summaryrefslogtreecommitdiffstats
path: root/src/video_core/renderer_vulkan
diff options
context:
space:
mode:
authorReinUsesLisp <reinuseslisp@airmail.cc>2020-12-30 06:25:23 +0100
committerReinUsesLisp <reinuseslisp@airmail.cc>2020-12-30 07:38:50 +0100
commit9764c13d6d2977903f407761b27d847c0056e1c4 (patch)
treef6f5d6d6379b0404147969e7d1f548ed3d49ca01 /src/video_core/renderer_vulkan
parentvideo_core: Add a delayed destruction ring abstraction (diff)
downloadyuzu-9764c13d6d2977903f407761b27d847c0056e1c4.tar
yuzu-9764c13d6d2977903f407761b27d847c0056e1c4.tar.gz
yuzu-9764c13d6d2977903f407761b27d847c0056e1c4.tar.bz2
yuzu-9764c13d6d2977903f407761b27d847c0056e1c4.tar.lz
yuzu-9764c13d6d2977903f407761b27d847c0056e1c4.tar.xz
yuzu-9764c13d6d2977903f407761b27d847c0056e1c4.tar.zst
yuzu-9764c13d6d2977903f407761b27d847c0056e1c4.zip
Diffstat (limited to 'src/video_core/renderer_vulkan')
-rw-r--r--src/video_core/renderer_vulkan/blit_image.cpp624
-rw-r--r--src/video_core/renderer_vulkan/blit_image.h97
-rw-r--r--src/video_core/renderer_vulkan/fixed_pipeline_state.cpp1
-rw-r--r--src/video_core/renderer_vulkan/fixed_pipeline_state.h1
-rw-r--r--src/video_core/renderer_vulkan/maxwell_to_vk.cpp30
-rw-r--r--src/video_core/renderer_vulkan/maxwell_to_vk.h2
-rw-r--r--src/video_core/renderer_vulkan/renderer_vulkan.cpp13
-rw-r--r--src/video_core/renderer_vulkan/renderer_vulkan.h3
-rw-r--r--src/video_core/renderer_vulkan/shaders/blit.frag24
-rw-r--r--src/video_core/renderer_vulkan/shaders/blit.vert28
-rw-r--r--src/video_core/renderer_vulkan/shaders/quad_array.comp37
-rw-r--r--src/video_core/renderer_vulkan/shaders/quad_indexed.comp50
-rw-r--r--src/video_core/renderer_vulkan/shaders/uint8.comp33
-rw-r--r--src/video_core/renderer_vulkan/vk_blit_screen.cpp301
-rw-r--r--src/video_core/renderer_vulkan/vk_blit_screen.h4
-rw-r--r--src/video_core/renderer_vulkan/vk_buffer_cache.cpp67
-rw-r--r--src/video_core/renderer_vulkan/vk_buffer_cache.h10
-rw-r--r--src/video_core/renderer_vulkan/vk_compute_pass.cpp327
-rw-r--r--src/video_core/renderer_vulkan/vk_compute_pass.h4
-rw-r--r--src/video_core/renderer_vulkan/vk_device.cpp110
-rw-r--r--src/video_core/renderer_vulkan/vk_device.h34
-rw-r--r--src/video_core/renderer_vulkan/vk_fence_manager.cpp7
-rw-r--r--src/video_core/renderer_vulkan/vk_fence_manager.h6
-rw-r--r--src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp64
-rw-r--r--src/video_core/renderer_vulkan/vk_graphics_pipeline.h26
-rw-r--r--src/video_core/renderer_vulkan/vk_image.cpp135
-rw-r--r--src/video_core/renderer_vulkan/vk_image.h84
-rw-r--r--src/video_core/renderer_vulkan/vk_memory_manager.cpp2
-rw-r--r--src/video_core/renderer_vulkan/vk_memory_manager.h20
-rw-r--r--src/video_core/renderer_vulkan/vk_pipeline_cache.cpp28
-rw-r--r--src/video_core/renderer_vulkan/vk_pipeline_cache.h16
-rw-r--r--src/video_core/renderer_vulkan/vk_rasterizer.cpp700
-rw-r--r--src/video_core/renderer_vulkan/vk_rasterizer.h131
-rw-r--r--src/video_core/renderer_vulkan/vk_renderpass_cache.cpp158
-rw-r--r--src/video_core/renderer_vulkan/vk_renderpass_cache.h70
-rw-r--r--src/video_core/renderer_vulkan/vk_sampler_cache.cpp83
-rw-r--r--src/video_core/renderer_vulkan/vk_sampler_cache.h29
-rw-r--r--src/video_core/renderer_vulkan/vk_scheduler.cpp79
-rw-r--r--src/video_core/renderer_vulkan/vk_scheduler.h14
-rw-r--r--src/video_core/renderer_vulkan/vk_shader_decompiler.cpp6
-rw-r--r--src/video_core/renderer_vulkan/vk_shader_decompiler.h8
-rw-r--r--src/video_core/renderer_vulkan/vk_shader_util.cpp11
-rw-r--r--src/video_core/renderer_vulkan/vk_shader_util.h4
-rw-r--r--src/video_core/renderer_vulkan/vk_state_tracker.cpp23
-rw-r--r--src/video_core/renderer_vulkan/vk_state_tracker.h8
-rw-r--r--src/video_core/renderer_vulkan/vk_stream_buffer.cpp20
-rw-r--r--src/video_core/renderer_vulkan/vk_stream_buffer.h12
-rw-r--r--src/video_core/renderer_vulkan/vk_texture_cache.cpp1473
-rw-r--r--src/video_core/renderer_vulkan/vk_texture_cache.h328
-rw-r--r--src/video_core/renderer_vulkan/vk_update_descriptor.h30
-rw-r--r--src/video_core/renderer_vulkan/wrapper.cpp69
-rw-r--r--src/video_core/renderer_vulkan/wrapper.h117
52 files changed, 2979 insertions, 2582 deletions
diff --git a/src/video_core/renderer_vulkan/blit_image.cpp b/src/video_core/renderer_vulkan/blit_image.cpp
new file mode 100644
index 000000000..87c8e5693
--- /dev/null
+++ b/src/video_core/renderer_vulkan/blit_image.cpp
@@ -0,0 +1,624 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+
+#include "video_core/host_shaders/convert_depth_to_float_frag_spv.h"
+#include "video_core/host_shaders/convert_float_to_depth_frag_spv.h"
+#include "video_core/host_shaders/full_screen_triangle_vert_spv.h"
+#include "video_core/host_shaders/vulkan_blit_color_float_frag_spv.h"
+#include "video_core/host_shaders/vulkan_blit_depth_stencil_frag_spv.h"
+#include "video_core/renderer_vulkan/blit_image.h"
+#include "video_core/renderer_vulkan/maxwell_to_vk.h"
+#include "video_core/renderer_vulkan/vk_device.h"
+#include "video_core/renderer_vulkan/vk_scheduler.h"
+#include "video_core/renderer_vulkan/vk_shader_util.h"
+#include "video_core/renderer_vulkan/vk_state_tracker.h"
+#include "video_core/renderer_vulkan/vk_texture_cache.h"
+#include "video_core/renderer_vulkan/vk_update_descriptor.h"
+#include "video_core/renderer_vulkan/wrapper.h"
+#include "video_core/surface.h"
+
+namespace Vulkan {
+
+using VideoCommon::ImageViewType;
+
+namespace {
+struct PushConstants {
+ std::array<float, 2> tex_scale;
+ std::array<float, 2> tex_offset;
+};
+
+template <u32 binding>
+inline constexpr VkDescriptorSetLayoutBinding TEXTURE_DESCRIPTOR_SET_LAYOUT_BINDING{
+ .binding = binding,
+ .descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER,
+ .descriptorCount = 1,
+ .stageFlags = VK_SHADER_STAGE_FRAGMENT_BIT,
+ .pImmutableSamplers = nullptr,
+};
+constexpr std::array TWO_TEXTURES_DESCRIPTOR_SET_LAYOUT_BINDINGS{
+ TEXTURE_DESCRIPTOR_SET_LAYOUT_BINDING<0>,
+ TEXTURE_DESCRIPTOR_SET_LAYOUT_BINDING<1>,
+};
+constexpr VkDescriptorSetLayoutCreateInfo ONE_TEXTURE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO{
+ .sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO,
+ .pNext = nullptr,
+ .flags = 0,
+ .bindingCount = 1,
+ .pBindings = &TEXTURE_DESCRIPTOR_SET_LAYOUT_BINDING<0>,
+};
+constexpr VkDescriptorSetLayoutCreateInfo TWO_TEXTURES_DESCRIPTOR_SET_LAYOUT_CREATE_INFO{
+ .sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO,
+ .pNext = nullptr,
+ .flags = 0,
+ .bindingCount = static_cast<u32>(TWO_TEXTURES_DESCRIPTOR_SET_LAYOUT_BINDINGS.size()),
+ .pBindings = TWO_TEXTURES_DESCRIPTOR_SET_LAYOUT_BINDINGS.data(),
+};
+constexpr VkPushConstantRange PUSH_CONSTANT_RANGE{
+ .stageFlags = VK_SHADER_STAGE_VERTEX_BIT,
+ .offset = 0,
+ .size = sizeof(PushConstants),
+};
+constexpr VkPipelineVertexInputStateCreateInfo PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO{
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO,
+ .pNext = nullptr,
+ .flags = 0,
+ .vertexBindingDescriptionCount = 0,
+ .pVertexBindingDescriptions = nullptr,
+ .vertexAttributeDescriptionCount = 0,
+ .pVertexAttributeDescriptions = nullptr,
+};
+constexpr VkPipelineInputAssemblyStateCreateInfo PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO{
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO,
+ .pNext = nullptr,
+ .flags = 0,
+ .topology = VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST,
+ .primitiveRestartEnable = VK_FALSE,
+};
+constexpr VkPipelineViewportStateCreateInfo PIPELINE_VIEWPORT_STATE_CREATE_INFO{
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO,
+ .pNext = nullptr,
+ .flags = 0,
+ .viewportCount = 1,
+ .pViewports = nullptr,
+ .scissorCount = 1,
+ .pScissors = nullptr,
+};
+constexpr VkPipelineRasterizationStateCreateInfo PIPELINE_RASTERIZATION_STATE_CREATE_INFO{
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO,
+ .pNext = nullptr,
+ .flags = 0,
+ .depthClampEnable = VK_FALSE,
+ .rasterizerDiscardEnable = VK_FALSE,
+ .polygonMode = VK_POLYGON_MODE_FILL,
+ .cullMode = VK_CULL_MODE_BACK_BIT,
+ .frontFace = VK_FRONT_FACE_CLOCKWISE,
+ .depthBiasEnable = VK_FALSE,
+ .depthBiasConstantFactor = 0.0f,
+ .depthBiasClamp = 0.0f,
+ .depthBiasSlopeFactor = 0.0f,
+ .lineWidth = 1.0f,
+};
+constexpr VkPipelineMultisampleStateCreateInfo PIPELINE_MULTISAMPLE_STATE_CREATE_INFO{
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO,
+ .pNext = nullptr,
+ .flags = 0,
+ .rasterizationSamples = VK_SAMPLE_COUNT_1_BIT,
+ .sampleShadingEnable = VK_FALSE,
+ .minSampleShading = 0.0f,
+ .pSampleMask = nullptr,
+ .alphaToCoverageEnable = VK_FALSE,
+ .alphaToOneEnable = VK_FALSE,
+};
+constexpr std::array DYNAMIC_STATES{
+ VK_DYNAMIC_STATE_VIEWPORT,
+ VK_DYNAMIC_STATE_SCISSOR,
+};
+constexpr VkPipelineDynamicStateCreateInfo PIPELINE_DYNAMIC_STATE_CREATE_INFO{
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_DYNAMIC_STATE_CREATE_INFO,
+ .pNext = nullptr,
+ .flags = 0,
+ .dynamicStateCount = static_cast<u32>(DYNAMIC_STATES.size()),
+ .pDynamicStates = DYNAMIC_STATES.data(),
+};
+constexpr VkPipelineColorBlendStateCreateInfo PIPELINE_COLOR_BLEND_STATE_EMPTY_CREATE_INFO{
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO,
+ .pNext = nullptr,
+ .flags = 0,
+ .logicOpEnable = VK_FALSE,
+ .logicOp = VK_LOGIC_OP_CLEAR,
+ .attachmentCount = 0,
+ .pAttachments = nullptr,
+ .blendConstants = {0.0f, 0.0f, 0.0f, 0.0f},
+};
+constexpr VkPipelineColorBlendAttachmentState PIPELINE_COLOR_BLEND_ATTACHMENT_STATE{
+ .blendEnable = VK_FALSE,
+ .srcColorBlendFactor = VK_BLEND_FACTOR_ZERO,
+ .dstColorBlendFactor = VK_BLEND_FACTOR_ZERO,
+ .colorBlendOp = VK_BLEND_OP_ADD,
+ .srcAlphaBlendFactor = VK_BLEND_FACTOR_ZERO,
+ .dstAlphaBlendFactor = VK_BLEND_FACTOR_ZERO,
+ .alphaBlendOp = VK_BLEND_OP_ADD,
+ .colorWriteMask = VK_COLOR_COMPONENT_R_BIT | VK_COLOR_COMPONENT_G_BIT |
+ VK_COLOR_COMPONENT_B_BIT | VK_COLOR_COMPONENT_A_BIT,
+};
+constexpr VkPipelineColorBlendStateCreateInfo PIPELINE_COLOR_BLEND_STATE_GENERIC_CREATE_INFO{
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO,
+ .pNext = nullptr,
+ .flags = 0,
+ .logicOpEnable = VK_FALSE,
+ .logicOp = VK_LOGIC_OP_CLEAR,
+ .attachmentCount = 1,
+ .pAttachments = &PIPELINE_COLOR_BLEND_ATTACHMENT_STATE,
+ .blendConstants = {0.0f, 0.0f, 0.0f, 0.0f},
+};
+constexpr VkPipelineDepthStencilStateCreateInfo PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO{
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO,
+ .pNext = nullptr,
+ .flags = 0,
+ .depthTestEnable = VK_TRUE,
+ .depthWriteEnable = VK_TRUE,
+ .depthCompareOp = VK_COMPARE_OP_ALWAYS,
+ .depthBoundsTestEnable = VK_FALSE,
+ .stencilTestEnable = VK_FALSE,
+ .front = VkStencilOpState{},
+ .back = VkStencilOpState{},
+ .minDepthBounds = 0.0f,
+ .maxDepthBounds = 0.0f,
+};
+
+template <VkFilter filter>
+inline constexpr VkSamplerCreateInfo SAMPLER_CREATE_INFO{
+ .sType = VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO,
+ .pNext = nullptr,
+ .flags = 0,
+ .magFilter = filter,
+ .minFilter = filter,
+ .mipmapMode = VK_SAMPLER_MIPMAP_MODE_NEAREST,
+ .addressModeU = VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER,
+ .addressModeV = VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER,
+ .addressModeW = VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER,
+ .mipLodBias = 0.0f,
+ .anisotropyEnable = VK_FALSE,
+ .maxAnisotropy = 0.0f,
+ .compareEnable = VK_FALSE,
+ .compareOp = VK_COMPARE_OP_NEVER,
+ .minLod = 0.0f,
+ .maxLod = 0.0f,
+ .borderColor = VK_BORDER_COLOR_FLOAT_OPAQUE_WHITE,
+ .unnormalizedCoordinates = VK_TRUE,
+};
+
+constexpr VkPipelineLayoutCreateInfo PipelineLayoutCreateInfo(
+ const VkDescriptorSetLayout* set_layout) {
+ return VkPipelineLayoutCreateInfo{
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO,
+ .pNext = nullptr,
+ .flags = 0,
+ .setLayoutCount = 1,
+ .pSetLayouts = set_layout,
+ .pushConstantRangeCount = 1,
+ .pPushConstantRanges = &PUSH_CONSTANT_RANGE,
+ };
+}
+
+constexpr VkPipelineShaderStageCreateInfo PipelineShaderStageCreateInfo(VkShaderStageFlagBits stage,
+ VkShaderModule shader) {
+ return VkPipelineShaderStageCreateInfo{
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO,
+ .pNext = nullptr,
+ .flags = 0,
+ .stage = stage,
+ .module = shader,
+ .pName = "main",
+ .pSpecializationInfo = nullptr,
+ };
+}
+
+constexpr std::array<VkPipelineShaderStageCreateInfo, 2> MakeStages(
+ VkShaderModule vertex_shader, VkShaderModule fragment_shader) {
+ return std::array{
+ PipelineShaderStageCreateInfo(VK_SHADER_STAGE_VERTEX_BIT, vertex_shader),
+ PipelineShaderStageCreateInfo(VK_SHADER_STAGE_FRAGMENT_BIT, fragment_shader),
+ };
+}
+
+void UpdateOneTextureDescriptorSet(const VKDevice& device, VkDescriptorSet descriptor_set,
+ VkSampler sampler, VkImageView image_view) {
+ const VkDescriptorImageInfo image_info{
+ .sampler = sampler,
+ .imageView = image_view,
+ .imageLayout = VK_IMAGE_LAYOUT_GENERAL,
+ };
+ const VkWriteDescriptorSet write_descriptor_set{
+ .sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET,
+ .pNext = nullptr,
+ .dstSet = descriptor_set,
+ .dstBinding = 0,
+ .dstArrayElement = 0,
+ .descriptorCount = 1,
+ .descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER,
+ .pImageInfo = &image_info,
+ .pBufferInfo = nullptr,
+ .pTexelBufferView = nullptr,
+ };
+ device.GetLogical().UpdateDescriptorSets(write_descriptor_set, nullptr);
+}
+
+void UpdateTwoTexturesDescriptorSet(const VKDevice& device, VkDescriptorSet descriptor_set,
+ VkSampler sampler, VkImageView image_view_0,
+ VkImageView image_view_1) {
+ const VkDescriptorImageInfo image_info_0{
+ .sampler = sampler,
+ .imageView = image_view_0,
+ .imageLayout = VK_IMAGE_LAYOUT_GENERAL,
+ };
+ const VkDescriptorImageInfo image_info_1{
+ .sampler = sampler,
+ .imageView = image_view_1,
+ .imageLayout = VK_IMAGE_LAYOUT_GENERAL,
+ };
+ const std::array write_descriptor_sets{
+ VkWriteDescriptorSet{
+ .sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET,
+ .pNext = nullptr,
+ .dstSet = descriptor_set,
+ .dstBinding = 0,
+ .dstArrayElement = 0,
+ .descriptorCount = 1,
+ .descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER,
+ .pImageInfo = &image_info_0,
+ .pBufferInfo = nullptr,
+ .pTexelBufferView = nullptr,
+ },
+ VkWriteDescriptorSet{
+ .sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET,
+ .pNext = nullptr,
+ .dstSet = descriptor_set,
+ .dstBinding = 1,
+ .dstArrayElement = 0,
+ .descriptorCount = 1,
+ .descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER,
+ .pImageInfo = &image_info_1,
+ .pBufferInfo = nullptr,
+ .pTexelBufferView = nullptr,
+ },
+ };
+ device.GetLogical().UpdateDescriptorSets(write_descriptor_sets, nullptr);
+}
+
+void BindBlitState(vk::CommandBuffer cmdbuf, VkPipelineLayout layout,
+ const std::array<Offset2D, 2>& dst_region,
+ const std::array<Offset2D, 2>& src_region) {
+ const VkOffset2D offset{
+ .x = std::min(dst_region[0].x, dst_region[1].x),
+ .y = std::min(dst_region[0].y, dst_region[1].y),
+ };
+ const VkExtent2D extent{
+ .width = static_cast<u32>(std::abs(dst_region[1].x - dst_region[0].x)),
+ .height = static_cast<u32>(std::abs(dst_region[1].y - dst_region[0].y)),
+ };
+ const VkViewport viewport{
+ .x = static_cast<float>(offset.x),
+ .y = static_cast<float>(offset.y),
+ .width = static_cast<float>(extent.width),
+ .height = static_cast<float>(extent.height),
+ .minDepth = 0.0f,
+ .maxDepth = 1.0f,
+ };
+ // TODO: Support scissored blits
+ const VkRect2D scissor{
+ .offset = offset,
+ .extent = extent,
+ };
+ const float scale_x = static_cast<float>(src_region[1].x - src_region[0].x);
+ const float scale_y = static_cast<float>(src_region[1].y - src_region[0].y);
+ const PushConstants push_constants{
+ .tex_scale = {scale_x, scale_y},
+ .tex_offset = {static_cast<float>(src_region[0].x), static_cast<float>(src_region[0].y)},
+ };
+ cmdbuf.SetViewport(0, viewport);
+ cmdbuf.SetScissor(0, scissor);
+ cmdbuf.PushConstants(layout, VK_SHADER_STAGE_VERTEX_BIT, push_constants);
+}
+
+} // Anonymous namespace
+
+BlitImageHelper::BlitImageHelper(const VKDevice& device_, VKScheduler& scheduler_,
+ StateTracker& state_tracker_, VKDescriptorPool& descriptor_pool)
+ : device{device_}, scheduler{scheduler_}, state_tracker{state_tracker_},
+ one_texture_set_layout(device.GetLogical().CreateDescriptorSetLayout(
+ ONE_TEXTURE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO)),
+ two_textures_set_layout(device.GetLogical().CreateDescriptorSetLayout(
+ TWO_TEXTURES_DESCRIPTOR_SET_LAYOUT_CREATE_INFO)),
+ one_texture_descriptor_allocator(descriptor_pool, *one_texture_set_layout),
+ two_textures_descriptor_allocator(descriptor_pool, *two_textures_set_layout),
+ one_texture_pipeline_layout(device.GetLogical().CreatePipelineLayout(
+ PipelineLayoutCreateInfo(one_texture_set_layout.address()))),
+ two_textures_pipeline_layout(device.GetLogical().CreatePipelineLayout(
+ PipelineLayoutCreateInfo(two_textures_set_layout.address()))),
+ full_screen_vert(BuildShader(device, FULL_SCREEN_TRIANGLE_VERT_SPV)),
+ blit_color_to_color_frag(BuildShader(device, VULKAN_BLIT_COLOR_FLOAT_FRAG_SPV)),
+ convert_depth_to_float_frag(BuildShader(device, CONVERT_DEPTH_TO_FLOAT_FRAG_SPV)),
+ convert_float_to_depth_frag(BuildShader(device, CONVERT_FLOAT_TO_DEPTH_FRAG_SPV)),
+ linear_sampler(device.GetLogical().CreateSampler(SAMPLER_CREATE_INFO<VK_FILTER_LINEAR>)),
+ nearest_sampler(device.GetLogical().CreateSampler(SAMPLER_CREATE_INFO<VK_FILTER_NEAREST>)) {
+ if (device.IsExtShaderStencilExportSupported()) {
+ blit_depth_stencil_frag = BuildShader(device, VULKAN_BLIT_DEPTH_STENCIL_FRAG_SPV);
+ }
+}
+
+BlitImageHelper::~BlitImageHelper() = default;
+
+void BlitImageHelper::BlitColor(const Framebuffer* dst_framebuffer, const ImageView& src_image_view,
+ const std::array<Offset2D, 2>& dst_region,
+ const std::array<Offset2D, 2>& src_region,
+ Tegra::Engines::Fermi2D::Filter filter,
+ Tegra::Engines::Fermi2D::Operation operation) {
+ const bool is_linear = filter == Tegra::Engines::Fermi2D::Filter::Bilinear;
+ const BlitImagePipelineKey key{
+ .renderpass = dst_framebuffer->RenderPass(),
+ .operation = operation,
+ };
+ const VkPipelineLayout layout = *one_texture_pipeline_layout;
+ const VkImageView src_view = src_image_view.Handle(ImageViewType::e2D);
+ const VkSampler sampler = is_linear ? *linear_sampler : *nearest_sampler;
+ const VkPipeline pipeline = FindOrEmplacePipeline(key);
+ const VkDescriptorSet descriptor_set = one_texture_descriptor_allocator.Commit();
+ scheduler.RequestRenderpass(dst_framebuffer);
+ scheduler.Record([dst_region, src_region, pipeline, layout, sampler, src_view, descriptor_set,
+ &device = device](vk::CommandBuffer cmdbuf) {
+ // TODO: Barriers
+ UpdateOneTextureDescriptorSet(device, descriptor_set, sampler, src_view);
+ cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
+ cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, layout, 0, descriptor_set,
+ nullptr);
+ BindBlitState(cmdbuf, layout, dst_region, src_region);
+ cmdbuf.Draw(3, 1, 0, 0);
+ });
+ scheduler.InvalidateState();
+}
+
+void BlitImageHelper::BlitDepthStencil(const Framebuffer* dst_framebuffer,
+ VkImageView src_depth_view, VkImageView src_stencil_view,
+ const std::array<Offset2D, 2>& dst_region,
+ const std::array<Offset2D, 2>& src_region,
+ Tegra::Engines::Fermi2D::Filter filter,
+ Tegra::Engines::Fermi2D::Operation operation) {
+ ASSERT(filter == Tegra::Engines::Fermi2D::Filter::Point);
+ ASSERT(operation == Tegra::Engines::Fermi2D::Operation::SrcCopy);
+
+ const VkPipelineLayout layout = *two_textures_pipeline_layout;
+ const VkSampler sampler = *nearest_sampler;
+ const VkPipeline pipeline = BlitDepthStencilPipeline(dst_framebuffer->RenderPass());
+ const VkDescriptorSet descriptor_set = two_textures_descriptor_allocator.Commit();
+ scheduler.RequestRenderpass(dst_framebuffer);
+ scheduler.Record([dst_region, src_region, pipeline, layout, sampler, src_depth_view,
+ src_stencil_view, descriptor_set,
+ &device = device](vk::CommandBuffer cmdbuf) {
+ // TODO: Barriers
+ UpdateTwoTexturesDescriptorSet(device, descriptor_set, sampler, src_depth_view,
+ src_stencil_view);
+ cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
+ cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, layout, 0, descriptor_set,
+ nullptr);
+ BindBlitState(cmdbuf, layout, dst_region, src_region);
+ cmdbuf.Draw(3, 1, 0, 0);
+ });
+ scheduler.InvalidateState();
+}
+
+void BlitImageHelper::ConvertD32ToR32(const Framebuffer* dst_framebuffer,
+ const ImageView& src_image_view) {
+ ConvertDepthToColorPipeline(convert_d32_to_r32_pipeline, dst_framebuffer->RenderPass());
+ Convert(*convert_d32_to_r32_pipeline, dst_framebuffer, src_image_view);
+}
+
+void BlitImageHelper::ConvertR32ToD32(const Framebuffer* dst_framebuffer,
+ const ImageView& src_image_view) {
+
+ ConvertColorToDepthPipeline(convert_r32_to_d32_pipeline, dst_framebuffer->RenderPass());
+ Convert(*convert_r32_to_d32_pipeline, dst_framebuffer, src_image_view);
+}
+
+void BlitImageHelper::ConvertD16ToR16(const Framebuffer* dst_framebuffer,
+ const ImageView& src_image_view) {
+ ConvertDepthToColorPipeline(convert_d16_to_r16_pipeline, dst_framebuffer->RenderPass());
+ Convert(*convert_d16_to_r16_pipeline, dst_framebuffer, src_image_view);
+}
+
+void BlitImageHelper::ConvertR16ToD16(const Framebuffer* dst_framebuffer,
+ const ImageView& src_image_view) {
+ ConvertColorToDepthPipeline(convert_r16_to_d16_pipeline, dst_framebuffer->RenderPass());
+ Convert(*convert_r16_to_d16_pipeline, dst_framebuffer, src_image_view);
+}
+
+void BlitImageHelper::Convert(VkPipeline pipeline, const Framebuffer* dst_framebuffer,
+ const ImageView& src_image_view) {
+ const VkPipelineLayout layout = *one_texture_pipeline_layout;
+ const VkImageView src_view = src_image_view.Handle(ImageViewType::e2D);
+ const VkSampler sampler = *nearest_sampler;
+ const VkDescriptorSet descriptor_set = one_texture_descriptor_allocator.Commit();
+ const VkExtent2D extent{
+ .width = src_image_view.size.width,
+ .height = src_image_view.size.height,
+ };
+ scheduler.RequestRenderpass(dst_framebuffer);
+ scheduler.Record([pipeline, layout, sampler, src_view, descriptor_set, extent,
+ &device = device](vk::CommandBuffer cmdbuf) {
+ const VkOffset2D offset{
+ .x = 0,
+ .y = 0,
+ };
+ const VkViewport viewport{
+ .x = 0.0f,
+ .y = 0.0f,
+ .width = static_cast<float>(extent.width),
+ .height = static_cast<float>(extent.height),
+ .minDepth = 0.0f,
+ .maxDepth = 0.0f,
+ };
+ const VkRect2D scissor{
+ .offset = offset,
+ .extent = extent,
+ };
+ const PushConstants push_constants{
+ .tex_scale = {viewport.width, viewport.height},
+ .tex_offset = {0.0f, 0.0f},
+ };
+ UpdateOneTextureDescriptorSet(device, descriptor_set, sampler, src_view);
+
+ // TODO: Barriers
+ cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
+ cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, layout, 0, descriptor_set,
+ nullptr);
+ cmdbuf.SetViewport(0, viewport);
+ cmdbuf.SetScissor(0, scissor);
+ cmdbuf.PushConstants(layout, VK_SHADER_STAGE_VERTEX_BIT, push_constants);
+ cmdbuf.Draw(3, 1, 0, 0);
+ });
+ scheduler.InvalidateState();
+}
+
+VkPipeline BlitImageHelper::FindOrEmplacePipeline(const BlitImagePipelineKey& key) {
+ const auto it = std::ranges::find(blit_color_keys, key);
+ if (it != blit_color_keys.end()) {
+ return *blit_color_pipelines[std::distance(blit_color_keys.begin(), it)];
+ }
+ blit_color_keys.push_back(key);
+
+ const std::array stages = MakeStages(*full_screen_vert, *blit_color_to_color_frag);
+ const VkPipelineColorBlendAttachmentState blend_attachment{
+ .blendEnable = VK_FALSE,
+ .srcColorBlendFactor = VK_BLEND_FACTOR_ZERO,
+ .dstColorBlendFactor = VK_BLEND_FACTOR_ZERO,
+ .colorBlendOp = VK_BLEND_OP_ADD,
+ .srcAlphaBlendFactor = VK_BLEND_FACTOR_ZERO,
+ .dstAlphaBlendFactor = VK_BLEND_FACTOR_ZERO,
+ .alphaBlendOp = VK_BLEND_OP_ADD,
+ .colorWriteMask = VK_COLOR_COMPONENT_R_BIT | VK_COLOR_COMPONENT_G_BIT |
+ VK_COLOR_COMPONENT_B_BIT | VK_COLOR_COMPONENT_A_BIT,
+ };
+ // TODO: programmable blending
+ const VkPipelineColorBlendStateCreateInfo color_blend_create_info{
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO,
+ .pNext = nullptr,
+ .flags = 0,
+ .logicOpEnable = VK_FALSE,
+ .logicOp = VK_LOGIC_OP_CLEAR,
+ .attachmentCount = 1,
+ .pAttachments = &blend_attachment,
+ .blendConstants = {0.0f, 0.0f, 0.0f, 0.0f},
+ };
+ blit_color_pipelines.push_back(device.GetLogical().CreateGraphicsPipeline({
+ .sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO,
+ .pNext = nullptr,
+ .flags = 0,
+ .stageCount = static_cast<u32>(stages.size()),
+ .pStages = stages.data(),
+ .pVertexInputState = &PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO,
+ .pInputAssemblyState = &PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO,
+ .pTessellationState = nullptr,
+ .pViewportState = &PIPELINE_VIEWPORT_STATE_CREATE_INFO,
+ .pRasterizationState = &PIPELINE_RASTERIZATION_STATE_CREATE_INFO,
+ .pMultisampleState = &PIPELINE_MULTISAMPLE_STATE_CREATE_INFO,
+ .pDepthStencilState = nullptr,
+ .pColorBlendState = &color_blend_create_info,
+ .pDynamicState = &PIPELINE_DYNAMIC_STATE_CREATE_INFO,
+ .layout = *one_texture_pipeline_layout,
+ .renderPass = key.renderpass,
+ .subpass = 0,
+ .basePipelineHandle = VK_NULL_HANDLE,
+ .basePipelineIndex = 0,
+ }));
+ return *blit_color_pipelines.back();
+}
+
+VkPipeline BlitImageHelper::BlitDepthStencilPipeline(VkRenderPass renderpass) {
+ if (blit_depth_stencil_pipeline) {
+ return *blit_depth_stencil_pipeline;
+ }
+ const std::array stages = MakeStages(*full_screen_vert, *blit_depth_stencil_frag);
+ blit_depth_stencil_pipeline = device.GetLogical().CreateGraphicsPipeline({
+ .sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO,
+ .pNext = nullptr,
+ .flags = 0,
+ .stageCount = static_cast<u32>(stages.size()),
+ .pStages = stages.data(),
+ .pVertexInputState = &PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO,
+ .pInputAssemblyState = &PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO,
+ .pTessellationState = nullptr,
+ .pViewportState = &PIPELINE_VIEWPORT_STATE_CREATE_INFO,
+ .pRasterizationState = &PIPELINE_RASTERIZATION_STATE_CREATE_INFO,
+ .pMultisampleState = &PIPELINE_MULTISAMPLE_STATE_CREATE_INFO,
+ .pDepthStencilState = &PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO,
+ .pColorBlendState = &PIPELINE_COLOR_BLEND_STATE_EMPTY_CREATE_INFO,
+ .pDynamicState = &PIPELINE_DYNAMIC_STATE_CREATE_INFO,
+ .layout = *two_textures_pipeline_layout,
+ .renderPass = renderpass,
+ .subpass = 0,
+ .basePipelineHandle = VK_NULL_HANDLE,
+ .basePipelineIndex = 0,
+ });
+ return *blit_depth_stencil_pipeline;
+}
+
+void BlitImageHelper::ConvertDepthToColorPipeline(vk::Pipeline& pipeline, VkRenderPass renderpass) {
+ if (pipeline) {
+ return;
+ }
+ const std::array stages = MakeStages(*full_screen_vert, *convert_depth_to_float_frag);
+ pipeline = device.GetLogical().CreateGraphicsPipeline({
+ .sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO,
+ .pNext = nullptr,
+ .flags = 0,
+ .stageCount = static_cast<u32>(stages.size()),
+ .pStages = stages.data(),
+ .pVertexInputState = &PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO,
+ .pInputAssemblyState = &PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO,
+ .pTessellationState = nullptr,
+ .pViewportState = &PIPELINE_VIEWPORT_STATE_CREATE_INFO,
+ .pRasterizationState = &PIPELINE_RASTERIZATION_STATE_CREATE_INFO,
+ .pMultisampleState = &PIPELINE_MULTISAMPLE_STATE_CREATE_INFO,
+ .pDepthStencilState = nullptr,
+ .pColorBlendState = &PIPELINE_COLOR_BLEND_STATE_GENERIC_CREATE_INFO,
+ .pDynamicState = &PIPELINE_DYNAMIC_STATE_CREATE_INFO,
+ .layout = *one_texture_pipeline_layout,
+ .renderPass = renderpass,
+ .subpass = 0,
+ .basePipelineHandle = VK_NULL_HANDLE,
+ .basePipelineIndex = 0,
+ });
+}
+
+void BlitImageHelper::ConvertColorToDepthPipeline(vk::Pipeline& pipeline, VkRenderPass renderpass) {
+ if (pipeline) {
+ return;
+ }
+ const std::array stages = MakeStages(*full_screen_vert, *convert_float_to_depth_frag);
+ pipeline = device.GetLogical().CreateGraphicsPipeline({
+ .sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO,
+ .pNext = nullptr,
+ .flags = 0,
+ .stageCount = static_cast<u32>(stages.size()),
+ .pStages = stages.data(),
+ .pVertexInputState = &PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO,
+ .pInputAssemblyState = &PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO,
+ .pTessellationState = nullptr,
+ .pViewportState = &PIPELINE_VIEWPORT_STATE_CREATE_INFO,
+ .pRasterizationState = &PIPELINE_RASTERIZATION_STATE_CREATE_INFO,
+ .pMultisampleState = &PIPELINE_MULTISAMPLE_STATE_CREATE_INFO,
+ .pDepthStencilState = &PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO,
+ .pColorBlendState = &PIPELINE_COLOR_BLEND_STATE_EMPTY_CREATE_INFO,
+ .pDynamicState = &PIPELINE_DYNAMIC_STATE_CREATE_INFO,
+ .layout = *one_texture_pipeline_layout,
+ .renderPass = renderpass,
+ .subpass = 0,
+ .basePipelineHandle = VK_NULL_HANDLE,
+ .basePipelineIndex = 0,
+ });
+}
+
+} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/blit_image.h b/src/video_core/renderer_vulkan/blit_image.h
new file mode 100644
index 000000000..2c2790bf9
--- /dev/null
+++ b/src/video_core/renderer_vulkan/blit_image.h
@@ -0,0 +1,97 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <compare>
+
+#include "video_core/engines/fermi_2d.h"
+#include "video_core/renderer_vulkan/vk_descriptor_pool.h"
+#include "video_core/renderer_vulkan/wrapper.h"
+#include "video_core/texture_cache/types.h"
+
+namespace Vulkan {
+
+using VideoCommon::Offset2D;
+
+class VKDevice;
+class VKScheduler;
+class StateTracker;
+
+class Framebuffer;
+class ImageView;
+
+struct BlitImagePipelineKey {
+ constexpr auto operator<=>(const BlitImagePipelineKey&) const noexcept = default;
+
+ VkRenderPass renderpass;
+ Tegra::Engines::Fermi2D::Operation operation;
+};
+
+class BlitImageHelper {
+public:
+ explicit BlitImageHelper(const VKDevice& device, VKScheduler& scheduler,
+ StateTracker& state_tracker, VKDescriptorPool& descriptor_pool);
+ ~BlitImageHelper();
+
+ void BlitColor(const Framebuffer* dst_framebuffer, const ImageView& src_image_view,
+ const std::array<Offset2D, 2>& dst_region,
+ const std::array<Offset2D, 2>& src_region,
+ Tegra::Engines::Fermi2D::Filter filter,
+ Tegra::Engines::Fermi2D::Operation operation);
+
+ void BlitDepthStencil(const Framebuffer* dst_framebuffer, VkImageView src_depth_view,
+ VkImageView src_stencil_view, const std::array<Offset2D, 2>& dst_region,
+ const std::array<Offset2D, 2>& src_region,
+ Tegra::Engines::Fermi2D::Filter filter,
+ Tegra::Engines::Fermi2D::Operation operation);
+
+ void ConvertD32ToR32(const Framebuffer* dst_framebuffer, const ImageView& src_image_view);
+
+ void ConvertR32ToD32(const Framebuffer* dst_framebuffer, const ImageView& src_image_view);
+
+ void ConvertD16ToR16(const Framebuffer* dst_framebuffer, const ImageView& src_image_view);
+
+ void ConvertR16ToD16(const Framebuffer* dst_framebuffer, const ImageView& src_image_view);
+
+private:
+ void Convert(VkPipeline pipeline, const Framebuffer* dst_framebuffer,
+ const ImageView& src_image_view);
+
+ [[nodiscard]] VkPipeline FindOrEmplacePipeline(const BlitImagePipelineKey& key);
+
+ [[nodiscard]] VkPipeline BlitDepthStencilPipeline(VkRenderPass renderpass);
+
+ void ConvertDepthToColorPipeline(vk::Pipeline& pipeline, VkRenderPass renderpass);
+
+ void ConvertColorToDepthPipeline(vk::Pipeline& pipeline, VkRenderPass renderpass);
+
+ const VKDevice& device;
+ VKScheduler& scheduler;
+ StateTracker& state_tracker;
+
+ vk::DescriptorSetLayout one_texture_set_layout;
+ vk::DescriptorSetLayout two_textures_set_layout;
+ DescriptorAllocator one_texture_descriptor_allocator;
+ DescriptorAllocator two_textures_descriptor_allocator;
+ vk::PipelineLayout one_texture_pipeline_layout;
+ vk::PipelineLayout two_textures_pipeline_layout;
+ vk::ShaderModule full_screen_vert;
+ vk::ShaderModule blit_color_to_color_frag;
+ vk::ShaderModule blit_depth_stencil_frag;
+ vk::ShaderModule convert_depth_to_float_frag;
+ vk::ShaderModule convert_float_to_depth_frag;
+ vk::Sampler linear_sampler;
+ vk::Sampler nearest_sampler;
+
+ std::vector<BlitImagePipelineKey> blit_color_keys;
+ std::vector<vk::Pipeline> blit_color_pipelines;
+ vk::Pipeline blit_depth_stencil_pipeline;
+ vk::Pipeline convert_d32_to_r32_pipeline;
+ vk::Pipeline convert_r32_to_d32_pipeline;
+ vk::Pipeline convert_d16_to_r16_pipeline;
+ vk::Pipeline convert_r16_to_d16_pipeline;
+};
+
+} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp b/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp
index 5ec43db11..67dd10500 100644
--- a/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp
+++ b/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp
@@ -60,6 +60,7 @@ void FixedPipelineState::Fill(const Maxwell& regs, bool has_extended_dynamic_sta
logic_op.Assign(PackLogicOp(regs.logic_op.operation));
rasterize_enable.Assign(regs.rasterize_enable != 0 ? 1 : 0);
topology.Assign(regs.draw.topology);
+ msaa_mode.Assign(regs.multisample_mode);
raw2 = 0;
const auto test_func =
diff --git a/src/video_core/renderer_vulkan/fixed_pipeline_state.h b/src/video_core/renderer_vulkan/fixed_pipeline_state.h
index c26b77790..7e95e6fce 100644
--- a/src/video_core/renderer_vulkan/fixed_pipeline_state.h
+++ b/src/video_core/renderer_vulkan/fixed_pipeline_state.h
@@ -186,6 +186,7 @@ struct FixedPipelineState {
BitField<19, 4, u32> logic_op;
BitField<23, 1, u32> rasterize_enable;
BitField<24, 4, Maxwell::PrimitiveTopology> topology;
+ BitField<28, 4, Tegra::Texture::MsaaMode> msaa_mode;
};
union {
u32 raw2;
diff --git a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
index 58e117eb3..40501e7fa 100644
--- a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
+++ b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
@@ -122,7 +122,7 @@ struct FormatTuple {
{VK_FORMAT_A8B8G8R8_SINT_PACK32, Attachable | Storage}, // A8B8G8R8_SINT
{VK_FORMAT_A8B8G8R8_UINT_PACK32, Attachable | Storage}, // A8B8G8R8_UINT
{VK_FORMAT_R5G6B5_UNORM_PACK16, Attachable}, // R5G6B5_UNORM
- {VK_FORMAT_B5G6R5_UNORM_PACK16, Attachable}, // B5G6R5_UNORM
+ {VK_FORMAT_B5G6R5_UNORM_PACK16}, // B5G6R5_UNORM
{VK_FORMAT_A1R5G5B5_UNORM_PACK16, Attachable}, // A1R5G5B5_UNORM
{VK_FORMAT_A2B10G10R10_UNORM_PACK32, Attachable | Storage}, // A2B10G10R10_UNORM
{VK_FORMAT_A2B10G10R10_UINT_PACK32, Attachable | Storage}, // A2B10G10R10_UINT
@@ -163,7 +163,7 @@ struct FormatTuple {
{VK_FORMAT_R16G16_UNORM, Attachable | Storage}, // R16G16_UNORM
{VK_FORMAT_R16G16_SFLOAT, Attachable | Storage}, // R16G16_FLOAT
{VK_FORMAT_UNDEFINED}, // R16G16_UINT
- {VK_FORMAT_UNDEFINED}, // R16G16_SINT
+ {VK_FORMAT_R16G16_SINT, Attachable | Storage}, // R16G16_SINT
{VK_FORMAT_R16G16_SNORM, Attachable | Storage}, // R16G16_SNORM
{VK_FORMAT_UNDEFINED}, // R32G32B32_FLOAT
{VK_FORMAT_R8G8B8A8_SRGB, Attachable}, // A8B8G8R8_SRGB
@@ -233,18 +233,20 @@ FormatInfo SurfaceFormat(const VKDevice& device, FormatType format_type, PixelFo
// Use A8B8G8R8_UNORM on hardware that doesn't support ASTC natively
if (!device.IsOptimalAstcSupported() && VideoCore::Surface::IsPixelFormatASTC(pixel_format)) {
- tuple.format = VideoCore::Surface::IsPixelFormatSRGB(pixel_format)
- ? VK_FORMAT_A8B8G8R8_SRGB_PACK32
- : VK_FORMAT_A8B8G8R8_UNORM_PACK32;
+ const bool is_srgb = VideoCore::Surface::IsPixelFormatSRGB(pixel_format);
+ tuple.format = is_srgb ? VK_FORMAT_A8B8G8R8_SRGB_PACK32 : VK_FORMAT_A8B8G8R8_UNORM_PACK32;
}
const bool attachable = tuple.usage & Attachable;
const bool storage = tuple.usage & Storage;
VkFormatFeatureFlags usage;
- if (format_type == FormatType::Buffer) {
+ switch (format_type) {
+ case FormatType::Buffer:
usage =
VK_FORMAT_FEATURE_STORAGE_TEXEL_BUFFER_BIT | VK_FORMAT_FEATURE_UNIFORM_TEXEL_BUFFER_BIT;
- } else {
+ break;
+ case FormatType::Linear:
+ case FormatType::Optimal:
usage = VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT | VK_FORMAT_FEATURE_TRANSFER_DST_BIT |
VK_FORMAT_FEATURE_TRANSFER_SRC_BIT;
if (attachable) {
@@ -254,6 +256,7 @@ FormatInfo SurfaceFormat(const VKDevice& device, FormatType format_type, PixelFo
if (storage) {
usage |= VK_FORMAT_FEATURE_STORAGE_IMAGE_BIT;
}
+ break;
}
return {device.GetSupportedFormat(tuple.format, usage, format_type), attachable, storage};
}
@@ -724,4 +727,17 @@ VkViewportCoordinateSwizzleNV ViewportSwizzle(Maxwell::ViewportSwizzle swizzle)
return {};
}
+VkSamplerReductionMode SamplerReduction(Tegra::Texture::SamplerReduction reduction) {
+ switch (reduction) {
+ case Tegra::Texture::SamplerReduction::WeightedAverage:
+ return VK_SAMPLER_REDUCTION_MODE_WEIGHTED_AVERAGE_EXT;
+ case Tegra::Texture::SamplerReduction::Min:
+ return VK_SAMPLER_REDUCTION_MODE_MIN_EXT;
+ case Tegra::Texture::SamplerReduction::Max:
+ return VK_SAMPLER_REDUCTION_MODE_MAX_EXT;
+ }
+ UNREACHABLE_MSG("Invalid sampler mode={}", static_cast<int>(reduction));
+ return VK_SAMPLER_REDUCTION_MODE_WEIGHTED_AVERAGE_EXT;
+}
+
} // namespace Vulkan::MaxwellToVK
diff --git a/src/video_core/renderer_vulkan/maxwell_to_vk.h b/src/video_core/renderer_vulkan/maxwell_to_vk.h
index 7e213452f..1a90f192e 100644
--- a/src/video_core/renderer_vulkan/maxwell_to_vk.h
+++ b/src/video_core/renderer_vulkan/maxwell_to_vk.h
@@ -61,4 +61,6 @@ VkComponentSwizzle SwizzleSource(Tegra::Texture::SwizzleSource swizzle);
VkViewportCoordinateSwizzleNV ViewportSwizzle(Maxwell::ViewportSwizzle swizzle);
+VkSamplerReductionMode SamplerReduction(Tegra::Texture::SamplerReduction reduction);
+
} // namespace Vulkan::MaxwellToVK
diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.cpp b/src/video_core/renderer_vulkan/renderer_vulkan.cpp
index ea4b7c1e6..7f521cb9b 100644
--- a/src/video_core/renderer_vulkan/renderer_vulkan.cpp
+++ b/src/video_core/renderer_vulkan/renderer_vulkan.cpp
@@ -92,9 +92,9 @@ Common::DynamicLibrary OpenVulkanLibrary() {
return library;
}
-std::pair<vk::Instance, u32> CreateInstance(
- Common::DynamicLibrary& library, vk::InstanceDispatch& dld,
- WindowSystemType window_type = WindowSystemType::Headless, bool enable_layers = false) {
+std::pair<vk::Instance, u32> CreateInstance(Common::DynamicLibrary& library,
+ vk::InstanceDispatch& dld, WindowSystemType window_type,
+ bool enable_debug_utils, bool enable_layers) {
if (!library.IsOpen()) {
LOG_ERROR(Render_Vulkan, "Vulkan library not available");
return {};
@@ -133,7 +133,7 @@ std::pair<vk::Instance, u32> CreateInstance(
if (window_type != Core::Frontend::WindowSystemType::Headless) {
extensions.push_back(VK_KHR_SURFACE_EXTENSION_NAME);
}
- if (enable_layers) {
+ if (enable_debug_utils) {
extensions.push_back(VK_EXT_DEBUG_UTILS_EXTENSION_NAME);
}
extensions.push_back(VK_KHR_GET_PHYSICAL_DEVICE_PROPERTIES_2_EXTENSION_NAME);
@@ -287,7 +287,7 @@ void RendererVulkan::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
bool RendererVulkan::Init() {
library = OpenVulkanLibrary();
std::tie(instance, instance_version) = CreateInstance(
- library, dld, render_window.GetWindowInfo().type, Settings::values.renderer_debug);
+ library, dld, render_window.GetWindowInfo().type, true, Settings::values.renderer_debug);
if (!instance || !CreateDebugCallback() || !CreateSurface() || !PickDevices()) {
return false;
}
@@ -447,7 +447,8 @@ void RendererVulkan::Report() const {
std::vector<std::string> RendererVulkan::EnumerateDevices() {
vk::InstanceDispatch dld;
Common::DynamicLibrary library = OpenVulkanLibrary();
- vk::Instance instance = CreateInstance(library, dld).first;
+ vk::Instance instance =
+ CreateInstance(library, dld, WindowSystemType::Headless, false, false).first;
if (!instance) {
return {};
}
diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.h b/src/video_core/renderer_vulkan/renderer_vulkan.h
index 977b86003..74642fba4 100644
--- a/src/video_core/renderer_vulkan/renderer_vulkan.h
+++ b/src/video_core/renderer_vulkan/renderer_vulkan.h
@@ -33,10 +33,9 @@ class VKDevice;
class VKMemoryManager;
class VKSwapchain;
class VKScheduler;
-class VKImage;
struct VKScreenInfo {
- VKImage* image{};
+ VkImageView image_view{};
u32 width{};
u32 height{};
bool is_srgb{};
diff --git a/src/video_core/renderer_vulkan/shaders/blit.frag b/src/video_core/renderer_vulkan/shaders/blit.frag
deleted file mode 100644
index a06ecd24a..000000000
--- a/src/video_core/renderer_vulkan/shaders/blit.frag
+++ /dev/null
@@ -1,24 +0,0 @@
-// Copyright 2019 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-/*
- * Build instructions:
- * $ glslangValidator -V $THIS_FILE -o output.spv
- * $ spirv-opt -O --strip-debug output.spv -o optimized.spv
- * $ xxd -i optimized.spv
- *
- * Then copy that bytecode to the C++ file
- */
-
-#version 460 core
-
-layout (location = 0) in vec2 frag_tex_coord;
-
-layout (location = 0) out vec4 color;
-
-layout (binding = 1) uniform sampler2D color_texture;
-
-void main() {
- color = texture(color_texture, frag_tex_coord);
-}
diff --git a/src/video_core/renderer_vulkan/shaders/blit.vert b/src/video_core/renderer_vulkan/shaders/blit.vert
deleted file mode 100644
index c64d9235a..000000000
--- a/src/video_core/renderer_vulkan/shaders/blit.vert
+++ /dev/null
@@ -1,28 +0,0 @@
-// Copyright 2019 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-/*
- * Build instructions:
- * $ glslangValidator -V $THIS_FILE -o output.spv
- * $ spirv-opt -O --strip-debug output.spv -o optimized.spv
- * $ xxd -i optimized.spv
- *
- * Then copy that bytecode to the C++ file
- */
-
-#version 460 core
-
-layout (location = 0) in vec2 vert_position;
-layout (location = 1) in vec2 vert_tex_coord;
-
-layout (location = 0) out vec2 frag_tex_coord;
-
-layout (set = 0, binding = 0) uniform MatrixBlock {
- mat4 modelview_matrix;
-};
-
-void main() {
- gl_Position = modelview_matrix * vec4(vert_position, 0.0, 1.0);
- frag_tex_coord = vert_tex_coord;
-}
diff --git a/src/video_core/renderer_vulkan/shaders/quad_array.comp b/src/video_core/renderer_vulkan/shaders/quad_array.comp
deleted file mode 100644
index 5a5703308..000000000
--- a/src/video_core/renderer_vulkan/shaders/quad_array.comp
+++ /dev/null
@@ -1,37 +0,0 @@
-// Copyright 2019 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-/*
- * Build instructions:
- * $ glslangValidator -V $THIS_FILE -o output.spv
- * $ spirv-opt -O --strip-debug output.spv -o optimized.spv
- * $ xxd -i optimized.spv
- *
- * Then copy that bytecode to the C++ file
- */
-
-#version 460 core
-
-layout (local_size_x = 1024) in;
-
-layout (std430, set = 0, binding = 0) buffer OutputBuffer {
- uint output_indexes[];
-};
-
-layout (push_constant) uniform PushConstants {
- uint first;
-};
-
-void main() {
- uint primitive = gl_GlobalInvocationID.x;
- if (primitive * 6 >= output_indexes.length()) {
- return;
- }
-
- const uint quad_map[6] = uint[](0, 1, 2, 0, 2, 3);
- for (uint vertex = 0; vertex < 6; ++vertex) {
- uint index = first + primitive * 4 + quad_map[vertex];
- output_indexes[primitive * 6 + vertex] = index;
- }
-}
diff --git a/src/video_core/renderer_vulkan/shaders/quad_indexed.comp b/src/video_core/renderer_vulkan/shaders/quad_indexed.comp
deleted file mode 100644
index 5a472ba9b..000000000
--- a/src/video_core/renderer_vulkan/shaders/quad_indexed.comp
+++ /dev/null
@@ -1,50 +0,0 @@
-// Copyright 2020 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-/*
- * Build instructions:
- * $ glslangValidator -V quad_indexed.comp -o output.spv
- * $ spirv-opt -O --strip-debug output.spv -o optimized.spv
- * $ xxd -i optimized.spv
- *
- * Then copy that bytecode to the C++ file
- */
-
-#version 460 core
-
-layout (local_size_x = 1024) in;
-
-layout (std430, set = 0, binding = 0) readonly buffer InputBuffer {
- uint input_indexes[];
-};
-
-layout (std430, set = 0, binding = 1) writeonly buffer OutputBuffer {
- uint output_indexes[];
-};
-
-layout (push_constant) uniform PushConstants {
- uint base_vertex;
- int index_shift; // 0: uint8, 1: uint16, 2: uint32
-};
-
-void main() {
- int primitive = int(gl_GlobalInvocationID.x);
- if (primitive * 6 >= output_indexes.length()) {
- return;
- }
-
- int index_size = 8 << index_shift;
- int flipped_shift = 2 - index_shift;
- int mask = (1 << flipped_shift) - 1;
-
- const int quad_swizzle[6] = int[](0, 1, 2, 0, 2, 3);
- for (uint vertex = 0; vertex < 6; ++vertex) {
- int offset = primitive * 4 + quad_swizzle[vertex];
- int int_offset = offset >> flipped_shift;
- int bit_offset = (offset & mask) * index_size;
- uint packed_input = input_indexes[int_offset];
- uint index = bitfieldExtract(packed_input, bit_offset, index_size);
- output_indexes[primitive * 6 + vertex] = index + base_vertex;
- }
-}
diff --git a/src/video_core/renderer_vulkan/shaders/uint8.comp b/src/video_core/renderer_vulkan/shaders/uint8.comp
deleted file mode 100644
index a320f3ae0..000000000
--- a/src/video_core/renderer_vulkan/shaders/uint8.comp
+++ /dev/null
@@ -1,33 +0,0 @@
-// Copyright 2019 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-/*
- * Build instructions:
- * $ glslangValidator -V $THIS_FILE -o output.spv
- * $ spirv-opt -O --strip-debug output.spv -o optimized.spv
- * $ xxd -i optimized.spv
- *
- * Then copy that bytecode to the C++ file
- */
-
-#version 460 core
-#extension GL_EXT_shader_16bit_storage : require
-#extension GL_EXT_shader_8bit_storage : require
-
-layout (local_size_x = 1024) in;
-
-layout (std430, set = 0, binding = 0) readonly buffer InputBuffer {
- uint8_t input_indexes[];
-};
-
-layout (std430, set = 0, binding = 1) writeonly buffer OutputBuffer {
- uint16_t output_indexes[];
-};
-
-void main() {
- uint id = gl_GlobalInvocationID.x;
- if (id < input_indexes.length()) {
- output_indexes[id] = uint16_t(input_indexes[id]);
- }
-}
diff --git a/src/video_core/renderer_vulkan/vk_blit_screen.cpp b/src/video_core/renderer_vulkan/vk_blit_screen.cpp
index b5b60309e..d3a83f22f 100644
--- a/src/video_core/renderer_vulkan/vk_blit_screen.cpp
+++ b/src/video_core/renderer_vulkan/vk_blit_screen.cpp
@@ -16,12 +16,12 @@
#include "core/frontend/emu_window.h"
#include "core/memory.h"
#include "video_core/gpu.h"
-#include "video_core/morton.h"
+#include "video_core/host_shaders/vulkan_present_frag_spv.h"
+#include "video_core/host_shaders/vulkan_present_vert_spv.h"
#include "video_core/rasterizer_interface.h"
#include "video_core/renderer_vulkan/renderer_vulkan.h"
#include "video_core/renderer_vulkan/vk_blit_screen.h"
#include "video_core/renderer_vulkan/vk_device.h"
-#include "video_core/renderer_vulkan/vk_image.h"
#include "video_core/renderer_vulkan/vk_master_semaphore.h"
#include "video_core/renderer_vulkan/vk_memory_manager.h"
#include "video_core/renderer_vulkan/vk_scheduler.h"
@@ -29,108 +29,12 @@
#include "video_core/renderer_vulkan/vk_swapchain.h"
#include "video_core/renderer_vulkan/wrapper.h"
#include "video_core/surface.h"
+#include "video_core/textures/decoders.h"
namespace Vulkan {
namespace {
-// Generated from the "shaders/" directory, read the instructions there.
