X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;f=src%2Famd%2Fvulkan%2Fradv_meta_resolve_fs.c;h=af4a6fe7a66be032d806d0c8af96f3eb6092f11a;hb=90819abb56f6b1a0cd4946b13b6caf24fb46e500;hp=0e4957b163cbeccb7d2add7de1503e004d4e9a63;hpb=0a8127bbfbe8a8af45dacf6c0998a8f0f1152f4b;p=mesa.git

diff --git a/src/amd/vulkan/radv_meta_resolve_fs.c b/src/amd/vulkan/radv_meta_resolve_fs.c
index 0e4957b163c..af4a6fe7a66 100644
--- a/src/amd/vulkan/radv_meta_resolve_fs.c
+++ b/src/amd/vulkan/radv_meta_resolve_fs.c
@@ -156,30 +156,23 @@ static const VkPipelineVertexInputStateCreateInfo normal_vi_create_info = {
 	.vertexAttributeDescriptionCount = 0,
 };
 
-static VkFormat pipeline_formats[] = {
-   VK_FORMAT_R8G8B8A8_UNORM,
-   VK_FORMAT_R8G8B8A8_UINT,
-   VK_FORMAT_R8G8B8A8_SINT,
-   VK_FORMAT_A2R10G10B10_UINT_PACK32,
-   VK_FORMAT_A2R10G10B10_SINT_PACK32,
-   VK_FORMAT_R16G16B16A16_UNORM,
-   VK_FORMAT_R16G16B16A16_SNORM,
-   VK_FORMAT_R16G16B16A16_UINT,
-   VK_FORMAT_R16G16B16A16_SINT,
-   VK_FORMAT_R32_SFLOAT,
-   VK_FORMAT_R32G32_SFLOAT,
-   VK_FORMAT_R32G32B32A32_SFLOAT
-};
-
 static VkResult
 create_resolve_pipeline(struct radv_device *device,
 			int samples_log2,
 			VkFormat format)
 {
+	mtx_lock(&device->meta_state.mtx);
+
+	unsigned fs_key = radv_format_meta_fs_key(format);
+	VkPipeline *pipeline = &device->meta_state.resolve_fragment.rc[samples_log2].pipeline[fs_key];
+	if (*pipeline) {
+		mtx_unlock(&device->meta_state.mtx);
+		return VK_SUCCESS;
+	}
+
 	VkResult result;
 	bool is_integer = false;
 	uint32_t samples = 1 << samples_log2;
-	unsigned fs_key = radv_format_meta_fs_key(format);
 	const VkPipelineVertexInputStateCreateInfo *vi_create_info;
 	vi_create_info = &normal_vi_create_info;
 	if (vk_format_is_int(format))
@@ -195,9 +188,6 @@ create_resolve_pipeline(struct radv_device *device,
 
 	assert(!*rp);
 
-	VkPipeline *pipeline = &device->meta_state.resolve_fragment.rc[samples_log2].pipeline[fs_key];
-	assert(!*pipeline);
-
 	VkPipelineShaderStageCreateInfo pipeline_shader_stages[] = {
 		{
 			.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO,
@@ -322,11 +312,12 @@ create_resolve_pipeline(struct radv_device *device,
 	ralloc_free(vs.nir);
 	ralloc_free(fs.nir);
 
+	mtx_unlock(&device->meta_state.mtx);
 	return result;
 }
 
 VkResult
-radv_device_init_meta_resolve_fragment_state(struct radv_device *device)
+radv_device_init_meta_resolve_fragment_state(struct radv_device *device, bool on_demand)
 {
 	VkResult res;
 
@@ -334,9 +325,12 @@ radv_device_init_meta_resolve_fragment_state(struct radv_device *device)
 	if (res != VK_SUCCESS)
 		goto fail;
 
+	if (on_demand)
+		return VK_SUCCESS;
+
 	for (uint32_t i = 0; i < MAX_SAMPLES_LOG2; ++i) {
-		for (unsigned j = 0; j < ARRAY_SIZE(pipeline_formats); ++j) {
-			res = create_resolve_pipeline(device, i, pipeline_formats[j]);
+		for (unsigned j = 0; j < NUM_META_FS_KEYS; ++j) {
+			res = create_resolve_pipeline(device, i, radv_fs_key_format_exemplars[j]);
 			if (res != VK_SUCCESS)
 				goto fail;
 		}
@@ -419,10 +413,18 @@ emit_resolve(struct radv_cmd_buffer *cmd_buffer,
 			      push_constants);
 
 	unsigned fs_key = radv_format_meta_fs_key(dest_iview->vk_format);
-	VkPipeline pipeline_h = device->meta_state.resolve_fragment.rc[samples_log2].pipeline[fs_key];
+	VkPipeline* pipeline = &device->meta_state.resolve_fragment.rc[samples_log2].pipeline[fs_key];
+
+	if (*pipeline == VK_NULL_HANDLE) {
+		VkResult ret = create_resolve_pipeline(device, samples_log2, radv_fs_key_format_exemplars[fs_key]);
+		if (ret != VK_SUCCESS) {
+			cmd_buffer->record_result = ret;
+			return;
+		}
+	}
 
 	radv_CmdBindPipeline(cmd_buffer_h, VK_PIPELINE_BIND_POINT_GRAPHICS,
-			     pipeline_h);
+			     *pipeline);
 
 	radv_CmdSetViewport(radv_cmd_buffer_to_handle(cmd_buffer), 0, 1, &(VkViewport) {
 		.x = dest_offset->x,
@@ -461,6 +463,14 @@ void radv_meta_resolve_fragment_image(struct radv_cmd_buffer *cmd_buffer,
 	radv_decompress_resolve_src(cmd_buffer, src_image, src_image_layout,
 				    region_count, regions);
 
+	if (!device->meta_state.resolve_fragment.rc[samples_log2].render_pass[fs_key][dst_layout]) {
+		VkResult ret = create_resolve_pipeline(device, samples_log2, radv_fs_key_format_exemplars[fs_key]);
+		if (ret != VK_SUCCESS) {
+			cmd_buffer->record_result = ret;
+			return;
+		}
+	}
+
 	rp = device->meta_state.resolve_fragment.rc[samples_log2].render_pass[fs_key][dst_layout];
 
 	radv_meta_save(&saved_state, cmd_buffer,
@@ -582,33 +592,20 @@ radv_cmd_buffer_resolve_subpass_fs(struct radv_cmd_buffer *cmd_buffer)
 	struct radv_meta_saved_state saved_state;
 	struct radv_subpass_barrier barrier;
 
-	/* FINISHME(perf): Skip clears for resolve attachments.
-	 *
-	 * From the Vulkan 1.0 spec:
-	 *
-	 *    If the first use of an attachment in a render pass is as a resolve
-	 *    attachment, then the loadOp is effectively ignored as the resolve is
-	 *    guaranteed to overwrite all pixels in the render area.
-	 */
-
-	if (!subpass->has_resolve)
-		return;
-
-	radv_meta_save(&saved_state, cmd_buffer,
-		       RADV_META_SAVE_GRAPHICS_PIPELINE |
-		       RADV_META_SAVE_CONSTANTS |
-		       RADV_META_SAVE_DESCRIPTORS);
-
 	/* Resolves happen before the end-of-subpass barriers get executed,
 	 * so we have to make the attachment shader-readable */
 	barrier.src_stage_mask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT;
-	barrier.src_access_mask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
-				  VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT;
+	barrier.src_access_mask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;
 	barrier.dst_access_mask = VK_ACCESS_INPUT_ATTACHMENT_READ_BIT;
 	radv_subpass_barrier(cmd_buffer, &barrier);
 
 	radv_decompress_resolve_subpass_src(cmd_buffer);
 
+	radv_meta_save(&saved_state, cmd_buffer,
+		       RADV_META_SAVE_GRAPHICS_PIPELINE |
+		       RADV_META_SAVE_CONSTANTS |
+		       RADV_META_SAVE_DESCRIPTORS);
+
 	for (uint32_t i = 0; i < subpass->color_count; ++i) {
 		struct radv_subpass_attachment src_att = subpass->color_attachments[i];
 		struct radv_subpass_attachment dest_att = subpass->resolve_attachments[i];