[TEST] Adjustments on pipeline hash collisions

src/video_core/renderer_vulkan/vk_render_pass_cache.cpp

@@ -145,18 +145,6 @@ VkRenderPass RenderPassCache::Get(const RenderPassKey& key) {
         .preserveAttachmentCount = 0,
         .pPreserveAttachments = nullptr,
     };
-    const VkSubpassDependency dependency{
-            .srcSubpass = 0,  // Current subpass
-            .dstSubpass = 0,  // Same subpass (self-dependency)
-            .srcStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT |
-                            VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT |
-                            VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT,
-            .dstStageMask = VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT,
-            .srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
-                             VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT,
-            .dstAccessMask = VK_ACCESS_SHADER_READ_BIT,
-            .dependencyFlags = VK_DEPENDENCY_BY_REGION_BIT
-    };
     pair->second = device->GetLogical().CreateRenderPass({
         .sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO,
         .pNext = nullptr,
@@ -165,8 +153,8 @@ VkRenderPass RenderPassCache::Get(const RenderPassKey& key) {
         .pAttachments = descriptions.empty() ? nullptr : descriptions.data(),
         .subpassCount = 1,
         .pSubpasses = &subpass,
-        .dependencyCount = 1,
-        .pDependencies = &dependency,
+        .dependencyCount = 0,
+        .pDependencies = nullptr,
     });
     return *pair->second;
 }

src/video_core/renderer_vulkan/vk_render_pass_cache.h

@@ -9,6 +9,7 @@
 #include <mutex>
 #include <ankerl/unordered_dense.h>
 
+#include "common/container_hash.h"
 #include "video_core/surface.h"
 #include "video_core/vulkan_common/vulkan_wrapper.h"
 
@@ -31,17 +32,26 @@ struct RenderPassKey {
 namespace std {
 template <>
 struct hash<Vulkan::RenderPassKey> {
+    static_assert(std::tuple_size_v<decltype(Vulkan::RenderPassKey::color_formats)> <= 8);
+    static_assert(static_cast<u32>(VideoCore::Surface::PixelFormat::Invalid) <= 0xFF);
+    static_assert(static_cast<u32>(VideoCore::Surface::PixelFormat::Max) <= 0xFF);
+    static_assert(VK_SAMPLE_COUNT_64_BIT <= 0xFF);
+
     [[nodiscard]] 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;
-        value ^= static_cast<size_t>(key.resolve_color) << 63;
-        value ^= static_cast<size_t>(key.color_clear_mask) << 54;
-        value ^= static_cast<size_t>(key.depth_stencil_clear) << 62;
-        value ^= static_cast<size_t>(key.color_discard_mask) << 24;
-        for (size_t i = 0; i < key.color_formats.size(); ++i) {
-            value ^= static_cast<size_t>(key.color_formats[i]) << (i * 6);
+        u64 formats = 0;
+        for (size_t index = 0; index < key.color_formats.size(); ++index) {
+            formats |= static_cast<u64>(key.color_formats[index]) << (index * 8);
         }
-        return value;
+        const u64 state = static_cast<u64>(key.depth_format) |
+                          (static_cast<u64>(key.samples) << 8) |
+                          (static_cast<u64>(key.color_clear_mask) << 16) |
+                          (static_cast<u64>(key.color_discard_mask) << 24) |
+                          (static_cast<u64>(key.resolve_color) << 32) |
+                          (static_cast<u64>(key.depth_stencil_clear) << 33);
+        size_t seed = 0;
+        Common::HashCombine(seed, formats);
+        Common::HashCombine(seed, state);
+        return seed;
     }
 };
 } // namespace std

src/video_core/renderer_vulkan/vk_texture_cache.cpp

@@ -628,18 +628,12 @@ void CopyBufferToImage(vk::CommandBuffer cmdbuf, VkBuffer src_buffer, VkImage im
             .subresourceRange = subresource_range,
     };
 
