Merge remote-tracking branch 'public/master' into vulkan

3 files changed, 233 insertions, 263 deletions

diff --git a/src/gallium/drivers/r600/evergreen_compute.c b/src/gallium/drivers/r600/evergreen_compute.c
index f4b669000dc..6f171487f92 100644
--- a/src/gallium/drivers/r600/evergreen_compute.c
+++ b/src/gallium/drivers/r600/evergreen_compute.c
@@ -83,29 +83,26 @@ writable images will consume TEX slots, VTX slots too because of linear indexing
 
 */
 
-struct r600_resource* r600_compute_buffer_alloc_vram(
-       struct r600_screen *screen,
-       unsigned size)
+struct r600_resource *r600_compute_buffer_alloc_vram(struct r600_screen *screen,
+						     unsigned size)
 {
-	struct pipe_resource * buffer = NULL;
+	struct pipe_resource *buffer = NULL;
 	assert(size);
 
-	buffer = pipe_buffer_create(
-		(struct pipe_screen*) screen,
-		PIPE_BIND_CUSTOM,
-		PIPE_USAGE_IMMUTABLE,
-		size);
+	buffer = pipe_buffer_create((struct pipe_screen*) screen,
+				    PIPE_BIND_CUSTOM,
+				    PIPE_USAGE_IMMUTABLE,
+				    size);
 
 	return (struct r600_resource *)buffer;
 }
 
 
-static void evergreen_set_rat(
-	struct r600_pipe_compute *pipe,
-	unsigned id,
-	struct r600_resource* bo,
-	int start,
-	int size)
+static void evergreen_set_rat(struct r600_pipe_compute *pipe,
+			      unsigned id,
+			      struct r600_resource *bo,
+			      int start,
+			      int size)
 {
 	struct pipe_surface rat_templ;
 	struct r600_surface *surf = NULL;
@@ -145,11 +142,10 @@ static void evergreen_set_rat(
 	evergreen_init_color_surface_rat(rctx, surf);
 }
 
-static void evergreen_cs_set_vertex_buffer(
-	struct r600_context * rctx,
-	unsigned vb_index,
-	unsigned offset,
-	struct pipe_resource * buffer)
+static void evergreen_cs_set_vertex_buffer(struct r600_context *rctx,
+					   unsigned vb_index,
+					   unsigned offset,
+					   struct pipe_resource *buffer)
 {
 	struct r600_vertexbuf_state *state = &rctx->cs_vertex_buffer_state;
 	struct pipe_vertex_buffer *vb = &state->vb[vb_index];
@@ -166,12 +162,11 @@ static void evergreen_cs_set_vertex_buffer(
 	r600_mark_atom_dirty(rctx, &state->atom);
 }
 
-static void evergreen_cs_set_constant_buffer(
-	struct r600_context * rctx,
-	unsigned cb_index,
-	unsigned offset,
-	unsigned size,
-	struct pipe_resource * buffer)
+static void evergreen_cs_set_constant_buffer(struct r600_context *rctx,
+					     unsigned cb_index,
+					     unsigned offset,
+					     unsigned size,
+					     struct pipe_resource *buffer)
 {
 	struct pipe_constant_buffer cb;
 	cb.buffer_size = size;
@@ -182,16 +177,6 @@ static void evergreen_cs_set_constant_buffer(
 	rctx->b.b.set_constant_buffer(&rctx->b.b, PIPE_SHADER_COMPUTE, cb_index, &cb);
 }
 
-static const struct u_resource_vtbl r600_global_buffer_vtbl =
-{
-	u_default_resource_get_handle, /* get_handle */
-	r600_compute_global_buffer_destroy, /* resource_destroy */
-	r600_compute_global_transfer_map, /* transfer_map */
-	r600_compute_global_transfer_flush_region,/* transfer_flush_region */
-	r600_compute_global_transfer_unmap, /* transfer_unmap */
-	r600_compute_global_transfer_inline_write /* transfer_inline_write */
-};
-
 /* We need to define these R600 registers here, because we can't include
  * evergreend.h and r600d.h.
  */
@@ -256,33 +241,32 @@ static void r600_destroy_shader(struct r600_bytecode *bc)
 	FREE(bc->bytecode);
 }
 
-void *evergreen_create_compute_state(
-	struct pipe_context *ctx_,
-	const const struct pipe_compute_state *cso)
+static void *evergreen_create_compute_state(struct pipe_context *ctx,
+					    const const struct pipe_compute_state *cso)
 {
-	struct r600_context *ctx = (struct r600_context *)ctx_;
+	struct r600_context *rctx = (struct r600_context *)ctx;
 	struct r600_pipe_compute *shader = CALLOC_STRUCT(r600_pipe_compute);
 #ifdef HAVE_OPENCL
-	const struct pipe_llvm_program_header * header;
+	const struct pipe_llvm_program_header *header;
 	const char *code;
 	void *p;
 	boolean use_kill;
 
-	COMPUTE_DBG(ctx->screen, "*** evergreen_create_compute_state\n");
+	COMPUTE_DBG(rctx->screen, "*** evergreen_create_compute_state\n");
 	header = cso->prog;
 	code = cso->prog + sizeof(struct pipe_llvm_program_header);
 	radeon_shader_binary_init(&shader->binary);
 	radeon_elf_read(code, header->num_bytes, &shader->binary);
 	r600_create_shader(&shader->bc, &shader->binary, &use_kill);
 
-	shader->code_bo = r600_compute_buffer_alloc_vram(ctx->screen,
+	shader->code_bo = r600_compute_buffer_alloc_vram(rctx->screen,
 							shader->bc.ndw * 4);
-	p = r600_buffer_map_sync_with_rings(&ctx->b, shader->code_bo, PIPE_TRANSFER_WRITE);
+	p = r600_buffer_map_sync_with_rings(&rctx->b, shader->code_bo, PIPE_TRANSFER_WRITE);
 	memcpy(p, shader->bc.bytecode, shader->bc.ndw * 4);
-	ctx->b.ws->buffer_unmap(shader->code_bo->buf);
+	rctx->b.ws->buffer_unmap(shader->code_bo->buf);
 #endif
 
-	shader->ctx = ctx;
+	shader->ctx = rctx;
 	shader->local_size = cso->req_local_mem;
 	shader->private_size = cso->req_private_mem;
 	shader->input_size = cso->req_input_mem;
@@ -290,12 +274,13 @@ void *evergreen_create_compute_state(
 	return shader;
 }
 
-void evergreen_delete_compute_state(struct pipe_context *ctx_, void* state)
+static void evergreen_delete_compute_state(struct pipe_context *ctx, void *state)
 {
-	struct r600_context *ctx = (struct r600_context *)ctx_;
-	COMPUTE_DBG(ctx->screen, "*** evergreen_delete_compute_state\n");
+	struct r600_context *rctx = (struct r600_context *)ctx;
 	struct r600_pipe_compute *shader = state;
 
+	COMPUTE_DBG(rctx->screen, "*** evergreen_delete_compute_state\n");
+
 	if (!shader)
 		return;
 
@@ -307,13 +292,13 @@ void evergreen_delete_compute_state(struct pipe_context *ctx_, void* state)
 	FREE(shader);
 }
 
-static void evergreen_bind_compute_state(struct pipe_context *ctx_, void *state)
+static void evergreen_bind_compute_state(struct pipe_context *ctx, void *state)
 {
-	struct r600_context *ctx = (struct r600_context *)ctx_;
+	struct r600_context *rctx = (struct r600_context *)ctx;
 
-	COMPUTE_DBG(ctx->screen, "*** evergreen_bind_compute_state\n");
+	COMPUTE_DBG(rctx->screen, "*** evergreen_bind_compute_state\n");
 
-	ctx->cs_shader_state.shader = (struct r600_pipe_compute *)state;
+	rctx->cs_shader_state.shader = (struct r600_pipe_compute *)state;
 }
 
 /* The kernel parameters are stored a vtx buffer (ID=0), besides the explicit
@@ -327,23 +312,20 @@ static void evergreen_bind_compute_state(struct pipe_context *ctx_, void *state)
  *             (x,y,z)
  * DWORDS 9+ : Kernel parameters
  */
-void evergreen_compute_upload_input(
-	struct pipe_context *ctx_,
-	const uint *block_layout,
-	const uint *grid_layout,
-	const void *input)
+static void evergreen_compute_upload_input(struct pipe_context *ctx,
+					   const struct pipe_grid_info *info)
 {
-	struct r600_context *ctx = (struct r600_context *)ctx_;
-	struct r600_pipe_compute *shader = ctx->cs_shader_state.shader;
+	struct r600_context *rctx = (struct r600_context *)ctx;
+	struct r600_pipe_compute *shader = rctx->cs_shader_state.shader;
 	unsigned i;
 	/* We need to reserve 9 dwords (36 bytes) for implicit kernel
 	 * parameters.
 	 */
 	unsigned input_size = shader->input_size + 36;
-	uint32_t * num_work_groups_start;
-	uint32_t * global_size_start;
-	uint32_t * local_size_start;
-	uint32_t * kernel_parameters_start;
+	uint32_t *num_work_groups_start;
+	uint32_t *global_size_start;
+	uint32_t *local_size_start;
+	uint32_t *kernel_parameters_start;
 	struct pipe_box box;
 	struct pipe_transfer *transfer = NULL;
 
@@ -354,12 +336,12 @@ void evergreen_compute_upload_input(
 	if (!shader->kernel_param) {
 		/* Add space for the grid dimensions */
 		shader->kernel_param = (struct r600_resource *)
-			pipe_buffer_create(ctx_->screen, PIPE_BIND_CUSTOM,
+			pipe_buffer_create(ctx->screen, PIPE_BIND_CUSTOM,
 					PIPE_USAGE_IMMUTABLE, input_size);
 	}
 
 	u_box_1d(0, input_size, &box);
-	num_work_groups_start = ctx_->transfer_map(ctx_,
+	num_work_groups_start = ctx->transfer_map(ctx,
 			(struct pipe_resource*)shader->kernel_param,
 			0, PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD_RANGE,
 			&box, &transfer);
@@ -368,34 +350,33 @@ void evergreen_compute_upload_input(
 	kernel_parameters_start = local_size_start + (3 * (sizeof(uint)) / 4);
 
 	/* Copy the work group size */
-	memcpy(num_work_groups_start, grid_layout, 3 * sizeof(uint));
+	memcpy(num_work_groups_start, info->grid, 3 * sizeof(uint));
 
 	/* Copy the global size */
 	for (i = 0; i < 3; i++) {
-		global_size_start[i] = grid_layout[i] * block_layout[i];
+		global_size_start[i] = info->grid[i] * info->block[i];
 	}
 
 	/* Copy the local dimensions */
-	memcpy(local_size_start, block_layout, 3 * sizeof(uint));
+	memcpy(local_size_start, info->block, 3 * sizeof(uint));
 
