4 files changed, 117 insertions, 7 deletions

diff --git a/src/video_core/renderer_vulkan/blit_image.cpp b/src/video_core/renderer_vulkan/blit_image.cpp
index 3f2b139e0..dd00d3edf 100644
--- a/src/video_core/renderer_vulkan/blit_image.cpp
+++ b/src/video_core/renderer_vulkan/blit_image.cpp
@@ -4,13 +4,13 @@
 #include <algorithm>
 
 #include "common/settings.h"
+#include "video_core/host_shaders/blit_color_float_frag_spv.h"
 #include "video_core/host_shaders/convert_abgr8_to_d24s8_frag_spv.h"
 #include "video_core/host_shaders/convert_d24s8_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"
 #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"
@@ -303,7 +303,7 @@ void UpdateTwoTexturesDescriptorSet(const Device& device, VkDescriptorSet descri
 }
 
 void BindBlitState(vk::CommandBuffer cmdbuf, VkPipelineLayout layout, const Region2D& dst_region,
-                   const Region2D& src_region) {
+                   const Region2D& src_region, const Extent3D& src_size = {1, 1, 1}) {
     const VkOffset2D offset{
         .x = std::min(dst_region.start.x, dst_region.end.x),
         .y = std::min(dst_region.start.y, dst_region.end.y),
@@ -325,12 +325,15 @@ void BindBlitState(vk::CommandBuffer cmdbuf, VkPipelineLayout layout, const Regi
         .offset = offset,
         .extent = extent,
     };
-    const float scale_x = static_cast<float>(src_region.end.x - src_region.start.x);
-    const float scale_y = static_cast<float>(src_region.end.y - src_region.start.y);
+    const float scale_x = static_cast<float>(src_region.end.x - src_region.start.x) /
+                          static_cast<float>(src_size.width);
+    const float scale_y = static_cast<float>(src_region.end.y - src_region.start.y) /
+                          static_cast<float>(src_size.height);
     const PushConstants push_constants{
         .tex_scale = {scale_x, scale_y},
-        .tex_offset = {static_cast<float>(src_region.start.x),
-                       static_cast<float>(src_region.start.y)},
+        .tex_offset = {static_cast<float>(src_region.start.x) / static_cast<float>(src_size.width),
+                       static_cast<float>(src_region.start.y) /
+                           static_cast<float>(src_size.height)},
     };
     cmdbuf.SetViewport(0, viewport);
     cmdbuf.SetScissor(0, scissor);
@@ -347,6 +350,51 @@ VkExtent2D GetConversionExtent(const ImageView& src_image_view) {
         .height = is_rescaled ? resolution.ScaleUp(height) : height,
     };
 }
+
+void TransitionImageLayout(vk::CommandBuffer& cmdbuf, VkImage image, VkImageLayout target_layout,
+                           VkImageLayout source_layout = VK_IMAGE_LAYOUT_GENERAL) {
+    constexpr VkFlags flags{VK_ACCESS_COLOR_ATTACHMENT_READ_BIT |
+                            VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT | VK_ACCESS_SHADER_READ_BIT};
+    const VkImageMemoryBarrier barrier{
+        .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
+        .pNext = nullptr,
+        .srcAccessMask = flags,
+        .dstAccessMask = flags,
+        .oldLayout = source_layout,
+        .newLayout = target_layout,
+        .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,
+        },
+    };
+    cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
+                           0, barrier);
+}
+
+void BeginRenderPass(vk::CommandBuffer& cmdbuf, const Framebuffer* framebuffer) {
+    const VkRenderPass render_pass = framebuffer->RenderPass();
+    const VkFramebuffer framebuffer_handle = framebuffer->Handle();
+    const VkExtent2D render_area = framebuffer->RenderArea();
+    const VkRenderPassBeginInfo renderpass_bi{
+        .sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO,
+        .pNext = nullptr,
+        .renderPass = render_pass,
+        .framebuffer = framebuffer_handle,
+        .renderArea{
+            .offset{},
+            .extent = render_area,
+        },
+        .clearValueCount = 0,
+        .pClearValues = nullptr,
+    };
+    cmdbuf.BeginRenderPass(renderpass_bi, VK_SUBPASS_CONTENTS_INLINE);
+}
 } // Anonymous namespace
 
 BlitImageHelper::BlitImageHelper(const Device& device_, Scheduler& scheduler_,
@@ -365,7 +413,7 @@ BlitImageHelper::BlitImageHelper(const Device& device_, Scheduler& scheduler_,
       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)),
+      blit_color_to_color_frag(BuildShader(device, BLIT_COLOR_FLOAT_FRAG_SPV)),
       blit_depth_stencil_frag(BuildShader(device, VULKAN_BLIT_DEPTH_STENCIL_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)),
@@ -404,6 +452,32 @@ void BlitImageHelper::BlitColor(const Framebuffer* dst_framebuffer, VkImageView
     scheduler.InvalidateState();
 }
 
