[Mesa-dev] [PATCH 1/7] radeonsi: import r600_streamout from drivers/radeon

Marek Olšák maraeo at gmail.com
Sat Oct 7 22:47:00 UTC 2017


From: Marek Olšák <marek.olsak at amd.com>

---
 src/gallium/drivers/radeon/Makefile.sources        |   1 -
 src/gallium/drivers/radeon/r600_pipe_common.c      |  12 --
 src/gallium/drivers/radeon/r600_pipe_common.h      |  57 -------
 src/gallium/drivers/radeon/r600_query.c            |   4 +
 src/gallium/drivers/radeonsi/Makefile.sources      |   1 +
 src/gallium/drivers/radeonsi/si_blit.c             |   4 +-
 src/gallium/drivers/radeonsi/si_descriptors.c      |  20 +--
 src/gallium/drivers/radeonsi/si_hw_context.c       |  13 +-
 src/gallium/drivers/radeonsi/si_pipe.c             |   1 +
 src/gallium/drivers/radeonsi/si_pipe.h             |  44 ++++++
 src/gallium/drivers/radeonsi/si_state.c            |   4 +-
 src/gallium/drivers/radeonsi/si_state.h            |  11 ++
 src/gallium/drivers/radeonsi/si_state_draw.c       |   6 +-
 src/gallium/drivers/radeonsi/si_state_shaders.c    |   4 +-
 .../si_state_streamout.c}                          | 176 +++++++++++----------
 15 files changed, 181 insertions(+), 177 deletions(-)
 rename src/gallium/drivers/{radeon/r600_streamout.c => radeonsi/si_state_streamout.c} (58%)

diff --git a/src/gallium/drivers/radeon/Makefile.sources b/src/gallium/drivers/radeon/Makefile.sources
index 5d38bb3..c32ebea 100644
--- a/src/gallium/drivers/radeon/Makefile.sources
+++ b/src/gallium/drivers/radeon/Makefile.sources
@@ -1,21 +1,20 @@
 C_SOURCES := \
 	cayman_msaa.c \
 	r600_buffer_common.c \
 	r600_cs.h \
 	r600_gpu_load.c \
 	r600_perfcounter.c \
 	r600_pipe_common.c \
 	r600_pipe_common.h \
 	r600_query.c \
 	r600_query.h \
-	r600_streamout.c \
 	r600_test_dma.c \
 	r600_texture.c \
 	radeon_uvd.c \
 	radeon_uvd.h \
 	radeon_vcn_dec.c \
 	radeon_vcn_dec.h \
 	radeon_vce_40_2_2.c \
 	radeon_vce_50.c \
 	radeon_vce_52.c \
 	radeon_vce.c \
diff --git a/src/gallium/drivers/radeon/r600_pipe_common.c b/src/gallium/drivers/radeon/r600_pipe_common.c
index da61580..a6008a1 100644
--- a/src/gallium/drivers/radeon/r600_pipe_common.c
+++ b/src/gallium/drivers/radeon/r600_pipe_common.c
@@ -289,35 +289,24 @@ void si_need_dma_space(struct r600_common_context *ctx, unsigned num_dw,
 
 static void r600_memory_barrier(struct pipe_context *ctx, unsigned flags)
 {
 }
 
 void si_preflush_suspend_features(struct r600_common_context *ctx)
 {
 	/* suspend queries */
 	if (!LIST_IS_EMPTY(&ctx->active_queries))
 		si_suspend_queries(ctx);
-
-	ctx->streamout.suspended = false;
-	if (ctx->streamout.begin_emitted) {
-		si_emit_streamout_end(ctx);
-		ctx->streamout.suspended = true;
-	}
 }
 
 void si_postflush_resume_features(struct r600_common_context *ctx)
 {
-	if (ctx->streamout.suspended) {
-		ctx->streamout.append_bitmask = ctx->streamout.enabled_mask;
-		si_streamout_buffers_dirty(ctx);
-	}
-
 	/* resume queries */
 	if (!LIST_IS_EMPTY(&ctx->active_queries))
 		si_resume_queries(ctx);
 }
 
 static void r600_add_fence_dependency(struct r600_common_context *rctx,
 				      struct pipe_fence_handle *fence)
 {
 	struct radeon_winsys *ws = rctx->ws;
 
@@ -640,21 +629,20 @@ bool si_common_context_init(struct r600_common_context *rctx,
 	if (rscreen->info.drm_major == 2 && rscreen->info.drm_minor >= 43) {
 		rctx->b.get_device_reset_status = r600_get_reset_status;
 		rctx->gpu_reset_counter =
 			rctx->ws->query_value(rctx->ws,
 					      RADEON_GPU_RESET_COUNTER);
 	}
 
 	rctx->b.set_device_reset_callback = r600_set_device_reset_callback;
 
 	si_init_context_texture_functions(rctx);
-	si_streamout_init(rctx);
 	si_init_query_functions(rctx);
 	si_init_msaa(&rctx->b);
 
 	if (rctx->chip_class == CIK ||
 	    rctx->chip_class == VI ||
 	    rctx->chip_class == GFX9) {
 		rctx->eop_bug_scratch = (struct r600_resource*)
 			pipe_buffer_create(&rscreen->b, 0, PIPE_USAGE_DEFAULT,
 					   16 * rscreen->info.num_render_backends);
 		if (!rctx->eop_bug_scratch)
diff --git a/src/gallium/drivers/radeon/r600_pipe_common.h b/src/gallium/drivers/radeon/r600_pipe_common.h
index e3cb1cf..b620e6b 100644
--- a/src/gallium/drivers/radeon/r600_pipe_common.h
+++ b/src/gallium/drivers/radeon/r600_pipe_common.h
@@ -490,57 +490,20 @@ struct r600_common_screen {
 				    struct radeon_bo_metadata *md);
 };
 
 /* This encapsulates a state or an operation which can emitted into the GPU
  * command stream. */
 struct r600_atom {
 	void (*emit)(struct r600_common_context *ctx, struct r600_atom *state);
 	unsigned short		id;
 };
 
-struct r600_so_target {
-	struct pipe_stream_output_target b;
-
-	/* The buffer where BUFFER_FILLED_SIZE is stored. */
-	struct r600_resource	*buf_filled_size;
-	unsigned		buf_filled_size_offset;
-	bool			buf_filled_size_valid;
-
-	unsigned		stride_in_dw;
-};
-
-struct r600_streamout {
-	struct r600_atom		begin_atom;
-	bool				begin_emitted;
-
-	unsigned			enabled_mask;
-	unsigned			num_targets;
-	struct r600_so_target		*targets[PIPE_MAX_SO_BUFFERS];
-
-	unsigned			append_bitmask;
-	bool				suspended;
-
-	/* External state which comes from the vertex shader,
-	 * it must be set explicitly when binding a shader. */
-	uint16_t			*stride_in_dw;
-	unsigned			enabled_stream_buffers_mask; /* stream0 buffers0-3 in 4 LSB */
-
-	/* The state of VGT_STRMOUT_BUFFER_(CONFIG|EN). */
-	unsigned			hw_enabled_mask;
-
-	/* The state of VGT_STRMOUT_(CONFIG|EN). */
-	struct r600_atom		enable_atom;
-	bool				streamout_enabled;
-	bool				prims_gen_query_enabled;
-	int				num_prims_gen_queries;
-};
-
 struct r600_ring {
 	struct radeon_winsys_cs		*cs;
 	void (*flush)(void *ctx, unsigned flags,
 		      struct pipe_fence_handle **fence);
 };
 
