draw: implement vertex color clamping, and disable SSE and PPC paths

(some little changes by Marek Olšák)

Squashed commit of the following:

commit 737c0c6b7d591ac0fc969a7590e1691eeef0ce5e
Author: Luca Barbieri <luca@luca-barbieri.com>
Date:   Fri Aug 27 02:13:57 2010 +0200

    draw: disable SSE and PPC paths (use LLVM instead)

    These paths don't support vertex clamping, and are anyway
    obsoleted by LLVM.

    If you want to re-enable them, add vertex clamping and test that it
    works with the ARB_color_buffer_float piglit tests.

commit fed3486a7ca0683b403913604a26ee49a3ef48c7
Author: Luca Barbieri <luca@luca-barbieri.com>
Date:   Thu Aug 26 18:27:38 2010 +0200

    draw_llvm: respect vertex color clamp

commit ef0efe9f3d1d0f9b40ebab78940491d2154277a9
Author: Luca Barbieri <luca@luca-barbieri.com>
Date:   Thu Aug 26 18:26:43 2010 +0200

    draw: respect vertex clamping in interpreter path

diff --git a/src/gallium/auxiliary/draw/draw_llvm.c b/src/gallium/auxiliary/draw/draw_llvm.c
index a5217c1d4ec8c1bf3cc76b5f1bf618ff76e271da..27c5f3bff5ac45fcc07ac6fc2640451ea531cc2c 100644 (file)
--- a/src/gallium/auxiliary/draw/draw_llvm.c
+++ b/src/gallium/auxiliary/draw/draw_llvm.c
@@ -438,7 +438,8 @@ generate_vs(struct draw_llvm *llvm,
             const LLVMValueRef (*inputs)[NUM_CHANNELS],
             LLVMValueRef system_values_array,
             LLVMValueRef context_ptr,
-            struct lp_build_sampler_soa *draw_sampler)
+            struct lp_build_sampler_soa *draw_sampler,
+            boolean clamp_vertex_color)
 {
    const struct tgsi_token *tokens = llvm->draw->vs.vertex_shader->state.tokens;
    struct lp_type vs_type;
@@ -474,6 +475,30 @@ generate_vs(struct draw_llvm *llvm,
                      outputs,
                      sampler,
                      &llvm->draw->vs.vertex_shader->info);
+
+   if(clamp_vertex_color)
+   {
+      LLVMValueRef out;
+      unsigned chan, attrib;
+      struct lp_build_context bld;
+      struct tgsi_shader_info* info = &llvm->draw->vs.vertex_shader->info;
+      lp_build_context_init(&bld, llvm->gallivm, vs_type);
+
+      for (attrib = 0; attrib < info->num_outputs; ++attrib) {
+         for(chan = 0; chan < NUM_CHANNELS; ++chan) {
+            if(outputs[attrib][chan]) {
+               switch (info->output_semantic_name[attrib]) {
+               case TGSI_SEMANTIC_COLOR:
+               case TGSI_SEMANTIC_BCOLOR:
+                  out = LLVMBuildLoad(builder, outputs[attrib][chan], "");
+                  out = lp_build_clamp(&bld, out, bld.zero, bld.one);
+                  LLVMBuildStore(builder, out, outputs[attrib][chan]);
+                  break;
+               }
+            }
+         }
+      }
+   }
 }
 
 #if DEBUG_STORE
@@ -1235,7 +1260,8 @@ draw_llvm_generate(struct draw_llvm *llvm, struct draw_llvm_variant *variant)
                   ptr_aos,
                   system_values_array,
                   context_ptr,
-                  sampler);
+                  sampler,
+                  variant->key.clamp_vertex_color);
 
       /* store original positions in clip before further manipulation */
       store_clip(gallivm, io, outputs);
@@ -1446,7 +1472,8 @@ draw_llvm_generate_elts(struct draw_llvm *llvm, struct draw_llvm_variant *varian
                   ptr_aos,
                   system_values_array,
                   context_ptr,
-                  sampler);
+                  sampler,
+                  variant->key.clamp_vertex_color);
 
       /* store original positions in clip before further manipulation */
       store_clip(gallivm, io, outputs);
@@ -1524,6 +1551,8 @@ draw_llvm_make_variant_key(struct draw_llvm *llvm, char *store)
 
    key = (struct draw_llvm_variant_key *)store;
 