 	/* Copy the kernel inputs */
-	memcpy(kernel_parameters_start, input, shader->input_size);
+	memcpy(kernel_parameters_start, info->input, shader->input_size);
 
 	for (i = 0; i < (input_size / 4); i++) {
-		COMPUTE_DBG(ctx->screen, "input %i : %u\n", i,
+		COMPUTE_DBG(rctx->screen, "input %i : %u\n", i,
 			((unsigned*)num_work_groups_start)[i]);
 	}
 
-	ctx_->transfer_unmap(ctx_, transfer);
+	ctx->transfer_unmap(ctx, transfer);
 
 	/* ID=0 is reserved for the parameters */
-	evergreen_cs_set_constant_buffer(ctx, 0, 0, input_size,
+	evergreen_cs_set_constant_buffer(rctx, 0, 0, input_size,
 			(struct pipe_resource*)shader->kernel_param);
 }
 
-static void evergreen_emit_direct_dispatch(
-		struct r600_context *rctx,
-		const uint *block_layout, const uint *grid_layout)
+static void evergreen_emit_dispatch(struct r600_context *rctx,
+				    const struct pipe_grid_info *info)
 {
 	int i;
 	struct radeon_winsys_cs *cs = rctx->b.gfx.cs;
@@ -411,15 +392,15 @@ static void evergreen_emit_direct_dispatch(
 
 	/* Calculate group_size/grid_size */
 	for (i = 0; i < 3; i++) {
-		group_size *= block_layout[i];
+		group_size *= info->block[i];
 	}
 
 	for (i = 0; i < 3; i++)	{
-		grid_size *= grid_layout[i];
+		grid_size *= info->grid[i];
 	}
 
 	/* num_waves = ceil((tg_size.x * tg_size.y, tg_size.z) / (16 * num_pipes)) */
-	num_waves = (block_layout[0] * block_layout[1] * block_layout[2] +
+	num_waves = (info->block[0] * info->block[1] * info->block[2] +
 			wave_divisor - 1) / wave_divisor;
 
 	COMPUTE_DBG(rctx->screen, "Using %u pipes, "
@@ -438,9 +419,9 @@ static void evergreen_emit_direct_dispatch(
 								group_size);
 
 	radeon_compute_set_context_reg_seq(cs, R_0286EC_SPI_COMPUTE_NUM_THREAD_X, 3);
-	radeon_emit(cs, block_layout[0]); /* R_0286EC_SPI_COMPUTE_NUM_THREAD_X */
-	radeon_emit(cs, block_layout[1]); /* R_0286F0_SPI_COMPUTE_NUM_THREAD_Y */
-	radeon_emit(cs, block_layout[2]); /* R_0286F4_SPI_COMPUTE_NUM_THREAD_Z */
+	radeon_emit(cs, info->block[0]); /* R_0286EC_SPI_COMPUTE_NUM_THREAD_X */
+	radeon_emit(cs, info->block[1]); /* R_0286F0_SPI_COMPUTE_NUM_THREAD_Y */
+	radeon_emit(cs, info->block[2]); /* R_0286F4_SPI_COMPUTE_NUM_THREAD_Z */
 
 	if (rctx->b.chip_class < CAYMAN) {
 		assert(lds_size <= 8192);
@@ -455,22 +436,22 @@ static void evergreen_emit_direct_dispatch(
 
 	/* Dispatch packet */
 	radeon_emit(cs, PKT3C(PKT3_DISPATCH_DIRECT, 3, 0));
-	radeon_emit(cs, grid_layout[0]);
-	radeon_emit(cs, grid_layout[1]);
-	radeon_emit(cs, grid_layout[2]);
+	radeon_emit(cs, info->grid[0]);
+	radeon_emit(cs, info->grid[1]);
+	radeon_emit(cs, info->grid[2]);
 	/* VGT_DISPATCH_INITIATOR = COMPUTE_SHADER_EN */
 	radeon_emit(cs, 1);
 }
 
-static void compute_emit_cs(struct r600_context *ctx, const uint *block_layout,
-		const uint *grid_layout)
+static void compute_emit_cs(struct r600_context *rctx,
+			    const struct pipe_grid_info *info)
 {
-	struct radeon_winsys_cs *cs = ctx->b.gfx.cs;
+	struct radeon_winsys_cs *cs = rctx->b.gfx.cs;
 	unsigned i;
 
 	/* make sure that the gfx ring is only one active */
-	if (ctx->b.dma.cs && ctx->b.dma.cs->cdw) {
-		ctx->b.dma.flush(ctx, RADEON_FLUSH_ASYNC, NULL);
+	if (rctx->b.dma.cs && rctx->b.dma.cs->cdw) {
+		rctx->b.dma.flush(rctx, RADEON_FLUSH_ASYNC, NULL);
 	}
 
 	/* Initialize all the compute-related registers.
@@ -478,20 +459,20 @@ static void compute_emit_cs(struct r600_context *ctx, const uint *block_layout,
 	 * See evergreen_init_atom_start_compute_cs() in this file for the list
 	 * of registers initialized by the start_compute_cs_cmd atom.
 	 */
-	r600_emit_command_buffer(cs, &ctx->start_compute_cs_cmd);
+	r600_emit_command_buffer(cs, &rctx->start_compute_cs_cmd);
 
 	/* emit config state */
-	if (ctx->b.chip_class == EVERGREEN)
-		r600_emit_atom(ctx, &ctx->config_state.atom);
+	if (rctx->b.chip_class == EVERGREEN)
+		r600_emit_atom(rctx, &rctx->config_state.atom);
 