 /* Saved CS data for debugging features. */
 struct radeon_saved_cs {
 	uint32_t			*ib;
 	unsigned			num_dw;
@@ -571,23 +534,20 @@ struct r600_common_context {
 
 	struct threaded_context		*tc;
 	struct u_suballocator		*allocator_zeroed_memory;
 	struct slab_child_pool		pool_transfers;
 	struct slab_child_pool		pool_transfers_unsync; /* for threaded_context */
 
 	/* Current unaccounted memory usage. */
 	uint64_t			vram;
 	uint64_t			gtt;
 
-	/* States. */
-	struct r600_streamout		streamout;
-
 	/* Additional context states. */
 	unsigned flags; /* flush flags */
 
 	/* Queries. */
 	/* Maintain the list of active queries for pausing between IBs. */
 	int				num_occlusion_queries;
 	int				num_perfect_occlusion_queries;
 	struct list_head		active_queries;
 	unsigned			num_cs_dw_queries_suspend;
 	/* Misc stats. */
@@ -783,31 +743,20 @@ unsigned si_end_counter(struct r600_common_screen *rscreen, unsigned type,
 
 /* r600_perfcounters.c */
 void si_perfcounters_destroy(struct r600_common_screen *rscreen);
 
 /* r600_query.c */
 void si_init_screen_query_functions(struct r600_common_screen *rscreen);
 void si_init_query_functions(struct r600_common_context *rctx);
 void si_suspend_queries(struct r600_common_context *ctx);
 void si_resume_queries(struct r600_common_context *ctx);
 
-/* r600_streamout.c */
-void si_streamout_buffers_dirty(struct r600_common_context *rctx);
-void si_common_set_streamout_targets(struct pipe_context *ctx,
-				     unsigned num_targets,
-				     struct pipe_stream_output_target **targets,
-				     const unsigned *offset);
-void si_emit_streamout_end(struct r600_common_context *rctx);
-void si_update_prims_generated_query_state(struct r600_common_context *rctx,
-					   unsigned type, int diff);
-void si_streamout_init(struct r600_common_context *rctx);
-
 /* r600_test_dma.c */
 void si_test_dma(struct r600_common_screen *rscreen);
 
 /* r600_texture.c */
 bool si_prepare_for_dma_blit(struct r600_common_context *rctx,
 			     struct r600_texture *rdst,
 			     unsigned dst_level, unsigned dstx,
 			     unsigned dsty, unsigned dstz,
 			     struct r600_texture *rsrc,
 			     unsigned src_level,
@@ -893,26 +842,20 @@ r600_context_add_resource_size(struct pipe_context *ctx, struct pipe_resource *r
 	struct r600_common_context *rctx = (struct r600_common_context *)ctx;
 	struct r600_resource *res = (struct r600_resource *)r;
 
 	if (res) {
 		/* Add memory usage for need_gfx_cs_space */
 		rctx->vram += res->vram_usage;
 		rctx->gtt += res->gart_usage;
 	}
 }
 
-static inline bool r600_get_strmout_en(struct r600_common_context *rctx)
-{
-	return rctx->streamout.streamout_enabled ||
-	       rctx->streamout.prims_gen_query_enabled;
-}
-
 #define     SQ_TEX_XY_FILTER_POINT                         0x00
 #define     SQ_TEX_XY_FILTER_BILINEAR                      0x01
 #define     SQ_TEX_XY_FILTER_ANISO_POINT                   0x02
 #define     SQ_TEX_XY_FILTER_ANISO_BILINEAR                0x03
 
 static inline unsigned eg_tex_filter(unsigned filter, unsigned max_aniso)
 {
 	if (filter == PIPE_TEX_FILTER_LINEAR)
 		return max_aniso > 1 ? SQ_TEX_XY_FILTER_ANISO_BILINEAR
 				     : SQ_TEX_XY_FILTER_BILINEAR;
diff --git a/src/gallium/drivers/radeon/r600_query.c b/src/gallium/drivers/radeon/r600_query.c
index adf3522..3abfe1e 100644
--- a/src/gallium/drivers/radeon/r600_query.c
+++ b/src/gallium/drivers/radeon/r600_query.c
@@ -22,20 +22,24 @@
  * USE OR OTHER DEALINGS IN THE SOFTWARE.
  */
 
 #include "r600_query.h"
 #include "r600_cs.h"
 #include "util/u_memory.h"
 #include "util/u_upload_mgr.h"
 #include "os/os_time.h"
 #include "tgsi/tgsi_text.h"
 
+/* TODO: remove this: */
+void si_update_prims_generated_query_state(struct r600_common_context *rctx,
+					   unsigned type, int diff);
+
 #define R600_MAX_STREAMS 4
 
 struct r600_hw_query_params {
 	unsigned start_offset;
 	unsigned end_offset;
 	unsigned fence_offset;
 	unsigned pair_stride;
 	unsigned pair_count;
 };
 
diff --git a/src/gallium/drivers/radeonsi/Makefile.sources b/src/gallium/drivers/radeonsi/Makefile.sources
index ed3e520..63cd7a3 100644
--- a/src/gallium/drivers/radeonsi/Makefile.sources
+++ b/src/gallium/drivers/radeonsi/Makefile.sources
@@ -23,13 +23,14 @@ C_SOURCES := \
 	si_shader.h \
 	si_shader_internal.h \
 	si_shader_nir.c \
 	si_shader_tgsi_alu.c \
 	si_shader_tgsi_mem.c \
 	si_shader_tgsi_setup.c \
 	si_state.c \
 	si_state_binning.c \
 	si_state_draw.c \
 	si_state_shaders.c \
+	si_state_streamout.c \
 	si_state_viewport.c \
 	si_state.h \
 	si_uvd.c
diff --git a/src/gallium/drivers/radeonsi/si_blit.c b/src/gallium/drivers/radeonsi/si_blit.c
index 4806e7c..03aa4f7 100644
--- a/src/gallium/drivers/radeonsi/si_blit.c
+++ b/src/gallium/drivers/radeonsi/si_blit.c
@@ -51,22 +51,22 @@ enum si_blitter_op /* bitmask */
 };
 
 static void si_blitter_begin(struct pipe_context *ctx, enum si_blitter_op op)
 {
 	struct si_context *sctx = (struct si_context *)ctx;
 
 	util_blitter_save_vertex_shader(sctx->blitter, sctx->vs_shader.cso);
 	util_blitter_save_tessctrl_shader(sctx->blitter, sctx->tcs_shader.cso);
 	util_blitter_save_tesseval_shader(sctx->blitter, sctx->tes_shader.cso);
 	util_blitter_save_geometry_shader(sctx->blitter, sctx->gs_shader.cso);
-	util_blitter_save_so_targets(sctx->blitter, sctx->b.streamout.num_targets,
-				     (struct pipe_stream_output_target**)sctx->b.streamout.targets);
+	util_blitter_save_so_targets(sctx->blitter, sctx->streamout.num_targets,
+				     (struct pipe_stream_output_target**)sctx->streamout.targets);
 	util_blitter_save_rasterizer(sctx->blitter, sctx->queued.named.rasterizer);
 
