radeonsi: move draw state into new handling

Split it out into si_state_draw.c

Signed-off-by: Christian König <[email protected]>

7 files changed, 589 insertions, 558 deletions

diff --git a/src/gallium/drivers/radeonsi/Makefile.sources b/src/gallium/drivers/radeonsi/Makefile.sources
index 8aeea8d574f..b5c193b6585 100644
--- a/src/gallium/drivers/radeonsi/Makefile.sources
+++ b/src/gallium/drivers/radeonsi/Makefile.sources
@@ -12,4 +12,5 @@ C_SOURCES := \
 	r600_translate.c \
 	r600_state_common.c \
 	radeonsi_pm4.c \
-	si_state.c
+	si_state.c \
+	si_state_draw.c
diff --git a/src/gallium/drivers/radeonsi/r600_state_common.c b/src/gallium/drivers/radeonsi/r600_state_common.c
index af8a0b49de9..d54ffc56f09 100644
--- a/src/gallium/drivers/radeonsi/r600_state_common.c
+++ b/src/gallium/drivers/radeonsi/r600_state_common.c
@@ -176,24 +176,6 @@ void *si_create_vertex_elements(struct pipe_context *ctx,
 	return v;
 }
 
-static void r600_update_alpha_ref(struct r600_context *rctx)
-{
-#if 0
-	unsigned alpha_ref;
-	struct r600_pipe_state rstate;
-
-	alpha_ref = rctx->alpha_ref;
-	rstate.nregs = 0;
-	if (rctx->export_16bpc)
-		alpha_ref &= ~0x1FFF;
-	r600_pipe_state_add_reg(&rstate, R_028438_SX_ALPHA_REF, alpha_ref, NULL, 0);
-
-	r600_context_pipe_state_set(rctx, &rstate);
-	rctx->alpha_ref_dirty = false;
-#endif
-}
-
-
 struct pipe_stream_output_target *
 r600_create_so_target(struct pipe_context *ctx,
 		      struct pipe_resource *buffer,
@@ -259,245 +241,6 @@ void r600_set_so_targets(struct pipe_context *ctx,
 	rctx->streamout_append_bitmask = append_bitmask;
 }
 