-	ctx->b.flags |= R600_CONTEXT_WAIT_3D_IDLE | R600_CONTEXT_FLUSH_AND_INV;
-	r600_flush_emit(ctx);
+	rctx->b.flags |= R600_CONTEXT_WAIT_3D_IDLE | R600_CONTEXT_FLUSH_AND_INV;
+	r600_flush_emit(rctx);
 
 	/* Emit colorbuffers. */
 	/* XXX support more than 8 colorbuffers (the offsets are not a multiple of 0x3C for CB8-11) */
-	for (i = 0; i < 8 && i < ctx->framebuffer.state.nr_cbufs; i++) {
-		struct r600_surface *cb = (struct r600_surface*)ctx->framebuffer.state.cbufs[i];
-		unsigned reloc = radeon_add_to_buffer_list(&ctx->b, &ctx->b.gfx,
+	for (i = 0; i < 8 && i < rctx->framebuffer.state.nr_cbufs; i++) {
+		struct r600_surface *cb = (struct r600_surface*)rctx->framebuffer.state.cbufs[i];
+		unsigned reloc = radeon_add_to_buffer_list(&rctx->b, &rctx->b.gfx,
 						       (struct r600_resource*)cb->base.texture,
 						       RADEON_USAGE_READWRITE,
 						       RADEON_PRIO_SHADER_RW_BUFFER);
@@ -520,51 +501,51 @@ static void compute_emit_cs(struct r600_context *ctx, const uint *block_layout,
 
 	/* Set CB_TARGET_MASK  XXX: Use cb_misc_state */
 	radeon_compute_set_context_reg(cs, R_028238_CB_TARGET_MASK,
-					ctx->compute_cb_target_mask);
+					rctx->compute_cb_target_mask);
 
 
 	/* Emit vertex buffer state */
-	ctx->cs_vertex_buffer_state.atom.num_dw = 12 * util_bitcount(ctx->cs_vertex_buffer_state.dirty_mask);
-	r600_emit_atom(ctx, &ctx->cs_vertex_buffer_state.atom);
+	rctx->cs_vertex_buffer_state.atom.num_dw = 12 * util_bitcount(rctx->cs_vertex_buffer_state.dirty_mask);
+	r600_emit_atom(rctx, &rctx->cs_vertex_buffer_state.atom);
 
 	/* Emit constant buffer state */
-	r600_emit_atom(ctx, &ctx->constbuf_state[PIPE_SHADER_COMPUTE].atom);
+	r600_emit_atom(rctx, &rctx->constbuf_state[PIPE_SHADER_COMPUTE].atom);
 
 	/* Emit sampler state */
-	r600_emit_atom(ctx, &ctx->samplers[PIPE_SHADER_COMPUTE].states.atom);
+	r600_emit_atom(rctx, &rctx->samplers[PIPE_SHADER_COMPUTE].states.atom);
 
 	/* Emit sampler view (texture resource) state */
-	r600_emit_atom(ctx, &ctx->samplers[PIPE_SHADER_COMPUTE].views.atom);
+	r600_emit_atom(rctx, &rctx->samplers[PIPE_SHADER_COMPUTE].views.atom);
 
 	/* Emit compute shader state */
-	r600_emit_atom(ctx, &ctx->cs_shader_state.atom);
+	r600_emit_atom(rctx, &rctx->cs_shader_state.atom);
 
 	/* Emit dispatch state and dispatch packet */
-	evergreen_emit_direct_dispatch(ctx, block_layout, grid_layout);
+	evergreen_emit_dispatch(rctx, info);
 
 	/* XXX evergreen_flush_emit() hardcodes the CP_COHER_SIZE to 0xffffffff
 	 */
-	ctx->b.flags |= R600_CONTEXT_INV_CONST_CACHE |
+	rctx->b.flags |= R600_CONTEXT_INV_CONST_CACHE |
 		      R600_CONTEXT_INV_VERTEX_CACHE |
 	              R600_CONTEXT_INV_TEX_CACHE;
-	r600_flush_emit(ctx);
-	ctx->b.flags = 0;
+	r600_flush_emit(rctx);
+	rctx->b.flags = 0;
 
-	if (ctx->b.chip_class >= CAYMAN) {
-		cs->buf[cs->cdw++] = PKT3(PKT3_EVENT_WRITE, 0, 0);
-		cs->buf[cs->cdw++] = EVENT_TYPE(EVENT_TYPE_CS_PARTIAL_FLUSH) | EVENT_INDEX(4);
+	if (rctx->b.chip_class >= CAYMAN) {
+		radeon_emit(cs, PKT3(PKT3_EVENT_WRITE, 0, 0));
+		radeon_emit(cs, EVENT_TYPE(EVENT_TYPE_CS_PARTIAL_FLUSH) | EVENT_INDEX(4));
 		/* DEALLOC_STATE prevents the GPU from hanging when a
 		 * SURFACE_SYNC packet is emitted some time after a DISPATCH_DIRECT
 		 * with any of the CB*_DEST_BASE_ENA or DB_DEST_BASE_ENA bits set.
 		 */
-		cs->buf[cs->cdw++] = PKT3C(PKT3_DEALLOC_STATE, 0, 0);
-		cs->buf[cs->cdw++] = 0;
+		radeon_emit(cs, PKT3C(PKT3_DEALLOC_STATE, 0, 0));
+		radeon_emit(cs, 0);
 	}
 
 #if 0
-	COMPUTE_DBG(ctx->screen, "cdw: %i\n", cs->cdw);
+	COMPUTE_DBG(rctx->screen, "cdw: %i\n", cs->cdw);
 	for (i = 0; i < cs->cdw; i++) {
-		COMPUTE_DBG(ctx->screen, "%4i : 0x%08X\n", i, cs->buf[i]);
+		COMPUTE_DBG(rctx->screen, "%4i : 0x%08X\n", i, cs->buf[i]);
 	}
 #endif
 
@@ -574,9 +555,8 @@ static void compute_emit_cs(struct r600_context *ctx, const uint *block_layout,
 /**
  * Emit function for r600_cs_shader_state atom
  */
-void evergreen_emit_cs_shader(
-		struct r600_context *rctx,
-		struct r600_atom *atom)
+void evergreen_emit_cs_shader(struct r600_context *rctx,
+			      struct r600_atom *atom)
 {
 	struct r600_cs_shader_state *state =
 					(struct r600_cs_shader_state*)atom;
@@ -604,35 +584,35 @@ void evergreen_emit_cs_shader(
 					      RADEON_PRIO_USER_SHADER));
 }
 
-static void evergreen_launch_grid(
-		struct pipe_context *ctx_, const struct pipe_grid_info *info)
+static void evergreen_launch_grid(struct pipe_context *ctx,
+				  const struct pipe_grid_info *info)
 {
-	struct r600_context *ctx = (struct r600_context *)ctx_;
+	struct r600_context *rctx = (struct r600_context *)ctx;
 #ifdef HAVE_OPENCL
-	struct r600_pipe_compute *shader = ctx->cs_shader_state.shader;
+	struct r600_pipe_compute *shader = rctx->cs_shader_state.shader;
 	boolean use_kill;
 
-	ctx->cs_shader_state.pc = info->pc;
+	rctx->cs_shader_state.pc = info->pc;
 	/* Get the config information for this kernel. */
 	r600_shader_binary_read_config(&shader->binary, &shader->bc,
                                   info->pc, &use_kill);
 #endif
 
-	COMPUTE_DBG(ctx->screen, "*** evergreen_launch_grid: pc = %u\n", info->pc);
+	COMPUTE_DBG(rctx->screen, "*** evergreen_launch_grid: pc = %u\n", info->pc);
 
 
-	evergreen_compute_upload_input(ctx_, info->block, info->grid, info->input);
-	compute_emit_cs(ctx, info->block, info->grid);
+	evergreen_compute_upload_input(ctx, info);
+	compute_emit_cs(rctx, info);
 }
 
-static void evergreen_set_compute_resources(struct pipe_context * ctx_,
-		unsigned start, unsigned count,
-		struct pipe_surface ** surfaces)
+static void evergreen_set_compute_resources(struct pipe_context *ctx,
+					    unsigned start, unsigned count,
+					    struct pipe_surface **surfaces)
 {
-	struct r600_context *ctx = (struct r600_context *)ctx_;
+	struct r600_context *rctx = (struct r600_context *)ctx;
 	struct r600_surface **resources = (struct r600_surface **)surfaces;
 
-	COMPUTE_DBG(ctx->screen, "*** evergreen_set_compute_resources: start = %u count = %u\n",
+	COMPUTE_DBG(rctx->screen, "*** evergreen_set_compute_resources: start = %u count = %u\n",
 			start, count);
 
 	for (unsigned i = 0; i < count; i++) {
@@ -646,31 +626,31 @@ static void evergreen_set_compute_resources(struct pipe_context * ctx_,
 			if (resources[i]->base.writable) {
 				assert(i+1 < 12);
 
-				evergreen_set_rat(ctx->cs_shader_state.shader, i+1,
+				evergreen_set_rat(rctx->cs_shader_state.shader, i+1,
 				(struct r600_resource *)resources[i]->base.texture,
 				buffer->chunk->start_in_dw*4,
 				resources[i]->base.texture->width0);
 			}
 
-			evergreen_cs_set_vertex_buffer(ctx, vtx_id,
+			evergreen_cs_set_vertex_buffer(rctx, vtx_id,
 					buffer->chunk->start_in_dw * 4,
 					resources[i]->base.texture);
 		}
 	}
 }
 
-static void evergreen_set_global_binding(
-	struct pipe_context *ctx_, unsigned first, unsigned n,
-	struct pipe_resource **resources,
-	uint32_t **handles)
+static void evergreen_set_global_binding(struct pipe_context *ctx,
+					 unsigned first, unsigned n,
+					 struct pipe_resource **resources,
+					 uint32_t **handles)
 {
-	struct r600_context *ctx = (struct r600_context *)ctx_;
-	struct compute_memory_pool *pool = ctx->screen->global_pool;
+	struct r600_context *rctx = (struct r600_context *)ctx;
+	struct compute_memory_pool *pool = rctx->screen->global_pool;
 	struct r600_resource_global **buffers =
 		(struct r600_resource_global **)resources;
 	unsigned i;
 
-	COMPUTE_DBG(ctx->screen, "*** evergreen_set_global_binding first = %u n = %u\n",
+	COMPUTE_DBG(rctx->screen, "*** evergreen_set_global_binding first = %u n = %u\n",
 			first, n);
 
 	if (!resources) {
@@ -687,7 +667,7 @@ static void evergreen_set_global_binding(
 			buffers[i]->chunk->status |= ITEM_FOR_PROMOTING;
 	}
 