-    cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT |
-                           VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT |
-                           VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, 0,
+    cmdbuf.PipelineBarrier(vk::PIPELINE_STAGE_GRAPHICS_COMPUTE, 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_LATE_FRAGMENT_TESTS_BIT |
-            VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT |
-            VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT,
-            0, nullptr, nullptr, write_barrier);
+    cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, vk::PIPELINE_STAGE_GRAPHICS_COMPUTE, 0,
+                           nullptr, nullptr, write_barrier);
 }
 
 [[nodiscard]] VkImageBlit MakeImageBlit(const Region2D& dst_region, const Region2D& src_region,
@@ -2838,11 +2832,25 @@ VkRenderPass Framebuffer::RenderPassVariant(u32 color_clear_mask, bool depth_ste
     if (color_clear_mask == 0 && !depth_stencil_clear && color_discard_mask == 0) {
         return renderpass;
     }
+    static_assert(NUM_RT <= 8);
+    const u32 variant_key = color_clear_mask | (color_discard_mask << 8) |
+                            (static_cast<u32>(depth_stencil_clear) << 16);
+    for (u32 index = 0; index < num_memoized_variants; ++index) {
+        if (variant_keys[index] == variant_key) {
+            return variant_render_passes[index];
+        }
+    }
     RenderPassKey key = render_pass_key;
     key.color_clear_mask = color_clear_mask;
     key.depth_stencil_clear = depth_stencil_clear;
     key.color_discard_mask = color_discard_mask;
-    return render_pass_cache->Get(key);
+    const VkRenderPass variant = render_pass_cache->Get(key);
+    if (num_memoized_variants < variant_keys.size()) {
+        variant_keys[num_memoized_variants] = variant_key;
+        variant_render_passes[num_memoized_variants] = variant;
+        ++num_memoized_variants;
+    }
+    return variant;
 }
 
 void TextureCacheRuntime::AccelerateImageUpload(

src/video_core/renderer_vulkan/vk_texture_cache.h

@@ -248,6 +248,8 @@ public:
     }
 
 private:
+    static constexpr size_t NUM_MEMOIZED_RENDER_PASS_VARIANTS = 8;
+
     vk::Framebuffer framebuffer;
     VkRenderPass renderpass{};
     VkExtent2D render_area{};
@@ -265,6 +267,9 @@ private:
     RenderPassKey render_pass_key{};
     RenderPassCache* render_pass_cache{nullptr};
     bool discard_msaa_color{};
+    mutable std::array<u32, NUM_MEMOIZED_RENDER_PASS_VARIANTS> variant_keys{};
+    mutable std::array<VkRenderPass, NUM_MEMOIZED_RENDER_PASS_VARIANTS> variant_render_passes{};
+    mutable u32 num_memoized_variants{};
 };
 
 class Image : public VideoCommon::ImageBase {

src/video_core/textures/decoders.cpp

@@ -36,8 +36,7 @@ constexpr u32 SWIZZLE_RUN_SHIFT = 4;
 constexpr u32 SWIZZLE_RUN_MASK = SWIZZLE_RUN_BYTES - 1;
 constexpr u32 SWIZZLE_RUN_INDEX_MASK = GOB_SIZE_X / SWIZZLE_RUN_BYTES - 1;
 
-static_assert((SWIZZLE_X_BITS & SWIZZLE_RUN_MASK) == SWIZZLE_RUN_MASK,
-              "A swizzled run is only contiguous while the low bits of X map to themselves");
+static_assert((SWIZZLE_X_BITS & SWIZZLE_RUN_MASK) == SWIZZLE_RUN_MASK);
 
 constexpr std::array<u32, GOB_SIZE_X / SWIZZLE_RUN_BYTES> SWIZZLE_X_RUN_TABLE = [] {
     std::array<u32, GOB_SIZE_X / SWIZZLE_RUN_BYTES> table{};