[texture_cache] Implement alpha copies and put barrier in compute stages

src/video_core/host_shaders/CMakeLists.txt

@@ -21,6 +21,8 @@ set(SHADER_FILES
     convert_abgr8_srgb_to_d24s8.frag
     convert_abgr8_to_d24s8.frag
     convert_abgr8_to_d32f.frag
+    convert_r8_to_abgr8.frag
+    convert_abgr8_to_r8.frag
     convert_d32f_to_abgr8.frag
     convert_d24s8_to_abgr8.frag
     convert_depth_to_float.frag

src/video_core/host_shaders/convert_abgr8_to_r8.frag

@@ -0,0 +1,19 @@
+// SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#version 450
+
+layout(binding = 0) uniform sampler2D tex;
+
+layout(location = 0) out float color;
+
+layout(push_constant) uniform PushConstants {
+    uniform vec2 tex_scale;
+    uniform vec2 tex_offset;
+};
+
+void main() {
+    vec2 coord = gl_FragCoord.xy / tex_scale * textureSize(tex, 0);
+    float red = texelFetch(tex, ivec2(coord), 0).r;
+    color = red;
+}

src/video_core/host_shaders/convert_r8_to_abgr8.frag

@@ -0,0 +1,20 @@
+// SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#version 450
+
+layout(binding = 0) uniform sampler2D tex;
+
+layout(push_constant) uniform PushConstants {
+    uniform vec2 tex_scale;
+    uniform vec2 tex_offset;
+};
+
+layout(location = 0) out vec4 color;
+
+void main() {
+    vec2 coord = gl_FragCoord.xy / tex_scale * textureSize(tex, 0);
+    float red = texelFetch(tex, ivec2(coord), 0).r;
+    color = vec4(red, red, red, 1.0);
+}
+

src/video_core/renderer_vulkan/blit_image.cpp

@@ -11,6 +11,7 @@
 #include "video_core/host_shaders/convert_abgr8_to_d32f_frag_spv.h"
 #include "video_core/host_shaders/convert_d24s8_to_abgr8_frag_spv.h"
 #include "video_core/host_shaders/convert_d32f_to_abgr8_frag_spv.h"
+#include "video_core/host_shaders/convert_r8_to_abgr8_frag_spv.h"
 #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/convert_s8d24_to_abgr8_frag_spv.h"
@@ -29,6 +30,7 @@
 #include "video_core/vulkan_common/vulkan_device.h"
 #include "video_core/vulkan_common/vulkan_wrapper.h"
 #include "video_core/host_shaders/convert_abgr8_srgb_to_d24s8_frag_spv.h"
+#include "video_core/host_shaders/convert_abgr8_to_r8_frag_spv.h"
 #include "video_core/host_shaders/convert_rgba8_to_bgra8_frag_spv.h"
 #include "video_core/host_shaders/convert_yuv420_to_rgb_comp_spv.h"
 #include "video_core/host_shaders/convert_rgb_to_yuv420_comp_spv.h"
@@ -444,7 +446,8 @@ BlitImageHelper::BlitImageHelper(const Device& device_, Scheduler& scheduler_,
           descriptor_pool.Allocator(*two_textures_set_layout, TEXTURE_DESCRIPTOR_BANK_INFO<2>)},
       one_texture_pipeline_layout(device.GetLogical().CreatePipelineLayout(PipelineLayoutCreateInfo(
           one_texture_set_layout.address(),
-          PUSH_CONSTANT_RANGE<VK_SHADER_STAGE_VERTEX_BIT, sizeof(PushConstants)>))),
+          PUSH_CONSTANT_RANGE<VK_SHADER_STAGE_VERTEX_BIT | VK_SHADER_STAGE_FRAGMENT_BIT, sizeof(PushConstants)>
+          ))),
       two_textures_pipeline_layout(
           device.GetLogical().CreatePipelineLayout(PipelineLayoutCreateInfo(
               two_textures_set_layout.address(),
@@ -461,6 +464,8 @@ BlitImageHelper::BlitImageHelper(const Device& device_, Scheduler& scheduler_,
       convert_float_to_depth_frag(BuildShader(device, CONVERT_FLOAT_TO_DEPTH_FRAG_SPV)),
       convert_abgr8_to_d24s8_frag(BuildShader(device, CONVERT_ABGR8_TO_D24S8_FRAG_SPV)),
       convert_abgr8_to_d32f_frag(BuildShader(device, CONVERT_ABGR8_TO_D32F_FRAG_SPV)),
+      convert_abgr8_to_r8_frag(BuildShader(device, CONVERT_ABGR8_TO_R8_FRAG_SPV)),
+      convert_r8_to_abgr8_frag(BuildShader(device, CONVERT_R8_TO_ABGR8_FRAG_SPV)),
       convert_d32f_to_abgr8_frag(BuildShader(device, CONVERT_D32F_TO_ABGR8_FRAG_SPV)),
       convert_d24s8_to_abgr8_frag(BuildShader(device, CONVERT_D24S8_TO_ABGR8_FRAG_SPV)),
       convert_s8d24_to_abgr8_frag(BuildShader(device, CONVERT_S8D24_TO_ABGR8_FRAG_SPV)),
@@ -588,6 +593,20 @@ void BlitImageHelper::ConvertR16ToD16(const Framebuffer* dst_framebuffer,
     Convert(*convert_r16_to_d16_pipeline, dst_framebuffer, src_image_view);
 }
 
