r600: move full state to radeon state atoms

9 files changed, 262 insertions, 192 deletions

diff --git a/src/mesa/drivers/dri/r600/r600_context.c b/src/mesa/drivers/dri/r600/r600_context.c
index 4489064c0d..6a066f3510 100644
--- a/src/mesa/drivers/dri/r600/r600_context.c
+++ b/src/mesa/drivers/dri/r600/r600_context.c
@@ -185,7 +185,7 @@ static void r600_vtbl_emit_cs_header(struct radeon_cs *cs, radeonContextPtr rmes
 
 static void r600_vtbl_pre_emit_atoms(radeonContextPtr radeon)
 {
-	/* to be enabled */
+	r700Start3D((context_t *)radeon);
 }
 
 static void r600_fallback(GLcontext *ctx, GLuint bit, GLboolean mode)
diff --git a/src/mesa/drivers/dri/r600/r600_context.h b/src/mesa/drivers/dri/r600/r600_context.h
index a9b080baa3..48be89b638 100644
--- a/src/mesa/drivers/dri/r600/r600_context.h
+++ b/src/mesa/drivers/dri/r600/r600_context.h
@@ -55,10 +55,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 struct r600_context;
 typedef struct r600_context context_t;
 
-extern GLboolean r700SendPSState(context_t *context);
-extern GLboolean r700SendVSState(context_t *context);
-extern GLboolean r700SendFSState(context_t *context);
-
 #include "main/mm.h"
 
 /************ DMA BUFFERS **************/
@@ -126,6 +122,16 @@ struct r600_hw_state {
 	struct radeon_state_atom vgt;
 	struct radeon_state_atom spi;
 	struct radeon_state_atom vpt;
+
+	struct radeon_state_atom fs;
+	struct radeon_state_atom vs;
+	struct radeon_state_atom ps;
+
+	struct radeon_state_atom vs_consts;
+	struct radeon_state_atom ps_consts;
+
+	struct radeon_state_atom vtx;
+	struct radeon_state_atom tx;
 };
 
 /**
@@ -168,22 +174,14 @@ do {							\
 	r600->radeon.hw.is_dirty = GL_TRUE;			\
 } while(0)
 
-extern GLboolean r700SendTextureState(context_t *context);
-
 extern GLboolean r700SyncSurf(context_t *context,
 			      struct radeon_bo *pbo,
 			      uint32_t read_domain,
 			      uint32_t write_domain,
 			      uint32_t sync_type);
 
-extern int       r700SetupStreams(GLcontext * ctx);
-extern void      r700SetupVTXConstants(GLcontext  * ctx, 
-				       unsigned int nStreamID,
-				       void *       pAos,
-				       unsigned int size,      /* number of elements in vector */
-				       unsigned int stride,
-				       unsigned int Count);    /* number of vectors in stream */
-
+extern void r700SetupStreams(GLcontext * ctx);
+extern void r700Start3D(context_t *context);
 extern void r600InitAtoms(context_t *context);
 
 #define RADEON_D_CAPTURE 0
diff --git a/src/mesa/drivers/dri/r600/r600_texstate.c b/src/mesa/drivers/dri/r600/r600_texstate.c
index 1057d7d8bb..7486d73760 100644
--- a/src/mesa/drivers/dri/r600/r600_texstate.c
+++ b/src/mesa/drivers/dri/r600/r600_texstate.c
@@ -60,6 +60,8 @@ void r600UpdateTextureState(GLcontext * ctx)
 	struct radeon_tex_obj *t;
 	GLuint    unit;
 
+	R600_STATECHANGE(context, tx);
+
 	for (unit = 0; unit < R700_MAX_TEXTURE_UNITS; unit++) {
 		texUnit = &ctx->Texture.Unit[unit];
 		t = radeon_tex_obj(ctx->Texture.Unit[unit]._Current);
diff --git a/src/mesa/drivers/dri/r600/r700_chip.c b/src/mesa/drivers/dri/r600/r700_chip.c
index 550fb73d53..b7c2be16ff 100644
--- a/src/mesa/drivers/dri/r600/r700_chip.c
+++ b/src/mesa/drivers/dri/r600/r700_chip.c
@@ -41,11 +41,12 @@
 
 #include "radeon_mipmap_tree.h"
 