 	if (op & SI_SAVE_FRAGMENT_STATE) {
 		util_blitter_save_blend(sctx->blitter, sctx->queued.named.blend);
 		util_blitter_save_depth_stencil_alpha(sctx->blitter, sctx->queued.named.dsa);
 		util_blitter_save_stencil_ref(sctx->blitter, &sctx->stencil_ref.state);
 		util_blitter_save_fragment_shader(sctx->blitter, sctx->ps_shader.cso);
 		util_blitter_save_sample_mask(sctx->blitter, sctx->sample_mask.sample_mask);
 		util_blitter_save_scissor(sctx->blitter, &sctx->scissors.states[0]);
 	}
diff --git a/src/gallium/drivers/radeonsi/si_descriptors.c b/src/gallium/drivers/radeonsi/si_descriptors.c
index dee8e71..dd1f1e9 100644
--- a/src/gallium/drivers/radeonsi/si_descriptors.c
+++ b/src/gallium/drivers/radeonsi/si_descriptors.c
@@ -1366,37 +1366,37 @@ void si_set_ring_buffer(struct pipe_context *ctx, uint slot,
 /* STREAMOUT BUFFERS */
 
 static void si_set_streamout_targets(struct pipe_context *ctx,
 				     unsigned num_targets,
 				     struct pipe_stream_output_target **targets,
 				     const unsigned *offsets)
 {
 	struct si_context *sctx = (struct si_context *)ctx;
 	struct si_buffer_resources *buffers = &sctx->rw_buffers;
 	struct si_descriptors *descs = &sctx->descriptors[SI_DESCS_RW_BUFFERS];
-	unsigned old_num_targets = sctx->b.streamout.num_targets;
+	unsigned old_num_targets = sctx->streamout.num_targets;
 	unsigned i, bufidx;
 
 	/* We are going to unbind the buffers. Mark which caches need to be flushed. */
-	if (sctx->b.streamout.num_targets && sctx->b.streamout.begin_emitted) {
+	if (sctx->streamout.num_targets && sctx->streamout.begin_emitted) {
 		/* Since streamout uses vector writes which go through TC L2
 		 * and most other clients can use TC L2 as well, we don't need
 		 * to flush it.
 		 *
 		 * The only cases which requires flushing it is VGT DMA index
 		 * fetching (on <= CIK) and indirect draw data, which are rare
 		 * cases. Thus, flag the TC L2 dirtiness in the resource and
 		 * handle it at draw call time.
 		 */
-		for (i = 0; i < sctx->b.streamout.num_targets; i++)
-			if (sctx->b.streamout.targets[i])
-				r600_resource(sctx->b.streamout.targets[i]->b.buffer)->TC_L2_dirty = true;
+		for (i = 0; i < sctx->streamout.num_targets; i++)
+			if (sctx->streamout.targets[i])
+				r600_resource(sctx->streamout.targets[i]->b.buffer)->TC_L2_dirty = true;
 
 		/* Invalidate the scalar cache in case a streamout buffer is
 		 * going to be used as a constant buffer.
 		 *
 		 * Invalidate TC L1, because streamout bypasses it (done by
 		 * setting GLC=1 in the store instruction), but it can contain
 		 * outdated data of streamout buffers.
 		 *
 		 * VS_PARTIAL_FLUSH is required if the buffers are going to be
 		 * used as an input immediately.
@@ -1643,25 +1643,25 @@ static void si_rebind_buffer(struct pipe_context *ctx, struct pipe_resource *buf
 			si_desc_reset_buffer_offset(ctx, descs->list + i*4,
 						    old_va, buf);
 			sctx->descriptors_dirty |= 1u << SI_DESCS_RW_BUFFERS;
 
 			radeon_add_to_buffer_list_check_mem(&sctx->b, &sctx->b.gfx,
 							    rbuffer, buffers->shader_usage,
 							    RADEON_PRIO_SHADER_RW_BUFFER,
 							    true);
 
 			/* Update the streamout state. */
-			if (sctx->b.streamout.begin_emitted)
-				si_emit_streamout_end(&sctx->b);
-			sctx->b.streamout.append_bitmask =
-					sctx->b.streamout.enabled_mask;
-			si_streamout_buffers_dirty(&sctx->b);
+			if (sctx->streamout.begin_emitted)
+				si_emit_streamout_end(sctx);
+			sctx->streamout.append_bitmask =
+					sctx->streamout.enabled_mask;
+			si_streamout_buffers_dirty(sctx);
 		}
 	}
 
 	/* Constant and shader buffers. */
 	if (rbuffer->bind_history & PIPE_BIND_CONSTANT_BUFFER) {
 		for (shader = 0; shader < SI_NUM_SHADERS; shader++)
 			si_reset_buffer_resources(sctx, &sctx->const_and_shader_buffers[shader],
 						  si_const_and_shader_buffer_descriptors_idx(shader),
 						  u_bit_consecutive(SI_NUM_SHADER_BUFFERS, SI_NUM_CONST_BUFFERS),
 						  buf, old_va,
diff --git a/src/gallium/drivers/radeonsi/si_hw_context.c b/src/gallium/drivers/radeonsi/si_hw_context.c
index 72da54e..317b50c 100644
--- a/src/gallium/drivers/radeonsi/si_hw_context.c
+++ b/src/gallium/drivers/radeonsi/si_hw_context.c
@@ -93,20 +93,26 @@ void si_context_gfx_flush(void *context, unsigned flags,
 	 */
 	if (radeon_emitted(ctx->b.dma.cs, 0)) {
 		assert(fence == NULL); /* internal flushes only */
 		ctx->b.dma.flush(ctx, flags, NULL);
 	}
 
 	ctx->gfx_flush_in_progress = true;
 
 	si_preflush_suspend_features(&ctx->b);
 
+	ctx->streamout.suspended = false;
+	if (ctx->streamout.begin_emitted) {
+		si_emit_streamout_end(ctx);
+		ctx->streamout.suspended = true;
+	}
+
 	ctx->b.flags |= SI_CONTEXT_CS_PARTIAL_FLUSH |
 			SI_CONTEXT_PS_PARTIAL_FLUSH;
 
 	/* DRM 3.1.0 doesn't flush TC for VI correctly. */
 	if (ctx->b.chip_class == VI && ctx->b.screen->info.drm_minor <= 1)
 		ctx->b.flags |= SI_CONTEXT_INV_GLOBAL_L2 |
 				SI_CONTEXT_INV_VMEM_L1;
 
 	si_emit_cache_flush(ctx);
 