-	if (compute_memory_finalize_pending(pool, ctx_) == -1) {
+	if (compute_memory_finalize_pending(pool, ctx) == -1) {
 		/* XXX: Unset */
 		return;
 	}
@@ -705,8 +685,8 @@ static void evergreen_set_global_binding(
 		*(handles[i]) = util_cpu_to_le32(handle);
 	}
 
-	evergreen_set_rat(ctx->cs_shader_state.shader, 0, pool->bo, 0, pool->size_in_dw * 4);
-	evergreen_cs_set_vertex_buffer(ctx, 1, 0,
+	evergreen_set_rat(rctx->cs_shader_state.shader, 0, pool->bo, 0, pool->size_in_dw * 4);
+	evergreen_cs_set_vertex_buffer(rctx, 1, 0,
 				(struct pipe_resource*)pool->bo);
 }
 
@@ -721,9 +701,9 @@ static void evergreen_set_global_binding(
  * functions evergreen_init_atom_start_cs or cayman_init_atom_start_cs depending
  * on the GPU family.
  */
-void evergreen_init_atom_start_compute_cs(struct r600_context *ctx)
+void evergreen_init_atom_start_compute_cs(struct r600_context *rctx)
 {
-	struct r600_command_buffer *cb = &ctx->start_compute_cs_cmd;
+	struct r600_command_buffer *cb = &rctx->start_compute_cs_cmd;
 	int num_threads;
 	int num_stack_entries;
 
@@ -742,7 +722,7 @@ void evergreen_init_atom_start_compute_cs(struct r600_context *ctx)
 	r600_store_value(cb, PKT3(PKT3_EVENT_WRITE, 0, 0));
 	r600_store_value(cb, EVENT_TYPE(EVENT_TYPE_CS_PARTIAL_FLUSH) | EVENT_INDEX(4));
 
-	switch (ctx->b.family) {
+	switch (rctx->b.family) {
 	case CHIP_CEDAR:
 	default:
 		num_threads = 128;
@@ -788,18 +768,18 @@ void evergreen_init_atom_start_compute_cs(struct r600_context *ctx)
 	}
 
 	/* Config Registers */
-	if (ctx->b.chip_class < CAYMAN)
-		evergreen_init_common_regs(ctx, cb, ctx->b.chip_class, ctx->b.family,
-					   ctx->screen->b.info.drm_minor);
+	if (rctx->b.chip_class < CAYMAN)
+		evergreen_init_common_regs(rctx, cb, rctx->b.chip_class, rctx->b.family,
+					   rctx->screen->b.info.drm_minor);
 	else
-		cayman_init_common_regs(cb, ctx->b.chip_class, ctx->b.family,
-					ctx->screen->b.info.drm_minor);
+		cayman_init_common_regs(cb, rctx->b.chip_class, rctx->b.family,
+					rctx->screen->b.info.drm_minor);
 
 	/* The primitive type always needs to be POINTLIST for compute. */
 	r600_store_config_reg(cb, R_008958_VGT_PRIMITIVE_TYPE,
 						V_008958_DI_PT_POINTLIST);
 
-	if (ctx->b.chip_class < CAYMAN) {
+	if (rctx->b.chip_class < CAYMAN) {
 
 		/* These registers control which simds can be used by each stage.
 		 * The default for these registers is 0xffffffff, which means
@@ -849,7 +829,7 @@ void evergreen_init_atom_start_compute_cs(struct r600_context *ctx)
 	 * allocate the appropriate amount of LDS dwords using the
 	 * CM_R_0288E8_SQ_LDS_ALLOC register.
 	 */
-	if (ctx->b.chip_class < CAYMAN) {
+	if (rctx->b.chip_class < CAYMAN) {
 		r600_store_config_reg(cb, R_008E2C_SQ_LDS_RESOURCE_MGMT,
 			S_008E2C_NUM_PS_LDS(0x0000) | S_008E2C_NUM_LS_LDS(8192));
 	} else {
@@ -860,7 +840,7 @@ void evergreen_init_atom_start_compute_cs(struct r600_context *ctx)
 
 	/* Context Registers */
 
-	if (ctx->b.chip_class < CAYMAN) {
+	if (rctx->b.chip_class < CAYMAN) {
 		/* workaround for hw issues with dyn gpr - must set all limits
 		 * to 240 instead of 0, 0x1e == 240 / 8
 		 */
@@ -902,86 +882,26 @@ void evergreen_init_atom_start_compute_cs(struct r600_context *ctx)
 	eg_store_loop_const(cb, R_03A200_SQ_LOOP_CONST_0 + (160 * 4), 0x1000FFF);
 }
 
-void evergreen_init_compute_state_functions(struct r600_context *ctx)
-{
-	ctx->b.b.create_compute_state = evergreen_create_compute_state;
-	ctx->b.b.delete_compute_state = evergreen_delete_compute_state;
-	ctx->b.b.bind_compute_state = evergreen_bind_compute_state;
-//	 ctx->context.create_sampler_view = evergreen_compute_create_sampler_view;
-	ctx->b.b.set_compute_resources = evergreen_set_compute_resources;
-	ctx->b.b.set_global_binding = evergreen_set_global_binding;
-	ctx->b.b.launch_grid = evergreen_launch_grid;
-
-}
-
-struct pipe_resource *r600_compute_global_buffer_create(
-	struct pipe_screen *screen,
-	const struct pipe_resource *templ)
+void evergreen_init_compute_state_functions(struct r600_context *rctx)
 {
-	struct r600_resource_global* result = NULL;
-	struct r600_screen* rscreen = NULL;
-	int size_in_dw = 0;
-
-	assert(templ->target == PIPE_BUFFER);
-	assert(templ->bind & PIPE_BIND_GLOBAL);
-	assert(templ->array_size == 1 || templ->array_size == 0);
-	assert(templ->depth0 == 1 || templ->depth0 == 0);
-	assert(templ->height0 == 1 || templ->height0 == 0);
-
-	result = (struct r600_resource_global*)
-	CALLOC(sizeof(struct r600_resource_global), 1);
-	rscreen = (struct r600_screen*)screen;
-
-	COMPUTE_DBG(rscreen, "*** r600_compute_global_buffer_create\n");
-	COMPUTE_DBG(rscreen, "width = %u array_size = %u\n", templ->width0,
-			templ->array_size);
-
-	result->base.b.vtbl = &r600_global_buffer_vtbl;
-	result->base.b.b = *templ;
-	result->base.b.b.screen = screen;
-	pipe_reference_init(&result->base.b.b.reference, 1);
-
-	size_in_dw = (templ->width0+3) / 4;
-
-	result->chunk = compute_memory_alloc(rscreen->global_pool, size_in_dw);
-
-	if (result->chunk == NULL)
-	{
-		free(result);
-		return NULL;
-	}
-
-	return &result->base.b.b;
-}
-
-void r600_compute_global_buffer_destroy(
-	struct pipe_screen *screen,
-	struct pipe_resource *res)
-{
-	struct r600_resource_global* buffer = NULL;
-	struct r600_screen* rscreen = NULL;
-
-	assert(res->target == PIPE_BUFFER);
-	assert(res->bind & PIPE_BIND_GLOBAL);
-
-	buffer = (struct r600_resource_global*)res;
-	rscreen = (struct r600_screen*)screen;
+	rctx->b.b.create_compute_state = evergreen_create_compute_state;
+	rctx->b.b.delete_compute_state = evergreen_delete_compute_state;
+	rctx->b.b.bind_compute_state = evergreen_bind_compute_state;
+//	 rctx->context.create_sampler_view = evergreen_compute_create_sampler_view;
+	rctx->b.b.set_compute_resources = evergreen_set_compute_resources;
+	rctx->b.b.set_global_binding = evergreen_set_global_binding;
+	rctx->b.b.launch_grid = evergreen_launch_grid;
 
-	compute_memory_free(rscreen->global_pool, buffer->chunk->id);
-
-	buffer->chunk = NULL;
-	free(res);
 }
 
-void *r600_compute_global_transfer_map(
-	struct pipe_context *ctx_,
-	struct pipe_resource *resource,
-	unsigned level,
-	unsigned usage,
-	const struct pipe_box *box,
-	struct pipe_transfer **ptransfer)
+static void *r600_compute_global_transfer_map(struct pipe_context *ctx,
+					      struct pipe_resource *resource,
+					      unsigned level,
+					      unsigned usage,
+					      const struct pipe_box *box,
+					      struct pipe_transfer **ptransfer)
 {
-	struct r600_context *rctx = (struct r600_context*)ctx_;
+	struct r600_context *rctx = (struct r600_context*)ctx;
 	struct compute_memory_pool *pool = rctx->screen->global_pool;
 	struct r600_resource_global* buffer =
 		(struct r600_resource_global*)resource;
@@ -991,7 +911,7 @@ void *r600_compute_global_transfer_map(
 	unsigned offset = box->x;
 
 	if (is_item_in_pool(item)) {
-		compute_memory_demote_item(pool, item, ctx_);
+		compute_memory_demote_item(pool, item, ctx);
 	}
 	else {
 		if (item->real_buffer == NULL) {
@@ -1021,13 +941,12 @@ void *r600_compute_global_transfer_map(
 	assert(box->z == 0);
 
 	///TODO: do it better, mapping is not possible if the pool is too big
-	return pipe_buffer_map_range(ctx_, dst,
+	return pipe_buffer_map_range(ctx, dst,
 			offset, box->width, usage, ptransfer);
 }
 
-void r600_compute_global_transfer_unmap(
-	struct pipe_context *ctx_,
-	struct pipe_transfer* transfer)
+static void r600_compute_global_transfer_unmap(struct pipe_context *ctx,
+					       struct pipe_transfer *transfer)
 {
 	/* struct r600_resource_global are not real resources, they just map
 	 * to an offset within the compute memory pool.  The function
@@ -1042,23 +961,88 @@ void r600_compute_global_transfer_unmap(
 	assert (!"This function should not be called");
 }
 
-void r600_compute_global_transfer_flush_region(
-	struct pipe_context *ctx_,
-	struct pipe_transfer *transfer,
-	const struct pipe_box *box)
+static void r600_compute_global_transfer_flush_region(struct pipe_context *ctx,
+						      struct pipe_transfer *transfer,
+						      const struct pipe_box *box)
 {
 	assert(0 && "TODO");
 }
 
