gallium/radeon: derive buffer placement and flags only at initialization

Invalidated buffers don't have to go through it.

Split r600_init_resource into r600_init_resource_fields and
r600_alloc_resource.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

5 files changed, 67 insertions, 49 deletions

diff --git a/src/gallium/drivers/r600/r600_state_common.c b/src/gallium/drivers/r600/r600_state_common.c
index 5234c04..6f8053a 100644
--- a/src/gallium/drivers/r600/r600_state_common.c
+++ b/src/gallium/drivers/r600/r600_state_common.c
@@ -2782,12 +2782,11 @@ static void r600_invalidate_buffer(struct pipe_context *ctx, struct pipe_resourc
 {
 	struct r600_context *rctx = (struct r600_context*)ctx;
 	struct r600_resource *rbuffer = r600_resource(buf);
-	unsigned i, shader, mask, alignment = rbuffer->buf->alignment;
+	unsigned i, shader, mask;
 	struct r600_pipe_sampler_view *view;
 
 	/* Reallocate the buffer in the same pipe_resource. */
-	r600_init_resource(&rctx->screen->b, rbuffer, rbuffer->b.b.width0,
-			   alignment);
+	r600_alloc_resource(&rctx->screen->b, rbuffer);
 
 	/* We changed the buffer, now we need to bind it where the old one was bound. */
 	/* Vertex buffers. */
diff --git a/src/gallium/drivers/radeon/r600_buffer_common.c b/src/gallium/drivers/radeon/r600_buffer_common.c
index 4480293..6a55de1 100644
--- a/src/gallium/drivers/radeon/r600_buffer_common.c
+++ b/src/gallium/drivers/radeon/r600_buffer_common.c
@@ -99,20 +99,23 @@ void *r600_buffer_map_sync_with_rings(struct r600_common_context *ctx,
 	return ctx->ws->buffer_map(resource->buf, NULL, usage);
 }
 
-bool r600_init_resource(struct r600_common_screen *rscreen,
-			struct r600_resource *res,
-			uint64_t size, unsigned alignment)
+void r600_init_resource_fields(struct r600_common_screen *rscreen,
+			       struct r600_resource *res,
+			       uint64_t size, unsigned alignment)
 {
 	struct r600_texture *rtex = (struct r600_texture*)res;
-	struct pb_buffer *old_buf, *new_buf;
-	enum radeon_bo_flag flags = 0;
+
+	res->bo_size = size;
+	res->bo_alignment = alignment;
+	res->flags = 0;
 
 	switch (res->b.b.usage) {
 	case PIPE_USAGE_STREAM:
-		flags = RADEON_FLAG_GTT_WC;
+		res->flags = RADEON_FLAG_GTT_WC;
 		/* fall through */
 	case PIPE_USAGE_STAGING:
-		/* Transfers are likely to occur more often with these resources. */
+		/* Transfers are likely to occur more often with these
+		 * resources. */
 		res->domains = RADEON_DOMAIN_GTT;
 		break;
 	case PIPE_USAGE_DYNAMIC:
@@ -122,60 +125,78 @@ bool r600_init_resource(struct r600_common_screen *rscreen,
 		if (rscreen->info.drm_major == 2 &&
 		    rscreen->info.drm_minor < 40) {
 			res->domains = RADEON_DOMAIN_GTT;
-			flags |= RADEON_FLAG_GTT_WC;
+			res->flags |= RADEON_FLAG_GTT_WC;
 			break;
 		}
-		flags |= RADEON_FLAG_CPU_ACCESS;
+		res->flags |= RADEON_FLAG_CPU_ACCESS;
 		/* fall through */
 	case PIPE_USAGE_DEFAULT:
 	case PIPE_USAGE_IMMUTABLE:
 	default:
-		/* Not listing GTT here improves performance in some apps. */
+		/* Not listing GTT here improves performance in some
+		 * apps. */
 		res->domains = RADEON_DOMAIN_VRAM;
-		flags |= RADEON_FLAG_GTT_WC;
+		res->flags |= RADEON_FLAG_GTT_WC;
 		break;
 	}
 
 	if (res->b.b.target == PIPE_BUFFER &&
 	    res->b.b.flags & (PIPE_RESOURCE_FLAG_MAP_PERSISTENT |
 			      PIPE_RESOURCE_FLAG_MAP_COHERENT)) {
-		/* Use GTT for all persistent mappings with older kernels,
-		 * because they didn't always flush the HDP cache before CS
-		 * execution.
+		/* Use GTT for all persistent mappings with older
+		 * kernels, because they didn't always flush the HDP
+		 * cache before CS execution.
 		 *
-		 * Write-combined CPU mappings are fine, the kernel ensures all CPU
-		 * writes finish before the GPU executes a command stream.
+		 * Write-combined CPU mappings are fine, the kernel
+		 * ensures all CPU writes finish before the GPU
+		 * executes a command stream.
 		 */
 		if (rscreen->info.drm_major == 2 &&
 		    rscreen->info.drm_minor < 40)
 			res->domains = RADEON_DOMAIN_GTT;
 		else if (res->domains & RADEON_DOMAIN_VRAM)
-			flags |= RADEON_FLAG_CPU_ACCESS;
+			res->flags |= RADEON_FLAG_CPU_ACCESS;
 	}
 
 	/* Tiled textures are unmappable. Always put them in VRAM. */
 	if (res->b.b.target != PIPE_BUFFER &&
 	    rtex->surface.level[0].mode >= RADEON_SURF_MODE_1D) {
 		res->domains = RADEON_DOMAIN_VRAM;
-		flags &= ~RADEON_FLAG_CPU_ACCESS;
-		flags |= RADEON_FLAG_NO_CPU_ACCESS |
+		res->flags &= ~RADEON_FLAG_CPU_ACCESS;
+		res->flags |= RADEON_FLAG_NO_CPU_ACCESS |
 			 RADEON_FLAG_GTT_WC;
 	}
 
-	/* If VRAM is just stolen system memory, allow both VRAM and GTT,
-	 * whichever has free space. If a buffer is evicted from VRAM to GTT,
-	 * it will stay there.
+	/* If VRAM is just stolen system memory, allow both VRAM and
+	 * GTT, whichever has free space. If a buffer is evicted from
+	 * VRAM to GTT, it will stay there.
 	 */
 	if (!rscreen->info.has_dedicated_vram &&
 	    res->domains == RADEON_DOMAIN_VRAM)
 		res->domains = RADEON_DOMAIN_VRAM_GTT;
 
 	if (rscreen->debug_flags & DBG_NO_WC)
-		flags &= ~RADEON_FLAG_GTT_WC;
+		res->flags &= ~RADEON_FLAG_GTT_WC;
+
+	/* Set expected VRAM and GART usage for the buffer. */
+	res->vram_usage = 0;
+	res->gart_usage = 0;
+
+	if (res->domains & RADEON_DOMAIN_VRAM)
+		res->vram_usage = size;
+	else if (res->domains & RADEON_DOMAIN_GTT)
+		res->gart_usage = size;
+}
+
+bool r600_alloc_resource(struct r600_common_screen *rscreen,
+			 struct r600_resource *res)
+{
+	struct pb_buffer *old_buf, *new_buf;
 
 	/* Allocate a new resource. */
-	new_buf = rscreen->ws->buffer_create(rscreen->ws, size, alignment,
-					     res->domains, flags);
+	new_buf = rscreen->ws->buffer_create(rscreen->ws, res->bo_size,
+					     res->bo_alignment,
+					     res->domains, res->flags);
 	if (!new_buf) {
 		return false;
 	}
@@ -197,15 +218,6 @@ bool r600_init_resource(struct r600_common_screen *rscreen,
 	util_range_set_empty(&res->valid_buffer_range);
 	res->TC_L2_dirty = false;
 
-	/* Set expected VRAM and GART usage for the buffer. */
-	res->vram_usage = 0;
-	res->gart_usage = 0;
-
-	if (res->domains & RADEON_DOMAIN_VRAM)
-		res->vram_usage = size;
-	else if (res->domains & RADEON_DOMAIN_GTT)
-		res->gart_usage = size;
-
 	/* Print debug information. */
 	if (rscreen->debug_flags & DBG_VM && res->b.b.target == PIPE_BUFFER) {
 		fprintf(stderr, "VM start=0x%"PRIX64"  end=0x%"PRIX64" | Buffer %"PRIu64" bytes\n",
@@ -516,7 +528,9 @@ struct pipe_resource *r600_buffer_create(struct pipe_screen *screen,
 	struct r600_common_screen *rscreen = (struct r600_common_screen*)screen;
 	struct r600_resource *rbuffer = r600_alloc_buffer_struct(screen, templ);
 
-	if (!r600_init_resource(rscreen, rbuffer, templ->width0, alignment)) {
+	r600_init_resource_fields(rscreen, rbuffer, templ->width0, alignment);
+
+	if (!r600_alloc_resource(rscreen, rbuffer)) {
 		FREE(rbuffer);
 		return NULL;
 	}
diff --git a/src/gallium/drivers/radeon/r600_pipe_common.h b/src/gallium/drivers/radeon/r600_pipe_common.h
index 5375044..1924535 100644
--- a/src/gallium/drivers/radeon/r600_pipe_common.h
+++ b/src/gallium/drivers/radeon/r600_pipe_common.h
@@ -174,8 +174,11 @@ struct r600_resource {
 	uint64_t			vram_usage;
 	uint64_t			gart_usage;
 
-	/* Resource state. */
+	/* Resource properties. */
+	uint64_t			bo_size;
+	unsigned			bo_alignment;
 	enum radeon_bo_domain		domains;
+	enum radeon_bo_flag		flags;
 
 	/* The buffer range which is initialized (with a write transfer,
 	 * streamout, DMA, or as a random access target). The rest of
@@ -653,9 +656,11 @@ void r600_buffer_subdata(struct pipe_context *ctx,
 			 struct pipe_resource *buffer,
 			 unsigned usage, unsigned offset,
 			 unsigned size, const void *data);
-bool r600_init_resource(struct r600_common_screen *rscreen,
-			struct r600_resource *res,
-			uint64_t size, unsigned alignment);
+void r600_init_resource_fields(struct r600_common_screen *rscreen,
+			       struct r600_resource *res,
+			       uint64_t size, unsigned alignment);
+bool r600_alloc_resource(struct r600_common_screen *rscreen,
+			 struct r600_resource *res);
 struct pipe_resource *r600_buffer_create(struct pipe_screen *screen,
 					 const struct pipe_resource *templ,
 					 unsigned alignment);
diff --git a/src/gallium/drivers/radeon/r600_texture.c b/src/gallium/drivers/radeon/r600_texture.c
index e19150b..fb3068a 100644
--- a/src/gallium/drivers/radeon/r600_texture.c
+++ b/src/gallium/drivers/radeon/r600_texture.c
@@ -1103,8 +1103,10 @@ r600_texture_create_object(struct pipe_screen *screen,
 
 	/* Now create the backing buffer. */
 	if (!buf) {
-		if (!r600_init_resource(rscreen, resource, rtex->size,
-					rtex->surface.bo_alignment)) {
+		r600_init_resource_fields(rscreen, resource, rtex->size,
+					  rtex->surface.bo_alignment);
+
+		if (!r600_alloc_resource(rscreen, resource)) {
 			FREE(rtex);
 			return NULL;
 		}
@@ -1418,8 +1420,7 @@ static void r600_texture_invalidate_storage(struct r600_common_context *rctx,
 	assert(rtex->surface.level[0].mode == RADEON_SURF_MODE_LINEAR_ALIGNED);
 
 	/* Reallocate the buffer in the same pipe_resource. */
-	r600_init_resource(rscreen, &rtex->resource, rtex->size,
-			   rtex->surface.bo_alignment);
+	r600_alloc_resource(rscreen, &rtex->resource);
 
 	/* Initialize the CMASK base address (needed even without CMASK). */
 	rtex->cmask.base_address_reg =
diff --git a/src/gallium/drivers/radeonsi/si_descriptors.c b/src/gallium/drivers/radeonsi/si_descriptors.c
index eb0e5fa..b80f4f2 100644
--- a/src/gallium/drivers/radeonsi/si_descriptors.c
+++ b/src/gallium/drivers/radeonsi/si_descriptors.c
@@ -1427,15 +1427,14 @@ static void si_invalidate_buffer(struct pipe_context *ctx, struct pipe_resource
 {
 	struct si_context *sctx = (struct si_context*)ctx;
 	struct r600_resource *rbuffer = r600_resource(buf);
-	unsigned i, shader, alignment = rbuffer->buf->alignment;
+	unsigned i, shader;
 	uint64_t old_va = rbuffer->gpu_address;
 	unsigned num_elems = sctx->vertex_elements ?
 				       sctx->vertex_elements->count : 0;
 	struct si_sampler_view *view;
 
 	/* Reallocate the buffer in the same pipe_resource. */
-	r600_init_resource(&sctx->screen->b, rbuffer, rbuffer->b.b.width0,
-			   alignment);
+	r600_alloc_resource(&sctx->screen->b, rbuffer);
 
 	/* We changed the buffer, now we need to bind it where the old one
 	 * was bound. This consists of 2 things: