Merge branch 'master' of ssh://git.freedesktop.org/git/mesa/mesa into r600_state_predict

8 files changed, 149 insertions, 122 deletions

diff --git a/src/mesa/drivers/dri/r600/r600_context.c b/src/mesa/drivers/dri/r600/r600_context.c
index 0b0c4f5049..6a066f3510 100644
--- a/src/mesa/drivers/dri/r600/r600_context.c
+++ b/src/mesa/drivers/dri/r600/r600_context.c
@@ -185,13 +185,7 @@ static void r600_vtbl_emit_cs_header(struct radeon_cs *cs, radeonContextPtr rmes
 
 static void r600_vtbl_pre_emit_atoms(radeonContextPtr radeon)
 {
-	context_t *context = (context_t *)radeon;
-
-	/* always emit CB base to prevent
-	 * lock ups on some chips.
-	 */
-	R600_STATECHANGE(context, cb_target);
-	r700Start3D(context);
+	r700Start3D((context_t *)radeon);
 }
 
 static void r600_fallback(GLcontext *ctx, GLuint bit, GLboolean mode)
diff --git a/src/mesa/drivers/dri/r600/r700_chip.c b/src/mesa/drivers/dri/r600/r700_chip.c
index 3f11cf2c98..16f1a3df36 100644
--- a/src/mesa/drivers/dri/r600/r700_chip.c
+++ b/src/mesa/drivers/dri/r600/r700_chip.c
@@ -230,6 +230,9 @@ static void r700SendVTXState(GLcontext *ctx, struct radeon_state_atom *atom)
     unsigned int i, j = 0;
     BATCH_LOCALS(&context->radeon);
 
+    if (context->radeon.tcl.aos_count == 0)
+	    return;
+
     BEGIN_BATCH_NO_AUTOSTATE(6);
     R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, 1));
     R600_OUT_BATCH(mmSQ_VTX_BASE_VTX_LOC - ASIC_CTL_CONST_BASE_INDEX);
@@ -989,11 +992,60 @@ static int check_always(GLcontext *ctx, struct radeon_state_atom *atom)
 	return atom->cmd_size;
 }
 
+static int check_cb(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	int count = 7;
+
+	if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)
+		count += 11;
+
+	return count;
+}
+
+static int check_blnd(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	unsigned int ui;
+	int count = 3;
+
+	if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)
+		count += 3;
+
+	if (context->radeon.radeonScreen->chip_family > CHIP_FAMILY_R600) {
+		for (ui = 0; ui < R700_MAX_RENDER_TARGETS; ui++) {
+                        if (r700->render_target[ui].enabled)
+				count += 3;
+		}
+	}
+
+	return count;
+}
+
+static int check_ucp(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	int i;
+	int count = 0;
+
+	for (i = 0; i < R700_MAX_UCP; i++) {
+		if (r700->ucp[i].enabled)
+			count += 6;
+	}
+	return count;
+}
+
 static int check_vtx(GLcontext *ctx, struct radeon_state_atom *atom)
 {
 	context_t *context = R700_CONTEXT(ctx);
+	int count = context->radeon.tcl.aos_count * 18;
+
+	if (count)
+		count += 6;
 
-	return context->radeon.tcl.aos_count * 18;
+	return count;
 }
 
 static int check_tx(GLcontext *ctx, struct radeon_state_atom *atom)
@@ -1014,16 +1066,24 @@ static int check_ps_consts(GLcontext *ctx, struct radeon_state_atom *atom)
 {
 	context_t *context = R700_CONTEXT(ctx);
 	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	int count = r700->ps.num_consts * 4;
 
-	return 2 + (r700->ps.num_consts * 4);
+	if (count)
+		count += 2;
+
+	return count;
 }
 
 static int check_vs_consts(GLcontext *ctx, struct radeon_state_atom *atom)
 {
 	context_t *context = R700_CONTEXT(ctx);
 	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+	int count = r700->vs.num_consts * 4;
+
+	if (count)
+		count += 2;
 
-	return 2 + (r700->vs.num_consts * 4);
+	return count;
 }
 
 #define ALLOC_STATE( ATOM, CHK, SZ, EMIT )				\
@@ -1056,12 +1116,12 @@ void r600InitAtoms(context_t *context)
 	ALLOC_STATE(aa, always, 12, r700SendAAState);
 	ALLOC_STATE(cl, always, 12, r700SendCLState);
 	ALLOC_STATE(gb, always, 6, r700SendGBState);
-	ALLOC_STATE(ucp, always, 36, r700SendUCPState);
+	ALLOC_STATE(ucp, ucp, (R700_MAX_UCP * 6), r700SendUCPState);
 	ALLOC_STATE(su, always, 9, r700SendSUState);
 	ALLOC_STATE(poly, always, 10, r700SendPolyState);
-	ALLOC_STATE(cb, always, 18, r700SendCBState);
+	ALLOC_STATE(cb, cb, 18, r700SendCBState);
 	ALLOC_STATE(clrcmp, always, 6, r700SendCBCLRCMPState);
-	ALLOC_STATE(blnd, always, 30, r700SendCBBlendState);
+	ALLOC_STATE(blnd, blnd, (6 + (R700_MAX_RENDER_TARGETS * 3)), r700SendCBBlendState);
 	ALLOC_STATE(blnd_clr, always, 6, r700SendCBBlendColorState);
 	ALLOC_STATE(cb_target, always, 25, r700SendRenderTargetState);
 	ALLOC_STATE(sx, always, 9, r700SendSXState);
@@ -1073,7 +1133,7 @@ void r600InitAtoms(context_t *context)
 	ALLOC_STATE(ps, always, 21, r700SendPSState);
 	ALLOC_STATE(vs_consts, vs_consts, (2 + (R700_MAX_DX9_CONSTS * 4)), r700SendVSConsts);
 	ALLOC_STATE(ps_consts, ps_consts, (2 + (R700_MAX_DX9_CONSTS * 4)), r700SendPSConsts);
-	ALLOC_STATE(vtx, vtx, (VERT_ATTRIB_MAX * 18), r700SendVTXState);
+	ALLOC_STATE(vtx, vtx, (6 + (VERT_ATTRIB_MAX * 18)), r700SendVTXState);
 	ALLOC_STATE(tx, tx, (R700_TEXTURE_NUMBERUNITS * 20), r700SendTexState);
 	ALLOC_STATE(tx_smplr, tx, (R700_TEXTURE_NUMBERUNITS * 5), r700SendTexSamplerState);
 	ALLOC_STATE(tx_brdr_clr, tx, (R700_TEXTURE_NUMBERUNITS * 6), r700SendTexBorderColorState);
diff --git a/src/mesa/drivers/dri/r600/r700_fragprog.c b/src/mesa/drivers/dri/r600/r700_fragprog.c
index 098b420dfc..e4a6d4cedf 100644
--- a/src/mesa/drivers/dri/r600/r700_fragprog.c
+++ b/src/mesa/drivers/dri/r600/r700_fragprog.c
@@ -121,13 +121,13 @@ void Map_Fragment_Program(r700_AssemblerBase         *pAsm,
 		pAsm->pR700Shader->depthIsExported = 1;
 	}
 
-    pAsm->pucOutMask = (unsigned char*) MALLOC(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;
     }
-	
-	pAsm->uFirstHelpReg = pAsm->number_used_registers;
+
+    pAsm->uFirstHelpReg = pAsm->number_used_registers;
 }
 
 GLboolean Find_Instruction_Dependencies_fp(struct r700_fragment_program *fp,
@@ -258,6 +258,19 @@ GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp,
 	return GL_TRUE;
 }
 
+void r700SelectFragmentShader(GLcontext *ctx)
+{
+    context_t *context = R700_CONTEXT(ctx);
+    struct r700_fragment_program *fp = (struct r700_fragment_program *)
+	    (ctx->FragmentProgram._Current);
+    if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)
+    {
+	    fp->r700AsmCode.bR6xx = 1;
+    }
+
+    r700TranslateFragmentShader(fp, &(fp->mesa_program));
+}
+
 void * r700GetActiveFpShaderBo(GLcontext * ctx)
 {
     struct r700_fragment_program *fp = (struct r700_fragment_program *)
@@ -283,7 +296,7 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx)
 
     if(GL_FALSE == fp->loaded)
     {
-        if(fp->r700Shader.bNeedsAssembly == GL_TRUE)
+	    if(fp->r700Shader.bNeedsAssembly == GL_TRUE)
 	    {
 		    Assemble( &(fp->r700Shader) );
 	    }
diff --git a/src/mesa/drivers/dri/r600/r700_fragprog.h b/src/mesa/drivers/dri/r600/r700_fragprog.h
index 9c7813e908..cbb108d212 100644
--- a/src/mesa/drivers/dri/r600/r700_fragprog.h
+++ b/src/mesa/drivers/dri/r600/r700_fragprog.h
@@ -49,13 +49,16 @@ struct r700_fragment_program
 
 /* Internal */
 void Map_Fragment_Program(r700_AssemblerBase         *pAsm,
-						  struct gl_fragment_program *mesa_fp);
+			  struct gl_fragment_program *mesa_fp);
 GLboolean Find_Instruction_Dependencies_fp(struct r700_fragment_program *fp,
-					                	   struct gl_fragment_program   *mesa_fp);
+					   struct gl_fragment_program   *mesa_fp);
+
+GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp,
+				      struct gl_fragment_program   *mesa_vp);
 
 /* Interface */
-extern GLboolean r700TranslateFragmentShader(struct r700_fragment_program *fp,
-							                 struct gl_fragment_program   *mesa_vp);
+extern void r700SelectFragmentShader(GLcontext *ctx);
+
 extern GLboolean r700SetupFragmentProgram(GLcontext * ctx);
 
 extern void *    r700GetActiveFpShaderBo(GLcontext * ctx);
diff --git a/src/mesa/drivers/dri/r600/r700_render.c b/src/mesa/drivers/dri/r600/r700_render.c
index 9b0b5443d6..d64e921bda 100644
--- a/src/mesa/drivers/dri/r600/r700_render.c
+++ b/src/mesa/drivers/dri/r600/r700_render.c
@@ -347,6 +347,13 @@ static GLboolean r700RunRender(GLcontext * ctx,
         fprintf(stderr, "%s: cs begin at %d\n",
                 __func__, context->radeon.cmdbuf.cs->cdw);
 
+    /* always emit CB base to prevent
+     * lock ups on some chips.
+     */
+    R600_STATECHANGE(context, cb_target);
+    /* mark vtx as dirty since it changes per-draw */
+    R600_STATECHANGE(context, vtx);
+
     r700UpdateShaders(ctx);
     r700SetScissor(context);
     r700SetupVertexProgram(ctx);
diff --git a/src/mesa/drivers/dri/r600/r700_state.c b/src/mesa/drivers/dri/r600/r700_state.c
index 15f40b2771..efa1daf2d3 100644
--- a/src/mesa/drivers/dri/r600/r700_state.c
+++ b/src/mesa/drivers/dri/r600/r700_state.c
@@ -71,65 +71,37 @@ void r700SetDefaultStates(context_t *context) //--------------------
 void r700UpdateShaders (GLcontext * ctx)  //----------------------------------
 {
     context_t *context = R700_CONTEXT(ctx);
-
     GLvector4f dummy_attrib[_TNL_ATTRIB_MAX];
     GLvector4f *temp_attrib[_TNL_ATTRIB_MAX];
     int i;
 
-    if (ctx->FragmentProgram._Current) {
-	    struct r700_fragment_program *fp = (struct r700_fragment_program *)
-		    (ctx->FragmentProgram._Current);
-	    if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)
-	    {
-		    fp->r700AsmCode.bR6xx = 1;
-	    }
-
-	    if(GL_FALSE == fp->translated)
-	    {
-		    if( GL_FALSE == r700TranslateFragmentShader(fp, &(fp->mesa_program)) )
-		    {
-			    //return GL_TRUE;
-		    }
-	    }
+    /* should only happenen once, just after context is created */
+    /* TODO: shouldn't we fallback to sw here? */
+    if (!ctx->FragmentProgram._Current) {
+	    _mesa_fprintf(stderr, "No ctx->FragmentProgram._Current!!\n");
+	    return;
     }
 
-    if (context->radeon.NewGLState) 
-    {
-	struct r700_vertex_program *vp;
-        context->radeon.NewGLState = 0;
-
-        for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) 
-        {
-            /* mat states from state var not array for sw */
-            dummy_attrib[i].stride = 0;
+    r700SelectFragmentShader(ctx);
 
-            temp_attrib[i] = TNL_CONTEXT(ctx)->vb.AttribPtr[i];
-            TNL_CONTEXT(ctx)->vb.AttribPtr[i] = &(dummy_attrib[i]);
-        }
-
-        _tnl_UpdateFixedFunctionProgram(ctx);
-
-        for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) 
-        {
-            TNL_CONTEXT(ctx)->vb.AttribPtr[i] = temp_attrib[i];
-        }
+    if (context->radeon.NewGLState) {
+	    for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
+		    /* mat states from state var not array for sw */
+		    dummy_attrib[i].stride = 0;
+	            temp_attrib[i] = TNL_CONTEXT(ctx)->vb.AttribPtr[i];
+		    TNL_CONTEXT(ctx)->vb.AttribPtr[i] = &(dummy_attrib[i]);
+	    }
 
-        r700SelectVertexShader(ctx);
-        vp = (struct r700_vertex_program *)ctx->VertexProgram._Current;
+	    _tnl_UpdateFixedFunctionProgram(ctx);
 
-        if (vp->translated == GL_FALSE) 
-        {
-            // TODO
-            //fprintf(stderr, "Failing back to sw-tcl\n");
-            //hw_tcl_on = future_hw_tcl_on = 0;
-            //r300ResetHwState(rmesa);
-            //
-            r700UpdateStateParameters(ctx, _NEW_PROGRAM);
-            return;
-        }
+	    for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
+		    TNL_CONTEXT(ctx)->vb.AttribPtr[i] = temp_attrib[i];
+	    }
     }
 
-    r700UpdateStateParameters(ctx, _NEW_PROGRAM);
+    r700SelectVertexShader(ctx);
+    r700UpdateStateParameters(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+    context->radeon.NewGLState = 0;
 }
 
 /*
@@ -176,45 +148,25 @@ void r700UpdateDrawBuffer(GLcontext * ctx) /* TODO */ //---------------------
 	r700SetDepthTarget(context);
 }
 
-static void r700FetchStateParameter(GLcontext * ctx,
-			                        const gl_state_index state[STATE_LENGTH],
-			                        GLfloat * value)
-{
-    /* TODO */
-}
-
 void r700UpdateStateParameters(GLcontext * ctx, GLuint new_state) //--------------------
 {
-	struct r700_fragment_program *fp;
+	struct r700_fragment_program *fp =
+		(struct r700_fragment_program *)ctx->FragmentProgram._Current;
 	struct gl_program_parameter_list *paramList;
-	GLuint i;
 
-	if (!(new_state & (_NEW_BUFFERS | _NEW_PROGRAM)))
+	if (!(new_state & (_NEW_BUFFERS | _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS)))
 		return;
 
-	fp = (struct r700_fragment_program *)ctx->FragmentProgram._Current;
-	if (!fp)
-    {
+	if (!ctx->FragmentProgram._Current || !fp)
 		return;
-    }
 
-	paramList = fp->mesa_program.Base.Parameters;
+	paramList = ctx->FragmentProgram._Current->Base.Parameters;
 
 	if (!paramList)
-    {
 		return;
-    }
 
-	for (i = 0; i < paramList->NumParameters; i++) 
-    {
-		if (paramList->Parameters[i].Type == PROGRAM_STATE_VAR) 
-        {
-			r700FetchStateParameter(ctx,
-						paramList->Parameters[i].
-						StateIndexes,
-						paramList->ParameterValues[i]);
-		}
-	}
+	_mesa_load_state_parameters(ctx, paramList);
+
 }
 
 /**
diff --git a/src/mesa/drivers/dri/r600/r700_vertprog.c b/src/mesa/drivers/dri/r600/r700_vertprog.c
index 550594e9df..f8f862b33a 100644
--- a/src/mesa/drivers/dri/r600/r700_vertprog.c
+++ b/src/mesa/drivers/dri/r600/r700_vertprog.c
@@ -296,16 +296,16 @@ void r700SelectVertexShader(GLcontext *ctx)
     context_t *context = R700_CONTEXT(ctx);
     struct r700_vertex_program *vpc
              = (struct r700_vertex_program *)ctx->VertexProgram._Current;
+    TNLcontext *tnl = TNL_CONTEXT(ctx);
+    struct vertex_buffer *vb = &tnl->vb;
+    unsigned int unBit;
+    unsigned int i;
+
     if (context->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV770)
     {
         vpc->r700AsmCode.bR6xx = 1;
     }
-    
-    TNLcontext *tnl = TNL_CONTEXT(ctx);
-	struct vertex_buffer *vb = &tnl->vb;
 
-    unsigned int unBit;
-    unsigned int i;
 	for(i=0; i<VERT_ATTRIB_MAX; i++)
 	{
 		unBit = 1 << i;
@@ -317,11 +317,9 @@ void r700SelectVertexShader(GLcontext *ctx)
 		}
 	}
 
-    if(GL_FALSE == vpc->translated)
-    {
-        r700TranslateVertexShader(vpc,
-		    					  &(vpc->mesa_program) );
-    }
+	if(GL_FALSE == vpc->translated) {
+		r700TranslateVertexShader(vpc, &(vpc->mesa_program) );
+	}
 }
 
 void * r700GetActiveVpShaderBo(GLcontext * ctx)
@@ -345,17 +343,17 @@ GLboolean r700SetupVertexProgram(GLcontext * ctx)
 
     if(GL_FALSE == vp->loaded)
     {
-        if(vp->r700Shader.bNeedsAssembly == GL_TRUE)
+	    if(vp->r700Shader.bNeedsAssembly == GL_TRUE)
 	    {
 		    Assemble( &(vp->r700Shader) );
 	    }
 
         /* Load vp to gpu */
-        r600EmitShader(ctx, 
-                       &(vp->shaderbo), 
+        r600EmitShader(ctx,
+                       &(vp->shaderbo),
                        (GLvoid *)(vp->r700Shader.pProgram),
                        vp->r700Shader.uShaderBinaryDWORDSize,
-                       "VS"); 
+                       "VS");
 
         vp->loaded = GL_TRUE;
     }
diff --git a/src/mesa/drivers/dri/r600/r700_vertprog.h b/src/mesa/drivers/dri/r600/r700_vertprog.h
index 6a9726a3d0..4c3b7ee453 100644
--- a/src/mesa/drivers/dri/r600/r700_vertprog.h
+++ b/src/mesa/drivers/dri/r600/r700_vertprog.h
@@ -62,23 +62,23 @@ struct r700_vertex_program
 
 //Internal
 unsigned int Map_Vertex_Output(r700_AssemblerBase       *pAsm, 
-					           struct gl_vertex_program *mesa_vp,
-					           unsigned int unStart);
+			       struct gl_vertex_program *mesa_vp,
+			       unsigned int unStart);
 unsigned int Map_Vertex_Input(r700_AssemblerBase       *pAsm, 
-					          struct gl_vertex_program *mesa_vp,
-					          unsigned int unStart);
+			      struct gl_vertex_program *mesa_vp,
+			      unsigned int unStart);
 GLboolean Process_Vertex_Program_Vfetch_Instructions(
-						struct r700_vertex_program *vp,
-						struct gl_vertex_program   *mesa_vp);
+	struct r700_vertex_program *vp,
+	struct gl_vertex_program   *mesa_vp);
 void Map_Vertex_Program(struct r700_vertex_program *vp,
-						struct gl_vertex_program   *mesa_vp);
+			struct gl_vertex_program   *mesa_vp);
 GLboolean Find_Instruction_Dependencies_vp(struct r700_vertex_program *vp,
-					                	   struct gl_vertex_program   *mesa_vp);
+					   struct gl_vertex_program   *mesa_vp);
 
-/* Interface */
-extern GLboolean r700TranslateVertexShader(struct r700_vertex_program *vp,
-							   struct gl_vertex_program   *mesa_vp);
+GLboolean r700TranslateVertexShader(struct r700_vertex_program *vp,
+				    struct gl_vertex_program   *mesa_vp);
 
+/* Interface */
 extern void r700SelectVertexShader(GLcontext *ctx);
 
 extern GLboolean r700SetupVertexProgram(GLcontext * ctx);