & (TILE_SIZE-1)) + (j >> 1);
          for (i = 0; i < 4; i++) {
             dest[i][j] = tile->data.color[y][x][i];
          }
diff --git a/src/gallium/drivers/softpipe/sp_quad_coverage.c b/src/gallium/drivers/softpipe/sp_quad_coverage.c
index ad907ec25f..ee29aa7dfe 100644
--- a/src/gallium/drivers/softpipe/sp_quad_coverage.c
+++ b/src/gallium/drivers/softpipe/sp_quad_coverage.c
@@ -47,19 +47,19 @@ coverage_quad(struct quad_stage *qs, struct quad_header *quad)
 {
    struct softpipe_context *softpipe = qs->softpipe;
 
-   if ((softpipe->rasterizer->poly_smooth && quad->prim == PRIM_TRI) ||
-       (softpipe->rasterizer->line_smooth && quad->prim == PRIM_LINE) ||
-       (softpipe->rasterizer->point_smooth && quad->prim == PRIM_POINT)) {
+   if ((softpipe->rasterizer->poly_smooth && quad->input.prim == PRIM_TRI) ||
+       (softpipe->rasterizer->line_smooth && quad->input.prim == PRIM_LINE) ||
+       (softpipe->rasterizer->point_smooth && quad->input.prim == PRIM_POINT)) {
       uint cbuf;
 
       /* loop over colorbuffer outputs */
       for (cbuf = 0; cbuf < softpipe->framebuffer.num_cbufs; cbuf++) {
-         float (*quadColor)[4] = quad->outputs.color[cbuf];
+         float (*quadColor)[4] = quad->output.color[cbuf];
          unsigned j;
          for (j = 0; j < QUAD_SIZE; j++) {
-            assert(quad->coverage[j] >= 0.0);
-            assert(quad->coverage[j] <= 1.0);
-         quadColor[3][j] *= quad->coverage[j];
+            assert(quad->input.coverage[j] >= 0.0);
+            assert(quad->input.coverage[j] <= 1.0);
+         quadColor[3][j] *= quad->input.coverage[j];
          }
       }
    }
diff --git a/src/gallium/drivers/softpipe/sp_quad_depth_test.c b/src/gallium/drivers/softpipe/sp_quad_depth_test.c
index 227cb2014e..523bd3e080 100644
--- a/src/gallium/drivers/softpipe/sp_quad_depth_test.c
+++ b/src/gallium/drivers/softpipe/sp_quad_depth_test.c
@@ -60,7 +60,7 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad)
    unsigned zmask = 0;
    unsigned j;
    struct softpipe_cached_tile *tile
-      = sp_get_cached_tile(softpipe, softpipe->zsbuf_cache, quad->x0, quad->y0);
+      = sp_get_cached_tile(softpipe, softpipe->zsbuf_cache, quad->input.x0, quad->input.y0);
 
    assert(ps); /* shouldn't get here if there's no zbuffer */
 
@@ -79,12 +79,12 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad)
          float scale = 65535.0;
 
          for (j = 0; j < QUAD_SIZE; j++) {
-            qzzzz[j] = (unsigned) (quad->outputs.depth[j] * scale);
+            qzzzz[j] = (unsigned) (quad->output.depth[j] * scale);
          }
 
          for (j = 0; j < QUAD_SIZE; j++) {
-            int x = quad->x0 % TILE_SIZE + (j & 1);
-            int y = quad->y0 % TILE_SIZE + (j >> 1);
+            int x = quad->input.x0 % TILE_SIZE + (j & 1);
+            int y = quad->input.y0 % TILE_SIZE + (j >> 1);
             bzzzz[j] = tile->data.depth16[y][x];
          }
       }
@@ -94,12 +94,12 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad)
          double scale = (double) (uint) ~0UL;
 
          for (j = 0; j < QUAD_SIZE; j++) {
-            qzzzz[j] = (unsigned) (quad->outputs.depth[j] * scale);
+            qzzzz[j] = (unsigned) (quad->output.depth[j] * scale);
          }
 
          for (j = 0; j < QUAD_SIZE; j++) {
-            int x = quad->x0 % TILE_SIZE + (j & 1);
-            int y = quad->y0 % TILE_SIZE + (j >> 1);
+            int x = quad->input.x0 % TILE_SIZE + (j & 1);
+            int y = quad->input.y0 % TILE_SIZE + (j >> 1);
             bzzzz[j] = tile->data.depth32[y][x];
          }
       }
@@ -111,12 +111,12 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad)
          float scale = (float) ((1 << 24) - 1);
 
          for (j = 0; j < QUAD_SIZE; j++) {
-            qzzzz[j] = (unsigned) (quad->outputs.depth[j] * scale);
+            qzzzz[j] = (unsigned) (quad->output.depth[j] * scale);
          }
 