-GLboolean r700SendTextureState(context_t *context)
+static void r700SendTextureState(GLcontext *ctx, struct radeon_state_atom *atom)
 {
-    unsigned int i;
+    context_t         *context = R700_CONTEXT(ctx);
     R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
     struct radeon_bo *bo = NULL;
+    unsigned int i;
     BATCH_LOCALS(&context->radeon);
 
     for (i=0; i<R700_TEXTURE_NUMBERUNITS; i++) {
@@ -101,15 +102,14 @@ GLboolean r700SendTextureState(context_t *context)
 		    }
 	    }
     }
-    return GL_TRUE;
 }
 
-void r700SetupVTXConstants(GLcontext  * ctx,
-			   unsigned int nStreamID,
-			   void *       pAos,
-			   unsigned int size,      /* number of elements in vector */
-			   unsigned int stride,
-			   unsigned int count)     /* number of vectors in stream */
+static void r700SetupVTXConstants(GLcontext  * ctx,
+				  unsigned int nStreamID,
+				  void *       pAos,
+				  unsigned int size,      /* number of elements in vector */
+				  unsigned int stride,
+				  unsigned int count)     /* number of vectors in stream */
 {
     context_t *context = R700_CONTEXT(ctx);
     struct radeon_aos * paos = (struct radeon_aos *)pAos;
@@ -170,19 +170,38 @@ void r700SetupVTXConstants(GLcontext  * ctx,
 
 }
 
-int r700SetupStreams(GLcontext * ctx)
+void r700SetupStreams(GLcontext *ctx)
 {
     context_t         *context = R700_CONTEXT(ctx);
-    BATCH_LOCALS(&context->radeon);
-
-    struct r700_vertex_program *vpc
+     struct r700_vertex_program *vpc
              = (struct r700_vertex_program *)ctx->VertexProgram._Current;
-
     TNLcontext *tnl = TNL_CONTEXT(ctx);
-	struct vertex_buffer *vb = &tnl->vb;
+    struct vertex_buffer *vb = &tnl->vb;
+    unsigned int i, j = 0;
+
+    R600_STATECHANGE(context, vtx);
+
+    for(i=0; i<VERT_ATTRIB_MAX; i++) {
+	    if(vpc->mesa_program.Base.InputsRead & (1 << i)) {
+		    rcommon_emit_vector(ctx,
+					&context->radeon.tcl.aos[j],
+					vb->AttribPtr[i]->data,
+					vb->AttribPtr[i]->size,
+					vb->AttribPtr[i]->stride,
+					vb->Count);
+		    j++;
+	    }
+    }
+    context->radeon.tcl.aos_count = j;
+}
 
-    unsigned int unBit;
+static void r700SendVTXState(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+    context_t         *context = R700_CONTEXT(ctx);
+    struct r700_vertex_program *vpc
+             = (struct r700_vertex_program *)ctx->VertexProgram._Current;
     unsigned int i, j = 0;
+    BATCH_LOCALS(&context->radeon);
 
     BEGIN_BATCH_NO_AUTOSTATE(6);
     R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_CTL_CONST, 1));
@@ -195,31 +214,18 @@ int r700SetupStreams(GLcontext * ctx)
     END_BATCH();
     COMMIT_BATCH();
 
-	for(i=0; i<VERT_ATTRIB_MAX; i++)
-	{
-		unBit = 1 << i;
-		if(vpc->mesa_program.Base.InputsRead & unBit)
-		{
-			rcommon_emit_vector(ctx,
-					    &context->radeon.tcl.aos[j],
-					    vb->AttribPtr[i]->data,
-					    vb->AttribPtr[i]->size,
-					    vb->AttribPtr[i]->stride,
-					    vb->Count);
-
-			/* currently aos are packed */
-			r700SetupVTXConstants(ctx,
-					      i,
-					      (void*)(&context->radeon.tcl.aos[j]),
-					      (unsigned int)context->radeon.tcl.aos[j].components,
-					      (unsigned int)context->radeon.tcl.aos[j].stride * 4,
-					      (unsigned int)context->radeon.tcl.aos[j].count);
-			j++;
-		}
-	}
-	context->radeon.tcl.aos_count = j;
-
-    return R600_FALLBACK_NONE;
+    for(i=0; i<VERT_ATTRIB_MAX; i++) {
+	    if(vpc->mesa_program.Base.InputsRead & (1 << i)) {
+		    /* currently aos are packed */
+		    r700SetupVTXConstants(ctx,
+					  i,
+					  (void*)(&context->radeon.tcl.aos[j]),
+					  (unsigned int)context->radeon.tcl.aos[j].components,
+					  (unsigned int)context->radeon.tcl.aos[j].stride * 4,
+					  (unsigned int)context->radeon.tcl.aos[j].count);
+		    j++;
+	    }
+    }
 }
 
 static void r700SendDepthTargetState(GLcontext *ctx, struct radeon_state_atom *atom)
@@ -310,8 +316,9 @@ static void r700SendRenderTargetState(GLcontext *ctx, struct radeon_state_atom *
 
 }
 
-GLboolean r700SendPSState(context_t *context)
+static void r700SendPSState(GLcontext *ctx, struct radeon_state_atom *atom)
 {
+	context_t *context = R700_CONTEXT(ctx);
 	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
 	struct radeon_bo * pbo;
 	BATCH_LOCALS(&context->radeon);
@@ -319,7 +326,7 @@ GLboolean r700SendPSState(context_t *context)
 	pbo = (struct radeon_bo *)r700GetActiveFpShaderBo(GL_CONTEXT(context));
 
 	if (!pbo)
-		return GL_FALSE;
+		return;
 
 	r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit);
 
@@ -340,13 +347,11 @@ GLboolean r700SendPSState(context_t *context)
 
 	COMMIT_BATCH();
 
-	r700->ps.dirty = GL_FALSE;
-
-	return GL_TRUE;
 }
 
-GLboolean r700SendVSState(context_t *context)
+static void r700SendVSState(GLcontext *ctx, struct radeon_state_atom *atom)
 {
+	context_t *context = R700_CONTEXT(ctx);
 	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
 	struct radeon_bo * pbo;
 	BATCH_LOCALS(&context->radeon);
@@ -354,7 +359,7 @@ GLboolean r700SendVSState(context_t *context)
 	pbo = (struct radeon_bo *)r700GetActiveVpShaderBo(GL_CONTEXT(context));
 
 	if (!pbo)
-		return GL_FALSE;
+		return;
 
 	r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit);
 
@@ -373,14 +378,11 @@ GLboolean r700SendVSState(context_t *context)
         END_BATCH();
 
 	COMMIT_BATCH();
-
-	r700->vs.dirty = GL_FALSE;
-
-	return GL_TRUE;
 }
 
-GLboolean r700SendFSState(context_t *context)
+static void r700SendFSState(GLcontext *ctx, struct radeon_state_atom *atom)
 {
+	context_t *context = R700_CONTEXT(ctx);
 	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
 	struct radeon_bo * pbo;
 	BATCH_LOCALS(&context->radeon);
@@ -397,7 +399,7 @@ GLboolean r700SendFSState(context_t *context)
 	/* XXX */
 
 	if (!pbo)
-		return GL_FALSE;
+		return;
 
 	r700SyncSurf(context, pbo, RADEON_GEM_DOMAIN_GTT, 0, SH_ACTION_ENA_bit);
 
@@ -417,9 +419,6 @@ GLboolean r700SendFSState(context_t *context)
 
 	COMMIT_BATCH();
 
-	r700->fs.dirty = GL_FALSE;
-
-	return GL_TRUE;
 }
 
 static void r700SendViewportState(GLcontext *ctx, struct radeon_state_atom *atom)
@@ -831,18 +830,103 @@ static void r700SendSCState(GLcontext *ctx, struct radeon_state_atom *atom)
 	COMMIT_BATCH();
 }
 
+static void r700SendPSConsts(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	int i;
+	BATCH_LOCALS(&context->radeon);
+
+	if (r700->ps.num_consts == 0)
+		return;
+
+	BEGIN_BATCH_NO_AUTOSTATE(2 + (r700->ps.num_consts * 4));
+	R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, (r700->ps.num_consts * 4)));
+	/* assembler map const from very beginning. */
+	R600_OUT_BATCH(SQ_ALU_CONSTANT_PS_OFFSET * 4);
+	for (i = 0; i < r700->ps.num_consts; i++) {
+		R600_OUT_BATCH(r700->ps.consts[i][0].u32All);
+		R600_OUT_BATCH(r700->ps.consts[i][1].u32All);
+		R600_OUT_BATCH(r700->ps.consts[i][2].u32All);
+		R600_OUT_BATCH(r700->ps.consts[i][3].u32All);
+	}
+	END_BATCH();
+	COMMIT_BATCH();
+}
+
+static void r700SendVSConsts(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	R700_CHIP_CONTEXT *r700 = R700_CONTEXT_STATES(context);
+	int i;
+	BATCH_LOCALS(&context->radeon);
+
+	if (r700->vs.num_consts == 0)
+		return;
+
+	BEGIN_BATCH_NO_AUTOSTATE(2 + (r700->vs.num_consts * 4));
+	R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, (r700->vs.num_consts * 4)));
+	/* assembler map const from very beginning. */
+	R600_OUT_BATCH(SQ_ALU_CONSTANT_VS_OFFSET * 4);
+	for (i = 0; i < r700->vs.num_consts; i++) {
+		R600_OUT_BATCH(r700->vs.consts[i][0].u32All);
+		R600_OUT_BATCH(r700->vs.consts[i][1].u32All);
+		R600_OUT_BATCH(r700->vs.consts[i][2].u32All);
+		R600_OUT_BATCH(r700->vs.consts[i][3].u32All);
+	}
+	END_BATCH();
+	COMMIT_BATCH();
+}
+
 static int check_always(GLcontext *ctx, struct radeon_state_atom *atom)
 {
 	return atom->cmd_size;
 }
 
-#define ALLOC_STATE( ATOM, SZ, EMIT )					\
+static int check_vtx(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+
+	return context->radeon.tcl.aos_count * 18;
+}
+
+static int check_tx(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+	context_t *context = R700_CONTEXT(ctx);
+	unsigned int i, count = 0;
+	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
+
+	for (i = 0; i < R700_TEXTURE_NUMBERUNITS; i++) {
+		radeonTexObj *t = r700->textures[i];
+		if (t)
+			count++;
+	}
+	return count * 31;
+}
+
+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);
+
+	return 2 + (r700->ps.num_consts * 4);
+}
+
+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);
+
+	return 2 + (r700->vs.num_consts * 4);
+}
+
+#define ALLOC_STATE( ATOM, CHK, SZ, EMIT )				\
 do {									\
 	context->atoms.ATOM.cmd_size = (SZ);				\
 	context->atoms.ATOM.cmd = NULL;					\
 	context->atoms.ATOM.name = #ATOM;				\
 	context->atoms.ATOM.idx = 0;					\
-	context->atoms.ATOM.check = check_always;			\
+	context->atoms.ATOM.check = check_##CHK;			\
 	context->atoms.ATOM.dirty = GL_FALSE;				\
 	context->atoms.ATOM.emit = (EMIT);				\
 	context->radeon.hw.max_state_size += (SZ);			\
@@ -851,26 +935,36 @@ do {									\
 
 void r600InitAtoms(context_t *context)
 {
-	/* FIXME: rough estimate for "large" const and shader state */
-	context->radeon.hw.max_state_size = 7500;
+	context->radeon.hw.max_state_size = 10 + 5 + 14; /* start 3d, idle, cb/db flush */
 
 	/* Setup the atom linked list */
 	make_empty_list(&context->radeon.hw.atomlist);
 	context->radeon.hw.atomlist.name = "atom-list";
 
-	ALLOC_STATE(sq, 34, r700SendSQConfig);
-	ALLOC_STATE(db, 27, r700SendDBState);
-	ALLOC_STATE(db_target, 19, r700SendDepthTargetState);
-	ALLOC_STATE(sc, 47, r700SendSCState);
-	ALLOC_STATE(cl, 18, r700SendCLState);
-	ALLOC_STATE(ucp, 36, r700SendUCPState);
-	ALLOC_STATE(su, 19, r700SendSUState);
-	ALLOC_STATE(cb, 39, r700SendCBState);
-	ALLOC_STATE(cb_target, 32, r700SendRenderTargetState);
-	ALLOC_STATE(sx, 9, r700SendSXState);
-	ALLOC_STATE(vgt, 41, r700SendVGTState);
-	ALLOC_STATE(spi, (59 + R700_MAX_SHADER_EXPORTS), r700SendSPIState);
-	ALLOC_STATE(vpt, 16, r700SendViewportState);
+	ALLOC_STATE(sq, always, 34, r700SendSQConfig);
+
+	ALLOC_STATE(db, always, 27, r700SendDBState);
+	ALLOC_STATE(db_target, always, 19, r700SendDepthTargetState);
+	ALLOC_STATE(sc, always, 47, r700SendSCState);
+	ALLOC_STATE(cl, always, 18, r700SendCLState);
+	ALLOC_STATE(ucp, always, 36, r700SendUCPState);
+	ALLOC_STATE(su, always, 19, r700SendSUState);
+	ALLOC_STATE(cb, always, 39, r700SendCBState);
+	ALLOC_STATE(cb_target, always, 32, r700SendRenderTargetState);
+	ALLOC_STATE(sx, always, 9, r700SendSXState);
+	ALLOC_STATE(vgt, always, 41, r700SendVGTState);
+	ALLOC_STATE(spi, always, (59 + R700_MAX_SHADER_EXPORTS), r700SendSPIState);
+	ALLOC_STATE(vpt, always, 16, r700SendViewportState);
+
+	ALLOC_STATE(fs, always, 18, r700SendFSState);
+	ALLOC_STATE(vs, always, 18, r700SendVSState);
+	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(tx, tx, (R700_TEXTURE_NUMBERUNITS * 31), r700SendTextureState);
 
 	context->radeon.hw.is_dirty = GL_TRUE;
 	context->radeon.hw.all_dirty = GL_TRUE;
diff --git a/src/mesa/drivers/dri/r600/r700_chip.h b/src/mesa/drivers/dri/r600/r700_chip.h
index c0ec4b0dd5..ae249e15fd 100644
--- a/src/mesa/drivers/dri/r600/r700_chip.h
+++ b/src/mesa/drivers/dri/r600/r700_chip.h
@@ -46,6 +46,7 @@
 #define R700_MAX_VIEWPORTS       16
 #define R700_MAX_SHADER_EXPORTS  32
 #define R700_MAX_UCP             6
+#define R700_MAX_DX9_CONSTS      256
 
 /* Enum not show in r600_*.h */
 
@@ -224,6 +225,8 @@ typedef struct _PS_STATE_STRUCT
 	union UINT_FLOAT         	SQ_PGM_EXPORTS_PS         ;  /* 0xA215 */
 	union UINT_FLOAT       	        SQ_PGM_CF_OFFSET_PS       ;  /* 0xA233 */
 	GLboolean                         dirty;
+	int                             num_consts;
+	union UINT_FLOAT                consts[R700_MAX_DX9_CONSTS][4];
 } PS_STATE_STRUCT;
 
 typedef struct _VS_STATE_STRUCT
@@ -232,6 +235,8 @@ typedef struct _VS_STATE_STRUCT
 	union UINT_FLOAT  		SQ_PGM_RESOURCES_VS       ;  /* 0xA21A */
 	union UINT_FLOAT       	        SQ_PGM_CF_OFFSET_VS       ;  /* 0xA234 */
 	GLboolean                         dirty;
+	int                             num_consts;
+	union UINT_FLOAT                consts[R700_MAX_DX9_CONSTS][4];
 } VS_STATE_STRUCT;
 
 typedef struct _GS_STATE_STRUCT
diff --git a/src/mesa/drivers/dri/r600/r700_fragprog.c b/src/mesa/drivers/dri/r600/r700_fragprog.c
index 394482594e..098b420dfc 100644
--- a/src/mesa/drivers/dri/r600/r700_fragprog.c
+++ b/src/mesa/drivers/dri/r600/r700_fragprog.c
@@ -270,7 +270,6 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx)
 {
     context_t *context = R700_CONTEXT(ctx);
     R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
-    BATCH_LOCALS(&context->radeon);
     struct r700_fragment_program *fp = (struct r700_fragment_program *)
 	                                   (ctx->FragmentProgram._Current);
     r700_AssemblerBase         *pAsm = &(fp->r700AsmCode);
@@ -280,6 +279,7 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx)
     unsigned int ui, i;
     unsigned int unNumOfReg;
     unsigned int unBit;
+    GLuint exportCount;
 
     if(GL_FALSE == fp->loaded)
     {
@@ -305,10 +305,15 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx)
     (context->chipobj.MemUse)(context, fp->shadercode.buf->id);
     */
 
-    R600_STATECHANGE(context, spi);
+    R600_STATECHANGE(context, ps);
+
+    r700->ps.SQ_PGM_RESOURCES_PS.u32All = 0;
+    SETbit(r700->ps.SQ_PGM_RESOURCES_PS.u32All, PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit);
 
     r700->ps.SQ_PGM_START_PS.u32All = 0; /* set from buffer obj */
 
+    R600_STATECHANGE(context, spi);
+
     unNumOfReg = fp->r700Shader.nRegs + 1;
 
     ui = (r700->SPI_PS_IN_CONTROL_0.u32All & NUM_INTERP_mask) / (1 << NUM_INTERP_shift);
@@ -325,8 +330,8 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx)
 
     ui = (unNumOfReg < ui) ? ui : unNumOfReg;
 