+void BlitImageHelper::ConvertR8ToABGR8(const Framebuffer* dst_framebuffer,
+                                  const ImageView& src_image_view) {
+    ConvertPipelineColorTargetEx(convert_r8_to_abgr8_pipeline, dst_framebuffer->RenderPass(),
+                                 convert_r8_to_abgr8_frag);
+    Convert(*convert_r8_to_abgr8_pipeline, dst_framebuffer, src_image_view);
+}
+
+void BlitImageHelper::ConvertABGR8ToR8(const Framebuffer* dst_framebuffer,
+                                      const ImageView& src_image_view) {
+    ConvertPipelineColorTargetEx(convert_abgr8_to_r8_pipeline, dst_framebuffer->RenderPass(),
+                                 convert_abgr8_to_r8_frag);
+    Convert(*convert_abgr8_to_r8_pipeline, dst_framebuffer, src_image_view);
+}
+
 void BlitImageHelper::ConvertABGR8ToD24S8(const Framebuffer* dst_framebuffer,
                                           const ImageView& src_image_view) {
     ConvertPipelineDepthTargetEx(convert_abgr8_to_d24s8_pipeline, dst_framebuffer->RenderPass(),
@@ -687,7 +706,7 @@ void BlitImageHelper::Convert(VkPipeline pipeline, const Framebuffer* dst_frameb
     const VkExtent2D extent = GetConversionExtent(src_image_view);
 
     scheduler.RequestRenderpass(dst_framebuffer);
-    scheduler.Record([pipeline, layout, sampler, src_view, extent, this](vk::CommandBuffer cmdbuf) {
+    scheduler.Record([pipeline, layout, sampler, src_view, extent, this, dst_framebuffer](vk::CommandBuffer cmdbuf) {
         const VkOffset2D offset{
             .x = 0,
             .y = 0,
@@ -704,10 +723,17 @@ void BlitImageHelper::Convert(VkPipeline pipeline, const Framebuffer* dst_frameb
             .offset = offset,
             .extent = extent,
         };
-        const PushConstants push_constants{
+        const PushConstants push_constants_vertex{
             .tex_scale = {viewport.width, viewport.height},
             .tex_offset = {0.0f, 0.0f},
         };
+        const PushConstants push_constants_frag{
+            .tex_scale = {
+                static_cast<float>(dst_framebuffer->RenderArea().width),
+                static_cast<float>(dst_framebuffer->RenderArea().height)
+            },
+            .tex_offset = {0.0f, 0.0f},
+        };
         const VkDescriptorSet descriptor_set = one_texture_descriptor_allocator.Commit();
         UpdateOneTextureDescriptorSet(device, descriptor_set, sampler, src_view);
 
@@ -717,7 +743,8 @@ void BlitImageHelper::Convert(VkPipeline pipeline, const Framebuffer* dst_frameb
                                   nullptr);
         cmdbuf.SetViewport(0, viewport);
         cmdbuf.SetScissor(0, scissor);
-        cmdbuf.PushConstants(layout, VK_SHADER_STAGE_VERTEX_BIT, push_constants);
+        cmdbuf.PushConstants(layout, VK_SHADER_STAGE_VERTEX_BIT, push_constants_vertex);
+        cmdbuf.PushConstants(layout, VK_SHADER_STAGE_FRAGMENT_BIT, push_constants_frag);
         cmdbuf.Draw(3, 1, 0, 0);
     });
     scheduler.InvalidateState();

src/video_core/renderer_vulkan/blit_image.h

@@ -65,6 +65,10 @@ public:
 
     void ConvertR16ToD16(const Framebuffer* dst_framebuffer, const ImageView& src_image_view);
 
+    void ConvertR8ToABGR8(const Framebuffer* dst_framebuffer, const ImageView& src_image_view);
+
+    void ConvertABGR8ToR8(const Framebuffer* dst_framebuffer, const ImageView& src_image_view);
+
     void ConvertABGR8ToD24S8(const Framebuffer* dst_framebuffer, const ImageView& src_image_view);
 
     void ConvertABGR8SRGBToD24S8(const Framebuffer* dst_framebuffer, const ImageView& src_image_view);
@@ -144,6 +148,8 @@ private:
     vk::ShaderModule convert_float_to_depth_frag;
     vk::ShaderModule convert_abgr8_to_d24s8_frag;
     vk::ShaderModule convert_abgr8_to_d32f_frag;
+    vk::ShaderModule convert_abgr8_to_r8_frag;
+    vk::ShaderModule convert_r8_to_abgr8_frag;
     vk::ShaderModule convert_d32f_to_abgr8_frag;
     vk::ShaderModule convert_d24s8_to_abgr8_frag;
     vk::ShaderModule convert_s8d24_to_abgr8_frag;
@@ -176,6 +182,8 @@ private:
     vk::Pipeline convert_d32f_to_abgr8_pipeline;
     vk::Pipeline convert_d24s8_to_abgr8_pipeline;
     vk::Pipeline convert_s8d24_to_abgr8_pipeline;
+    vk::Pipeline convert_r8_to_abgr8_pipeline;
+    vk::Pipeline convert_abgr8_to_r8_pipeline;
     vk::Pipeline convert_abgr8_srgb_to_d24s8_pipeline;
     vk::Pipeline convert_rgba_to_bgra_pipeline;
     vk::Pipeline convert_yuv420_to_rgb_pipeline;

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -518,6 +518,14 @@ void RasterizerVulkan::DispatchCompute() {
     }
     const std::array<u32, 3> dim{qmd.grid_dim_x, qmd.grid_dim_y, qmd.grid_dim_z};
     scheduler.RequestOutsideRenderPassOperationContext();
+    static constexpr VkMemoryBarrier READ_BARRIER{
+        .sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER,
+        .pNext = nullptr,
+        .srcAccessMask = VK_ACCESS_MEMORY_WRITE_BIT,
+        .dstAccessMask = VK_ACCESS_MEMORY_READ_BIT,
+    };
+    scheduler.Record([](vk::CommandBuffer cmdbuf) { cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
+                               0, READ_BARRIER); });
     scheduler.Record([dim](vk::CommandBuffer cmdbuf) { cmdbuf.Dispatch(dim[0], dim[1], dim[2]); });
 }
 

src/video_core/renderer_vulkan/vk_texture_cache.cpp

@@ -1219,8 +1219,16 @@ void TextureCacheRuntime::ConvertImage(Framebuffer* dst, ImageView& dst_view, Im
             return blit_image_helper.ConvertABGR8ToD24S8(dst, src_view);
         }
         break;
-
+    case PixelFormat::R8_UNORM:
+        if (src_view.format == PixelFormat::A8B8G8R8_UNORM) {
+            return blit_image_helper.ConvertABGR8ToR8(dst, src_view);
+        }
+        break;
     case PixelFormat::A8B8G8R8_UNORM:
+        if (src_view.format == PixelFormat::R8_UNORM) {
+            return blit_image_helper.ConvertR8ToABGR8(dst, src_view);
+        }
+        break;
     case PixelFormat::A8B8G8R8_SNORM:
     case PixelFormat::A8B8G8R8_SINT:
     case PixelFormat::A8B8G8R8_UINT:
@@ -1232,7 +1240,6 @@ void TextureCacheRuntime::ConvertImage(Framebuffer* dst, ImageView& dst_view, Im
     case PixelFormat::A2R10G10B10_UNORM:
     case PixelFormat::A1B5G5R5_UNORM:
     case PixelFormat::A5B5G5R1_UNORM:
-    case PixelFormat::R8_UNORM:
     case PixelFormat::R8_SNORM:
     case PixelFormat::R8_SINT:
     case PixelFormat::R8_UINT:
@@ -1325,6 +1332,7 @@ void TextureCacheRuntime::ConvertImage(Framebuffer* dst, ImageView& dst_view, Im
     default:
         break;
     }
+    LOG_WARNING(Render_Vulkan, "Unimplemented texture conversion from {} to {} format type", src_view.format, dst_view.format);
 }
 
 VkFormat TextureCacheRuntime::GetSupportedFormat(VkFormat requested_format,

src/video_core/surface.cpp

@@ -237,6 +237,37 @@ SurfaceType GetFormatType(PixelFormat pixel_format) {
     return SurfaceType::Invalid;
 }
 
+bool HasAlpha(PixelFormat pixel_format) {
+    switch (pixel_format) {
+        case PixelFormat::A8B8G8R8_UNORM:
+        case PixelFormat::A8B8G8R8_SNORM:
+        case PixelFormat::A8B8G8R8_SINT:
+        case PixelFormat::A8B8G8R8_UINT:
+        case PixelFormat::A1R5G5B5_UNORM:
+        case PixelFormat::A2B10G10R10_UNORM:
+        case PixelFormat::A2B10G10R10_UINT:
+        case PixelFormat::A2R10G10B10_UNORM:
+        case PixelFormat::A1B5G5R5_UNORM:
+        case PixelFormat::A5B5G5R1_UNORM:
+        case PixelFormat::R16G16B16A16_FLOAT:
+        case PixelFormat::R16G16B16A16_UNORM:
+        case PixelFormat::R16G16B16A16_SNORM:
+        case PixelFormat::R16G16B16A16_SINT:
+        case PixelFormat::R16G16B16A16_UINT:
+        case PixelFormat::R32G32B32A32_UINT:
+        case PixelFormat::BC1_RGBA_UNORM:
+        case PixelFormat::B8G8R8A8_UNORM:
+        case PixelFormat::R32G32B32A32_FLOAT:
+        case PixelFormat::R32G32B32A32_SINT:
+        case PixelFormat::A8B8G8R8_SRGB:
+        case PixelFormat::BC1_RGBA_SRGB:
+        case PixelFormat::A4B4G4R4_UNORM:
+            return true;
+        default:
+            return false;
+    }
+}
+
 bool IsPixelFormatASTC(PixelFormat format) {
     switch (format) {
     case PixelFormat::ASTC_2D_4X4_UNORM:

src/video_core/surface.h

@@ -503,6 +503,8 @@ PixelFormat PixelFormatFromGPUPixelFormat(Service::android::PixelFormat format);
 
 SurfaceType GetFormatType(PixelFormat pixel_format);
 
+bool HasAlpha(PixelFormat pixel_format);
+
 bool IsPixelFormatASTC(PixelFormat format);
 
 bool IsPixelFormatBCn(PixelFormat format);

src/video_core/texture_cache/texture_cache.h

@@ -26,6 +26,7 @@ namespace VideoCommon {
 using Tegra::Texture::TICEntry;
 using Tegra::Texture::TSCEntry;
 using VideoCore::Surface::GetFormatType;
+using VideoCore::Surface::HasAlpha;
 using VideoCore::Surface::PixelFormat;
 using VideoCore::Surface::SurfaceType;
 using namespace Common::Literals;
@@ -2398,7 +2399,7 @@ void TextureCache<P>::CopyImage(ImageId dst_id, ImageId src_id, std::vector<Imag
     }
     const auto dst_format_type = GetFormatType(dst.info.format);
     const auto src_format_type = GetFormatType(src.info.format);
-    if (src_format_type == dst_format_type) {
+    if (src_format_type == dst_format_type && HasAlpha(src.info.format) == HasAlpha(dst.info.format)) {
         if constexpr (HAS_EMULATED_COPIES) {
             if (!runtime.CanImageBeCopied(dst, src)) {
                 return runtime.EmulateCopyImage(dst, src, copies);
@@ -2406,8 +2407,8 @@ void TextureCache<P>::CopyImage(ImageId dst_id, ImageId src_id, std::vector<Imag
         }
         return runtime.CopyImage(dst, src, copies);
     }
-    UNIMPLEMENTED_IF(dst.info.type != ImageType::e2D);
+    UNIMPLEMENTED_IF(dst.info.type != ImageType::e2D && dst.info.type != ImageType::Linear);
-    UNIMPLEMENTED_IF(src.info.type != ImageType::e2D);
+    UNIMPLEMENTED_IF(src.info.type != ImageType::e2D && dst.info.type != ImageType::Linear);
     if (runtime.ShouldReinterpret(dst, src)) {
         return runtime.ReinterpretImage(dst, src, copies);
     }