          for (j = 0; j < QUAD_SIZE; j++) {
-            int x = quad->x0 % TILE_SIZE + (j & 1);
-            int y = quad->y0 % TILE_SIZE + (j >> 1);
+            int x = quad->input.x0 % TILE_SIZE + (j & 1);
+            int y = quad->input.y0 % TILE_SIZE + (j >> 1);
             bzzzz[j] = tile->data.depth32[y][x] & 0xffffff;
          }
       }
@@ -128,12 +128,12 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad)
          float scale = (float) ((1 << 24) - 1);
 
          for (j = 0; j < QUAD_SIZE; j++) {
-            qzzzz[j] = (unsigned) (quad->outputs.depth[j] * scale);
+            qzzzz[j] = (unsigned) (quad->output.depth[j] * scale);
          }
 
          for (j = 0; j < QUAD_SIZE; j++) {
-            int x = quad->x0 % TILE_SIZE + (j & 1);
-            int y = quad->y0 % TILE_SIZE + (j >> 1);
+            int x = quad->input.x0 % TILE_SIZE + (j & 1);
+            int y = quad->input.y0 % TILE_SIZE + (j >> 1);
             bzzzz[j] = tile->data.depth32[y][x] >> 8;
          }
       }
@@ -192,14 +192,14 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad)
       assert(0);
    }
 
-   quad->mask &= zmask;
+   quad->inout.mask &= zmask;
 
    if (softpipe->depth_stencil->depth.writemask) {
       
       /* This is also efficient with sse / spe instructions: 
        */
       for (j = 0; j < QUAD_SIZE; j++) {
-	 if (quad->mask & (1 << j)) {
+	 if (quad->inout.mask & (1 << j)) {
 	    bzzzz[j] = qzzzz[j];
 	 }
       }
@@ -208,8 +208,8 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad)
       switch (format) {
       case PIPE_FORMAT_Z16_UNORM:
          for (j = 0; j < QUAD_SIZE; j++) {
-            int x = quad->x0 % TILE_SIZE + (j & 1);
-            int y = quad->y0 % TILE_SIZE + (j >> 1);
+            int x = quad->input.x0 % TILE_SIZE + (j & 1);
+            int y = quad->input.y0 % TILE_SIZE + (j >> 1);
             tile->data.depth16[y][x] = (ushort) bzzzz[j];
          }
          break;
@@ -218,15 +218,15 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad)
          /* (yes, this falls through to a different case than above) */
       case PIPE_FORMAT_Z32_UNORM:
          for (j = 0; j < QUAD_SIZE; j++) {
-            int x = quad->x0 % TILE_SIZE + (j & 1);
-            int y = quad->y0 % TILE_SIZE + (j >> 1);
+            int x = quad->input.x0 % TILE_SIZE + (j & 1);
+            int y = quad->input.y0 % TILE_SIZE + (j >> 1);
             tile->data.depth32[y][x] = bzzzz[j];
          }
          break;
       case PIPE_FORMAT_S8Z24_UNORM:
          for (j = 0; j < QUAD_SIZE; j++) {
-            int x = quad->x0 % TILE_SIZE + (j & 1);
-            int y = quad->y0 % TILE_SIZE + (j >> 1);
+            int x = quad->input.x0 % TILE_SIZE + (j & 1);
+            int y = quad->input.y0 % TILE_SIZE + (j >> 1);
             uint s8z24 = tile->data.depth32[y][x];
             s8z24 = (s8z24 & 0xff000000) | bzzzz[j];
             tile->data.depth32[y][x] = s8z24;
@@ -234,8 +234,8 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad)
          break;
       case PIPE_FORMAT_Z24S8_UNORM:
          for (j = 0; j < QUAD_SIZE; j++) {
-            int x = quad->x0 % TILE_SIZE + (j & 1);
-            int y = quad->y0 % TILE_SIZE + (j >> 1);
+            int x = quad->input.x0 % TILE_SIZE + (j & 1);
+            int y = quad->input.y0 % TILE_SIZE + (j >> 1);
             uint z24s8 = tile->data.depth32[y][x];
             z24s8 = (z24s8 & 0xff) | (bzzzz[j] << 8);
             tile->data.depth32[y][x] = z24s8;
@@ -243,8 +243,8 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad)
          break;
       case PIPE_FORMAT_Z24X8_UNORM:
          for (j = 0; j < QUAD_SIZE; j++) {
-            int x = quad->x0 % TILE_SIZE + (j & 1);
-            int y = quad->y0 % TILE_SIZE + (j >> 1);
+            int x = quad->input.x0 % TILE_SIZE + (j & 1);
+            int y = quad->input.y0 % TILE_SIZE + (j >> 1);
             tile->data.depth32[y][x] = bzzzz[j] << 8;
          }
          break;
@@ -260,7 +260,7 @@ depth_test_quad(struct quad_stage *qs, struct quad_header *quad)
 {
    sp_depth_test_quad(qs, quad);
 
-   if (quad->mask)
+   if (quad->inout.mask)
       qs->next->run(qs->next, quad);
 }
 
diff --git a/src/gallium/drivers/softpipe/sp_quad_earlyz.c b/src/gallium/drivers/softpipe/sp_quad_earlyz.c
index 5a66a86699..6e2dde304e 100644
--- a/src/gallium/drivers/softpipe/sp_quad_earlyz.c
+++ b/src/gallium/drivers/softpipe/sp_quad_earlyz.c
@@ -45,16 +45,16 @@ earlyz_quad(
    struct quad_stage    *qs,
    struct quad_header   *quad )
 {
-   const float fx = (float) quad->x0;
-   const float fy = (float) quad->y0;
+   const float fx = (float) quad->input.x0;
+   const float fy = (float) quad->input.y0;
    const float dzdx = quad->posCoef->dadx[2];
    const float dzdy = quad->posCoef->dady[2];
    const float z0 = quad->posCoef->a0[2] + dzdx * fx + dzdy * fy;
 
-   quad->outputs.depth[0] = z0;
-   quad->outputs.depth[1] = z0 + dzdx;
-   quad->outputs.depth[2] = z0 + dzdy;
-   quad->outputs.depth[3] = z0 + dzdx + dzdy;
+   quad->output.depth[0] = z0;
+   quad->output.depth[1] = z0 + dzdx;
+   quad->output.depth[2] = z0 + dzdy;
+   quad->output.depth[3] = z0 + dzdx + dzdy;
 
    qs->next->run( qs->next, quad );
 }
diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c
index 5499ba5361..1f0cb3e035 100644
--- a/src/gallium/drivers/softpipe/sp_quad_fs.c
+++ b/src/gallium/drivers/softpipe/sp_quad_fs.c
@@ -85,7 +85,7 @@ shade_quad(
    machine->InterpCoefs = quad->coef;
 
    /* run shader */
-   quad->mask &= softpipe->fs->run( softpipe->fs, 
+   quad->inout.mask &= softpipe->fs->run( softpipe->fs, 
 				    &qss->machine,
 				    quad );
 
@@ -101,16 +101,16 @@ shade_quad(
          case TGSI_SEMANTIC_COLOR:
             {
                uint cbuf = sem_index[i];
-               memcpy(quad->outputs.color[cbuf],
+               memcpy(quad->output.color[cbuf],
                       &machine->Outputs[i].xyzw[0].f[0],
-                      sizeof(quad->outputs.color[0]) );
+                      sizeof(quad->output.color[0]) );
             }
             break;
          case TGSI_SEMANTIC_POSITION:
             {
                uint j;
                for (j = 0; j < 4; j++) {
-                  quad->outputs.depth[j] = machine->Outputs[0].xyzw[2].f[j];
+                  quad->output.depth[j] = machine->Outputs[0].xyzw[2].f[j];
                }
                z_written = TRUE;
             }
@@ -122,20 +122,20 @@ shade_quad(
    if (!z_written) {
       /* compute Z values now, as in the quad earlyz stage */
       /* XXX we should really only do this if the earlyz stage is not used */
-      const float fx = (float) quad->x0;
-      const float fy = (float) quad->y0;
+      const float fx = (float) quad->input.x0;
+      const float fy = (float) quad->input.y0;
       const float dzdx = quad->posCoef->dadx[2];
       const float dzdy = quad->posCoef->dady[2];
       const float z0 = quad->posCoef->a0[2] + dzdx * fx + dzdy * fy;
 
-      quad->outputs.depth[0] = z0;
-      quad->outputs.depth[1] = z0 + dzdx;
-      quad->outputs.depth[2] = z0 + dzdy;
-      quad->outputs.depth[3] = z0 + dzdx + dzdy;
+      quad->output.depth[0] = z0;
+      quad->output.depth[1] = z0 + dzdx;
+      quad->output.depth[2] = z0 + dzdy;
+      quad->output.depth[3] = z0 + dzdx + dzdy;
    }
 
    /* shader may cull fragments */
-   if( quad->mask ) {
+   if( quad->inout.mask ) {
       qs->next->run( qs->next, quad );
    }
 }
diff --git a/src/gallium/drivers/softpipe/sp_quad_occlusion.c b/src/gallium/drivers/softpipe/sp_quad_occlusion.c
index db13e73ae3..169bd82876 100644
--- a/src/gallium/drivers/softpipe/sp_quad_occlusion.c
+++ b/src/gallium/drivers/softpipe/sp_quad_occlusion.c
@@ -54,7 +54,7 @@ occlusion_count_quad(struct quad_stage *qs, struct quad_header *quad)
 {
    struct softpipe_context *softpipe = qs->softpipe;
 
