XxX: IMgae WIP

6 files changed, 77 insertions, 36 deletions

diff --git a/src/gallium/drivers/r600/evergreen_compute.c b/src/gallium/drivers/r600/evergreen_compute.c
index f76fc9cc25e..e8833c01fb1 100644
--- a/src/gallium/drivers/r600/evergreen_compute.c
+++ b/src/gallium/drivers/r600/evergreen_compute.c
@@ -462,6 +462,9 @@ static void compute_emit_cs(struct r600_context *ctx, const uint *block_layout,
 	/* Emit constant buffer state */
 	r600_emit_atom(ctx, &ctx->constbuf_state[PIPE_SHADER_COMPUTE].atom);
 
+	/* Emit sampler view state */
+	r600_emit_atom(ctx, &ctx->samplers[PIPE_SHADER_COMPUTE].views.atom);
+
 	/* Emit compute shader state */
 	r600_emit_atom(ctx, &ctx->cs_shader_state.atom);
 
@@ -580,29 +583,31 @@ static void evergreen_set_compute_resources(struct pipe_context * ctx_,
 	struct r600_context *ctx = (struct r600_context *)ctx_;
 	struct r600_surface **resources = (struct r600_surface **)surfaces;
 
+	if (!surfaces) { return; }
 	COMPUTE_DBG(ctx->screen, "*** evergreen_set_compute_resources: start = %u count = %u\n",
 			start, count);
 
 	for (int i = 0; i < count; i++)	{
-		/* The First two vertex buffers are reserved for parameters and
-		 * global buffers. */
-		unsigned vtx_id = 2 + i;
-		if (resources[i]) {
-			struct r600_resource_global *buffer =
-				(struct r600_resource_global*)
-				resources[i]->base.texture;
-			if (resources[i]->base.writable) {
-				assert(i+1 < 12);
-
-				evergreen_set_rat(ctx->cs_shader_state.shader, i+1,
-				(struct r600_resource *)resources[i]->base.texture,
-				buffer->chunk->start_in_dw*4,
-				resources[i]->base.texture->width0);
+		struct pipe_surface *surface = surfaces[i];
+		if (surface) {
+			struct r600_surface *rsurface =
+					(struct r600_surface*)surface;
+			struct pipe_resource *tex = surface->texture;
+			if (tex->target == PIPE_TEXTURE_2D) {
+				#if 0
+				pipe_sampler_view info;
+				struct pipe_sampler_view *sampler_view;
+				// Create the sampler view
+				u_sampler_view_default_template(&info, tex, tex->format);
+				sampler_view = ctx_->create_sampler_view(ctx_, tex, &info);
+				#endif
+				// Setup the color surface as a RAT
+				ctx->framebuffer.state.cbufs[1 + i] = surface;
+				evergreen_init_color_surface_rat(ctx, rsurface);
+				ctx->compute_cb_target_mask |= (0xf << (i * 4));
+				ctx->framebuffer.state.nr_cbufs =
+					MAX2(i + 1, ctx->framebuffer.state.nr_cbufs);
 			}
-
-			evergreen_cs_set_vertex_buffer(ctx, vtx_id,
-					buffer->chunk->start_in_dw * 4,
-					resources[i]->base.texture);
 		}
 	}
 }
@@ -631,7 +636,7 @@ static void evergreen_bind_compute_sampler_states(
 	unsigned num_samplers,
 	void **samplers_)
 {
-	struct compute_sampler_state ** samplers =
+	struct compute_sampler_state **samplers =
 		(struct compute_sampler_state **)samplers_;
 
 	for (int i = 0; i < num_samplers; i++) {
@@ -874,8 +879,8 @@ void evergreen_init_compute_state_functions(struct r600_context *ctx)
 	ctx->context.bind_compute_state = evergreen_bind_compute_state;
 //	 ctx->context.create_sampler_view = evergreen_compute_create_sampler_view;
 	ctx->context.set_compute_resources = evergreen_set_compute_resources;
-	ctx->context.set_compute_sampler_views = evergreen_set_cs_sampler_view;
-	ctx->context.bind_compute_sampler_states = evergreen_bind_compute_sampler_states;
+//	ctx->context.set_compute_sampler_views = evergreen_set_cs_sampler_view;
+//	ctx->context.bind_compute_sampler_states = evergreen_bind_compute_sampler_states;
 	ctx->context.set_global_binding = evergreen_set_global_binding;
 	ctx->context.launch_grid = evergreen_launch_grid;
 
diff --git a/src/gallium/drivers/r600/evergreen_state.c b/src/gallium/drivers/r600/evergreen_state.c
index 2d8aad86af3..09aadada8c6 100644
--- a/src/gallium/drivers/r600/evergreen_state.c
+++ b/src/gallium/drivers/r600/evergreen_state.c
@@ -2574,7 +2574,8 @@ static void evergreen_emit_cs_constant_buffers(struct r600_context *rctx, struct
 
 static void evergreen_emit_sampler_views(struct r600_context *rctx,
 					 struct r600_samplerview_state *state,
-					 unsigned resource_id_base)
+					 unsigned resource_id_base,
+					 unsigned pkt_flags)
 {
 	struct radeon_winsys_cs *cs = rctx->rings.gfx.cs;
 	uint32_t dirty_mask = state->dirty_mask;
@@ -2587,17 +2588,17 @@ static void evergreen_emit_sampler_views(struct r600_context *rctx,
 		rview = state->views[resource_index];
 		assert(rview);
 
-		r600_write_value(cs, PKT3(PKT3_SET_RESOURCE, 8, 0));
+		r600_write_value(cs, PKT3(PKT3_SET_RESOURCE, 8, 0) | pkt_flags);
 		r600_write_value(cs, (resource_id_base + resource_index) * 8);
 		r600_write_array(cs, 8, rview->tex_resource_words);
 
 		reloc = r600_context_bo_reloc(rctx, &rctx->rings.gfx, rview->tex_resource,
 					      RADEON_USAGE_READ);
-		r600_write_value(cs, PKT3(PKT3_NOP, 0, 0));
+		r600_write_value(cs, PKT3(PKT3_NOP, 0, 0) | pkt_flags);
 		r600_write_value(cs, reloc);
 
 		if (!rview->skip_mip_address_reloc) {
-			r600_write_value(cs, PKT3(PKT3_NOP, 0, 0));
+			r600_write_value(cs, PKT3(PKT3_NOP, 0, 0) | pkt_flags);
 			r600_write_value(cs, reloc);
 		}
 	}
@@ -2606,17 +2607,29 @@ static void evergreen_emit_sampler_views(struct r600_context *rctx,
 
 static void evergreen_emit_vs_sampler_views(struct r600_context *rctx, struct r600_atom *atom)
 {
-	evergreen_emit_sampler_views(rctx, &rctx->samplers[PIPE_SHADER_VERTEX].views, 176 + R600_MAX_CONST_BUFFERS);
+	evergreen_emit_sampler_views(rctx, &rctx->samplers[PIPE_SHADER_VERTEX].views,
+				     176 + R600_MAX_CONST_BUFFERS,
+				     0 /* PKT3 flags */);
 }
 
 static void evergreen_emit_gs_sampler_views(struct r600_context *rctx, struct r600_atom *atom)
 {
-	evergreen_emit_sampler_views(rctx, &rctx->samplers[PIPE_SHADER_GEOMETRY].views, 336 + R600_MAX_CONST_BUFFERS);
+	evergreen_emit_sampler_views(rctx, &rctx->samplers[PIPE_SHADER_GEOMETRY].views,
+				     336 + R600_MAX_CONST_BUFFERS,
+				     0 /* PKT3 flags */);
 }
 
 static void evergreen_emit_ps_sampler_views(struct r600_context *rctx, struct r600_atom *atom)
 {
-	evergreen_emit_sampler_views(rctx, &rctx->samplers[PIPE_SHADER_FRAGMENT].views, R600_MAX_CONST_BUFFERS);
+	evergreen_emit_sampler_views(rctx, &rctx->samplers[PIPE_SHADER_FRAGMENT].views,
+				     R600_MAX_CONST_BUFFERS,
+				     0 /* PKT3 flags */);
+}
+
+static void evergreen_emit_cs_sampler_views(struct r600_context *rctx, struct r600_atom *atom)
+{
+	evergreen_emit_sampler_views(rctx, &rctx->samplers[PIPE_SHADER_COMPUTE].views,
+				     816 + 2, RADEON_CP_PACKET3_COMPUTE_MODE);
 }
 
 static void evergreen_emit_sampler_states(struct r600_context *rctx,
@@ -3840,6 +3853,7 @@ void evergreen_init_state_functions(struct r600_context *rctx)
 	r600_init_atom(rctx, &rctx->samplers[PIPE_SHADER_VERTEX].views.atom, id++, evergreen_emit_vs_sampler_views, 0);
 	r600_init_atom(rctx, &rctx->samplers[PIPE_SHADER_GEOMETRY].views.atom, id++, evergreen_emit_gs_sampler_views, 0);
 	r600_init_atom(rctx, &rctx->samplers[PIPE_SHADER_FRAGMENT].views.atom, id++, evergreen_emit_ps_sampler_views, 0);
+	r600_init_atom(rctx, &rctx->samplers[PIPE_SHADER_COMPUTE].views.atom, id++, evergreen_emit_cs_sampler_views, 0);
 
 	r600_init_atom(rctx, &rctx->vgt_state.atom, id++, r600_emit_vgt_state, 7);
 
diff --git a/src/gallium/drivers/r600/r600_pipe.h b/src/gallium/drivers/r600/r600_pipe.h
index 349a6cb6cfb..683f48d3bbb 100644
--- a/src/gallium/drivers/r600/r600_pipe.h
+++ b/src/gallium/drivers/r600/r600_pipe.h
@@ -35,7 +35,7 @@
 #include "r600_public.h"
 #include "r600_resource.h"
 
-#define R600_NUM_ATOMS 41
+#define R600_NUM_ATOMS 42
 
 /* the number of CS dwords for flushing and drawing */
 #define R600_MAX_FLUSH_CS_DWORDS	16
diff --git a/src/gallium/drivers/r600/r600_state_common.c b/src/gallium/drivers/r600/r600_state_common.c
index ea5a4e7627c..a9156c13af7 100644
--- a/src/gallium/drivers/r600/r600_state_common.c
+++ b/src/gallium/drivers/r600/r600_state_common.c
@@ -447,6 +447,11 @@ static void r600_bind_ps_sampler_states(struct pipe_context *ctx, unsigned count
 	r600_bind_sampler_states(ctx, PIPE_SHADER_FRAGMENT, 0, count, states);
 }
 
+static void r600_bind_cs_sampler_states(struct pipe_context *ctx, unsigned count, void **states)
+{
+	r600_bind_sampler_states(ctx, PIPE_SHADER_COMPUTE, 0, count, states);
+}
+
 static void r600_delete_sampler_state(struct pipe_context *ctx, void *state)
 {
 	free(state);
@@ -661,6 +666,12 @@ static void r600_set_ps_sampler_views(struct pipe_context *ctx, unsigned count,
 	r600_set_sampler_views(ctx, PIPE_SHADER_FRAGMENT, 0, count, views);
 }
 
+static void r600_set_cs_sampler_views(struct pipe_context *ctx, unsigned count,
+				      struct pipe_sampler_view **views)
+{
+	r600_set_sampler_views(ctx, PIPE_SHADER_COMPUTE, 0, count, views);
+}
+
 static void r600_set_viewport_states(struct pipe_context *ctx,
                                      unsigned start_slot,
                                      unsigned num_viewports,
@@ -1741,6 +1752,7 @@ void r600_init_common_state_functions(struct r600_context *rctx)
 	rctx->context.bind_vertex_elements_state = r600_bind_vertex_elements;
 	rctx->context.bind_vertex_sampler_states = r600_bind_vs_sampler_states;
 	rctx->context.bind_vs_state = r600_bind_vs_state;
+	rctx->context.bind_compute_sampler_states = r600_bind_cs_sampler_states;
 	rctx->context.delete_blend_state = r600_delete_blend_state;
 	rctx->context.delete_depth_stencil_alpha_state = r600_delete_dsa_state;
 	rctx->context.delete_fs_state = r600_delete_ps_state;
@@ -1758,6 +1770,7 @@ void r600_init_common_state_functions(struct r600_context *rctx)
 	rctx->context.set_index_buffer = r600_set_index_buffer;
 	rctx->context.set_fragment_sampler_views = r600_set_ps_sampler_views;
 	rctx->context.set_vertex_sampler_views = r600_set_vs_sampler_views;
+	rctx->context.set_compute_sampler_views = r600_set_cs_sampler_views;
 	rctx->context.sampler_view_destroy = r600_sampler_view_destroy;
 	rctx->context.texture_barrier = r600_texture_barrier;
 	rctx->context.create_stream_output_target = r600_create_so_target;
diff --git a/src/gallium/state_trackers/clover/core/memory.cpp b/src/gallium/state_trackers/clover/core/memory.cpp
index 63050b3d589..29785250264 100644
--- a/src/gallium/state_trackers/clover/core/memory.cpp
+++ b/src/gallium/state_trackers/clover/core/memory.cpp
@@ -174,7 +174,7 @@ image2d::image2d(clover::context &ctx, cl_mem_flags flags,
                  const cl_image_format *format, size_t width,
                  size_t height, size_t row_pitch,
                  void *host_ptr) :
-   image(ctx, flags, format, width, height, 0,
+   image(ctx, flags, format, width, height, 1,
          row_pitch, 0, height * row_pitch, host_ptr) {
 }
 
diff --git a/src/gallium/state_trackers/clover/llvm/invocation.cpp b/src/gallium/state_trackers/clover/llvm/invocation.cpp
index 9fae83fe457..e8d33f2dc92 100644
--- a/src/gallium/state_trackers/clover/llvm/invocation.cpp
+++ b/src/gallium/state_trackers/clover/llvm/invocation.cpp
@@ -251,6 +251,7 @@ namespace {
       }
 #endif
 
+      mod->dump();
       // Add a function internalizer pass.
       //
       // By default, the function internalizer pass will look for a function
@@ -329,12 +330,20 @@ namespace {
                // target.  I think we need to ask clang what these are.  For now,
                // pretend everything is in the global address space.
                llvm::Type *elem_type = arg_type->getPointerElementType();
-               if (elem_type->isStructTy() &&
-                   elem_type->getStructName().equals("opencl.image2d_t")) {
-                  args.push_back(module::argument(module::argument::image2d_wr,
-                                 arg_size, target_size, target_align,
-                                 module::argument::zero_ext));
-                  continue;
+               if (elem_type->isStructTy()) {
+                  const llvm::StringRef &type_name = elem_type->getStructName();
+
+                  if (type_name.equals("opencl.image2d_t")) {
+                     args.push_back(module::argument(module::argument::image2d_rd,
+                                    arg_size, target_size, target_align,
+                                    module::argument::zero_ext));
+                     continue;
+                  } else if (type_name.equals("opencl.sampler_t")) {
+                     args.push_back(module::argument(module::argument::sampler,
+                                    arg_size, target_size, target_align,
+                                    module::argument::zero_ext));
+                     continue;
+                  }
                }
                unsigned address_space = llvm::cast<llvm::PointerType>(arg_type)->getAddressSpace();
                switch (address_space) {