-constexpr u8 blit_vertex_code[] = {
- 0x03, 0x02, 0x23, 0x07, 0x00, 0x00, 0x01, 0x00, 0x07, 0x00, 0x08, 0x00, 0x27, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x11, 0x00, 0x02, 0x00, 0x01, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x06, 0x00,
- 0x01, 0x00, 0x00, 0x00, 0x47, 0x4c, 0x53, 0x4c, 0x2e, 0x73, 0x74, 0x64, 0x2e, 0x34, 0x35, 0x30,
- 0x00, 0x00, 0x00, 0x00, 0x0e, 0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
- 0x0f, 0x00, 0x09, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x6d, 0x61, 0x69, 0x6e,
- 0x00, 0x00, 0x00, 0x00, 0x0d, 0x00, 0x00, 0x00, 0x19, 0x00, 0x00, 0x00, 0x24, 0x00, 0x00, 0x00,
- 0x25, 0x00, 0x00, 0x00, 0x48, 0x00, 0x05, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
- 0x0b, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x48, 0x00, 0x05, 0x00, 0x0b, 0x00, 0x00, 0x00,
- 0x01, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x48, 0x00, 0x05, 0x00,
- 0x0b, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00,
- 0x48, 0x00, 0x05, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00,
- 0x04, 0x00, 0x00, 0x00, 0x47, 0x00, 0x03, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00,
- 0x48, 0x00, 0x04, 0x00, 0x11, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00,
- 0x48, 0x00, 0x05, 0x00, 0x11, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x23, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x48, 0x00, 0x05, 0x00, 0x11, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
- 0x07, 0x00, 0x00, 0x00, 0x10, 0x00, 0x00, 0x00, 0x47, 0x00, 0x03, 0x00, 0x11, 0x00, 0x00, 0x00,
- 0x02, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x13, 0x00, 0x00, 0x00, 0x22, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x13, 0x00, 0x00, 0x00, 0x21, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x19, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x24, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x25, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x00, 0x00,
- 0x01, 0x00, 0x00, 0x00, 0x13, 0x00, 0x02, 0x00, 0x02, 0x00, 0x00, 0x00, 0x21, 0x00, 0x03, 0x00,
- 0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x16, 0x00, 0x03, 0x00, 0x06, 0x00, 0x00, 0x00,
- 0x20, 0x00, 0x00, 0x00, 0x17, 0x00, 0x04, 0x00, 0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00,
- 0x04, 0x00, 0x00, 0x00, 0x15, 0x00, 0x04, 0x00, 0x08, 0x00, 0x00, 0x00, 0x20, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x08, 0x00, 0x00, 0x00, 0x09, 0x00, 0x00, 0x00,
- 0x01, 0x00, 0x00, 0x00, 0x1c, 0x00, 0x04, 0x00, 0x0a, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00,
- 0x09, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x06, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x0a, 0x00, 0x00, 0x00, 0x0a, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00,
- 0x0c, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00,
- 0x0c, 0x00, 0x00, 0x00, 0x0d, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x15, 0x00, 0x04, 0x00,
- 0x0e, 0x00, 0x00, 0x00, 0x20, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00,
- 0x0e, 0x00, 0x00, 0x00, 0x0f, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x18, 0x00, 0x04, 0x00,
- 0x10, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x03, 0x00,
- 0x11, 0x00, 0x00, 0x00, 0x10, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x12, 0x00, 0x00, 0x00,
- 0x02, 0x00, 0x00, 0x00, 0x11, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x12, 0x00, 0x00, 0x00,
- 0x13, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x14, 0x00, 0x00, 0x00,
- 0x02, 0x00, 0x00, 0x00, 0x10, 0x00, 0x00, 0x00, 0x17, 0x00, 0x04, 0x00, 0x17, 0x00, 0x00, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x18, 0x00, 0x00, 0x00,
- 0x01, 0x00, 0x00, 0x00, 0x17, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x18, 0x00, 0x00, 0x00,
- 0x19, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00,
- 0x1b, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00,
- 0x1c, 0x00, 0x00, 0x00, 0x00, 0x00, 0x80, 0x3f, 0x20, 0x00, 0x04, 0x00, 0x21, 0x00, 0x00, 0x00,
- 0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x23, 0x00, 0x00, 0x00,
- 0x03, 0x00, 0x00, 0x00, 0x17, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x23, 0x00, 0x00, 0x00,
- 0x24, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x18, 0x00, 0x00, 0x00,
- 0x25, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x36, 0x00, 0x05, 0x00, 0x02, 0x00, 0x00, 0x00,
- 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00,
- 0x05, 0x00, 0x00, 0x00, 0x41, 0x00, 0x05, 0x00, 0x14, 0x00, 0x00, 0x00, 0x15, 0x00, 0x00, 0x00,
- 0x13, 0x00, 0x00, 0x00, 0x0f, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00, 0x10, 0x00, 0x00, 0x00,
- 0x16, 0x00, 0x00, 0x00, 0x15, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00, 0x17, 0x00, 0x00, 0x00,
- 0x1a, 0x00, 0x00, 0x00, 0x19, 0x00, 0x00, 0x00, 0x51, 0x00, 0x05, 0x00, 0x06, 0x00, 0x00, 0x00,
- 0x1d, 0x00, 0x00, 0x00, 0x1a, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x51, 0x00, 0x05, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x00, 0x00, 0x1a, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
- 0x50, 0x00, 0x07, 0x00, 0x07, 0x00, 0x00, 0x00, 0x1f, 0x00, 0x00, 0x00, 0x1d, 0x00, 0x00, 0x00,
- 0x1e, 0x00, 0x00, 0x00, 0x1b, 0x00, 0x00, 0x00, 0x1c, 0x00, 0x00, 0x00, 0x91, 0x00, 0x05, 0x00,
- 0x07, 0x00, 0x00, 0x00, 0x20, 0x00, 0x00, 0x00, 0x16, 0x00, 0x00, 0x00, 0x1f, 0x00, 0x00, 0x00,
- 0x41, 0x00, 0x05, 0x00, 0x21, 0x00, 0x00, 0x00, 0x22, 0x00, 0x00, 0x00, 0x0d, 0x00, 0x00, 0x00,
- 0x0f, 0x00, 0x00, 0x00, 0x3e, 0x00, 0x03, 0x00, 0x22, 0x00, 0x00, 0x00, 0x20, 0x00, 0x00, 0x00,
- 0x3d, 0x00, 0x04, 0x00, 0x17, 0x00, 0x00, 0x00, 0x26, 0x00, 0x00, 0x00, 0x25, 0x00, 0x00, 0x00,
- 0x3e, 0x00, 0x03, 0x00, 0x24, 0x00, 0x00, 0x00, 0x26, 0x00, 0x00, 0x00, 0xfd, 0x00, 0x01, 0x00,
- 0x38, 0x00, 0x01, 0x00};
-
-constexpr u8 blit_fragment_code[] = {
- 0x03, 0x02, 0x23, 0x07, 0x00, 0x00, 0x01, 0x00, 0x07, 0x00, 0x08, 0x00, 0x14, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x11, 0x00, 0x02, 0x00, 0x01, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x06, 0x00,
- 0x01, 0x00, 0x00, 0x00, 0x47, 0x4c, 0x53, 0x4c, 0x2e, 0x73, 0x74, 0x64, 0x2e, 0x34, 0x35, 0x30,
- 0x00, 0x00, 0x00, 0x00, 0x0e, 0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
- 0x0f, 0x00, 0x07, 0x00, 0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x6d, 0x61, 0x69, 0x6e,
- 0x00, 0x00, 0x00, 0x00, 0x09, 0x00, 0x00, 0x00, 0x11, 0x00, 0x00, 0x00, 0x10, 0x00, 0x03, 0x00,
- 0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x09, 0x00, 0x00, 0x00,
- 0x1e, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x0d, 0x00, 0x00, 0x00,
- 0x22, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x0d, 0x00, 0x00, 0x00,
- 0x21, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x11, 0x00, 0x00, 0x00,
- 0x1e, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x13, 0x00, 0x02, 0x00, 0x02, 0x00, 0x00, 0x00,
- 0x21, 0x00, 0x03, 0x00, 0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x16, 0x00, 0x03, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x20, 0x00, 0x00, 0x00, 0x17, 0x00, 0x04, 0x00, 0x07, 0x00, 0x00, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x08, 0x00, 0x00, 0x00,
- 0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x08, 0x00, 0x00, 0x00,
- 0x09, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x19, 0x00, 0x09, 0x00, 0x0a, 0x00, 0x00, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x1b, 0x00, 0x03, 0x00,
- 0x0b, 0x00, 0x00, 0x00, 0x0a, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x0c, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x0c, 0x00, 0x00, 0x00,
- 0x0d, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x17, 0x00, 0x04, 0x00, 0x0f, 0x00, 0x00, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x10, 0x00, 0x00, 0x00,
- 0x01, 0x00, 0x00, 0x00, 0x0f, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x10, 0x00, 0x00, 0x00,
- 0x11, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x36, 0x00, 0x05, 0x00, 0x02, 0x00, 0x00, 0x00,
- 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00,
- 0x05, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x0e, 0x00, 0x00, 0x00,
- 0x0d, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00, 0x0f, 0x00, 0x00, 0x00, 0x12, 0x00, 0x00, 0x00,
- 0x11, 0x00, 0x00, 0x00, 0x57, 0x00, 0x05, 0x00, 0x07, 0x00, 0x00, 0x00, 0x13, 0x00, 0x00, 0x00,
- 0x0e, 0x00, 0x00, 0x00, 0x12, 0x00, 0x00, 0x00, 0x3e, 0x00, 0x03, 0x00, 0x09, 0x00, 0x00, 0x00,
- 0x13, 0x00, 0x00, 0x00, 0xfd, 0x00, 0x01, 0x00, 0x38, 0x00, 0x01, 0x00};
-
struct ScreenRectVertex {
ScreenRectVertex() = default;
explicit ScreenRectVertex(f32 x, f32 y, f32 u, f32 v) : position{{x, y}}, tex_coord{{u, v}} {}
@@ -173,9 +77,9 @@ constexpr std::array<f32, 4 * 4> MakeOrthographicMatrix(f32 width, f32 height) {
// clang-format on
}
-std::size_t GetBytesPerPixel(const Tegra::FramebufferConfig& framebuffer) {
+u32 GetBytesPerPixel(const Tegra::FramebufferConfig& framebuffer) {
using namespace VideoCore::Surface;
- return GetBytesPerPixel(PixelFormatFromGPUPixelFormat(framebuffer.pixel_format));
+ return BytesPerBlock(PixelFormatFromGPUPixelFormat(framebuffer.pixel_format));
}
std::size_t GetSizeInBytes(const Tegra::FramebufferConfig& framebuffer) {
@@ -239,34 +143,30 @@ VkSemaphore VKBlitScreen::Draw(const Tegra::FramebufferConfig& framebuffer, bool
scheduler.Wait(resource_ticks[image_index]);
resource_ticks[image_index] = scheduler.CurrentTick();
- VKImage* blit_image = use_accelerated ? screen_info.image : raw_images[image_index].get();
-
- UpdateDescriptorSet(image_index, blit_image->GetPresentView());
+ UpdateDescriptorSet(image_index,
+ use_accelerated ? screen_info.image_view : *raw_image_views[image_index]);
BufferData data;
SetUniformData(data, framebuffer);
SetVertexData(data, framebuffer);
auto map = buffer_commit->Map();
- std::memcpy(map.GetAddress(), &data, sizeof(data));
+ std::memcpy(map.Address(), &data, sizeof(data));
if (!use_accelerated) {
const u64 image_offset = GetRawImageOffset(framebuffer, image_index);
- const auto pixel_format =
- VideoCore::Surface::PixelFormatFromGPUPixelFormat(framebuffer.pixel_format);
const VAddr framebuffer_addr = framebuffer.address + framebuffer.offset;
- const auto host_ptr = cpu_memory.GetPointer(framebuffer_addr);
- rasterizer.FlushRegion(ToCacheAddr(host_ptr), GetSizeInBytes(framebuffer));
+ const u8* const host_ptr = cpu_memory.GetPointer(framebuffer_addr);
+ const size_t size_bytes = GetSizeInBytes(framebuffer);
+ rasterizer.FlushRegion(ToCacheAddr(host_ptr), size_bytes);
// TODO(Rodrigo): Read this from HLE
constexpr u32 block_height_log2 = 4;
- VideoCore::MortonSwizzle(VideoCore::MortonSwizzleMode::MortonToLinear, pixel_format,
- framebuffer.stride, block_height_log2, framebuffer.height, 0, 1, 1,
- map.GetAddress() + image_offset, host_ptr);
-
- blit_image->Transition(0, 1, 0, 1, VK_PIPELINE_STAGE_TRANSFER_BIT,
- VK_ACCESS_TRANSFER_WRITE_BIT, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL);
+ const u32 bytes_per_pixel = GetBytesPerPixel(framebuffer);
+ Tegra::Texture::UnswizzleTexture(
+ std::span(map.Address() + image_offset, size_bytes), std::span(host_ptr, size_bytes),
+ bytes_per_pixel, framebuffer.width, framebuffer.height, 1, block_height_log2, 0);
const VkBufferImageCopy copy{
.bufferOffset = image_offset,
@@ -288,15 +188,44 @@ VkSemaphore VKBlitScreen::Draw(const Tegra::FramebufferConfig& framebuffer, bool
},
};
scheduler.Record(
- [buffer = *buffer, image = *blit_image->GetHandle(), copy](vk::CommandBuffer cmdbuf) {
- cmdbuf.CopyBufferToImage(buffer, image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, copy);
+ [buffer = *buffer, image = *raw_images[image_index], copy](vk::CommandBuffer cmdbuf) {
+ const VkImageMemoryBarrier base_barrier{
+ .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
+ .pNext = nullptr,
+ .srcAccessMask = 0,
+ .dstAccessMask = 0,
+ .oldLayout = VK_IMAGE_LAYOUT_GENERAL,
+ .newLayout = VK_IMAGE_LAYOUT_GENERAL,
+ .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+ .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+ .image = image,
+ .subresourceRange =
+ {
+ .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
+ .baseMipLevel = 0,
+ .levelCount = 1,
+ .baseArrayLayer = 0,
+ .layerCount = 1,
+ },
+ };
+ VkImageMemoryBarrier read_barrier = base_barrier;
+ read_barrier.srcAccessMask = VK_ACCESS_HOST_WRITE_BIT;
+ read_barrier.dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
+ read_barrier.oldLayout = VK_IMAGE_LAYOUT_UNDEFINED;
+
+ VkImageMemoryBarrier write_barrier = base_barrier;
+ write_barrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
+ write_barrier.dstAccessMask = VK_ACCESS_SHADER_READ_BIT;
+
+ cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_HOST_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT,
+ 0, read_barrier);
+ cmdbuf.CopyBufferToImage(buffer, image, VK_IMAGE_LAYOUT_GENERAL, copy);
+ cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT,
+ VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT, 0, write_barrier);
});
}
map.Release();
- blit_image->Transition(0, 1, 0, 1, VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT,
- VK_ACCESS_SHADER_READ_BIT, VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL);
-
scheduler.Record([renderpass = *renderpass, framebuffer = *framebuffers[image_index],
descriptor_set = descriptor_sets[image_index], buffer = *buffer,
size = swapchain.GetSize(), pipeline = *pipeline,
@@ -304,31 +233,31 @@ VkSemaphore VKBlitScreen::Draw(const Tegra::FramebufferConfig& framebuffer, bool
const VkClearValue clear_color{
.color = {.float32 = {0.0f, 0.0f, 0.0f, 0.0f}},
};
-
- VkRenderPassBeginInfo renderpass_bi;
- renderpass_bi.sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO;
- renderpass_bi.pNext = nullptr;
- renderpass_bi.renderPass = renderpass;
- renderpass_bi.framebuffer = framebuffer;
- renderpass_bi.renderArea.offset.x = 0;
- renderpass_bi.renderArea.offset.y = 0;
- renderpass_bi.renderArea.extent = size;
- renderpass_bi.clearValueCount = 1;
- renderpass_bi.pClearValues = &clear_color;
-
- VkViewport viewport;
- viewport.x = 0.0f;
- viewport.y = 0.0f;
- viewport.width = static_cast<float>(size.width);
- viewport.height = static_cast<float>(size.height);
- viewport.minDepth = 0.0f;
- viewport.maxDepth = 1.0f;
-
- VkRect2D scissor;
- scissor.offset.x = 0;
- scissor.offset.y = 0;
- scissor.extent = size;
-
+ const VkRenderPassBeginInfo renderpass_bi{
+ .sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO,
+ .pNext = nullptr,
+ .renderPass = renderpass,
+ .framebuffer = framebuffer,
+ .renderArea =
+ {
+ .offset = {0, 0},
+ .extent = size,
+ },
+ .clearValueCount = 1,
+ .pClearValues = &clear_color,
+ };
+ const VkViewport viewport{
+ .x = 0.0f,
+ .y = 0.0f,
+ .width = static_cast<float>(size.width),
+ .height = static_cast<float>(size.height),
+ .minDepth = 0.0f,
+ .maxDepth = 1.0f,
+ };
+ const VkRect2D scissor{
+ .offset = {0, 0},
+ .extent = size,
+ };
cmdbuf.BeginRenderPass(renderpass_bi, VK_SUBPASS_CONTENTS_INLINE);
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
cmdbuf.SetViewport(0, viewport);
@@ -372,8 +301,8 @@ void VKBlitScreen::RefreshResources(const Tegra::FramebufferConfig& framebuffer)
}
void VKBlitScreen::CreateShaders() {
- vertex_shader = BuildShader(device, sizeof(blit_vertex_code), blit_vertex_code);
- fragment_shader = BuildShader(device, sizeof(blit_fragment_code), blit_fragment_code);
+ vertex_shader = BuildShader(device, VULKAN_PRESENT_VERT_SPV);
+ fragment_shader = BuildShader(device, VULKAN_PRESENT_FRAG_SPV);
}
void VKBlitScreen::CreateSemaphores() {
@@ -420,7 +349,7 @@ void VKBlitScreen::CreateRenderPass() {
const VkAttachmentReference color_attachment_ref{
.attachment = 0,
- .layout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL,
+ .layout = VK_IMAGE_LAYOUT_GENERAL,
};
const VkSubpassDescription subpass_description{
@@ -735,34 +664,56 @@ void VKBlitScreen::CreateStagingBuffer(const Tegra::FramebufferConfig& framebuff
void VKBlitScreen::CreateRawImages(const Tegra::FramebufferConfig& framebuffer) {
raw_images.resize(image_count);
+ raw_image_views.resize(image_count);
raw_buffer_commits.resize(image_count);
- const VkImageCreateInfo ci{
- .sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO,
- .pNext = nullptr,
- .flags = 0,
- .imageType = VK_IMAGE_TYPE_2D,
- .format = GetFormat(framebuffer),
- .extent =
- {
- .width = framebuffer.width,
- .height = framebuffer.height,
- .depth = 1,
- },
- .mipLevels = 1,
- .arrayLayers = 1,
- .samples = VK_SAMPLE_COUNT_1_BIT,
- .tiling = VK_IMAGE_TILING_LINEAR,
- .usage = VK_IMAGE_USAGE_TRANSFER_DST_BIT | VK_IMAGE_USAGE_SAMPLED_BIT,
- .sharingMode = VK_SHARING_MODE_EXCLUSIVE,
- .queueFamilyIndexCount = 0,
- .pQueueFamilyIndices = nullptr,
- .initialLayout = VK_IMAGE_LAYOUT_UNDEFINED,
- };
-
- for (std::size_t i = 0; i < image_count; ++i) {
- raw_images[i] = std::make_unique<VKImage>(device, scheduler, ci, VK_IMAGE_ASPECT_COLOR_BIT);
- raw_buffer_commits[i] = memory_manager.Commit(raw_images[i]->GetHandle(), false);
+ for (size_t i = 0; i < image_count; ++i) {
+ raw_images[i] = device.GetLogical().CreateImage(VkImageCreateInfo{
+ .sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO,
+ .pNext = nullptr,
+ .flags = 0,
+ .imageType = VK_IMAGE_TYPE_2D,
+ .format = GetFormat(framebuffer),
+ .extent =
+ {
+ .width = framebuffer.width,
+ .height = framebuffer.height,
+ .depth = 1,
+ },
+ .mipLevels = 1,
+ .arrayLayers = 1,
+ .samples = VK_SAMPLE_COUNT_1_BIT,
+ .tiling = VK_IMAGE_TILING_LINEAR,
+ .usage = VK_IMAGE_USAGE_TRANSFER_DST_BIT | VK_IMAGE_USAGE_SAMPLED_BIT,
+ .sharingMode = VK_SHARING_MODE_EXCLUSIVE,
+ .queueFamilyIndexCount = 0,
+ .pQueueFamilyIndices = nullptr,
+ .initialLayout = VK_IMAGE_LAYOUT_UNDEFINED,
+ });
+ raw_buffer_commits[i] = memory_manager.Commit(raw_images[i], false);
+ raw_image_views[i] = device.GetLogical().CreateImageView(VkImageViewCreateInfo{
+ .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
+ .pNext = nullptr,
+ .flags = 0,
+ .image = *raw_images[i],
+ .viewType = VK_IMAGE_VIEW_TYPE_2D,
+ .format = GetFormat(framebuffer),
+ .components =
+ {
+ .r = VK_COMPONENT_SWIZZLE_IDENTITY,
+ .g = VK_COMPONENT_SWIZZLE_IDENTITY,
+ .b = VK_COMPONENT_SWIZZLE_IDENTITY,
+ .a = VK_COMPONENT_SWIZZLE_IDENTITY,
+ },
+ .subresourceRange =
+ {
+ .aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
+ .baseMipLevel = 0,
+ .levelCount = 1,
+ .baseArrayLayer = 0,
+ .layerCount = 1,
+ },
+ });
}
}
@@ -789,7 +740,7 @@ void VKBlitScreen::UpdateDescriptorSet(std::size_t image_index, VkImageView imag
const VkDescriptorImageInfo image_info{
.sampler = *sampler,
.imageView = image_view,
- .imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL,
+ .imageLayout = VK_IMAGE_LAYOUT_GENERAL,
};
const VkWriteDescriptorSet sampler_write{
diff --git a/src/video_core/renderer_vulkan/vk_blit_screen.h b/src/video_core/renderer_vulkan/vk_blit_screen.h
index 8f2839214..2ee374247 100644
--- a/src/video_core/renderer_vulkan/vk_blit_screen.h
+++ b/src/video_core/renderer_vulkan/vk_blit_screen.h
@@ -35,7 +35,6 @@ struct ScreenInfo;
class RasterizerVulkan;
class VKDevice;
-class VKImage;
class VKScheduler;
class VKSwapchain;
@@ -110,7 +109,8 @@ private:
std::vector<u64> resource_ticks;
std::vector<vk::Semaphore> semaphores;
- std::vector<std::unique_ptr<VKImage>> raw_images;
+ std::vector<vk::Image> raw_images;
+ std::vector<vk::ImageView> raw_image_views;
std::vector<VKMemoryCommit> raw_buffer_commits;
u32 raw_width = 0;
u32 raw_height = 0;
diff --git a/src/video_core/renderer_vulkan/vk_buffer_cache.cpp b/src/video_core/renderer_vulkan/vk_buffer_cache.cpp
index 444d3fb93..10d296c2f 100644
--- a/src/video_core/renderer_vulkan/vk_buffer_cache.cpp
+++ b/src/video_core/renderer_vulkan/vk_buffer_cache.cpp
@@ -31,15 +31,19 @@ constexpr VkAccessFlags UPLOAD_ACCESS_BARRIERS =
VK_ACCESS_TRANSFER_READ_BIT | VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_UNIFORM_READ_BIT |
VK_ACCESS_VERTEX_ATTRIBUTE_READ_BIT | VK_ACCESS_INDEX_READ_BIT;
+constexpr VkAccessFlags TRANSFORM_FEEDBACK_WRITE_ACCESS =
+ VK_ACCESS_TRANSFORM_FEEDBACK_WRITE_BIT_EXT | VK_ACCESS_TRANSFORM_FEEDBACK_COUNTER_WRITE_BIT_EXT;
+
std::unique_ptr<VKStreamBuffer> CreateStreamBuffer(const VKDevice& device, VKScheduler& scheduler) {
- return std::make_unique<VKStreamBuffer>(device, scheduler, BUFFER_USAGE);
+ return std::make_unique<VKStreamBuffer>(device, scheduler);
}
} // Anonymous namespace
-Buffer::Buffer(const VKDevice& device, VKMemoryManager& memory_manager, VKScheduler& scheduler_,
+Buffer::Buffer(const VKDevice& device_, VKMemoryManager& memory_manager, VKScheduler& scheduler_,
VKStagingBufferPool& staging_pool_, VAddr cpu_addr_, std::size_t size_)
- : BufferBlock{cpu_addr_, size_}, scheduler{scheduler_}, staging_pool{staging_pool_} {
+ : BufferBlock{cpu_addr_, size_}, device{device_}, scheduler{scheduler_}, staging_pool{
+ staging_pool_} {
const VkBufferCreateInfo ci{
.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
.pNext = nullptr,
@@ -64,24 +68,39 @@ void Buffer::Upload(std::size_t offset, std::size_t data_size, const u8* data) {
scheduler.RequestOutsideRenderPassOperationContext();
const VkBuffer handle = Handle();
- scheduler.Record(
- [staging = *staging.handle, handle, offset, data_size](vk::CommandBuffer cmdbuf) {
- cmdbuf.CopyBuffer(staging, handle, VkBufferCopy{0, offset, data_size});
-
- const VkBufferMemoryBarrier barrier{
- .sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER,
- .pNext = nullptr,
- .srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
- .dstAccessMask = UPLOAD_ACCESS_BARRIERS,
- .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
- .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
- .buffer = handle,
- .offset = offset,
- .size = data_size,
- };
- cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, UPLOAD_PIPELINE_STAGE, 0, {},
- barrier, {});
- });
+ scheduler.Record([staging = *staging.handle, handle, offset, data_size,
+ &device = device](vk::CommandBuffer cmdbuf) {
+ const VkBufferMemoryBarrier read_barrier{
+ .sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER,
+ .pNext = nullptr,
+ .srcAccessMask =
+ VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_TRANSFER_WRITE_BIT |
+ VK_ACCESS_HOST_WRITE_BIT |
+ (device.IsExtTransformFeedbackSupported() ? TRANSFORM_FEEDBACK_WRITE_ACCESS : 0),
+ .dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT,
+ .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+ .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+ .buffer = handle,
+ .offset = offset,
+ .size = data_size,
+ };
+ const VkBufferMemoryBarrier write_barrier{
+ .sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER,
+ .pNext = nullptr,
+ .srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
+ .dstAccessMask = UPLOAD_ACCESS_BARRIERS,
+ .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+ .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+ .buffer = handle,
+ .offset = offset,
+ .size = data_size,
+ };
+ cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT,
+ 0, read_barrier);
+ cmdbuf.CopyBuffer(staging, handle, VkBufferCopy{0, offset, data_size});
+ cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, UPLOAD_PIPELINE_STAGE, 0,
+ write_barrier);
+ });
}
void Buffer::Download(std::size_t offset, std::size_t data_size, u8* data) {
@@ -150,8 +169,10 @@ void Buffer::CopyFrom(const Buffer& src, std::size_t src_offset, std::size_t dst
VKBufferCache::VKBufferCache(VideoCore::RasterizerInterface& rasterizer_,
Tegra::MemoryManager& gpu_memory_, Core::Memory::Memory& cpu_memory_,
const VKDevice& device_, VKMemoryManager& memory_manager_,
- VKScheduler& scheduler_, VKStagingBufferPool& staging_pool_)
- : BufferCache{rasterizer_, gpu_memory_, cpu_memory_, CreateStreamBuffer(device_, scheduler_)},
+ VKScheduler& scheduler_, VKStreamBuffer& stream_buffer_,
+ VKStagingBufferPool& staging_pool_)
+ : VideoCommon::BufferCache<Buffer, VkBuffer, VKStreamBuffer>{rasterizer_, gpu_memory_,
+ cpu_memory_, stream_buffer_},
device{device_}, memory_manager{memory_manager_}, scheduler{scheduler_}, staging_pool{
staging_pool_} {}
diff --git a/src/video_core/renderer_vulkan/vk_buffer_cache.h b/src/video_core/renderer_vulkan/vk_buffer_cache.h
index 6008b8373..daf498222 100644
--- a/src/video_core/renderer_vulkan/vk_buffer_cache.h
+++ b/src/video_core/renderer_vulkan/vk_buffer_cache.h
@@ -41,6 +41,7 @@ public:
}
private:
+ const VKDevice& device;
VKScheduler& scheduler;
VKStagingBufferPool& staging_pool;
@@ -49,10 +50,11 @@ private:
class VKBufferCache final : public VideoCommon::BufferCache<Buffer, VkBuffer, VKStreamBuffer> {
public:
- explicit VKBufferCache(VideoCore::RasterizerInterface& rasterizer_,
- Tegra::MemoryManager& gpu_memory_, Core::Memory::Memory& cpu_memory_,
- const VKDevice& device_, VKMemoryManager& memory_manager_,
- VKScheduler& scheduler_, VKStagingBufferPool& staging_pool_);
+ explicit VKBufferCache(VideoCore::RasterizerInterface& rasterizer,
+ Tegra::MemoryManager& gpu_memory, Core::Memory::Memory& cpu_memory,
+ const VKDevice& device, VKMemoryManager& memory_manager,
+ VKScheduler& scheduler, VKStreamBuffer& stream_buffer,
+ VKStagingBufferPool& staging_pool);
~VKBufferCache();
BufferInfo GetEmptyBuffer(std::size_t size) override;
diff --git a/src/video_core/renderer_vulkan/vk_compute_pass.cpp b/src/video_core/renderer_vulkan/vk_compute_pass.cpp
index 1ac7e2a30..2c030e910 100644
--- a/src/video_core/renderer_vulkan/vk_compute_pass.cpp
+++ b/src/video_core/renderer_vulkan/vk_compute_pass.cpp
@@ -10,6 +10,9 @@
#include "common/alignment.h"
#include "common/assert.h"
#include "common/common_types.h"
+#include "video_core/host_shaders/vulkan_quad_array_comp_spv.h"
+#include "video_core/host_shaders/vulkan_quad_indexed_comp_spv.h"
+#include "video_core/host_shaders/vulkan_uint8_comp_spv.h"
#include "video_core/renderer_vulkan/vk_compute_pass.h"
#include "video_core/renderer_vulkan/vk_descriptor_pool.h"
#include "video_core/renderer_vulkan/vk_device.h"
@@ -22,99 +25,6 @@ namespace Vulkan {
namespace {
-// Quad array SPIR-V module. Generated from the "shaders/" directory, read the instructions there.
-constexpr u8 quad_array[] = {
- 0x03, 0x02, 0x23, 0x07, 0x00, 0x00, 0x01, 0x00, 0x07, 0x00, 0x08, 0x00, 0x54, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x11, 0x00, 0x02, 0x00, 0x01, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x06, 0x00,
- 0x01, 0x00, 0x00, 0x00, 0x47, 0x4c, 0x53, 0x4c, 0x2e, 0x73, 0x74, 0x64, 0x2e, 0x34, 0x35, 0x30,
- 0x00, 0x00, 0x00, 0x00, 0x0e, 0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
- 0x0f, 0x00, 0x06, 0x00, 0x05, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x6d, 0x61, 0x69, 0x6e,
- 0x00, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x10, 0x00, 0x06, 0x00, 0x04, 0x00, 0x00, 0x00,
- 0x11, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
- 0x47, 0x00, 0x04, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x1c, 0x00, 0x00, 0x00,
- 0x47, 0x00, 0x04, 0x00, 0x13, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00,
- 0x48, 0x00, 0x05, 0x00, 0x14, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x23, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x03, 0x00, 0x14, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00,
- 0x47, 0x00, 0x04, 0x00, 0x16, 0x00, 0x00, 0x00, 0x22, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
- 0x47, 0x00, 0x04, 0x00, 0x16, 0x00, 0x00, 0x00, 0x21, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
- 0x48, 0x00, 0x05, 0x00, 0x29, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x23, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x03, 0x00, 0x29, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00,
- 0x47, 0x00, 0x04, 0x00, 0x4a, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x19, 0x00, 0x00, 0x00,
- 0x13, 0x00, 0x02, 0x00, 0x02, 0x00, 0x00, 0x00, 0x21, 0x00, 0x03, 0x00, 0x03, 0x00, 0x00, 0x00,
- 0x02, 0x00, 0x00, 0x00, 0x15, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x20, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x07, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x17, 0x00, 0x04, 0x00, 0x09, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00,
- 0x03, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x0a, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
- 0x09, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x0a, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00,
- 0x01, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x0c, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x0d, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x11, 0x00, 0x00, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x1d, 0x00, 0x03, 0x00, 0x13, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00,
- 0x1e, 0x00, 0x03, 0x00, 0x14, 0x00, 0x00, 0x00, 0x13, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00,
- 0x15, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x14, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00,
- 0x15, 0x00, 0x00, 0x00, 0x16, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x15, 0x00, 0x04, 0x00,
- 0x18, 0x00, 0x00, 0x00, 0x20, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x14, 0x00, 0x02, 0x00,
- 0x1b, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x03, 0x00, 0x29, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00,
- 0x20, 0x00, 0x04, 0x00, 0x2a, 0x00, 0x00, 0x00, 0x09, 0x00, 0x00, 0x00, 0x29, 0x00, 0x00, 0x00,
- 0x3b, 0x00, 0x04, 0x00, 0x2a, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x00, 0x00, 0x09, 0x00, 0x00, 0x00,
- 0x2b, 0x00, 0x04, 0x00, 0x18, 0x00, 0x00, 0x00, 0x2c, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
- 0x20, 0x00, 0x04, 0x00, 0x2d, 0x00, 0x00, 0x00, 0x09, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00,
- 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x31, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00,
- 0x1c, 0x00, 0x04, 0x00, 0x34, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x11, 0x00, 0x00, 0x00,
- 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x35, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
- 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x36, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00,
- 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x37, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00,
- 0x2c, 0x00, 0x09, 0x00, 0x34, 0x00, 0x00, 0x00, 0x38, 0x00, 0x00, 0x00, 0x0c, 0x00, 0x00, 0x00,
- 0x35, 0x00, 0x00, 0x00, 0x36, 0x00, 0x00, 0x00, 0x0c, 0x00, 0x00, 0x00, 0x36, 0x00, 0x00, 0x00,
- 0x37, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x3a, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00,
- 0x34, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x44, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x18, 0x00, 0x00, 0x00, 0x47, 0x00, 0x00, 0x00,
- 0x01, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x49, 0x00, 0x00, 0x00,
- 0x00, 0x04, 0x00, 0x00, 0x2c, 0x00, 0x06, 0x00, 0x09, 0x00, 0x00, 0x00, 0x4a, 0x00, 0x00, 0x00,
- 0x49, 0x00, 0x00, 0x00, 0x35, 0x00, 0x00, 0x00, 0x35, 0x00, 0x00, 0x00, 0x36, 0x00, 0x05, 0x00,
- 0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00,
- 0xf8, 0x00, 0x02, 0x00, 0x05, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x3a, 0x00, 0x00, 0x00,
- 0x3b, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0xf9, 0x00, 0x02, 0x00, 0x4c, 0x00, 0x00, 0x00,
- 0xf8, 0x00, 0x02, 0x00, 0x4c, 0x00, 0x00, 0x00, 0xf6, 0x00, 0x04, 0x00, 0x4b, 0x00, 0x00, 0x00,
- 0x4e, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xf9, 0x00, 0x02, 0x00, 0x4d, 0x00, 0x00, 0x00,
- 0xf8, 0x00, 0x02, 0x00, 0x4d, 0x00, 0x00, 0x00, 0x41, 0x00, 0x05, 0x00, 0x0d, 0x00, 0x00, 0x00,
- 0x0e, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x0c, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x0f, 0x00, 0x00, 0x00, 0x0e, 0x00, 0x00, 0x00, 0x84, 0x00, 0x05, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x12, 0x00, 0x00, 0x00, 0x0f, 0x00, 0x00, 0x00, 0x11, 0x00, 0x00, 0x00,
- 0x44, 0x00, 0x05, 0x00, 0x06, 0x00, 0x00, 0x00, 0x17, 0x00, 0x00, 0x00, 0x16, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x7c, 0x00, 0x04, 0x00, 0x18, 0x00, 0x00, 0x00, 0x19, 0x00, 0x00, 0x00,
- 0x17, 0x00, 0x00, 0x00, 0x7c, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x1a, 0x00, 0x00, 0x00,
- 0x19, 0x00, 0x00, 0x00, 0xae, 0x00, 0x05, 0x00, 0x1b, 0x00, 0x00, 0x00, 0x1c, 0x00, 0x00, 0x00,
- 0x12, 0x00, 0x00, 0x00, 0x1a, 0x00, 0x00, 0x00, 0xf7, 0x00, 0x03, 0x00, 0x1e, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0xfa, 0x00, 0x04, 0x00, 0x1c, 0x00, 0x00, 0x00, 0x1d, 0x00, 0x00, 0x00,
- 0x1e, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x1d, 0x00, 0x00, 0x00, 0xf9, 0x00, 0x02, 0x00,
- 0x4b, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x1e, 0x00, 0x00, 0x00, 0xf9, 0x00, 0x02, 0x00,
- 0x21, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x21, 0x00, 0x00, 0x00, 0xf5, 0x00, 0x07, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x53, 0x00, 0x00, 0x00, 0x0c, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x00, 0x00,
- 0x48, 0x00, 0x00, 0x00, 0x22, 0x00, 0x00, 0x00, 0xb0, 0x00, 0x05, 0x00, 0x1b, 0x00, 0x00, 0x00,
- 0x27, 0x00, 0x00, 0x00, 0x53, 0x00, 0x00, 0x00, 0x11, 0x00, 0x00, 0x00, 0xf6, 0x00, 0x04, 0x00,
- 0x23, 0x00, 0x00, 0x00, 0x22, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xfa, 0x00, 0x04, 0x00,
- 0x27, 0x00, 0x00, 0x00, 0x22, 0x00, 0x00, 0x00, 0x23, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00,
- 0x22, 0x00, 0x00, 0x00, 0x41, 0x00, 0x05, 0x00, 0x2d, 0x00, 0x00, 0x00, 0x2e, 0x00, 0x00, 0x00,
- 0x2b, 0x00, 0x00, 0x00, 0x2c, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00,
- 0x2f, 0x00, 0x00, 0x00, 0x2e, 0x00, 0x00, 0x00, 0x84, 0x00, 0x05, 0x00, 0x06, 0x00, 0x00, 0x00,
- 0x32, 0x00, 0x00, 0x00, 0x0f, 0x00, 0x00, 0x00, 0x31, 0x00, 0x00, 0x00, 0x80, 0x00, 0x05, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x33, 0x00, 0x00, 0x00, 0x2f, 0x00, 0x00, 0x00, 0x32, 0x00, 0x00, 0x00,
- 0x3e, 0x00, 0x03, 0x00, 0x3b, 0x00, 0x00, 0x00, 0x38, 0x00, 0x00, 0x00, 0x41, 0x00, 0x05, 0x00,
- 0x07, 0x00, 0x00, 0x00, 0x3c, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x00, 0x00, 0x53, 0x00, 0x00, 0x00,
- 0x3d, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x00, 0x00, 0x3c, 0x00, 0x00, 0x00,
- 0x80, 0x00, 0x05, 0x00, 0x06, 0x00, 0x00, 0x00, 0x3e, 0x00, 0x00, 0x00, 0x33, 0x00, 0x00, 0x00,
- 0x3d, 0x00, 0x00, 0x00, 0x80, 0x00, 0x05, 0x00, 0x06, 0x00, 0x00, 0x00, 0x42, 0x00, 0x00, 0x00,
- 0x12, 0x00, 0x00, 0x00, 0x53, 0x00, 0x00, 0x00, 0x41, 0x00, 0x06, 0x00, 0x44, 0x00, 0x00, 0x00,
- 0x45, 0x00, 0x00, 0x00, 0x16, 0x00, 0x00, 0x00, 0x2c, 0x00, 0x00, 0x00, 0x42, 0x00, 0x00, 0x00,
- 0x3e, 0x00, 0x03, 0x00, 0x45, 0x00, 0x00, 0x00, 0x3e, 0x00, 0x00, 0x00, 0x80, 0x00, 0x05, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x48, 0x00, 0x00, 0x00, 0x53, 0x00, 0x00, 0x00, 0x47, 0x00, 0x00, 0x00,
- 0xf9, 0x00, 0x02, 0x00, 0x21, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x23, 0x00, 0x00, 0x00,
- 0xf9, 0x00, 0x02, 0x00, 0x4b, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x4e, 0x00, 0x00, 0x00,
- 0xf9, 0x00, 0x02, 0x00, 0x4c, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x4b, 0x00, 0x00, 0x00,
- 0xfd, 0x00, 0x01, 0x00, 0x38, 0x00, 0x01, 0x00,
-};
-
VkDescriptorSetLayoutBinding BuildQuadArrayPassDescriptorSetLayoutBinding() {
return {
.binding = 0,
@@ -144,208 +54,6 @@ VkPushConstantRange BuildComputePushConstantRange(std::size_t size) {
};
}
-// Uint8 SPIR-V module. Generated from the "shaders/" directory.
-constexpr u8 uint8_pass[] = {
- 0x03, 0x02, 0x23, 0x07, 0x00, 0x00, 0x01, 0x00, 0x07, 0x00, 0x08, 0x00, 0x2f, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x11, 0x00, 0x02, 0x00, 0x01, 0x00, 0x00, 0x00, 0x11, 0x00, 0x02, 0x00,
- 0x51, 0x11, 0x00, 0x00, 0x11, 0x00, 0x02, 0x00, 0x61, 0x11, 0x00, 0x00, 0x0a, 0x00, 0x07, 0x00,
- 0x53, 0x50, 0x56, 0x5f, 0x4b, 0x48, 0x52, 0x5f, 0x31, 0x36, 0x62, 0x69, 0x74, 0x5f, 0x73, 0x74,
- 0x6f, 0x72, 0x61, 0x67, 0x65, 0x00, 0x00, 0x00, 0x0a, 0x00, 0x07, 0x00, 0x53, 0x50, 0x56, 0x5f,
- 0x4b, 0x48, 0x52, 0x5f, 0x38, 0x62, 0x69, 0x74, 0x5f, 0x73, 0x74, 0x6f, 0x72, 0x61, 0x67, 0x65,
- 0x00, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x06, 0x00, 0x01, 0x00, 0x00, 0x00, 0x47, 0x4c, 0x53, 0x4c,
- 0x2e, 0x73, 0x74, 0x64, 0x2e, 0x34, 0x35, 0x30, 0x00, 0x00, 0x00, 0x00, 0x0e, 0x00, 0x03, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x0f, 0x00, 0x06, 0x00, 0x05, 0x00, 0x00, 0x00,
- 0x04, 0x00, 0x00, 0x00, 0x6d, 0x61, 0x69, 0x6e, 0x00, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00,
- 0x10, 0x00, 0x06, 0x00, 0x04, 0x00, 0x00, 0x00, 0x11, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00,
- 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x0b, 0x00, 0x00, 0x00,
- 0x0b, 0x00, 0x00, 0x00, 0x1c, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x12, 0x00, 0x00, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x48, 0x00, 0x04, 0x00, 0x13, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x18, 0x00, 0x00, 0x00, 0x48, 0x00, 0x05, 0x00, 0x13, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x23, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x03, 0x00,
- 0x13, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x15, 0x00, 0x00, 0x00,
- 0x22, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x15, 0x00, 0x00, 0x00,
- 0x21, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x1f, 0x00, 0x00, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x48, 0x00, 0x04, 0x00, 0x20, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x19, 0x00, 0x00, 0x00, 0x48, 0x00, 0x05, 0x00, 0x20, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x23, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x03, 0x00,
- 0x20, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x22, 0x00, 0x00, 0x00,
- 0x22, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x22, 0x00, 0x00, 0x00,
- 0x21, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x2e, 0x00, 0x00, 0x00,
- 0x0b, 0x00, 0x00, 0x00, 0x19, 0x00, 0x00, 0x00, 0x13, 0x00, 0x02, 0x00, 0x02, 0x00, 0x00, 0x00,
- 0x21, 0x00, 0x03, 0x00, 0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x15, 0x00, 0x04, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x20, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00,
- 0x07, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x17, 0x00, 0x04, 0x00,
- 0x09, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00,
- 0x0a, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x09, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00,
- 0x0a, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x0c, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00,
- 0x0d, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x15, 0x00, 0x04, 0x00,
- 0x11, 0x00, 0x00, 0x00, 0x08, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x1d, 0x00, 0x03, 0x00,
- 0x12, 0x00, 0x00, 0x00, 0x11, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x03, 0x00, 0x13, 0x00, 0x00, 0x00,
- 0x12, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x14, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00,
- 0x13, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x14, 0x00, 0x00, 0x00, 0x15, 0x00, 0x00, 0x00,
- 0x02, 0x00, 0x00, 0x00, 0x15, 0x00, 0x04, 0x00, 0x17, 0x00, 0x00, 0x00, 0x20, 0x00, 0x00, 0x00,
- 0x01, 0x00, 0x00, 0x00, 0x14, 0x00, 0x02, 0x00, 0x1a, 0x00, 0x00, 0x00, 0x15, 0x00, 0x04, 0x00,
- 0x1e, 0x00, 0x00, 0x00, 0x10, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x1d, 0x00, 0x03, 0x00,
- 0x1f, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x03, 0x00, 0x20, 0x00, 0x00, 0x00,
- 0x1f, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x21, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00,
- 0x20, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x21, 0x00, 0x00, 0x00, 0x22, 0x00, 0x00, 0x00,
- 0x02, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x17, 0x00, 0x00, 0x00, 0x23, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x26, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00,
- 0x11, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x2a, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00,
- 0x1e, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x2c, 0x00, 0x00, 0x00,
- 0x00, 0x04, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x2d, 0x00, 0x00, 0x00,
- 0x01, 0x00, 0x00, 0x00, 0x2c, 0x00, 0x06, 0x00, 0x09, 0x00, 0x00, 0x00, 0x2e, 0x00, 0x00, 0x00,
- 0x2c, 0x00, 0x00, 0x00, 0x2d, 0x00, 0x00, 0x00, 0x2d, 0x00, 0x00, 0x00, 0x36, 0x00, 0x05, 0x00,
- 0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00,
- 0xf8, 0x00, 0x02, 0x00, 0x05, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x07, 0x00, 0x00, 0x00,
- 0x08, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x41, 0x00, 0x05, 0x00, 0x0d, 0x00, 0x00, 0x00,
- 0x0e, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x0c, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x0f, 0x00, 0x00, 0x00, 0x0e, 0x00, 0x00, 0x00, 0x3e, 0x00, 0x03, 0x00,
- 0x08, 0x00, 0x00, 0x00, 0x0f, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00,
- 0x10, 0x00, 0x00, 0x00, 0x08, 0x00, 0x00, 0x00, 0x44, 0x00, 0x05, 0x00, 0x06, 0x00, 0x00, 0x00,
- 0x16, 0x00, 0x00, 0x00, 0x15, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x7c, 0x00, 0x04, 0x00,
- 0x17, 0x00, 0x00, 0x00, 0x18, 0x00, 0x00, 0x00, 0x16, 0x00, 0x00, 0x00, 0x7c, 0x00, 0x04, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x19, 0x00, 0x00, 0x00, 0x18, 0x00, 0x00, 0x00, 0xb0, 0x00, 0x05, 0x00,
- 0x1a, 0x00, 0x00, 0x00, 0x1b, 0x00, 0x00, 0x00, 0x10, 0x00, 0x00, 0x00, 0x19, 0x00, 0x00, 0x00,
- 0xf7, 0x00, 0x03, 0x00, 0x1d, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xfa, 0x00, 0x04, 0x00,
- 0x1b, 0x00, 0x00, 0x00, 0x1c, 0x00, 0x00, 0x00, 0x1d, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00,
- 0x1c, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x24, 0x00, 0x00, 0x00,
- 0x08, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x25, 0x00, 0x00, 0x00,
- 0x08, 0x00, 0x00, 0x00, 0x41, 0x00, 0x06, 0x00, 0x26, 0x00, 0x00, 0x00, 0x27, 0x00, 0x00, 0x00,
- 0x15, 0x00, 0x00, 0x00, 0x23, 0x00, 0x00, 0x00, 0x25, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00,
- 0x11, 0x00, 0x00, 0x00, 0x28, 0x00, 0x00, 0x00, 0x27, 0x00, 0x00, 0x00, 0x71, 0x00, 0x04, 0x00,
- 0x1e, 0x00, 0x00, 0x00, 0x29, 0x00, 0x00, 0x00, 0x28, 0x00, 0x00, 0x00, 0x41, 0x00, 0x06, 0x00,
- 0x2a, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x00, 0x00, 0x22, 0x00, 0x00, 0x00, 0x23, 0x00, 0x00, 0x00,
- 0x24, 0x00, 0x00, 0x00, 0x3e, 0x00, 0x03, 0x00, 0x2b, 0x00, 0x00, 0x00, 0x29, 0x00, 0x00, 0x00,
- 0xf9, 0x00, 0x02, 0x00, 0x1d, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x1d, 0x00, 0x00, 0x00,
- 0xfd, 0x00, 0x01, 0x00, 0x38, 0x00, 0x01, 0x00,
-};
-
-// Quad indexed SPIR-V module. Generated from the "shaders/" directory.
-constexpr u8 QUAD_INDEXED_SPV[] = {
- 0x03, 0x02, 0x23, 0x07, 0x00, 0x00, 0x01, 0x00, 0x07, 0x00, 0x08, 0x00, 0x7c, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x11, 0x00, 0x02, 0x00, 0x01, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x06, 0x00,
- 0x01, 0x00, 0x00, 0x00, 0x47, 0x4c, 0x53, 0x4c, 0x2e, 0x73, 0x74, 0x64, 0x2e, 0x34, 0x35, 0x30,
- 0x00, 0x00, 0x00, 0x00, 0x0e, 0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
- 0x0f, 0x00, 0x06, 0x00, 0x05, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x6d, 0x61, 0x69, 0x6e,
- 0x00, 0x00, 0x00, 0x00, 0x0c, 0x00, 0x00, 0x00, 0x10, 0x00, 0x06, 0x00, 0x04, 0x00, 0x00, 0x00,
- 0x11, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
- 0x47, 0x00, 0x04, 0x00, 0x0c, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x1c, 0x00, 0x00, 0x00,
- 0x47, 0x00, 0x04, 0x00, 0x15, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00,
- 0x48, 0x00, 0x04, 0x00, 0x16, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x19, 0x00, 0x00, 0x00,
- 0x48, 0x00, 0x05, 0x00, 0x16, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x23, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x03, 0x00, 0x16, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00,
- 0x47, 0x00, 0x04, 0x00, 0x18, 0x00, 0x00, 0x00, 0x22, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
- 0x47, 0x00, 0x04, 0x00, 0x18, 0x00, 0x00, 0x00, 0x21, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
- 0x48, 0x00, 0x05, 0x00, 0x22, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x23, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x48, 0x00, 0x05, 0x00, 0x22, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
- 0x23, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x47, 0x00, 0x03, 0x00, 0x22, 0x00, 0x00, 0x00,
- 0x02, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x56, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00,
- 0x04, 0x00, 0x00, 0x00, 0x48, 0x00, 0x04, 0x00, 0x57, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
- 0x18, 0x00, 0x00, 0x00, 0x48, 0x00, 0x05, 0x00, 0x57, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
- 0x23, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x03, 0x00, 0x57, 0x00, 0x00, 0x00,
- 0x03, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x59, 0x00, 0x00, 0x00, 0x22, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x59, 0x00, 0x00, 0x00, 0x21, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x72, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00,
- 0x19, 0x00, 0x00, 0x00, 0x13, 0x00, 0x02, 0x00, 0x02, 0x00, 0x00, 0x00, 0x21, 0x00, 0x03, 0x00,
- 0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x15, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00,
- 0x20, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x07, 0x00, 0x00, 0x00,
- 0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x15, 0x00, 0x04, 0x00, 0x09, 0x00, 0x00, 0x00,
- 0x20, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x17, 0x00, 0x04, 0x00, 0x0a, 0x00, 0x00, 0x00,
- 0x09, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x0b, 0x00, 0x00, 0x00,
- 0x01, 0x00, 0x00, 0x00, 0x0a, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x0b, 0x00, 0x00, 0x00,
- 0x0c, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x09, 0x00, 0x00, 0x00,
- 0x0d, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x0e, 0x00, 0x00, 0x00,
- 0x01, 0x00, 0x00, 0x00, 0x09, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00,
- 0x13, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x1d, 0x00, 0x03, 0x00, 0x15, 0x00, 0x00, 0x00,
- 0x09, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x03, 0x00, 0x16, 0x00, 0x00, 0x00, 0x15, 0x00, 0x00, 0x00,
- 0x20, 0x00, 0x04, 0x00, 0x17, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x16, 0x00, 0x00, 0x00,
- 0x3b, 0x00, 0x04, 0x00, 0x17, 0x00, 0x00, 0x00, 0x18, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00,
- 0x14, 0x00, 0x02, 0x00, 0x1b, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00,
- 0x21, 0x00, 0x00, 0x00, 0x08, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x04, 0x00, 0x22, 0x00, 0x00, 0x00,
- 0x09, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x23, 0x00, 0x00, 0x00,
- 0x09, 0x00, 0x00, 0x00, 0x22, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x23, 0x00, 0x00, 0x00,
- 0x24, 0x00, 0x00, 0x00, 0x09, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00,
- 0x25, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x26, 0x00, 0x00, 0x00,
- 0x09, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00,
- 0x2b, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x09, 0x00, 0x00, 0x00,
- 0x3b, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00,
- 0x3f, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x1c, 0x00, 0x04, 0x00, 0x41, 0x00, 0x00, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00,
- 0x42, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00,
- 0x43, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x2c, 0x00, 0x09, 0x00, 0x41, 0x00, 0x00, 0x00,
- 0x44, 0x00, 0x00, 0x00, 0x42, 0x00, 0x00, 0x00, 0x25, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x00, 0x00,
- 0x42, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x00, 0x00, 0x43, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00,
- 0x46, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x41, 0x00, 0x00, 0x00, 0x1d, 0x00, 0x03, 0x00,
- 0x56, 0x00, 0x00, 0x00, 0x09, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x03, 0x00, 0x57, 0x00, 0x00, 0x00,
- 0x56, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x58, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00,
- 0x57, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x58, 0x00, 0x00, 0x00, 0x59, 0x00, 0x00, 0x00,
- 0x02, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x5b, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00,
- 0x09, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x69, 0x00, 0x00, 0x00, 0x09, 0x00, 0x00, 0x00,
- 0x09, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x09, 0x00, 0x00, 0x00, 0x70, 0x00, 0x00, 0x00,
- 0x00, 0x04, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x09, 0x00, 0x00, 0x00, 0x71, 0x00, 0x00, 0x00,
- 0x01, 0x00, 0x00, 0x00, 0x2c, 0x00, 0x06, 0x00, 0x0a, 0x00, 0x00, 0x00, 0x72, 0x00, 0x00, 0x00,
- 0x70, 0x00, 0x00, 0x00, 0x71, 0x00, 0x00, 0x00, 0x71, 0x00, 0x00, 0x00, 0x36, 0x00, 0x05, 0x00,
- 0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00,
- 0xf8, 0x00, 0x02, 0x00, 0x05, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x46, 0x00, 0x00, 0x00,
- 0x47, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0xf9, 0x00, 0x02, 0x00, 0x74, 0x00, 0x00, 0x00,
- 0xf8, 0x00, 0x02, 0x00, 0x74, 0x00, 0x00, 0x00, 0xf6, 0x00, 0x04, 0x00, 0x73, 0x00, 0x00, 0x00,
- 0x76, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xf9, 0x00, 0x02, 0x00, 0x75, 0x00, 0x00, 0x00,
- 0xf8, 0x00, 0x02, 0x00, 0x75, 0x00, 0x00, 0x00, 0x41, 0x00, 0x05, 0x00, 0x0e, 0x00, 0x00, 0x00,
- 0x0f, 0x00, 0x00, 0x00, 0x0c, 0x00, 0x00, 0x00, 0x0d, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00,
- 0x09, 0x00, 0x00, 0x00, 0x10, 0x00, 0x00, 0x00, 0x0f, 0x00, 0x00, 0x00, 0x7c, 0x00, 0x04, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x11, 0x00, 0x00, 0x00, 0x10, 0x00, 0x00, 0x00, 0x84, 0x00, 0x05, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x14, 0x00, 0x00, 0x00, 0x11, 0x00, 0x00, 0x00, 0x13, 0x00, 0x00, 0x00,
- 0x44, 0x00, 0x05, 0x00, 0x09, 0x00, 0x00, 0x00, 0x19, 0x00, 0x00, 0x00, 0x18, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0x7c, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x1a, 0x00, 0x00, 0x00,
- 0x19, 0x00, 0x00, 0x00, 0xaf, 0x00, 0x05, 0x00, 0x1b, 0x00, 0x00, 0x00, 0x1c, 0x00, 0x00, 0x00,
- 0x14, 0x00, 0x00, 0x00, 0x1a, 0x00, 0x00, 0x00, 0xf7, 0x00, 0x03, 0x00, 0x1e, 0x00, 0x00, 0x00,
- 0x00, 0x00, 0x00, 0x00, 0xfa, 0x00, 0x04, 0x00, 0x1c, 0x00, 0x00, 0x00, 0x1d, 0x00, 0x00, 0x00,
- 0x1e, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x1d, 0x00, 0x00, 0x00, 0xf9, 0x00, 0x02, 0x00,
- 0x73, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x1e, 0x00, 0x00, 0x00, 0x41, 0x00, 0x05, 0x00,
- 0x26, 0x00, 0x00, 0x00, 0x27, 0x00, 0x00, 0x00, 0x24, 0x00, 0x00, 0x00, 0x25, 0x00, 0x00, 0x00,
- 0x3d, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x28, 0x00, 0x00, 0x00, 0x27, 0x00, 0x00, 0x00,
- 0xc4, 0x00, 0x05, 0x00, 0x06, 0x00, 0x00, 0x00, 0x29, 0x00, 0x00, 0x00, 0x21, 0x00, 0x00, 0x00,
- 0x28, 0x00, 0x00, 0x00, 0x82, 0x00, 0x05, 0x00, 0x06, 0x00, 0x00, 0x00, 0x2e, 0x00, 0x00, 0x00,
- 0x2b, 0x00, 0x00, 0x00, 0x28, 0x00, 0x00, 0x00, 0xc4, 0x00, 0x05, 0x00, 0x06, 0x00, 0x00, 0x00,
- 0x31, 0x00, 0x00, 0x00, 0x25, 0x00, 0x00, 0x00, 0x2e, 0x00, 0x00, 0x00, 0x82, 0x00, 0x05, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x32, 0x00, 0x00, 0x00, 0x31, 0x00, 0x00, 0x00, 0x25, 0x00, 0x00, 0x00,
- 0xf9, 0x00, 0x02, 0x00, 0x35, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x35, 0x00, 0x00, 0x00,
- 0xf5, 0x00, 0x07, 0x00, 0x09, 0x00, 0x00, 0x00, 0x7b, 0x00, 0x00, 0x00, 0x0d, 0x00, 0x00, 0x00,
- 0x1e, 0x00, 0x00, 0x00, 0x6f, 0x00, 0x00, 0x00, 0x36, 0x00, 0x00, 0x00, 0xb0, 0x00, 0x05, 0x00,
- 0x1b, 0x00, 0x00, 0x00, 0x3c, 0x00, 0x00, 0x00, 0x7b, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x00, 0x00,
- 0xf6, 0x00, 0x04, 0x00, 0x37, 0x00, 0x00, 0x00, 0x36, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
- 0xfa, 0x00, 0x04, 0x00, 0x3c, 0x00, 0x00, 0x00, 0x36, 0x00, 0x00, 0x00, 0x37, 0x00, 0x00, 0x00,
- 0xf8, 0x00, 0x02, 0x00, 0x36, 0x00, 0x00, 0x00, 0x84, 0x00, 0x05, 0x00, 0x06, 0x00, 0x00, 0x00,
- 0x40, 0x00, 0x00, 0x00, 0x11, 0x00, 0x00, 0x00, 0x3f, 0x00, 0x00, 0x00, 0x3e, 0x00, 0x03, 0x00,
- 0x47, 0x00, 0x00, 0x00, 0x44, 0x00, 0x00, 0x00, 0x41, 0x00, 0x05, 0x00, 0x07, 0x00, 0x00, 0x00,
- 0x48, 0x00, 0x00, 0x00, 0x47, 0x00, 0x00, 0x00, 0x7b, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x49, 0x00, 0x00, 0x00, 0x48, 0x00, 0x00, 0x00, 0x80, 0x00, 0x05, 0x00,
- 0x06, 0x00, 0x00, 0x00, 0x4a, 0x00, 0x00, 0x00, 0x40, 0x00, 0x00, 0x00, 0x49, 0x00, 0x00, 0x00,
- 0xc3, 0x00, 0x05, 0x00, 0x06, 0x00, 0x00, 0x00, 0x4e, 0x00, 0x00, 0x00, 0x4a, 0x00, 0x00, 0x00,
- 0x2e, 0x00, 0x00, 0x00, 0xc7, 0x00, 0x05, 0x00, 0x06, 0x00, 0x00, 0x00, 0x52, 0x00, 0x00, 0x00,
- 0x4a, 0x00, 0x00, 0x00, 0x32, 0x00, 0x00, 0x00, 0x84, 0x00, 0x05, 0x00, 0x06, 0x00, 0x00, 0x00,
- 0x54, 0x00, 0x00, 0x00, 0x52, 0x00, 0x00, 0x00, 0x29, 0x00, 0x00, 0x00, 0x41, 0x00, 0x06, 0x00,
- 0x5b, 0x00, 0x00, 0x00, 0x5c, 0x00, 0x00, 0x00, 0x59, 0x00, 0x00, 0x00, 0x42, 0x00, 0x00, 0x00,
- 0x4e, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00, 0x09, 0x00, 0x00, 0x00, 0x5d, 0x00, 0x00, 0x00,
- 0x5c, 0x00, 0x00, 0x00, 0xcb, 0x00, 0x06, 0x00, 0x09, 0x00, 0x00, 0x00, 0x62, 0x00, 0x00, 0x00,
- 0x5d, 0x00, 0x00, 0x00, 0x54, 0x00, 0x00, 0x00, 0x29, 0x00, 0x00, 0x00, 0x7c, 0x00, 0x04, 0x00,
- 0x09, 0x00, 0x00, 0x00, 0x65, 0x00, 0x00, 0x00, 0x14, 0x00, 0x00, 0x00, 0x80, 0x00, 0x05, 0x00,
- 0x09, 0x00, 0x00, 0x00, 0x67, 0x00, 0x00, 0x00, 0x65, 0x00, 0x00, 0x00, 0x7b, 0x00, 0x00, 0x00,
- 0x41, 0x00, 0x05, 0x00, 0x69, 0x00, 0x00, 0x00, 0x6a, 0x00, 0x00, 0x00, 0x24, 0x00, 0x00, 0x00,
- 0x42, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00, 0x09, 0x00, 0x00, 0x00, 0x6b, 0x00, 0x00, 0x00,
- 0x6a, 0x00, 0x00, 0x00, 0x80, 0x00, 0x05, 0x00, 0x09, 0x00, 0x00, 0x00, 0x6c, 0x00, 0x00, 0x00,
- 0x62, 0x00, 0x00, 0x00, 0x6b, 0x00, 0x00, 0x00, 0x41, 0x00, 0x06, 0x00, 0x5b, 0x00, 0x00, 0x00,
- 0x6d, 0x00, 0x00, 0x00, 0x18, 0x00, 0x00, 0x00, 0x42, 0x00, 0x00, 0x00, 0x67, 0x00, 0x00, 0x00,
- 0x3e, 0x00, 0x03, 0x00, 0x6d, 0x00, 0x00, 0x00, 0x6c, 0x00, 0x00, 0x00, 0x80, 0x00, 0x05, 0x00,
- 0x09, 0x00, 0x00, 0x00, 0x6f, 0x00, 0x00, 0x00, 0x7b, 0x00, 0x00, 0x00, 0x25, 0x00, 0x00, 0x00,
- 0xf9, 0x00, 0x02, 0x00, 0x35, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x37, 0x00, 0x00, 0x00,
- 0xf9, 0x00, 0x02, 0x00, 0x73, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x76, 0x00, 0x00, 0x00,
- 0xf9, 0x00, 0x02, 0x00, 0x74, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x73, 0x00, 0x00, 0x00,
- 0xfd, 0x00, 0x01, 0x00, 0x38, 0x00, 0x01, 0x00,
-};
-
std::array<VkDescriptorSetLayoutBinding, 2> BuildInputOutputDescriptorSetBindings() {
return {{
{
@@ -381,8 +89,8 @@ VkDescriptorUpdateTemplateEntryKHR BuildInputOutputDescriptorUpdateTemplate() {
VKComputePass::VKComputePass(const VKDevice& device, VKDescriptorPool& descriptor_pool,
vk::Span<VkDescriptorSetLayoutBinding> bindings,
vk::Span<VkDescriptorUpdateTemplateEntryKHR> templates,
- vk::Span<VkPushConstantRange> push_constants, std::size_t code_size,
- const u8* code) {
+ vk::Span<VkPushConstantRange> push_constants,
+ std::span<const u32> code) {
descriptor_set_layout = device.GetLogical().CreateDescriptorSetLayout({
.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO,
.pNext = nullptr,
@@ -390,7 +98,6 @@ VKComputePass::VKComputePass(const VKDevice& device, VKDescriptorPool& descripto
.bindingCount = bindings.size(),
.pBindings = bindings.data(),
});
-
layout = device.GetLogical().CreatePipelineLayout({
.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO,
.pNext = nullptr,
@@ -400,7 +107,6 @@ VKComputePass::VKComputePass(const VKDevice& device, VKDescriptorPool& descripto
.pushConstantRangeCount = push_constants.size(),
.pPushConstantRanges = push_constants.data(),
});
-
if (!templates.empty()) {
descriptor_template = device.GetLogical().CreateDescriptorUpdateTemplateKHR({
.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_UPDATE_TEMPLATE_CREATE_INFO_KHR,
@@ -417,18 +123,13 @@ VKComputePass::VKComputePass(const VKDevice& device, VKDescriptorPool& descripto
descriptor_allocator.emplace(descriptor_pool, *descriptor_set_layout);
}
-
- auto code_copy = std::make_unique<u32[]>(code_size / sizeof(u32) + 1);
- std::memcpy(code_copy.get(), code, code_size);
-
module = device.GetLogical().CreateShaderModule({
.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
- .codeSize = code_size,
- .pCode = code_copy.get(),
+ .codeSize = static_cast<u32>(code.size_bytes()),
+ .pCode = code.data(),
});
-
pipeline = device.GetLogical().CreateComputePipeline({
.sType = VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO,
.pNext = nullptr,
@@ -467,7 +168,7 @@ QuadArrayPass::QuadArrayPass(const VKDevice& device_, VKScheduler& scheduler_,
VKUpdateDescriptorQueue& update_descriptor_queue_)
: VKComputePass(device_, descriptor_pool_, BuildQuadArrayPassDescriptorSetLayoutBinding(),
BuildQuadArrayPassDescriptorUpdateTemplateEntry(),
- BuildComputePushConstantRange(sizeof(u32)), std::size(quad_array), quad_array),
+ BuildComputePushConstantRange(sizeof(u32)), VULKAN_QUAD_ARRAY_COMP_SPV),
scheduler{scheduler_}, staging_buffer_pool{staging_buffer_pool_},
update_descriptor_queue{update_descriptor_queue_} {}
@@ -510,12 +211,11 @@ std::pair<VkBuffer, VkDeviceSize> QuadArrayPass::Assemble(u32 num_vertices, u32
return {*buffer.handle, 0};
}
-Uint8Pass::Uint8Pass(const VKDevice& device_, VKScheduler& scheduler_,
- VKDescriptorPool& descriptor_pool_, VKStagingBufferPool& staging_buffer_pool_,
+Uint8Pass::Uint8Pass(const VKDevice& device, VKScheduler& scheduler_,
+ VKDescriptorPool& descriptor_pool, VKStagingBufferPool& staging_buffer_pool_,
VKUpdateDescriptorQueue& update_descriptor_queue_)
- : VKComputePass(device_, descriptor_pool_, BuildInputOutputDescriptorSetBindings(),
- BuildInputOutputDescriptorUpdateTemplate(), {}, std::size(uint8_pass),
- uint8_pass),
+ : VKComputePass(device, descriptor_pool, BuildInputOutputDescriptorSetBindings(),
+ BuildInputOutputDescriptorUpdateTemplate(), {}, VULKAN_UINT8_COMP_SPV),
scheduler{scheduler_}, staging_buffer_pool{staging_buffer_pool_},
update_descriptor_queue{update_descriptor_queue_} {}
@@ -561,8 +261,7 @@ QuadIndexedPass::QuadIndexedPass(const VKDevice& device_, VKScheduler& scheduler
VKUpdateDescriptorQueue& update_descriptor_queue_)
: VKComputePass(device_, descriptor_pool_, BuildInputOutputDescriptorSetBindings(),
BuildInputOutputDescriptorUpdateTemplate(),
- BuildComputePushConstantRange(sizeof(u32) * 2), std::size(QUAD_INDEXED_SPV),
- QUAD_INDEXED_SPV),
+ BuildComputePushConstantRange(sizeof(u32) * 2), VULKAN_QUAD_INDEXED_COMP_SPV),
scheduler{scheduler_}, staging_buffer_pool{staging_buffer_pool_},
update_descriptor_queue{update_descriptor_queue_} {}
diff --git a/src/video_core/renderer_vulkan/vk_compute_pass.h b/src/video_core/renderer_vulkan/vk_compute_pass.h
index 2dc87902c..abdf61e2c 100644
--- a/src/video_core/renderer_vulkan/vk_compute_pass.h
+++ b/src/video_core/renderer_vulkan/vk_compute_pass.h
@@ -5,6 +5,7 @@
#pragma once
#include <optional>
+#include <span>
#include <utility>
#include "common/common_types.h"
@@ -24,8 +25,7 @@ public:
explicit VKComputePass(const VKDevice& device, VKDescriptorPool& descriptor_pool,
vk::Span<VkDescriptorSetLayoutBinding> bindings,
vk::Span<VkDescriptorUpdateTemplateEntryKHR> templates,
- vk::Span<VkPushConstantRange> push_constants, std::size_t code_size,
- const u8* code);
+ vk::Span<VkPushConstantRange> push_constants, std::span<const u32> code);
~VKComputePass();
protected:
diff --git a/src/video_core/renderer_vulkan/vk_device.cpp b/src/video_core/renderer_vulkan/vk_device.cpp
index ce3846195..370a63f74 100644
--- a/src/video_core/renderer_vulkan/vk_device.cpp
+++ b/src/video_core/renderer_vulkan/vk_device.cpp
@@ -46,6 +46,7 @@ constexpr std::array REQUIRED_EXTENSIONS{
VK_KHR_DRIVER_PROPERTIES_EXTENSION_NAME,
VK_KHR_DESCRIPTOR_UPDATE_TEMPLATE_EXTENSION_NAME,
VK_KHR_TIMELINE_SEMAPHORE_EXTENSION_NAME,
+ VK_KHR_SAMPLER_MIRROR_CLAMP_TO_EDGE_EXTENSION_NAME,
VK_EXT_VERTEX_ATTRIBUTE_DIVISOR_EXTENSION_NAME,
VK_EXT_SHADER_SUBGROUP_BALLOT_EXTENSION_NAME,
VK_EXT_SHADER_SUBGROUP_VOTE_EXTENSION_NAME,
@@ -122,6 +123,7 @@ std::unordered_map<VkFormat, VkFormatProperties> GetFormatProperties(
VK_FORMAT_R16G16_UNORM,
VK_FORMAT_R16G16_SNORM,
VK_FORMAT_R16G16_SFLOAT,
+ VK_FORMAT_R16G16_SINT,
VK_FORMAT_R16_UNORM,
VK_FORMAT_R16_UINT,
VK_FORMAT_R8G8B8A8_SRGB,
@@ -161,18 +163,32 @@ std::unordered_map<VkFormat, VkFormatProperties> GetFormatProperties(
VK_FORMAT_BC2_SRGB_BLOCK,
VK_FORMAT_BC3_SRGB_BLOCK,
VK_FORMAT_BC7_SRGB_BLOCK,
+ VK_FORMAT_ASTC_4x4_UNORM_BLOCK,
VK_FORMAT_ASTC_4x4_SRGB_BLOCK,
- VK_FORMAT_ASTC_8x8_SRGB_BLOCK,
- VK_FORMAT_ASTC_8x5_SRGB_BLOCK,
+ VK_FORMAT_ASTC_5x4_UNORM_BLOCK,
VK_FORMAT_ASTC_5x4_SRGB_BLOCK,
VK_FORMAT_ASTC_5x5_UNORM_BLOCK,
VK_FORMAT_ASTC_5x5_SRGB_BLOCK,
- VK_FORMAT_ASTC_10x8_UNORM_BLOCK,
- VK_FORMAT_ASTC_10x8_SRGB_BLOCK,
+ VK_FORMAT_ASTC_6x5_UNORM_BLOCK,
+ VK_FORMAT_ASTC_6x5_SRGB_BLOCK,
VK_FORMAT_ASTC_6x6_UNORM_BLOCK,
VK_FORMAT_ASTC_6x6_SRGB_BLOCK,
+ VK_FORMAT_ASTC_8x5_UNORM_BLOCK,
+ VK_FORMAT_ASTC_8x5_SRGB_BLOCK,
+ VK_FORMAT_ASTC_8x6_UNORM_BLOCK,
+ VK_FORMAT_ASTC_8x6_SRGB_BLOCK,
+ VK_FORMAT_ASTC_8x8_UNORM_BLOCK,
+ VK_FORMAT_ASTC_8x8_SRGB_BLOCK,
+ VK_FORMAT_ASTC_10x5_UNORM_BLOCK,
+ VK_FORMAT_ASTC_10x5_SRGB_BLOCK,
+ VK_FORMAT_ASTC_10x6_UNORM_BLOCK,
+ VK_FORMAT_ASTC_10x6_SRGB_BLOCK,
+ VK_FORMAT_ASTC_10x8_UNORM_BLOCK,
+ VK_FORMAT_ASTC_10x8_SRGB_BLOCK,
VK_FORMAT_ASTC_10x10_UNORM_BLOCK,
VK_FORMAT_ASTC_10x10_SRGB_BLOCK,
+ VK_FORMAT_ASTC_12x10_UNORM_BLOCK,
+ VK_FORMAT_ASTC_12x10_SRGB_BLOCK,
VK_FORMAT_ASTC_12x12_UNORM_BLOCK,
VK_FORMAT_ASTC_12x12_SRGB_BLOCK,
VK_FORMAT_ASTC_8x6_UNORM_BLOCK,
@@ -192,7 +208,7 @@ std::unordered_map<VkFormat, VkFormatProperties> GetFormatProperties(
VKDevice::VKDevice(VkInstance instance_, u32 instance_version_, vk::PhysicalDevice physical_,
VkSurfaceKHR surface, const vk::InstanceDispatch& dld_)
- : dld{dld_}, physical{physical_}, properties{physical.GetProperties()},
+ : instance{instance_}, dld{dld_}, physical{physical_}, properties{physical.GetProperties()},
instance_version{instance_version_}, format_properties{GetFormatProperties(physical, dld)} {
SetupFamilies(surface);
SetupFeatures();
@@ -214,7 +230,7 @@ bool VKDevice::Create() {
features2.features = {
.robustBufferAccess = false,
.fullDrawIndexUint32 = false,
- .imageCubeArray = false,
+ .imageCubeArray = true,
.independentBlend = true,
.geometryShader = true,
.tessellationShader = true,
@@ -242,7 +258,7 @@ bool VKDevice::Create() {
.shaderTessellationAndGeometryPointSize = false,
.shaderImageGatherExtended = true,
.shaderStorageImageExtendedFormats = false,
- .shaderStorageImageMultisample = false,
+ .shaderStorageImageMultisample = true,
.shaderStorageImageReadWithoutFormat = is_formatless_image_load_supported,
.shaderStorageImageWriteWithoutFormat = true,
.shaderUniformBufferArrayDynamicIndexing = false,
@@ -268,7 +284,6 @@ bool VKDevice::Create() {
.variableMultisampleRate = false,
.inheritedQueries = false,
};
-
VkPhysicalDeviceTimelineSemaphoreFeaturesKHR timeline_semaphore{
.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_TIMELINE_SEMAPHORE_FEATURES_KHR,
.pNext = nullptr,
@@ -380,6 +395,20 @@ bool VKDevice::Create() {
LOG_INFO(Render_Vulkan, "Device doesn't support extended dynamic state");
}
+ VkPhysicalDeviceRobustness2FeaturesEXT robustness2;
+ if (ext_robustness2) {
+ robustness2 = {
+ .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_ROBUSTNESS_2_FEATURES_EXT,
+ .pNext = nullptr,
+ .robustBufferAccess2 = false,
+ .robustImageAccess2 = true,
+ .nullDescriptor = true,
+ };
+ SetNext(next, robustness2);
+ } else {
+ LOG_INFO(Render_Vulkan, "Device doesn't support robustness2");
+ }
+
if (!ext_depth_range_unrestricted) {
LOG_INFO(Render_Vulkan, "Device doesn't support depth range unrestricted");
}
@@ -405,7 +434,14 @@ bool VKDevice::Create() {
}
CollectTelemetryParameters();
+ CollectToolingInfo();
+ if (ext_extended_dynamic_state && driver_id == VK_DRIVER_ID_MESA_RADV) {
+ LOG_WARNING(
+ Render_Vulkan,
+ "Blacklisting RADV for VK_EXT_extended_dynamic state, likely due to a bug in yuzu");
+ ext_extended_dynamic_state = false;
+ }
if (ext_extended_dynamic_state && IsRDNA(properties.deviceName, driver_id)) {
// AMD's proprietary driver supports VK_EXT_extended_dynamic_state but on RDNA devices it
// seems to cause stability issues
@@ -458,7 +494,7 @@ void VKDevice::ReportLoss() const {
LOG_CRITICAL(Render_Vulkan, "Device loss occured!");
// Wait for the log to flush and for Nsight Aftermath to dump the results
- std::this_thread::sleep_for(std::chrono::seconds{3});
+ std::this_thread::sleep_for(std::chrono::seconds{15});
}
void VKDevice::SaveShader(const std::vector<u32>& spirv) const {
@@ -499,6 +535,16 @@ bool VKDevice::IsOptimalAstcSupported(const VkPhysicalDeviceFeatures& features)
return true;
}
+bool VKDevice::TestDepthStencilBlits() const {
+ static constexpr VkFormatFeatureFlags required_features =
+ VK_FORMAT_FEATURE_BLIT_SRC_BIT | VK_FORMAT_FEATURE_BLIT_DST_BIT;
+ const auto test_features = [](VkFormatProperties props) {
+ return (props.optimalTilingFeatures & required_features) == required_features;
+ };
+ return test_features(format_properties.at(VK_FORMAT_D32_SFLOAT_S8_UINT)) &&
+ test_features(format_properties.at(VK_FORMAT_D24_UNORM_S8_UINT));
+}
+
bool VKDevice::IsFormatSupported(VkFormat wanted_format, VkFormatFeatureFlags wanted_usage,
FormatType format_type) const {
const auto it = format_properties.find(wanted_format);
@@ -569,6 +615,7 @@ bool VKDevice::IsSuitable(vk::PhysicalDevice physical, VkSurfaceKHR surface) {
const auto features{physical.GetFeatures()};
const std::array feature_report = {
std::make_pair(features.vertexPipelineStoresAndAtomics, "vertexPipelineStoresAndAtomics"),
+ std::make_pair(features.imageCubeArray, "imageCubeArray"),
std::make_pair(features.independentBlend, "independentBlend"),
std::make_pair(features.depthClamp, "depthClamp"),
std::make_pair(features.samplerAnisotropy, "samplerAnisotropy"),
@@ -580,6 +627,7 @@ bool VKDevice::IsSuitable(vk::PhysicalDevice physical, VkSurfaceKHR surface) {
std::make_pair(features.occlusionQueryPrecise, "occlusionQueryPrecise"),
std::make_pair(features.fragmentStoresAndAtomics, "fragmentStoresAndAtomics"),
std::make_pair(features.shaderImageGatherExtended, "shaderImageGatherExtended"),
+ std::make_pair(features.shaderStorageImageMultisample, "shaderStorageImageMultisample"),
std::make_pair(features.shaderStorageImageWriteWithoutFormat,
"shaderStorageImageWriteWithoutFormat"),
};
@@ -608,6 +656,7 @@ std::vector<const char*> VKDevice::LoadExtensions() {
bool has_ext_transform_feedback{};
bool has_ext_custom_border_color{};
bool has_ext_extended_dynamic_state{};
+ bool has_ext_robustness2{};
for (const VkExtensionProperties& extension : physical.EnumerateDeviceExtensionProperties()) {
const auto test = [&](std::optional<std::reference_wrapper<bool>> status, const char* name,
bool push) {
@@ -627,11 +676,15 @@ std::vector<const char*> VKDevice::LoadExtensions() {
test(has_khr_shader_float16_int8, VK_KHR_SHADER_FLOAT16_INT8_EXTENSION_NAME, false);
test(ext_depth_range_unrestricted, VK_EXT_DEPTH_RANGE_UNRESTRICTED_EXTENSION_NAME, true);
test(ext_index_type_uint8, VK_EXT_INDEX_TYPE_UINT8_EXTENSION_NAME, true);
+ test(ext_sampler_filter_minmax, VK_EXT_SAMPLER_FILTER_MINMAX_EXTENSION_NAME, true);
test(ext_shader_viewport_index_layer, VK_EXT_SHADER_VIEWPORT_INDEX_LAYER_EXTENSION_NAME,
true);
+ test(ext_tooling_info, VK_EXT_TOOLING_INFO_EXTENSION_NAME, true);
+ test(ext_shader_stencil_export, VK_EXT_SHADER_STENCIL_EXPORT_EXTENSION_NAME, true);
test(has_ext_transform_feedback, VK_EXT_TRANSFORM_FEEDBACK_EXTENSION_NAME, false);
test(has_ext_custom_border_color, VK_EXT_CUSTOM_BORDER_COLOR_EXTENSION_NAME, false);
test(has_ext_extended_dynamic_state, VK_EXT_EXTENDED_DYNAMIC_STATE_EXTENSION_NAME, false);
+ test(has_ext_robustness2, VK_EXT_ROBUSTNESS_2_EXTENSION_NAME, false);
if (instance_version >= VK_API_VERSION_1_1) {
test(has_ext_subgroup_size_control, VK_EXT_SUBGROUP_SIZE_CONTROL_EXTENSION_NAME, false);
}
@@ -733,6 +786,18 @@ std::vector<const char*> VKDevice::LoadExtensions() {
}
}
+ if (has_ext_robustness2) {
+ VkPhysicalDeviceRobustness2FeaturesEXT robustness2;
+ robustness2.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_ROBUSTNESS_2_FEATURES_EXT;
+ robustness2.pNext = nullptr;
+ features.pNext = &robustness2;
+ physical.GetFeatures2KHR(features);
+ if (robustness2.nullDescriptor && robustness2.robustImageAccess2) {
+ extensions.push_back(VK_EXT_ROBUSTNESS_2_EXTENSION_NAME);
+ ext_robustness2 = true;
+ }
+ }
+
return extensions;
}
@@ -764,6 +829,7 @@ void VKDevice::SetupFamilies(VkSurfaceKHR surface) {
void VKDevice::SetupFeatures() {
const auto supported_features{physical.GetFeatures()};
is_formatless_image_load_supported = supported_features.shaderStorageImageReadWithoutFormat;
+ is_blit_depth_stencil_supported = TestDepthStencilBlits();
is_optimal_astc_supported = IsOptimalAstcSupported(supported_features);
}
@@ -794,6 +860,32 @@ void VKDevice::CollectTelemetryParameters() {
}
}
+void VKDevice::CollectToolingInfo() {
+ if (!ext_tooling_info) {
+ return;
+ }
+ const auto vkGetPhysicalDeviceToolPropertiesEXT =
+ reinterpret_cast<PFN_vkGetPhysicalDeviceToolPropertiesEXT>(
+ dld.vkGetInstanceProcAddr(instance, "vkGetPhysicalDeviceToolPropertiesEXT"));
+ if (!vkGetPhysicalDeviceToolPropertiesEXT) {
+ return;
+ }
+ u32 tool_count = 0;
+ if (vkGetPhysicalDeviceToolPropertiesEXT(physical, &tool_count, nullptr) != VK_SUCCESS) {
+ return;
+ }
+ std::vector<VkPhysicalDeviceToolPropertiesEXT> tools(tool_count);
+ if (vkGetPhysicalDeviceToolPropertiesEXT(physical, &tool_count, tools.data()) != VK_SUCCESS) {
+ return;
+ }
+ for (const VkPhysicalDeviceToolPropertiesEXT& tool : tools) {
+ const std::string_view name = tool.name;
+ LOG_INFO(Render_Vulkan, "{}", name);
+ has_renderdoc = has_renderdoc || name == "RenderDoc";
+ has_nsight_graphics = has_nsight_graphics || name == "NVIDIA Nsight Graphics";
+ }
+}
+
std::vector<VkDeviceQueueCreateInfo> VKDevice::GetDeviceQueueCreateInfos() const {
static constexpr float QUEUE_PRIORITY = 1.0f;
diff --git a/src/video_core/renderer_vulkan/vk_device.h b/src/video_core/renderer_vulkan/vk_device.h
index 4286673d9..995dcfc0f 100644
--- a/src/video_core/renderer_vulkan/vk_device.h
+++ b/src/video_core/renderer_vulkan/vk_device.h
@@ -157,6 +157,11 @@ public:
return is_formatless_image_load_supported;
}
+ /// Returns true when blitting from and to depth stencil images is supported.
+ bool IsBlitDepthStencilSupported() const {
+ return is_blit_depth_stencil_supported;
+ }
+
/// Returns true if the device supports VK_NV_viewport_swizzle.
bool IsNvViewportSwizzleSupported() const {
return nv_viewport_swizzle;
@@ -172,6 +177,11 @@ public:
return ext_index_type_uint8;
}
+ /// Returns true if the device supports VK_EXT_sampler_filter_minmax.
+ bool IsExtSamplerFilterMinmaxSupported() const {
+ return ext_sampler_filter_minmax;
+ }
+
/// Returns true if the device supports VK_EXT_depth_range_unrestricted.
bool IsExtDepthRangeUnrestrictedSupported() const {
return ext_depth_range_unrestricted;
@@ -197,6 +207,16 @@ public:
return ext_extended_dynamic_state;
}
+ /// Returns true if the device supports VK_EXT_shader_stencil_export.
+ bool IsExtShaderStencilExportSupported() const {
+ return ext_shader_stencil_export;
+ }
+
+ /// Returns true when a known debugging tool is attached.
+ bool HasDebuggingToolAttached() const {
+ return has_renderdoc || has_nsight_graphics;
+ }
+
/// Returns the vendor name reported from Vulkan.
std::string_view GetVendorName() const {
return vendor_name;
@@ -228,16 +248,23 @@ private:
/// Collects telemetry information from the device.
void CollectTelemetryParameters();
+ /// Collects information about attached tools.
+ void CollectToolingInfo();
+
/// Returns a list of queue initialization descriptors.
std::vector<VkDeviceQueueCreateInfo> GetDeviceQueueCreateInfos() const;
/// Returns true if ASTC textures are natively supported.
bool IsOptimalAstcSupported(const VkPhysicalDeviceFeatures& features) const;
+ /// Returns true if the device natively supports blitting depth stencil images.
+ bool TestDepthStencilBlits() const;
+
/// Returns true if a format is supported.
bool IsFormatSupported(VkFormat wanted_format, VkFormatFeatureFlags wanted_usage,
FormatType format_type) const;
+ VkInstance instance; ///< Vulkan instance.
vk::DeviceDispatch dld; ///< Device function pointers.
vk::PhysicalDevice physical; ///< Physical device.
VkPhysicalDeviceProperties properties; ///< Device properties.
@@ -253,15 +280,22 @@ private:
bool is_float16_supported{}; ///< Support for float16 arithmetics.
bool is_warp_potentially_bigger{}; ///< Host warp size can be bigger than guest.
bool is_formatless_image_load_supported{}; ///< Support for shader image read without format.
+ bool is_blit_depth_stencil_supported{}; ///< Support for blitting from and to depth stencil.
bool nv_viewport_swizzle{}; ///< Support for VK_NV_viewport_swizzle.
bool khr_uniform_buffer_standard_layout{}; ///< Support for std430 on UBOs.
bool ext_index_type_uint8{}; ///< Support for VK_EXT_index_type_uint8.
+ bool ext_sampler_filter_minmax{}; ///< Support for VK_EXT_sampler_filter_minmax.
bool ext_depth_range_unrestricted{}; ///< Support for VK_EXT_depth_range_unrestricted.
bool ext_shader_viewport_index_layer{}; ///< Support for VK_EXT_shader_viewport_index_layer.
+ bool ext_tooling_info{}; ///< Support for VK_EXT_tooling_info.
bool ext_transform_feedback{}; ///< Support for VK_EXT_transform_feedback.
bool ext_custom_border_color{}; ///< Support for VK_EXT_custom_border_color.
bool ext_extended_dynamic_state{}; ///< Support for VK_EXT_extended_dynamic_state.
+ bool ext_robustness2{}; ///< Support for VK_EXT_robustness2.
+ bool ext_shader_stencil_export{}; ///< Support for VK_EXT_shader_stencil_export.
bool nv_device_diagnostics_config{}; ///< Support for VK_NV_device_diagnostics_config.
+ bool has_renderdoc{}; ///< Has RenderDoc attached
+ bool has_nsight_graphics{}; ///< Has Nsight Graphics attached
// Asynchronous Graphics Pipeline setting
bool use_asynchronous_shaders{}; ///< Setting to use asynchronous shaders/graphics pipeline
diff --git a/src/video_core/renderer_vulkan/vk_fence_manager.cpp b/src/video_core/renderer_vulkan/vk_fence_manager.cpp
index 0bcaee714..774a12a53 100644
--- a/src/video_core/renderer_vulkan/vk_fence_manager.cpp
+++ b/src/video_core/renderer_vulkan/vk_fence_manager.cpp
@@ -73,10 +73,9 @@ bool InnerFence::IsEventSignalled() const {
}
VKFenceManager::VKFenceManager(VideoCore::RasterizerInterface& rasterizer_, Tegra::GPU& gpu_,
- Tegra::MemoryManager& memory_manager_,
- VKTextureCache& texture_cache_, VKBufferCache& buffer_cache_,
- VKQueryCache& query_cache_, const VKDevice& device_,
- VKScheduler& scheduler_)
+ Tegra::MemoryManager& memory_manager_, TextureCache& texture_cache_,
+ VKBufferCache& buffer_cache_, VKQueryCache& query_cache_,
+ const VKDevice& device_, VKScheduler& scheduler_)
: GenericFenceManager{rasterizer_, gpu_, texture_cache_, buffer_cache_, query_cache_},
device{device_}, scheduler{scheduler_} {}
diff --git a/src/video_core/renderer_vulkan/vk_fence_manager.h b/src/video_core/renderer_vulkan/vk_fence_manager.h
index c8547cc24..c2869e8e3 100644
--- a/src/video_core/renderer_vulkan/vk_fence_manager.h
+++ b/src/video_core/renderer_vulkan/vk_fence_manager.h
@@ -8,6 +8,7 @@
#include "video_core/fence_manager.h"
#include "video_core/renderer_vulkan/vk_buffer_cache.h"
+#include "video_core/renderer_vulkan/vk_texture_cache.h"
#include "video_core/renderer_vulkan/wrapper.h"
namespace Core {
@@ -24,7 +25,6 @@ class VKBufferCache;
class VKDevice;
class VKQueryCache;
class VKScheduler;
-class VKTextureCache;
class InnerFence : public VideoCommon::FenceBase {
public:
@@ -51,12 +51,12 @@ private:
using Fence = std::shared_ptr<InnerFence>;
using GenericFenceManager =
- VideoCommon::FenceManager<Fence, VKTextureCache, VKBufferCache, VKQueryCache>;
+ VideoCommon::FenceManager<Fence, TextureCache, VKBufferCache, VKQueryCache>;
class VKFenceManager final : public GenericFenceManager {
public:
explicit VKFenceManager(VideoCore::RasterizerInterface& rasterizer_, Tegra::GPU& gpu_,
- Tegra::MemoryManager& memory_manager_, VKTextureCache& texture_cache_,
+ Tegra::MemoryManager& memory_manager_, TextureCache& texture_cache_,
VKBufferCache& buffer_cache_, VKQueryCache& query_cache_,
const VKDevice& device_, VKScheduler& scheduler_);
diff --git a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp
index 970979fa1..7979df3a8 100644
--- a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp
+++ b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp
@@ -15,7 +15,6 @@
#include "video_core/renderer_vulkan/vk_device.h"
#include "video_core/renderer_vulkan/vk_graphics_pipeline.h"
#include "video_core/renderer_vulkan/vk_pipeline_cache.h"
-#include "video_core/renderer_vulkan/vk_renderpass_cache.h"
#include "video_core/renderer_vulkan/vk_scheduler.h"
#include "video_core/renderer_vulkan/vk_update_descriptor.h"
#include "video_core/renderer_vulkan/wrapper.h"
@@ -69,23 +68,45 @@ VkViewportSwizzleNV UnpackViewportSwizzle(u16 swizzle) {
};
}
+VkSampleCountFlagBits ConvertMsaaMode(Tegra::Texture::MsaaMode msaa_mode) {
+ switch (msaa_mode) {
+ case Tegra::Texture::MsaaMode::Msaa1x1:
+ return VK_SAMPLE_COUNT_1_BIT;
+ case Tegra::Texture::MsaaMode::Msaa2x1:
+ case Tegra::Texture::MsaaMode::Msaa2x1_D3D:
+ return VK_SAMPLE_COUNT_2_BIT;
+ case Tegra::Texture::MsaaMode::Msaa2x2:
+ case Tegra::Texture::MsaaMode::Msaa2x2_VC4:
+ case Tegra::Texture::MsaaMode::Msaa2x2_VC12:
+ return VK_SAMPLE_COUNT_4_BIT;
+ case Tegra::Texture::MsaaMode::Msaa4x2:
+ case Tegra::Texture::MsaaMode::Msaa4x2_D3D:
+ case Tegra::Texture::MsaaMode::Msaa4x2_VC8:
+ case Tegra::Texture::MsaaMode::Msaa4x2_VC24:
+ return VK_SAMPLE_COUNT_8_BIT;
+ case Tegra::Texture::MsaaMode::Msaa4x4:
+ return VK_SAMPLE_COUNT_16_BIT;
+ default:
+ UNREACHABLE_MSG("Invalid msaa_mode={}", static_cast<int>(msaa_mode));
+ return VK_SAMPLE_COUNT_1_BIT;
+ }
+}
+
} // Anonymous namespace
VKGraphicsPipeline::VKGraphicsPipeline(const VKDevice& device_, VKScheduler& scheduler_,
VKDescriptorPool& descriptor_pool_,
VKUpdateDescriptorQueue& update_descriptor_queue_,
- VKRenderPassCache& renderpass_cache_,
- const GraphicsPipelineCacheKey& key_,
- vk::Span<VkDescriptorSetLayoutBinding> bindings_,
- const SPIRVProgram& program_)
- : device{device_}, scheduler{scheduler_}, cache_key{key_}, hash{cache_key.Hash()},
- descriptor_set_layout{CreateDescriptorSetLayout(bindings_)},
+ const GraphicsPipelineCacheKey& key,
+ vk::Span<VkDescriptorSetLayoutBinding> bindings,
+ const SPIRVProgram& program, u32 num_color_buffers)
+ : device{device_}, scheduler{scheduler_}, cache_key{key}, hash{cache_key.Hash()},
+ descriptor_set_layout{CreateDescriptorSetLayout(bindings)},
descriptor_allocator{descriptor_pool_, *descriptor_set_layout},
update_descriptor_queue{update_descriptor_queue_}, layout{CreatePipelineLayout()},
- descriptor_template{CreateDescriptorUpdateTemplate(program_)}, modules{CreateShaderModules(
- program_)},
- renderpass{renderpass_cache_.GetRenderPass(cache_key.renderpass_params)},
- pipeline{CreatePipeline(cache_key.renderpass_params, program_)} {}
+ descriptor_template{CreateDescriptorUpdateTemplate(program)},
+ modules(CreateShaderModules(program)),
+ pipeline(CreatePipeline(program, cache_key.renderpass, num_color_buffers)) {}
VKGraphicsPipeline::~VKGraphicsPipeline() = default;
@@ -179,8 +200,9 @@ std::vector<vk::ShaderModule> VKGraphicsPipeline::CreateShaderModules(
return shader_modules;
}
-vk::Pipeline VKGraphicsPipeline::CreatePipeline(const RenderPassParams& renderpass_params,
- const SPIRVProgram& program) const {
+vk::Pipeline VKGraphicsPipeline::CreatePipeline(const SPIRVProgram& program,
+ VkRenderPass renderpass,
+ u32 num_color_buffers) const {
const auto& state = cache_key.fixed_state;
const auto& viewport_swizzles = state.viewport_swizzles;
@@ -290,8 +312,7 @@ vk::Pipeline VKGraphicsPipeline::CreatePipeline(const RenderPassParams& renderpa
};
std::array<VkViewportSwizzleNV, Maxwell::NumViewports> swizzles;
- std::transform(viewport_swizzles.begin(), viewport_swizzles.end(), swizzles.begin(),
- UnpackViewportSwizzle);
+ std::ranges::transform(viewport_swizzles, swizzles.begin(), UnpackViewportSwizzle);
VkPipelineViewportSwizzleStateCreateInfoNV swizzle_ci{
.sType = VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_SWIZZLE_STATE_CREATE_INFO_NV,
.pNext = nullptr,
@@ -326,7 +347,7 @@ vk::Pipeline VKGraphicsPipeline::CreatePipeline(const RenderPassParams& renderpa
.sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
- .rasterizationSamples = VK_SAMPLE_COUNT_1_BIT,
+ .rasterizationSamples = ConvertMsaaMode(state.msaa_mode),
.sampleShadingEnable = VK_FALSE,
.minSampleShading = 0.0f,
.pSampleMask = nullptr,
@@ -352,8 +373,7 @@ vk::Pipeline VKGraphicsPipeline::CreatePipeline(const RenderPassParams& renderpa
};
std::array<VkPipelineColorBlendAttachmentState, Maxwell::NumRenderTargets> cb_attachments;
- const auto num_attachments = static_cast<std::size_t>(renderpass_params.num_color_attachments);
- for (std::size_t index = 0; index < num_attachments; ++index) {
+ for (std::size_t index = 0; index < num_color_buffers; ++index) {
static constexpr std::array COMPONENT_TABLE{
VK_COLOR_COMPONENT_R_BIT,
VK_COLOR_COMPONENT_G_BIT,
@@ -387,7 +407,7 @@ vk::Pipeline VKGraphicsPipeline::CreatePipeline(const RenderPassParams& renderpa
.flags = 0,
.logicOpEnable = VK_FALSE,
.logicOp = VK_LOGIC_OP_COPY,
- .attachmentCount = static_cast<u32>(num_attachments),
+ .attachmentCount = num_color_buffers,
.pAttachments = cb_attachments.data(),
.blendConstants = {},
};
@@ -447,8 +467,7 @@ vk::Pipeline VKGraphicsPipeline::CreatePipeline(const RenderPassParams& renderpa
stage_ci.pNext = &subgroup_size_ci;
}
}
-
- const VkGraphicsPipelineCreateInfo ci{
+ return device.GetLogical().CreateGraphicsPipeline(VkGraphicsPipelineCreateInfo{
.sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
@@ -468,8 +487,7 @@ vk::Pipeline VKGraphicsPipeline::CreatePipeline(const RenderPassParams& renderpa
.subpass = 0,
.basePipelineHandle = nullptr,
.basePipelineIndex = 0,
- };
- return device.GetLogical().CreateGraphicsPipeline(ci);
+ });
}
} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_graphics_pipeline.h b/src/video_core/renderer_vulkan/vk_graphics_pipeline.h
index 3fb31d55a..214d06b4c 100644
--- a/src/video_core/renderer_vulkan/vk_graphics_pipeline.h
+++ b/src/video_core/renderer_vulkan/vk_graphics_pipeline.h
@@ -8,10 +8,10 @@
#include <optional>
#include <vector>
+#include "common/common_types.h"
#include "video_core/engines/maxwell_3d.h"
#include "video_core/renderer_vulkan/fixed_pipeline_state.h"
#include "video_core/renderer_vulkan/vk_descriptor_pool.h"
-#include "video_core/renderer_vulkan/vk_renderpass_cache.h"
#include "video_core/renderer_vulkan/vk_shader_decompiler.h"
#include "video_core/renderer_vulkan/wrapper.h"
@@ -20,8 +20,7 @@ namespace Vulkan {
using Maxwell = Tegra::Engines::Maxwell3D::Regs;
struct GraphicsPipelineCacheKey {
- RenderPassParams renderpass_params;
- u32 padding;
+ VkRenderPass renderpass;
std::array<GPUVAddr, Maxwell::MaxShaderProgram> shaders;
FixedPipelineState fixed_state;
@@ -34,7 +33,7 @@ struct GraphicsPipelineCacheKey {
}
std::size_t Size() const noexcept {
- return sizeof(renderpass_params) + sizeof(padding) + sizeof(shaders) + fixed_state.Size();
+ return sizeof(renderpass) + sizeof(shaders) + fixed_state.Size();
}
};
static_assert(std::has_unique_object_representations_v<GraphicsPipelineCacheKey>);
@@ -43,7 +42,6 @@ static_assert(std::is_trivially_constructible_v<GraphicsPipelineCacheKey>);
class VKDescriptorPool;
class VKDevice;
-class VKRenderPassCache;
class VKScheduler;
class VKUpdateDescriptorQueue;
@@ -52,12 +50,11 @@ using SPIRVProgram = std::array<std::optional<SPIRVShader>, Maxwell::MaxShaderSt
class VKGraphicsPipeline final {
public:
explicit VKGraphicsPipeline(const VKDevice& device_, VKScheduler& scheduler_,
- VKDescriptorPool& descriptor_pool_,
+ VKDescriptorPool& descriptor_pool,
VKUpdateDescriptorQueue& update_descriptor_queue_,
- VKRenderPassCache& renderpass_cache_,
- const GraphicsPipelineCacheKey& key_,
- vk::Span<VkDescriptorSetLayoutBinding> bindings_,
- const SPIRVProgram& program_);
+ const GraphicsPipelineCacheKey& key,
+ vk::Span<VkDescriptorSetLayoutBinding> bindings,
+ const SPIRVProgram& program, u32 num_color_buffers);
~VKGraphicsPipeline();
VkDescriptorSet CommitDescriptorSet();
@@ -70,10 +67,6 @@ public:
return *layout;
}
- VkRenderPass GetRenderPass() const {
- return renderpass;
- }
-
GraphicsPipelineCacheKey GetCacheKey() const {
return cache_key;
}
@@ -89,8 +82,8 @@ private:
std::vector<vk::ShaderModule> CreateShaderModules(const SPIRVProgram& program) const;
- vk::Pipeline CreatePipeline(const RenderPassParams& renderpass_params,
- const SPIRVProgram& program) const;
+ vk::Pipeline CreatePipeline(const SPIRVProgram& program, VkRenderPass renderpass,
+ u32 num_color_buffers) const;
const VKDevice& device;
VKScheduler& scheduler;
@@ -104,7 +97,6 @@ private:
vk::DescriptorUpdateTemplateKHR descriptor_template;
std::vector<vk::ShaderModule> modules;
- VkRenderPass renderpass;
vk::Pipeline pipeline;
};
diff --git a/src/video_core/renderer_vulkan/vk_image.cpp b/src/video_core/renderer_vulkan/vk_image.cpp
deleted file mode 100644
index 072d14e3b..000000000
--- a/src/video_core/renderer_vulkan/vk_image.cpp
+++ /dev/null
@@ -1,135 +0,0 @@
-// Copyright 2018 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#include <memory>
-#include <vector>
-
-#include "common/assert.h"
-#include "video_core/renderer_vulkan/vk_device.h"
-#include "video_core/renderer_vulkan/vk_image.h"
-#include "video_core/renderer_vulkan/vk_scheduler.h"
-#include "video_core/renderer_vulkan/wrapper.h"
-
-namespace Vulkan {
-
-VKImage::VKImage(const VKDevice& device_, VKScheduler& scheduler_,
- const VkImageCreateInfo& image_ci_, VkImageAspectFlags aspect_mask_)
- : device{device_}, scheduler{scheduler_}, format{image_ci_.format}, aspect_mask{aspect_mask_},
- image_num_layers{image_ci_.arrayLayers}, image_num_levels{image_ci_.mipLevels} {
- UNIMPLEMENTED_IF_MSG(image_ci_.queueFamilyIndexCount != 0,
- "Queue family tracking is not implemented");
-
- image = device_.GetLogical().CreateImage(image_ci_);
-
- const u32 num_ranges = image_num_layers * image_num_levels;
- barriers.resize(num_ranges);
- subrange_states.resize(num_ranges, {{}, image_ci_.initialLayout});
-}
-
-VKImage::~VKImage() = default;
-
-void VKImage::Transition(u32 base_layer, u32 num_layers, u32 base_level, u32 num_levels,
- VkPipelineStageFlags new_stage_mask, VkAccessFlags new_access,
- VkImageLayout new_layout) {
- if (!HasChanged(base_layer, num_layers, base_level, num_levels, new_access, new_layout)) {
- return;
- }
-
- std::size_t cursor = 0;
- for (u32 layer_it = 0; layer_it < num_layers; ++layer_it) {
- for (u32 level_it = 0; level_it < num_levels; ++level_it, ++cursor) {
- const u32 layer = base_layer + layer_it;
- const u32 level = base_level + level_it;
- auto& state = GetSubrangeState(layer, level);
- auto& barrier = barriers[cursor];
- barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
- barrier.pNext = nullptr;
- barrier.srcAccessMask = state.access;
- barrier.dstAccessMask = new_access;
- barrier.oldLayout = state.layout;
- barrier.newLayout = new_layout;
- barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
- barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
- barrier.image = *image;
- barrier.subresourceRange.aspectMask = aspect_mask;
- barrier.subresourceRange.baseMipLevel = level;
- barrier.subresourceRange.levelCount = 1;
- barrier.subresourceRange.baseArrayLayer = layer;
- barrier.subresourceRange.layerCount = 1;
- state.access = new_access;
- state.layout = new_layout;
- }
- }
-
- scheduler.RequestOutsideRenderPassOperationContext();
-
- scheduler.Record([barriers = barriers, cursor](vk::CommandBuffer cmdbuf) {
- // TODO(Rodrigo): Implement a way to use the latest stage across subresources.
- cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
- VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, 0, {}, {},
- vk::Span(barriers.data(), cursor));
- });
-}
-
-bool VKImage::HasChanged(u32 base_layer, u32 num_layers, u32 base_level, u32 num_levels,
- VkAccessFlags new_access, VkImageLayout new_layout) noexcept {
- const bool is_full_range = base_layer == 0 && num_layers == image_num_layers &&
- base_level == 0 && num_levels == image_num_levels;
- if (!is_full_range) {
- state_diverged = true;
- }
-
- if (!state_diverged) {
- auto& state = GetSubrangeState(0, 0);
- if (state.access != new_access || state.layout != new_layout) {
- return true;
- }
- }
-
- for (u32 layer_it = 0; layer_it < num_layers; ++layer_it) {
- for (u32 level_it = 0; level_it < num_levels; ++level_it) {
- const u32 layer = base_layer + layer_it;
- const u32 level = base_level + level_it;
- auto& state = GetSubrangeState(layer, level);
- if (state.access != new_access || state.layout != new_layout) {
- return true;
- }
- }
- }
- return false;
-}
-
-void VKImage::CreatePresentView() {
- // Image type has to be 2D to be presented.
- present_view = device.GetLogical().CreateImageView({
- .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
- .pNext = nullptr,
- .flags = 0,
- .image = *image,
- .viewType = VK_IMAGE_VIEW_TYPE_2D,
- .format = format,
- .components =
- {
- .r = VK_COMPONENT_SWIZZLE_IDENTITY,
- .g = VK_COMPONENT_SWIZZLE_IDENTITY,
- .b = VK_COMPONENT_SWIZZLE_IDENTITY,
- .a = VK_COMPONENT_SWIZZLE_IDENTITY,
- },
- .subresourceRange =
- {
- .aspectMask = aspect_mask,
- .baseMipLevel = 0,
- .levelCount = 1,
- .baseArrayLayer = 0,
- .layerCount = 1,
- },
- });
-}
-
-VKImage::SubrangeState& VKImage::GetSubrangeState(u32 layer, u32 level) noexcept {
- return subrange_states[static_cast<std::size_t>(layer * image_num_levels) +
- static_cast<std::size_t>(level)];
-}
-
-} // namespace Vulkan \ No newline at end of file
diff --git a/src/video_core/renderer_vulkan/vk_image.h b/src/video_core/renderer_vulkan/vk_image.h
deleted file mode 100644
index 287ab90ca..000000000
--- a/src/video_core/renderer_vulkan/vk_image.h
+++ /dev/null
@@ -1,84 +0,0 @@
-// Copyright 2018 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include <memory>
-#include <vector>
-
-#include "common/common_types.h"
-#include "video_core/renderer_vulkan/wrapper.h"
-
-namespace Vulkan {
-
-class VKDevice;
-class VKScheduler;
-
-class VKImage {
-public:
- explicit VKImage(const VKDevice& device_, VKScheduler& scheduler_,
- const VkImageCreateInfo& image_ci_, VkImageAspectFlags aspect_mask_);
- ~VKImage();
-
- /// Records in the passed command buffer an image transition and updates the state of the image.
- void Transition(u32 base_layer, u32 num_layers, u32 base_level, u32 num_levels,
- VkPipelineStageFlags new_stage_mask, VkAccessFlags new_access,
- VkImageLayout new_layout);
-
- /// Returns a view compatible with presentation, the image has to be 2D.
- VkImageView GetPresentView() {
- if (!present_view) {
- CreatePresentView();
- }
- return *present_view;
- }
-
- /// Returns the Vulkan image handler.
- const vk::Image& GetHandle() const {
- return image;
- }
-
- /// Returns the Vulkan format for this image.
- VkFormat GetFormat() const {
- return format;
- }
-
- /// Returns the Vulkan aspect mask.
- VkImageAspectFlags GetAspectMask() const {
- return aspect_mask;
- }
-
-private:
- struct SubrangeState final {
- VkAccessFlags access = 0; ///< Current access bits.
- VkImageLayout layout = VK_IMAGE_LAYOUT_UNDEFINED; ///< Current image layout.
- };
-
- bool HasChanged(u32 base_layer, u32 num_layers, u32 base_level, u32 num_levels,
- VkAccessFlags new_access, VkImageLayout new_layout) noexcept;
-
- /// Creates a presentation view.
- void CreatePresentView();
-
- /// Returns the subrange state for a layer and layer.
- SubrangeState& GetSubrangeState(u32 layer, u32 level) noexcept;
-
- const VKDevice& device; ///< Device handler.
- VKScheduler& scheduler; ///< Device scheduler.
-
- const VkFormat format; ///< Vulkan format.
- const VkImageAspectFlags aspect_mask; ///< Vulkan aspect mask.
- const u32 image_num_layers; ///< Number of layers.
- const u32 image_num_levels; ///< Number of mipmap levels.
-
- vk::Image image; ///< Image handle.
- vk::ImageView present_view; ///< Image view compatible with presentation.
-
- std::vector<VkImageMemoryBarrier> barriers; ///< Pool of barriers.
- std::vector<SubrangeState> subrange_states; ///< Current subrange state.
-
- bool state_diverged = false; ///< True when subresources mismatch in layout.
-};
-
-} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_memory_manager.cpp b/src/video_core/renderer_vulkan/vk_memory_manager.cpp
index be53d450f..56b24b70f 100644
--- a/src/video_core/renderer_vulkan/vk_memory_manager.cpp
+++ b/src/video_core/renderer_vulkan/vk_memory_manager.cpp
@@ -216,7 +216,7 @@ VKMemoryCommitImpl::~VKMemoryCommitImpl() {
}
MemoryMap VKMemoryCommitImpl::Map(u64 size, u64 offset_) const {
- return MemoryMap{this, memory.Map(interval.first + offset_, size)};
+ return MemoryMap(this, std::span<u8>(memory.Map(interval.first + offset_, size), size));
}
void VKMemoryCommitImpl::Unmap() const {
diff --git a/src/video_core/renderer_vulkan/vk_memory_manager.h b/src/video_core/renderer_vulkan/vk_memory_manager.h
index 39f903ec8..318f8b43e 100644
--- a/src/video_core/renderer_vulkan/vk_memory_manager.h
+++ b/src/video_core/renderer_vulkan/vk_memory_manager.h
@@ -5,6 +5,7 @@
#pragma once
#include <memory>
+#include <span>
#include <utility>
#include <vector>
#include "common/common_types.h"
@@ -93,8 +94,8 @@ private:
/// Holds ownership of a memory map.
class MemoryMap final {
public:
- explicit MemoryMap(const VKMemoryCommitImpl* commit_, u8* address_)
- : commit{commit_}, address{address_} {}
+ explicit MemoryMap(const VKMemoryCommitImpl* commit_, std::span<u8> span_)
+ : commit{commit_}, span{span_} {}
~MemoryMap() {
if (commit) {
@@ -108,19 +109,24 @@ public:
commit = nullptr;
}
+ /// Returns a span to the memory map.
+ [[nodiscard]] std::span<u8> Span() const noexcept {
+ return span;
+ }
+
/// Returns the address of the memory map.
- u8* GetAddress() const {
- return address;
+ [[nodiscard]] u8* Address() const noexcept {
+ return span.data();
}
/// Returns the address of the memory map;
- operator u8*() const {
- return address;
+ [[nodiscard]] operator u8*() const noexcept {
+ return span.data();
}
private:
const VKMemoryCommitImpl* commit{}; ///< Mapped memory commit.
- u8* address{}; ///< Address to the mapped memory.
+ std::span<u8> span; ///< Address to the mapped memory.
};
} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
index 3fb264d03..083796d05 100644
--- a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
+++ b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
@@ -8,6 +8,7 @@
#include <vector>
#include "common/bit_cast.h"
+#include "common/cityhash.h"
#include "common/microprofile.h"
#include "core/core.h"
#include "core/memory.h"
@@ -22,7 +23,6 @@
#include "video_core/renderer_vulkan/vk_graphics_pipeline.h"
#include "video_core/renderer_vulkan/vk_pipeline_cache.h"
#include "video_core/renderer_vulkan/vk_rasterizer.h"
-#include "video_core/renderer_vulkan/vk_renderpass_cache.h"
#include "video_core/renderer_vulkan/vk_scheduler.h"
#include "video_core/renderer_vulkan/vk_update_descriptor.h"
#include "video_core/renderer_vulkan/wrapper.h"
@@ -52,7 +52,9 @@ constexpr VkDescriptorType STORAGE_TEXEL_BUFFER = VK_DESCRIPTOR_TYPE_STORAGE_TEX
constexpr VkDescriptorType STORAGE_IMAGE = VK_DESCRIPTOR_TYPE_STORAGE_IMAGE;
constexpr VideoCommon::Shader::CompilerSettings compiler_settings{
- VideoCommon::Shader::CompileDepth::FullDecompile};
+ .depth = VideoCommon::Shader::CompileDepth::FullDecompile,
+ .disable_else_derivation = true,
+};
constexpr std::size_t GetStageFromProgram(std::size_t program) {
return program == 0 ? 0 : program - 1;
@@ -149,12 +151,11 @@ VKPipelineCache::VKPipelineCache(RasterizerVulkan& rasterizer_, Tegra::GPU& gpu_
Tegra::Engines::KeplerCompute& kepler_compute_,
Tegra::MemoryManager& gpu_memory_, const VKDevice& device_,
VKScheduler& scheduler_, VKDescriptorPool& descriptor_pool_,
- VKUpdateDescriptorQueue& update_descriptor_queue_,
- VKRenderPassCache& renderpass_cache_)
- : ShaderCache{rasterizer_}, gpu{gpu_}, maxwell3d{maxwell3d_}, kepler_compute{kepler_compute_},
- gpu_memory{gpu_memory_}, device{device_}, scheduler{scheduler_},
- descriptor_pool{descriptor_pool_}, update_descriptor_queue{update_descriptor_queue_},
- renderpass_cache{renderpass_cache_} {}
+ VKUpdateDescriptorQueue& update_descriptor_queue_)
+ : VideoCommon::ShaderCache<Shader>{rasterizer_}, gpu{gpu_}, maxwell3d{maxwell3d_},
+ kepler_compute{kepler_compute_}, gpu_memory{gpu_memory_}, device{device_},
+ scheduler{scheduler_}, descriptor_pool{descriptor_pool_}, update_descriptor_queue{
+ update_descriptor_queue_} {}
VKPipelineCache::~VKPipelineCache() = default;
@@ -199,7 +200,8 @@ std::array<Shader*, Maxwell::MaxShaderProgram> VKPipelineCache::GetShaders() {
}
VKGraphicsPipeline* VKPipelineCache::GetGraphicsPipeline(
- const GraphicsPipelineCacheKey& key, VideoCommon::Shader::AsyncShaders& async_shaders) {
+ const GraphicsPipelineCacheKey& key, u32 num_color_buffers,
+ VideoCommon::Shader::AsyncShaders& async_shaders) {
MICROPROFILE_SCOPE(Vulkan_PipelineCache);
if (last_graphics_pipeline && last_graphics_key == key) {
@@ -215,8 +217,8 @@ VKGraphicsPipeline* VKPipelineCache::GetGraphicsPipeline(
LOG_INFO(Render_Vulkan, "Compile 0x{:016X}", key.Hash());
const auto [program, bindings] = DecompileShaders(key.fixed_state);
async_shaders.QueueVulkanShader(this, device, scheduler, descriptor_pool,
- update_descriptor_queue, renderpass_cache, bindings,
- program, key);
+ update_descriptor_queue, bindings, program, key,
+ num_color_buffers);
}
last_graphics_pipeline = pair->second.get();
return last_graphics_pipeline;
@@ -229,8 +231,8 @@ VKGraphicsPipeline* VKPipelineCache::GetGraphicsPipeline(
LOG_INFO(Render_Vulkan, "Compile 0x{:016X}", key.Hash());
const auto [program, bindings] = DecompileShaders(key.fixed_state);
entry = std::make_unique<VKGraphicsPipeline>(device, scheduler, descriptor_pool,
- update_descriptor_queue, renderpass_cache, key,
- bindings, program);
+ update_descriptor_queue, key, bindings,
+ program, num_color_buffers);
gpu.ShaderNotify().MarkShaderComplete();
}
last_graphics_pipeline = entry.get();
diff --git a/src/video_core/renderer_vulkan/vk_pipeline_cache.h b/src/video_core/renderer_vulkan/vk_pipeline_cache.h
index 9e1f8fcbb..fbaa8257c 100644
--- a/src/video_core/renderer_vulkan/vk_pipeline_cache.h
+++ b/src/video_core/renderer_vulkan/vk_pipeline_cache.h
@@ -19,7 +19,6 @@
#include "video_core/engines/maxwell_3d.h"
#include "video_core/renderer_vulkan/fixed_pipeline_state.h"
#include "video_core/renderer_vulkan/vk_graphics_pipeline.h"
-#include "video_core/renderer_vulkan/vk_renderpass_cache.h"
#include "video_core/renderer_vulkan/vk_shader_decompiler.h"
#include "video_core/renderer_vulkan/wrapper.h"
#include "video_core/shader/async_shaders.h"
@@ -119,18 +118,18 @@ private:
class VKPipelineCache final : public VideoCommon::ShaderCache<Shader> {
public:
- explicit VKPipelineCache(RasterizerVulkan& rasterizer_, Tegra::GPU& gpu_,
- Tegra::Engines::Maxwell3D& maxwell3d_,
- Tegra::Engines::KeplerCompute& kepler_compute_,
- Tegra::MemoryManager& gpu_memory_, const VKDevice& device_,
- VKScheduler& scheduler_, VKDescriptorPool& descriptor_pool_,
- VKUpdateDescriptorQueue& update_descriptor_queue_,
- VKRenderPassCache& renderpass_cache_);
+ explicit VKPipelineCache(RasterizerVulkan& rasterizer, Tegra::GPU& gpu,
+ Tegra::Engines::Maxwell3D& maxwell3d,
+ Tegra::Engines::KeplerCompute& kepler_compute,
+ Tegra::MemoryManager& gpu_memory, const VKDevice& device,
+ VKScheduler& scheduler, VKDescriptorPool& descriptor_pool,
+ VKUpdateDescriptorQueue& update_descriptor_queue);
~VKPipelineCache() override;
std::array<Shader*, Maxwell::MaxShaderProgram> GetShaders();
VKGraphicsPipeline* GetGraphicsPipeline(const GraphicsPipelineCacheKey& key,
+ u32 num_color_buffers,
VideoCommon::Shader::AsyncShaders& async_shaders);
VKComputePipeline& GetComputePipeline(const ComputePipelineCacheKey& key);
@@ -153,7 +152,6 @@ private:
VKScheduler& scheduler;
VKDescriptorPool& descriptor_pool;
VKUpdateDescriptorQueue& update_descriptor_queue;
- VKRenderPassCache& renderpass_cache;
std::unique_ptr<Shader> null_shader;
std::unique_ptr<Shader> null_kernel;
diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp
index f93986aab..04c5c859c 100644
--- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp
+++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp
@@ -19,6 +19,7 @@
#include "core/settings.h"
#include "video_core/engines/kepler_compute.h"
#include "video_core/engines/maxwell_3d.h"
+#include "video_core/renderer_vulkan/blit_image.h"
#include "video_core/renderer_vulkan/fixed_pipeline_state.h"
#include "video_core/renderer_vulkan/maxwell_to_vk.h"
#include "video_core/renderer_vulkan/renderer_vulkan.h"
@@ -30,8 +31,6 @@
#include "video_core/renderer_vulkan/vk_graphics_pipeline.h"
#include "video_core/renderer_vulkan/vk_pipeline_cache.h"
#include "video_core/renderer_vulkan/vk_rasterizer.h"
-#include "video_core/renderer_vulkan/vk_renderpass_cache.h"
-#include "video_core/renderer_vulkan/vk_sampler_cache.h"
#include "video_core/renderer_vulkan/vk_scheduler.h"
#include "video_core/renderer_vulkan/vk_staging_buffer_pool.h"
#include "video_core/renderer_vulkan/vk_state_tracker.h"
@@ -39,10 +38,13 @@
#include "video_core/renderer_vulkan/vk_update_descriptor.h"
#include "video_core/renderer_vulkan/wrapper.h"
#include "video_core/shader_cache.h"
+#include "video_core/texture_cache/texture_cache.h"
namespace Vulkan {
using Maxwell = Tegra::Engines::Maxwell3D::Regs;
+using VideoCommon::ImageViewId;
+using VideoCommon::ImageViewType;
MICROPROFILE_DEFINE(Vulkan_WaitForWorker, "Vulkan", "Wait for worker", MP_RGB(255, 192, 192));
MICROPROFILE_DEFINE(Vulkan_Drawing, "Vulkan", "Record drawing", MP_RGB(192, 128, 128));
@@ -58,9 +60,9 @@ MICROPROFILE_DEFINE(Vulkan_PipelineCache, "Vulkan", "Pipeline cache", MP_RGB(192
namespace {
-constexpr auto ComputeShaderIndex = static_cast<std::size_t>(Tegra::Engines::ShaderType::Compute);
+constexpr auto COMPUTE_SHADER_INDEX = static_cast<size_t>(Tegra::Engines::ShaderType::Compute);
-VkViewport GetViewportState(const VKDevice& device, const Maxwell& regs, std::size_t index) {
+VkViewport GetViewportState(const VKDevice& device, const Maxwell& regs, size_t index) {
const auto& src = regs.viewport_transform[index];
const float width = src.scale_x * 2.0f;
const float height = src.scale_y * 2.0f;
@@ -83,7 +85,7 @@ VkViewport GetViewportState(const VKDevice& device, const Maxwell& regs, std::si
return viewport;
}
-VkRect2D GetScissorState(const Maxwell& regs, std::size_t index) {
+VkRect2D GetScissorState(const Maxwell& regs, size_t index) {
const auto& src = regs.scissor_test[index];
VkRect2D scissor;
if (src.enable) {
@@ -103,98 +105,122 @@ VkRect2D GetScissorState(const Maxwell& regs, std::size_t index) {
std::array<GPUVAddr, Maxwell::MaxShaderProgram> GetShaderAddresses(
const std::array<Shader*, Maxwell::MaxShaderProgram>& shaders) {
std::array<GPUVAddr, Maxwell::MaxShaderProgram> addresses;
- for (std::size_t i = 0; i < std::size(addresses); ++i) {
+ for (size_t i = 0; i < std::size(addresses); ++i) {
addresses[i] = shaders[i] ? shaders[i]->GetGpuAddr() : 0;
}
return addresses;
}
-void TransitionImages(const std::vector<ImageView>& views, VkPipelineStageFlags pipeline_stage,
- VkAccessFlags access) {
- for (auto& [view, layout] : views) {
- view->Transition(*layout, pipeline_stage, access);
+struct TextureHandle {
+ constexpr TextureHandle(u32 data, bool via_header_index) {
+ const Tegra::Texture::TextureHandle handle{data};
+ image = handle.tic_id;
+ sampler = via_header_index ? image : handle.tsc_id.Value();
}
-}
+
+ u32 image;
+ u32 sampler;
+};
template <typename Engine, typename Entry>
-Tegra::Texture::FullTextureInfo GetTextureInfo(const Engine& engine, const Entry& entry,
- std::size_t stage, std::size_t index = 0) {
- const auto stage_type = static_cast<Tegra::Engines::ShaderType>(stage);
+TextureHandle GetTextureInfo(const Engine& engine, bool via_header_index, const Entry& entry,
+ size_t stage, size_t index = 0) {
+ const auto shader_type = static_cast<Tegra::Engines::ShaderType>(stage);
if constexpr (std::is_same_v<Entry, SamplerEntry>) {
if (entry.is_separated) {
const u32 buffer_1 = entry.buffer;
const u32 buffer_2 = entry.secondary_buffer;
const u32 offset_1 = entry.offset;
const u32 offset_2 = entry.secondary_offset;
- const u32 handle_1 = engine.AccessConstBuffer32(stage_type, buffer_1, offset_1);
- const u32 handle_2 = engine.AccessConstBuffer32(stage_type, buffer_2, offset_2);
- return engine.GetTextureInfo(Tegra::Texture::TextureHandle{handle_1 | handle_2});
+ const u32 handle_1 = engine.AccessConstBuffer32(shader_type, buffer_1, offset_1);
+ const u32 handle_2 = engine.AccessConstBuffer32(shader_type, buffer_2, offset_2);
+ return TextureHandle(handle_1 | handle_2, via_header_index);
}
}
if (entry.is_bindless) {
- const auto tex_handle = engine.AccessConstBuffer32(stage_type, entry.buffer, entry.offset);
- return engine.GetTextureInfo(Tegra::Texture::TextureHandle{tex_handle});
- }
- const auto& gpu_profile = engine.AccessGuestDriverProfile();
- const u32 entry_offset = static_cast<u32>(index * gpu_profile.GetTextureHandlerSize());
- const u32 offset = entry.offset + entry_offset;
- if constexpr (std::is_same_v<Engine, Tegra::Engines::Maxwell3D>) {
- return engine.GetStageTexture(stage_type, offset);
- } else {
- return engine.GetTexture(offset);
- }
-}
-
-/// @brief Determine if an attachment to be updated has to preserve contents
-/// @param is_clear True when a clear is being executed
-/// @param regs 3D registers
-/// @return True when the contents have to be preserved
-bool HasToPreserveColorContents(bool is_clear, const Maxwell& regs) {
- if (!is_clear) {
- return true;
- }
- // First we have to make sure all clear masks are enabled.
- if (!regs.clear_buffers.R || !regs.clear_buffers.G || !regs.clear_buffers.B ||
- !regs.clear_buffers.A) {
- return true;
- }
- // If scissors are disabled, the whole screen is cleared
- if (!regs.clear_flags.scissor) {
- return false;
+ const u32 raw = engine.AccessConstBuffer32(shader_type, entry.buffer, entry.offset);
+ return TextureHandle(raw, via_header_index);
}
- // Then we have to confirm scissor testing clears the whole image
- const std::size_t index = regs.clear_buffers.RT;
- const auto& scissor = regs.scissor_test[0];
- return scissor.min_x > 0 || scissor.min_y > 0 || scissor.max_x < regs.rt[index].width ||
- scissor.max_y < regs.rt[index].height;
+ const u32 buffer = engine.GetBoundBuffer();
+ const u64 offset = (entry.offset + index) * sizeof(u32);
+ return TextureHandle(engine.AccessConstBuffer32(shader_type, buffer, offset), via_header_index);
}
-/// @brief Determine if an attachment to be updated has to preserve contents
-/// @param is_clear True when a clear is being executed
-/// @param regs 3D registers
-/// @return True when the contents have to be preserved
-bool HasToPreserveDepthContents(bool is_clear, const Maxwell& regs) {
- // If we are not clearing, the contents have to be preserved
- if (!is_clear) {
- return true;
- }
- // For depth stencil clears we only have to confirm scissor test covers the whole image
- if (!regs.clear_flags.scissor) {
- return false;
- }
- // Make sure the clear cover the whole image
- const auto& scissor = regs.scissor_test[0];
- return scissor.min_x > 0 || scissor.min_y > 0 || scissor.max_x < regs.zeta_width ||
- scissor.max_y < regs.zeta_height;
-}
-
-template <std::size_t N>
+template <size_t N>
std::array<VkDeviceSize, N> ExpandStrides(const std::array<u16, N>& strides) {
std::array<VkDeviceSize, N> expanded;
std::copy(strides.begin(), strides.end(), expanded.begin());
return expanded;
}
+ImageViewType ImageViewTypeFromEntry(const SamplerEntry& entry) {
+ if (entry.is_buffer) {
+ return ImageViewType::e2D;
+ }
+ switch (entry.type) {
+ case Tegra::Shader::TextureType::Texture1D:
+ return entry.is_array ? ImageViewType::e1DArray : ImageViewType::e1D;
+ case Tegra::Shader::TextureType::Texture2D:
+ return entry.is_array ? ImageViewType::e2DArray : ImageViewType::e2D;
+ case Tegra::Shader::TextureType::Texture3D:
+ return ImageViewType::e3D;
+ case Tegra::Shader::TextureType::TextureCube:
+ return entry.is_array ? ImageViewType::CubeArray : ImageViewType::Cube;
+ }
+ UNREACHABLE();
+ return ImageViewType::e2D;
+}
+
+ImageViewType ImageViewTypeFromEntry(const ImageEntry& entry) {
+ switch (entry.type) {
+ case Tegra::Shader::ImageType::Texture1D:
+ return ImageViewType::e1D;
+ case Tegra::Shader::ImageType::Texture1DArray:
+ return ImageViewType::e1DArray;
+ case Tegra::Shader::ImageType::Texture2D:
+ return ImageViewType::e2D;
+ case Tegra::Shader::ImageType::Texture2DArray:
+ return ImageViewType::e2DArray;
+ case Tegra::Shader::ImageType::Texture3D:
+ return ImageViewType::e3D;
+ case Tegra::Shader::ImageType::TextureBuffer:
+ return ImageViewType::Buffer;
+ }
+ UNREACHABLE();
+ return ImageViewType::e2D;
+}
+
+void PushImageDescriptors(const ShaderEntries& entries, TextureCache& texture_cache,
+ VKUpdateDescriptorQueue& update_descriptor_queue,
+ ImageViewId*& image_view_id_ptr, VkSampler*& sampler_ptr) {
+ for ([[maybe_unused]] const auto& entry : entries.uniform_texels) {
+ const ImageViewId image_view_id = *image_view_id_ptr++;
+ const ImageView& image_view = texture_cache.GetImageView(image_view_id);
+ update_descriptor_queue.AddTexelBuffer(image_view.BufferView());
+ }
+ for (const auto& entry : entries.samplers) {
+ for (size_t i = 0; i < entry.size; ++i) {
+ const VkSampler sampler = *sampler_ptr++;
+ const ImageViewId image_view_id = *image_view_id_ptr++;
+ const ImageView& image_view = texture_cache.GetImageView(image_view_id);
+ const VkImageView handle = image_view.Handle(ImageViewTypeFromEntry(entry));
+ update_descriptor_queue.AddSampledImage(handle, sampler);
+ }
+ }
+ for ([[maybe_unused]] const auto& entry : entries.storage_texels) {
+ const ImageViewId image_view_id = *image_view_id_ptr++;
+ const ImageView& image_view = texture_cache.GetImageView(image_view_id);
+ update_descriptor_queue.AddTexelBuffer(image_view.BufferView());
+ }
+ for (const auto& entry : entries.images) {
+ // TODO: Mark as modified
+ const ImageViewId image_view_id = *image_view_id_ptr++;
+ const ImageView& image_view = texture_cache.GetImageView(image_view_id);
+ const VkImageView handle = image_view.Handle(ImageViewTypeFromEntry(entry));
+ update_descriptor_queue.AddImage(handle);
+ }
+}
+
} // Anonymous namespace
class BufferBindings final {
@@ -290,7 +316,7 @@ public:
private:
// Some of these fields are intentionally left uninitialized to avoid initializing them twice.
struct {
- std::size_t num_buffers = 0;
+ size_t num_buffers = 0;
std::array<VkBuffer, Maxwell::NumVertexArrays> buffers;
std::array<VkDeviceSize, Maxwell::NumVertexArrays> offsets;
std::array<VkDeviceSize, Maxwell::NumVertexArrays> sizes;
@@ -303,7 +329,7 @@ private:
VkIndexType type;
} index;
- template <std::size_t N>
+ template <size_t N>
void BindStatic(const VKDevice& device, VKScheduler& scheduler) const {
if (device.IsExtExtendedDynamicStateSupported()) {
if (index.buffer) {
@@ -320,7 +346,7 @@ private:
}
}
- template <std::size_t N, bool is_indexed, bool has_extended_dynamic_state>
+ template <size_t N, bool is_indexed, bool has_extended_dynamic_state>
void BindStatic(VKScheduler& scheduler) const {
static_assert(N <= Maxwell::NumVertexArrays);
if constexpr (N == 0) {
@@ -385,20 +411,23 @@ RasterizerVulkan::RasterizerVulkan(Core::Frontend::EmuWindow& emu_window_, Tegra
Core::Memory::Memory& cpu_memory_, VKScreenInfo& screen_info_,
const VKDevice& device_, VKMemoryManager& memory_manager_,
StateTracker& state_tracker_, VKScheduler& scheduler_)
- : RasterizerAccelerated(cpu_memory_), gpu(gpu_), gpu_memory(gpu_memory_),
- maxwell3d(gpu.Maxwell3D()), kepler_compute(gpu.KeplerCompute()), screen_info(screen_info_),
- device(device_), memory_manager(memory_manager_), state_tracker(state_tracker_),
- scheduler(scheduler_), staging_pool(device, memory_manager, scheduler),
- descriptor_pool(device, scheduler_), update_descriptor_queue(device, scheduler),
- renderpass_cache(device),
+ : RasterizerAccelerated{cpu_memory_}, gpu{gpu_},
+ gpu_memory{gpu_memory_}, maxwell3d{gpu.Maxwell3D()}, kepler_compute{gpu.KeplerCompute()},
+ screen_info{screen_info_}, device{device_}, memory_manager{memory_manager_},
+ state_tracker{state_tracker_}, scheduler{scheduler_}, stream_buffer(device, scheduler),
+ staging_pool(device, memory_manager, scheduler), descriptor_pool(device, scheduler),
+ update_descriptor_queue(device, scheduler),
+ blit_image(device, scheduler, state_tracker, descriptor_pool),
quad_array_pass(device, scheduler, descriptor_pool, staging_pool, update_descriptor_queue),
quad_indexed_pass(device, scheduler, descriptor_pool, staging_pool, update_descriptor_queue),
uint8_pass(device, scheduler, descriptor_pool, staging_pool, update_descriptor_queue),
- texture_cache(*this, maxwell3d, gpu_memory, device, memory_manager, scheduler, staging_pool),
+ texture_cache_runtime{device, scheduler, memory_manager, staging_pool, blit_image},
+ texture_cache(texture_cache_runtime, *this, maxwell3d, kepler_compute, gpu_memory),
pipeline_cache(*this, gpu, maxwell3d, kepler_compute, gpu_memory, device, scheduler,
- descriptor_pool, update_descriptor_queue, renderpass_cache),
- buffer_cache(*this, gpu_memory, cpu_memory_, device, memory_manager, scheduler, staging_pool),
- sampler_cache(device), query_cache(*this, maxwell3d, gpu_memory, device, scheduler),
+ descriptor_pool, update_descriptor_queue),
+ buffer_cache(*this, gpu_memory, cpu_memory_, device, memory_manager, scheduler, stream_buffer,
+ staging_pool),
+ query_cache{*this, maxwell3d, gpu_memory, device, scheduler},
fence_manager(*this, gpu, gpu_memory, texture_cache, buffer_cache, query_cache, device,
scheduler),
wfi_event(device.GetLogical().CreateEvent()), async_shaders(emu_window_) {
@@ -427,9 +456,10 @@ void RasterizerVulkan::Draw(bool is_indexed, bool is_instanced) {
const DrawParameters draw_params =
SetupGeometry(key.fixed_state, buffer_bindings, is_indexed, is_instanced);
- update_descriptor_queue.Acquire();
- sampled_views.clear();
- image_views.clear();
+ auto lock = texture_cache.AcquireLock();
+ texture_cache.SynchronizeGraphicsDescriptors();
+
+ texture_cache.UpdateRenderTargets(false);
const auto shaders = pipeline_cache.GetShaders();
key.shaders = GetShaderAddresses(shaders);
@@ -437,30 +467,24 @@ void RasterizerVulkan::Draw(bool is_indexed, bool is_instanced) {
buffer_cache.Unmap();
- const Texceptions texceptions = UpdateAttachments(false);
- SetupImageTransitions(texceptions, color_attachments, zeta_attachment);
-
- key.renderpass_params = GetRenderPassParams(texceptions);
- key.padding = 0;
+ const Framebuffer* const framebuffer = texture_cache.GetFramebuffer();
+ key.renderpass = framebuffer->RenderPass();
- auto* pipeline = pipeline_cache.GetGraphicsPipeline(key, async_shaders);
+ auto* const pipeline =
+ pipeline_cache.GetGraphicsPipeline(key, framebuffer->NumColorBuffers(), async_shaders);
if (pipeline == nullptr || pipeline->GetHandle() == VK_NULL_HANDLE) {
// Async graphics pipeline was not ready.
return;
}
- scheduler.BindGraphicsPipeline(pipeline->GetHandle());
-
- const auto renderpass = pipeline->GetRenderPass();
- const auto [framebuffer, render_area] = ConfigureFramebuffers(renderpass);
- scheduler.RequestRenderpass(renderpass, framebuffer, render_area);
-
- UpdateDynamicStates();
-
buffer_bindings.Bind(device, scheduler);
BeginTransformFeedback();
+ scheduler.RequestRenderpass(framebuffer);
+ scheduler.BindGraphicsPipeline(pipeline->GetHandle());
+ UpdateDynamicStates();
+
const auto pipeline_layout = pipeline->GetLayout();
const auto descriptor_set = pipeline->CommitDescriptorSet();
scheduler.Record([pipeline_layout, descriptor_set, draw_params](vk::CommandBuffer cmdbuf) {
@@ -481,9 +505,6 @@ void RasterizerVulkan::Clear() {
return;
}
- sampled_views.clear();
- image_views.clear();
-
query_cache.UpdateCounters();
const auto& regs = maxwell3d.regs;
@@ -495,20 +516,24 @@ void RasterizerVulkan::Clear() {
return;
}
- [[maybe_unused]] const auto texceptions = UpdateAttachments(true);
- DEBUG_ASSERT(texceptions.none());
- SetupImageTransitions(0, color_attachments, zeta_attachment);
+ auto lock = texture_cache.AcquireLock();
+ texture_cache.UpdateRenderTargets(true);
+ const Framebuffer* const framebuffer = texture_cache.GetFramebuffer();
+ const VkExtent2D render_area = framebuffer->RenderArea();
+ scheduler.RequestRenderpass(framebuffer);
- const VkRenderPass renderpass = renderpass_cache.GetRenderPass(GetRenderPassParams(0));
- const auto [framebuffer, render_area] = ConfigureFramebuffers(renderpass);
- scheduler.RequestRenderpass(renderpass, framebuffer, render_area);
-
- VkClearRect clear_rect;
- clear_rect.baseArrayLayer = regs.clear_buffers.layer;
- clear_rect.layerCount = 1;
- clear_rect.rect = GetScissorState(regs, 0);
- clear_rect.rect.extent.width = std::min(clear_rect.rect.extent.width, render_area.width);
- clear_rect.rect.extent.height = std::min(clear_rect.rect.extent.height, render_area.height);
+ VkClearRect clear_rect{
+ .rect = GetScissorState(regs, 0),
+ .baseArrayLayer = regs.clear_buffers.layer,
+ .layerCount = 1,
+ };
+ if (clear_rect.rect.extent.width == 0 || clear_rect.rect.extent.height == 0) {
+ return;
+ }
+ clear_rect.rect.extent = VkExtent2D{
+ .width = std::min(clear_rect.rect.extent.width, render_area.width),
+ .height = std::min(clear_rect.rect.extent.height, render_area.height),
+ };
if (use_color) {
VkClearValue clear_value;
@@ -549,9 +574,6 @@ void RasterizerVulkan::Clear() {
void RasterizerVulkan::DispatchCompute(GPUVAddr code_addr) {
MICROPROFILE_SCOPE(Vulkan_Compute);
- update_descriptor_queue.Acquire();
- sampled_views.clear();
- image_views.clear();
query_cache.UpdateCounters();
@@ -570,29 +592,43 @@ void RasterizerVulkan::DispatchCompute(GPUVAddr code_addr) {
// Compute dispatches can't be executed inside a renderpass
scheduler.RequestOutsideRenderPassOperationContext();
- buffer_cache.Map(CalculateComputeStreamBufferSize());
+ image_view_indices.clear();
+ sampler_handles.clear();
+
+ auto lock = texture_cache.AcquireLock();
+ texture_cache.SynchronizeComputeDescriptors();
const auto& entries = pipeline.GetEntries();
- SetupComputeConstBuffers(entries);
- SetupComputeGlobalBuffers(entries);
SetupComputeUniformTexels(entries);
SetupComputeTextures(entries);
SetupComputeStorageTexels(entries);
SetupComputeImages(entries);
- buffer_cache.Unmap();
+ const std::span indices_span(image_view_indices.data(), image_view_indices.size());
+ texture_cache.FillComputeImageViews(indices_span, image_view_ids);
- TransitionImages(sampled_views, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
- VK_ACCESS_SHADER_READ_BIT);
- TransitionImages(image_views, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
- VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT);
+ buffer_cache.Map(CalculateComputeStreamBufferSize());
+ update_descriptor_queue.Acquire();
+
+ SetupComputeConstBuffers(entries);
+ SetupComputeGlobalBuffers(entries);
+
+ ImageViewId* image_view_id_ptr = image_view_ids.data();
+ VkSampler* sampler_ptr = sampler_handles.data();
+ PushImageDescriptors(entries, texture_cache, update_descriptor_queue, image_view_id_ptr,
+ sampler_ptr);
+
+ buffer_cache.Unmap();
+
+ const VkPipeline pipeline_handle = pipeline.GetHandle();
+ const VkPipelineLayout pipeline_layout = pipeline.GetLayout();
+ const VkDescriptorSet descriptor_set = pipeline.CommitDescriptorSet();
scheduler.Record([grid_x = launch_desc.grid_dim_x, grid_y = launch_desc.grid_dim_y,
- grid_z = launch_desc.grid_dim_z, pipeline_handle = pipeline.GetHandle(),
- layout = pipeline.GetLayout(),
- descriptor_set = pipeline.CommitDescriptorSet()](vk::CommandBuffer cmdbuf) {
+ grid_z = launch_desc.grid_dim_z, pipeline_handle, pipeline_layout,
+ descriptor_set](vk::CommandBuffer cmdbuf) {
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, pipeline_handle);
- cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_COMPUTE, layout, DESCRIPTOR_SET,
+ cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_COMPUTE, pipeline_layout, DESCRIPTOR_SET,
descriptor_set, {});
cmdbuf.Dispatch(grid_x, grid_y, grid_z);
});
@@ -613,7 +649,10 @@ void RasterizerVulkan::FlushRegion(VAddr addr, u64 size) {
if (addr == 0 || size == 0) {
return;
}
- texture_cache.FlushRegion(addr, size);
+ {
+ auto lock = texture_cache.AcquireLock();
+ texture_cache.DownloadMemory(addr, size);
+ }
buffer_cache.FlushRegion(addr, size);
query_cache.FlushRegion(addr, size);
}
@@ -622,14 +661,18 @@ bool RasterizerVulkan::MustFlushRegion(VAddr addr, u64 size) {
if (!Settings::IsGPULevelHigh()) {
return buffer_cache.MustFlushRegion(addr, size);
}
- return texture_cache.MustFlushRegion(addr, size) || buffer_cache.MustFlushRegion(addr, size);
+ return texture_cache.IsRegionGpuModified(addr, size) ||
+ buffer_cache.MustFlushRegion(addr, size);
}
void RasterizerVulkan::InvalidateRegion(VAddr addr, u64 size) {
if (addr == 0 || size == 0) {
return;
}
- texture_cache.InvalidateRegion(addr, size);
+ {
+ auto lock = texture_cache.AcquireLock();
+ texture_cache.WriteMemory(addr, size);
+ }
pipeline_cache.InvalidateRegion(addr, size);
buffer_cache.InvalidateRegion(addr, size);
query_cache.InvalidateRegion(addr, size);
@@ -639,17 +682,28 @@ void RasterizerVulkan::OnCPUWrite(VAddr addr, u64 size) {
if (addr == 0 || size == 0) {
return;
}
- texture_cache.OnCPUWrite(addr, size);
+ {
+ auto lock = texture_cache.AcquireLock();
+ texture_cache.WriteMemory(addr, size);
+ }
pipeline_cache.OnCPUWrite(addr, size);
buffer_cache.OnCPUWrite(addr, size);
}
void RasterizerVulkan::SyncGuestHost() {
- texture_cache.SyncGuestHost();
buffer_cache.SyncGuestHost();
pipeline_cache.SyncGuestHost();
}
+void RasterizerVulkan::UnmapMemory(VAddr addr, u64 size) {
+ {
+ auto lock = texture_cache.AcquireLock();
+ texture_cache.UnmapMemory(addr, size);
+ }
+ buffer_cache.OnCPUWrite(addr, size);
+ pipeline_cache.OnCPUWrite(addr, size);
+}
+
void RasterizerVulkan::SignalSemaphore(GPUVAddr addr, u32 value) {
if (!gpu.IsAsync()) {
gpu_memory.Write<u32>(addr, value);
@@ -700,6 +754,14 @@ void RasterizerVulkan::WaitForIdle() {
});
}
+void RasterizerVulkan::FragmentBarrier() {
+ // We already put barriers when a render pass finishes
+}
+
+void RasterizerVulkan::TiledCacheBarrier() {
+ // TODO: Implementing tiled barriers requires rewriting a good chunk of the Vulkan backend
+}
+
void RasterizerVulkan::FlushCommands() {
if (draw_counter > 0) {
draw_counter = 0;
@@ -710,14 +772,20 @@ void RasterizerVulkan::FlushCommands() {
void RasterizerVulkan::TickFrame() {
draw_counter = 0;
update_descriptor_queue.TickFrame();
+ fence_manager.TickFrame();
buffer_cache.TickFrame();
staging_pool.TickFrame();
+ {
+ auto lock = texture_cache.AcquireLock();
+ texture_cache.TickFrame();
+ }
}
-bool RasterizerVulkan::AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Regs::Surface& src,
- const Tegra::Engines::Fermi2D::Regs::Surface& dst,
+bool RasterizerVulkan::AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Surface& src,
+ const Tegra::Engines::Fermi2D::Surface& dst,
const Tegra::Engines::Fermi2D::Config& copy_config) {
- texture_cache.DoFermiCopy(src, dst, copy_config);
+ auto lock = texture_cache.AcquireLock();
+ texture_cache.BlitImage(dst, src, copy_config);
return true;
}
@@ -727,20 +795,16 @@ bool RasterizerVulkan::AccelerateDisplay(const Tegra::FramebufferConfig& config,
return false;
}
- const auto surface{texture_cache.TryFindFramebufferSurface(framebuffer_addr)};
- if (!surface) {
+ auto lock = texture_cache.AcquireLock();
+ ImageView* const image_view = texture_cache.TryFindFramebufferImageView(framebuffer_addr);
+ if (!image_view) {
return false;
}
- // Verify that the cached surface is the same size and format as the requested framebuffer
- const auto& params{surface->GetSurfaceParams()};
- ASSERT_MSG(params.width == config.width, "Framebuffer width is different");
- ASSERT_MSG(params.height == config.height, "Framebuffer height is different");
-
- screen_info.image = &surface->GetImage();
- screen_info.width = params.width;
- screen_info.height = params.height;
- screen_info.is_srgb = surface->GetSurfaceParams().srgb_conversion;
+ screen_info.image_view = image_view->Handle(VideoCommon::ImageViewType::e2D);
+ screen_info.width = image_view->size.width;
+ screen_info.height = image_view->size.height;
+ screen_info.is_srgb = VideoCore::Surface::IsPixelFormatSRGB(image_view->format);
return true;
}
@@ -765,103 +829,6 @@ void RasterizerVulkan::FlushWork() {
draw_counter = 0;
}
-RasterizerVulkan::Texceptions RasterizerVulkan::UpdateAttachments(bool is_clear) {
- MICROPROFILE_SCOPE(Vulkan_RenderTargets);
-
- const auto& regs = maxwell3d.regs;
- auto& dirty = maxwell3d.dirty.flags;
- const bool update_rendertargets = dirty[VideoCommon::Dirty::RenderTargets];
- dirty[VideoCommon::Dirty::RenderTargets] = false;
-
- texture_cache.GuardRenderTargets(true);
-
- Texceptions texceptions;
- for (std::size_t rt = 0; rt < Maxwell::NumRenderTargets; ++rt) {
- if (update_rendertargets) {
- const bool preserve_contents = HasToPreserveColorContents(is_clear, regs);
- color_attachments[rt] = texture_cache.GetColorBufferSurface(rt, preserve_contents);
- }
- if (color_attachments[rt] && WalkAttachmentOverlaps(*color_attachments[rt])) {
- texceptions[rt] = true;
- }
- }
-
- if (update_rendertargets) {
- const bool preserve_contents = HasToPreserveDepthContents(is_clear, regs);
- zeta_attachment = texture_cache.GetDepthBufferSurface(preserve_contents);
- }
- if (zeta_attachment && WalkAttachmentOverlaps(*zeta_attachment)) {
- texceptions[ZETA_TEXCEPTION_INDEX] = true;
- }
-
- texture_cache.GuardRenderTargets(false);
-
- return texceptions;
-}
-
-bool RasterizerVulkan::WalkAttachmentOverlaps(const CachedSurfaceView& attachment) {
- bool overlap = false;
- for (auto& [view, layout] : sampled_views) {
- if (!attachment.IsSameSurface(*view)) {
- continue;
- }
- overlap = true;
- *layout = VK_IMAGE_LAYOUT_GENERAL;
- }
- return overlap;
-}
-
-std::tuple<VkFramebuffer, VkExtent2D> RasterizerVulkan::ConfigureFramebuffers(
- VkRenderPass renderpass) {
- FramebufferCacheKey key{
- .renderpass = renderpass,
- .width = std::numeric_limits<u32>::max(),
- .height = std::numeric_limits<u32>::max(),
- .layers = std::numeric_limits<u32>::max(),
- .views = {},
- };
-
- const auto try_push = [&key](const View& view) {
- if (!view) {
- return false;
- }
- key.views.push_back(view->GetAttachment());
- key.width = std::min(key.width, view->GetWidth());
- key.height = std::min(key.height, view->GetHeight());
- key.layers = std::min(key.layers, view->GetNumLayers());
- return true;
- };
-
- const auto& regs = maxwell3d.regs;
- const std::size_t num_attachments = static_cast<std::size_t>(regs.rt_control.count);
- for (std::size_t index = 0; index < num_attachments; ++index) {
- if (try_push(color_attachments[index])) {
- texture_cache.MarkColorBufferInUse(index);
- }
- }
- if (try_push(zeta_attachment)) {
- texture_cache.MarkDepthBufferInUse();
- }
-
- const auto [fbentry, is_cache_miss] = framebuffer_cache.try_emplace(key);
- auto& framebuffer = fbentry->second;
- if (is_cache_miss) {
- framebuffer = device.GetLogical().CreateFramebuffer({
- .sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO,
- .pNext = nullptr,
- .flags = 0,
- .renderPass = key.renderpass,
- .attachmentCount = static_cast<u32>(key.views.size()),
- .pAttachments = key.views.data(),
- .width = key.width,
- .height = key.height,
- .layers = key.layers,
- });
- }
-
- return {*framebuffer, VkExtent2D{key.width, key.height}};
-}
-
RasterizerVulkan::DrawParameters RasterizerVulkan::SetupGeometry(FixedPipelineState& fixed_state,
BufferBindings& buffer_bindings,
bool is_indexed,
@@ -885,50 +852,37 @@ RasterizerVulkan::DrawParameters RasterizerVulkan::SetupGeometry(FixedPipelineSt
void RasterizerVulkan::SetupShaderDescriptors(
const std::array<Shader*, Maxwell::MaxShaderProgram>& shaders) {
- texture_cache.GuardSamplers(true);
-
- for (std::size_t stage = 0; stage < Maxwell::MaxShaderStage; ++stage) {
- // Skip VertexA stage
+ image_view_indices.clear();
+ sampler_handles.clear();
+ for (size_t stage = 0; stage < Maxwell::MaxShaderStage; ++stage) {
Shader* const shader = shaders[stage + 1];
if (!shader) {
continue;
}
const auto& entries = shader->GetEntries();
- SetupGraphicsConstBuffers(entries, stage);
- SetupGraphicsGlobalBuffers(entries, stage);
SetupGraphicsUniformTexels(entries, stage);
SetupGraphicsTextures(entries, stage);
SetupGraphicsStorageTexels(entries, stage);
SetupGraphicsImages(entries, stage);
}
- texture_cache.GuardSamplers(false);
-}
+ const std::span indices_span(image_view_indices.data(), image_view_indices.size());
+ texture_cache.FillGraphicsImageViews(indices_span, image_view_ids);
-void RasterizerVulkan::SetupImageTransitions(Texceptions texceptions, const ColorAttachments& color,
- const ZetaAttachment& zeta) {
- TransitionImages(sampled_views, VK_PIPELINE_STAGE_ALL_GRAPHICS_BIT, VK_ACCESS_SHADER_READ_BIT);
- TransitionImages(image_views, VK_PIPELINE_STAGE_ALL_GRAPHICS_BIT,
- VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT);
+ update_descriptor_queue.Acquire();
- for (std::size_t rt = 0; rt < color.size(); ++rt) {
- const auto color_attachment = color[rt];
- if (color_attachment == nullptr) {
+ ImageViewId* image_view_id_ptr = image_view_ids.data();
+ VkSampler* sampler_ptr = sampler_handles.data();
+ for (size_t stage = 0; stage < Maxwell::MaxShaderStage; ++stage) {
+ // Skip VertexA stage
+ Shader* const shader = shaders[stage + 1];
+ if (!shader) {
continue;
}
- const auto image_layout =
- texceptions[rt] ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL;
- color_attachment->Transition(image_layout, VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT,
- VK_ACCESS_COLOR_ATTACHMENT_READ_BIT |
- VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT);
- }
-
- if (zeta != nullptr) {
- const auto image_layout = texceptions[ZETA_TEXCEPTION_INDEX]
- ? VK_IMAGE_LAYOUT_GENERAL
- : VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL;
- zeta->Transition(image_layout, VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT,
- VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT |
- VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT);
+ const auto& entries = shader->GetEntries();
+ SetupGraphicsConstBuffers(entries, stage);
+ SetupGraphicsGlobalBuffers(entries, stage);
+ PushImageDescriptors(entries, texture_cache, update_descriptor_queue, image_view_id_ptr,
+ sampler_ptr);
}
}
@@ -1000,7 +954,7 @@ void RasterizerVulkan::EndTransformFeedback() {
void RasterizerVulkan::SetupVertexArrays(BufferBindings& buffer_bindings) {
const auto& regs = maxwell3d.regs;
- for (std::size_t index = 0; index < Maxwell::NumVertexArrays; ++index) {
+ for (size_t index = 0; index < Maxwell::NumVertexArrays; ++index) {
const auto& vertex_array = regs.vertex_array[index];
if (!vertex_array.IsEnabled()) {
continue;
@@ -1009,7 +963,7 @@ void RasterizerVulkan::SetupVertexArrays(BufferBindings& buffer_bindings) {
const GPUVAddr end{regs.vertex_array_limit[index].LimitAddress()};
ASSERT(end >= start);
- const std::size_t size = end - start;
+ const size_t size = end - start;
if (size == 0) {
buffer_bindings.AddVertexBinding(DefaultBuffer(), 0, DEFAULT_BUFFER_SIZE, 0);
continue;
@@ -1070,7 +1024,7 @@ void RasterizerVulkan::SetupIndexBuffer(BufferBindings& buffer_bindings, DrawPar
}
}
-void RasterizerVulkan::SetupGraphicsConstBuffers(const ShaderEntries& entries, std::size_t stage) {
+void RasterizerVulkan::SetupGraphicsConstBuffers(const ShaderEntries& entries, size_t stage) {
MICROPROFILE_SCOPE(Vulkan_ConstBuffers);
const auto& shader_stage = maxwell3d.state.shader_stages[stage];
for (const auto& entry : entries.const_buffers) {
@@ -1078,7 +1032,7 @@ void RasterizerVulkan::SetupGraphicsConstBuffers(const ShaderEntries& entries, s
}
}
-void RasterizerVulkan::SetupGraphicsGlobalBuffers(const ShaderEntries& entries, std::size_t stage) {
+void RasterizerVulkan::SetupGraphicsGlobalBuffers(const ShaderEntries& entries, size_t stage) {
MICROPROFILE_SCOPE(Vulkan_GlobalBuffers);
const auto& cbufs{maxwell3d.state.shader_stages[stage]};
@@ -1088,37 +1042,49 @@ void RasterizerVulkan::SetupGraphicsGlobalBuffers(const ShaderEntries& entries,
}
}
-void RasterizerVulkan::SetupGraphicsUniformTexels(const ShaderEntries& entries, std::size_t stage) {
+void RasterizerVulkan::SetupGraphicsUniformTexels(const ShaderEntries& entries, size_t stage) {
MICROPROFILE_SCOPE(Vulkan_Textures);
+ const auto& regs = maxwell3d.regs;
+ const bool via_header_index = regs.sampler_index == Maxwell::SamplerIndex::ViaHeaderIndex;
for (const auto& entry : entries.uniform_texels) {
- const auto image = GetTextureInfo(maxwell3d, entry, stage).tic;
- SetupUniformTexels(image, entry);
+ const TextureHandle handle = GetTextureInfo(maxwell3d, via_header_index, entry, stage);
+ image_view_indices.push_back(handle.image);
}
}
-void RasterizerVulkan::SetupGraphicsTextures(const ShaderEntries& entries, std::size_t stage) {
+void RasterizerVulkan::SetupGraphicsTextures(const ShaderEntries& entries, size_t stage) {
MICROPROFILE_SCOPE(Vulkan_Textures);
+ const auto& regs = maxwell3d.regs;
+ const bool via_header_index = regs.sampler_index == Maxwell::SamplerIndex::ViaHeaderIndex;
for (const auto& entry : entries.samplers) {
- for (std::size_t i = 0; i < entry.size; ++i) {
- const auto texture = GetTextureInfo(maxwell3d, entry, stage, i);
- SetupTexture(texture, entry);
+ for (size_t index = 0; index < entry.size; ++index) {
+ const TextureHandle handle =
+ GetTextureInfo(maxwell3d, via_header_index, entry, stage, index);
+ image_view_indices.push_back(handle.image);
+
+ Sampler* const sampler = texture_cache.GetGraphicsSampler(handle.sampler);
+ sampler_handles.push_back(sampler->Handle());
}
}
}
-void RasterizerVulkan::SetupGraphicsStorageTexels(const ShaderEntries& entries, std::size_t stage) {
+void RasterizerVulkan::SetupGraphicsStorageTexels(const ShaderEntries& entries, size_t stage) {
MICROPROFILE_SCOPE(Vulkan_Textures);
+ const auto& regs = maxwell3d.regs;
+ const bool via_header_index = regs.sampler_index == Maxwell::SamplerIndex::ViaHeaderIndex;
for (const auto& entry : entries.storage_texels) {
- const auto image = GetTextureInfo(maxwell3d, entry, stage).tic;
- SetupStorageTexel(image, entry);
+ const TextureHandle handle = GetTextureInfo(maxwell3d, via_header_index, entry, stage);
+ image_view_indices.push_back(handle.image);
}
}
-void RasterizerVulkan::SetupGraphicsImages(const ShaderEntries& entries, std::size_t stage) {
+void RasterizerVulkan::SetupGraphicsImages(const ShaderEntries& entries, size_t stage) {
MICROPROFILE_SCOPE(Vulkan_Images);
+ const auto& regs = maxwell3d.regs;
+ const bool via_header_index = regs.sampler_index == Maxwell::SamplerIndex::ViaHeaderIndex;
for (const auto& entry : entries.images) {
- const auto tic = GetTextureInfo(maxwell3d, entry, stage).tic;
- SetupImage(tic, entry);
+ const TextureHandle handle = GetTextureInfo(maxwell3d, via_header_index, entry, stage);
+ image_view_indices.push_back(handle.image);
}
}
@@ -1128,11 +1094,12 @@ void RasterizerVulkan::SetupComputeConstBuffers(const ShaderEntries& entries) {
for (const auto& entry : entries.const_buffers) {
const auto& config = launch_desc.const_buffer_config[entry.GetIndex()];
const std::bitset<8> mask = launch_desc.const_buffer_enable_mask.Value();
- Tegra::Engines::ConstBufferInfo buffer;
- buffer.address = config.Address();
- buffer.size = config.size;
- buffer.enabled = mask[entry.GetIndex()];
- SetupConstBuffer(entry, buffer);
+ const Tegra::Engines::ConstBufferInfo info{
+ .address = config.Address(),
+ .size = config.size,
+ .enabled = mask[entry.GetIndex()],
+ };
+ SetupConstBuffer(entry, info);
}
}
@@ -1147,35 +1114,46 @@ void RasterizerVulkan::SetupComputeGlobalBuffers(const ShaderEntries& entries) {
void RasterizerVulkan::SetupComputeUniformTexels(const ShaderEntries& entries) {
MICROPROFILE_SCOPE(Vulkan_Textures);
+ const bool via_header_index = kepler_compute.launch_description.linked_tsc;
for (const auto& entry : entries.uniform_texels) {
- const auto image = GetTextureInfo(kepler_compute, entry, ComputeShaderIndex).tic;
- SetupUniformTexels(image, entry);
+ const TextureHandle handle =
+ GetTextureInfo(kepler_compute, via_header_index, entry, COMPUTE_SHADER_INDEX);
+ image_view_indices.push_back(handle.image);
}
}
void RasterizerVulkan::SetupComputeTextures(const ShaderEntries& entries) {
MICROPROFILE_SCOPE(Vulkan_Textures);
+ const bool via_header_index = kepler_compute.launch_description.linked_tsc;
for (const auto& entry : entries.samplers) {
- for (std::size_t i = 0; i < entry.size; ++i) {
- const auto texture = GetTextureInfo(kepler_compute, entry, ComputeShaderIndex, i);
- SetupTexture(texture, entry);
+ for (size_t index = 0; index < entry.size; ++index) {
+ const TextureHandle handle = GetTextureInfo(kepler_compute, via_header_index, entry,
+ COMPUTE_SHADER_INDEX, index);
+ image_view_indices.push_back(handle.image);
+
+ Sampler* const sampler = texture_cache.GetComputeSampler(handle.sampler);
+ sampler_handles.push_back(sampler->Handle());
}
}
}
void RasterizerVulkan::SetupComputeStorageTexels(const ShaderEntries& entries) {
MICROPROFILE_SCOPE(Vulkan_Textures);
+ const bool via_header_index = kepler_compute.launch_description.linked_tsc;
for (const auto& entry : entries.storage_texels) {
- const auto image = GetTextureInfo(kepler_compute, entry, ComputeShaderIndex).tic;
- SetupStorageTexel(image, entry);
+ const TextureHandle handle =
+ GetTextureInfo(kepler_compute, via_header_index, entry, COMPUTE_SHADER_INDEX);
+ image_view_indices.push_back(handle.image);
}
}
void RasterizerVulkan::SetupComputeImages(const ShaderEntries& entries) {
MICROPROFILE_SCOPE(Vulkan_Images);
+ const bool via_header_index = kepler_compute.launch_description.linked_tsc;
for (const auto& entry : entries.images) {
- const auto tic = GetTextureInfo(kepler_compute, entry, ComputeShaderIndex).tic;
- SetupImage(tic, entry);
+ const TextureHandle handle =
+ GetTextureInfo(kepler_compute, via_header_index, entry, COMPUTE_SHADER_INDEX);
+ image_view_indices.push_back(handle.image);
}
}
@@ -1186,14 +1164,12 @@ void RasterizerVulkan::SetupConstBuffer(const ConstBufferEntry& entry,
update_descriptor_queue.AddBuffer(DefaultBuffer(), 0, DEFAULT_BUFFER_SIZE);
return;
}
-
// Align the size to avoid bad std140 interactions
- const std::size_t size =
- Common::AlignUp(CalculateConstBufferSize(entry, buffer), 4 * sizeof(float));
+ const size_t size = Common::AlignUp(CalculateConstBufferSize(entry, buffer), 4 * sizeof(float));
ASSERT(size <= MaxConstbufferSize);
- const auto info =
- buffer_cache.UploadMemory(buffer.address, size, device.GetUniformBufferAlignment());
+ const u64 alignment = device.GetUniformBufferAlignment();
+ const auto info = buffer_cache.UploadMemory(buffer.address, size, alignment);
update_descriptor_queue.AddBuffer(info.handle, info.offset, size);
}
@@ -1206,7 +1182,7 @@ void RasterizerVulkan::SetupGlobalBuffer(const GlobalBufferEntry& entry, GPUVAdd
// because Vulkan doesn't like empty buffers.
// Note: Do *not* use DefaultBuffer() here, storage buffers can be written breaking the
// default buffer.
- static constexpr std::size_t dummy_size = 4;
+ static constexpr size_t dummy_size = 4;
const auto info = buffer_cache.GetEmptyBuffer(dummy_size);
update_descriptor_queue.AddBuffer(info.handle, info.offset, dummy_size);
return;
@@ -1217,55 +1193,6 @@ void RasterizerVulkan::SetupGlobalBuffer(const GlobalBufferEntry& entry, GPUVAdd
update_descriptor_queue.AddBuffer(info.handle, info.offset, size);
}
-void RasterizerVulkan::SetupUniformTexels(const Tegra::Texture::TICEntry& tic,
- const UniformTexelEntry& entry) {
- const auto view = texture_cache.GetTextureSurface(tic, entry);
- ASSERT(view->IsBufferView());
-
- update_descriptor_queue.AddTexelBuffer(view->GetBufferView());
-}
-
-void RasterizerVulkan::SetupTexture(const Tegra::Texture::FullTextureInfo& texture,
- const SamplerEntry& entry) {
- auto view = texture_cache.GetTextureSurface(texture.tic, entry);
- ASSERT(!view->IsBufferView());
-
- const VkImageView image_view = view->GetImageView(texture.tic.x_source, texture.tic.y_source,
- texture.tic.z_source, texture.tic.w_source);
- const auto sampler = sampler_cache.GetSampler(texture.tsc);
- update_descriptor_queue.AddSampledImage(sampler, image_view);
-
- VkImageLayout* const image_layout = update_descriptor_queue.LastImageLayout();
- *image_layout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL;
- sampled_views.push_back(ImageView{std::move(view), image_layout});
-}
-
-void RasterizerVulkan::SetupStorageTexel(const Tegra::Texture::TICEntry& tic,
- const StorageTexelEntry& entry) {
- const auto view = texture_cache.GetImageSurface(tic, entry);
- ASSERT(view->IsBufferView());
-
- update_descriptor_queue.AddTexelBuffer(view->GetBufferView());
-}
-
-void RasterizerVulkan::SetupImage(const Tegra::Texture::TICEntry& tic, const ImageEntry& entry) {
- auto view = texture_cache.GetImageSurface(tic, entry);
-
- if (entry.is_written) {
- view->MarkAsModified(texture_cache.Tick());
- }
-
- UNIMPLEMENTED_IF(tic.IsBuffer());
-
- const VkImageView image_view =
- view->GetImageView(tic.x_source, tic.y_source, tic.z_source, tic.w_source);
- update_descriptor_queue.AddImage(image_view);
-
- VkImageLayout* const image_layout = update_descriptor_queue.LastImageLayout();
- *image_layout = VK_IMAGE_LAYOUT_GENERAL;
- image_views.push_back(ImageView{std::move(view), image_layout});
-}
-
void RasterizerVulkan::UpdateViewportsState(Tegra::Engines::Maxwell3D::Regs& regs) {
if (!state_tracker.TouchViewports()) {
return;
@@ -1457,8 +1384,8 @@ void RasterizerVulkan::UpdateStencilTestEnable(Tegra::Engines::Maxwell3D::Regs&
});
}
-std::size_t RasterizerVulkan::CalculateGraphicsStreamBufferSize(bool is_indexed) const {
- std::size_t size = CalculateVertexArraysSize();
+size_t RasterizerVulkan::CalculateGraphicsStreamBufferSize(bool is_indexed) const {
+ size_t size = CalculateVertexArraysSize();
if (is_indexed) {
size = Common::AlignUp(size, 4) + CalculateIndexBufferSize();
}
@@ -1466,15 +1393,15 @@ std::size_t RasterizerVulkan::CalculateGraphicsStreamBufferSize(bool is_indexed)
return size;
}
-std::size_t RasterizerVulkan::CalculateComputeStreamBufferSize() const {
+size_t RasterizerVulkan::CalculateComputeStreamBufferSize() const {
return Tegra::Engines::KeplerCompute::NumConstBuffers *
(Maxwell::MaxConstBufferSize + device.GetUniformBufferAlignment());
}
-std::size_t RasterizerVulkan::CalculateVertexArraysSize() const {
+size_t RasterizerVulkan::CalculateVertexArraysSize() const {
const auto& regs = maxwell3d.regs;
- std::size_t size = 0;
+ size_t size = 0;
for (u32 index = 0; index < Maxwell::NumVertexArrays; ++index) {
// This implementation assumes that all attributes are used in the shader.
const GPUVAddr start{regs.vertex_array[index].StartAddress()};
@@ -1486,12 +1413,12 @@ std::size_t RasterizerVulkan::CalculateVertexArraysSize() const {
return size;
}
-std::size_t RasterizerVulkan::CalculateIndexBufferSize() const {
- return static_cast<std::size_t>(maxwell3d.regs.index_array.count) *
- static_cast<std::size_t>(maxwell3d.regs.index_array.FormatSizeInBytes());
+size_t RasterizerVulkan::CalculateIndexBufferSize() const {
+ return static_cast<size_t>(maxwell3d.regs.index_array.count) *
+ static_cast<size_t>(maxwell3d.regs.index_array.FormatSizeInBytes());
}
-std::size_t RasterizerVulkan::CalculateConstBufferSize(
+size_t RasterizerVulkan::CalculateConstBufferSize(
const ConstBufferEntry& entry, const Tegra::Engines::ConstBufferInfo& buffer) const {
if (entry.IsIndirect()) {
// Buffer is accessed indirectly, so upload the entire thing
@@ -1502,37 +1429,10 @@ std::size_t RasterizerVulkan::CalculateConstBufferSize(
}
}
-RenderPassParams RasterizerVulkan::GetRenderPassParams(Texceptions texceptions) const {
- const auto& regs = maxwell3d.regs;
- const std::size_t num_attachments = static_cast<std::size_t>(regs.rt_control.count);
-
- RenderPassParams params;
- params.color_formats = {};
- std::size_t color_texceptions = 0;
-
- std::size_t index = 0;
- for (std::size_t rt = 0; rt < num_attachments; ++rt) {
- const auto& rendertarget = regs.rt[rt];
- if (rendertarget.Address() == 0 || rendertarget.format == Tegra::RenderTargetFormat::NONE) {
- continue;
- }
- params.color_formats[index] = static_cast<u8>(rendertarget.format);
- color_texceptions |= (texceptions[rt] ? 1ULL : 0ULL) << index;
- ++index;
- }
- params.num_color_attachments = static_cast<u8>(index);
- params.texceptions = static_cast<u8>(color_texceptions);
-
- params.zeta_format = regs.zeta_enable ? static_cast<u8>(regs.zeta.format) : 0;
- params.zeta_texception = texceptions[ZETA_TEXCEPTION_INDEX];
- return params;
-}
-
VkBuffer RasterizerVulkan::DefaultBuffer() {
if (default_buffer) {
return *default_buffer;
}
-
default_buffer = device.GetLogical().CreateBuffer({
.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
.pNext = nullptr,
diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.h b/src/video_core/renderer_vulkan/vk_rasterizer.h
index 30ec58eb4..990f9e031 100644
--- a/src/video_core/renderer_vulkan/vk_rasterizer.h
+++ b/src/video_core/renderer_vulkan/vk_rasterizer.h
@@ -11,11 +11,11 @@
#include <vector>
#include <boost/container/static_vector.hpp>
-#include <boost/functional/hash.hpp>
#include "common/common_types.h"
#include "video_core/rasterizer_accelerated.h"
#include "video_core/rasterizer_interface.h"
+#include "video_core/renderer_vulkan/blit_image.h"
#include "video_core/renderer_vulkan/fixed_pipeline_state.h"
#include "video_core/renderer_vulkan/vk_buffer_cache.h"
#include "video_core/renderer_vulkan/vk_compute_pass.h"
@@ -24,10 +24,9 @@
#include "video_core/renderer_vulkan/vk_memory_manager.h"
#include "video_core/renderer_vulkan/vk_pipeline_cache.h"
#include "video_core/renderer_vulkan/vk_query_cache.h"
-#include "video_core/renderer_vulkan/vk_renderpass_cache.h"
-#include "video_core/renderer_vulkan/vk_sampler_cache.h"
#include "video_core/renderer_vulkan/vk_scheduler.h"
#include "video_core/renderer_vulkan/vk_staging_buffer_pool.h"
+#include "video_core/renderer_vulkan/vk_stream_buffer.h"
#include "video_core/renderer_vulkan/vk_texture_cache.h"
#include "video_core/renderer_vulkan/vk_update_descriptor.h"
#include "video_core/renderer_vulkan/wrapper.h"
@@ -49,60 +48,9 @@ namespace Vulkan {
struct VKScreenInfo;
-using ImageViewsPack = boost::container::static_vector<VkImageView, Maxwell::NumRenderTargets + 1>;
-
-struct FramebufferCacheKey {
- VkRenderPass renderpass{};
- u32 width = 0;
- u32 height = 0;
- u32 layers = 0;
- ImageViewsPack views;
-
- std::size_t Hash() const noexcept {
- std::size_t hash = 0;
- boost::hash_combine(hash, static_cast<VkRenderPass>(renderpass));
- for (const auto& view : views) {
- boost::hash_combine(hash, static_cast<VkImageView>(view));
- }
- boost::hash_combine(hash, width);
- boost::hash_combine(hash, height);
- boost::hash_combine(hash, layers);
- return hash;
- }
-
- bool operator==(const FramebufferCacheKey& rhs) const noexcept {
- return std::tie(renderpass, views, width, height, layers) ==
- std::tie(rhs.renderpass, rhs.views, rhs.width, rhs.height, rhs.layers);
- }
-
- bool operator!=(const FramebufferCacheKey& rhs) const noexcept {
- return !operator==(rhs);
- }
-};
-
-} // namespace Vulkan
-
-namespace std {
-
-template <>
-struct hash<Vulkan::FramebufferCacheKey> {
- std::size_t operator()(const Vulkan::FramebufferCacheKey& k) const noexcept {
- return k.Hash();
- }
-};
-
-} // namespace std
-
-namespace Vulkan {
-
class StateTracker;
class BufferBindings;
-struct ImageView {
- View view;
- VkImageLayout* layout = nullptr;
-};
-
class RasterizerVulkan final : public VideoCore::RasterizerAccelerated {
public:
explicit RasterizerVulkan(Core::Frontend::EmuWindow& emu_window_, Tegra::GPU& gpu_,
@@ -123,15 +71,18 @@ public:
void InvalidateRegion(VAddr addr, u64 size) override;
void OnCPUWrite(VAddr addr, u64 size) override;
void SyncGuestHost() override;
+ void UnmapMemory(VAddr addr, u64 size) override;
void SignalSemaphore(GPUVAddr addr, u32 value) override;
void SignalSyncPoint(u32 value) override;
void ReleaseFences() override;
void FlushAndInvalidateRegion(VAddr addr, u64 size) override;
void WaitForIdle() override;
+ void FragmentBarrier() override;
+ void TiledCacheBarrier() override;
void FlushCommands() override;
void TickFrame() override;
- bool AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Regs::Surface& src,
- const Tegra::Engines::Fermi2D::Regs::Surface& dst,
+ bool AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Surface& src,
+ const Tegra::Engines::Fermi2D::Surface& dst,
const Tegra::Engines::Fermi2D::Config& copy_config) override;
bool AccelerateDisplay(const Tegra::FramebufferConfig& config, VAddr framebuffer_addr,
u32 pixel_stride) override;
@@ -145,11 +96,17 @@ public:
}
/// Maximum supported size that a constbuffer can have in bytes.
- static constexpr std::size_t MaxConstbufferSize = 0x10000;
+ static constexpr size_t MaxConstbufferSize = 0x10000;
static_assert(MaxConstbufferSize % (4 * sizeof(float)) == 0,
"The maximum size of a constbuffer must be a multiple of the size of GLvec4");
private:
+ static constexpr size_t MAX_TEXTURES = 192;
+ static constexpr size_t MAX_IMAGES = 48;
+ static constexpr size_t MAX_IMAGE_VIEWS = MAX_TEXTURES + MAX_IMAGES;
+
+ static constexpr VkDeviceSize DEFAULT_BUFFER_SIZE = 4 * sizeof(float);
+
struct DrawParameters {
void Draw(vk::CommandBuffer cmdbuf) const;
@@ -160,23 +117,8 @@ private:
bool is_indexed = 0;
};
- using ColorAttachments = std::array<View, Maxwell::NumRenderTargets>;
- using ZetaAttachment = View;
-
- using Texceptions = std::bitset<Maxwell::NumRenderTargets + 1>;
-
- static constexpr std::size_t ZETA_TEXCEPTION_INDEX = 8;
- static constexpr VkDeviceSize DEFAULT_BUFFER_SIZE = 4 * sizeof(float);
-
void FlushWork();
- /// @brief Updates the currently bound attachments
- /// @param is_clear True when the framebuffer is updated as a clear
- /// @return Bitfield of attachments being used as sampled textures
- Texceptions UpdateAttachments(bool is_clear);
-
- std::tuple<VkFramebuffer, VkExtent2D> ConfigureFramebuffers(VkRenderPass renderpass);
-
/// Setups geometry buffers and state.
DrawParameters SetupGeometry(FixedPipelineState& fixed_state, BufferBindings& buffer_bindings,
bool is_indexed, bool is_instanced);
@@ -184,17 +126,12 @@ private:
/// Setup descriptors in the graphics pipeline.
void SetupShaderDescriptors(const std::array<Shader*, Maxwell::MaxShaderProgram>& shaders);
- void SetupImageTransitions(Texceptions texceptions, const ColorAttachments& color,
- const ZetaAttachment& zeta);
-
void UpdateDynamicStates();
void BeginTransformFeedback();
void EndTransformFeedback();
- bool WalkAttachmentOverlaps(const CachedSurfaceView& attachment);
-
void SetupVertexArrays(BufferBindings& buffer_bindings);
void SetupIndexBuffer(BufferBindings& buffer_bindings, DrawParameters& params, bool is_indexed);
@@ -240,14 +177,6 @@ private:
void SetupGlobalBuffer(const GlobalBufferEntry& entry, GPUVAddr address);
- void SetupUniformTexels(const Tegra::Texture::TICEntry& image, const UniformTexelEntry& entry);
-
- void SetupTexture(const Tegra::Texture::FullTextureInfo& texture, const SamplerEntry& entry);
-
- void SetupStorageTexel(const Tegra::Texture::TICEntry& tic, const StorageTexelEntry& entry);
-
- void SetupImage(const Tegra::Texture::TICEntry& tic, const ImageEntry& entry);
-
void UpdateViewportsState(Tegra::Engines::Maxwell3D::Regs& regs);
void UpdateScissorsState(Tegra::Engines::Maxwell3D::Regs& regs);
void UpdateDepthBias(Tegra::Engines::Maxwell3D::Regs& regs);
@@ -264,18 +193,16 @@ private:
void UpdateStencilOp(Tegra::Engines::Maxwell3D::Regs& regs);
void UpdateStencilTestEnable(Tegra::Engines::Maxwell3D::Regs& regs);
- std::size_t CalculateGraphicsStreamBufferSize(bool is_indexed) const;
-
- std::size_t CalculateComputeStreamBufferSize() const;
+ size_t CalculateGraphicsStreamBufferSize(bool is_indexed) const;
- std::size_t CalculateVertexArraysSize() const;
+ size_t CalculateComputeStreamBufferSize() const;
- std::size_t CalculateIndexBufferSize() const;
+ size_t CalculateVertexArraysSize() const;
- std::size_t CalculateConstBufferSize(const ConstBufferEntry& entry,
- const Tegra::Engines::ConstBufferInfo& buffer) const;
+ size_t CalculateIndexBufferSize() const;
- RenderPassParams GetRenderPassParams(Texceptions texceptions) const;
+ size_t CalculateConstBufferSize(const ConstBufferEntry& entry,
+ const Tegra::Engines::ConstBufferInfo& buffer) const;
VkBuffer DefaultBuffer();
@@ -290,18 +217,19 @@ private:
StateTracker& state_tracker;
VKScheduler& scheduler;
+ VKStreamBuffer stream_buffer;
VKStagingBufferPool staging_pool;
VKDescriptorPool descriptor_pool;
VKUpdateDescriptorQueue update_descriptor_queue;
- VKRenderPassCache renderpass_cache;
+ BlitImageHelper blit_image;
QuadArrayPass quad_array_pass;
QuadIndexedPass quad_indexed_pass;
Uint8Pass uint8_pass;
- VKTextureCache texture_cache;
+ TextureCacheRuntime texture_cache_runtime;
+ TextureCache texture_cache;
VKPipelineCache pipeline_cache;
VKBufferCache buffer_cache;
- VKSamplerCache sampler_cache;
VKQueryCache query_cache;
VKFenceManager fence_manager;
@@ -310,16 +238,11 @@ private:
vk::Event wfi_event;
VideoCommon::Shader::AsyncShaders async_shaders;
- ColorAttachments color_attachments;
- ZetaAttachment zeta_attachment;
-
- std::vector<ImageView> sampled_views;
- std::vector<ImageView> image_views;
+ boost::container::static_vector<u32, MAX_IMAGE_VIEWS> image_view_indices;
+ std::array<VideoCommon::ImageViewId, MAX_IMAGE_VIEWS> image_view_ids;
+ boost::container::static_vector<VkSampler, MAX_TEXTURES> sampler_handles;
u32 draw_counter = 0;
-
- // TODO(Rodrigo): Invalidate on image destruction
- std::unordered_map<FramebufferCacheKey, vk::Framebuffer> framebuffer_cache;
};
} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_renderpass_cache.cpp b/src/video_core/renderer_vulkan/vk_renderpass_cache.cpp
deleted file mode 100644
index e812c7dd6..000000000
--- a/src/video_core/renderer_vulkan/vk_renderpass_cache.cpp
+++ /dev/null
@@ -1,158 +0,0 @@
-// Copyright 2019 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#include <cstring>
-#include <memory>
-#include <vector>
-
-#include "common/cityhash.h"
-#include "video_core/engines/maxwell_3d.h"
-#include "video_core/renderer_vulkan/maxwell_to_vk.h"
-#include "video_core/renderer_vulkan/vk_device.h"
-#include "video_core/renderer_vulkan/vk_renderpass_cache.h"
-#include "video_core/renderer_vulkan/wrapper.h"
-
-namespace Vulkan {
-
-std::size_t RenderPassParams::Hash() const noexcept {
- const u64 hash = Common::CityHash64(reinterpret_cast<const char*>(this), sizeof *this);
- return static_cast<std::size_t>(hash);
-}
-
-bool RenderPassParams::operator==(const RenderPassParams& rhs) const noexcept {
- return std::memcmp(&rhs, this, sizeof *this) == 0;
-}
-
-VKRenderPassCache::VKRenderPassCache(const VKDevice& device_) : device{device_} {}
-
-VKRenderPassCache::~VKRenderPassCache() = default;
-
-VkRenderPass VKRenderPassCache::GetRenderPass(const RenderPassParams& params) {
- const auto [pair, is_cache_miss] = cache.try_emplace(params);
- auto& entry = pair->second;
- if (is_cache_miss) {
- entry = CreateRenderPass(params);
- }
- return *entry;
-}
-
-vk::RenderPass VKRenderPassCache::CreateRenderPass(const RenderPassParams& params) const {
- using namespace VideoCore::Surface;
- const std::size_t num_attachments = static_cast<std::size_t>(params.num_color_attachments);
-
- std::vector<VkAttachmentDescription> descriptors;
- descriptors.reserve(num_attachments);
-
- std::vector<VkAttachmentReference> color_references;
- color_references.reserve(num_attachments);
-
- for (std::size_t rt = 0; rt < num_attachments; ++rt) {
- const auto guest_format = static_cast<Tegra::RenderTargetFormat>(params.color_formats[rt]);
- const PixelFormat pixel_format = PixelFormatFromRenderTargetFormat(guest_format);
- const auto format = MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, pixel_format);
- ASSERT_MSG(format.attachable, "Trying to attach a non-attachable format with format={}",
- static_cast<int>(pixel_format));
-
- // TODO(Rodrigo): Add MAY_ALIAS_BIT when it's needed.
- const VkImageLayout color_layout = ((params.texceptions >> rt) & 1) != 0
- ? VK_IMAGE_LAYOUT_GENERAL
- : VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL;
- descriptors.push_back({
- .flags = VK_ATTACHMENT_DESCRIPTION_MAY_ALIAS_BIT,
- .format = format.format,
- .samples = VK_SAMPLE_COUNT_1_BIT,
- .loadOp = VK_ATTACHMENT_LOAD_OP_LOAD,
- .storeOp = VK_ATTACHMENT_STORE_OP_STORE,
- .stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE,
- .stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE,
- .initialLayout = color_layout,
- .finalLayout = color_layout,
- });
-
- color_references.push_back({
- .attachment = static_cast<u32>(rt),
- .layout = color_layout,
- });
- }
-
- VkAttachmentReference zeta_attachment_ref;
- const bool has_zeta = params.zeta_format != 0;
- if (has_zeta) {
- const auto guest_format = static_cast<Tegra::DepthFormat>(params.zeta_format);
- const PixelFormat pixel_format = PixelFormatFromDepthFormat(guest_format);
- const auto format = MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, pixel_format);
- ASSERT_MSG(format.attachable, "Trying to attach a non-attachable format with format={}",
- static_cast<int>(pixel_format));
-
- const VkImageLayout zeta_layout = params.zeta_texception != 0
- ? VK_IMAGE_LAYOUT_GENERAL
- : VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL;
- descriptors.push_back({
- .flags = 0,
- .format = format.format,
- .samples = VK_SAMPLE_COUNT_1_BIT,
- .loadOp = VK_ATTACHMENT_LOAD_OP_LOAD,
- .storeOp = VK_ATTACHMENT_STORE_OP_STORE,
- .stencilLoadOp = VK_ATTACHMENT_LOAD_OP_LOAD,
- .stencilStoreOp = VK_ATTACHMENT_STORE_OP_STORE,
- .initialLayout = zeta_layout,
- .finalLayout = zeta_layout,
- });
-
- zeta_attachment_ref = {
- .attachment = static_cast<u32>(num_attachments),
- .layout = zeta_layout,
- };
- }
-
- const VkSubpassDescription subpass_description{
- .flags = 0,
- .pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS,
- .inputAttachmentCount = 0,
- .pInputAttachments = nullptr,
- .colorAttachmentCount = static_cast<u32>(color_references.size()),
- .pColorAttachments = color_references.data(),
- .pResolveAttachments = nullptr,
- .pDepthStencilAttachment = has_zeta ? &zeta_attachment_ref : nullptr,
- .preserveAttachmentCount = 0,
- .pPreserveAttachments = nullptr,
- };
-
- VkAccessFlags access = 0;
- VkPipelineStageFlags stage = 0;
- if (!color_references.empty()) {
- access |= VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT;
- stage |= VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT;
- }
-
- if (has_zeta) {
- access |= VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT |
- VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT;
- stage |= VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT;
- }
-
- const VkSubpassDependency subpass_dependency{
- .srcSubpass = VK_SUBPASS_EXTERNAL,
- .dstSubpass = 0,
- .srcStageMask = stage,
- .dstStageMask = stage,
- .srcAccessMask = 0,
- .dstAccessMask = access,
- .dependencyFlags = 0,
- };
-
- return device.GetLogical().CreateRenderPass({
- .sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO,
- .pNext = nullptr,
- .flags = 0,
- .attachmentCount = static_cast<u32>(descriptors.size()),
- .pAttachments = descriptors.data(),
- .subpassCount = 1,
- .pSubpasses = &subpass_description,
- .dependencyCount = 1,
- .pDependencies = &subpass_dependency,
- });
-}
-
-} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_renderpass_cache.h b/src/video_core/renderer_vulkan/vk_renderpass_cache.h
deleted file mode 100644
index 652ecef7b..000000000
--- a/src/video_core/renderer_vulkan/vk_renderpass_cache.h
+++ /dev/null
@@ -1,70 +0,0 @@
-// Copyright 2019 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include <type_traits>
-#include <unordered_map>
-
-#include <boost/container/static_vector.hpp>
-#include <boost/functional/hash.hpp>
-
-#include "video_core/engines/maxwell_3d.h"
-#include "video_core/renderer_vulkan/wrapper.h"
-#include "video_core/surface.h"
-
-namespace Vulkan {
-
-class VKDevice;
-
-struct RenderPassParams {
- std::array<u8, Tegra::Engines::Maxwell3D::Regs::NumRenderTargets> color_formats;
- u8 num_color_attachments;
- u8 texceptions;
-
- u8 zeta_format;
- u8 zeta_texception;
-
- std::size_t Hash() const noexcept;
-
- bool operator==(const RenderPassParams& rhs) const noexcept;
-
- bool operator!=(const RenderPassParams& rhs) const noexcept {
- return !operator==(rhs);
- }
-};
-static_assert(std::has_unique_object_representations_v<RenderPassParams>);
-static_assert(std::is_trivially_copyable_v<RenderPassParams>);
-static_assert(std::is_trivially_constructible_v<RenderPassParams>);
-
-} // namespace Vulkan
-
-namespace std {
-
-template <>
-struct hash<Vulkan::RenderPassParams> {
- std::size_t operator()(const Vulkan::RenderPassParams& k) const noexcept {
- return k.Hash();
- }
-};
-
-} // namespace std
-
-namespace Vulkan {
-
-class VKRenderPassCache final {
-public:
- explicit VKRenderPassCache(const VKDevice& device_);
- ~VKRenderPassCache();
-
- VkRenderPass GetRenderPass(const RenderPassParams& params);
-
-private:
- vk::RenderPass CreateRenderPass(const RenderPassParams& params) const;
-
- const VKDevice& device;
- std::unordered_map<RenderPassParams, vk::RenderPass> cache;
-};
-
-} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_sampler_cache.cpp b/src/video_core/renderer_vulkan/vk_sampler_cache.cpp
deleted file mode 100644
index b859691fa..000000000
--- a/src/video_core/renderer_vulkan/vk_sampler_cache.cpp
+++ /dev/null
@@ -1,83 +0,0 @@
-// Copyright 2019 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#include <unordered_map>
-
-#include "video_core/renderer_vulkan/maxwell_to_vk.h"
-#include "video_core/renderer_vulkan/vk_sampler_cache.h"
-#include "video_core/renderer_vulkan/wrapper.h"
-#include "video_core/textures/texture.h"
-
-using Tegra::Texture::TextureMipmapFilter;
-
-namespace Vulkan {
-
-namespace {
-
-VkBorderColor ConvertBorderColor(std::array<float, 4> color) {
- // TODO(Rodrigo): Manage integer border colors
- if (color == std::array<float, 4>{0, 0, 0, 0}) {
- return VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK;
- } else if (color == std::array<float, 4>{0, 0, 0, 1}) {
- return VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK;
- } else if (color == std::array<float, 4>{1, 1, 1, 1}) {
- return VK_BORDER_COLOR_FLOAT_OPAQUE_WHITE;
- }
- if (color[0] + color[1] + color[2] > 1.35f) {
- // If color elements are brighter than roughly 0.5 average, use white border
- return VK_BORDER_COLOR_FLOAT_OPAQUE_WHITE;
- } else if (color[3] > 0.5f) {
- return VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK;
- } else {
- return VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK;
- }
-}
-
-} // Anonymous namespace
-
-VKSamplerCache::VKSamplerCache(const VKDevice& device_) : device{device_} {}
-
-VKSamplerCache::~VKSamplerCache() = default;
-
-vk::Sampler VKSamplerCache::CreateSampler(const Tegra::Texture::TSCEntry& tsc) const {
- const bool arbitrary_borders = device.IsExtCustomBorderColorSupported();
- const std::array color = tsc.GetBorderColor();
-
- VkSamplerCustomBorderColorCreateInfoEXT border{
- .sType = VK_STRUCTURE_TYPE_SAMPLER_CUSTOM_BORDER_COLOR_CREATE_INFO_EXT,
- .pNext = nullptr,
- .customBorderColor = {},
- .format = VK_FORMAT_UNDEFINED,
- };
- std::memcpy(&border.customBorderColor, color.data(), sizeof(color));
-
- return device.GetLogical().CreateSampler({
- .sType = VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO,
- .pNext = arbitrary_borders ? &border : nullptr,
- .flags = 0,
- .magFilter = MaxwellToVK::Sampler::Filter(tsc.mag_filter),
- .minFilter = MaxwellToVK::Sampler::Filter(tsc.min_filter),
- .mipmapMode = MaxwellToVK::Sampler::MipmapMode(tsc.mipmap_filter),
- .addressModeU = MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_u, tsc.mag_filter),
- .addressModeV = MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_v, tsc.mag_filter),
- .addressModeW = MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_p, tsc.mag_filter),
- .mipLodBias = tsc.GetLodBias(),
- .anisotropyEnable =
- static_cast<VkBool32>(tsc.GetMaxAnisotropy() > 1.0f ? VK_TRUE : VK_FALSE),
- .maxAnisotropy = tsc.GetMaxAnisotropy(),
- .compareEnable = tsc.depth_compare_enabled,
- .compareOp = MaxwellToVK::Sampler::DepthCompareFunction(tsc.depth_compare_func),
- .minLod = tsc.mipmap_filter == TextureMipmapFilter::None ? 0.0f : tsc.GetMinLod(),
- .maxLod = tsc.mipmap_filter == TextureMipmapFilter::None ? 0.25f : tsc.GetMaxLod(),
- .borderColor =
- arbitrary_borders ? VK_BORDER_COLOR_INT_CUSTOM_EXT : ConvertBorderColor(color),
- .unnormalizedCoordinates = VK_FALSE,
- });
-}
-
-VkSampler VKSamplerCache::ToSamplerType(const vk::Sampler& sampler) const {
- return *sampler;
-}
-
-} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_sampler_cache.h b/src/video_core/renderer_vulkan/vk_sampler_cache.h
deleted file mode 100644
index 3f22c4610..000000000
--- a/src/video_core/renderer_vulkan/vk_sampler_cache.h
+++ /dev/null
@@ -1,29 +0,0 @@
-// Copyright 2019 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include "video_core/renderer_vulkan/wrapper.h"
-#include "video_core/sampler_cache.h"
-#include "video_core/textures/texture.h"
-
-namespace Vulkan {
-
-class VKDevice;
-
-class VKSamplerCache final : public VideoCommon::SamplerCache<VkSampler, vk::Sampler> {
-public:
- explicit VKSamplerCache(const VKDevice& device_);
- ~VKSamplerCache();
-
-protected:
- vk::Sampler CreateSampler(const Tegra::Texture::TSCEntry& tsc) const override;
-
- VkSampler ToSamplerType(const vk::Sampler& sampler) const override;
-
-private:
- const VKDevice& device;
-};
-
-} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_scheduler.cpp b/src/video_core/renderer_vulkan/vk_scheduler.cpp
index 1a483dc71..c104c6fe3 100644
--- a/src/video_core/renderer_vulkan/vk_scheduler.cpp
+++ b/src/video_core/renderer_vulkan/vk_scheduler.cpp
@@ -16,6 +16,7 @@
#include "video_core/renderer_vulkan/vk_query_cache.h"
#include "video_core/renderer_vulkan/vk_scheduler.h"
#include "video_core/renderer_vulkan/vk_state_tracker.h"
+#include "video_core/renderer_vulkan/vk_texture_cache.h"
#include "video_core/renderer_vulkan/wrapper.h"
namespace Vulkan {
@@ -96,38 +97,39 @@ void VKScheduler::DispatchWork() {
AcquireNewChunk();
}
-void VKScheduler::RequestRenderpass(VkRenderPass renderpass, VkFramebuffer framebuffer,
- VkExtent2D render_area) {
- if (renderpass == state.renderpass && framebuffer == state.framebuffer &&
+void VKScheduler::RequestRenderpass(const Framebuffer* framebuffer) {
+ const VkRenderPass renderpass = framebuffer->RenderPass();
+ const VkFramebuffer framebuffer_handle = framebuffer->Handle();
+ const VkExtent2D render_area = framebuffer->RenderArea();
+ if (renderpass == state.renderpass && framebuffer_handle == state.framebuffer &&
render_area.width == state.render_area.width &&
render_area.height == state.render_area.height) {
return;
}
- const bool end_renderpass = state.renderpass != nullptr;
+ EndRenderPass();
state.renderpass = renderpass;
- state.framebuffer = framebuffer;
+ state.framebuffer = framebuffer_handle;
state.render_area = render_area;
- const VkRenderPassBeginInfo renderpass_bi{
- .sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO,
- .pNext = nullptr,
- .renderPass = renderpass,
- .framebuffer = framebuffer,
- .renderArea =
- {
- .offset = {.x = 0, .y = 0},
- .extent = render_area,
- },
- .clearValueCount = 0,
- .pClearValues = nullptr,
- };
-
- Record([renderpass_bi, end_renderpass](vk::CommandBuffer cmdbuf) {
- if (end_renderpass) {
- cmdbuf.EndRenderPass();
- }
+ Record([renderpass, framebuffer_handle, render_area](vk::CommandBuffer cmdbuf) {
+ const VkRenderPassBeginInfo renderpass_bi{
+ .sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO,
+ .pNext = nullptr,
+ .renderPass = renderpass,
+ .framebuffer = framebuffer_handle,
+ .renderArea =
+ {
+ .offset = {.x = 0, .y = 0},
+ .extent = render_area,
+ },
+ .clearValueCount = 0,
+ .pClearValues = nullptr,
+ };
cmdbuf.BeginRenderPass(renderpass_bi, VK_SUBPASS_CONTENTS_INLINE);
});
+ num_renderpass_images = framebuffer->NumImages();
+ renderpass_images = framebuffer->Images();
+ renderpass_image_ranges = framebuffer->ImageRanges();
}
void VKScheduler::RequestOutsideRenderPassOperationContext() {
@@ -241,8 +243,37 @@ void VKScheduler::EndRenderPass() {
if (!state.renderpass) {
return;
}
+ Record([num_images = num_renderpass_images, images = renderpass_images,
+ ranges = renderpass_image_ranges](vk::CommandBuffer cmdbuf) {
+ std::array<VkImageMemoryBarrier, 9> barriers;
+ for (size_t i = 0; i < num_images; ++i) {
+ barriers[i] = VkImageMemoryBarrier{
+ .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
+ .pNext = nullptr,
+ .srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
+ VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT,
+ .dstAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT |
+ VK_ACCESS_COLOR_ATTACHMENT_READ_BIT |
+ VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
+ VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT |
+ VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT,
+ .oldLayout = VK_IMAGE_LAYOUT_GENERAL,
+ .newLayout = VK_IMAGE_LAYOUT_GENERAL,
+ .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+ .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+ .image = images[i],
+ .subresourceRange = ranges[i],
+ };
+ }
+ cmdbuf.EndRenderPass();
+ cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT |
+ VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT |
+ VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT,
+ VK_PIPELINE_STAGE_ALL_GRAPHICS_BIT, 0, nullptr, nullptr,
+ vk::Span(barriers.data(), num_images));
+ });
state.renderpass = nullptr;
- Record([](vk::CommandBuffer cmdbuf) { cmdbuf.EndRenderPass(); });
+ num_renderpass_images = 0;
}
void VKScheduler::AcquireNewChunk() {
diff --git a/src/video_core/renderer_vulkan/vk_scheduler.h b/src/video_core/renderer_vulkan/vk_scheduler.h
index 6d3a5da0b..0a36c8fad 100644
--- a/src/video_core/renderer_vulkan/vk_scheduler.h
+++ b/src/video_core/renderer_vulkan/vk_scheduler.h
@@ -17,6 +17,7 @@
namespace Vulkan {
class CommandPool;
+class Framebuffer;
class MasterSemaphore;
class StateTracker;
class VKDevice;
@@ -52,8 +53,7 @@ public:
void DispatchWork();
/// Requests to begin a renderpass.
- void RequestRenderpass(VkRenderPass renderpass, VkFramebuffer framebuffer,
- VkExtent2D render_area);
+ void RequestRenderpass(const Framebuffer* framebuffer);
/// Requests the current executino context to be able to execute operations only allowed outside
/// of a renderpass.
@@ -62,6 +62,9 @@ public:
/// Binds a pipeline to the current execution context.
void BindGraphicsPipeline(VkPipeline pipeline);
+ /// Invalidates current command buffer state except for render passes
+ void InvalidateState();
+
/// Assigns the query cache.
void SetQueryCache(VKQueryCache& query_cache_) {
query_cache = &query_cache_;
@@ -170,8 +173,6 @@ private:
void AllocateNewContext();
- void InvalidateState();
-
void EndPendingOperations();
void EndRenderPass();
@@ -192,6 +193,11 @@ private:
std::thread worker_thread;
State state;
+
+ u32 num_renderpass_images = 0;
+ std::array<VkImage, 9> renderpass_images{};
+ std::array<VkImageSubresourceRange, 9> renderpass_image_ranges{};
+
Common::SPSCQueue<std::unique_ptr<CommandChunk>> chunk_queue;
Common::SPSCQueue<std::unique_ptr<CommandChunk>> chunk_reserve;
std::mutex mutex;
diff --git a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
index 72954d0e3..09d6f9f35 100644
--- a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
+++ b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
@@ -102,7 +102,7 @@ struct GenericVaryingDescription {
bool is_scalar = false;
};
-spv::Dim GetSamplerDim(const Sampler& sampler) {
+spv::Dim GetSamplerDim(const SamplerEntry& sampler) {
ASSERT(!sampler.is_buffer);
switch (sampler.type) {
case Tegra::Shader::TextureType::Texture1D:
@@ -119,7 +119,7 @@ spv::Dim GetSamplerDim(const Sampler& sampler) {
}
}
-std::pair<spv::Dim, bool> GetImageDim(const Image& image) {
+std::pair<spv::Dim, bool> GetImageDim(const ImageEntry& image) {
switch (image.type) {
case Tegra::Shader::ImageType::Texture1D:
return {spv::Dim::Dim1D, false};
@@ -980,7 +980,7 @@ private:
return binding;
}
- void DeclareImage(const Image& image, u32& binding) {
+ void DeclareImage(const ImageEntry& image, u32& binding) {
const auto [dim, arrayed] = GetImageDim(image);
constexpr int depth = 0;
constexpr bool ms = false;
diff --git a/src/video_core/renderer_vulkan/vk_shader_decompiler.h b/src/video_core/renderer_vulkan/vk_shader_decompiler.h
index df1812514..ad91ad5de 100644
--- a/src/video_core/renderer_vulkan/vk_shader_decompiler.h
+++ b/src/video_core/renderer_vulkan/vk_shader_decompiler.h
@@ -21,10 +21,10 @@ class VKDevice;
namespace Vulkan {
using Maxwell = Tegra::Engines::Maxwell3D::Regs;
-using UniformTexelEntry = VideoCommon::Shader::Sampler;
-using SamplerEntry = VideoCommon::Shader::Sampler;
-using StorageTexelEntry = VideoCommon::Shader::Image;
-using ImageEntry = VideoCommon::Shader::Image;
+using UniformTexelEntry = VideoCommon::Shader::SamplerEntry;
+using SamplerEntry = VideoCommon::Shader::SamplerEntry;
+using StorageTexelEntry = VideoCommon::Shader::ImageEntry;
+using ImageEntry = VideoCommon::Shader::ImageEntry;
constexpr u32 DESCRIPTOR_SET = 0;
diff --git a/src/video_core/renderer_vulkan/vk_shader_util.cpp b/src/video_core/renderer_vulkan/vk_shader_util.cpp
index c1a218d76..38a0be7f2 100644
--- a/src/video_core/renderer_vulkan/vk_shader_util.cpp
+++ b/src/video_core/renderer_vulkan/vk_shader_util.cpp
@@ -13,18 +13,13 @@
namespace Vulkan {
-vk::ShaderModule BuildShader(const VKDevice& device, std::size_t code_size, const u8* code_data) {
- // Avoid undefined behavior by copying to a staging allocation
- ASSERT(code_size % sizeof(u32) == 0);
- const auto data = std::make_unique<u32[]>(code_size / sizeof(u32));
- std::memcpy(data.get(), code_data, code_size);
-
+vk::ShaderModule BuildShader(const VKDevice& device, std::span<const u32> code) {
return device.GetLogical().CreateShaderModule({
.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
- .codeSize = code_size,
- .pCode = data.get(),
+ .codeSize = static_cast<u32>(code.size_bytes()),
+ .pCode = code.data(),
});
}
diff --git a/src/video_core/renderer_vulkan/vk_shader_util.h b/src/video_core/renderer_vulkan/vk_shader_util.h
index d1d3f3cae..dce34a140 100644
--- a/src/video_core/renderer_vulkan/vk_shader_util.h
+++ b/src/video_core/renderer_vulkan/vk_shader_util.h
@@ -4,6 +4,8 @@
#pragma once
+#include <span>
+
#include "common/common_types.h"
#include "video_core/renderer_vulkan/wrapper.h"
@@ -11,6 +13,6 @@ namespace Vulkan {
class VKDevice;
-vk::ShaderModule BuildShader(const VKDevice& device, std::size_t code_size, const u8* code_data);
+vk::ShaderModule BuildShader(const VKDevice& device, std::span<const u32> code);
} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_state_tracker.cpp b/src/video_core/renderer_vulkan/vk_state_tracker.cpp
index 50164cc08..1779a2e30 100644
--- a/src/video_core/renderer_vulkan/vk_state_tracker.cpp
+++ b/src/video_core/renderer_vulkan/vk_state_tracker.cpp
@@ -3,6 +3,7 @@
// Refer to the license.txt file included.
#include <algorithm>
+#include <array>
#include <cstddef>
#include <iterator>
@@ -29,21 +30,15 @@ using Table = Maxwell3D::DirtyState::Table;
using Flags = Maxwell3D::DirtyState::Flags;
Flags MakeInvalidationFlags() {
+ static constexpr std::array INVALIDATION_FLAGS{
+ Viewports, Scissors, DepthBias, BlendConstants, DepthBounds,
+ StencilProperties, CullMode, DepthBoundsEnable, DepthTestEnable, DepthWriteEnable,
+ DepthCompareOp, FrontFace, StencilOp, StencilTestEnable,
+ };
Flags flags{};
- flags[Viewports] = true;
- flags[Scissors] = true;
- flags[DepthBias] = true;
- flags[BlendConstants] = true;
- flags[DepthBounds] = true;
- flags[StencilProperties] = true;
- flags[CullMode] = true;
- flags[DepthBoundsEnable] = true;
- flags[DepthTestEnable] = true;
- flags[DepthWriteEnable] = true;
- flags[DepthCompareOp] = true;
- flags[FrontFace] = true;
- flags[StencilOp] = true;
- flags[StencilTestEnable] = true;
+ for (const int flag : INVALIDATION_FLAGS) {
+ flags[flag] = true;
+ }
return flags;
}
diff --git a/src/video_core/renderer_vulkan/vk_state_tracker.h b/src/video_core/renderer_vulkan/vk_state_tracker.h
index 1de789e57..c335d2bdf 100644
--- a/src/video_core/renderer_vulkan/vk_state_tracker.h
+++ b/src/video_core/renderer_vulkan/vk_state_tracker.h
@@ -52,6 +52,14 @@ public:
current_topology = INVALID_TOPOLOGY;
}
+ void InvalidateViewports() {
+ flags[Dirty::Viewports] = true;
+ }
+
+ void InvalidateScissors() {
+ flags[Dirty::Scissors] = true;
+ }
+
bool TouchViewports() {
return Exchange(Dirty::Viewports, false);
}
diff --git a/src/video_core/renderer_vulkan/vk_stream_buffer.cpp b/src/video_core/renderer_vulkan/vk_stream_buffer.cpp
index 1b59612b9..419cb154d 100644
--- a/src/video_core/renderer_vulkan/vk_stream_buffer.cpp
+++ b/src/video_core/renderer_vulkan/vk_stream_buffer.cpp
@@ -19,6 +19,10 @@ namespace Vulkan {
namespace {
+constexpr VkBufferUsageFlags BUFFER_USAGE =
+ VK_BUFFER_USAGE_VERTEX_BUFFER_BIT | VK_BUFFER_USAGE_INDEX_BUFFER_BIT |
+ VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT;
+
constexpr u64 WATCHES_INITIAL_RESERVE = 0x4000;
constexpr u64 WATCHES_RESERVE_CHUNK = 0x1000;
@@ -56,17 +60,16 @@ u32 GetMemoryType(const VkPhysicalDeviceMemoryProperties& properties,
} // Anonymous namespace
-VKStreamBuffer::VKStreamBuffer(const VKDevice& device_, VKScheduler& scheduler_,
- VkBufferUsageFlags usage)
+VKStreamBuffer::VKStreamBuffer(const VKDevice& device_, VKScheduler& scheduler_)
: device{device_}, scheduler{scheduler_} {
- CreateBuffers(usage);
+ CreateBuffers();
ReserveWatches(current_watches, WATCHES_INITIAL_RESERVE);
ReserveWatches(previous_watches, WATCHES_INITIAL_RESERVE);
}
VKStreamBuffer::~VKStreamBuffer() = default;
-std::tuple<u8*, u64, bool> VKStreamBuffer::Map(u64 size, u64 alignment) {
+std::pair<u8*, u64> VKStreamBuffer::Map(u64 size, u64 alignment) {
ASSERT(size <= stream_buffer_size);
mapped_size = size;
@@ -76,7 +79,6 @@ std::tuple<u8*, u64, bool> VKStreamBuffer::Map(u64 size, u64 alignment) {
WaitPendingOperations(offset);
- bool invalidated = false;
if (offset + size > stream_buffer_size) {
// The buffer would overflow, save the amount of used watches and reset the state.
invalidation_mark = current_watch_cursor;
@@ -90,11 +92,9 @@ std::tuple<u8*, u64, bool> VKStreamBuffer::Map(u64 size, u64 alignment) {
// Ensure that we don't wait for uncommitted fences.
scheduler.Flush();
-
- invalidated = true;
}
- return {memory.Map(offset, size), offset, invalidated};
+ return std::make_pair(memory.Map(offset, size), offset);
}
void VKStreamBuffer::Unmap(u64 size) {
@@ -113,7 +113,7 @@ void VKStreamBuffer::Unmap(u64 size) {
watch.tick = scheduler.CurrentTick();
}
-void VKStreamBuffer::CreateBuffers(VkBufferUsageFlags usage) {
+void VKStreamBuffer::CreateBuffers() {
const auto memory_properties = device.GetPhysical().GetMemoryProperties();
const u32 preferred_type = GetMemoryType(memory_properties);
const u32 preferred_heap = memory_properties.memoryTypes[preferred_type].heapIndex;
@@ -127,7 +127,7 @@ void VKStreamBuffer::CreateBuffers(VkBufferUsageFlags usage) {
.pNext = nullptr,
.flags = 0,
.size = std::min(PREFERRED_STREAM_BUFFER_SIZE, allocable_size),
- .usage = usage,
+ .usage = BUFFER_USAGE,
.sharingMode = VK_SHARING_MODE_EXCLUSIVE,
.queueFamilyIndexCount = 0,
.pQueueFamilyIndices = nullptr,
diff --git a/src/video_core/renderer_vulkan/vk_stream_buffer.h b/src/video_core/renderer_vulkan/vk_stream_buffer.h
index 5e15ad78f..1428f77bf 100644
--- a/src/video_core/renderer_vulkan/vk_stream_buffer.h
+++ b/src/video_core/renderer_vulkan/vk_stream_buffer.h
@@ -5,7 +5,7 @@
#pragma once
#include <optional>
-#include <tuple>
+#include <utility>
#include <vector>
#include "common/common_types.h"
@@ -19,17 +19,15 @@ class VKScheduler;
class VKStreamBuffer final {
public:
- explicit VKStreamBuffer(const VKDevice& device, VKScheduler& scheduler,
- VkBufferUsageFlags usage);
+ explicit VKStreamBuffer(const VKDevice& device, VKScheduler& scheduler);
~VKStreamBuffer();
/**
* Reserves a region of memory from the stream buffer.
* @param size Size to reserve.
- * @returns A tuple in the following order: Raw memory pointer (with offset added), buffer
- * offset and a boolean that's true when buffer has been invalidated.
+ * @returns A pair of a raw memory pointer (with offset added), and the buffer offset
*/
- std::tuple<u8*, u64, bool> Map(u64 size, u64 alignment);
+ std::pair<u8*, u64> Map(u64 size, u64 alignment);
/// Ensures that "size" bytes of memory are available to the GPU, potentially recording a copy.
void Unmap(u64 size);
@@ -49,7 +47,7 @@ private:
};
/// Creates Vulkan buffer handles committing the required the required memory.
- void CreateBuffers(VkBufferUsageFlags usage);
+ void CreateBuffers();
/// Increases the amount of watches available.
void ReserveWatches(std::vector<Watch>& watches, std::size_t grow_size);
diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.cpp b/src/video_core/renderer_vulkan/vk_texture_cache.cpp
index ae2e3322c..261808391 100644
--- a/src/video_core/renderer_vulkan/vk_texture_cache.cpp
+++ b/src/video_core/renderer_vulkan/vk_texture_cache.cpp
@@ -4,614 +4,1103 @@
#include <algorithm>
#include <array>
-#include <cstddef>
-#include <cstring>
-#include <memory>
-#include <variant>
+#include <span>
#include <vector>
-#include "common/assert.h"
-#include "common/common_types.h"
-#include "core/core.h"
-#include "video_core/engines/maxwell_3d.h"
-#include "video_core/morton.h"
+#include "video_core/engines/fermi_2d.h"
+#include "video_core/renderer_vulkan/blit_image.h"
#include "video_core/renderer_vulkan/maxwell_to_vk.h"
#include "video_core/renderer_vulkan/vk_device.h"
-#include "video_core/renderer_vulkan/vk_memory_manager.h"
-#include "video_core/renderer_vulkan/vk_rasterizer.h"
#include "video_core/renderer_vulkan/vk_scheduler.h"
#include "video_core/renderer_vulkan/vk_staging_buffer_pool.h"
#include "video_core/renderer_vulkan/vk_texture_cache.h"
#include "video_core/renderer_vulkan/wrapper.h"
-#include "video_core/surface.h"
namespace Vulkan {
-using VideoCore::MortonSwizzle;
-using VideoCore::MortonSwizzleMode;
-
+using Tegra::Engines::Fermi2D;
using Tegra::Texture::SwizzleSource;
-using VideoCore::Surface::PixelFormat;
-using VideoCore::Surface::SurfaceTarget;
+using Tegra::Texture::TextureMipmapFilter;
+using VideoCommon::BufferImageCopy;
+using VideoCommon::ImageInfo;
+using VideoCommon::ImageType;
+using VideoCommon::SubresourceRange;
+using VideoCore::Surface::IsPixelFormatASTC;
namespace {
-VkImageType SurfaceTargetToImage(SurfaceTarget target) {
- switch (target) {
- case SurfaceTarget::Texture1D:
- case SurfaceTarget::Texture1DArray:
+constexpr std::array ATTACHMENT_REFERENCES{
+ VkAttachmentReference{0, VK_IMAGE_LAYOUT_GENERAL},
+ VkAttachmentReference{1, VK_IMAGE_LAYOUT_GENERAL},
+ VkAttachmentReference{2, VK_IMAGE_LAYOUT_GENERAL},
+ VkAttachmentReference{3, VK_IMAGE_LAYOUT_GENERAL},
+ VkAttachmentReference{4, VK_IMAGE_LAYOUT_GENERAL},
+ VkAttachmentReference{5, VK_IMAGE_LAYOUT_GENERAL},
+ VkAttachmentReference{6, VK_IMAGE_LAYOUT_GENERAL},
+ VkAttachmentReference{7, VK_IMAGE_LAYOUT_GENERAL},
+ VkAttachmentReference{8, VK_IMAGE_LAYOUT_GENERAL},
+};
+
+constexpr VkBorderColor ConvertBorderColor(const std::array<float, 4>& color) {
+ if (color == std::array<float, 4>{0, 0, 0, 0}) {
+ return VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK;
+ } else if (color == std::array<float, 4>{0, 0, 0, 1}) {
+ return VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK;
+ } else if (color == std::array<float, 4>{1, 1, 1, 1}) {
+ return VK_BORDER_COLOR_FLOAT_OPAQUE_WHITE;
+ }
+ if (color[0] + color[1] + color[2] > 1.35f) {
+ // If color elements are brighter than roughly 0.5 average, use white border
+ return VK_BORDER_COLOR_FLOAT_OPAQUE_WHITE;
+ } else if (color[3] > 0.5f) {
+ return VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK;
+ } else {
+ return VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK;
+ }
+}
+
+[[nodiscard]] VkImageType ConvertImageType(const ImageType type) {
+ switch (type) {
+ case ImageType::e1D:
return VK_IMAGE_TYPE_1D;
- case SurfaceTarget::Texture2D:
- case SurfaceTarget::Texture2DArray:
- case SurfaceTarget::TextureCubemap:
- case SurfaceTarget::TextureCubeArray:
+ case ImageType::e2D:
+ case ImageType::Linear:
return VK_IMAGE_TYPE_2D;
- case SurfaceTarget::Texture3D:
+ case ImageType::e3D:
return VK_IMAGE_TYPE_3D;
- case SurfaceTarget::TextureBuffer:
- UNREACHABLE();
- return {};
+ case ImageType::Buffer:
+ break;
}
- UNREACHABLE_MSG("Unknown texture target={}", target);
+ UNREACHABLE_MSG("Invalid image type={}", type);
return {};
}
-VkImageAspectFlags PixelFormatToImageAspect(PixelFormat pixel_format) {
- if (pixel_format < PixelFormat::MaxColorFormat) {
- return VK_IMAGE_ASPECT_COLOR_BIT;
- } else if (pixel_format < PixelFormat::MaxDepthFormat) {
- return VK_IMAGE_ASPECT_DEPTH_BIT;
- } else if (pixel_format < PixelFormat::MaxDepthStencilFormat) {
- return VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT;
- } else {
- UNREACHABLE_MSG("Invalid pixel format={}", pixel_format);
- return VK_IMAGE_ASPECT_COLOR_BIT;
+[[nodiscard]] VkSampleCountFlagBits ConvertSampleCount(u32 num_samples) {
+ switch (num_samples) {
+ case 1:
+ return VK_SAMPLE_COUNT_1_BIT;
+ case 2:
+ return VK_SAMPLE_COUNT_2_BIT;
+ case 4:
+ return VK_SAMPLE_COUNT_4_BIT;
+ case 8:
+ return VK_SAMPLE_COUNT_8_BIT;
+ case 16:
+ return VK_SAMPLE_COUNT_16_BIT;
+ default:
+ UNREACHABLE_MSG("Invalid number of samples={}", num_samples);
+ return VK_SAMPLE_COUNT_1_BIT;
}
}
-VkImageViewType GetImageViewType(SurfaceTarget target) {
- switch (target) {
- case SurfaceTarget::Texture1D:
- return VK_IMAGE_VIEW_TYPE_1D;
- case SurfaceTarget::Texture2D:
- return VK_IMAGE_VIEW_TYPE_2D;
- case SurfaceTarget::Texture3D:
- return VK_IMAGE_VIEW_TYPE_3D;
- case SurfaceTarget::Texture1DArray:
- return VK_IMAGE_VIEW_TYPE_1D_ARRAY;
- case SurfaceTarget::Texture2DArray:
- return VK_IMAGE_VIEW_TYPE_2D_ARRAY;
- case SurfaceTarget::TextureCubemap:
- return VK_IMAGE_VIEW_TYPE_CUBE;
- case SurfaceTarget::TextureCubeArray:
- return VK_IMAGE_VIEW_TYPE_CUBE_ARRAY;
- case SurfaceTarget::TextureBuffer:
- break;
+[[nodiscard]] VkImageCreateInfo MakeImageCreateInfo(const VKDevice& device, const ImageInfo& info) {
+ const auto format_info = MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, info.format);
+ VkImageCreateFlags flags = VK_IMAGE_CREATE_MUTABLE_FORMAT_BIT;
+ if (info.type == ImageType::e2D && info.resources.layers >= 6 &&
+ info.size.width == info.size.height) {
+ flags |= VK_IMAGE_CREATE_CUBE_COMPATIBLE_BIT;
}
- UNREACHABLE();
- return {};
-}
-
-vk::Buffer CreateBuffer(const VKDevice& device, const SurfaceParams& params,
- std::size_t host_memory_size) {
- // TODO(Rodrigo): Move texture buffer creation to the buffer cache
- return device.GetLogical().CreateBuffer({
- .sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
+ if (info.type == ImageType::e3D) {
+ flags |= VK_IMAGE_CREATE_2D_ARRAY_COMPATIBLE_BIT;
+ }
+ VkImageUsageFlags usage = VK_IMAGE_USAGE_TRANSFER_SRC_BIT | VK_IMAGE_USAGE_TRANSFER_DST_BIT |
+ VK_IMAGE_USAGE_SAMPLED_BIT;
+ if (format_info.attachable) {
+ switch (VideoCore::Surface::GetFormatType(info.format)) {
+ case VideoCore::Surface::SurfaceType::ColorTexture:
+ usage |= VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT;
+ break;
+ case VideoCore::Surface::SurfaceType::Depth:
+ case VideoCore::Surface::SurfaceType::DepthStencil:
+ usage |= VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT;
+ break;
+ default:
+ UNREACHABLE_MSG("Invalid surface type");
+ }
+ }
+ if (format_info.storage) {
+ usage |= VK_IMAGE_USAGE_STORAGE_BIT;
+ }
+ const auto [samples_x, samples_y] = VideoCommon::SamplesLog2(info.num_samples);
+ return VkImageCreateInfo{
+ .sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO,
.pNext = nullptr,
- .flags = 0,
- .size = static_cast<VkDeviceSize>(host_memory_size),
- .usage = VK_BUFFER_USAGE_UNIFORM_TEXEL_BUFFER_BIT |
- VK_BUFFER_USAGE_STORAGE_TEXEL_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_SRC_BIT |
- VK_BUFFER_USAGE_TRANSFER_DST_BIT,
+ .flags = flags,
+ .imageType = ConvertImageType(info.type),
+ .format = format_info.format,
+ .extent =
+ {
+ .width = info.size.width >> samples_x,
+ .height = info.size.height >> samples_y,
+ .depth = info.size.depth,
+ },
+ .mipLevels = static_cast<u32>(info.resources.levels),
+ .arrayLayers = static_cast<u32>(info.resources.layers),
+ .samples = ConvertSampleCount(info.num_samples),
+ .tiling = VK_IMAGE_TILING_OPTIMAL,
+ .usage = usage,
.sharingMode = VK_SHARING_MODE_EXCLUSIVE,
.queueFamilyIndexCount = 0,
.pQueueFamilyIndices = nullptr,
- });
-}
-
-VkBufferViewCreateInfo GenerateBufferViewCreateInfo(const VKDevice& device,
- const SurfaceParams& params, VkBuffer buffer,
- std::size_t host_memory_size) {
- ASSERT(params.IsBuffer());
-
- return {
- .sType = VK_STRUCTURE_TYPE_BUFFER_VIEW_CREATE_INFO,
- .pNext = nullptr,
- .flags = 0,
- .buffer = buffer,
- .format =
- MaxwellToVK::SurfaceFormat(device, FormatType::Buffer, params.pixel_format).format,
- .offset = 0,
- .range = static_cast<VkDeviceSize>(host_memory_size),
+ .initialLayout = VK_IMAGE_LAYOUT_UNDEFINED,
};
}
-VkImageCreateInfo GenerateImageCreateInfo(const VKDevice& device, const SurfaceParams& params) {
- ASSERT(!params.IsBuffer());
-
- const auto [format, attachable, storage] =
- MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, params.pixel_format);
+[[nodiscard]] vk::Image MakeImage(const VKDevice& device, const ImageInfo& info) {
+ if (info.type == ImageType::Buffer) {
+ return vk::Image{};
+ }
+ return device.GetLogical().CreateImage(MakeImageCreateInfo(device, info));
+}
- VkImageCreateInfo ci{
- .sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO,
+[[nodiscard]] vk::Buffer MakeBuffer(const VKDevice& device, const ImageInfo& info) {
+ if (info.type != ImageType::Buffer) {
+ return vk::Buffer{};
+ }
+ const size_t bytes_per_block = VideoCore::Surface::BytesPerBlock(info.format);
+ return device.GetLogical().CreateBuffer(VkBufferCreateInfo{
+ .sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
- .imageType = SurfaceTargetToImage(params.target),
- .format = format,
- .extent = {},
- .mipLevels = params.num_levels,
- .arrayLayers = static_cast<u32>(params.GetNumLayers()),
- .samples = VK_SAMPLE_COUNT_1_BIT,
- .tiling = VK_IMAGE_TILING_OPTIMAL,
- .usage = VK_IMAGE_USAGE_SAMPLED_BIT | VK_IMAGE_USAGE_TRANSFER_DST_BIT |
- VK_IMAGE_USAGE_TRANSFER_SRC_BIT,
+ .size = info.size.width * bytes_per_block,
+ .usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT |
+ VK_BUFFER_USAGE_UNIFORM_TEXEL_BUFFER_BIT |
+ VK_BUFFER_USAGE_STORAGE_TEXEL_BUFFER_BIT,
.sharingMode = VK_SHARING_MODE_EXCLUSIVE,
.queueFamilyIndexCount = 0,
.pQueueFamilyIndices = nullptr,
- .initialLayout = VK_IMAGE_LAYOUT_UNDEFINED,
- };
- if (attachable) {
- ci.usage |= params.IsPixelFormatZeta() ? VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT
- : VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT;
- }
- if (storage) {
- ci.usage |= VK_IMAGE_USAGE_STORAGE_BIT;
- }
-
- switch (params.target) {
- case SurfaceTarget::TextureCubemap:
- case SurfaceTarget::TextureCubeArray:
- ci.flags |= VK_IMAGE_CREATE_CUBE_COMPATIBLE_BIT;
- [[fallthrough]];
- case SurfaceTarget::Texture1D:
- case SurfaceTarget::Texture1DArray:
- case SurfaceTarget::Texture2D:
- case SurfaceTarget::Texture2DArray:
- ci.extent = {params.width, params.height, 1};
- break;
- case SurfaceTarget::Texture3D:
- ci.flags |= VK_IMAGE_CREATE_2D_ARRAY_COMPATIBLE_BIT;
- ci.extent = {params.width, params.height, params.depth};
- break;
- case SurfaceTarget::TextureBuffer:
- UNREACHABLE();
- }
-
- return ci;
+ });
}
-u32 EncodeSwizzle(SwizzleSource x_source, SwizzleSource y_source, SwizzleSource z_source,
- SwizzleSource w_source) {
- return (static_cast<u32>(x_source) << 24) | (static_cast<u32>(y_source) << 16) |
- (static_cast<u32>(z_source) << 8) | static_cast<u32>(w_source);
+[[nodiscard]] VkImageAspectFlags ImageAspectMask(PixelFormat format) {
+ switch (VideoCore::Surface::GetFormatType(format)) {
+ case VideoCore::Surface::SurfaceType::ColorTexture:
+ return VK_IMAGE_ASPECT_COLOR_BIT;
+ case VideoCore::Surface::SurfaceType::Depth:
+ return VK_IMAGE_ASPECT_DEPTH_BIT;
+ case VideoCore::Surface::SurfaceType::DepthStencil:
+ return VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT;
+ default:
+ UNREACHABLE_MSG("Invalid surface type");
+ return VkImageAspectFlags{};
+ }
}
-} // Anonymous namespace
-
-CachedSurface::CachedSurface(const VKDevice& device_, VKMemoryManager& memory_manager_,
- VKScheduler& scheduler_, VKStagingBufferPool& staging_pool_,
- GPUVAddr gpu_addr_, const SurfaceParams& params_)
- : SurfaceBase<View>{gpu_addr_, params_, device_.IsOptimalAstcSupported()}, device{device_},
- memory_manager{memory_manager_}, scheduler{scheduler_}, staging_pool{staging_pool_} {
- if (params.IsBuffer()) {
- buffer = CreateBuffer(device, params, host_memory_size);
- commit = memory_manager.Commit(buffer, false);
-
- const auto buffer_view_ci =
- GenerateBufferViewCreateInfo(device, params, *buffer, host_memory_size);
- format = buffer_view_ci.format;
-
- buffer_view = device.GetLogical().CreateBufferView(buffer_view_ci);
- } else {
- const auto image_ci = GenerateImageCreateInfo(device, params);
- format = image_ci.format;
-
- image.emplace(device, scheduler, image_ci, PixelFormatToImageAspect(params.pixel_format));
- commit = memory_manager.Commit(image->GetHandle(), false);
+[[nodiscard]] VkImageAspectFlags ImageViewAspectMask(const VideoCommon::ImageViewInfo& info) {
+ if (info.IsRenderTarget()) {
+ return ImageAspectMask(info.format);
}
-
- // TODO(Rodrigo): Move this to a virtual function.