+void BlitImageHelper::BlitColor(const Framebuffer* dst_framebuffer, VkImageView src_image_view,
+                                VkImage src_image, VkSampler src_sampler,
+                                const Region2D& dst_region, const Region2D& src_region,
+                                const Extent3D& src_size) {
+    const BlitImagePipelineKey key{
+        .renderpass = dst_framebuffer->RenderPass(),
+        .operation = Tegra::Engines::Fermi2D::Operation::SrcCopy,
+    };
+    const VkPipelineLayout layout = *one_texture_pipeline_layout;
+    const VkPipeline pipeline = FindOrEmplaceColorPipeline(key);
+    scheduler.RequestOutsideRenderPassOperationContext();
+    scheduler.Record([this, dst_framebuffer, src_image_view, src_image, src_sampler, dst_region,
+                      src_region, src_size, pipeline, layout](vk::CommandBuffer cmdbuf) {
+        TransitionImageLayout(cmdbuf, src_image, VK_IMAGE_LAYOUT_READ_ONLY_OPTIMAL);
+        BeginRenderPass(cmdbuf, dst_framebuffer);
+        const VkDescriptorSet descriptor_set = one_texture_descriptor_allocator.Commit();
+        UpdateOneTextureDescriptorSet(device, descriptor_set, src_sampler, src_image_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, src_size);
+        cmdbuf.Draw(3, 1, 0, 0);
+        cmdbuf.EndRenderPass();
+    });
+}
+
 void BlitImageHelper::BlitDepthStencil(const Framebuffer* dst_framebuffer,
                                        VkImageView src_depth_view, VkImageView src_stencil_view,
                                        const Region2D& dst_region, const Region2D& src_region,
diff --git a/src/video_core/renderer_vulkan/blit_image.h b/src/video_core/renderer_vulkan/blit_image.h
index 5df679fb4..be8a9a2f6 100644
--- a/src/video_core/renderer_vulkan/blit_image.h
+++ b/src/video_core/renderer_vulkan/blit_image.h
@@ -10,6 +10,8 @@
 
 namespace Vulkan {
 
+using VideoCommon::Extent3D;
+using VideoCommon::Offset2D;
 using VideoCommon::Region2D;
 
 class Device;
@@ -36,6 +38,10 @@ public:
                    Tegra::Engines::Fermi2D::Filter filter,
                    Tegra::Engines::Fermi2D::Operation operation);
 
+    void BlitColor(const Framebuffer* dst_framebuffer, VkImageView src_image_view,
+                   VkImage src_image, VkSampler src_sampler, const Region2D& dst_region,
+                   const Region2D& src_region, const Extent3D& src_size);
+
     void BlitDepthStencil(const Framebuffer* dst_framebuffer, VkImageView src_depth_view,
                           VkImageView src_stencil_view, const Region2D& dst_region,
                           const Region2D& src_region, Tegra::Engines::Fermi2D::Filter filter,
diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp
index b75b8eec6..86ef0daeb 100644
--- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp
+++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp
@@ -266,6 +266,35 @@ void RasterizerVulkan::DrawIndirect() {
     buffer_cache.SetDrawIndirect(nullptr);
 }
 
+void RasterizerVulkan::DrawTexture() {
+    MICROPROFILE_SCOPE(Vulkan_Drawing);
+
+    SCOPE_EXIT({ gpu.TickWork(); });
+    FlushWork();
+
+    query_cache.UpdateCounters();
+
+    texture_cache.SynchronizeGraphicsDescriptors();
+    texture_cache.UpdateRenderTargets(false);
+
+    UpdateDynamicStates();
+
+    const auto& draw_texture_state = maxwell3d->draw_manager->GetDrawTextureState();
+    const auto& sampler = texture_cache.GetGraphicsSampler(draw_texture_state.src_sampler);
+    const auto& texture = texture_cache.GetImageView(draw_texture_state.src_texture);
+    Region2D dst_region = {Offset2D{.x = static_cast<s32>(draw_texture_state.dst_x0),
+                                    .y = static_cast<s32>(draw_texture_state.dst_y0)},
+                           Offset2D{.x = static_cast<s32>(draw_texture_state.dst_x1),
+                                    .y = static_cast<s32>(draw_texture_state.dst_y1)}};
+    Region2D src_region = {Offset2D{.x = static_cast<s32>(draw_texture_state.src_x0),
+                                    .y = static_cast<s32>(draw_texture_state.src_y0)},
+                           Offset2D{.x = static_cast<s32>(draw_texture_state.src_x1),
+                                    .y = static_cast<s32>(draw_texture_state.src_y1)}};
+    blit_image.BlitColor(texture_cache.GetFramebuffer(), texture.RenderTarget(),
+                         texture.ImageHandle(), sampler->Handle(), dst_region, src_region,
+                         texture.size);
+}
+
 void RasterizerVulkan::Clear(u32 layer_count) {
     MICROPROFILE_SCOPE(Vulkan_Clearing);
 
diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.h b/src/video_core/renderer_vulkan/vk_rasterizer.h
index 472cc64d9..a0508b57c 100644
--- a/src/video_core/renderer_vulkan/vk_rasterizer.h
+++ b/src/video_core/renderer_vulkan/vk_rasterizer.h
@@ -66,6 +66,7 @@ public:
 
     void Draw(bool is_indexed, u32 instance_count) override;
     void DrawIndirect() override;
+    void DrawTexture() override;
     void Clear(u32 layer_count) override;
     void DispatchCompute() override;
     void ResetCounter(VideoCore::QueryType type) override;