-   softpipe->occlusion_count += count_bits(quad->mask);
+   softpipe->occlusion_count += count_bits(quad->inout.mask);
 
    qs->next->run(qs->next, quad);
 }
diff --git a/src/gallium/drivers/softpipe/sp_quad_output.c b/src/gallium/drivers/softpipe/sp_quad_output.c
index b64646a449..d05e12d1d9 100644
--- a/src/gallium/drivers/softpipe/sp_quad_output.c
+++ b/src/gallium/drivers/softpipe/sp_quad_output.c
@@ -41,8 +41,8 @@ static void
 output_quad(struct quad_stage *qs, struct quad_header *quad)
 {
    /* in-tile pos: */
-   const int itx = quad->x0 % TILE_SIZE;
-   const int ity = quad->y0 % TILE_SIZE;
+   const int itx = quad->input.x0 % TILE_SIZE;
+   const int ity = quad->input.y0 % TILE_SIZE;
 
    struct softpipe_context *softpipe = qs->softpipe;
    uint cbuf;
@@ -52,13 +52,13 @@ output_quad(struct quad_stage *qs, struct quad_header *quad)
       struct softpipe_cached_tile *tile
          = sp_get_cached_tile(softpipe,
                               softpipe->cbuf_cache[cbuf],
-                              quad->x0, quad->y0);
-      float (*quadColor)[4] = quad->outputs.color[cbuf];
+                              quad->input.x0, quad->input.y0);
+      float (*quadColor)[4] = quad->output.color[cbuf];
       int i, j;
 
       /* get/swizzle dest colors */
       for (j = 0; j < QUAD_SIZE; j++) {
-         if (quad->mask & (1 << j)) {
+         if (quad->inout.mask & (1 << j)) {
             int x = itx + (j & 1);
             int y = ity + (j >> 1);
             for (i = 0; i < 4; i++) { /* loop over color chans */
diff --git a/src/gallium/drivers/softpipe/sp_quad_stencil.c b/src/gallium/drivers/softpipe/sp_quad_stencil.c
index ce9562e07c..abb5487748 100644
--- a/src/gallium/drivers/softpipe/sp_quad_stencil.c
+++ b/src/gallium/drivers/softpipe/sp_quad_stencil.c
@@ -206,9 +206,9 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad)
    ubyte ref, wrtMask, valMask;
    ubyte stencilVals[QUAD_SIZE];
    struct softpipe_cached_tile *tile
-      = sp_get_cached_tile(softpipe, softpipe->zsbuf_cache, quad->x0, quad->y0);
+      = sp_get_cached_tile(softpipe, softpipe->zsbuf_cache, quad->input.x0, quad->input.y0);
    uint j;
-   uint face = quad->facing;
+   uint face = quad->input.facing;
 
    if (!softpipe->depth_stencil->stencil[1].enabled) {
       /* single-sided stencil test, use front (face=0) state */
@@ -231,22 +231,22 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad)
    switch (ps->format) {
    case PIPE_FORMAT_S8Z24_UNORM:
       for (j = 0; j < QUAD_SIZE; j++) {
-         int x = quad->x0 % TILE_SIZE + (j & 1);
-         int y = quad->y0 % TILE_SIZE + (j >> 1);
+         int x = quad->input.x0 % TILE_SIZE + (j & 1);
+         int y = quad->input.y0 % TILE_SIZE + (j >> 1);
          stencilVals[j] = tile->data.depth32[y][x] >> 24;
       }
       break;
    case PIPE_FORMAT_Z24S8_UNORM:
       for (j = 0; j < QUAD_SIZE; j++) {
-         int x = quad->x0 % TILE_SIZE + (j & 1);
-         int y = quad->y0 % TILE_SIZE + (j >> 1);
+         int x = quad->input.x0 % TILE_SIZE + (j & 1);
+         int y = quad->input.y0 % TILE_SIZE + (j >> 1);
          stencilVals[j] = tile->data.depth32[y][x] & 0xff;
       }
       break;
    case PIPE_FORMAT_S8_UNORM:
       for (j = 0; j < QUAD_SIZE; j++) {
-         int x = quad->x0 % TILE_SIZE + (j & 1);
-         int y = quad->y0 % TILE_SIZE + (j >> 1);
+         int x = quad->input.x0 % TILE_SIZE + (j & 1);
+         int y = quad->input.y0 % TILE_SIZE + (j >> 1);
          stencilVals[j] = tile->data.stencil8[y][x];
       }
       break;
@@ -258,35 +258,35 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad)
    {
       unsigned passMask, failMask;
       passMask = do_stencil_test(stencilVals, func, ref, valMask);
-      failMask = quad->mask & ~passMask;
-      quad->mask &= passMask;
+      failMask = quad->inout.mask & ~passMask;
+      quad->inout.mask &= passMask;
 
       if (failOp != PIPE_STENCIL_OP_KEEP) {
          apply_stencil_op(stencilVals, failMask, failOp, ref, wrtMask);
       }
    }
 
-   if (quad->mask) {
+   if (quad->inout.mask) {
       /* now the pixels that passed the stencil test are depth tested */
       if (softpipe->depth_stencil->depth.enabled) {
-         const unsigned origMask = quad->mask;
+         const unsigned origMask = quad->inout.mask;
 
          sp_depth_test_quad(qs, quad);  /* quad->mask is updated */
 
          /* update stencil buffer values according to z pass/fail result */
          if (zFailOp != PIPE_STENCIL_OP_KEEP) {
-            const unsigned failMask = origMask & ~quad->mask;
+            const unsigned failMask = origMask & ~quad->inout.mask;
             apply_stencil_op(stencilVals, failMask, zFailOp, ref, wrtMask);
          }
 
          if (zPassOp != PIPE_STENCIL_OP_KEEP) {
-            const unsigned passMask = origMask & quad->mask;
+            const unsigned passMask = origMask & quad->inout.mask;
             apply_stencil_op(stencilVals, passMask, zPassOp, ref, wrtMask);
          }
       }
       else {
          /* no depth test, apply Zpass operator to stencil buffer values */
-         apply_stencil_op(stencilVals, quad->mask, zPassOp, ref, wrtMask);
+         apply_stencil_op(stencilVals, quad->inout.mask, zPassOp, ref, wrtMask);
       }
 
    }
@@ -295,8 +295,8 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad)
    switch (ps->format) {
    case PIPE_FORMAT_S8Z24_UNORM:
       for (j = 0; j < QUAD_SIZE; j++) {
-         int x = quad->x0 % TILE_SIZE + (j & 1);
-         int y = quad->y0 % TILE_SIZE + (j >> 1);
+         int x = quad->input.x0 % TILE_SIZE + (j & 1);
+         int y = quad->input.y0 % TILE_SIZE + (j >> 1);
          uint s8z24 = tile->data.depth32[y][x];
          s8z24 = (stencilVals[j] << 24) | (s8z24 & 0xffffff);
          tile->data.depth32[y][x] = s8z24;
@@ -304,8 +304,8 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad)
       break;
    case PIPE_FORMAT_Z24S8_UNORM:
       for (j = 0; j < QUAD_SIZE; j++) {
-         int x = quad->x0 % TILE_SIZE + (j & 1);
-         int y = quad->y0 % TILE_SIZE + (j >> 1);
+         int x = quad->input.x0 % TILE_SIZE + (j & 1);
+         int y = quad->input.y0 % TILE_SIZE + (j >> 1);
          uint z24s8 = tile->data.depth32[y][x];
          z24s8 = (z24s8 & 0xffffff00) | stencilVals[j];
          tile->data.depth32[y][x] = z24s8;
@@ -313,8 +313,8 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad)
       break;
    case PIPE_FORMAT_S8_UNORM:
       for (j = 0; j < QUAD_SIZE; j++) {
-         int x = quad->x0 % TILE_SIZE + (j & 1);
-         int y = quad->y0 % TILE_SIZE + (j >> 1);
+         int x = quad->input.x0 % TILE_SIZE + (j & 1);
+         int y = quad->input.y0 % TILE_SIZE + (j >> 1);
          tile->data.stencil8[y][x] = stencilVals[j];
       }
       break;
@@ -322,7 +322,7 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad)
       assert(0);
    }
 
-   if (quad->mask)
+   if (quad->inout.mask)
       qs->next->run(qs->next, quad);
 }
 
diff --git a/src/gallium/drivers/softpipe/sp_quad_stipple.c b/src/gallium/drivers/softpipe/sp_quad_stipple.c
index a39ecc2e9d..ccf37f6be5 100644
--- a/src/gallium/drivers/softpipe/sp_quad_stipple.c
+++ b/src/gallium/drivers/softpipe/sp_quad_stipple.c
@@ -19,17 +19,17 @@ stipple_quad(struct quad_stage *qs, struct quad_header *quad)
    static const uint bit31 = 1 << 31;
    static const uint bit30 = 1 << 30;
 
-   if (quad->prim == PRIM_TRI) {
+   if (quad->input.prim == PRIM_TRI) {
       struct softpipe_context *softpipe = qs->softpipe;
       /* need to invert Y to index into OpenGL's stipple pattern */
       int y0, y1;
       uint stipple0, stipple1;
       if (softpipe->rasterizer->origin_lower_left) {
-         y0 = softpipe->framebuffer.height - 1 - quad->y0;
+         y0 = softpipe->framebuffer.height - 1 - quad->input.y0;
          y1 = y0 - 1;
       }
       else {
-         y0 = quad->y0;
+         y0 = quad->input.y0;
          y1 = y0 + 1;
       }
       stipple0 = softpipe->poly_stipple.stipple[y0 % 32];
@@ -37,18 +37,18 @@ stipple_quad(struct quad_stage *qs, struct quad_header *quad)
 
 #if 1
       {
-      const int col0 = quad->x0 % 32;
+      const int col0 = quad->input.x0 % 32;
       if ((stipple0 & (bit31 >> col0)) == 0)
-         quad->mask &= ~MASK_TOP_LEFT;
+         quad->inout.mask &= ~MASK_TOP_LEFT;
 
       if ((stipple0 & (bit30 >> col0)) == 0)
-         quad->mask &= ~MASK_TOP_RIGHT;
+         quad->inout.mask &= ~MASK_TOP_RIGHT;
 
       if ((stipple1 & (bit31 >> col0)) == 0)
-         quad->mask &= ~MASK_BOTTOM_LEFT;
+         quad->inout.mask &= ~MASK_BOTTOM_LEFT;
 
       if ((stipple1 & (bit30 >> col0)) == 0)
-         quad->mask &= ~MASK_BOTTOM_RIGHT;
+         quad->inout.mask &= ~MASK_BOTTOM_RIGHT;
       }
 #else
       /* We'd like to use this code, but we'd need to redefine
@@ -56,11 +56,11 @@ stipple_quad(struct quad_stage *qs, struct quad_header *quad)
        * and similarly for the BOTTOM bits.  But that may have undesirable
        * side effects elsewhere.
        */
-      const int col0 = 30 - (quad->x0 % 32);
-      quad->mask &= (((stipple0 >> col0) & 0x3) | 
+      const int col0 = 30 - (quad->input.x0 % 32);
+      quad->inout.mask &= (((stipple0 >> col0) & 0x3) | 
                      (((stipple1 >> col0) & 0x3) << 2));
 #endif
-      if (!quad->mask)
+      if (!quad->inout.mask)
          return;
    }
 
diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c
index 72298dfbb0..965d62b29e 100644
--- a/src/gallium/drivers/softpipe/sp_setup.c
+++ b/src/gallium/drivers/softpipe/sp_setup.c
@@ -298,22 +298,22 @@ quad_clip( struct setup_context *setup, struct quad_header *quad )
    const int miny = (int) cliprect->miny;
    const int maxy = (int) cliprect->maxy;
 
-   if (quad->x0 >= maxx ||
-       quad->y0 >= maxy ||
-       quad->x0 + 1 < minx ||
-       quad->y0 + 1 < miny) {
+   if (quad->input.x0 >= maxx ||
+       quad->input.y0 >= maxy ||
+       quad->input.x0 + 1 < minx ||
+       quad->input.y0 + 1 < miny) {
       /* totally clipped */
-      quad->mask = 0x0;
+      quad->inout.mask = 0x0;
       return;
    }
-   if (quad->x0 < minx)
-      quad->mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT);
-   if (quad->y0 < miny)
-      quad->mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT);
-   if (quad->x0 == maxx - 1)
-      quad->mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT);
-   if (quad->y0 == maxy - 1)
-      quad->mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT);
+   if (quad->input.x0 < minx)
+      quad->inout.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT);
+   if (quad->input.y0 < miny)
+      quad->inout.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT);
+   if (quad->input.x0 == maxx - 1)
+      quad->inout.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT);
+   if (quad->input.y0 == maxy - 1)
+      quad->inout.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT);
 }
 
 
@@ -324,7 +324,7 @@ static INLINE void
 clip_emit_quad( struct setup_context *setup, struct quad_header *quad, uint thread )
 {
    quad_clip( setup, quad );
-   if (quad->mask) {
+   if (quad->inout.mask) {
       struct softpipe_context *sp = setup->softpipe;
 
       sp->quad[thread].first->run( sp->quad[thread].first, quad );
@@ -355,9 +355,9 @@ emit_quad( struct setup_context *setup, int x, int y, unsigned mask, struct quad
 {
    struct softpipe_context *sp = setup->softpipe;
 
-   quad->x0 = x;
-   quad->y0 = y;
-   quad->mask = mask;
+   quad->input.x0 = x;
+   quad->input.y0 = y;
+   quad->inout.mask = mask;
 #if DEBUG_FRAGS
    if (mask & 1) setup->numFragsEmitted++;
    if (mask & 2) setup->numFragsEmitted++;
@@ -366,7 +366,7 @@ emit_quad( struct setup_context *setup, int x, int y, unsigned mask, struct quad
 #endif
    sp->quad[thread].first->run( sp->quad[thread].first, quad );
 #if DEBUG_FRAGS
-   mask = quad->mask;
+   mask = quad->inout.mask;
    if (mask & 1) setup->numFragsWritten++;
    if (mask & 2) setup->numFragsWritten++;
    if (mask & 4) setup->numFragsWritten++;
@@ -577,7 +577,7 @@ static boolean setup_sort_vertices( struct setup_context *setup,
     *  - the GLSL gl_FrontFacing fragment attribute (bool)
     *  - two-sided stencil test
     */
-   setup->quad.facing = (det > 0.0) ^ (setup->softpipe->rasterizer->front_winding == PIPE_WINDING_CW);
+   setup->quad.input.facing = (det > 0.0) ^ (setup->softpipe->rasterizer->front_winding == PIPE_WINDING_CW);
 
    return TRUE;
 }
@@ -774,7 +774,7 @@ static void setup_tri_coefficients( struct setup_context *setup )
 
       if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) {
          /* FOG.y = front/back facing  XXX fix this */
-         setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing;
+         setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.input.facing;
          setup->coef[fragSlot].dadx[1] = 0.0;
          setup->coef[fragSlot].dady[1] = 0.0;
       }
@@ -944,7 +944,7 @@ void setup_tri( struct setup_context *setup,
    setup_tri_coefficients( setup );
    setup_tri_edges( setup );
 
-   setup->quad.prim = PRIM_TRI;
+   setup->quad.input.prim = PRIM_TRI;
 
    setup->span.y = 0;
    setup->span.y_flags = 0;
@@ -1085,7 +1085,7 @@ setup_line_coefficients(struct setup_context *setup,
 
       if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) {
          /* FOG.y = front/back facing  XXX fix this */
-         setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing;
+         setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.input.facing;
          setup->coef[fragSlot].dadx[1] = 0.0;
          setup->coef[fragSlot].dady[1] = 0.0;
       }
@@ -1106,20 +1106,20 @@ plot(struct setup_context *setup, int x, int y)
    const int quadY = y - iy;
    const int mask = (1 << ix) << (2 * iy);
 
-   if (quadX != setup->quad.x0 ||
-       quadY != setup->quad.y0)
+   if (quadX != setup->quad.input.x0 ||
+       quadY != setup->quad.input.y0)
    {
       /* flush prev quad, start new quad */
 
-      if (setup->quad.x0 != -1)
+      if (setup->quad.input.x0 != -1)
          CLIP_EMIT_QUAD(setup);
 
-      setup->quad.x0 = quadX;
-      setup->quad.y0 = quadY;
-      setup->quad.mask = 0x0;
+      setup->quad.input.x0 = quadX;
+      setup->quad.input.y0 = quadY;
+      setup->quad.inout.mask = 0x0;
    }
 
-   setup->quad.mask |= mask;
+   setup->quad.inout.mask |= mask;
 }
 
 
@@ -1180,16 +1180,16 @@ setup_line(struct setup_context *setup,
    assert(dx >= 0);
    assert(dy >= 0);
 
-   setup->quad.x0 = setup->quad.y0 = -1;
-   setup->quad.mask = 0x0;
-   setup->quad.prim = PRIM_LINE;
+   setup->quad.input.x0 = setup->quad.input.y0 = -1;
+   setup->quad.inout.mask = 0x0;
+   setup->quad.input.prim = PRIM_LINE;
    /* XXX temporary: set coverage to 1.0 so the line appears
     * if AA mode happens to be enabled.
     */
-   setup->quad.coverage[0] =
-   setup->quad.coverage[1] =
-   setup->quad.coverage[2] =
-   setup->quad.coverage[3] = 1.0;
+   setup->quad.input.coverage[0] =
+   setup->quad.input.coverage[1] =
+   setup->quad.input.coverage[2] =
+   setup->quad.input.coverage[3] = 1.0;
 
    if (dx > dy) {
       /*** X-major line ***/
@@ -1233,7 +1233,7 @@ setup_line(struct setup_context *setup,
    }
 
    /* draw final quad */
-   if (setup->quad.mask) {
+   if (setup->quad.inout.mask) {
       CLIP_EMIT_QUAD(setup);
    }
 
@@ -1331,21 +1331,21 @@ setup_point( struct setup_context *setup,
 
       if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) {
          /* FOG.y = front/back facing  XXX fix this */
-         setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing;
+         setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.input.facing;
          setup->coef[fragSlot].dadx[1] = 0.0;
          setup->coef[fragSlot].dady[1] = 0.0;
       }
    }
 
-   setup->quad.prim = PRIM_POINT;
+   setup->quad.input.prim = PRIM_POINT;
 
    if (halfSize <= 0.5 && !round) {
       /* special case for 1-pixel points */
       const int ix = ((int) x) & 1;
       const int iy = ((int) y) & 1;
-      setup->quad.x0 = (int) x - ix;
-      setup->quad.y0 = (int) y - iy;
-      setup->quad.mask = (1 << ix) << (2 * iy);
+      setup->quad.input.x0 = (int) x - ix;
+      setup->quad.input.y0 = (int) y - iy;
+      setup->quad.inout.mask = (1 << ix) << (2 * iy);
       CLIP_EMIT_QUAD(setup);
    }
    else {
@@ -1366,15 +1366,15 @@ setup_point( struct setup_context *setup,
             for (ix = ixmin; ix <= ixmax; ix += 2) {
                float dx, dy, dist2, cover;
 
-               setup->quad.mask = 0x0;
+               setup->quad.inout.mask = 0x0;
 
                dx = (ix + 0.5f) - x;
                dy = (iy + 0.5f) - y;
                dist2 = dx * dx + dy * dy;
                if (dist2 <= rmax2) {
                   cover = 1.0F - (dist2 - rmin2) * cscale;
-                  setup->quad.coverage[QUAD_TOP_LEFT] = MIN2(cover, 1.0f);
-                  setup->quad.mask |= MASK_TOP_LEFT;
+                  setup->quad.input.coverage[QUAD_TOP_LEFT] = MIN2(cover, 1.0f);
+                  setup->quad.inout.mask |= MASK_TOP_LEFT;
                }
 
                dx = (ix + 1.5f) - x;
@@ -1382,8 +1382,8 @@ setup_point( struct setup_context *setup,
                dist2 = dx * dx + dy * dy;
                if (dist2 <= rmax2) {
                   cover = 1.0F - (dist2 - rmin2) * cscale;
-                  setup->quad.coverage[QUAD_TOP_RIGHT] = MIN2(cover, 1.0f);
-                  setup->quad.mask |= MASK_TOP_RIGHT;
+                  setup->quad.input.coverage[QUAD_TOP_RIGHT] = MIN2(cover, 1.0f);
+                  setup->quad.inout.mask |= MASK_TOP_RIGHT;
                }
 
                dx = (ix + 0.5f) - x;
@@ -1391,8 +1391,8 @@ setup_point( struct setup_context *setup,
                dist2 = dx * dx + dy * dy;
                if (dist2 <= rmax2) {
                   cover = 1.0F - (dist2 - rmin2) * cscale;
-                  setup->quad.coverage[QUAD_BOTTOM_LEFT] = MIN2(cover, 1.0f);
-                  setup->quad.mask |= MASK_BOTTOM_LEFT;
+                  setup->quad.input.coverage[QUAD_BOTTOM_LEFT] = MIN2(cover, 1.0f);
+                  setup->quad.inout.mask |= MASK_BOTTOM_LEFT;
                }
 
                dx = (ix + 1.5f) - x;
@@ -1400,13 +1400,13 @@ setup_point( struct setup_context *setup,
                dist2 = dx * dx + dy * dy;
                if (dist2 <= rmax2) {
                   cover = 1.0F - (dist2 - rmin2) * cscale;
-                  setup->quad.coverage[QUAD_BOTTOM_RIGHT] = MIN2(cover, 1.0f);
-                  setup->quad.mask |= MASK_BOTTOM_RIGHT;
+                  setup->quad.input.coverage[QUAD_BOTTOM_RIGHT] = MIN2(cover, 1.0f);
+                  setup->quad.inout.mask |= MASK_BOTTOM_RIGHT;
                }
 
-               if (setup->quad.mask) {
-                  setup->quad.x0 = ix;
-                  setup->quad.y0 = iy;
+               if (setup->quad.inout.mask) {
+                  setup->quad.input.x0 = ix;
+                  setup->quad.input.y0 = iy;
                   CLIP_EMIT_QUAD(setup);
                }
             }
@@ -1451,9 +1451,9 @@ setup_point( struct setup_context *setup,
                   mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT);
                }
 
-               setup->quad.mask = mask;
-               setup->quad.x0 = ix;
-               setup->quad.y0 = iy;
+               setup->quad.inout.mask = mask;
+               setup->quad.input.x0 = ix;
+               setup->quad.input.y0 = iy;
                CLIP_EMIT_QUAD(setup);
             }
          }