- u32 num_layers = 1;
- if (params.is_layered || params.target == SurfaceTarget::Texture3D) {
- num_layers = params.depth;
+ const bool is_first = info.Swizzle()[0] == SwizzleSource::R;
+ switch (info.format) {
+ case PixelFormat::D24_UNORM_S8_UINT:
+ case PixelFormat::D32_FLOAT_S8_UINT:
+ return is_first ? VK_IMAGE_ASPECT_DEPTH_BIT : VK_IMAGE_ASPECT_STENCIL_BIT;
+ case PixelFormat::S8_UINT_D24_UNORM:
+ return is_first ? VK_IMAGE_ASPECT_STENCIL_BIT : VK_IMAGE_ASPECT_DEPTH_BIT;
+ case PixelFormat::D16_UNORM:
+ case PixelFormat::D32_FLOAT:
+ return VK_IMAGE_ASPECT_DEPTH_BIT;
+ default:
+ return VK_IMAGE_ASPECT_COLOR_BIT;
}
- main_view = CreateView(ViewParams(params.target, 0, num_layers, 0, params.num_levels));
}
-CachedSurface::~CachedSurface() = default;
-
-void CachedSurface::UploadTexture(const std::vector<u8>& staging_buffer) {
- // To upload data we have to be outside of a renderpass
- scheduler.RequestOutsideRenderPassOperationContext();
+[[nodiscard]] VkAttachmentDescription AttachmentDescription(const VKDevice& device,
+ const ImageView* image_view) {
+ const auto pixel_format = image_view->format;
+ return VkAttachmentDescription{
+ .flags = VK_ATTACHMENT_DESCRIPTION_MAY_ALIAS_BIT,
+ .format = MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, pixel_format).format,
+ .samples = image_view->Samples(),
+ .loadOp = VK_ATTACHMENT_LOAD_OP_LOAD,
+ .storeOp = VK_ATTACHMENT_STORE_OP_STORE,
+ .stencilLoadOp = VK_ATTACHMENT_LOAD_OP_LOAD,
+ .stencilStoreOp = VK_ATTACHMENT_STORE_OP_STORE,
+ .initialLayout = VK_IMAGE_LAYOUT_GENERAL,
+ .finalLayout = VK_IMAGE_LAYOUT_GENERAL,
+ };
+}
- if (params.IsBuffer()) {
- UploadBuffer(staging_buffer);
- } else {
- UploadImage(staging_buffer);
+[[nodiscard]] VkComponentSwizzle ComponentSwizzle(SwizzleSource swizzle) {
+ switch (swizzle) {
+ case SwizzleSource::Zero:
+ return VK_COMPONENT_SWIZZLE_ZERO;
+ case SwizzleSource::R:
+ return VK_COMPONENT_SWIZZLE_R;
+ case SwizzleSource::G:
+ return VK_COMPONENT_SWIZZLE_G;
+ case SwizzleSource::B:
+ return VK_COMPONENT_SWIZZLE_B;
+ case SwizzleSource::A:
+ return VK_COMPONENT_SWIZZLE_A;
+ case SwizzleSource::OneFloat:
+ case SwizzleSource::OneInt:
+ return VK_COMPONENT_SWIZZLE_ONE;
}
+ UNREACHABLE_MSG("Invalid swizzle={}", swizzle);
+ return VK_COMPONENT_SWIZZLE_ZERO;
}
-void CachedSurface::DownloadTexture(std::vector<u8>& staging_buffer) {
- UNIMPLEMENTED_IF(params.IsBuffer());
-
- if (params.pixel_format == PixelFormat::A1B5G5R5_UNORM) {
- LOG_WARNING(Render_Vulkan, "A1B5G5R5 flushing is stubbed");
+[[nodiscard]] VkImageViewType ImageViewType(VideoCommon::ImageViewType type) {
+ switch (type) {
+ case VideoCommon::ImageViewType::e1D:
+ return VK_IMAGE_VIEW_TYPE_1D;
+ case VideoCommon::ImageViewType::e2D:
+ return VK_IMAGE_VIEW_TYPE_2D;
+ case VideoCommon::ImageViewType::Cube:
+ return VK_IMAGE_VIEW_TYPE_CUBE;
+ case VideoCommon::ImageViewType::e3D:
+ return VK_IMAGE_VIEW_TYPE_3D;
+ case VideoCommon::ImageViewType::e1DArray:
+ return VK_IMAGE_VIEW_TYPE_1D_ARRAY;
+ case VideoCommon::ImageViewType::e2DArray:
+ return VK_IMAGE_VIEW_TYPE_2D_ARRAY;
+ case VideoCommon::ImageViewType::CubeArray:
+ return VK_IMAGE_VIEW_TYPE_CUBE_ARRAY;
+ case VideoCommon::ImageViewType::Rect:
+ LOG_WARNING(Render_Vulkan, "Unnormalized image view type not supported");
+ return VK_IMAGE_VIEW_TYPE_2D;
+ case VideoCommon::ImageViewType::Buffer:
+ UNREACHABLE_MSG("Texture buffers can't be image views");
+ return VK_IMAGE_VIEW_TYPE_1D;
}
+ UNREACHABLE_MSG("Invalid image view type={}", type);
+ return VK_IMAGE_VIEW_TYPE_2D;
+}
- // We can't copy images to buffers inside a renderpass
- scheduler.RequestOutsideRenderPassOperationContext();
+[[nodiscard]] VkImageSubresourceLayers MakeImageSubresourceLayers(
+ VideoCommon::SubresourceLayers subresource, VkImageAspectFlags aspect_mask) {
+ return VkImageSubresourceLayers{
+ .aspectMask = aspect_mask,
+ .mipLevel = static_cast<u32>(subresource.base_level),
+ .baseArrayLayer = static_cast<u32>(subresource.base_layer),
+ .layerCount = static_cast<u32>(subresource.num_layers),
+ };
+}
- FullTransition(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_ACCESS_TRANSFER_READ_BIT,
- VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL);
+[[nodiscard]] VkOffset3D MakeOffset3D(VideoCommon::Offset3D offset3d) {
+ return VkOffset3D{
+ .x = offset3d.x,
+ .y = offset3d.y,
+ .z = offset3d.z,
+ };
+}
- const auto& unused_buffer = staging_pool.GetUnusedBuffer(host_memory_size, true);
- // TODO(Rodrigo): Do this in a single copy
- for (u32 level = 0; level < params.num_levels; ++level) {
- scheduler.Record([image = *image->GetHandle(), buffer = *unused_buffer.handle,
- copy = GetBufferImageCopy(level)](vk::CommandBuffer cmdbuf) {
- cmdbuf.CopyImageToBuffer(image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, buffer, copy);
- });
- }
- scheduler.Finish();
+[[nodiscard]] VkExtent3D MakeExtent3D(VideoCommon::Extent3D extent3d) {
+ return VkExtent3D{
+ .width = static_cast<u32>(extent3d.width),
+ .height = static_cast<u32>(extent3d.height),
+ .depth = static_cast<u32>(extent3d.depth),
+ };
+}
- // TODO(Rodrigo): Use an intern buffer for staging buffers and avoid this unnecessary memcpy.
- std::memcpy(staging_buffer.data(), unused_buffer.commit->Map(host_memory_size),
- host_memory_size);
+[[nodiscard]] VkImageCopy MakeImageCopy(const VideoCommon::ImageCopy& copy,
+ VkImageAspectFlags aspect_mask) noexcept {
+ return VkImageCopy{
+ .srcSubresource = MakeImageSubresourceLayers(copy.src_subresource, aspect_mask),
+ .srcOffset = MakeOffset3D(copy.src_offset),
+ .dstSubresource = MakeImageSubresourceLayers(copy.dst_subresource, aspect_mask),
+ .dstOffset = MakeOffset3D(copy.dst_offset),
+ .extent = MakeExtent3D(copy.extent),
+ };
}
-void CachedSurface::DecorateSurfaceName() {
- // TODO(Rodrigo): Add name decorations
+[[nodiscard]] std::vector<VkBufferCopy> TransformBufferCopies(
+ std::span<const VideoCommon::BufferCopy> copies, size_t buffer_offset) {
+ std::vector<VkBufferCopy> result(copies.size());
+ std::ranges::transform(
+ copies, result.begin(), [buffer_offset](const VideoCommon::BufferCopy& copy) {
+ return VkBufferCopy{
+ .srcOffset = static_cast<VkDeviceSize>(copy.src_offset + buffer_offset),
+ .dstOffset = static_cast<VkDeviceSize>(copy.dst_offset),
+ .size = static_cast<VkDeviceSize>(copy.size),
+ };
+ });
+ return result;
}
-View CachedSurface::CreateView(const ViewParams& view_params) {
- // TODO(Rodrigo): Add name decorations
- return views[view_params] = std::make_shared<CachedSurfaceView>(device, *this, view_params);
+[[nodiscard]] std::vector<VkBufferImageCopy> TransformBufferImageCopies(
+ std::span<const BufferImageCopy> copies, size_t buffer_offset, VkImageAspectFlags aspect_mask) {
+ struct Maker {
+ VkBufferImageCopy operator()(const BufferImageCopy& copy) const {
+ return VkBufferImageCopy{
+ .bufferOffset = copy.buffer_offset + buffer_offset,
+ .bufferRowLength = copy.buffer_row_length,
+ .bufferImageHeight = copy.buffer_image_height,
+ .imageSubresource =
+ {
+ .aspectMask = aspect_mask,
+ .mipLevel = static_cast<u32>(copy.image_subresource.base_level),
+ .baseArrayLayer = static_cast<u32>(copy.image_subresource.base_layer),
+ .layerCount = static_cast<u32>(copy.image_subresource.num_layers),
+ },
+ .imageOffset =
+ {
+ .x = copy.image_offset.x,
+ .y = copy.image_offset.y,
+ .z = copy.image_offset.z,
+ },
+ .imageExtent =
+ {
+ .width = copy.image_extent.width,
+ .height = copy.image_extent.height,
+ .depth = copy.image_extent.depth,
+ },
+ };
+ }
+ size_t buffer_offset;
+ VkImageAspectFlags aspect_mask;
+ };
+ if (aspect_mask == (VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT)) {
+ std::vector<VkBufferImageCopy> result(copies.size() * 2);
+ std::ranges::transform(copies, result.begin(),
+ Maker{buffer_offset, VK_IMAGE_ASPECT_DEPTH_BIT});
+ std::ranges::transform(copies, result.begin() + copies.size(),
+ Maker{buffer_offset, VK_IMAGE_ASPECT_STENCIL_BIT});
+ return result;
+ } else {
+ std::vector<VkBufferImageCopy> result(copies.size());
+ std::ranges::transform(copies, result.begin(), Maker{buffer_offset, aspect_mask});
+ return result;
+ }
}
-void CachedSurface::UploadBuffer(const std::vector<u8>& staging_buffer) {
- const auto& src_buffer = staging_pool.GetUnusedBuffer(host_memory_size, true);
- std::memcpy(src_buffer.commit->Map(host_memory_size), staging_buffer.data(), host_memory_size);
+[[nodiscard]] VkImageSubresourceRange MakeSubresourceRange(VkImageAspectFlags aspect_mask,
+ const SubresourceRange& range) {
+ return VkImageSubresourceRange{
+ .aspectMask = aspect_mask,
+ .baseMipLevel = static_cast<u32>(range.base.level),
+ .levelCount = static_cast<u32>(range.extent.levels),
+ .baseArrayLayer = static_cast<u32>(range.base.layer),
+ .layerCount = static_cast<u32>(range.extent.layers),
+ };
+}
- scheduler.Record([src_buffer = *src_buffer.handle, dst_buffer = *buffer,
- size = host_memory_size](vk::CommandBuffer cmdbuf) {
- VkBufferCopy copy;
- copy.srcOffset = 0;
- copy.dstOffset = 0;
- copy.size = size;
- cmdbuf.CopyBuffer(src_buffer, dst_buffer, copy);
+[[nodiscard]] VkImageSubresourceRange MakeSubresourceRange(const ImageView* image_view) {
+ SubresourceRange range = image_view->range;
+ if (True(image_view->flags & VideoCommon::ImageViewFlagBits::Slice)) {
+ // Slice image views always affect a single layer, but their subresource range corresponds
+ // to the slice. Override the value to affect a single layer.
+ range.base.layer = 0;
+ range.extent.layers = 1;
+ }
+ return MakeSubresourceRange(ImageAspectMask(image_view->format), range);
+}
- VkBufferMemoryBarrier barrier;
- barrier.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER;
- barrier.pNext = nullptr;
- barrier.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT;
- barrier.dstAccessMask = VK_ACCESS_SHADER_READ_BIT;
- barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; // They'll be ignored anyway
- barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED;
- barrier.buffer = dst_buffer;
- barrier.offset = 0;
- barrier.size = size;
- cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_VERTEX_SHADER_BIT,
- 0, {}, barrier, {});
- });
+[[nodiscard]] VkImageSubresourceLayers MakeSubresourceLayers(const ImageView* image_view) {
+ return VkImageSubresourceLayers{
+ .aspectMask = ImageAspectMask(image_view->format),
+ .mipLevel = static_cast<u32>(image_view->range.base.level),
+ .baseArrayLayer = static_cast<u32>(image_view->range.base.layer),
+ .layerCount = static_cast<u32>(image_view->range.extent.layers),
+ };
}
-void CachedSurface::UploadImage(const std::vector<u8>& staging_buffer) {
- const auto& src_buffer = staging_pool.GetUnusedBuffer(host_memory_size, true);
- std::memcpy(src_buffer.commit->Map(host_memory_size), staging_buffer.data(), host_memory_size);
-
- FullTransition(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_ACCESS_TRANSFER_WRITE_BIT,
- VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL);
-
- for (u32 level = 0; level < params.num_levels; ++level) {
- const VkBufferImageCopy copy = GetBufferImageCopy(level);
- if (image->GetAspectMask() == (VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT)) {
- scheduler.Record([buffer = *src_buffer.handle, image = *image->GetHandle(),
- copy](vk::CommandBuffer cmdbuf) {
- std::array<VkBufferImageCopy, 2> copies = {copy, copy};
- copies[0].imageSubresource.aspectMask = VK_IMAGE_ASPECT_DEPTH_BIT;
- copies[1].imageSubresource.aspectMask = VK_IMAGE_ASPECT_STENCIL_BIT;
- cmdbuf.CopyBufferToImage(buffer, image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
- copies);
- });
- } else {
- scheduler.Record([buffer = *src_buffer.handle, image = *image->GetHandle(),
- copy](vk::CommandBuffer cmdbuf) {
- cmdbuf.CopyBufferToImage(buffer, image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, copy);
- });
- }
+[[nodiscard]] constexpr SwizzleSource ConvertGreenRed(SwizzleSource value) {
+ switch (value) {
+ case SwizzleSource::G:
+ return SwizzleSource::R;
+ default:
+ return value;
}
}
-VkBufferImageCopy CachedSurface::GetBufferImageCopy(u32 level) const {
- return {
- .bufferOffset = params.GetHostMipmapLevelOffset(level, is_converted),
- .bufferRowLength = 0,
- .bufferImageHeight = 0,
- .imageSubresource =
+void CopyBufferToImage(vk::CommandBuffer cmdbuf, VkBuffer src_buffer, VkImage image,
+ VkImageAspectFlags aspect_mask, bool is_initialized,
+ std::span<const VkBufferImageCopy> copies) {
+ static constexpr VkAccessFlags ACCESS_FLAGS = VK_ACCESS_SHADER_WRITE_BIT |
+ VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
+ VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT;
+ const VkImageMemoryBarrier read_barrier{
+ .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
+ .pNext = nullptr,
+ .srcAccessMask = ACCESS_FLAGS,
+ .dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
+ .oldLayout = is_initialized ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_UNDEFINED,
+ .newLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
+ .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+ .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+ .image = image,
+ .subresourceRange =
{
- .aspectMask = image->GetAspectMask(),
- .mipLevel = level,
+ .aspectMask = aspect_mask,
+ .baseMipLevel = 0,
+ .levelCount = VK_REMAINING_MIP_LEVELS,
.baseArrayLayer = 0,
- .layerCount = static_cast<u32>(params.GetNumLayers()),
+ .layerCount = VK_REMAINING_ARRAY_LAYERS,
},
- .imageOffset = {.x = 0, .y = 0, .z = 0},
- .imageExtent =
+ };
+ const VkImageMemoryBarrier write_barrier{
+ .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
+ .pNext = nullptr,
+ .srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
+ .dstAccessMask = ACCESS_FLAGS,
+ .oldLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
+ .newLayout = VK_IMAGE_LAYOUT_GENERAL,
+ .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+ .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+ .image = image,
+ .subresourceRange =
{
- .width = params.GetMipWidth(level),
- .height = params.GetMipHeight(level),
- .depth = params.target == SurfaceTarget::Texture3D ? params.GetMipDepth(level) : 1U,
+ .aspectMask = aspect_mask,
+ .baseMipLevel = 0,
+ .levelCount = VK_REMAINING_MIP_LEVELS,
+ .baseArrayLayer = 0,
+ .layerCount = VK_REMAINING_ARRAY_LAYERS,
},
};
+ cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0,
+ read_barrier);
+ cmdbuf.CopyBufferToImage(src_buffer, image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, copies);
+ // TODO: Move this to another API
+ cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, 0,
+ write_barrier);
}
-VkImageSubresourceRange CachedSurface::GetImageSubresourceRange() const {
- return {image->GetAspectMask(), 0, params.num_levels, 0,
- static_cast<u32>(params.GetNumLayers())};
+[[nodiscard]] VkImageBlit MakeImageBlit(const std::array<Offset2D, 2>& dst_region,
+ const std::array<Offset2D, 2>& src_region,
+ const VkImageSubresourceLayers& dst_layers,
+ const VkImageSubresourceLayers& src_layers) {
+ return VkImageBlit{
+ .srcSubresource = src_layers,
+ .srcOffsets =
+ {
+ {
+ .x = src_region[0].x,
+ .y = src_region[0].y,
+ .z = 0,
+ },
+ {
+ .x = src_region[1].x,
+ .y = src_region[1].y,
+ .z = 1,
+ },
+ },
+ .dstSubresource = dst_layers,
+ .dstOffsets =
+ {
+ {
+ .x = dst_region[0].x,
+ .y = dst_region[0].y,
+ .z = 0,
+ },
+ {
+ .x = dst_region[1].x,
+ .y = dst_region[1].y,
+ .z = 1,
+ },
+ },
+ };
}
-CachedSurfaceView::CachedSurfaceView(const VKDevice& device_, CachedSurface& surface_,
- const ViewParams& view_params_)
- : ViewBase{view_params_}, surface_params{surface_.GetSurfaceParams()},
- image{surface_.GetImageHandle()}, buffer_view{surface_.GetBufferViewHandle()},
- aspect_mask{surface_.GetAspectMask()}, device{device_}, surface{surface_},
- base_level{view_params_.base_level}, num_levels{view_params_.num_levels},
- image_view_type{image ? GetImageViewType(view_params_.target) : VK_IMAGE_VIEW_TYPE_1D} {
- if (image_view_type == VK_IMAGE_VIEW_TYPE_3D) {
- base_layer = 0;
- num_layers = 1;
- base_slice = view_params_.base_layer;
- num_slices = view_params_.num_layers;
- } else {
- base_layer = view_params_.base_layer;
- num_layers = view_params_.num_layers;
- }
+[[nodiscard]] VkImageResolve MakeImageResolve(const std::array<Offset2D, 2>& dst_region,
+ const std::array<Offset2D, 2>& src_region,
+ const VkImageSubresourceLayers& dst_layers,
+ const VkImageSubresourceLayers& src_layers) {
+ return VkImageResolve{
+ .srcSubresource = src_layers,
+ .srcOffset =
+ {
+ .x = src_region[0].x,
+ .y = src_region[0].y,
+ .z = 0,
+ },
+ .dstSubresource = dst_layers,
+ .dstOffset =
+ {
+ .x = dst_region[0].x,
+ .y = dst_region[0].y,
+ .z = 0,
+ },
+ .extent =
+ {
+ .width = static_cast<u32>(dst_region[1].x - dst_region[0].x),
+ .height = static_cast<u32>(dst_region[1].y - dst_region[0].y),
+ .depth = 1,
+ },
+ };
}
-CachedSurfaceView::~CachedSurfaceView() = default;
-
-VkImageView CachedSurfaceView::GetImageView(SwizzleSource x_source, SwizzleSource y_source,
- SwizzleSource z_source, SwizzleSource w_source) {
- const u32 new_swizzle = EncodeSwizzle(x_source, y_source, z_source, w_source);
- if (last_image_view && last_swizzle == new_swizzle) {
- return last_image_view;
+struct RangedBarrierRange {
+ u32 min_mip = std::numeric_limits<u32>::max();
+ u32 max_mip = std::numeric_limits<u32>::min();
+ u32 min_layer = std::numeric_limits<u32>::max();
+ u32 max_layer = std::numeric_limits<u32>::min();
+
+ void AddLayers(const VkImageSubresourceLayers& layers) {
+ min_mip = std::min(min_mip, layers.mipLevel);
+ max_mip = std::max(max_mip, layers.mipLevel + 1);
+ min_layer = std::min(min_layer, layers.baseArrayLayer);
+ max_layer = std::max(max_layer, layers.baseArrayLayer + layers.layerCount);
}
- last_swizzle = new_swizzle;
- const auto [entry, is_cache_miss] = view_cache.try_emplace(new_swizzle);
- auto& image_view = entry->second;
- if (!is_cache_miss) {
- return last_image_view = *image_view;
+ VkImageSubresourceRange SubresourceRange(VkImageAspectFlags aspect_mask) const noexcept {
+ return VkImageSubresourceRange{
+ .aspectMask = aspect_mask,
+ .baseMipLevel = min_mip,
+ .levelCount = max_mip - min_mip,
+ .baseArrayLayer = min_layer,
+ .layerCount = max_layer - min_layer,
+ };
}
+};
- std::array swizzle{MaxwellToVK::SwizzleSource(x_source), MaxwellToVK::SwizzleSource(y_source),
- MaxwellToVK::SwizzleSource(z_source), MaxwellToVK::SwizzleSource(w_source)};
- if (surface_params.pixel_format == PixelFormat::A1B5G5R5_UNORM) {
- // A1B5G5R5 is implemented as A1R5G5B5, we have to change the swizzle here.
- std::swap(swizzle[0], swizzle[2]);
- }
+} // Anonymous namespace
- // Games can sample depth or stencil values on textures. This is decided by the swizzle value on
- // hardware. To emulate this on Vulkan we specify it in the aspect.
- VkImageAspectFlags aspect = aspect_mask;
- if (aspect == (VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT)) {
- UNIMPLEMENTED_IF(x_source != SwizzleSource::R && x_source != SwizzleSource::G);
- const bool is_first = x_source == SwizzleSource::R;
- switch (surface_params.pixel_format) {
- case PixelFormat::D24_UNORM_S8_UINT:
- case PixelFormat::D32_FLOAT_S8_UINT:
- aspect = is_first ? VK_IMAGE_ASPECT_DEPTH_BIT : VK_IMAGE_ASPECT_STENCIL_BIT;
- break;
- case PixelFormat::S8_UINT_D24_UNORM:
- aspect = is_first ? VK_IMAGE_ASPECT_STENCIL_BIT : VK_IMAGE_ASPECT_DEPTH_BIT;
- break;
- default:
- aspect = VK_IMAGE_ASPECT_DEPTH_BIT;
- UNIMPLEMENTED();
- }
+void TextureCacheRuntime::Finish() {
+ scheduler.Finish();
+}
- // Make sure we sample the first component
- std::transform(
- swizzle.begin(), swizzle.end(), swizzle.begin(), [](VkComponentSwizzle component) {
- return component == VK_COMPONENT_SWIZZLE_G ? VK_COMPONENT_SWIZZLE_R : component;
- });
- }
+ImageBufferMap TextureCacheRuntime::MapUploadBuffer(size_t size) {
+ const auto& buffer = staging_buffer_pool.GetUnusedBuffer(size, true);
+ return ImageBufferMap{
+ .handle = *buffer.handle,
+ .map = buffer.commit->Map(size),
+ };
+}
- if (image_view_type == VK_IMAGE_VIEW_TYPE_3D) {
- ASSERT(base_slice == 0);
- ASSERT(num_slices == surface_params.depth);
+void TextureCacheRuntime::BlitImage(Framebuffer* dst_framebuffer, ImageView& dst, ImageView& src,
+ const std::array<Offset2D, 2>& dst_region,
+ const std::array<Offset2D, 2>& src_region,
+ Tegra::Engines::Fermi2D::Filter filter,
+ Tegra::Engines::Fermi2D::Operation operation) {
+ const VkImageAspectFlags aspect_mask = ImageAspectMask(src.format);
+ const bool is_dst_msaa = dst.Samples() != VK_SAMPLE_COUNT_1_BIT;
+ const bool is_src_msaa = src.Samples() != VK_SAMPLE_COUNT_1_BIT;
+ ASSERT(aspect_mask == ImageAspectMask(dst.format));
+ if (aspect_mask == VK_IMAGE_ASPECT_COLOR_BIT && !is_src_msaa && !is_dst_msaa) {
+ blit_image_helper.BlitColor(dst_framebuffer, src, dst_region, src_region, filter,
+ operation);
+ return;
}
-
- image_view = device.GetLogical().CreateImageView({
- .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
- .pNext = nullptr,
- .flags = 0,
- .image = surface.GetImageHandle(),
- .viewType = image_view_type,
- .format = surface.GetImage().GetFormat(),
- .components =
- {
- .r = swizzle[0],
- .g = swizzle[1],
- .b = swizzle[2],
- .a = swizzle[3],
+ if (aspect_mask == (VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT)) {
+ if (!device.IsBlitDepthStencilSupported()) {
+ UNIMPLEMENTED_IF(is_src_msaa || is_dst_msaa);
+ blit_image_helper.BlitDepthStencil(dst_framebuffer, src.DepthView(), src.StencilView(),
+ dst_region, src_region, filter, operation);
+ return;
+ }
+ }
+ ASSERT(src.ImageFormat() == dst.ImageFormat());
+ ASSERT(!(is_dst_msaa && !is_src_msaa));
+ ASSERT(operation == Fermi2D::Operation::SrcCopy);
+
+ const VkImage dst_image = dst.ImageHandle();
+ const VkImage src_image = src.ImageHandle();
+ const VkImageSubresourceLayers dst_layers = MakeSubresourceLayers(&dst);
+ const VkImageSubresourceLayers src_layers = MakeSubresourceLayers(&src);
+ const bool is_resolve = is_src_msaa && !is_dst_msaa;
+ scheduler.RequestOutsideRenderPassOperationContext();
+ scheduler.Record([filter, dst_region, src_region, dst_image, src_image, dst_layers, src_layers,
+ aspect_mask, is_resolve](vk::CommandBuffer cmdbuf) {
+ const std::array read_barriers{
+ VkImageMemoryBarrier{
+ .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
+ .pNext = nullptr,
+ .srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT |
+ VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT |
+ VK_ACCESS_TRANSFER_WRITE_BIT,
+ .dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT,
+ .oldLayout = VK_IMAGE_LAYOUT_GENERAL,
+ .newLayout = VK_IMAGE_LAYOUT_GENERAL,
+ .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+ .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+ .image = src_image,
+ .subresourceRange{
+ .aspectMask = aspect_mask,
+ .baseMipLevel = 0,
+ .levelCount = VK_REMAINING_MIP_LEVELS,
+ .baseArrayLayer = 0,
+ .layerCount = VK_REMAINING_ARRAY_LAYERS,
+ },
},
- .subresourceRange =
- {
- .aspectMask = aspect,
- .baseMipLevel = base_level,
- .levelCount = num_levels,
- .baseArrayLayer = base_layer,
- .layerCount = num_layers,
+ VkImageMemoryBarrier{
+ .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
+ .pNext = nullptr,
+ .srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT |
+ VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT |
+ VK_ACCESS_TRANSFER_WRITE_BIT,
+ .dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
+ .oldLayout = VK_IMAGE_LAYOUT_GENERAL,
+ .newLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
+ .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+ .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+ .image = dst_image,
+ .subresourceRange{
+ .aspectMask = aspect_mask,
+ .baseMipLevel = 0,
+ .levelCount = VK_REMAINING_MIP_LEVELS,
+ .baseArrayLayer = 0,
+ .layerCount = VK_REMAINING_ARRAY_LAYERS,
+ },
+ },
+ };
+ VkImageMemoryBarrier write_barrier{
+ .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
+ .pNext = nullptr,
+ .srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
+ .dstAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT |
+ VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT |
+ VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT |
+ VK_ACCESS_TRANSFER_READ_BIT | VK_ACCESS_TRANSFER_WRITE_BIT,
+ .oldLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
+ .newLayout = VK_IMAGE_LAYOUT_GENERAL,
+ .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+ .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+ .image = dst_image,
+ .subresourceRange{
+ .aspectMask = aspect_mask,
+ .baseMipLevel = 0,
+ .levelCount = VK_REMAINING_MIP_LEVELS,
+ .baseArrayLayer = 0,
+ .layerCount = VK_REMAINING_ARRAY_LAYERS,
},
+ };
+ cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT,
+ 0, nullptr, nullptr, read_barriers);
+ if (is_resolve) {
+ cmdbuf.ResolveImage(src_image, VK_IMAGE_LAYOUT_GENERAL, dst_image,
+ VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
+ MakeImageResolve(dst_region, src_region, dst_layers, src_layers));
+ } else {
+ const bool is_linear = filter == Fermi2D::Filter::Bilinear;
+ const VkFilter vk_filter = is_linear ? VK_FILTER_LINEAR : VK_FILTER_NEAREST;
+ cmdbuf.BlitImage(
+ src_image, VK_IMAGE_LAYOUT_GENERAL, dst_image, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
+ MakeImageBlit(dst_region, src_region, dst_layers, src_layers), vk_filter);
+ }
+ cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
+ 0, write_barrier);
});
-
- return last_image_view = *image_view;
}
-VkImageView CachedSurfaceView::GetAttachment() {
- if (render_target) {
- return *render_target;
+void TextureCacheRuntime::ConvertImage(Framebuffer* dst, ImageView& dst_view, ImageView& src_view) {
+ switch (dst_view.format) {
+ case PixelFormat::R16_UNORM:
+ if (src_view.format == PixelFormat::D16_UNORM) {
+ return blit_image_helper.ConvertD16ToR16(dst, src_view);
+ }
+ break;
+ case PixelFormat::R32_FLOAT:
+ if (src_view.format == PixelFormat::D32_FLOAT) {
+ return blit_image_helper.ConvertD32ToR32(dst, src_view);
+ }
+ break;
+ case PixelFormat::D16_UNORM:
+ if (src_view.format == PixelFormat::R16_UNORM) {
+ return blit_image_helper.ConvertR16ToD16(dst, src_view);
+ }
+ break;
+ case PixelFormat::D32_FLOAT:
+ if (src_view.format == PixelFormat::R32_FLOAT) {
+ return blit_image_helper.ConvertR32ToD32(dst, src_view);
+ }
+ break;
+ default:
+ break;
}
+ UNIMPLEMENTED_MSG("Unimplemented format copy from {} to {}", src_view.format, dst_view.format);
+}
- VkImageViewCreateInfo ci{
- .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
- .pNext = nullptr,
- .flags = 0,
- .image = surface.GetImageHandle(),
- .viewType = VK_IMAGE_VIEW_TYPE_1D,
- .format = surface.GetImage().GetFormat(),
- .components =
- {
- .r = VK_COMPONENT_SWIZZLE_IDENTITY,
- .g = VK_COMPONENT_SWIZZLE_IDENTITY,
- .b = VK_COMPONENT_SWIZZLE_IDENTITY,
- .a = VK_COMPONENT_SWIZZLE_IDENTITY,
+void TextureCacheRuntime::CopyImage(Image& dst, Image& src,
+ std::span<const VideoCommon::ImageCopy> copies) {
+ std::vector<VkImageCopy> vk_copies(copies.size());
+ const VkImageAspectFlags aspect_mask = dst.AspectMask();
+ ASSERT(aspect_mask == src.AspectMask());
+
+ std::ranges::transform(copies, vk_copies.begin(), [aspect_mask](const auto& copy) {
+ return MakeImageCopy(copy, aspect_mask);
+ });
+ const VkImage dst_image = dst.Handle();
+ const VkImage src_image = src.Handle();
+ scheduler.RequestOutsideRenderPassOperationContext();
+ scheduler.Record([dst_image, src_image, aspect_mask, vk_copies](vk::CommandBuffer cmdbuf) {
+ RangedBarrierRange dst_range;
+ RangedBarrierRange src_range;
+ for (const VkImageCopy& copy : vk_copies) {
+ dst_range.AddLayers(copy.dstSubresource);
+ src_range.AddLayers(copy.srcSubresource);
+ }
+ const std::array read_barriers{
+ VkImageMemoryBarrier{
+ .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
+ .pNext = nullptr,
+ .srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
+ VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT |
+ VK_ACCESS_TRANSFER_WRITE_BIT,
+ .dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
+ .oldLayout = VK_IMAGE_LAYOUT_GENERAL,
+ .newLayout = VK_IMAGE_LAYOUT_GENERAL,
+ .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+ .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+ .image = src_image,
+ .subresourceRange = src_range.SubresourceRange(aspect_mask),
},
- .subresourceRange =
- {
- .aspectMask = aspect_mask,
- .baseMipLevel = base_level,
- .levelCount = num_levels,
- .baseArrayLayer = 0,
- .layerCount = 0,
+ VkImageMemoryBarrier{
+ .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
+ .pNext = nullptr,
+ .srcAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT |
+ VK_ACCESS_COLOR_ATTACHMENT_READ_BIT |
+ VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
+ VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT |
+ VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT |
+ VK_ACCESS_TRANSFER_READ_BIT | VK_ACCESS_TRANSFER_WRITE_BIT,
+ .dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
+ .oldLayout = VK_IMAGE_LAYOUT_GENERAL,
+ .newLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
+ .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+ .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+ .image = dst_image,
+ .subresourceRange = dst_range.SubresourceRange(aspect_mask),
},
- };
- if (image_view_type == VK_IMAGE_VIEW_TYPE_3D) {
- ci.viewType = num_slices > 1 ? VK_IMAGE_VIEW_TYPE_2D_ARRAY : VK_IMAGE_VIEW_TYPE_2D;
- ci.subresourceRange.baseArrayLayer = base_slice;
- ci.subresourceRange.layerCount = num_slices;
+ };
+ const VkImageMemoryBarrier write_barrier{
+ .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
+ .pNext = nullptr,
+ .srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
+ .dstAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT |
+ VK_ACCESS_COLOR_ATTACHMENT_READ_BIT |
+ VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
+ VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT |
+ VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT |
+ VK_ACCESS_TRANSFER_READ_BIT | VK_ACCESS_TRANSFER_WRITE_BIT,
+ .oldLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
+ .newLayout = VK_IMAGE_LAYOUT_GENERAL,
+ .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+ .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+ .image = dst_image,
+ .subresourceRange = dst_range.SubresourceRange(aspect_mask),
+ };
+ cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT,
+ 0, {}, {}, read_barriers);
+ cmdbuf.CopyImage(src_image, VK_IMAGE_LAYOUT_GENERAL, dst_image,
+ VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, vk_copies);
+ cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
+ 0, write_barrier);
+ });
+}
+
+Image::Image(TextureCacheRuntime& runtime, const ImageInfo& info_, GPUVAddr gpu_addr_,
+ VAddr cpu_addr_)
+ : VideoCommon::ImageBase(info_, gpu_addr_, cpu_addr_), scheduler{&runtime.scheduler},
+ image(MakeImage(runtime.device, info)), buffer(MakeBuffer(runtime.device, info)),
+ aspect_mask(ImageAspectMask(info.format)) {
+ if (image) {
+ commit = runtime.memory_manager.Commit(image, false);
} else {
- ci.viewType = image_view_type;
- ci.subresourceRange.baseArrayLayer = base_layer;
- ci.subresourceRange.layerCount = num_layers;
+ commit = runtime.memory_manager.Commit(buffer, false);
+ }
+ if (IsPixelFormatASTC(info.format) && !runtime.device.IsOptimalAstcSupported()) {
+ flags |= VideoCommon::ImageFlagBits::Converted;
+ }
+ if (runtime.device.HasDebuggingToolAttached()) {
+ if (image) {
+ image.SetObjectNameEXT(VideoCommon::Name(*this).c_str());
+ } else {
+ buffer.SetObjectNameEXT(VideoCommon::Name(*this).c_str());
+ }
}
- render_target = device.GetLogical().CreateImageView(ci);
- return *render_target;
}
-VKTextureCache::VKTextureCache(VideoCore::RasterizerInterface& rasterizer_,
- Tegra::Engines::Maxwell3D& maxwell3d_,
- Tegra::MemoryManager& gpu_memory_, const VKDevice& device_,
- VKMemoryManager& memory_manager_, VKScheduler& scheduler_,
- VKStagingBufferPool& staging_pool_)
- : TextureCache(rasterizer_, maxwell3d_, gpu_memory_, device_.IsOptimalAstcSupported()),
- device{device_}, memory_manager{memory_manager_}, scheduler{scheduler_}, staging_pool{
- staging_pool_} {}
-
-VKTextureCache::~VKTextureCache() = default;
-
-Surface VKTextureCache::CreateSurface(GPUVAddr gpu_addr, const SurfaceParams& params) {
- return std::make_shared<CachedSurface>(device, memory_manager, scheduler, staging_pool,
- gpu_addr, params);
+void Image::UploadMemory(const ImageBufferMap& map, size_t buffer_offset,
+ std::span<const BufferImageCopy> copies) {
+ // TODO: Move this to another API
+ scheduler->RequestOutsideRenderPassOperationContext();
+ std::vector vk_copies = TransformBufferImageCopies(copies, buffer_offset, aspect_mask);
+ const VkBuffer src_buffer = map.handle;
+ const VkImage vk_image = *image;
+ const VkImageAspectFlags vk_aspect_mask = aspect_mask;
+ const bool is_initialized = std::exchange(initialized, true);
+ scheduler->Record([src_buffer, vk_image, vk_aspect_mask, is_initialized,
+ vk_copies](vk::CommandBuffer cmdbuf) {
+ CopyBufferToImage(cmdbuf, src_buffer, vk_image, vk_aspect_mask, is_initialized, vk_copies);
+ });
}
-void VKTextureCache::ImageCopy(Surface& src_surface, Surface& dst_surface,
- const VideoCommon::CopyParams& copy_params) {
- const bool src_3d = src_surface->GetSurfaceParams().target == SurfaceTarget::Texture3D;
- const bool dst_3d = dst_surface->GetSurfaceParams().target == SurfaceTarget::Texture3D;
- UNIMPLEMENTED_IF(src_3d);
+void Image::UploadMemory(const ImageBufferMap& map, size_t buffer_offset,
+ std::span<const VideoCommon::BufferCopy> copies) {
+ // TODO: Move this to another API
+ scheduler->RequestOutsideRenderPassOperationContext();
+ std::vector vk_copies = TransformBufferCopies(copies, buffer_offset);
+ const VkBuffer src_buffer = map.handle;
+ const VkBuffer dst_buffer = *buffer;
+ scheduler->Record([src_buffer, dst_buffer, vk_copies](vk::CommandBuffer cmdbuf) {
+ // TODO: Barriers
+ cmdbuf.CopyBuffer(src_buffer, dst_buffer, vk_copies);
+ });
+}
- // The texture cache handles depth in OpenGL terms, we have to handle it as subresource and
- // dimension respectively.
- const u32 dst_base_layer = dst_3d ? 0 : copy_params.dest_z;
- const u32 dst_offset_z = dst_3d ? copy_params.dest_z : 0;
+void Image::DownloadMemory(const ImageBufferMap& map, size_t buffer_offset,
+ std::span<const BufferImageCopy> copies) {
+ std::vector vk_copies = TransformBufferImageCopies(copies, buffer_offset, aspect_mask);
+ scheduler->Record([buffer = map.handle, image = *image, aspect_mask = aspect_mask,
+ vk_copies](vk::CommandBuffer cmdbuf) {
+ // TODO: Barriers
+ cmdbuf.CopyImageToBuffer(image, VK_IMAGE_LAYOUT_GENERAL, buffer, vk_copies);
+ });
+}
- const u32 extent_z = dst_3d ? copy_params.depth : 1;
- const u32 num_layers = dst_3d ? 1 : copy_params.depth;
+ImageView::ImageView(TextureCacheRuntime& runtime, const VideoCommon::ImageViewInfo& info,
+ ImageId image_id_, Image& image)
+ : VideoCommon::ImageViewBase{info, image.info, image_id_}, device{&runtime.device},
+ image_handle{image.Handle()}, image_format{image.info.format}, samples{ConvertSampleCount(
+ image.info.num_samples)} {
+ const VkImageAspectFlags aspect_mask = ImageViewAspectMask(info);
+ std::array<SwizzleSource, 4> swizzle{
+ SwizzleSource::R,
+ SwizzleSource::G,
+ SwizzleSource::B,
+ SwizzleSource::A,
+ };
+ if (!info.IsRenderTarget()) {
+ swizzle = info.Swizzle();
+ if ((aspect_mask & (VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT)) != 0) {
+ std::ranges::transform(swizzle, swizzle.begin(), ConvertGreenRed);
+ }
+ }
+ const VkFormat vk_format =
+ MaxwellToVK::SurfaceFormat(*device, FormatType::Optimal, format).format;
+ const VkImageViewCreateInfo create_info{
+ .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
+ .pNext = nullptr,
+ .flags = 0,
+ .image = image.Handle(),
+ .viewType = VkImageViewType{},
+ .format = vk_format,
+ .components{
+ .r = ComponentSwizzle(swizzle[0]),
+ .g = ComponentSwizzle(swizzle[1]),
+ .b = ComponentSwizzle(swizzle[2]),
+ .a = ComponentSwizzle(swizzle[3]),
+ },
+ .subresourceRange = MakeSubresourceRange(aspect_mask, info.range),
+ };
+ const auto create = [&](VideoCommon::ImageViewType view_type, std::optional<u32> num_layers) {
+ VkImageViewCreateInfo ci{create_info};
+ ci.viewType = ImageViewType(view_type);
+ if (num_layers) {
+ ci.subresourceRange.layerCount = *num_layers;
+ }
+ vk::ImageView handle = device->GetLogical().CreateImageView(ci);
+ if (device->HasDebuggingToolAttached()) {
+ handle.SetObjectNameEXT(VideoCommon::Name(*this, view_type).c_str());
+ }
+ image_views[static_cast<size_t>(view_type)] = std::move(handle);
+ };
+ switch (info.type) {
+ case VideoCommon::ImageViewType::e1D:
+ case VideoCommon::ImageViewType::e1DArray:
+ create(VideoCommon::ImageViewType::e1D, 1);
+ create(VideoCommon::ImageViewType::e1DArray, std::nullopt);
+ render_target = Handle(VideoCommon::ImageViewType::e1DArray);
+ break;
+ case VideoCommon::ImageViewType::e2D:
+ case VideoCommon::ImageViewType::e2DArray:
+ create(VideoCommon::ImageViewType::e2D, 1);
+ create(VideoCommon::ImageViewType::e2DArray, std::nullopt);
+ render_target = Handle(VideoCommon::ImageViewType::e2DArray);
+ break;
+ case VideoCommon::ImageViewType::e3D:
+ create(VideoCommon::ImageViewType::e3D, std::nullopt);
+ render_target = Handle(VideoCommon::ImageViewType::e3D);
+ break;
+ case VideoCommon::ImageViewType::Cube:
+ case VideoCommon::ImageViewType::CubeArray:
+ create(VideoCommon::ImageViewType::Cube, 6);
+ create(VideoCommon::ImageViewType::CubeArray, std::nullopt);
+ break;
+ case VideoCommon::ImageViewType::Rect:
+ UNIMPLEMENTED();
+ break;
+ case VideoCommon::ImageViewType::Buffer:
+ buffer_view = device->GetLogical().CreateBufferView(VkBufferViewCreateInfo{
+ .sType = VK_STRUCTURE_TYPE_BUFFER_VIEW_CREATE_INFO,
+ .pNext = nullptr,
+ .flags = 0,
+ .buffer = image.Buffer(),
+ .format = vk_format,
+ .offset = 0, // TODO: Redesign buffer cache to support this
+ .range = image.guest_size_bytes,
+ });
+ break;
+ }
+}
- // We can't copy inside a renderpass
- scheduler.RequestOutsideRenderPassOperationContext();
+ImageView::ImageView(TextureCacheRuntime&, const VideoCommon::NullImageParams& params)
+ : VideoCommon::ImageViewBase{params} {}
- src_surface->Transition(copy_params.source_z, copy_params.depth, copy_params.source_level, 1,
- VK_PIPELINE_STAGE_TRANSFER_BIT, VK_ACCESS_TRANSFER_READ_BIT,
- VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL);
- dst_surface->Transition(dst_base_layer, num_layers, copy_params.dest_level, 1,
- VK_PIPELINE_STAGE_TRANSFER_BIT, VK_ACCESS_TRANSFER_WRITE_BIT,
- VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL);
+VkImageView ImageView::DepthView() {
+ if (depth_view) {
+ return *depth_view;
+ }
+ depth_view = MakeDepthStencilView(VK_IMAGE_ASPECT_DEPTH_BIT);
+ return *depth_view;
+}
- const VkImageCopy copy{
- .srcSubresource =
- {
- .aspectMask = src_surface->GetAspectMask(),
- .mipLevel = copy_params.source_level,
- .baseArrayLayer = copy_params.source_z,
- .layerCount = num_layers,
- },
- .srcOffset =
- {
- .x = static_cast<s32>(copy_params.source_x),
- .y = static_cast<s32>(copy_params.source_y),
- .z = 0,
- },
- .dstSubresource =
- {
- .aspectMask = dst_surface->GetAspectMask(),
- .mipLevel = copy_params.dest_level,
- .baseArrayLayer = dst_base_layer,
- .layerCount = num_layers,
- },
- .dstOffset =
- {
- .x = static_cast<s32>(copy_params.dest_x),
- .y = static_cast<s32>(copy_params.dest_y),
- .z = static_cast<s32>(dst_offset_z),
- },
- .extent =
- {
- .width = copy_params.width,
- .height = copy_params.height,
- .depth = extent_z,
- },
- };
+VkImageView ImageView::StencilView() {
+ if (stencil_view) {
+ return *stencil_view;
+ }
+ stencil_view = MakeDepthStencilView(VK_IMAGE_ASPECT_STENCIL_BIT);
+ return *stencil_view;
+}
- const VkImage src_image = src_surface->GetImageHandle();
- const VkImage dst_image = dst_surface->GetImageHandle();
- scheduler.Record([src_image, dst_image, copy](vk::CommandBuffer cmdbuf) {
- cmdbuf.CopyImage(src_image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, dst_image,
- VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, copy);
+vk::ImageView ImageView::MakeDepthStencilView(VkImageAspectFlags aspect_mask) {
+ return device->GetLogical().CreateImageView({
+ .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
+ .pNext = nullptr,
+ .flags = 0,
+ .image = image_handle,
+ .viewType = ImageViewType(type),
+ .format = MaxwellToVK::SurfaceFormat(*device, FormatType::Optimal, format).format,
+ .components{
+ .r = VK_COMPONENT_SWIZZLE_IDENTITY,
+ .g = VK_COMPONENT_SWIZZLE_IDENTITY,
+ .b = VK_COMPONENT_SWIZZLE_IDENTITY,
+ .a = VK_COMPONENT_SWIZZLE_IDENTITY,
+ },
+ .subresourceRange = MakeSubresourceRange(aspect_mask, range),
});
}
-void VKTextureCache::ImageBlit(View& src_view, View& dst_view,
- const Tegra::Engines::Fermi2D::Config& copy_config) {
- // We can't blit inside a renderpass
- scheduler.RequestOutsideRenderPassOperationContext();
-
- src_view->Transition(VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, VK_PIPELINE_STAGE_TRANSFER_BIT,
- VK_ACCESS_TRANSFER_READ_BIT);
- dst_view->Transition(VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, VK_PIPELINE_STAGE_TRANSFER_BIT,
- VK_ACCESS_TRANSFER_WRITE_BIT);
-
- VkImageBlit blit;
- blit.srcSubresource = src_view->GetImageSubresourceLayers();
- blit.srcOffsets[0].x = copy_config.src_rect.left;
- blit.srcOffsets[0].y = copy_config.src_rect.top;
- blit.srcOffsets[0].z = 0;
- blit.srcOffsets[1].x = copy_config.src_rect.right;
- blit.srcOffsets[1].y = copy_config.src_rect.bottom;
- blit.srcOffsets[1].z = 1;
- blit.dstSubresource = dst_view->GetImageSubresourceLayers();
- blit.dstOffsets[0].x = copy_config.dst_rect.left;
- blit.dstOffsets[0].y = copy_config.dst_rect.top;
- blit.dstOffsets[0].z = 0;
- blit.dstOffsets[1].x = copy_config.dst_rect.right;
- blit.dstOffsets[1].y = copy_config.dst_rect.bottom;
- blit.dstOffsets[1].z = 1;
-
- const bool is_linear = copy_config.filter == Tegra::Engines::Fermi2D::Filter::Linear;
-
- scheduler.Record([src_image = src_view->GetImage(), dst_image = dst_view->GetImage(), blit,
- is_linear](vk::CommandBuffer cmdbuf) {
- cmdbuf.BlitImage(src_image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, dst_image,
- VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, blit,
- is_linear ? VK_FILTER_LINEAR : VK_FILTER_NEAREST);
+Sampler::Sampler(TextureCacheRuntime& runtime, const Tegra::Texture::TSCEntry& tsc) {
+ const auto& device = runtime.device;
+ const bool arbitrary_borders = runtime.device.IsExtCustomBorderColorSupported();
+ const std::array<float, 4> color = tsc.BorderColor();
+ // C++20 bit_cast
+ VkClearColorValue border_color;
+ std::memcpy(&border_color, &color, sizeof(color));
+ const VkSamplerCustomBorderColorCreateInfoEXT border_ci{
+ .sType = VK_STRUCTURE_TYPE_SAMPLER_CUSTOM_BORDER_COLOR_CREATE_INFO_EXT,
+ .pNext = nullptr,
+ .customBorderColor = border_color,
+ .format = VK_FORMAT_UNDEFINED,
+ };
+ const void* pnext = nullptr;
+ if (arbitrary_borders) {
+ pnext = &border_ci;
+ }
+ const VkSamplerReductionModeCreateInfoEXT reduction_ci{
+ .sType = VK_STRUCTURE_TYPE_SAMPLER_REDUCTION_MODE_CREATE_INFO_EXT,
+ .pNext = pnext,
+ .reductionMode = MaxwellToVK::SamplerReduction(tsc.reduction_filter),
+ };
+ if (runtime.device.IsExtSamplerFilterMinmaxSupported()) {
+ pnext = &reduction_ci;
+ } else if (reduction_ci.reductionMode != VK_SAMPLER_REDUCTION_MODE_WEIGHTED_AVERAGE_EXT) {
+ LOG_WARNING(Render_Vulkan, "VK_EXT_sampler_filter_minmax is required");
+ }
+ // Some games have samplers with garbage. Sanitize them here.
+ const float max_anisotropy = std::clamp(tsc.MaxAnisotropy(), 1.0f, 16.0f);
+ sampler = device.GetLogical().CreateSampler(VkSamplerCreateInfo{
+ .sType = VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO,
+ .pNext = pnext,
+ .flags = 0,
+ .magFilter = MaxwellToVK::Sampler::Filter(tsc.mag_filter),
+ .minFilter = MaxwellToVK::Sampler::Filter(tsc.min_filter),
+ .mipmapMode = MaxwellToVK::Sampler::MipmapMode(tsc.mipmap_filter),
+ .addressModeU = MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_u, tsc.mag_filter),
+ .addressModeV = MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_v, tsc.mag_filter),
+ .addressModeW = MaxwellToVK::Sampler::WrapMode(device, tsc.wrap_p, tsc.mag_filter),
+ .mipLodBias = tsc.LodBias(),
+ .anisotropyEnable = static_cast<VkBool32>(max_anisotropy > 1.0f ? VK_TRUE : VK_FALSE),
+ .maxAnisotropy = max_anisotropy,
+ .compareEnable = tsc.depth_compare_enabled,
+ .compareOp = MaxwellToVK::Sampler::DepthCompareFunction(tsc.depth_compare_func),
+ .minLod = tsc.mipmap_filter == TextureMipmapFilter::None ? 0.0f : tsc.MinLod(),
+ .maxLod = tsc.mipmap_filter == TextureMipmapFilter::None ? 0.25f : tsc.MaxLod(),
+ .borderColor =
+ arbitrary_borders ? VK_BORDER_COLOR_INT_CUSTOM_EXT : ConvertBorderColor(color),
+ .unnormalizedCoordinates = VK_FALSE,
});
}
-void VKTextureCache::BufferCopy(Surface& src_surface, Surface& dst_surface) {
- // Currently unimplemented. PBO copies should be dropped and we should use a render pass to
- // convert from color to depth and viceversa.
- LOG_WARNING(Render_Vulkan, "Unimplemented");
+Framebuffer::Framebuffer(TextureCacheRuntime& runtime, std::span<ImageView*, NUM_RT> color_buffers,
+ ImageView* depth_buffer, const VideoCommon::RenderTargets& key) {
+ std::vector<VkAttachmentDescription> descriptions;
+ std::vector<VkImageView> attachments;
+ RenderPassKey renderpass_key{};
+ s32 num_layers = 1;
+
+ for (size_t index = 0; index < NUM_RT; ++index) {
+ const ImageView* const color_buffer = color_buffers[index];
+ if (!color_buffer) {
+ renderpass_key.color_formats[index] = PixelFormat::Invalid;
+ continue;
+ }
+ descriptions.push_back(AttachmentDescription(runtime.device, color_buffer));
+ attachments.push_back(color_buffer->RenderTarget());
+ renderpass_key.color_formats[index] = color_buffer->format;
+ num_layers = std::max(num_layers, color_buffer->range.extent.layers);
+ images[num_images] = color_buffer->ImageHandle();
+ image_ranges[num_images] = MakeSubresourceRange(color_buffer);
+ samples = color_buffer->Samples();
+ ++num_images;
+ }
+ const size_t num_colors = attachments.size();
+ const VkAttachmentReference* depth_attachment =
+ depth_buffer ? &ATTACHMENT_REFERENCES[num_colors] : nullptr;
+ if (depth_buffer) {
+ descriptions.push_back(AttachmentDescription(runtime.device, depth_buffer));
+ attachments.push_back(depth_buffer->RenderTarget());
+ renderpass_key.depth_format = depth_buffer->format;
+ num_layers = std::max(num_layers, depth_buffer->range.extent.layers);
+ images[num_images] = depth_buffer->ImageHandle();
+ image_ranges[num_images] = MakeSubresourceRange(depth_buffer);
+ samples = depth_buffer->Samples();
+ ++num_images;
+ } else {
+ renderpass_key.depth_format = PixelFormat::Invalid;
+ }
+ renderpass_key.samples = samples;
+
+ const auto& device = runtime.device.GetLogical();
+ const auto [cache_pair, is_new] = runtime.renderpass_cache.try_emplace(renderpass_key);
+ if (is_new) {
+ const VkSubpassDescription subpass{
+ .flags = 0,
+ .pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS,
+ .inputAttachmentCount = 0,
+ .pInputAttachments = nullptr,
+ .colorAttachmentCount = static_cast<u32>(num_colors),
+ .pColorAttachments = num_colors != 0 ? ATTACHMENT_REFERENCES.data() : nullptr,
+ .pResolveAttachments = nullptr,
+ .pDepthStencilAttachment = depth_attachment,
+ .preserveAttachmentCount = 0,
+ .pPreserveAttachments = nullptr,
+ };
+ cache_pair->second = device.CreateRenderPass(VkRenderPassCreateInfo{
+ .sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO,
+ .pNext = nullptr,
+ .flags = 0,
+ .attachmentCount = static_cast<u32>(descriptions.size()),
+ .pAttachments = descriptions.data(),
+ .subpassCount = 1,
+ .pSubpasses = &subpass,
+ .dependencyCount = 0,
+ .pDependencies = nullptr,
+ });
+ }
+ renderpass = *cache_pair->second;
+ render_area = VkExtent2D{
+ .width = key.size.width,
+ .height = key.size.height,
+ };
+ num_color_buffers = static_cast<u32>(num_colors);
+ framebuffer = device.CreateFramebuffer(VkFramebufferCreateInfo{
+ .sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO,
+ .pNext = nullptr,
+ .flags = 0,
+ .renderPass = renderpass,
+ .attachmentCount = static_cast<u32>(attachments.size()),
+ .pAttachments = attachments.data(),
+ .width = key.size.width,
+ .height = key.size.height,
+ .layers = static_cast<u32>(num_layers),
+ });
+ if (runtime.device.HasDebuggingToolAttached()) {
+ framebuffer.SetObjectNameEXT(VideoCommon::Name(key).c_str());
+ }
}
} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.h b/src/video_core/renderer_vulkan/vk_texture_cache.h
index b0be4cb0f..edc3d80c0 100644
--- a/src/video_core/renderer_vulkan/vk_texture_cache.h
+++ b/src/video_core/renderer_vulkan/vk_texture_cache.h
@@ -4,217 +4,265 @@
#pragma once
-#include <memory>
-#include <unordered_map>
+#include <compare>
+#include <span>
-#include "common/common_types.h"
-#include "video_core/renderer_vulkan/vk_image.h"
#include "video_core/renderer_vulkan/vk_memory_manager.h"
-#include "video_core/renderer_vulkan/vk_scheduler.h"
#include "video_core/renderer_vulkan/wrapper.h"
-#include "video_core/texture_cache/surface_base.h"
#include "video_core/texture_cache/texture_cache.h"
-namespace VideoCore {
-class RasterizerInterface;
-}
-
namespace Vulkan {
-class RasterizerVulkan;
+using VideoCommon::ImageId;
+using VideoCommon::NUM_RT;
+using VideoCommon::Offset2D;
+using VideoCommon::RenderTargets;
+using VideoCore::Surface::PixelFormat;
+
class VKDevice;
class VKScheduler;
class VKStagingBufferPool;
-class CachedSurfaceView;
-class CachedSurface;
+class BlitImageHelper;
+class Image;
+class ImageView;
+class Framebuffer;
-using Surface = std::shared_ptr<CachedSurface>;
-using View = std::shared_ptr<CachedSurfaceView>;
-using TextureCacheBase = VideoCommon::TextureCache<Surface, View>;
+struct RenderPassKey {
+ constexpr auto operator<=>(const RenderPassKey&) const noexcept = default;
-using VideoCommon::SurfaceParams;
-using VideoCommon::ViewParams;
+ std::array<PixelFormat, NUM_RT> color_formats;
+ PixelFormat depth_format;
+ VkSampleCountFlagBits samples;
+};
-class CachedSurface final : public VideoCommon::SurfaceBase<View> {
- friend CachedSurfaceView;
+} // namespace Vulkan
-public:
- explicit CachedSurface(const VKDevice& device_, VKMemoryManager& memory_manager_,
- VKScheduler& scheduler_, VKStagingBufferPool& staging_pool_,
- GPUVAddr gpu_addr_, const SurfaceParams& params_);
- ~CachedSurface();
+namespace std {
+template <>
+struct hash<Vulkan::RenderPassKey> {
+ [[nodiscard]] constexpr size_t operator()(const Vulkan::RenderPassKey& key) const noexcept {
+ size_t value = static_cast<size_t>(key.depth_format) << 48;
+ value ^= static_cast<size_t>(key.samples) << 52;
+ for (size_t i = 0; i < key.color_formats.size(); ++i) {
+ value ^= static_cast<size_t>(key.color_formats[i]) << (i * 6);
+ }
+ return value;
+ }
+};
+} // namespace std
- void UploadTexture(const std::vector<u8>& staging_buffer) override;
- void DownloadTexture(std::vector<u8>& staging_buffer) override;
+namespace Vulkan {
- void FullTransition(VkPipelineStageFlags new_stage_mask, VkAccessFlags new_access,
- VkImageLayout new_layout) {
- image->Transition(0, static_cast<u32>(params.GetNumLayers()), 0, params.num_levels,
- new_stage_mask, new_access, new_layout);
+struct ImageBufferMap {
+ [[nodiscard]] VkBuffer Handle() const noexcept {
+ return handle;
}
- void Transition(u32 base_layer, u32 num_layers, u32 base_level, u32 num_levels,
- VkPipelineStageFlags new_stage_mask, VkAccessFlags new_access,
- VkImageLayout new_layout) {
- image->Transition(base_layer, num_layers, base_level, num_levels, new_stage_mask,
- new_access, new_layout);
+ [[nodiscard]] std::span<u8> Span() const noexcept {
+ return map.Span();
}
- VKImage& GetImage() {
- return *image;
- }
+ VkBuffer handle;
+ MemoryMap map;
+};
- const VKImage& GetImage() const {
- return *image;
- }
+struct TextureCacheRuntime {
+ const VKDevice& device;
+ VKScheduler& scheduler;
+ VKMemoryManager& memory_manager;
+ VKStagingBufferPool& staging_buffer_pool;
+ BlitImageHelper& blit_image_helper;
+ std::unordered_map<RenderPassKey, vk::RenderPass> renderpass_cache;
+
+ void Finish();
- VkImage GetImageHandle() const {
- return *image->GetHandle();
+ [[nodiscard]] ImageBufferMap MapUploadBuffer(size_t size);
+
+ [[nodiscard]] ImageBufferMap MapDownloadBuffer(size_t size) {
+ // TODO: Have a special function for this
+ return MapUploadBuffer(size);
}
- VkImageAspectFlags GetAspectMask() const {
- return image->GetAspectMask();
+ void BlitImage(Framebuffer* dst_framebuffer, ImageView& dst, ImageView& src,
+ const std::array<Offset2D, 2>& dst_region,
+ const std::array<Offset2D, 2>& src_region,
+ Tegra::Engines::Fermi2D::Filter filter,
+ Tegra::Engines::Fermi2D::Operation operation);
+
+ void CopyImage(Image& dst, Image& src, std::span<const VideoCommon::ImageCopy> copies);
+
+ void ConvertImage(Framebuffer* dst, ImageView& dst_view, ImageView& src_view);
+
+ [[nodiscard]] bool CanAccelerateImageUpload(Image&) const noexcept {
+ return false;
}
- VkBufferView GetBufferViewHandle() const {
- return *buffer_view;
+ void AccelerateImageUpload(Image&, const ImageBufferMap&, size_t,
+ std::span<const VideoCommon::SwizzleParameters>) {
+ UNREACHABLE();
}
-protected:
- void DecorateSurfaceName() override;
+ void InsertUploadMemoryBarrier() {}
+};
- View CreateView(const ViewParams& view_params) override;
+class Image : public VideoCommon::ImageBase {
+public:
+ explicit Image(TextureCacheRuntime&, const VideoCommon::ImageInfo& info, GPUVAddr gpu_addr,
+ VAddr cpu_addr);
-private:
- void UploadBuffer(const std::vector<u8>& staging_buffer);
+ void UploadMemory(const ImageBufferMap& map, size_t buffer_offset,
+ std::span<const VideoCommon::BufferImageCopy> copies);
- void UploadImage(const std::vector<u8>& staging_buffer);
+ void UploadMemory(const ImageBufferMap& map, size_t buffer_offset,
+ std::span<const VideoCommon::BufferCopy> copies);
- VkBufferImageCopy GetBufferImageCopy(u32 level) const;
+ void DownloadMemory(const ImageBufferMap& map, size_t buffer_offset,
+ std::span<const VideoCommon::BufferImageCopy> copies);
- VkImageSubresourceRange GetImageSubresourceRange() const;
+ [[nodiscard]] VkImage Handle() const noexcept {
+ return *image;
+ }
- const VKDevice& device;
- VKMemoryManager& memory_manager;
- VKScheduler& scheduler;
- VKStagingBufferPool& staging_pool;
+ [[nodiscard]] VkBuffer Buffer() const noexcept {
+ return *buffer;
+ }
+
+ [[nodiscard]] VkImageCreateFlags AspectMask() const noexcept {
+ return aspect_mask;
+ }
- std::optional<VKImage> image;
+private:
+ VKScheduler* scheduler;
+ vk::Image image;
vk::Buffer buffer;
- vk::BufferView buffer_view;
VKMemoryCommit commit;
-
- VkFormat format = VK_FORMAT_UNDEFINED;
+ VkImageAspectFlags aspect_mask = 0;
+ bool initialized = false;
};
-class CachedSurfaceView final : public VideoCommon::ViewBase {
+class ImageView : public VideoCommon::ImageViewBase {
public:
- explicit CachedSurfaceView(const VKDevice& device_, CachedSurface& surface_,
- const ViewParams& view_params_);
- ~CachedSurfaceView();
+ explicit ImageView(TextureCacheRuntime&, const VideoCommon::ImageViewInfo&, ImageId, Image&);
+ explicit ImageView(TextureCacheRuntime&, const VideoCommon::NullImageParams&);
- VkImageView GetImageView(Tegra::Texture::SwizzleSource x_source,
- Tegra::Texture::SwizzleSource y_source,
- Tegra::Texture::SwizzleSource z_source,
- Tegra::Texture::SwizzleSource w_source);
+ [[nodiscard]] VkImageView DepthView();
- VkImageView GetAttachment();
+ [[nodiscard]] VkImageView StencilView();
- bool IsSameSurface(const CachedSurfaceView& rhs) const {
- return &surface == &rhs.surface;
+ [[nodiscard]] VkImageView Handle(VideoCommon::ImageViewType query_type) const noexcept {
+ return *image_views[static_cast<size_t>(query_type)];
}
- u32 GetWidth() const {
- return surface_params.GetMipWidth(base_level);
+ [[nodiscard]] VkBufferView BufferView() const noexcept {
+ return *buffer_view;
}
- u32 GetHeight() const {
- return surface_params.GetMipHeight(base_level);
+ [[nodiscard]] VkImage ImageHandle() const noexcept {
+ return image_handle;
}
- u32 GetNumLayers() const {
- return num_layers;
+ [[nodiscard]] VkImageView RenderTarget() const noexcept {
+ return render_target;
}
- bool IsBufferView() const {
- return buffer_view;
+ [[nodiscard]] PixelFormat ImageFormat() const noexcept {
+ return image_format;
}
- VkImage GetImage() const {
- return image;
+ [[nodiscard]] VkSampleCountFlagBits Samples() const noexcept {
+ return samples;
}
- VkBufferView GetBufferView() const {
- return buffer_view;
- }
+private:
+ [[nodiscard]] vk::ImageView MakeDepthStencilView(VkImageAspectFlags aspect_mask);
- VkImageSubresourceRange GetImageSubresourceRange() const {
- return {aspect_mask, base_level, num_levels, base_layer, num_layers};
- }
+ const VKDevice* device = nullptr;
+ std::array<vk::ImageView, VideoCommon::NUM_IMAGE_VIEW_TYPES> image_views;
+ vk::ImageView depth_view;
+ vk::ImageView stencil_view;
+ vk::BufferView buffer_view;
+ VkImage image_handle = VK_NULL_HANDLE;
+ VkImageView render_target = VK_NULL_HANDLE;
+ PixelFormat image_format = PixelFormat::Invalid;
+ VkSampleCountFlagBits samples = VK_SAMPLE_COUNT_1_BIT;
+};
- VkImageSubresourceLayers GetImageSubresourceLayers() const {
- return {surface.GetAspectMask(), base_level, base_layer, num_layers};
- }
+class ImageAlloc : public VideoCommon::ImageAllocBase {};
- void Transition(VkImageLayout new_layout, VkPipelineStageFlags new_stage_mask,
- VkAccessFlags new_access) const {
- surface.Transition(base_layer, num_layers, base_level, num_levels, new_stage_mask,
- new_access, new_layout);
- }
+class Sampler {
+public:
+ explicit Sampler(TextureCacheRuntime&, const Tegra::Texture::TSCEntry&);
- void MarkAsModified(u64 tick) {
- surface.MarkAsModified(true, tick);
+ [[nodiscard]] VkSampler Handle() const noexcept {
+ return *sampler;
}
private:
- // Store a copy of these values to avoid double dereference when reading them
- const SurfaceParams surface_params;
- const VkImage image;
- const VkBufferView buffer_view;
- const VkImageAspectFlags aspect_mask;
-
- const VKDevice& device;
- CachedSurface& surface;
- const u32 base_level;
- const u32 num_levels;
- const VkImageViewType image_view_type;
- u32 base_layer = 0;
- u32 num_layers = 0;
- u32 base_slice = 0;
- u32 num_slices = 0;
-
- VkImageView last_image_view = nullptr;
- u32 last_swizzle = 0;
-
- vk::ImageView render_target;
- std::unordered_map<u32, vk::ImageView> view_cache;
+ vk::Sampler sampler;
};
-class VKTextureCache final : public TextureCacheBase {
+class Framebuffer {
public:
- explicit VKTextureCache(VideoCore::RasterizerInterface& rasterizer_,
- Tegra::Engines::Maxwell3D& maxwell3d_,
- Tegra::MemoryManager& gpu_memory_, const VKDevice& device_,
- VKMemoryManager& memory_manager_, VKScheduler& scheduler_,
- VKStagingBufferPool& staging_pool_);
- ~VKTextureCache();
+ explicit Framebuffer(TextureCacheRuntime&, std::span<ImageView*, NUM_RT> color_buffers,
+ ImageView* depth_buffer, const VideoCommon::RenderTargets& key);
-private:
- Surface CreateSurface(GPUVAddr gpu_addr, const SurfaceParams& params) override;
+ [[nodiscard]] VkFramebuffer Handle() const noexcept {
+ return *framebuffer;
+ }
- void ImageCopy(Surface& src_surface, Surface& dst_surface,
- const VideoCommon::CopyParams& copy_params) override;
+ [[nodiscard]] VkRenderPass RenderPass() const noexcept {
+ return renderpass;
+ }
- void ImageBlit(View& src_view, View& dst_view,
- const Tegra::Engines::Fermi2D::Config& copy_config) override;
+ [[nodiscard]] VkExtent2D RenderArea() const noexcept {
+ return render_area;
+ }
- void BufferCopy(Surface& src_surface, Surface& dst_surface) override;
+ [[nodiscard]] VkSampleCountFlagBits Samples() const noexcept {
+ return samples;
+ }
- const VKDevice& device;
- VKMemoryManager& memory_manager;
- VKScheduler& scheduler;
- VKStagingBufferPool& staging_pool;
+ [[nodiscard]] u32 NumColorBuffers() const noexcept {
+ return num_color_buffers;
+ }
+
+ [[nodiscard]] u32 NumImages() const noexcept {
+ return num_images;
+ }
+
+ [[nodiscard]] const std::array<VkImage, 9>& Images() const noexcept {
+ return images;
+ }
+
+ [[nodiscard]] const std::array<VkImageSubresourceRange, 9>& ImageRanges() const noexcept {
+ return image_ranges;
+ }
+
+private:
+ vk::Framebuffer framebuffer;
+ VkRenderPass renderpass{};
+ VkExtent2D render_area{};
+ VkSampleCountFlagBits samples = VK_SAMPLE_COUNT_1_BIT;
+ u32 num_color_buffers = 0;
+ u32 num_images = 0;
+ std::array<VkImage, 9> images{};
+ std::array<VkImageSubresourceRange, 9> image_ranges{};
+};
+
+struct TextureCacheParams {
+ static constexpr bool ENABLE_VALIDATION = true;
+ static constexpr bool FRAMEBUFFER_BLITS = false;
+ static constexpr bool HAS_EMULATED_COPIES = false;
+
+ using Runtime = Vulkan::TextureCacheRuntime;
+ using Image = Vulkan::Image;
+ using ImageAlloc = Vulkan::ImageAlloc;
+ using ImageView = Vulkan::ImageView;
+ using Sampler = Vulkan::Sampler;
+ using Framebuffer = Vulkan::Framebuffer;
};
+using TextureCache = VideoCommon::TextureCache<TextureCacheParams>;
+
} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_update_descriptor.h b/src/video_core/renderer_vulkan/vk_update_descriptor.h
index f7e3c9821..f098a8540 100644
--- a/src/video_core/renderer_vulkan/vk_update_descriptor.h
+++ b/src/video_core/renderer_vulkan/vk_update_descriptor.h
@@ -40,30 +40,34 @@ public:
void Send(VkDescriptorUpdateTemplateKHR update_template, VkDescriptorSet set);
- void AddSampledImage(VkSampler sampler, VkImageView image_view) {
- payload.emplace_back(VkDescriptorImageInfo{sampler, image_view, {}});
+ void AddSampledImage(VkImageView image_view, VkSampler sampler) {
+ payload.emplace_back(VkDescriptorImageInfo{
+ .sampler = sampler,
+ .imageView = image_view,
+ .imageLayout = VK_IMAGE_LAYOUT_GENERAL,
+ });
}
void AddImage(VkImageView image_view) {
- payload.emplace_back(VkDescriptorImageInfo{{}, image_view, {}});
+ payload.emplace_back(VkDescriptorImageInfo{
+ .sampler = VK_NULL_HANDLE,
+ .imageView = image_view,
+ .imageLayout = VK_IMAGE_LAYOUT_GENERAL,
+ });
}
- void AddBuffer(VkBuffer buffer, u64 offset, std::size_t size) {
- payload.emplace_back(VkDescriptorBufferInfo{buffer, offset, size});
+ void AddBuffer(VkBuffer buffer, u64 offset, size_t size) {
+ payload.emplace_back(VkDescriptorBufferInfo{
+ .buffer = buffer,
+ .offset = offset,
+ .range = size,
+ });
}
void AddTexelBuffer(VkBufferView texel_buffer) {
payload.emplace_back(texel_buffer);
}
- VkImageLayout* LastImageLayout() {
- return &payload.back().image.imageLayout;
- }
-
- const VkImageLayout* LastImageLayout() const {
- return &payload.back().image.imageLayout;
- }
-
private:
const VKDevice& device;
VKScheduler& scheduler;
diff --git a/src/video_core/renderer_vulkan/wrapper.cpp b/src/video_core/renderer_vulkan/wrapper.cpp
index 1eced809e..2a21e850d 100644
--- a/src/video_core/renderer_vulkan/wrapper.cpp
+++ b/src/video_core/renderer_vulkan/wrapper.cpp
@@ -81,6 +81,7 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
X(vkCmdBeginQuery);
X(vkCmdBeginRenderPass);
X(vkCmdBeginTransformFeedbackEXT);
+ X(vkCmdBeginDebugUtilsLabelEXT);
X(vkCmdBindDescriptorSets);
X(vkCmdBindIndexBuffer);
X(vkCmdBindPipeline);
@@ -98,6 +99,7 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
X(vkCmdEndQuery);
X(vkCmdEndRenderPass);
X(vkCmdEndTransformFeedbackEXT);
+ X(vkCmdEndDebugUtilsLabelEXT);
X(vkCmdFillBuffer);
X(vkCmdPipelineBarrier);
X(vkCmdPushConstants);
@@ -121,6 +123,7 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
X(vkCmdSetPrimitiveTopologyEXT);
X(vkCmdSetStencilOpEXT);
X(vkCmdSetStencilTestEnableEXT);
+ X(vkCmdResolveImage);
X(vkCreateBuffer);
X(vkCreateBufferView);
X(vkCreateCommandPool);
@@ -176,6 +179,8 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
X(vkQueueSubmit);
X(vkResetFences);
X(vkResetQueryPoolEXT);
+ X(vkSetDebugUtilsObjectNameEXT);
+ X(vkSetDebugUtilsObjectTagEXT);
X(vkUnmapMemory);
X(vkUpdateDescriptorSetWithTemplateKHR);
X(vkUpdateDescriptorSets);
@@ -184,6 +189,19 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
#undef X
}
+template <typename T>
+void SetObjectName(const DeviceDispatch* dld, VkDevice device, T handle, VkObjectType type,
+ const char* name) {
+ const VkDebugUtilsObjectNameInfoEXT name_info{
+ .sType = VK_STRUCTURE_TYPE_DEBUG_UTILS_OBJECT_NAME_INFO_EXT,
+ .pNext = nullptr,
+ .objectType = VK_OBJECT_TYPE_IMAGE,
+ .objectHandle = reinterpret_cast<u64>(handle),
+ .pObjectName = name,
+ };
+ Check(dld->vkSetDebugUtilsObjectNameEXT(device, &name_info));
+}
+
} // Anonymous namespace
bool Load(InstanceDispatch& dld) noexcept {
@@ -476,8 +494,7 @@ DebugCallback Instance::TryCreateDebugCallback(
VK_DEBUG_UTILS_MESSAGE_SEVERITY_INFO_BIT_EXT |
VK_DEBUG_UTILS_MESSAGE_SEVERITY_VERBOSE_BIT_EXT,
.messageType = VK_DEBUG_UTILS_MESSAGE_TYPE_GENERAL_BIT_EXT |
- VK_DEBUG_UTILS_MESSAGE_TYPE_VALIDATION_BIT_EXT |
- VK_DEBUG_UTILS_MESSAGE_TYPE_PERFORMANCE_BIT_EXT,
+ VK_DEBUG_UTILS_MESSAGE_TYPE_VALIDATION_BIT_EXT,
.pfnUserCallback = callback,
.pUserData = nullptr,
};
@@ -493,10 +510,38 @@ void Buffer::BindMemory(VkDeviceMemory memory, VkDeviceSize offset) const {
Check(dld->vkBindBufferMemory(owner, handle, memory, offset));
}
+void Buffer::SetObjectNameEXT(const char* name) const {
+ SetObjectName(dld, owner, handle, VK_OBJECT_TYPE_BUFFER, name);
+}
+
+void BufferView::SetObjectNameEXT(const char* name) const {
+ SetObjectName(dld, owner, handle, VK_OBJECT_TYPE_BUFFER_VIEW, name);
+}
+
void Image::BindMemory(VkDeviceMemory memory, VkDeviceSize offset) const {
Check(dld->vkBindImageMemory(owner, handle, memory, offset));
}
+void Image::SetObjectNameEXT(const char* name) const {
+ SetObjectName(dld, owner, handle, VK_OBJECT_TYPE_IMAGE, name);
+}
+
+void ImageView::SetObjectNameEXT(const char* name) const {
+ SetObjectName(dld, owner, handle, VK_OBJECT_TYPE_IMAGE_VIEW, name);
+}
+
+void DeviceMemory::SetObjectNameEXT(const char* name) const {
+ SetObjectName(dld, owner, handle, VK_OBJECT_TYPE_DEVICE_MEMORY, name);
+}
+
+void Fence::SetObjectNameEXT(const char* name) const {
+ SetObjectName(dld, owner, handle, VK_OBJECT_TYPE_FENCE, name);
+}
+
+void Framebuffer::SetObjectNameEXT(const char* name) const {
+ SetObjectName(dld, owner, handle, VK_OBJECT_TYPE_FRAMEBUFFER, name);
+}
+
DescriptorSets DescriptorPool::Allocate(const VkDescriptorSetAllocateInfo& ai) const {
const std::size_t num = ai.descriptorSetCount;
std::unique_ptr sets = std::make_unique<VkDescriptorSet[]>(num);
@@ -510,6 +555,10 @@ DescriptorSets DescriptorPool::Allocate(const VkDescriptorSetAllocateInfo& ai) c
}
}
+void DescriptorPool::SetObjectNameEXT(const char* name) const {
+ SetObjectName(dld, owner, handle, VK_OBJECT_TYPE_DESCRIPTOR_POOL, name);
+}
+
CommandBuffers CommandPool::Allocate(std::size_t num_buffers, VkCommandBufferLevel level) const {
const VkCommandBufferAllocateInfo ai{
.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO,
@@ -530,6 +579,10 @@ CommandBuffers CommandPool::Allocate(std::size_t num_buffers, VkCommandBufferLev
}
}
+void CommandPool::SetObjectNameEXT(const char* name) const {
+ SetObjectName(dld, owner, handle, VK_OBJECT_TYPE_COMMAND_POOL, name);
+}
+
std::vector<VkImage> SwapchainKHR::GetImages() const {
u32 num;
Check(dld->vkGetSwapchainImagesKHR(owner, handle, &num, nullptr));
@@ -538,6 +591,18 @@ std::vector<VkImage> SwapchainKHR::GetImages() const {
return images;
}
+void Event::SetObjectNameEXT(const char* name) const {
+ SetObjectName(dld, owner, handle, VK_OBJECT_TYPE_EVENT, name);
+}
+
+void ShaderModule::SetObjectNameEXT(const char* name) const {
+ SetObjectName(dld, owner, handle, VK_OBJECT_TYPE_SHADER_MODULE, name);
+}
+
+void Semaphore::SetObjectNameEXT(const char* name) const {
+ SetObjectName(dld, owner, handle, VK_OBJECT_TYPE_SEMAPHORE, name);
+}
+
Device Device::Create(VkPhysicalDevice physical_device, Span<VkDeviceQueueCreateInfo> queues_ci,
Span<const char*> enabled_extensions, const void* next,
DeviceDispatch& dispatch) noexcept {
diff --git a/src/video_core/renderer_vulkan/wrapper.h b/src/video_core/renderer_vulkan/wrapper.h
index 76f790eab..f9a184e00 100644
--- a/src/video_core/renderer_vulkan/wrapper.h
+++ b/src/video_core/renderer_vulkan/wrapper.h
@@ -9,6 +9,7 @@
#include <limits>
#include <memory>
#include <optional>
+#include <span>
#include <type_traits>
#include <utility>
#include <vector>
@@ -18,6 +19,10 @@
#include "common/common_types.h"
+#ifdef _MSC_VER
+#pragma warning(disable : 26812) // Disable prefer enum class over enum
+#endif
+
namespace Vulkan::vk {
/**
@@ -41,6 +46,9 @@ public:
/// Construct an empty span.
constexpr Span() noexcept = default;
+ /// Construct an empty span
+ constexpr Span(std::nullptr_t) noexcept {}
+
/// Construct a span from a single element.
constexpr Span(const T& value) noexcept : ptr{&value}, num{1} {}
@@ -177,6 +185,7 @@ struct DeviceDispatch : public InstanceDispatch {
PFN_vkCmdBeginQuery vkCmdBeginQuery;
PFN_vkCmdBeginRenderPass vkCmdBeginRenderPass;
PFN_vkCmdBeginTransformFeedbackEXT vkCmdBeginTransformFeedbackEXT;
+ PFN_vkCmdBeginDebugUtilsLabelEXT vkCmdBeginDebugUtilsLabelEXT;
PFN_vkCmdBindDescriptorSets vkCmdBindDescriptorSets;
PFN_vkCmdBindIndexBuffer vkCmdBindIndexBuffer;
PFN_vkCmdBindPipeline vkCmdBindPipeline;
@@ -194,6 +203,7 @@ struct DeviceDispatch : public InstanceDispatch {
PFN_vkCmdEndQuery vkCmdEndQuery;
PFN_vkCmdEndRenderPass vkCmdEndRenderPass;
PFN_vkCmdEndTransformFeedbackEXT vkCmdEndTransformFeedbackEXT;
+ PFN_vkCmdEndDebugUtilsLabelEXT vkCmdEndDebugUtilsLabelEXT;
PFN_vkCmdFillBuffer vkCmdFillBuffer;
PFN_vkCmdPipelineBarrier vkCmdPipelineBarrier;
PFN_vkCmdPushConstants vkCmdPushConstants;
@@ -217,6 +227,7 @@ struct DeviceDispatch : public InstanceDispatch {
PFN_vkCmdSetPrimitiveTopologyEXT vkCmdSetPrimitiveTopologyEXT;
PFN_vkCmdSetStencilOpEXT vkCmdSetStencilOpEXT;
PFN_vkCmdSetStencilTestEnableEXT vkCmdSetStencilTestEnableEXT;
+ PFN_vkCmdResolveImage vkCmdResolveImage;
PFN_vkCreateBuffer vkCreateBuffer;
PFN_vkCreateBufferView vkCreateBufferView;
PFN_vkCreateCommandPool vkCreateCommandPool;
@@ -272,6 +283,8 @@ struct DeviceDispatch : public InstanceDispatch {
PFN_vkQueueSubmit vkQueueSubmit;
PFN_vkResetFences vkResetFences;
PFN_vkResetQueryPoolEXT vkResetQueryPoolEXT;
+ PFN_vkSetDebugUtilsObjectNameEXT vkSetDebugUtilsObjectNameEXT;
+ PFN_vkSetDebugUtilsObjectTagEXT vkSetDebugUtilsObjectTagEXT;
PFN_vkUnmapMemory vkUnmapMemory;
PFN_vkUpdateDescriptorSetWithTemplateKHR vkUpdateDescriptorSetWithTemplateKHR;
PFN_vkUpdateDescriptorSets vkUpdateDescriptorSets;
@@ -542,18 +555,14 @@ private:
const DeviceDispatch* dld = nullptr;
};
-using BufferView = Handle<VkBufferView, VkDevice, DeviceDispatch>;
using DebugCallback = Handle<VkDebugUtilsMessengerEXT, VkInstance, InstanceDispatch>;
using DescriptorSetLayout = Handle<VkDescriptorSetLayout, VkDevice, DeviceDispatch>;
using DescriptorUpdateTemplateKHR = Handle<VkDescriptorUpdateTemplateKHR, VkDevice, DeviceDispatch>;
-using Framebuffer = Handle<VkFramebuffer, VkDevice, DeviceDispatch>;
-using ImageView = Handle<VkImageView, VkDevice, DeviceDispatch>;
using Pipeline = Handle<VkPipeline, VkDevice, DeviceDispatch>;
using PipelineLayout = Handle<VkPipelineLayout, VkDevice, DeviceDispatch>;
using QueryPool = Handle<VkQueryPool, VkDevice, DeviceDispatch>;
using RenderPass = Handle<VkRenderPass, VkDevice, DeviceDispatch>;
using Sampler = Handle<VkSampler, VkDevice, DeviceDispatch>;
-using ShaderModule = Handle<VkShaderModule, VkDevice, DeviceDispatch>;
using SurfaceKHR = Handle<VkSurfaceKHR, VkInstance, InstanceDispatch>;
using DescriptorSets = PoolAllocations<VkDescriptorSet, VkDescriptorPool>;
@@ -605,6 +614,17 @@ class Buffer : public Handle<VkBuffer, VkDevice, DeviceDispatch> {
public:
/// Attaches a memory allocation.
void BindMemory(VkDeviceMemory memory, VkDeviceSize offset) const;
+
+ /// Set object name.
+ void SetObjectNameEXT(const char* name) const;
+};
+
+class BufferView : public Handle<VkBufferView, VkDevice, DeviceDispatch> {
+ using Handle<VkBufferView, VkDevice, DeviceDispatch>::Handle;
+
+public:
+ /// Set object name.
+ void SetObjectNameEXT(const char* name) const;
};
class Image : public Handle<VkImage, VkDevice, DeviceDispatch> {
@@ -613,12 +633,26 @@ class Image : public Handle<VkImage, VkDevice, DeviceDispatch> {
public:
/// Attaches a memory allocation.
void BindMemory(VkDeviceMemory memory, VkDeviceSize offset) const;
+
+ /// Set object name.
+ void SetObjectNameEXT(const char* name) const;
+};
+
+class ImageView : public Handle<VkImageView, VkDevice, DeviceDispatch> {
+ using Handle<VkImageView, VkDevice, DeviceDispatch>::Handle;
+
+public:
+ /// Set object name.
+ void SetObjectNameEXT(const char* name) const;
};
class DeviceMemory : public Handle<VkDeviceMemory, VkDevice, DeviceDispatch> {
using Handle<VkDeviceMemory, VkDevice, DeviceDispatch>::Handle;
public:
+ /// Set object name.
+ void SetObjectNameEXT(const char* name) const;
+
u8* Map(VkDeviceSize offset, VkDeviceSize size) const {
void* data;
Check(dld->vkMapMemory(owner, handle, offset, size, 0, &data));
@@ -634,6 +668,9 @@ class Fence : public Handle<VkFence, VkDevice, DeviceDispatch> {
using Handle<VkFence, VkDevice, DeviceDispatch>::Handle;
public:
+ /// Set object name.
+ void SetObjectNameEXT(const char* name) const;
+
VkResult Wait(u64 timeout = std::numeric_limits<u64>::max()) const noexcept {
return dld->vkWaitForFences(owner, 1, &handle, true, timeout);
}
@@ -647,11 +684,22 @@ public:
}
};
+class Framebuffer : public Handle<VkFramebuffer, VkDevice, DeviceDispatch> {
+ using Handle<VkFramebuffer, VkDevice, DeviceDispatch>::Handle;
+
+public:
+ /// Set object name.
+ void SetObjectNameEXT(const char* name) const;
+};
+
class DescriptorPool : public Handle<VkDescriptorPool, VkDevice, DeviceDispatch> {
using Handle<VkDescriptorPool, VkDevice, DeviceDispatch>::Handle;
public:
DescriptorSets Allocate(const VkDescriptorSetAllocateInfo& ai) const;
+
+ /// Set object name.
+ void SetObjectNameEXT(const char* name) const;
};
class CommandPool : public Handle<VkCommandPool, VkDevice, DeviceDispatch> {
@@ -660,6 +708,9 @@ class CommandPool : public Handle<VkCommandPool, VkDevice, DeviceDispatch> {
public:
CommandBuffers Allocate(std::size_t num_buffers,
VkCommandBufferLevel level = VK_COMMAND_BUFFER_LEVEL_PRIMARY) const;
+
+ /// Set object name.
+ void SetObjectNameEXT(const char* name) const;
};
class SwapchainKHR : public Handle<VkSwapchainKHR, VkDevice, DeviceDispatch> {
@@ -673,15 +724,29 @@ class Event : public Handle<VkEvent, VkDevice, DeviceDispatch> {
using Handle<VkEvent, VkDevice, DeviceDispatch>::Handle;
public:
+ /// Set object name.
+ void SetObjectNameEXT(const char* name) const;
+
VkResult GetStatus() const noexcept {
return dld->vkGetEventStatus(owner, handle);
}
};
+class ShaderModule : public Handle<VkShaderModule, VkDevice, DeviceDispatch> {
+ using Handle<VkShaderModule, VkDevice, DeviceDispatch>::Handle;
+
+public:
+ /// Set object name.
+ void SetObjectNameEXT(const char* name) const;
+};
+
class Semaphore : public Handle<VkSemaphore, VkDevice, DeviceDispatch> {
using Handle<VkSemaphore, VkDevice, DeviceDispatch>::Handle;
public:
+ /// Set object name.
+ void SetObjectNameEXT(const char* name) const;
+
[[nodiscard]] u64 GetCounter() const {
u64 value;
Check(dld->vkGetSemaphoreCounterValueKHR(owner, handle, &value));
@@ -932,6 +997,12 @@ public:
regions.data(), filter);
}
+ void ResolveImage(VkImage src_image, VkImageLayout src_layout, VkImage dst_image,
+ VkImageLayout dst_layout, Span<VkImageResolve> regions) {
+ dld->vkCmdResolveImage(handle, src_image, src_layout, dst_image, dst_layout, regions.size(),
+ regions.data());
+ }
+
void Dispatch(u32 x, u32 y, u32 z) const noexcept {
dld->vkCmdDispatch(handle, x, y, z);
}
@@ -946,6 +1017,23 @@ public:
image_barriers.size(), image_barriers.data());
}
+ void PipelineBarrier(VkPipelineStageFlags src_stage_mask, VkPipelineStageFlags dst_stage_mask,
+ VkDependencyFlags dependency_flags = 0) const noexcept {
+ PipelineBarrier(src_stage_mask, dst_stage_mask, dependency_flags, {}, {}, {});
+ }
+
+ void PipelineBarrier(VkPipelineStageFlags src_stage_mask, VkPipelineStageFlags dst_stage_mask,
+ VkDependencyFlags dependency_flags,
+ const VkBufferMemoryBarrier& buffer_barrier) const noexcept {
+ PipelineBarrier(src_stage_mask, dst_stage_mask, dependency_flags, {}, buffer_barrier, {});
+ }
+
+ void PipelineBarrier(VkPipelineStageFlags src_stage_mask, VkPipelineStageFlags dst_stage_mask,
+ VkDependencyFlags dependency_flags,
+ const VkImageMemoryBarrier& image_barrier) const noexcept {
+ PipelineBarrier(src_stage_mask, dst_stage_mask, dependency_flags, {}, {}, image_barrier);
+ }
+
void CopyBufferToImage(VkBuffer src_buffer, VkImage dst_image, VkImageLayout dst_image_layout,
Span<VkBufferImageCopy> regions) const noexcept {
dld->vkCmdCopyBufferToImage(handle, src_buffer, dst_image, dst_image_layout, regions.size(),
@@ -979,6 +1067,13 @@ public:
dld->vkCmdPushConstants(handle, layout, flags, offset, size, values);
}
+ template <typename T>
+ void PushConstants(VkPipelineLayout layout, VkShaderStageFlags flags,
+ const T& data) const noexcept {
+ static_assert(std::is_trivially_copyable_v<T>, "<data> is not trivially copyable");
+ dld->vkCmdPushConstants(handle, layout, flags, 0, static_cast<u32>(sizeof(T)), &data);
+ }
+
void SetViewport(u32 first, Span<VkViewport> viewports) const noexcept {
dld->vkCmdSetViewport(handle, first, viewports.size(), viewports.data());
}
@@ -1088,6 +1183,20 @@ public:
counter_buffers, counter_buffer_offsets);
}
+ void BeginDebugUtilsLabelEXT(const char* label, std::span<float, 4> color) const noexcept {
+ const VkDebugUtilsLabelEXT label_info{
+ .sType = VK_STRUCTURE_TYPE_DEBUG_UTILS_LABEL_EXT,
+ .pNext = nullptr,
+ .pLabelName = label,
+ .color{color[0], color[1], color[2], color[3]},
+ };
+ dld->vkCmdBeginDebugUtilsLabelEXT(handle, &label_info);
+ }
+
+ void EndDebugUtilsLabelEXT() const noexcept {
+ dld->vkCmdEndDebugUtilsLabelEXT(handle);
+ }
+
private:
VkCommandBuffer handle;
const DeviceDispatch* dld;