-static void r600_vertex_buffer_update(struct r600_context *rctx)
-{
-	struct pipe_context *ctx = &rctx->context;
-	struct r600_pipe_state *rstate = &rctx->vs_user_data;
-	struct r600_resource *rbuffer, *t_list_buffer;
-	struct pipe_vertex_buffer *vertex_buffer;
-	unsigned i, count, offset;
-	uint32_t *ptr;
-	uint64_t va;
-
-	r600_inval_vertex_cache(rctx);
-
-	/* bind vertex buffer once */
-	count = rctx->nr_vertex_buffers;
-	assert(count <= 256 / 4);
-
-	t_list_buffer = (struct r600_resource*)
-		pipe_buffer_create(ctx->screen, PIPE_BIND_CUSTOM,
-				   PIPE_USAGE_IMMUTABLE, 4 * 4 * count);
-	if (t_list_buffer == NULL)
-		return;
-
-	ptr = (uint32_t*)rctx->ws->buffer_map(t_list_buffer->cs_buf,
-					      rctx->cs,
-					      PIPE_TRANSFER_WRITE);
-
-	for (i = 0 ; i < count; i++, ptr += 4) {
-		struct pipe_vertex_element *velem = &rctx->vertex_elements->elements[i];
-		const struct util_format_description *desc;
-		unsigned data_format, num_format;
-		int first_non_void;
-
-		/* bind vertex buffer once */
-		vertex_buffer = &rctx->vertex_buffer[i];
-		rbuffer = (struct r600_resource*)vertex_buffer->buffer;
-		offset = 0;
-		if (vertex_buffer == NULL || rbuffer == NULL)
-			continue;
-		offset += vertex_buffer->buffer_offset;
-
-		va = r600_resource_va(ctx->screen, (void*)rbuffer);
-		va += offset;
-
-		desc = util_format_description(velem->src_format);
-		first_non_void = util_format_get_first_non_void_channel(velem->src_format);
-		data_format = si_translate_vertexformat(ctx->screen,
-							velem->src_format,
-							desc, first_non_void);
-
-		switch (desc->channel[first_non_void].type) {
-		case UTIL_FORMAT_TYPE_FIXED:
-			num_format = V_008F0C_BUF_NUM_FORMAT_USCALED; /* XXX */
-			break;
-		case UTIL_FORMAT_TYPE_SIGNED:
-			num_format = V_008F0C_BUF_NUM_FORMAT_SNORM;
-			break;
-		case UTIL_FORMAT_TYPE_UNSIGNED:
-			num_format = V_008F0C_BUF_NUM_FORMAT_UNORM;
-			break;
-		case UTIL_FORMAT_TYPE_FLOAT:
-		default:
-			num_format = V_008F14_IMG_NUM_FORMAT_FLOAT;
-		}
-
-		/* Fill in T# buffer resource description */
-		ptr[0] = va & 0xFFFFFFFF;
-		ptr[1] = (S_008F04_BASE_ADDRESS_HI(va >> 32) |
-			  S_008F04_STRIDE(vertex_buffer->stride));
-		if (vertex_buffer->stride > 0)
-			ptr[2] = ((vertex_buffer->buffer->width0 - offset) /
-				  vertex_buffer->stride);
-		else
-			ptr[2] = vertex_buffer->buffer->width0 - offset;
-		ptr[3] = (S_008F0C_DST_SEL_X(si_map_swizzle(desc->swizzle[0])) |
-			  S_008F0C_DST_SEL_Y(si_map_swizzle(desc->swizzle[1])) |
-			  S_008F0C_DST_SEL_Z(si_map_swizzle(desc->swizzle[2])) |
-			  S_008F0C_DST_SEL_W(si_map_swizzle(desc->swizzle[3])) |
-			  S_008F0C_NUM_FORMAT(num_format) |
-			  S_008F0C_DATA_FORMAT(data_format));
-
-		r600_context_bo_reloc(rctx, rbuffer, RADEON_USAGE_READ);
-	}
-
-	rstate->nregs = 0;
-
-	va = r600_resource_va(ctx->screen, (void*)t_list_buffer);
-	r600_pipe_state_add_reg(rstate,
-				R_00B148_SPI_SHADER_USER_DATA_VS_6,
-				va, t_list_buffer, RADEON_USAGE_READ);
-	r600_pipe_state_add_reg(rstate,
-				R_00B14C_SPI_SHADER_USER_DATA_VS_7,
-				va >> 32,
-				NULL, 0);
-
-	r600_context_pipe_state_set(rctx, rstate);
-}
-
-static void si_update_derived_state(struct r600_context *rctx)
-{
-	struct pipe_context * ctx = (struct pipe_context*)rctx;
-
-	if (!rctx->blitter->running) {
-		if (rctx->have_depth_fb || rctx->have_depth_texture)
-			r600_flush_depth_textures(rctx);
-	}
-
-	if ((rctx->ps_shader->shader.fs_write_all &&
-	     (rctx->ps_shader->shader.nr_cbufs != rctx->framebuffer.nr_cbufs)) ||
-	    (rctx->sprite_coord_enable &&
-	     (rctx->ps_shader->sprite_coord_enable != rctx->sprite_coord_enable))) {
-		si_pipe_shader_destroy(&rctx->context, rctx->ps_shader);
-	}
-
-	if (rctx->alpha_ref_dirty) {
-		r600_update_alpha_ref(rctx);
-	}
-
-	if (!rctx->vs_shader->bo) {
-		si_pipe_shader_vs(ctx, rctx->vs_shader);
-	}
-
-	if (!rctx->ps_shader->bo) {
-		si_pipe_shader_ps(ctx, rctx->ps_shader);
-	}
-
-	if (rctx->shader_dirty) {
-		si_update_spi_map(rctx);
-		rctx->shader_dirty = false;
-	}
-}
-
-void r600_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *dinfo)
-{
-	struct r600_context *rctx = (struct r600_context *)ctx;
-	struct si_state_dsa *dsa = rctx->queued.named.dsa;
-	struct pipe_draw_info info = *dinfo;
-	struct r600_draw rdraw = {};
-	struct pipe_index_buffer ib = {};
-	struct r600_block *dirty_block = NULL, *next_block = NULL;
-	struct r600_atom *state = NULL, *next_state = NULL;
-	int i;
-
-	if ((!info.count && (info.indexed || !info.count_from_stream_output)) ||
-	    (info.indexed && !rctx->index_buffer.buffer)) {
-		return;
-	}
-
-	if (!rctx->ps_shader || !rctx->vs_shader)
-		return;
-
-	si_update_derived_state(rctx);
-
-	r600_vertex_buffer_update(rctx);
-
-	rdraw.vgt_num_indices = info.count;
-	rdraw.vgt_num_instances = info.instance_count;
-
-	if (info.indexed) {
-		/* Initialize the index buffer struct. */
-		pipe_resource_reference(&ib.buffer, rctx->index_buffer.buffer);
-		ib.index_size = rctx->index_buffer.index_size;
-		ib.offset = rctx->index_buffer.offset + info.start * ib.index_size;
-
-		/* Translate or upload, if needed. */
-		r600_translate_index_buffer(rctx, &ib, info.count);
-
-		if (ib.user_buffer) {
-			r600_upload_index_buffer(rctx, &ib, info.count);
-		}
-
-		/* Initialize the r600_draw struct with index buffer info. */
-		if (ib.index_size == 4) {
-			rdraw.vgt_index_type = V_028A7C_VGT_INDEX_32 |
-				(R600_BIG_ENDIAN ? V_028A7C_VGT_DMA_SWAP_32_BIT : 0);
-		} else {
-			rdraw.vgt_index_type = V_028A7C_VGT_INDEX_16 |
-				(R600_BIG_ENDIAN ? V_028A7C_VGT_DMA_SWAP_16_BIT : 0);
-		}
-		rdraw.indices = (struct r600_resource*)ib.buffer;
-		rdraw.indices_bo_offset = ib.offset;
-		rdraw.vgt_draw_initiator = V_0287F0_DI_SRC_SEL_DMA;
-	} else {
-		info.index_bias = info.start;
-		rdraw.vgt_draw_initiator = V_0287F0_DI_SRC_SEL_AUTO_INDEX;
-		if (info.count_from_stream_output) {
-			rdraw.vgt_draw_initiator |= S_0287F0_USE_OPAQUE(1);
-
-			r600_context_draw_opaque_count(rctx, (struct r600_so_target*)info.count_from_stream_output);
-		}
-	}
-
-	rctx->vs_shader_so_strides = rctx->vs_shader->so_strides;
-
-	if (!si_update_draw_info_state(rctx, &info))
-		return;
-
-	rdraw.db_render_override = dsa->db_render_override;
-	rdraw.db_render_control = dsa->db_render_control;
-
-	/* Emit states. */
-	rctx->pm4_dirty_cdwords += si_pm4_dirty_dw(rctx);
-
-	r600_need_cs_space(rctx, 0, TRUE);
-
-	LIST_FOR_EACH_ENTRY_SAFE(state, next_state, &rctx->dirty_states, head) {
-		r600_emit_atom(rctx, state);
-	}
-	LIST_FOR_EACH_ENTRY_SAFE(dirty_block, next_block, &rctx->dirty,list) {
-		r600_context_block_emit_dirty(rctx, dirty_block);
-	}
-	si_pm4_emit_dirty(rctx);
-	rctx->pm4_dirty_cdwords = 0;
-
-	/* Enable stream out if needed. */
-	if (rctx->streamout_start) {
-		r600_context_streamout_begin(rctx);
-		rctx->streamout_start = FALSE;
-	}
-
-	for (i = 0; i < NUM_TEX_UNITS; i++) {
-		if (rctx->ps_samplers.views[i])
-			r600_context_bo_reloc(rctx,
-					      (struct r600_resource*)rctx->ps_samplers.views[i]->base.texture,
-					      RADEON_USAGE_READ);
-	}
-
-	si_context_draw(rctx, &rdraw);
-
-	rctx->flags |= R600_CONTEXT_DST_CACHES_DIRTY | R600_CONTEXT_DRAW_PENDING;
-
-	if (rctx->framebuffer.zsbuf)
-	{
-		struct pipe_resource *tex = rctx->framebuffer.zsbuf->texture;
-		((struct r600_resource_texture *)tex)->dirty_db = TRUE;
-	}
-
-	pipe_resource_reference(&ib.buffer, NULL);
-}
-
 void _r600_pipe_state_add_reg(struct r600_context *ctx,
 			      struct r600_pipe_state *state,
 			      uint32_t offset, uint32_t value,
diff --git a/src/gallium/drivers/radeonsi/radeonsi_pipe.c b/src/gallium/drivers/radeonsi/radeonsi_pipe.c
index 4e695c3ffdc..e13d9bd69ed 100644
--- a/src/gallium/drivers/radeonsi/radeonsi_pipe.c
+++ b/src/gallium/drivers/radeonsi/radeonsi_pipe.c
@@ -213,7 +213,6 @@ static struct pipe_context *r600_create_context(struct pipe_screen *screen, void
 	r600_init_query_functions(rctx);
 	r600_init_context_resource_functions(rctx);
 	r600_init_surface_functions(rctx);
-	rctx->context.draw_vbo = r600_draw_vbo;
 
 	rctx->context.create_video_decoder = vl_create_decoder;
 	rctx->context.create_video_buffer = vl_video_buffer_create;
diff --git a/src/gallium/drivers/radeonsi/radeonsi_pipe.h b/src/gallium/drivers/radeonsi/radeonsi_pipe.h
index f78d8ef3af1..a0dcc411664 100644
--- a/src/gallium/drivers/radeonsi/radeonsi_pipe.h
+++ b/src/gallium/drivers/radeonsi/radeonsi_pipe.h
@@ -163,9 +163,8 @@ struct r600_context {
 	/* for saving when using blitter */
 	struct pipe_stencil_ref		stencil_ref;
 	struct r600_pipe_state		config;
-	struct si_pipe_shader 	*ps_shader;
-	struct si_pipe_shader 	*vs_shader;
-	struct r600_pipe_state		vs_user_data;
+	struct si_pipe_shader		*ps_shader;
+	struct si_pipe_shader		*vs_shader;
 	struct pipe_query		*current_render_cond;
 	unsigned			current_render_cond_mode;
 	struct pipe_query		*saved_render_cond;
@@ -253,8 +252,6 @@ static INLINE void r600_atom_dirty(struct r600_context *rctx, struct r600_atom *
 
 /* evergreen_state.c */
 void cayman_init_state_functions(struct r600_context *rctx);
-void si_pipe_shader_ps(struct pipe_context *ctx, struct si_pipe_shader *shader);
-void si_pipe_shader_vs(struct pipe_context *ctx, struct si_pipe_shader *shader);
 uint32_t si_translate_vertexformat(struct pipe_screen *screen,
 				   enum pipe_format format,
 				   const struct util_format_description *desc,
@@ -339,7 +336,6 @@ void r600_set_so_targets(struct pipe_context *ctx,
 			 unsigned num_targets,
 			 struct pipe_stream_output_target **targets,
 			 unsigned append_bitmask);
-void r600_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *info);
 