-void r600_compute_global_transfer_inline_write(
-	struct pipe_context *pipe,
-	struct pipe_resource *resource,
-	unsigned level,
-	unsigned usage,
-	const struct pipe_box *box,
-	const void *data,
-	unsigned stride,
-	unsigned layer_stride)
+static void r600_compute_global_transfer_inline_write(struct pipe_context *pipe,
+						      struct pipe_resource *resource,
+						      unsigned level,
+						      unsigned usage,
+						      const struct pipe_box *box,
+						      const void *data,
+						      unsigned stride,
+						      unsigned layer_stride)
 {
 	assert(0 && "TODO");
 }
+
+static void r600_compute_global_buffer_destroy(struct pipe_screen *screen,
+					       struct pipe_resource *res)
+{
+	struct r600_resource_global* buffer = NULL;
+	struct r600_screen* rscreen = NULL;
+
+	assert(res->target == PIPE_BUFFER);
+	assert(res->bind & PIPE_BIND_GLOBAL);
+
+	buffer = (struct r600_resource_global*)res;
+	rscreen = (struct r600_screen*)screen;
+
+	compute_memory_free(rscreen->global_pool, buffer->chunk->id);
+
+	buffer->chunk = NULL;
+	free(res);
+}
+
+static const struct u_resource_vtbl r600_global_buffer_vtbl =
+{
+	u_default_resource_get_handle, /* get_handle */
+	r600_compute_global_buffer_destroy, /* resource_destroy */
+	r600_compute_global_transfer_map, /* transfer_map */
+	r600_compute_global_transfer_flush_region,/* transfer_flush_region */
+	r600_compute_global_transfer_unmap, /* transfer_unmap */
+	r600_compute_global_transfer_inline_write /* transfer_inline_write */
+};
+
+struct pipe_resource *r600_compute_global_buffer_create(struct pipe_screen *screen,
+							const struct pipe_resource *templ)
+{
+	struct r600_resource_global* result = NULL;
+	struct r600_screen* rscreen = NULL;
+	int size_in_dw = 0;
+
+	assert(templ->target == PIPE_BUFFER);
+	assert(templ->bind & PIPE_BIND_GLOBAL);
+	assert(templ->array_size == 1 || templ->array_size == 0);
+	assert(templ->depth0 == 1 || templ->depth0 == 0);
+	assert(templ->height0 == 1 || templ->height0 == 0);
+
+	result = (struct r600_resource_global*)
+	CALLOC(sizeof(struct r600_resource_global), 1);
+	rscreen = (struct r600_screen*)screen;
+
+	COMPUTE_DBG(rscreen, "*** r600_compute_global_buffer_create\n");
+	COMPUTE_DBG(rscreen, "width = %u array_size = %u\n", templ->width0,
+			templ->array_size);
+
+	result->base.b.vtbl = &r600_global_buffer_vtbl;
+	result->base.b.b = *templ;
+	result->base.b.b.screen = screen;
+	pipe_reference_init(&result->base.b.b.reference, 1);
+
+	size_in_dw = (templ->width0+3) / 4;
+
+	result->chunk = compute_memory_alloc(rscreen->global_pool, size_in_dw);
+
+	if (result->chunk == NULL)
+	{
+		free(result);
+		return NULL;
+	}
+
+	return &result->base.b.b;
+}
diff --git a/src/gallium/drivers/r600/evergreen_compute.h b/src/gallium/drivers/r600/evergreen_compute.h
index e4d3a38e415..3c178870d91 100644
--- a/src/gallium/drivers/r600/evergreen_compute.h
+++ b/src/gallium/drivers/r600/evergreen_compute.h
@@ -38,26 +38,11 @@ struct r600_resource_global {
 	struct compute_memory_item *chunk;
 };
 
-void *evergreen_create_compute_state(struct pipe_context *ctx, const struct pipe_compute_state *cso);
-void evergreen_delete_compute_state(struct pipe_context *ctx, void *state);
-void evergreen_compute_upload_input(struct pipe_context *context, const uint *block_layout, const uint *grid_layout, const void *input);
 void evergreen_init_atom_start_compute_cs(struct r600_context *rctx);
 void evergreen_init_compute_state_functions(struct r600_context *rctx);
 void evergreen_emit_cs_shader(struct r600_context *rctx, struct r600_atom * atom);
 
 struct r600_resource* r600_compute_buffer_alloc_vram(struct r600_screen *screen, unsigned size);
 struct pipe_resource *r600_compute_global_buffer_create(struct pipe_screen *screen, const struct pipe_resource *templ);
-void r600_compute_global_buffer_destroy(struct pipe_screen *screen, struct pipe_resource *res);
-void *r600_compute_global_transfer_map(
-	struct pipe_context *ctx_,
-	struct pipe_resource *resource,
-	unsigned level,
-	unsigned usage,
-	const struct pipe_box *box,
-	struct pipe_transfer **ptransfer);
-void r600_compute_global_transfer_unmap(struct pipe_context *ctx, struct pipe_transfer* transfer);
-void r600_compute_global_transfer_flush_region( struct pipe_context *, struct pipe_transfer *, const struct pipe_box *);
-void r600_compute_global_transfer_inline_write( struct pipe_context *, struct pipe_resource *, unsigned level,
-                                                unsigned usage, const struct pipe_box *, const void *data, unsigned stride, unsigned layer_stride);
 
 #endif
diff --git a/src/gallium/drivers/r600/r600_pipe.c b/src/gallium/drivers/r600/r600_pipe.c
index c97e34121e3..36b808fbbca 100644
--- a/src/gallium/drivers/r600/r600_pipe.c
+++ b/src/gallium/drivers/r600/r600_pipe.c
@@ -281,6 +281,7 @@ static int r600_get_param(struct pipe_screen* pscreen, enum pipe_cap param)
 	case PIPE_CAP_INVALIDATE_BUFFER:
 	case PIPE_CAP_SURFACE_REINTERPRET_BLOCKS:
 	case PIPE_CAP_QUERY_MEMORY_INFO:
+	case PIPE_CAP_FRAMEBUFFER_NO_ATTACHMENT:
 		return 1;
 
 	case PIPE_CAP_DEVICE_RESET_STATUS_QUERY: