1 files changed, 101 insertions, 84 deletions

diff --git a/src/mesa/drivers/dri/r600/r700_vertprog.c b/src/mesa/drivers/dri/r600/r700_vertprog.c
index d12c39c9f7..782f151f5a 100644
--- a/src/mesa/drivers/dri/r600/r700_vertprog.c
+++ b/src/mesa/drivers/dri/r600/r700_vertprog.c
@@ -111,6 +111,15 @@ unsigned int Map_Vertex_Output(r700_AssemblerBase       *pAsm,
 		}
 	}
 
+    for(i=VERT_RESULT_VAR0; i<VERT_RESULT_MAX; i++)
+	{
+		unBit = 1 << i;
+		if(mesa_vp->Base.OutputsWritten & unBit)
+		{
+			pAsm->ucVP_OutputMap[i] = unTotal++;
+		}
+	}
+
 	return (unTotal - unStart);
 }
 
@@ -179,7 +188,8 @@ GLboolean Process_Vertex_Program_Vfetch_Instructions2(
                                       context->stream_desc[i].size,
                                       context->stream_desc[i].element,
                                       context->stream_desc[i]._signed,
-                                      context->stream_desc[i].normalize,						            
+                                      context->stream_desc[i].normalize,
+                                      context->stream_desc[i].format,
                                      &vtxFetchMethod);
     }
 
@@ -203,22 +213,11 @@ void Map_Vertex_Program(GLcontext *ctx,
 	pAsm->number_used_registers += num_inputs;
 
 	// Create VFETCH instructions for inputs
-	if(1 == vp->uiVersion) 
-    {
-	    if (GL_TRUE != Process_Vertex_Program_Vfetch_Instructions(vp, mesa_vp) ) 
-	    {
-		    radeon_error("Calling Process_Vertex_Program_Vfetch_Instructions return error. \n");
-		    return; 
-	    }
-    }
-    else
-    {
-        if (GL_TRUE != Process_Vertex_Program_Vfetch_Instructions2(ctx, vp, mesa_vp) ) 
-	    {
-		    radeon_error("Calling Process_Vertex_Program_Vfetch_Instructions2 return error. \n");
-		    return; 
-	    }
-    }
+        if (GL_TRUE != Process_Vertex_Program_Vfetch_Instructions2(ctx, vp, mesa_vp) )
+	{
+		radeon_error("Calling Process_Vertex_Program_Vfetch_Instructions2 return error. \n");
+		return;
+	}
 
 	// Map Outputs
 	pAsm->number_of_exports = Map_Vertex_Output(pAsm, mesa_vp, pAsm->number_used_registers);
@@ -228,7 +227,7 @@ void Map_Vertex_Program(GLcontext *ctx,
 	pAsm->number_used_registers += pAsm->number_of_exports;
 
     pAsm->pucOutMask = (unsigned char*) MALLOC(pAsm->number_of_exports);
-    
+
     for(ui=0; ui<pAsm->number_of_exports; ui++)
     {
         pAsm->pucOutMask[ui] = 0x0;
@@ -245,7 +244,9 @@ void Map_Vertex_Program(GLcontext *ctx,
     {   /* fix func t_vp uses NumTemporaries */
         pAsm->number_used_registers += mesa_vp->Base.NumTemporaries;
     }
-	
+
+    pAsm->flag_reg_index = pAsm->number_used_registers++;
+
     pAsm->uFirstHelpReg = pAsm->number_used_registers;
 }
 
@@ -300,18 +301,13 @@ GLboolean Find_Instruction_Dependencies_vp(struct r700_vertex_program *vp,
 }
 
 struct r700_vertex_program* r700TranslateVertexShader(GLcontext *ctx,
-						struct gl_vertex_program *mesa_vp,
-                        GLint nVer)
+						      struct gl_vertex_program *mesa_vp)
 {
 	context_t *context = R700_CONTEXT(ctx);
 	struct r700_vertex_program *vp;
-	TNLcontext *tnl = TNL_CONTEXT(ctx);
-	struct vertex_buffer *vb = &tnl->vb;
-	unsigned int unBit;
 	unsigned int i;
 
 	vp = _mesa_calloc(sizeof(*vp));
-    vp->uiVersion = nVer;
 	vp->mesa_program = (struct gl_vertex_program *)_mesa_clone_program(ctx, &mesa_vp->Base);
 
 	if (mesa_vp->IsPositionInvariant)
@@ -319,29 +315,14 @@ struct r700_vertex_program* r700TranslateVertexShader(GLcontext *ctx,
                 _mesa_insert_mvp_code(ctx, vp->mesa_program);
         }
 
