Don't always declare frag shader INPUT[0] as fragment position.

We were doing this for the sake of softpipe and the tgsi intergrepter since
we always need the fragment position and W-coordinate information in order
to compute fragment interpolants.
But that's not appropriate for hardware drivers.
The tgsi interpreter now get x,y,w information from a separate tgsi_exec_vector
variable setup by softpipe.
The new pipe_shader_state->input_map[] defines how vert shader outputs map
to frag shader inputs.  It may go away though, since one can also examine
the semantic label on frag shader input[0] to figure things out.

3 files changed, 45 insertions, 51 deletions

diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c
index 4ec10badad..33372b0f39 100644
--- a/src/mesa/state_tracker/st_atom_shader.c
+++ b/src/mesa/state_tracker/st_atom_shader.c
@@ -151,8 +151,7 @@ find_translated_vp(struct st_context *st,
 {
    static const GLuint UNUSED = ~0;
    struct translated_vertex_program *xvp;
-   const GLbitfield fragInputsRead
-      = stfp->Base.Base.InputsRead | FRAG_BIT_WPOS;
+   const GLbitfield fragInputsRead = stfp->Base.Base.InputsRead;
 
    /*
     * Translate fragment program if needed.
@@ -206,6 +205,7 @@ find_translated_vp(struct st_context *st,
    if (xvp->serialNo != stvp->serialNo) {
       GLuint outAttr, dummySlot;
       const GLbitfield outputsWritten = stvp->Base.Base.OutputsWritten;
+      GLuint numVpOuts = 0;
 
       /* Compute mapping of vertex program outputs to slots, which depends
        * on the fragment program's input->slot mapping.
@@ -214,11 +214,24 @@ find_translated_vp(struct st_context *st,
          /* set default: */
          xvp->output_to_slot[outAttr] = UNUSED;
 
-         if (outputsWritten & (1 << outAttr)) {
+         if (outAttr == VERT_RESULT_HPOS) {
+            /* always put xformed position into slot zero */
+            xvp->output_to_slot[VERT_RESULT_HPOS] = 0;
+            numVpOuts++;
+         }
+         else if (outputsWritten & (1 << outAttr)) {
             /* see if the frag prog wants this vert output */
-            GLint fpIn = vp_out_to_fp_in(outAttr);
-            if (fpIn >= 0) {
-               xvp->output_to_slot[outAttr] = stfp->input_to_slot[fpIn];
+            GLint fpInAttrib = vp_out_to_fp_in(outAttr);
+            if (fpInAttrib >= 0) {
+               GLuint fpInSlot = stfp->input_to_slot[fpInAttrib];
+               GLuint vpOutSlot = stfp->fs->state.input_map[fpInSlot];
+               xvp->output_to_slot[outAttr] = vpOutSlot;
+               numVpOuts++;
+            }
+            else if (outAttr == VERT_RESULT_BFC0 ||
+                     outAttr == VERT_RESULT_BFC1) {
+               /* backface colors go into last slots */
+               xvp->output_to_slot[outAttr] = numVpOuts++;
             }
          }
       }
diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c
index b392edf16d..27dab5b9c0 100644
--- a/src/mesa/state_tracker/st_mesa_to_tgsi.c
+++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c
@@ -675,44 +675,22 @@ tgsi_translate_mesa_program(
    if (procType == TGSI_PROCESSOR_FRAGMENT) {
       for (i = 0; i < numInputs; i++) {
          struct tgsi_full_declaration fulldecl;
-         switch (inputSemanticName[i]) {
-         case TGSI_SEMANTIC_POSITION:
-            /* Fragment XY pos */
-            fulldecl = make_input_decl(i,
-                                       GL_TRUE, TGSI_INTERPOLATE_CONSTANT,
-                                       TGSI_WRITEMASK_XY,
-                                       GL_TRUE, TGSI_SEMANTIC_POSITION, 0 );
-            ti += tgsi_build_full_declaration(
-                                              &fulldecl,
-                                              &tokens[ti],
-                                              header,
-                                              maxTokens - ti );
-            /* Fragment ZW pos */
-            fulldecl = make_input_decl(i,
-                                       GL_TRUE, TGSI_INTERPOLATE_LINEAR,
-                                       TGSI_WRITEMASK_ZW,
-                                       GL_TRUE, TGSI_SEMANTIC_POSITION, 0 );
-            ti += tgsi_build_full_declaration(&fulldecl,
-                                              &tokens[ti],
-                                              header,
-                                              maxTokens - ti );
-            break;
-         default:
-            fulldecl = make_input_decl(i,
-                                       GL_TRUE, interpMode[i],
-                                       TGSI_WRITEMASK_XYZW,
-                                       GL_TRUE, inputSemanticName[i],
-                                       inputSemanticIndex[i]);
-            ti += tgsi_build_full_declaration(&fulldecl,
-                                              &tokens[ti],
-                                              header,
-                                              maxTokens - ti );
-            break;
-         }
+         fulldecl = make_input_decl(i,
+                                    GL_TRUE, interpMode[i],
+                                    TGSI_WRITEMASK_XYZW,
+                                    GL_TRUE, inputSemanticName[i],
+                                    inputSemanticIndex[i]);
+         ti += tgsi_build_full_declaration(&fulldecl,
+                                           &tokens[ti],
+                                           header,
+                                           maxTokens - ti );
       }
    }
    else {
       /* vertex prog */
+      /* XXX: this could probaby be merged with the clause above.
+       * the only difference is the semantic tags.
+       */
       for (i = 0; i < numInputs; i++) {
          struct tgsi_full_declaration fulldecl;
          fulldecl = make_input_decl(i,
diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c
index e64bf14d56..fe22233c93 100644
--- a/src/mesa/state_tracker/st_program.c
+++ b/src/mesa/state_tracker/st_program.c
@@ -47,7 +47,7 @@
 #include "st_mesa_to_tgsi.h"
 
 
-#define TGSI_DEBUG 0
+#define TGSI_DEBUG 01
 
 
 /**
@@ -283,16 +283,17 @@ st_translate_fragment_program(struct st_context *st,
    const struct cso_fragment_shader *cso;
    GLuint interpMode[16];  /* XXX size? */
    GLuint attr;
-   GLbitfield inputsRead = stfp->Base.Base.InputsRead;
-
-   /* For software rendering, we always need the fragment input position
-    * in order to calculate interpolated values.
-    * For i915, we always want to emit the semantic info for position.
-    */
-   inputsRead |= FRAG_BIT_WPOS;
+   const GLbitfield inputsRead = stfp->Base.Base.InputsRead;
+   GLuint vslot = 0;
 
    memset(&fs, 0, sizeof(fs));
 
+   /* which vertex output goes to the first fragment input: */
+   if (inputsRead & FRAG_BIT_WPOS)
+      vslot = 0;
+   else
+      vslot = 1;
+
    /*
     * Convert Mesa program inputs to TGSI input register semantics.
     */
@@ -300,15 +301,17 @@ st_translate_fragment_program(struct st_context *st,
       if (inputsRead & (1 << attr)) {
          const GLuint slot = fs.num_inputs;
 
-         fs.num_inputs++;
-
          defaultInputMapping[attr] = slot;
 
+         fs.input_map[slot] = vslot++;
+
+         fs.num_inputs++;
+
          switch (attr) {
          case FRAG_ATTRIB_WPOS:
             fs.input_semantic_name[slot] = TGSI_SEMANTIC_POSITION;
             fs.input_semantic_index[slot] = 0;
-            interpMode[slot] = TGSI_INTERPOLATE_CONSTANT;
+            interpMode[slot] = TGSI_INTERPOLATE_LINEAR;
             break;
          case FRAG_ATTRIB_COL0:
             fs.input_semantic_name[slot] = TGSI_SEMANTIC_COLOR;