-    SETfield(r700->ps.SQ_PGM_RESOURCES_PS.u32All, ui, NUM_GPRS_shift, NUM_GPRS_mask); 
-    
+    SETfield(r700->ps.SQ_PGM_RESOURCES_PS.u32All, ui, NUM_GPRS_shift, NUM_GPRS_mask);
+
     CLEARbit(r700->ps.SQ_PGM_RESOURCES_PS.u32All, UNCACHED_FIRST_INST_bit);
 
     if(fp->r700Shader.uStackSize) /* we don't use branch for now, it should be zero. */
@@ -338,6 +343,8 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx)
     SETfield(r700->ps.SQ_PGM_EXPORTS_PS.u32All, fp->r700Shader.exportMode,
              EXPORT_MODE_shift, EXPORT_MODE_mask);
 
+    R600_STATECHANGE(context, db);
+
     if(fp->r700Shader.killIsUsed)
     {
 	    SETbit(r700->DB_SHADER_CONTROL.u32All, KILL_ENABLE_bit);
@@ -349,42 +356,13 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx)
 
     if(fp->r700Shader.depthIsExported)
     {
-	    SETbit(r700->DB_SHADER_CONTROL.u32All, Z_EXPORT_ENABLE_bit); 
+	    SETbit(r700->DB_SHADER_CONTROL.u32All, Z_EXPORT_ENABLE_bit);
     }
     else
     {
         CLEARbit(r700->DB_SHADER_CONTROL.u32All, Z_EXPORT_ENABLE_bit);
     }
 
-    /* sent out shader constants. */
-    paramList = fp->mesa_program.Base.Parameters;
-
-    if(NULL != paramList)
-    {
-        _mesa_load_state_parameters(ctx, paramList);
-
-        unNumParamData = paramList->NumParameters * 4;
-
-        BEGIN_BATCH_NO_AUTOSTATE(2 + unNumParamData);
-
-        R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, unNumParamData));
-
-        /* assembler map const from very beginning. */
-        R600_OUT_BATCH(SQ_ALU_CONSTANT_PS_OFFSET * 4);
-
-        unNumParamData = paramList->NumParameters;
-
-        for(ui=0; ui<unNumParamData; ui++)
-        {
-            R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][0])));
-            R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][1])));
-            R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][2])));
-            R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][3])));
-        }
-        END_BATCH();
-        COMMIT_BATCH();
-    }
-
     // emit ps input map
     unBit = 1 << FRAG_ATTRIB_WPOS;
     if(mesa_fp->Base.InputsRead & unBit)
@@ -451,6 +429,34 @@ GLboolean r700SetupFragmentProgram(GLcontext * ctx)
 	    }
     }
 
+    R600_STATECHANGE(context, cb);
+    exportCount = (r700->ps.SQ_PGM_EXPORTS_PS.u32All & EXPORT_MODE_mask) / (1 << EXPORT_MODE_shift);
+    r700->CB_SHADER_CONTROL.u32All = (1 << exportCount) - 1;
+
+    /* sent out shader constants. */
+    paramList = fp->mesa_program.Base.Parameters;
+
+    if(NULL != paramList) {
+	    _mesa_load_state_parameters(ctx, paramList);
+
+	    if (paramList->NumParameters > R700_MAX_DX9_CONSTS)
+		    return GL_FALSE;
+
+	    R600_STATECHANGE(context, ps_consts);
+
+	    r700->ps.num_consts = paramList->NumParameters;
+
+	    unNumParamData = paramList->NumParameters;
+
+	    for(ui=0; ui<unNumParamData; ui++) {
+		    r700->ps.consts[ui][0].f32All = paramList->ParameterValues[ui][0];
+		    r700->ps.consts[ui][1].f32All = paramList->ParameterValues[ui][1];
+		    r700->ps.consts[ui][2].f32All = paramList->ParameterValues[ui][2];
+		    r700->ps.consts[ui][3].f32All = paramList->ParameterValues[ui][3];
+	    }
+    } else
+	    r700->ps.num_consts = 0;
+
     return GL_TRUE;
 }
 
diff --git a/src/mesa/drivers/dri/r600/r700_render.c b/src/mesa/drivers/dri/r600/r700_render.c
index f0cd357c76..c8b405f555 100644
--- a/src/mesa/drivers/dri/r600/r700_render.c
+++ b/src/mesa/drivers/dri/r600/r700_render.c
@@ -55,7 +55,6 @@
 
 void r700WaitForIdle(context_t *context);
 void r700WaitForIdleClean(context_t *context);
-void r700Start3D(context_t *context);
 GLboolean r700SendTextureState(context_t *context);
 static unsigned int r700PrimitiveType(int prim);
 void r600UpdateTextureState(GLcontext * ctx);
@@ -116,39 +115,6 @@ void r700Start3D(context_t *context)
     r700WaitForIdleClean(context);
 }
 
-static GLboolean r700SetupShaders(GLcontext * ctx)
-{
-    context_t *context = R700_CONTEXT(ctx);
-
-    R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
-
-    GLuint exportCount;
-
-    r700->ps.SQ_PGM_RESOURCES_PS.u32All = 0;
-    r700->vs.SQ_PGM_RESOURCES_VS.u32All = 0;
-
-    SETbit(r700->ps.SQ_PGM_RESOURCES_PS.u32All, PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit);
-    SETbit(r700->vs.SQ_PGM_RESOURCES_VS.u32All, PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit);
-
-    r700SetupVertexProgram(ctx);
-
-    r700SetupFragmentProgram(ctx);
-
-    exportCount = (r700->ps.SQ_PGM_EXPORTS_PS.u32All & EXPORT_MODE_mask) / (1 << EXPORT_MODE_shift);
-    r700->CB_SHADER_CONTROL.u32All = (1 << exportCount) - 1;
-
-    r600UpdateTextureState(ctx);
-
-    r700SendFSState(context); // FIXME just a place holder for now
-    r700SendPSState(context);
-    r700SendVSState(context);
-
-    r700SendTextureState(context);
-    r700SetupStreams(ctx);
-
-    return GL_TRUE;
-}
-
 GLboolean r700SyncSurf(context_t *context,
 		       struct radeon_bo *pbo,
 		       uint32_t read_domain,
@@ -333,7 +299,7 @@ static void r700RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim
 }
 
 static GLboolean r700RunRender(GLcontext * ctx,
-			                   struct tnl_pipeline_stage *stage)
+			       struct tnl_pipeline_stage *stage)
 {
     context_t *context = R700_CONTEXT(ctx);
     radeonContextPtr radeon = &context->radeon;
@@ -347,12 +313,15 @@ static GLboolean r700RunRender(GLcontext * ctx,
 
     /* just an estimate, need to properly calculate this */
     rcommonEnsureCmdBufSpace(&context->radeon,
-			     radeon->hw.max_state_size + ind_count + 1000, __FUNCTION__);
+			     radeon->hw.max_state_size + ind_count, __FUNCTION__);
 
-    r700Start3D(context);
     r700UpdateShaders(ctx);
     r700SetScissor(context);
-    r700SetupShaders(ctx);
+    r700SetupVertexProgram(ctx);
+    r700SetupFragmentProgram(ctx);
+    r600UpdateTextureState(ctx);
+    r700SetupStreams(ctx);
+
     radeonEmitState(radeon);
 
     /* richard test code */
diff --git a/src/mesa/drivers/dri/r600/r700_state.c b/src/mesa/drivers/dri/r600/r700_state.c
index 97c7d7c385..0f25102a79 100644
--- a/src/mesa/drivers/dri/r600/r700_state.c
+++ b/src/mesa/drivers/dri/r600/r700_state.c
@@ -168,7 +168,6 @@ void r700UpdateViewportOffset(GLcontext * ctx) //------------------
 void r700UpdateDrawBuffer(GLcontext * ctx) /* TODO */ //---------------------
 {
 	context_t *context = R700_CONTEXT(ctx);
-	R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
 