-	if( 1 == nVer )
+	for(i=0; i<context->nNumActiveAos; i++)
 	{
-	    for(i=0; i<VERT_ATTRIB_MAX; i++)
-	    {
-		unBit = 1 << i;
-		if(vp->mesa_program->Base.InputsRead & unBit) /* ctx->Array.ArrayObj->xxxxxxx */
-		{
-			vp->aos_desc[i].size   = vb->AttribPtr[i]->size;
-			vp->aos_desc[i].stride = vb->AttribPtr[i]->size * sizeof(GL_FLOAT);/* when emit array, data is packed. vb->AttribPtr[i]->stride;*/
-			vp->aos_desc[i].type   = GL_FLOAT;
-		}
-	    }
+		vp->aos_desc[i].size   = context->stream_desc[i].size;
+		vp->aos_desc[i].stride = context->stream_desc[i].stride;
+		vp->aos_desc[i].type   = context->stream_desc[i].type;
+		vp->aos_desc[i].format = context->stream_desc[i].format;
 	}
-	else
-	{
-		for(i=0; i<context->nNumActiveAos; i++)
-		{
-			vp->aos_desc[i].size   = context->stream_desc[i].size;
-			vp->aos_desc[i].stride = context->stream_desc[i].stride;
-			vp->aos_desc[i].type   = context->stream_desc[i].type;
-		}
-	}	
-	
+
 	if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)
 	{
 		vp->r700AsmCode.bR6xx = 1;
@@ -354,20 +335,36 @@ struct r700_vertex_program* r700TranslateVertexShader(GLcontext *ctx,
 	if(GL_FALSE == Find_Instruction_Dependencies_vp(vp, vp->mesa_program))
 	{
 		return NULL;
+	}
+
+    InitShaderProgram(&(vp->r700AsmCode));
+
+    for(i=0; i < MAX_SAMPLERS; i++)
+    {
+        vp->r700AsmCode.SamplerUnits[i] = vp->mesa_program->Base.SamplerUnits[i];
     }
 
-	if(GL_FALSE == AssembleInstr(vp->mesa_program->Base.NumInstructions,
-                                 &(vp->mesa_program->Base.Instructions[0]), 
+    vp->r700AsmCode.unCurNumILInsts = vp->mesa_program->Base.NumInstructions;
+
+	if(GL_FALSE == AssembleInstr(0,
+                                 0,
+                                 vp->mesa_program->Base.NumInstructions,
+                                 &(vp->mesa_program->Base.Instructions[0]),
                                  &(vp->r700AsmCode)) )
 	{
 		return NULL;
-	} 
+	}
 
     if(GL_FALSE == Process_Vertex_Exports(&(vp->r700AsmCode), vp->mesa_program->Base.OutputsWritten) )
     {
         return NULL;
     }
 
+    if( GL_FALSE == RelocProgram(&(vp->r700AsmCode), &(vp->mesa_program->Base)) )
+    {
+        return GL_FALSE;
+    }
+
     vp->r700Shader.nRegs = (vp->r700AsmCode.number_used_registers == 0) ? 0 
                          : (vp->r700AsmCode.number_used_registers - 1);
 
@@ -378,14 +375,11 @@ struct r700_vertex_program* r700TranslateVertexShader(GLcontext *ctx,
 	return vp;
 }
 
-void r700SelectVertexShader(GLcontext *ctx, GLint nVersion)
+void r700SelectVertexShader(GLcontext *ctx)
 {
     context_t *context = R700_CONTEXT(ctx);
     struct r700_vertex_program_cont *vpc;
     struct r700_vertex_program *vp;
-    TNLcontext *tnl = TNL_CONTEXT(ctx);
-    struct vertex_buffer *vb = &tnl->vb;
-    unsigned int unBit;
     unsigned int i;
     GLboolean match;
     GLbitfield InputsRead;
@@ -396,47 +390,28 @@ void r700SelectVertexShader(GLcontext *ctx, GLint nVersion)
     if (vpc->mesa_program.IsPositionInvariant)
     {
 	InputsRead |= VERT_BIT_POS;
-    } 
-    
+    }
+
     for (vp = vpc->progs; vp; vp = vp->next)
     {
-	if (vp->uiVersion != nVersion ) 
-	    continue;
-	match = GL_TRUE;	
-	if ( 1 == nVersion ) 
+	match = GL_TRUE;
+	for(i=0; i<context->nNumActiveAos; i++)
 	{
-	    for(i=0; i<VERT_ATTRIB_MAX; i++)
-	    {
-		unBit = 1 << i;
-		if(InputsRead & unBit)
+		if (vp->aos_desc[i].size != context->stream_desc[i].size ||
+		    vp->aos_desc[i].format != context->stream_desc[i].format)
 		{
-		    if (vp->aos_desc[i].size != vb->AttribPtr[i]->size)
-		    {
 			match = GL_FALSE;
 			break;
-		    }
 		}
-	    }
 	}
-	else
-	{
-	    for(i=0; i<context->nNumActiveAos; i++)
-	    {
-		if (vp->aos_desc[i].size != context->stream_desc[i].size)
-		{
-		    match = GL_FALSE;
-		    break;
-		}
-	    }
-	}	
-	if (match) 
+	if (match)
 	{
 		context->selected_vp = vp;
 		return;
 	}
     }
 
-    vp = r700TranslateVertexShader(ctx, &(vpc->mesa_program), nVersion);
+    vp = r700TranslateVertexShader(ctx, &(vpc->mesa_program));
     if(!vp)
     {
 	radeon_error("Failed to translate vertex shader. \n");
@@ -526,6 +501,7 @@ static void r700TranslateAttrib(GLcontext *ctx, GLuint unLoc, int count, const s
 	pStreamDesc->size = input->Size;
 	pStreamDesc->dst_loc = context->nNumActiveAos;
 	pStreamDesc->element = unLoc;
+	pStreamDesc->format = input->Format;
 
 	switch (pStreamDesc->type) 
 	{ //GetSurfaceFormat
@@ -570,6 +546,11 @@ void r700SetVertexFormat(GLcontext *ctx, const struct gl_client_array *arrays[],
     unsigned int unBit = mesa_vp->Base.InputsRead;
     context->nNumActiveAos = 0;
 
+    if (mesa_vp->IsPositionInvariant)
+    {
+        unBit |= VERT_BIT_POS;
+    }
+
     while(unBit) 
     {
         if(unBit & 1)
@@ -580,6 +561,7 @@ void r700SetVertexFormat(GLcontext *ctx, const struct gl_client_array *arrays[],
         unBit >>= 1;
         ++unLoc;
     }
+    context->radeon.tcl.aos_count = context->nNumActiveAos;
 }
 
 void * r700GetActiveVpShaderBo(GLcontext * ctx)
@@ -661,6 +643,12 @@ GLboolean r700SetupVertexProgram(GLcontext * ctx)
     paramList = vp->mesa_program->Base.Parameters;
 
     if(NULL != paramList) {
+        /* vp->mesa_program was cloned, not updated by glsl shader api. */
+        /* _mesa_reference_program has already checked glsl shProg is ok and set ctx->VertexProgem._Current */
+        /* so, use ctx->VertexProgem._Current */       
+        struct gl_program_parameter_list *paramListOrginal = 
+                         paramListOrginal = ctx->VertexProgram._Current->Base.Parameters;
+         
 	    _mesa_load_state_parameters(ctx, paramList);
 
 	    if (paramList->NumParameters > R700_MAX_DX9_CONSTS)
@@ -673,13 +661,42 @@ GLboolean r700SetupVertexProgram(GLcontext * ctx)
 	    unNumParamData = paramList->NumParameters;
 
 	    for(ui=0; ui<unNumParamData; ui++) {
-		    r700->vs.consts[ui][0].f32All = paramList->ParameterValues[ui][0];
-		    r700->vs.consts[ui][1].f32All = paramList->ParameterValues[ui][1];
-		    r700->vs.consts[ui][2].f32All = paramList->ParameterValues[ui][2];
-		    r700->vs.consts[ui][3].f32All = paramList->ParameterValues[ui][3];
+            if(paramList->Parameters[ui].Type == PROGRAM_UNIFORM) 
+            {
+                r700->vs.consts[ui][0].f32All = paramListOrginal->ParameterValues[ui][0];
+		        r700->vs.consts[ui][1].f32All = paramListOrginal->ParameterValues[ui][1];
+		        r700->vs.consts[ui][2].f32All = paramListOrginal->ParameterValues[ui][2];
+		        r700->vs.consts[ui][3].f32All = paramListOrginal->ParameterValues[ui][3];
+            }
+            else
+            {
+		        r700->vs.consts[ui][0].f32All = paramList->ParameterValues[ui][0];
+		        r700->vs.consts[ui][1].f32All = paramList->ParameterValues[ui][1];
+		        r700->vs.consts[ui][2].f32All = paramList->ParameterValues[ui][2];
+		        r700->vs.consts[ui][3].f32All = paramList->ParameterValues[ui][3];
+            }
 	    }
     } else
 	    r700->vs.num_consts = 0;
 
+    COMPILED_SUB * pCompiledSub;
+    GLuint uj;
+    GLuint unConstOffset = r700->vs.num_consts;
+    for(ui=0; ui<vp->r700AsmCode.unNumPresub; ui++)
+    {
+        pCompiledSub = vp->r700AsmCode.presubs[ui].pCompiledSub;
+
+        r700->vs.num_consts += pCompiledSub->NumParameters;
+
+        for(uj=0; uj<pCompiledSub->NumParameters; uj++)
+        {
+            r700->vs.consts[uj + unConstOffset][0].f32All = pCompiledSub->ParameterValues[uj][0];
+		    r700->vs.consts[uj + unConstOffset][1].f32All = pCompiledSub->ParameterValues[uj][1];
+		    r700->vs.consts[uj + unConstOffset][2].f32All = pCompiledSub->ParameterValues[uj][2];
+		    r700->vs.consts[uj + unConstOffset][3].f32All = pCompiledSub->ParameterValues[uj][3];
+        }
+        unConstOffset += pCompiledSub->NumParameters;
+    }
+
     return GL_TRUE;
 }