 /*
  * common helpers
diff --git a/src/gallium/drivers/radeonsi/si_state.c b/src/gallium/drivers/radeonsi/si_state.c
index 32e59515b0e..034f46187af 100644
--- a/src/gallium/drivers/radeonsi/si_state.c
+++ b/src/gallium/drivers/radeonsi/si_state.c
@@ -26,6 +26,7 @@
 
 #include "util/u_memory.h"
 #include "util/u_framebuffer.h"
+#include "util/u_blitter.h"
 #include "tgsi/tgsi_parse.h"
 #include "radeonsi_pipe.h"
 #include "si_state.h"
@@ -1362,173 +1363,6 @@ static void si_delete_ps_shader(struct pipe_context *ctx, void *state)
 	free(shader);
 }
 
-void si_pipe_shader_vs(struct pipe_context *ctx, struct si_pipe_shader *shader)
-{
-	struct r600_context *rctx = (struct r600_context *)ctx;
-	struct si_pm4_state *pm4;
-	unsigned num_sgprs, num_user_sgprs;
-	unsigned nparams, i;
-	uint64_t va;
-
-	if (si_pipe_shader_create(ctx, shader))
-		return;
-
-	si_pm4_delete_state(rctx, vs, shader->pm4);
-	pm4 = shader->pm4 = CALLOC_STRUCT(si_pm4_state);
-
-	si_pm4_inval_shader_cache(pm4);
-
-	/* Certain attributes (position, psize, etc.) don't count as params.
-	 * VS is required to export at least one param and r600_shader_from_tgsi()
-	 * takes care of adding a dummy export.
-	 */
-	for (nparams = 0, i = 0 ; i < shader->shader.noutput; i++) {
-		if (shader->shader.output[i].name != TGSI_SEMANTIC_POSITION)
-			nparams++;
-	}
-	if (nparams < 1)
-		nparams = 1;
-
-	si_pm4_set_reg(pm4, R_0286C4_SPI_VS_OUT_CONFIG,
-		       S_0286C4_VS_EXPORT_COUNT(nparams - 1));
-
-	si_pm4_set_reg(pm4, R_02870C_SPI_SHADER_POS_FORMAT,
-		       S_02870C_POS0_EXPORT_FORMAT(V_02870C_SPI_SHADER_4COMP) |
-		       S_02870C_POS1_EXPORT_FORMAT(V_02870C_SPI_SHADER_NONE) |
-		       S_02870C_POS2_EXPORT_FORMAT(V_02870C_SPI_SHADER_NONE) |
-		       S_02870C_POS3_EXPORT_FORMAT(V_02870C_SPI_SHADER_NONE));
-
-	va = r600_resource_va(ctx->screen, (void *)shader->bo);
-	si_pm4_add_bo(pm4, shader->bo, RADEON_USAGE_READ);
-	si_pm4_set_reg(pm4, R_00B120_SPI_SHADER_PGM_LO_VS, va >> 8);
-	si_pm4_set_reg(pm4, R_00B124_SPI_SHADER_PGM_HI_VS, va >> 40);
-
-	num_user_sgprs = 8;
-	num_sgprs = shader->num_sgprs;
-	if (num_user_sgprs > num_sgprs)
-		num_sgprs = num_user_sgprs;
-	/* Last 2 reserved SGPRs are used for VCC */
-	num_sgprs += 2;
-	assert(num_sgprs <= 104);
-
-	si_pm4_set_reg(pm4, R_00B128_SPI_SHADER_PGM_RSRC1_VS,
-		       S_00B128_VGPRS((shader->num_vgprs - 1) / 4) |
-		       S_00B128_SGPRS((num_sgprs - 1) / 8));
-	si_pm4_set_reg(pm4, R_00B12C_SPI_SHADER_PGM_RSRC2_VS,
-		       S_00B12C_USER_SGPR(num_user_sgprs));
-
-	si_pm4_bind_state(rctx, vs, shader->pm4);
-}
-
-void si_pipe_shader_ps(struct pipe_context *ctx, struct si_pipe_shader *shader)
-{
-	struct r600_context *rctx = (struct r600_context *)ctx;
-	struct si_pm4_state *pm4;
-	unsigned i, exports_ps, num_cout, spi_ps_in_control, db_shader_control;
-	unsigned num_sgprs, num_user_sgprs;
-	int ninterp = 0;
-	boolean have_linear = FALSE, have_centroid = FALSE, have_perspective = FALSE;
-	unsigned spi_baryc_cntl;
-	uint64_t va;
-
-	if (si_pipe_shader_create(ctx, shader))
-		return;
-
-	si_pm4_delete_state(rctx, ps, shader->pm4);
-	pm4 = shader->pm4 = CALLOC_STRUCT(si_pm4_state);
-
-	si_pm4_inval_shader_cache(pm4);
-
-	db_shader_control = S_02880C_Z_ORDER(V_02880C_EARLY_Z_THEN_LATE_Z);
-	for (i = 0; i < shader->shader.ninput; i++) {
-		ninterp++;
-		/* XXX: Flat shading hangs the GPU */
-		if (shader->shader.input[i].interpolate == TGSI_INTERPOLATE_CONSTANT ||
-		    (shader->shader.input[i].interpolate == TGSI_INTERPOLATE_COLOR &&
-		     rctx->queued.named.rasterizer->flatshade))
-			have_linear = TRUE;
-		if (shader->shader.input[i].interpolate == TGSI_INTERPOLATE_LINEAR)
-			have_linear = TRUE;
-		if (shader->shader.input[i].interpolate == TGSI_INTERPOLATE_PERSPECTIVE)
-			have_perspective = TRUE;
-		if (shader->shader.input[i].centroid)
-			have_centroid = TRUE;
-	}
-
-	for (i = 0; i < shader->shader.noutput; i++) {
-		if (shader->shader.output[i].name == TGSI_SEMANTIC_POSITION)
-			db_shader_control |= S_02880C_Z_EXPORT_ENABLE(1);
-		if (shader->shader.output[i].name == TGSI_SEMANTIC_STENCIL)
-			db_shader_control |= 0; // XXX OP_VAL or TEST_VAL?
-	}
-	if (shader->shader.uses_kill)
-		db_shader_control |= S_02880C_KILL_ENABLE(1);
-
-	exports_ps = 0;
-	num_cout = 0;
-	for (i = 0; i < shader->shader.noutput; i++) {
-		if (shader->shader.output[i].name == TGSI_SEMANTIC_POSITION ||
-		    shader->shader.output[i].name == TGSI_SEMANTIC_STENCIL)
-			exports_ps |= 1;
-		else if (shader->shader.output[i].name == TGSI_SEMANTIC_COLOR) {
-			if (shader->shader.fs_write_all)
-				num_cout = shader->shader.nr_cbufs;
-			else
-				num_cout++;
-		}
-	}
-	if (!exports_ps) {
-		/* always at least export 1 component per pixel */
-		exports_ps = 2;
-	}
-
-	spi_ps_in_control = S_0286D8_NUM_INTERP(ninterp);
-
-	spi_baryc_cntl = 0;
-	if (have_perspective)
-		spi_baryc_cntl |= have_centroid ?
-			S_0286E0_PERSP_CENTROID_CNTL(1) : S_0286E0_PERSP_CENTER_CNTL(1);
-	if (have_linear)
-		spi_baryc_cntl |= have_centroid ?
-			S_0286E0_LINEAR_CENTROID_CNTL(1) : S_0286E0_LINEAR_CENTER_CNTL(1);
-
-	si_pm4_set_reg(pm4, R_0286E0_SPI_BARYC_CNTL, spi_baryc_cntl);
-	si_pm4_set_reg(pm4, R_0286CC_SPI_PS_INPUT_ENA, shader->spi_ps_input_ena);
-	si_pm4_set_reg(pm4, R_0286D0_SPI_PS_INPUT_ADDR, shader->spi_ps_input_ena);
-	si_pm4_set_reg(pm4, R_0286D8_SPI_PS_IN_CONTROL, spi_ps_in_control);
-
-	/* XXX: Depends on Z buffer format? */
-	si_pm4_set_reg(pm4, R_028710_SPI_SHADER_Z_FORMAT, 0);
-
-	/* XXX: Depends on color buffer format? */
-	si_pm4_set_reg(pm4, R_028714_SPI_SHADER_COL_FORMAT,
-		       S_028714_COL0_EXPORT_FORMAT(V_028714_SPI_SHADER_32_ABGR));
-
-	va = r600_resource_va(ctx->screen, (void *)shader->bo);
-	si_pm4_add_bo(pm4, shader->bo, RADEON_USAGE_READ);
-	si_pm4_set_reg(pm4, R_00B020_SPI_SHADER_PGM_LO_PS, va >> 8);
-	si_pm4_set_reg(pm4, R_00B024_SPI_SHADER_PGM_HI_PS, va >> 40);
-
-	num_user_sgprs = 6;
-	num_sgprs = shader->num_sgprs;
-	if (num_user_sgprs > num_sgprs)
-		num_sgprs = num_user_sgprs;
-	/* Last 2 reserved SGPRs are used for VCC */
-	num_sgprs += 2;
-	assert(num_sgprs <= 104);
-
-	si_pm4_set_reg(pm4, R_00B028_SPI_SHADER_PGM_RSRC1_PS,
-		       S_00B028_VGPRS((shader->num_vgprs - 1) / 4) |
-		       S_00B028_SGPRS((num_sgprs - 1) / 8));
-	si_pm4_set_reg(pm4, R_00B02C_SPI_SHADER_PGM_RSRC2_PS,
-		       S_00B02C_USER_SGPR(num_user_sgprs));
-
-	si_pm4_set_reg(pm4, R_02880C_DB_SHADER_CONTROL, db_shader_control);
-
-	shader->sprite_coord_enable = rctx->sprite_coord_enable;
-	si_pm4_bind_state(rctx, ps, shader->pm4);
-}
-
 /*
  * Samplers
  */
@@ -1562,12 +1396,16 @@ static void si_set_ps_sampler_view(struct pipe_context *ctx, unsigned count,
 	ptr = rctx->ws->buffer_map(bo->cs_buf, rctx->cs, PIPE_TRANSFER_WRITE);
 
 	for (i = 0; i < count; i++, ptr += sizeof(resource[0]->state)) {
+		struct r600_resource_texture *tex = (void *)resource[i]->base.texture;
+
 		pipe_sampler_view_reference(
 			(struct pipe_sampler_view **)&rctx->ps_samplers.views[i],
 			views[i]);
 
+		si_pm4_add_bo(pm4, &tex->resource, RADEON_USAGE_READ);
+
 		if (resource[i]) {
-			if (((struct r600_resource_texture *)resource[i]->base.texture)->depth)
+			if (tex->depth)
 				has_depth = 1;
 
 			memcpy(ptr, resource[i]->state, sizeof(resource[0]->state));
@@ -1726,6 +1564,8 @@ void si_init_state_functions(struct r600_context *rctx)
 	rctx->context.set_fragment_sampler_views = si_set_ps_sampler_view;
 
 	rctx->context.set_constant_buffer = si_set_constant_buffer;
+
+	rctx->context.draw_vbo = si_draw_vbo;
 }
 
 void si_init_config(struct r600_context *rctx)
@@ -1767,125 +1607,3 @@ void si_init_config(struct r600_context *rctx)
 
 	si_pm4_set_state(rctx, init, pm4);
 }
-
-static unsigned si_conv_pipe_prim(unsigned pprim)
-{
-        static const unsigned prim_conv[] = {
-		[PIPE_PRIM_POINTS]			= V_008958_DI_PT_POINTLIST,
-		[PIPE_PRIM_LINES]			= V_008958_DI_PT_LINELIST,
-		[PIPE_PRIM_LINE_LOOP]			= V_008958_DI_PT_LINELOOP,
-		[PIPE_PRIM_LINE_STRIP]			= V_008958_DI_PT_LINESTRIP,
-		[PIPE_PRIM_TRIANGLES]			= V_008958_DI_PT_TRILIST,
-		[PIPE_PRIM_TRIANGLE_STRIP]		= V_008958_DI_PT_TRISTRIP,
-		[PIPE_PRIM_TRIANGLE_FAN]		= V_008958_DI_PT_TRIFAN,
-		[PIPE_PRIM_QUADS]			= V_008958_DI_PT_QUADLIST,
-		[PIPE_PRIM_QUAD_STRIP]			= V_008958_DI_PT_QUADSTRIP,
-		[PIPE_PRIM_POLYGON]			= V_008958_DI_PT_POLYGON,
-		[PIPE_PRIM_LINES_ADJACENCY]		= ~0,
-		[PIPE_PRIM_LINE_STRIP_ADJACENCY]	= ~0,
-		[PIPE_PRIM_TRIANGLES_ADJACENCY]		= ~0,
-		[PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY]	= ~0
-        };
-	unsigned result = prim_conv[pprim];
-        if (result == ~0) {
-		R600_ERR("unsupported primitive type %d\n", pprim);
-        }
-	return result;
-}
-
-bool si_update_draw_info_state(struct r600_context *rctx,
-			       const struct pipe_draw_info *info)
-{
-	struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state);
-	unsigned prim = si_conv_pipe_prim(info->mode);
-	unsigned ls_mask = 0;
-
-	if (pm4 == NULL)
-		return false;
-
-	if (prim == ~0) {
-		FREE(pm4);
-		return false;
-	}
-
-	si_pm4_set_reg(pm4, R_008958_VGT_PRIMITIVE_TYPE, prim);
-	si_pm4_set_reg(pm4, R_028400_VGT_MAX_VTX_INDX, ~0);
-	si_pm4_set_reg(pm4, R_028404_VGT_MIN_VTX_INDX, 0);
-	si_pm4_set_reg(pm4, R_028408_VGT_INDX_OFFSET, info->index_bias);
-	si_pm4_set_reg(pm4, R_02840C_VGT_MULTI_PRIM_IB_RESET_INDX, info->restart_index);
-	si_pm4_set_reg(pm4, R_028A94_VGT_MULTI_PRIM_IB_RESET_EN, info->primitive_restart);
-#if 0
-	si_pm4_set_reg(pm4, R_03CFF0_SQ_VTX_BASE_VTX_LOC, 0);
-	si_pm4_set_reg(pm4, R_03CFF4_SQ_VTX_START_INST_LOC, info->start_instance);
-#endif
-
-        if (prim == V_008958_DI_PT_LINELIST)
-                ls_mask = 1;
-        else if (prim == V_008958_DI_PT_LINESTRIP)
-                ls_mask = 2;
-	si_pm4_set_reg(pm4, R_028A0C_PA_SC_LINE_STIPPLE,
-		       S_028A0C_AUTO_RESET_CNTL(ls_mask) |
-		       rctx->pa_sc_line_stipple);
-
-        if (info->mode == PIPE_PRIM_QUADS || info->mode == PIPE_PRIM_QUAD_STRIP || info->mode == PIPE_PRIM_POLYGON) {
-		si_pm4_set_reg(pm4, R_028814_PA_SU_SC_MODE_CNTL,
-			       S_028814_PROVOKING_VTX_LAST(1) | rctx->pa_su_sc_mode_cntl);
-        } else {
-		si_pm4_set_reg(pm4, R_028814_PA_SU_SC_MODE_CNTL, rctx->pa_su_sc_mode_cntl);
-        }
-	si_pm4_set_reg(pm4, R_02881C_PA_CL_VS_OUT_CNTL,
-		       prim == PIPE_PRIM_POINTS ? rctx->pa_cl_vs_out_cntl : 0
-		       /*| (rctx->rasterizer->clip_plane_enable &
-		       rctx->vs_shader->shader.clip_dist_write)*/);
-	si_pm4_set_reg(pm4, R_028810_PA_CL_CLIP_CNTL, rctx->pa_cl_clip_cntl
-			/*| (rctx->vs_shader->shader.clip_dist_write ||
-			rctx->vs_shader->shader.vs_prohibit_ucps ?
-			0 : rctx->rasterizer->clip_plane_enable & 0x3F)*/);
-
-	si_pm4_set_state(rctx, draw_info, pm4);
-	return true;
-}
-
-void si_update_spi_map(struct r600_context *rctx)
-{
-	struct si_shader *ps = &rctx->ps_shader->shader;
-	struct si_shader *vs = &rctx->vs_shader->shader;
-	struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state);
-	unsigned i, j, tmp;
-
-	for (i = 0; i < ps->ninput; i++) {
-		tmp = 0;
-
-#if 0
-		/* XXX: Flat shading hangs the GPU */
-		if (ps->input[i].name == TGSI_SEMANTIC_POSITION ||
-		    ps->input[i].interpolate == TGSI_INTERPOLATE_CONSTANT ||
-		    (ps->input[i].interpolate == TGSI_INTERPOLATE_COLOR &&
-		     rctx->rasterizer && rctx->rasterizer->flatshade)) {
-			tmp |= S_028644_FLAT_SHADE(1);
-		}
-#endif
-
-		if (ps->input[i].name == TGSI_SEMANTIC_GENERIC &&
-		    rctx->sprite_coord_enable & (1 << ps->input[i].sid)) {
-			tmp |= S_028644_PT_SPRITE_TEX(1);
-		}
-
-		for (j = 0; j < vs->noutput; j++) {
-			if (ps->input[i].name == vs->output[j].name &&
-			    ps->input[i].sid == vs->output[j].sid) {
-				tmp |= S_028644_OFFSET(vs->output[j].param_offset);
-				break;
-			}
-		}
-
-		if (j == vs->noutput) {
-			/* No corresponding output found, load defaults into input */
-			tmp |= S_028644_OFFSET(0x20);
-		}
-
-		si_pm4_set_reg(pm4, R_028644_SPI_PS_INPUT_CNTL_0 + i * 4, tmp);
-	}
-
-	si_pm4_set_state(rctx, spi, pm4);
-}
diff --git a/src/gallium/drivers/radeonsi/si_state.h b/src/gallium/drivers/radeonsi/si_state.h
index f0ddbc6ca06..cb993774f4c 100644
--- a/src/gallium/drivers/radeonsi/si_state.h
+++ b/src/gallium/drivers/radeonsi/si_state.h
@@ -122,6 +122,7 @@ union si_state {
 		struct si_pm4_state		*ps_sampler;
 		struct si_pm4_state		*ps_const;
 		struct si_pm4_state		*spi;
+		struct si_pm4_state		*vertex_buffers;
 		struct si_pm4_state		*draw_info;
 	} named;
 	struct si_pm4_state	*array[0];
@@ -154,10 +155,11 @@ union si_state {
 		} \
 	} while(0);
 
+/* si_state.c */
 void si_init_state_functions(struct r600_context *rctx);
 void si_init_config(struct r600_context *rctx);
-bool si_update_draw_info_state(struct r600_context *rctx,
-			       const struct pipe_draw_info *info);
-void si_update_spi_map(struct r600_context *rctx);
+
+/* si_state_draw.c */
+void si_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *dinfo);
 
 #endif
diff --git a/src/gallium/drivers/radeonsi/si_state_draw.c b/src/gallium/drivers/radeonsi/si_state_draw.c
new file mode 100644
index 00000000000..211956ef787
--- /dev/null
+++ b/src/gallium/drivers/radeonsi/si_state_draw.c
@@ -0,0 +1,572 @@
+/*
+ * Copyright 2012 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ *      Christian König <[email protected]>
+ */
+
+#include "util/u_memory.h"
+#include "util/u_framebuffer.h"
+#include "util/u_blitter.h"
+#include "tgsi/tgsi_parse.h"
+#include "radeonsi_pipe.h"
+#include "si_state.h"
+#include "sid.h"
+
+/*
+ * Shaders
+ */
+
+static void si_pipe_shader_vs(struct pipe_context *ctx, struct si_pipe_shader *shader)
+{
+	struct r600_context *rctx = (struct r600_context *)ctx;
+	struct si_pm4_state *pm4;
+	unsigned num_sgprs, num_user_sgprs;
+	unsigned nparams, i;
+	uint64_t va;
+
+	if (si_pipe_shader_create(ctx, shader))
+		return;
+
+	si_pm4_delete_state(rctx, vs, shader->pm4);
+	pm4 = shader->pm4 = CALLOC_STRUCT(si_pm4_state);
+
+	si_pm4_inval_shader_cache(pm4);
+
+	/* Certain attributes (position, psize, etc.) don't count as params.
+	 * VS is required to export at least one param and r600_shader_from_tgsi()
+	 * takes care of adding a dummy export.
+	 */
+	for (nparams = 0, i = 0 ; i < shader->shader.noutput; i++) {
+		if (shader->shader.output[i].name != TGSI_SEMANTIC_POSITION)
+			nparams++;
+	}
+	if (nparams < 1)
+		nparams = 1;
+
+	si_pm4_set_reg(pm4, R_0286C4_SPI_VS_OUT_CONFIG,
+		       S_0286C4_VS_EXPORT_COUNT(nparams - 1));
+
+	si_pm4_set_reg(pm4, R_02870C_SPI_SHADER_POS_FORMAT,
+		       S_02870C_POS0_EXPORT_FORMAT(V_02870C_SPI_SHADER_4COMP) |
+		       S_02870C_POS1_EXPORT_FORMAT(V_02870C_SPI_SHADER_NONE) |
+		       S_02870C_POS2_EXPORT_FORMAT(V_02870C_SPI_SHADER_NONE) |
+		       S_02870C_POS3_EXPORT_FORMAT(V_02870C_SPI_SHADER_NONE));
+
+	va = r600_resource_va(ctx->screen, (void *)shader->bo);
+	si_pm4_add_bo(pm4, shader->bo, RADEON_USAGE_READ);
+	si_pm4_set_reg(pm4, R_00B120_SPI_SHADER_PGM_LO_VS, va >> 8);
+	si_pm4_set_reg(pm4, R_00B124_SPI_SHADER_PGM_HI_VS, va >> 40);
+
+	num_user_sgprs = 8;
+	num_sgprs = shader->num_sgprs;
+	if (num_user_sgprs > num_sgprs)
+		num_sgprs = num_user_sgprs;
+	/* Last 2 reserved SGPRs are used for VCC */
+	num_sgprs += 2;
+	assert(num_sgprs <= 104);
+
+	si_pm4_set_reg(pm4, R_00B128_SPI_SHADER_PGM_RSRC1_VS,
+		       S_00B128_VGPRS((shader->num_vgprs - 1) / 4) |
+		       S_00B128_SGPRS((num_sgprs - 1) / 8));
+	si_pm4_set_reg(pm4, R_00B12C_SPI_SHADER_PGM_RSRC2_VS,
+		       S_00B12C_USER_SGPR(num_user_sgprs));
+
+	si_pm4_bind_state(rctx, vs, shader->pm4);
+}
+
+static void si_pipe_shader_ps(struct pipe_context *ctx, struct si_pipe_shader *shader)
+{
+	struct r600_context *rctx = (struct r600_context *)ctx;
+	struct si_pm4_state *pm4;
+	unsigned i, exports_ps, num_cout, spi_ps_in_control, db_shader_control;
+	unsigned num_sgprs, num_user_sgprs;
+	int ninterp = 0;
+	boolean have_linear = FALSE, have_centroid = FALSE, have_perspective = FALSE;
+	unsigned spi_baryc_cntl;
+	uint64_t va;
+
+	if (si_pipe_shader_create(ctx, shader))
+		return;
+
+	si_pm4_delete_state(rctx, ps, shader->pm4);
+	pm4 = shader->pm4 = CALLOC_STRUCT(si_pm4_state);
+
+	si_pm4_inval_shader_cache(pm4);
+
+	db_shader_control = S_02880C_Z_ORDER(V_02880C_EARLY_Z_THEN_LATE_Z);
+	for (i = 0; i < shader->shader.ninput; i++) {
+		ninterp++;
+		/* XXX: Flat shading hangs the GPU */
+		if (shader->shader.input[i].interpolate == TGSI_INTERPOLATE_CONSTANT ||
+		    (shader->shader.input[i].interpolate == TGSI_INTERPOLATE_COLOR &&
+		     rctx->queued.named.rasterizer->flatshade))
+			have_linear = TRUE;
+		if (shader->shader.input[i].interpolate == TGSI_INTERPOLATE_LINEAR)
+			have_linear = TRUE;
+		if (shader->shader.input[i].interpolate == TGSI_INTERPOLATE_PERSPECTIVE)
+			have_perspective = TRUE;
+		if (shader->shader.input[i].centroid)
+			have_centroid = TRUE;
+	}
+
+	for (i = 0; i < shader->shader.noutput; i++) {
+		if (shader->shader.output[i].name == TGSI_SEMANTIC_POSITION)
+			db_shader_control |= S_02880C_Z_EXPORT_ENABLE(1);
+		if (shader->shader.output[i].name == TGSI_SEMANTIC_STENCIL)
+			db_shader_control |= 0; // XXX OP_VAL or TEST_VAL?
+	}
+	if (shader->shader.uses_kill)
+		db_shader_control |= S_02880C_KILL_ENABLE(1);
+
+	exports_ps = 0;
+	num_cout = 0;
+	for (i = 0; i < shader->shader.noutput; i++) {
+		if (shader->shader.output[i].name == TGSI_SEMANTIC_POSITION ||
+		    shader->shader.output[i].name == TGSI_SEMANTIC_STENCIL)
+			exports_ps |= 1;
+		else if (shader->shader.output[i].name == TGSI_SEMANTIC_COLOR) {
+			if (shader->shader.fs_write_all)
+				num_cout = shader->shader.nr_cbufs;
+			else
+				num_cout++;
+		}
+	}
+	if (!exports_ps) {
+		/* always at least export 1 component per pixel */
+		exports_ps = 2;
+	}
+
+	spi_ps_in_control = S_0286D8_NUM_INTERP(ninterp);
+
+	spi_baryc_cntl = 0;
+	if (have_perspective)
+		spi_baryc_cntl |= have_centroid ?
+			S_0286E0_PERSP_CENTROID_CNTL(1) : S_0286E0_PERSP_CENTER_CNTL(1);
+	if (have_linear)
+		spi_baryc_cntl |= have_centroid ?
+			S_0286E0_LINEAR_CENTROID_CNTL(1) : S_0286E0_LINEAR_CENTER_CNTL(1);
+
+	si_pm4_set_reg(pm4, R_0286E0_SPI_BARYC_CNTL, spi_baryc_cntl);
+	si_pm4_set_reg(pm4, R_0286CC_SPI_PS_INPUT_ENA, shader->spi_ps_input_ena);
+	si_pm4_set_reg(pm4, R_0286D0_SPI_PS_INPUT_ADDR, shader->spi_ps_input_ena);
+	si_pm4_set_reg(pm4, R_0286D8_SPI_PS_IN_CONTROL, spi_ps_in_control);
+
+	/* XXX: Depends on Z buffer format? */
+	si_pm4_set_reg(pm4, R_028710_SPI_SHADER_Z_FORMAT, 0);
+
+	/* XXX: Depends on color buffer format? */
+	si_pm4_set_reg(pm4, R_028714_SPI_SHADER_COL_FORMAT,
+		       S_028714_COL0_EXPORT_FORMAT(V_028714_SPI_SHADER_32_ABGR));
+
+	va = r600_resource_va(ctx->screen, (void *)shader->bo);
+	si_pm4_add_bo(pm4, shader->bo, RADEON_USAGE_READ);
+	si_pm4_set_reg(pm4, R_00B020_SPI_SHADER_PGM_LO_PS, va >> 8);
+	si_pm4_set_reg(pm4, R_00B024_SPI_SHADER_PGM_HI_PS, va >> 40);
+
+	num_user_sgprs = 6;
+	num_sgprs = shader->num_sgprs;
+	if (num_user_sgprs > num_sgprs)
+		num_sgprs = num_user_sgprs;
+	/* Last 2 reserved SGPRs are used for VCC */
+	num_sgprs += 2;
+	assert(num_sgprs <= 104);
+
+	si_pm4_set_reg(pm4, R_00B028_SPI_SHADER_PGM_RSRC1_PS,
+		       S_00B028_VGPRS((shader->num_vgprs - 1) / 4) |
+		       S_00B028_SGPRS((num_sgprs - 1) / 8));
+	si_pm4_set_reg(pm4, R_00B02C_SPI_SHADER_PGM_RSRC2_PS,
+		       S_00B02C_USER_SGPR(num_user_sgprs));
+
+	si_pm4_set_reg(pm4, R_02880C_DB_SHADER_CONTROL, db_shader_control);
+
+	shader->sprite_coord_enable = rctx->sprite_coord_enable;
+	si_pm4_bind_state(rctx, ps, shader->pm4);
+}
+
+/*
+ * Drawing
+ */
+
+static unsigned si_conv_pipe_prim(unsigned pprim)
+{
+        static const unsigned prim_conv[] = {
+		[PIPE_PRIM_POINTS]			= V_008958_DI_PT_POINTLIST,
+		[PIPE_PRIM_LINES]			= V_008958_DI_PT_LINELIST,
+		[PIPE_PRIM_LINE_LOOP]			= V_008958_DI_PT_LINELOOP,
+		[PIPE_PRIM_LINE_STRIP]			= V_008958_DI_PT_LINESTRIP,
+		[PIPE_PRIM_TRIANGLES]			= V_008958_DI_PT_TRILIST,
+		[PIPE_PRIM_TRIANGLE_STRIP]		= V_008958_DI_PT_TRISTRIP,
+		[PIPE_PRIM_TRIANGLE_FAN]		= V_008958_DI_PT_TRIFAN,
+		[PIPE_PRIM_QUADS]			= V_008958_DI_PT_QUADLIST,
+		[PIPE_PRIM_QUAD_STRIP]			= V_008958_DI_PT_QUADSTRIP,
+		[PIPE_PRIM_POLYGON]			= V_008958_DI_PT_POLYGON,
+		[PIPE_PRIM_LINES_ADJACENCY]		= ~0,
+		[PIPE_PRIM_LINE_STRIP_ADJACENCY]	= ~0,
+		[PIPE_PRIM_TRIANGLES_ADJACENCY]		= ~0,
+		[PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY]	= ~0
+        };
+	unsigned result = prim_conv[pprim];
+        if (result == ~0) {
+		R600_ERR("unsupported primitive type %d\n", pprim);
+        }
+	return result;
+}
+
+static bool si_update_draw_info_state(struct r600_context *rctx,
+			       const struct pipe_draw_info *info)
+{
+	struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state);
+	unsigned prim = si_conv_pipe_prim(info->mode);
+	unsigned ls_mask = 0;
+
+	if (pm4 == NULL)
+		return false;
+
+	if (prim == ~0) {
+		FREE(pm4);
+		return false;
+	}
+
+	si_pm4_set_reg(pm4, R_008958_VGT_PRIMITIVE_TYPE, prim);
+	si_pm4_set_reg(pm4, R_028400_VGT_MAX_VTX_INDX, ~0);
+	si_pm4_set_reg(pm4, R_028404_VGT_MIN_VTX_INDX, 0);
+	si_pm4_set_reg(pm4, R_028408_VGT_INDX_OFFSET, info->index_bias);
+	si_pm4_set_reg(pm4, R_02840C_VGT_MULTI_PRIM_IB_RESET_INDX, info->restart_index);
+	si_pm4_set_reg(pm4, R_028A94_VGT_MULTI_PRIM_IB_RESET_EN, info->primitive_restart);
+#if 0
+	si_pm4_set_reg(pm4, R_03CFF0_SQ_VTX_BASE_VTX_LOC, 0);
+	si_pm4_set_reg(pm4, R_03CFF4_SQ_VTX_START_INST_LOC, info->start_instance);
+#endif
+
+        if (prim == V_008958_DI_PT_LINELIST)
+                ls_mask = 1;
+        else if (prim == V_008958_DI_PT_LINESTRIP)
+                ls_mask = 2;
+	si_pm4_set_reg(pm4, R_028A0C_PA_SC_LINE_STIPPLE,
+		       S_028A0C_AUTO_RESET_CNTL(ls_mask) |
+		       rctx->pa_sc_line_stipple);
+
+        if (info->mode == PIPE_PRIM_QUADS || info->mode == PIPE_PRIM_QUAD_STRIP || info->mode == PIPE_PRIM_POLYGON) {
+		si_pm4_set_reg(pm4, R_028814_PA_SU_SC_MODE_CNTL,
+			       S_028814_PROVOKING_VTX_LAST(1) | rctx->pa_su_sc_mode_cntl);
+        } else {
+		si_pm4_set_reg(pm4, R_028814_PA_SU_SC_MODE_CNTL, rctx->pa_su_sc_mode_cntl);
+        }
+	si_pm4_set_reg(pm4, R_02881C_PA_CL_VS_OUT_CNTL,
+		       prim == PIPE_PRIM_POINTS ? rctx->pa_cl_vs_out_cntl : 0
+		       /*| (rctx->rasterizer->clip_plane_enable &
+		       rctx->vs_shader->shader.clip_dist_write)*/);
+	si_pm4_set_reg(pm4, R_028810_PA_CL_CLIP_CNTL, rctx->pa_cl_clip_cntl
+			/*| (rctx->vs_shader->shader.clip_dist_write ||
+			rctx->vs_shader->shader.vs_prohibit_ucps ?
+			0 : rctx->rasterizer->clip_plane_enable & 0x3F)*/);
+
+	si_pm4_set_state(rctx, draw_info, pm4);
+	return true;
+}
+
+static void si_update_alpha_ref(struct r600_context *rctx)
+{
+#if 0
+        unsigned alpha_ref;
+        struct r600_pipe_state rstate;
+
+        alpha_ref = rctx->alpha_ref;
+        rstate.nregs = 0;
+        if (rctx->export_16bpc)
+                alpha_ref &= ~0x1FFF;
+        si_pm4_set_reg(&rstate, R_028438_SX_ALPHA_REF, alpha_ref);
+
+	si_pm4_set_state(rctx, TODO, pm4);
+        rctx->alpha_ref_dirty = false;
+#endif
+}
+
+static void si_update_spi_map(struct r600_context *rctx)
+{
+	struct si_shader *ps = &rctx->ps_shader->shader;
+	struct si_shader *vs = &rctx->vs_shader->shader;
+	struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state);
+	unsigned i, j, tmp;
+
+	for (i = 0; i < ps->ninput; i++) {
+		tmp = 0;
+
+#if 0
+		/* XXX: Flat shading hangs the GPU */
+		if (ps->input[i].name == TGSI_SEMANTIC_POSITION ||
+		    ps->input[i].interpolate == TGSI_INTERPOLATE_CONSTANT ||
+		    (ps->input[i].interpolate == TGSI_INTERPOLATE_COLOR &&
+		     rctx->rasterizer && rctx->rasterizer->flatshade)) {
+			tmp |= S_028644_FLAT_SHADE(1);
+		}
+#endif
+
+		if (ps->input[i].name == TGSI_SEMANTIC_GENERIC &&
+		    rctx->sprite_coord_enable & (1 << ps->input[i].sid)) {
+			tmp |= S_028644_PT_SPRITE_TEX(1);
+		}
+
+		for (j = 0; j < vs->noutput; j++) {
+			if (ps->input[i].name == vs->output[j].name &&
+			    ps->input[i].sid == vs->output[j].sid) {
+				tmp |= S_028644_OFFSET(vs->output[j].param_offset);
+				break;
+			}
+		}
+
+		if (j == vs->noutput) {
+			/* No corresponding output found, load defaults into input */
+			tmp |= S_028644_OFFSET(0x20);
+		}
+
+		si_pm4_set_reg(pm4, R_028644_SPI_PS_INPUT_CNTL_0 + i * 4, tmp);
+	}
+
+	si_pm4_set_state(rctx, spi, pm4);
+}
+
+static void si_update_derived_state(struct r600_context *rctx)
+{
+	struct pipe_context * ctx = (struct pipe_context*)rctx;
+
+	if (!rctx->blitter->running) {
+		if (rctx->have_depth_fb || rctx->have_depth_texture)
+			r600_flush_depth_textures(rctx);
+	}
+
+	if ((rctx->ps_shader->shader.fs_write_all &&
+	     (rctx->ps_shader->shader.nr_cbufs != rctx->framebuffer.nr_cbufs)) ||
+	    (rctx->sprite_coord_enable &&
+	     (rctx->ps_shader->sprite_coord_enable != rctx->sprite_coord_enable))) {
+		si_pipe_shader_destroy(&rctx->context, rctx->ps_shader);
+	}
+
+	if (rctx->alpha_ref_dirty) {
+		si_update_alpha_ref(rctx);
+	}
+
+	if (!rctx->vs_shader->bo) {
+		si_pipe_shader_vs(ctx, rctx->vs_shader);
+	}
+
+	if (!rctx->ps_shader->bo) {
+		si_pipe_shader_ps(ctx, rctx->ps_shader);
+	}
+
+	if (rctx->shader_dirty) {
+		si_update_spi_map(rctx);
+		rctx->shader_dirty = false;
+	}
+}
+
+static void si_vertex_buffer_update(struct r600_context *rctx)
+{
+	struct pipe_context *ctx = &rctx->context;
+	struct r600_resource *rbuffer, *t_list_buffer;
+	struct pipe_vertex_buffer *vertex_buffer;
+	struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state);
+	unsigned i, count, offset;
+	uint32_t *ptr;
+	uint64_t va;
+
+	si_pm4_inval_vertex_cache(pm4);
+
+	/* bind vertex buffer once */
+	count = rctx->nr_vertex_buffers;
+	assert(count <= 256 / 4);
+
+	t_list_buffer = (struct r600_resource*)
+		pipe_buffer_create(ctx->screen, PIPE_BIND_CUSTOM,
+				   PIPE_USAGE_IMMUTABLE, 4 * 4 * count);
+	if (t_list_buffer == NULL) {
+		FREE(pm4);
+		return;
+	}
+
+	ptr = (uint32_t*)rctx->ws->buffer_map(t_list_buffer->cs_buf,
+					      rctx->cs,
+					      PIPE_TRANSFER_WRITE);
+
+	for (i = 0 ; i < count; i++, ptr += 4) {
+		struct pipe_vertex_element *velem = &rctx->vertex_elements->elements[i];
+		const struct util_format_description *desc;
+		unsigned data_format, num_format;
+		int first_non_void;
+
+		/* bind vertex buffer once */
+		vertex_buffer = &rctx->vertex_buffer[i];
+		rbuffer = (struct r600_resource*)vertex_buffer->buffer;
+		offset = 0;
+		if (vertex_buffer == NULL || rbuffer == NULL)
+			continue;
+		offset += vertex_buffer->buffer_offset;
+
+		va = r600_resource_va(ctx->screen, (void*)rbuffer);
+		va += offset;
+
+		desc = util_format_description(velem->src_format);
+		first_non_void = util_format_get_first_non_void_channel(velem->src_format);
+		data_format = si_translate_vertexformat(ctx->screen,
+							velem->src_format,
+							desc, first_non_void);
+
+		switch (desc->channel[first_non_void].type) {
+		case UTIL_FORMAT_TYPE_FIXED:
+			num_format = V_008F0C_BUF_NUM_FORMAT_USCALED; /* XXX */
+			break;
+		case UTIL_FORMAT_TYPE_SIGNED:
+			num_format = V_008F0C_BUF_NUM_FORMAT_SNORM;
+			break;
+		case UTIL_FORMAT_TYPE_UNSIGNED:
+			num_format = V_008F0C_BUF_NUM_FORMAT_UNORM;
+			break;
+		case UTIL_FORMAT_TYPE_FLOAT:
+		default:
+			num_format = V_008F14_IMG_NUM_FORMAT_FLOAT;
+		}
+
+		/* Fill in T# buffer resource description */
+		ptr[0] = va & 0xFFFFFFFF;
+		ptr[1] = (S_008F04_BASE_ADDRESS_HI(va >> 32) |
+			  S_008F04_STRIDE(vertex_buffer->stride));
+		if (vertex_buffer->stride > 0)
+			ptr[2] = ((vertex_buffer->buffer->width0 - offset) /
+				  vertex_buffer->stride);
+		else
+			ptr[2] = vertex_buffer->buffer->width0 - offset;
+		ptr[3] = (S_008F0C_DST_SEL_X(si_map_swizzle(desc->swizzle[0])) |
+			  S_008F0C_DST_SEL_Y(si_map_swizzle(desc->swizzle[1])) |
+			  S_008F0C_DST_SEL_Z(si_map_swizzle(desc->swizzle[2])) |
+			  S_008F0C_DST_SEL_W(si_map_swizzle(desc->swizzle[3])) |
+			  S_008F0C_NUM_FORMAT(num_format) |
+			  S_008F0C_DATA_FORMAT(data_format));
+
+		si_pm4_add_bo(pm4, rbuffer, RADEON_USAGE_READ);
+	}
+
+	va = r600_resource_va(ctx->screen, (void*)t_list_buffer);
+	si_pm4_add_bo(pm4, t_list_buffer, RADEON_USAGE_READ);
+	si_pm4_set_reg(pm4, R_00B148_SPI_SHADER_USER_DATA_VS_6, va);
+	si_pm4_set_reg(pm4, R_00B14C_SPI_SHADER_USER_DATA_VS_7, va >> 32);
+	si_pm4_set_state(rctx, vertex_buffers, pm4);
+}
+
+void si_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *dinfo)
+{
+	struct r600_context *rctx = (struct r600_context *)ctx;
+	struct si_state_dsa *dsa = rctx->queued.named.dsa;
+	struct pipe_draw_info info = *dinfo;
+	struct r600_draw rdraw = {};
+	struct pipe_index_buffer ib = {};
+	struct r600_block *dirty_block = NULL, *next_block = NULL;
+	struct r600_atom *state = NULL, *next_state = NULL;
+
+	if ((!info.count && (info.indexed || !info.count_from_stream_output)) ||
+	    (info.indexed && !rctx->index_buffer.buffer)) {
+		return;
+	}
+
+	if (!rctx->ps_shader || !rctx->vs_shader)
+		return;
+
+	si_update_derived_state(rctx);
+	si_vertex_buffer_update(rctx);
+
+	rdraw.vgt_num_indices = info.count;
+	rdraw.vgt_num_instances = info.instance_count;
+
+	if (info.indexed) {
+		/* Initialize the index buffer struct. */
+		pipe_resource_reference(&ib.buffer, rctx->index_buffer.buffer);
+		ib.index_size = rctx->index_buffer.index_size;
+		ib.offset = rctx->index_buffer.offset + info.start * ib.index_size;
+
+		/* Translate or upload, if needed. */
+		r600_translate_index_buffer(rctx, &ib, info.count);
+
+		if (ib.user_buffer) {
+			r600_upload_index_buffer(rctx, &ib, info.count);
+		}
+
+		/* Initialize the r600_draw struct with index buffer info. */
+		if (ib.index_size == 4) {
+			rdraw.vgt_index_type = V_028A7C_VGT_INDEX_32 |
+				(R600_BIG_ENDIAN ? V_028A7C_VGT_DMA_SWAP_32_BIT : 0);
+		} else {
+			rdraw.vgt_index_type = V_028A7C_VGT_INDEX_16 |
+				(R600_BIG_ENDIAN ? V_028A7C_VGT_DMA_SWAP_16_BIT : 0);
+		}
+		rdraw.indices = (struct r600_resource*)ib.buffer;
+		rdraw.indices_bo_offset = ib.offset;
+		rdraw.vgt_draw_initiator = V_0287F0_DI_SRC_SEL_DMA;
+	} else {
+		info.index_bias = info.start;
+		rdraw.vgt_draw_initiator = V_0287F0_DI_SRC_SEL_AUTO_INDEX;
+		if (info.count_from_stream_output) {
+			rdraw.vgt_draw_initiator |= S_0287F0_USE_OPAQUE(1);
+
+			r600_context_draw_opaque_count(rctx, (struct r600_so_target*)info.count_from_stream_output);
+		}
+	}
+
+	rctx->vs_shader_so_strides = rctx->vs_shader->so_strides;
+
+	if (!si_update_draw_info_state(rctx, &info))
+		return;
+
+	rdraw.db_render_override = dsa->db_render_override;
+	rdraw.db_render_control = dsa->db_render_control;
+
+	/* Emit states. */
+	rctx->pm4_dirty_cdwords += si_pm4_dirty_dw(rctx);
+
+	r600_need_cs_space(rctx, 0, TRUE);
+
+	LIST_FOR_EACH_ENTRY_SAFE(state, next_state, &rctx->dirty_states, head) {
+		r600_emit_atom(rctx, state);
+	}
+	LIST_FOR_EACH_ENTRY_SAFE(dirty_block, next_block, &rctx->dirty,list) {
+		r600_context_block_emit_dirty(rctx, dirty_block);
+	}
+	si_pm4_emit_dirty(rctx);
+	rctx->pm4_dirty_cdwords = 0;
+
+	/* Enable stream out if needed. */
+	if (rctx->streamout_start) {
+		r600_context_streamout_begin(rctx);
+		rctx->streamout_start = FALSE;
+	}
+
+	si_context_draw(rctx, &rdraw);
+
+	rctx->flags |= R600_CONTEXT_DST_CACHES_DIRTY | R600_CONTEXT_DRAW_PENDING;
+
+	if (rctx->framebuffer.zsbuf)
+	{
+		struct pipe_resource *tex = rctx->framebuffer.zsbuf->texture;
+		((struct r600_resource_texture *)tex)->dirty_db = TRUE;
+	}
+
+	pipe_resource_reference(&ib.buffer, NULL);
+}