 	R600_STATECHANGE(context, cb_target);
 	R600_STATECHANGE(context, db_target);
@@ -1422,7 +1421,6 @@ static void r700SetRenderTarget(context_t *context, int id)
 
     rrb = radeon_get_colorbuffer(&context->radeon);
     if (!rrb || !rrb->bo) {
-	    fprintf(stderr, "no rrb\n");
 	    return;
     }
 
diff --git a/src/mesa/drivers/dri/r600/r700_vertprog.c b/src/mesa/drivers/dri/r600/r700_vertprog.c
index 14da4ed1d3..550594e9df 100644
--- a/src/mesa/drivers/dri/r600/r700_vertprog.c
+++ b/src/mesa/drivers/dri/r600/r700_vertprog.c
@@ -336,7 +336,6 @@ GLboolean r700SetupVertexProgram(GLcontext * ctx)
 {
     context_t *context = R700_CONTEXT(ctx);
     R700_CHIP_CONTEXT *r700 = (R700_CHIP_CONTEXT*)(&context->hw);
-    BATCH_LOCALS(&context->radeon);
     struct r700_vertex_program *vp
              = (struct r700_vertex_program *)ctx->VertexProgram._Current;
 
@@ -368,10 +367,14 @@ GLboolean r700SetupVertexProgram(GLcontext * ctx)
     (context->chipobj.MemUse)(context, vp->shadercode.buf->id);
     */
 
-    R600_STATECHANGE(context, spi);
+    R600_STATECHANGE(context, vs);
+    R600_STATECHANGE(context, fs); /* hack */
+
+    r700->vs.SQ_PGM_RESOURCES_VS.u32All = 0;
+    SETbit(r700->vs.SQ_PGM_RESOURCES_VS.u32All, PGM_RESOURCES__PRIME_CACHE_ON_DRAW_bit);
+
+    r700->vs.SQ_PGM_START_VS.u32All = 0; /* set from buffer object. */
 
-    r700->vs.SQ_PGM_START_VS.u32All = 0; /* set from buffer object. */ 
-    
     SETfield(r700->vs.SQ_PGM_RESOURCES_VS.u32All, vp->r700Shader.nRegs + 1,
              NUM_GPRS_shift, NUM_GPRS_mask);
 
@@ -381,9 +384,12 @@ GLboolean r700SetupVertexProgram(GLcontext * ctx)
                  STACK_SIZE_shift, STACK_SIZE_mask);
     }
 
-    SETfield(r700->SPI_VS_OUT_CONFIG.u32All, vp->r700Shader.nParamExports ? (vp->r700Shader.nParamExports - 1) : 0,
+    R600_STATECHANGE(context, spi);
+
+    SETfield(r700->SPI_VS_OUT_CONFIG.u32All,
+	     vp->r700Shader.nParamExports ? (vp->r700Shader.nParamExports - 1) : 0,
              VS_EXPORT_COUNT_shift, VS_EXPORT_COUNT_mask);
-	SETfield(r700->SPI_PS_IN_CONTROL_0.u32All, vp->r700Shader.nParamExports,
+    SETfield(r700->SPI_PS_IN_CONTROL_0.u32All, vp->r700Shader.nParamExports,
              NUM_INTERP_shift, NUM_INTERP_mask);
 
     /*
@@ -394,34 +400,26 @@ GLboolean r700SetupVertexProgram(GLcontext * ctx)
     /* sent out shader constants. */
     paramList = vp->mesa_program.Base.Parameters;
 
-    if(NULL != paramList)
-    {
-        _mesa_load_state_parameters(ctx, paramList);
+    if(NULL != paramList) {
+	    _mesa_load_state_parameters(ctx, paramList);
 
-        unNumParamData = paramList->NumParameters * 4;
+	    if (paramList->NumParameters > R700_MAX_DX9_CONSTS)
+		    return GL_FALSE;
 
-        BEGIN_BATCH_NO_AUTOSTATE(unNumParamData + 2);
+	    R600_STATECHANGE(context, vs_consts);
 
-        R600_OUT_BATCH(CP_PACKET3(R600_IT_SET_ALU_CONST, unNumParamData));
-        /* assembler map const from very beginning. */
-        R600_OUT_BATCH(SQ_ALU_CONSTANT_VS_OFFSET * 4);
+	    r700->vs.num_consts = paramList->NumParameters;
 
-        unNumParamData = paramList->NumParameters;
+	    unNumParamData = paramList->NumParameters;
 
-        for(ui=0; ui<unNumParamData; ui++)
-        {
-            R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][0])));
-            R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][1])));
-            R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][2])));
-            R600_OUT_BATCH(*((unsigned int*)&(paramList->ParameterValues[ui][3])));
-        }
-        END_BATCH();
-        COMMIT_BATCH();
-    }
+	    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];
+	    }
+    } else
+	    r700->vs.num_consts = 0;
 
     return GL_TRUE;
 }
-
-
-
-