@@ -236,37 +242,42 @@ void si_begin_new_cs(struct si_context *ctx)
 		si_mark_atom_dirty(ctx, &ctx->sample_mask.atom);
 	si_mark_atom_dirty(ctx, &ctx->cb_render_state);
 	/* CLEAR_STATE sets zeros. */
 	if (!has_clear_state || ctx->blend_color.any_nonzeros)
 		si_mark_atom_dirty(ctx, &ctx->blend_color.atom);
 	si_mark_atom_dirty(ctx, &ctx->db_render_state);
 	if (ctx->b.chip_class >= GFX9)
 		si_mark_atom_dirty(ctx, &ctx->dpbb_state);
 	si_mark_atom_dirty(ctx, &ctx->stencil_ref.atom);
 	si_mark_atom_dirty(ctx, &ctx->spi_map);
-	si_mark_atom_dirty(ctx, &ctx->b.streamout.enable_atom);
+	si_mark_atom_dirty(ctx, &ctx->streamout.enable_atom);
 	si_mark_atom_dirty(ctx, &ctx->b.render_cond_atom);
 	si_all_descriptors_begin_new_cs(ctx);
 	si_all_resident_buffers_begin_new_cs(ctx);
 
 	ctx->scissors.dirty_mask = (1 << SI_MAX_VIEWPORTS) - 1;
 	ctx->viewports.dirty_mask = (1 << SI_MAX_VIEWPORTS) - 1;
 	ctx->viewports.depth_range_dirty_mask = (1 << SI_MAX_VIEWPORTS) - 1;
 	si_mark_atom_dirty(ctx, &ctx->scissors.atom);
 	si_mark_atom_dirty(ctx, &ctx->viewports.atom);
 
 	si_mark_atom_dirty(ctx, &ctx->scratch_state);
 	if (ctx->scratch_buffer) {
 		r600_context_add_resource_size(&ctx->b.b,
 					       &ctx->scratch_buffer->b.b);
 	}
 
+	if (ctx->streamout.suspended) {
+		ctx->streamout.append_bitmask = ctx->streamout.enabled_mask;
+		si_streamout_buffers_dirty(ctx);
+	}
+
 	si_postflush_resume_features(&ctx->b);
 
 	assert(!ctx->b.gfx.cs->prev_dw);
 	ctx->b.initial_gfx_cs_size = ctx->b.gfx.cs->current.cdw;
 
 	/* Invalidate various draw states so that they are emitted before
 	 * the first draw call. */
 	si_invalidate_draw_sh_constants(ctx);
 	ctx->last_index_size = -1;
 	ctx->last_primitive_restart_en = -1;
diff --git a/src/gallium/drivers/radeonsi/si_pipe.c b/src/gallium/drivers/radeonsi/si_pipe.c
index d0b90e7..b9840ad 100644
--- a/src/gallium/drivers/radeonsi/si_pipe.c
+++ b/src/gallium/drivers/radeonsi/si_pipe.c
@@ -198,20 +198,21 @@ static struct pipe_context *si_create_context(struct pipe_screen *screen,
 	if (!si_common_context_init(&sctx->b, &sscreen->b, flags))
 		goto fail;
 
 	if (sscreen->b.info.drm_major == 3)
 		sctx->b.b.get_device_reset_status = si_amdgpu_get_reset_status;
 
 	si_init_blit_functions(sctx);
 	si_init_compute_functions(sctx);
 	si_init_cp_dma_functions(sctx);
 	si_init_debug_functions(sctx);
+	si_init_streamout_functions(sctx);
 
 	if (sscreen->b.info.has_hw_decode) {
 		sctx->b.b.create_video_codec = si_uvd_create_decoder;
 		sctx->b.b.create_video_buffer = si_video_buffer_create;
 	} else {
 		sctx->b.b.create_video_codec = vl_create_decoder;
 		sctx->b.b.create_video_buffer = vl_video_buffer_create;
 	}
 
 	sctx->b.gfx.cs = ws->cs_create(sctx->b.ctx, RING_GFX,
diff --git a/src/gallium/drivers/radeonsi/si_pipe.h b/src/gallium/drivers/radeonsi/si_pipe.h
index cf36100..4e54b7e 100644
--- a/src/gallium/drivers/radeonsi/si_pipe.h
+++ b/src/gallium/drivers/radeonsi/si_pipe.h
@@ -248,20 +248,57 @@ struct si_clip_state {
 struct si_sample_locs {
 	struct r600_atom	atom;
 	unsigned		nr_samples;
 };
 
 struct si_sample_mask {
 	struct r600_atom	atom;
 	uint16_t		sample_mask;
 };
 
+struct si_streamout_target {
+	struct pipe_stream_output_target b;
+
+	/* The buffer where BUFFER_FILLED_SIZE is stored. */
+	struct r600_resource	*buf_filled_size;
+	unsigned		buf_filled_size_offset;
+	bool			buf_filled_size_valid;
+
+	unsigned		stride_in_dw;
+};
+
+struct si_streamout {
+	struct r600_atom		begin_atom;
+	bool				begin_emitted;
+
+	unsigned			enabled_mask;
+	unsigned			num_targets;
+	struct si_streamout_target	*targets[PIPE_MAX_SO_BUFFERS];
+
+	unsigned			append_bitmask;
+	bool				suspended;
+
+	/* External state which comes from the vertex shader,
+	 * it must be set explicitly when binding a shader. */
+	uint16_t			*stride_in_dw;
+	unsigned			enabled_stream_buffers_mask; /* stream0 buffers0-3 in 4 LSB */
+
+	/* The state of VGT_STRMOUT_BUFFER_(CONFIG|EN). */
+	unsigned			hw_enabled_mask;
+
+	/* The state of VGT_STRMOUT_(CONFIG|EN). */
+	struct r600_atom		enable_atom;
+	bool				streamout_enabled;
+	bool				prims_gen_query_enabled;
+	int				num_prims_gen_queries;
+};
+
 /* A shader state consists of the shader selector, which is a constant state
  * object shared by multiple contexts and shouldn't be modified, and
  * the current shader variant selected for this context.
  */
 struct si_shader_ctx_state {
 	struct si_shader_selector	*cso;
 	struct si_shader		*current;
 };
 
 #define SI_NUM_VGT_PARAM_KEY_BITS 12
@@ -352,20 +389,21 @@ struct si_context {
 	struct si_sample_mask		sample_mask;
 	struct r600_atom		cb_render_state;
 	unsigned			last_cb_target_mask;
 	struct si_blend_color		blend_color;
 	struct r600_atom		clip_regs;
 	struct si_clip_state		clip_state;
 	struct si_shader_data		shader_pointers;
 	struct si_stencil_ref		stencil_ref;
 	struct r600_atom		spi_map;
 	struct si_scissors		scissors;
+	struct si_streamout		streamout;
 	struct si_viewports		viewports;
 
 	/* Precomputed states. */
 	struct si_pm4_state		*init_config;
 	struct si_pm4_state		*init_config_gs_rings;
 	bool				init_config_has_vgt_flush;
 	struct si_pm4_state		*vgt_shader_config[4];
 
 	/* shaders */
 	struct si_shader_ctx_state	ps_shader;
@@ -637,20 +675,26 @@ static inline struct tgsi_shader_info *si_get_vs_info(struct si_context *sctx)
 
 static inline struct si_shader* si_get_vs_state(struct si_context *sctx)
 {
 	if (sctx->gs_shader.cso)
 		return sctx->gs_shader.cso->gs_copy_shader;
 
 	struct si_shader_ctx_state *vs = si_get_vs(sctx);
 	return vs->current ? vs->current : NULL;
 }
 
+static inline bool si_get_strmout_en(struct si_context *sctx)
+{
+	return sctx->streamout.streamout_enabled ||
+	       sctx->streamout.prims_gen_query_enabled;
+}
+
 static inline unsigned
 si_optimal_tcc_alignment(struct si_context *sctx, unsigned upload_size)
 {
 	unsigned alignment, tcc_cache_line_size;
 
 	/* If the upload size is less than the cache line size (e.g. 16, 32),
 	 * the whole thing will fit into a cache line if we align it to its size.
 	 * The idea is that multiple small uploads can share a cache line.
 	 * If the upload size is greater, align it to the cache line size.
 	 */
diff --git a/src/gallium/drivers/radeonsi/si_state.c b/src/gallium/drivers/radeonsi/si_state.c
index 99c3ca3..82f3962 100644
--- a/src/gallium/drivers/radeonsi/si_state.c
+++ b/src/gallium/drivers/radeonsi/si_state.c
@@ -4400,22 +4400,22 @@ static void si_need_gfx_cs_space(struct pipe_context *ctx, unsigned num_dw,
 				 bool include_draw_vbo)
 {
 	si_need_cs_space((struct si_context*)ctx);
 }
 
 static void si_init_config(struct si_context *sctx);
 
 void si_init_state_functions(struct si_context *sctx)
 {
 	si_init_external_atom(sctx, &sctx->b.render_cond_atom, &sctx->atoms.s.render_cond);
-	si_init_external_atom(sctx, &sctx->b.streamout.begin_atom, &sctx->atoms.s.streamout_begin);
-	si_init_external_atom(sctx, &sctx->b.streamout.enable_atom, &sctx->atoms.s.streamout_enable);
+	si_init_external_atom(sctx, &sctx->streamout.begin_atom, &sctx->atoms.s.streamout_begin);
+	si_init_external_atom(sctx, &sctx->streamout.enable_atom, &sctx->atoms.s.streamout_enable);
 	si_init_external_atom(sctx, &sctx->scissors.atom, &sctx->atoms.s.scissors);
 	si_init_external_atom(sctx, &sctx->viewports.atom, &sctx->atoms.s.viewports);
 
 	si_init_atom(sctx, &sctx->framebuffer.atom, &sctx->atoms.s.framebuffer, si_emit_framebuffer_state);
 	si_init_atom(sctx, &sctx->msaa_sample_locs.atom, &sctx->atoms.s.msaa_sample_locs, si_emit_msaa_sample_locs);
 	si_init_atom(sctx, &sctx->db_render_state, &sctx->atoms.s.db_render_state, si_emit_db_render_state);
 	si_init_atom(sctx, &sctx->dpbb_state, &sctx->atoms.s.dpbb_state, si_emit_dpbb_state);
 	si_init_atom(sctx, &sctx->msaa_config, &sctx->atoms.s.msaa_config, si_emit_msaa_config);
 	si_init_atom(sctx, &sctx->sample_mask.atom, &sctx->atoms.s.sample_mask, si_emit_sample_mask);
 	si_init_atom(sctx, &sctx->cb_render_state, &sctx->atoms.s.cb_render_state, si_emit_cb_render_state);
diff --git a/src/gallium/drivers/radeonsi/si_state.h b/src/gallium/drivers/radeonsi/si_state.h
index 03e2a17..9d29878 100644
--- a/src/gallium/drivers/radeonsi/si_state.h
+++ b/src/gallium/drivers/radeonsi/si_state.h
@@ -416,20 +416,31 @@ void si_emit_cache_flush(struct si_context *sctx);
 void si_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *dinfo);
 void si_draw_rectangle(struct blitter_context *blitter,
 		       void *vertex_elements_cso,
 		       blitter_get_vs_func get_vs,
 		       int x1, int y1, int x2, int y2,
 		       float depth, unsigned num_instances,
 		       enum blitter_attrib_type type,
 		       const union blitter_attrib *attrib);
 void si_trace_emit(struct si_context *sctx);
 
+/* si_state_streamout.c */
+void si_streamout_buffers_dirty(struct si_context *sctx);
+void si_common_set_streamout_targets(struct pipe_context *ctx,
+				     unsigned num_targets,
+				     struct pipe_stream_output_target **targets,
+				     const unsigned *offset);
+void si_emit_streamout_end(struct si_context *sctx);
+void si_update_prims_generated_query_state(struct si_context *sctx,
+					   unsigned type, int diff);
+void si_init_streamout_functions(struct si_context *sctx);
+
 
 static inline unsigned
 si_tile_mode_index(struct r600_texture *rtex, unsigned level, bool stencil)
 {
 	if (stencil)
 		return rtex->surface.u.legacy.stencil_tiling_index[level];
 	else
 		return rtex->surface.u.legacy.tiling_index[level];
 }
 
diff --git a/src/gallium/drivers/radeonsi/si_state_draw.c b/src/gallium/drivers/radeonsi/si_state_draw.c
index 6eab4cb..9468fde 100644
--- a/src/gallium/drivers/radeonsi/si_state_draw.c
+++ b/src/gallium/drivers/radeonsi/si_state_draw.c
@@ -645,22 +645,22 @@ static void si_emit_draw_packets(struct si_context *sctx,
 				 unsigned index_offset)
 {
 	struct pipe_draw_indirect_info *indirect = info->indirect;
 	struct radeon_winsys_cs *cs = sctx->b.gfx.cs;
 	unsigned sh_base_reg = sctx->shader_pointers.sh_base[PIPE_SHADER_VERTEX];
 	bool render_cond_bit = sctx->b.render_cond && !sctx->b.render_cond_force_off;
 	uint32_t index_max_size = 0;
 	uint64_t index_va = 0;
 
 	if (info->count_from_stream_output) {
-		struct r600_so_target *t =
-			(struct r600_so_target*)info->count_from_stream_output;
+		struct si_streamout_target *t =
+			(struct si_streamout_target*)info->count_from_stream_output;
 		uint64_t va = t->buf_filled_size->gpu_address +
 			      t->buf_filled_size_offset;
 
 		radeon_set_context_reg(cs, R_028B30_VGT_STRMOUT_DRAW_OPAQUE_VERTEX_STRIDE,
 				       t->stride_in_dw);
 
 		radeon_emit(cs, PKT3(PKT3_COPY_DATA, 4, 0));
 		radeon_emit(cs, COPY_DATA_SRC_SEL(COPY_DATA_MEM) |
 			    COPY_DATA_DST_SEL(COPY_DATA_REG) |
 			    COPY_DATA_WR_CONFIRM);
@@ -1479,21 +1479,21 @@ void si_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *info)
 	if (unlikely(sctx->current_saved_cs)) {
 		si_trace_emit(sctx);
 		si_log_draw_state(sctx, sctx->b.log);
 	}
 
 	/* Workaround for a VGT hang when streamout is enabled.
 	 * It must be done after drawing. */
 	if ((sctx->b.family == CHIP_HAWAII ||
 	     sctx->b.family == CHIP_TONGA ||
 	     sctx->b.family == CHIP_FIJI) &&
-	    r600_get_strmout_en(&sctx->b)) {
+	    si_get_strmout_en(sctx)) {
 		sctx->b.flags |= SI_CONTEXT_VGT_STREAMOUT_SYNC;
 	}
 
 	if (unlikely(sctx->decompression_enabled)) {
 		sctx->b.num_decompress_calls++;
 	} else {
 		sctx->b.num_draw_calls++;
 		if (sctx->framebuffer.state.nr_cbufs > 1)
 			sctx->b.num_mrt_draw_calls++;
 		if (info->primitive_restart)
diff --git a/src/gallium/drivers/radeonsi/si_state_shaders.c b/src/gallium/drivers/radeonsi/si_state_shaders.c
index dbaa2dc..9340328 100644
--- a/src/gallium/drivers/radeonsi/si_state_shaders.c
+++ b/src/gallium/drivers/radeonsi/si_state_shaders.c
@@ -2245,23 +2245,23 @@ static void *si_create_shader_selector(struct pipe_context *ctx,
 	return sel;
 }
 
 static void si_update_streamout_state(struct si_context *sctx)
 {
 	struct si_shader_selector *shader_with_so = si_get_vs(sctx)->cso;
 
 	if (!shader_with_so)
 		return;
 
-	sctx->b.streamout.enabled_stream_buffers_mask =
+	sctx->streamout.enabled_stream_buffers_mask =
 		shader_with_so->enabled_streamout_buffer_mask;
-	sctx->b.streamout.stride_in_dw = shader_with_so->so.stride;
+	sctx->streamout.stride_in_dw = shader_with_so->so.stride;
 }
 
 static void si_update_clip_regs(struct si_context *sctx,
 				struct si_shader_selector *old_hw_vs,
 				struct si_shader *old_hw_vs_variant,
 				struct si_shader_selector *next_hw_vs,
 				struct si_shader *next_hw_vs_variant)
 {
 	if (next_hw_vs &&
 	    (!old_hw_vs ||
diff --git a/src/gallium/drivers/radeon/r600_streamout.c b/src/gallium/drivers/radeonsi/si_state_streamout.c
similarity index 58%
rename from src/gallium/drivers/radeon/r600_streamout.c
rename to src/gallium/drivers/radeonsi/si_state_streamout.c
index 5c14b1b..42a83d4 100644
--- a/src/gallium/drivers/radeon/r600_streamout.c
+++ b/src/gallium/drivers/radeonsi/si_state_streamout.c
@@ -17,294 +17,296 @@
  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  * THE AUTHORS OR COPYRIGHT HOLDERS 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: Marek Olšák <maraeo at gmail.com>
  *
  */
 
-#include "r600_pipe_common.h"
-#include "r600_cs.h"
+#include "si_pipe.h"
+#include "si_state.h"
+#include "radeon/r600_cs.h"
 
 #include "util/u_memory.h"
 
-static void r600_set_streamout_enable(struct r600_common_context *rctx, bool enable);
+static void si_set_streamout_enable(struct si_context *sctx, bool enable);
 
 static struct pipe_stream_output_target *
-r600_create_so_target(struct pipe_context *ctx,
-		      struct pipe_resource *buffer,
-		      unsigned buffer_offset,
-		      unsigned buffer_size)
+si_create_so_target(struct pipe_context *ctx,
+		    struct pipe_resource *buffer,
+		    unsigned buffer_offset,
+		    unsigned buffer_size)
 {
-	struct r600_common_context *rctx = (struct r600_common_context *)ctx;
-	struct r600_so_target *t;
+	struct si_context *sctx = (struct si_context *)ctx;
+	struct si_streamout_target *t;
 	struct r600_resource *rbuffer = (struct r600_resource*)buffer;
 
-	t = CALLOC_STRUCT(r600_so_target);
+	t = CALLOC_STRUCT(si_streamout_target);
 	if (!t) {
 		return NULL;
 	}
 
-	u_suballocator_alloc(rctx->allocator_zeroed_memory, 4, 4,
+	u_suballocator_alloc(sctx->b.allocator_zeroed_memory, 4, 4,
 			     &t->buf_filled_size_offset,
 			     (struct pipe_resource**)&t->buf_filled_size);
 	if (!t->buf_filled_size) {
 		FREE(t);
 		return NULL;
 	}
 
 	t->b.reference.count = 1;
 	t->b.context = ctx;
 	pipe_resource_reference(&t->b.buffer, buffer);
 	t->b.buffer_offset = buffer_offset;
 	t->b.buffer_size = buffer_size;
 
 	util_range_add(&rbuffer->valid_buffer_range, buffer_offset,
 		       buffer_offset + buffer_size);
 	return &t->b;
 }
 
-static void r600_so_target_destroy(struct pipe_context *ctx,
-				   struct pipe_stream_output_target *target)
+static void si_so_target_destroy(struct pipe_context *ctx,
+				 struct pipe_stream_output_target *target)
 {
-	struct r600_so_target *t = (struct r600_so_target*)target;
+	struct si_streamout_target *t = (struct si_streamout_target*)target;
 	pipe_resource_reference(&t->b.buffer, NULL);
 	r600_resource_reference(&t->buf_filled_size, NULL);
 	FREE(t);
 }
 
-void si_streamout_buffers_dirty(struct r600_common_context *rctx)
+void si_streamout_buffers_dirty(struct si_context *sctx)
 {
-	if (!rctx->streamout.enabled_mask)
+	if (!sctx->streamout.enabled_mask)
 		return;
 
-	rctx->set_atom_dirty(rctx, &rctx->streamout.begin_atom, true);
-	r600_set_streamout_enable(rctx, true);
+	si_mark_atom_dirty(sctx, &sctx->streamout.begin_atom);
+	si_set_streamout_enable(sctx, true);
 }
 
 void si_common_set_streamout_targets(struct pipe_context *ctx,
 				     unsigned num_targets,
 				     struct pipe_stream_output_target **targets,
 				     const unsigned *offsets)
 {
-	struct r600_common_context *rctx = (struct r600_common_context *)ctx;
+	struct si_context *sctx = (struct si_context *)ctx;
 	unsigned i;
         unsigned enabled_mask = 0, append_bitmask = 0;
 
 	/* Stop streamout. */
-	if (rctx->streamout.num_targets && rctx->streamout.begin_emitted) {
-		si_emit_streamout_end(rctx);
+	if (sctx->streamout.num_targets && sctx->streamout.begin_emitted) {
+		si_emit_streamout_end(sctx);
 	}
 
 	/* Set the new targets. */
 	for (i = 0; i < num_targets; i++) {
-		pipe_so_target_reference((struct pipe_stream_output_target**)&rctx->streamout.targets[i], targets[i]);
+		pipe_so_target_reference((struct pipe_stream_output_target**)&sctx->streamout.targets[i], targets[i]);
 		if (!targets[i])
 			continue;
 
 		r600_context_add_resource_size(ctx, targets[i]->buffer);
 		enabled_mask |= 1 << i;
 		if (offsets[i] == ((unsigned)-1))
 			append_bitmask |= 1 << i;
 	}
-	for (; i < rctx->streamout.num_targets; i++) {
-		pipe_so_target_reference((struct pipe_stream_output_target**)&rctx->streamout.targets[i], NULL);
+	for (; i < sctx->streamout.num_targets; i++) {
+		pipe_so_target_reference((struct pipe_stream_output_target**)&sctx->streamout.targets[i], NULL);
 	}
 
-	rctx->streamout.enabled_mask = enabled_mask;
+	sctx->streamout.enabled_mask = enabled_mask;
 
-	rctx->streamout.num_targets = num_targets;
-	rctx->streamout.append_bitmask = append_bitmask;
+	sctx->streamout.num_targets = num_targets;
+	sctx->streamout.append_bitmask = append_bitmask;
 
 	if (num_targets) {
-		si_streamout_buffers_dirty(rctx);
+		si_streamout_buffers_dirty(sctx);
 	} else {
-		rctx->set_atom_dirty(rctx, &rctx->streamout.begin_atom, false);
-		r600_set_streamout_enable(rctx, false);
+		si_set_atom_dirty(sctx, &sctx->streamout.begin_atom, false);
+		si_set_streamout_enable(sctx, false);
 	}
 }
 
-static void r600_flush_vgt_streamout(struct r600_common_context *rctx)
+static void si_flush_vgt_streamout(struct si_context *sctx)
 {
-	struct radeon_winsys_cs *cs = rctx->gfx.cs;
+	struct radeon_winsys_cs *cs = sctx->b.gfx.cs;
 	unsigned reg_strmout_cntl;
 
 	/* The register is at different places on different ASICs. */
-	if (rctx->chip_class >= CIK) {
+	if (sctx->b.chip_class >= CIK) {
 		reg_strmout_cntl = R_0300FC_CP_STRMOUT_CNTL;
 		radeon_set_uconfig_reg(cs, reg_strmout_cntl, 0);
 	} else {
 		reg_strmout_cntl = R_0084FC_CP_STRMOUT_CNTL;
 		radeon_set_config_reg(cs, reg_strmout_cntl, 0);
 	}
 
 	radeon_emit(cs, PKT3(PKT3_EVENT_WRITE, 0, 0));
 	radeon_emit(cs, EVENT_TYPE(EVENT_TYPE_SO_VGTSTREAMOUT_FLUSH) | EVENT_INDEX(0));
 
 	radeon_emit(cs, PKT3(PKT3_WAIT_REG_MEM, 5, 0));
 	radeon_emit(cs, WAIT_REG_MEM_EQUAL); /* wait until the register is equal to the reference value */
 	radeon_emit(cs, reg_strmout_cntl >> 2);  /* register */
 	radeon_emit(cs, 0);
 	radeon_emit(cs, S_008490_OFFSET_UPDATE_DONE(1)); /* reference value */
 	radeon_emit(cs, S_008490_OFFSET_UPDATE_DONE(1)); /* mask */
 	radeon_emit(cs, 4); /* poll interval */
 }
 
-static void r600_emit_streamout_begin(struct r600_common_context *rctx, struct r600_atom *atom)
+static void si_emit_streamout_begin(struct r600_common_context *rctx, struct r600_atom *atom)
 {
-	struct radeon_winsys_cs *cs = rctx->gfx.cs;
-	struct r600_so_target **t = rctx->streamout.targets;
-	uint16_t *stride_in_dw = rctx->streamout.stride_in_dw;
+	struct si_context *sctx = (struct si_context*)rctx;
+	struct radeon_winsys_cs *cs = sctx->b.gfx.cs;
+	struct si_streamout_target **t = sctx->streamout.targets;
+	uint16_t *stride_in_dw = sctx->streamout.stride_in_dw;
 	unsigned i;
 
-	r600_flush_vgt_streamout(rctx);
+	si_flush_vgt_streamout(sctx);
 
-	for (i = 0; i < rctx->streamout.num_targets; i++) {
+	for (i = 0; i < sctx->streamout.num_targets; i++) {
 		if (!t[i])
 			continue;
 
 		t[i]->stride_in_dw = stride_in_dw[i];
 
 		/* SI binds streamout buffers as shader resources.
 		 * VGT only counts primitives and tells the shader
 		 * through SGPRs what to do. */
 		radeon_set_context_reg_seq(cs, R_028AD0_VGT_STRMOUT_BUFFER_SIZE_0 + 16*i, 2);
 		radeon_emit(cs, (t[i]->b.buffer_offset +
 				 t[i]->b.buffer_size) >> 2);	/* BUFFER_SIZE (in DW) */
 		radeon_emit(cs, stride_in_dw[i]);		/* VTX_STRIDE (in DW) */
 
-		if (rctx->streamout.append_bitmask & (1 << i) && t[i]->buf_filled_size_valid) {
+		if (sctx->streamout.append_bitmask & (1 << i) && t[i]->buf_filled_size_valid) {
 			uint64_t va = t[i]->buf_filled_size->gpu_address +
 				      t[i]->buf_filled_size_offset;
 
 			/* Append. */
 			radeon_emit(cs, PKT3(PKT3_STRMOUT_BUFFER_UPDATE, 4, 0));
 			radeon_emit(cs, STRMOUT_SELECT_BUFFER(i) |
 				    STRMOUT_OFFSET_SOURCE(STRMOUT_OFFSET_FROM_MEM)); /* control */
 			radeon_emit(cs, 0); /* unused */
 			radeon_emit(cs, 0); /* unused */
 			radeon_emit(cs, va); /* src address lo */
 			radeon_emit(cs, va >> 32); /* src address hi */
 
-			r600_emit_reloc(rctx,  &rctx->gfx, t[i]->buf_filled_size,
+			r600_emit_reloc(&sctx->b,  &sctx->b.gfx, t[i]->buf_filled_size,
 					RADEON_USAGE_READ, RADEON_PRIO_SO_FILLED_SIZE);
 		} else {
 			/* Start from the beginning. */
 			radeon_emit(cs, PKT3(PKT3_STRMOUT_BUFFER_UPDATE, 4, 0));
 			radeon_emit(cs, STRMOUT_SELECT_BUFFER(i) |
 				    STRMOUT_OFFSET_SOURCE(STRMOUT_OFFSET_FROM_PACKET)); /* control */
 			radeon_emit(cs, 0); /* unused */
 			radeon_emit(cs, 0); /* unused */
 			radeon_emit(cs, t[i]->b.buffer_offset >> 2); /* buffer offset in DW */
 			radeon_emit(cs, 0); /* unused */
 		}
 	}
 
-	rctx->streamout.begin_emitted = true;
+	sctx->streamout.begin_emitted = true;
 }
 
-void si_emit_streamout_end(struct r600_common_context *rctx)
+void si_emit_streamout_end(struct si_context *sctx)
 {
-	struct radeon_winsys_cs *cs = rctx->gfx.cs;
-	struct r600_so_target **t = rctx->streamout.targets;
+	struct radeon_winsys_cs *cs = sctx->b.gfx.cs;
+	struct si_streamout_target **t = sctx->streamout.targets;
 	unsigned i;
 	uint64_t va;
 
-	r600_flush_vgt_streamout(rctx);
+	si_flush_vgt_streamout(sctx);
 
-	for (i = 0; i < rctx->streamout.num_targets; i++) {
+	for (i = 0; i < sctx->streamout.num_targets; i++) {
 		if (!t[i])
 			continue;
 
 		va = t[i]->buf_filled_size->gpu_address + t[i]->buf_filled_size_offset;
 		radeon_emit(cs, PKT3(PKT3_STRMOUT_BUFFER_UPDATE, 4, 0));
 		radeon_emit(cs, STRMOUT_SELECT_BUFFER(i) |
 			    STRMOUT_OFFSET_SOURCE(STRMOUT_OFFSET_NONE) |
 			    STRMOUT_STORE_BUFFER_FILLED_SIZE); /* control */
 		radeon_emit(cs, va);     /* dst address lo */
 		radeon_emit(cs, va >> 32); /* dst address hi */
 		radeon_emit(cs, 0); /* unused */
 		radeon_emit(cs, 0); /* unused */
 
-		r600_emit_reloc(rctx,  &rctx->gfx, t[i]->buf_filled_size,
+		r600_emit_reloc(&sctx->b,  &sctx->b.gfx, t[i]->buf_filled_size,
 				RADEON_USAGE_WRITE, RADEON_PRIO_SO_FILLED_SIZE);
 
 		/* Zero the buffer size. The counters (primitives generated,
 		 * primitives emitted) may be enabled even if there is not
 		 * buffer bound. This ensures that the primitives-emitted query
 		 * won't increment. */
 		radeon_set_context_reg(cs, R_028AD0_VGT_STRMOUT_BUFFER_SIZE_0 + 16*i, 0);
 
 		t[i]->buf_filled_size_valid = true;
 	}
 
-	rctx->streamout.begin_emitted = false;
-	rctx->flags |= R600_CONTEXT_STREAMOUT_FLUSH;
+	sctx->streamout.begin_emitted = false;
+	sctx->b.flags |= R600_CONTEXT_STREAMOUT_FLUSH;
 }
 
 /* STREAMOUT CONFIG DERIVED STATE
  *
  * Streamout must be enabled for the PRIMITIVES_GENERATED query to work.
  * The buffer mask is an independent state, so no writes occur if there
  * are no buffers bound.
  */
 
-static void r600_emit_streamout_enable(struct r600_common_context *rctx,
-				       struct r600_atom *atom)
+static void si_emit_streamout_enable(struct r600_common_context *rctx,
+				     struct r600_atom *atom)
 {
-	radeon_set_context_reg_seq(rctx->gfx.cs, R_028B94_VGT_STRMOUT_CONFIG, 2);
-	radeon_emit(rctx->gfx.cs,
-		    S_028B94_STREAMOUT_0_EN(r600_get_strmout_en(rctx)) |
+	struct si_context *sctx = (struct si_context*)rctx;
+
+	radeon_set_context_reg_seq(sctx->b.gfx.cs, R_028B94_VGT_STRMOUT_CONFIG, 2);
+	radeon_emit(sctx->b.gfx.cs,
+		    S_028B94_STREAMOUT_0_EN(si_get_strmout_en(sctx)) |
 		    S_028B94_RAST_STREAM(0) |
-		    S_028B94_STREAMOUT_1_EN(r600_get_strmout_en(rctx)) |
-		    S_028B94_STREAMOUT_2_EN(r600_get_strmout_en(rctx)) |
-		    S_028B94_STREAMOUT_3_EN(r600_get_strmout_en(rctx)));
-	radeon_emit(rctx->gfx.cs,
-		    rctx->streamout.hw_enabled_mask &
-		    rctx->streamout.enabled_stream_buffers_mask);
+		    S_028B94_STREAMOUT_1_EN(si_get_strmout_en(sctx)) |
+		    S_028B94_STREAMOUT_2_EN(si_get_strmout_en(sctx)) |
+		    S_028B94_STREAMOUT_3_EN(si_get_strmout_en(sctx)));
+	radeon_emit(sctx->b.gfx.cs,
+		    sctx->streamout.hw_enabled_mask &
+		    sctx->streamout.enabled_stream_buffers_mask);
 }
 
-static void r600_set_streamout_enable(struct r600_common_context *rctx, bool enable)
+static void si_set_streamout_enable(struct si_context *sctx, bool enable)
 {
-	bool old_strmout_en = r600_get_strmout_en(rctx);
-	unsigned old_hw_enabled_mask = rctx->streamout.hw_enabled_mask;
+	bool old_strmout_en = si_get_strmout_en(sctx);
+	unsigned old_hw_enabled_mask = sctx->streamout.hw_enabled_mask;
 
-	rctx->streamout.streamout_enabled = enable;
+	sctx->streamout.streamout_enabled = enable;
 
-	rctx->streamout.hw_enabled_mask = rctx->streamout.enabled_mask |
-					  (rctx->streamout.enabled_mask << 4) |
-					  (rctx->streamout.enabled_mask << 8) |
-					  (rctx->streamout.enabled_mask << 12);
+	sctx->streamout.hw_enabled_mask = sctx->streamout.enabled_mask |
+					  (sctx->streamout.enabled_mask << 4) |
+					  (sctx->streamout.enabled_mask << 8) |
+					  (sctx->streamout.enabled_mask << 12);
 
-	if ((old_strmout_en != r600_get_strmout_en(rctx)) ||
-            (old_hw_enabled_mask != rctx->streamout.hw_enabled_mask)) {
-		rctx->set_atom_dirty(rctx, &rctx->streamout.enable_atom, true);
-	}
+	if ((old_strmout_en != si_get_strmout_en(sctx)) ||
+            (old_hw_enabled_mask != sctx->streamout.hw_enabled_mask))
+		si_mark_atom_dirty(sctx, &sctx->streamout.enable_atom);
 }
 
-void si_update_prims_generated_query_state(struct r600_common_context *rctx,
+void si_update_prims_generated_query_state(struct si_context *sctx,
 					   unsigned type, int diff)
 {
 	if (type == PIPE_QUERY_PRIMITIVES_GENERATED) {
-		bool old_strmout_en = r600_get_strmout_en(rctx);
+		bool old_strmout_en = si_get_strmout_en(sctx);
 
-		rctx->streamout.num_prims_gen_queries += diff;
-		assert(rctx->streamout.num_prims_gen_queries >= 0);
+		sctx->streamout.num_prims_gen_queries += diff;
+		assert(sctx->streamout.num_prims_gen_queries >= 0);
 
-		rctx->streamout.prims_gen_query_enabled =
-			rctx->streamout.num_prims_gen_queries != 0;
+		sctx->streamout.prims_gen_query_enabled =
+			sctx->streamout.num_prims_gen_queries != 0;
 
-		if (old_strmout_en != r600_get_strmout_en(rctx)) {
-			rctx->set_atom_dirty(rctx, &rctx->streamout.enable_atom, true);
-		}
+		if (old_strmout_en != si_get_strmout_en(sctx))
+			si_mark_atom_dirty(sctx, &sctx->streamout.enable_atom);
 	}
 }
 
-void si_streamout_init(struct r600_common_context *rctx)
+void si_init_streamout_functions(struct si_context *sctx)
 {
-	rctx->b.create_stream_output_target = r600_create_so_target;
-	rctx->b.stream_output_target_destroy = r600_so_target_destroy;
-	rctx->streamout.begin_atom.emit = r600_emit_streamout_begin;
-	rctx->streamout.enable_atom.emit = r600_emit_streamout_enable;
+	sctx->b.b.create_stream_output_target = si_create_so_target;
+	sctx->b.b.stream_output_target_destroy = si_so_target_destroy;
+	sctx->streamout.begin_atom.emit = si_emit_streamout_begin;
+	sctx->streamout.enable_atom.emit = si_emit_streamout_enable;
 }
-- 
2.7.4



More information about the mesa-dev mailing list