+   key->clamp_vertex_color = llvm->draw->rasterizer->clamp_vertex_color; /**/
+
    /* Presumably all variants of the shader should have the same
     * number of vertex elements - ie the number of shader inputs.
     */

diff --git a/src/gallium/auxiliary/draw/draw_llvm.h b/src/gallium/auxiliary/draw/draw_llvm.h
index e8623e7bcdc6c2c085054d08c94f672ddd977173..873a272e98b35959cf502dc908b0049f21d20268 100644 (file)
--- a/src/gallium/auxiliary/draw/draw_llvm.h
+++ b/src/gallium/auxiliary/draw/draw_llvm.h
@@ -162,6 +162,7 @@ struct draw_llvm_variant_key
 {
    unsigned nr_vertex_elements:8;
    unsigned nr_samplers:8;
+   unsigned clamp_vertex_color:1;
    unsigned clip_xy:1;
    unsigned clip_z:1;
    unsigned clip_user:1;
@@ -169,7 +170,7 @@ struct draw_llvm_variant_key
    unsigned bypass_viewport:1;
    unsigned need_edgeflags:1;
    unsigned nr_planes:4;
-   unsigned pad:6;
+   unsigned pad:5;
 
    /* Variable number of vertex elements:
     */

diff --git a/src/gallium/auxiliary/draw/draw_vs.c b/src/gallium/auxiliary/draw/draw_vs.c
index 7caad6f0053a77d2d0cfc76de426c7b40e490197..1763dbc199f4c8a41a6ab75d706e289e95e0bc65 100644 (file)
--- a/src/gallium/auxiliary/draw/draw_vs.c
+++ b/src/gallium/auxiliary/draw/draw_vs.c
@@ -104,10 +104,17 @@ draw_create_vertex_shader(struct draw_context *draw,
    }
 
    if (!draw->pt.middle.llvm) {
+#if 0
+/* these paths don't support vertex clamping
+ * TODO: either add it, or remove them completely
+ * use LLVM instead if you want performance
+ * use exec instead if you want debugging/more correctness
+ */
 #if defined(PIPE_ARCH_X86)
       vs = draw_create_vs_sse( draw, shader );
 #elif defined(PIPE_ARCH_PPC)
       vs = draw_create_vs_ppc( draw, shader );
+#endif
 #endif
    }
 #if HAVE_LLVM

diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c
index c41d7c42a018e20ff58aafe47b89894210eec249..d9c4209a42b42419d9ee8223f6b266816ac33b39 100644 (file)
--- a/src/gallium/auxiliary/draw/draw_vs_exec.c
+++ b/src/gallium/auxiliary/draw/draw_vs_exec.c
@@ -95,6 +95,7 @@ vs_exec_run_linear( struct draw_vertex_shader *shader,
    struct tgsi_exec_machine *machine = evs->machine;
    unsigned int i, j;
    unsigned slot;
+   boolean clamp_vertex_color = shader->draw->rasterizer->clamp_vertex_color;
 
    tgsi_exec_set_constant_buffers(machine, PIPE_MAX_CONSTANT_BUFFERS,
                                   constants, const_size);
@@ -151,11 +152,22 @@ vs_exec_run_linear( struct draw_vertex_shader *shader,
        */
       for (j = 0; j < max_vertices; j++) {
          for (slot = 0; slot < shader->info.num_outputs; slot++) {
-            output[slot][0] = machine->Outputs[slot].xyzw[0].f[j];
-            output[slot][1] = machine->Outputs[slot].xyzw[1].f[j];
-            output[slot][2] = machine->Outputs[slot].xyzw[2].f[j];
-            output[slot][3] = machine->Outputs[slot].xyzw[3].f[j];
-
+            unsigned name = shader->info.output_semantic_name[slot];
+            if(clamp_vertex_color &&
+                  (name == TGSI_SEMANTIC_COLOR || name == TGSI_SEMANTIC_BCOLOR))
+            {
+               output[slot][0] = CLAMP(machine->Outputs[slot].xyzw[0].f[j], 0.0f, 1.0f);
+               output[slot][1] = CLAMP(machine->Outputs[slot].xyzw[1].f[j], 0.0f, 1.0f);
+               output[slot][2] = CLAMP(machine->Outputs[slot].xyzw[2].f[j], 0.0f, 1.0f);
+               output[slot][3] = CLAMP(machine->Outputs[slot].xyzw[3].f[j], 0.0f, 1.0f);
+            }
+            else
+            {
+               output[slot][0] = machine->Outputs[slot].xyzw[0].f[j];
+               output[slot][1] = machine->Outputs[slot].xyzw[1].f[j];
+               output[slot][2] = machine->Outputs[slot].xyzw[2].f[j];
+               output[slot][3] = machine->Outputs[slot].xyzw[3].f[j];
+            }
          }
 
 #if 0