Update for new hades

app/src/main/cpp/skyline/common/dirty_tracking.h

@@ -33,7 +33,7 @@ namespace skyline::dirty {
      * @tparam OverlapPoolSize Size of the pool used to store handles when there are multiple bound to the same subresource
      * @note This class is *NOT* thread-safe
      */
-    template<size_t ManagedResourceSize, size_t Granularity, size_t OverlapPoolSize = 0x1000>
+    template<size_t ManagedResourceSize, size_t Granularity, size_t OverlapPoolSize = 0x2000>
     class Manager {
       private:
         struct BindingState {

app/src/main/cpp/skyline/gpu/interconnect/maxwell_3d/active_state.cpp

@@ -218,7 +218,8 @@ namespace skyline::gpu::interconnect::maxwell3d {
                      viewport.offsetX, viewport.offsetY, viewport.scaleX, viewport.scaleY, viewport.swizzle,
                      viewportClip,
                      windowOrigin,
-                     viewportScaleOffsetEnable);
+                     viewportScaleOffsetEnable,
+                     surfaceClip);
     }
 
     ViewportState::ViewportState(dirty::Handle dirtyHandle, DirtyManager &manager, const EngineRegisters &engine, u32 index) : engine{manager, dirtyHandle, engine}, index{index} {}
@@ -258,14 +259,20 @@ namespace skyline::gpu::interconnect::maxwell3d {
         if (index != 0 && !ctx.gpu.traits.supportsMultipleViewports)
             return;
 
-        if (!engine->viewportScaleOffsetEnable)
+        if (!engine->viewportScaleOffsetEnable) {
-            // https://github.com/Ryujinx/Ryujinx/pull/3328
+            builder.SetViewport(index, vk::Viewport{
-            Logger::Warn("Viewport scale/offset disable is unimplemented");
+                .x = static_cast<float>(engine->surfaceClip.horizontal.x),
-
+                .y = static_cast<float>(engine->surfaceClip.vertical.y),
-        if (engine->viewport.scaleX == 0.0f || engine->viewport.scaleY == 0.0f)
+                .width = engine->surfaceClip.horizontal.width ? static_cast<float>(engine->surfaceClip.horizontal.width) : 1.0f,
+                .height = engine->surfaceClip.vertical.height ? static_cast<float>(engine->surfaceClip.vertical.height) : 1.0f,
+                .minDepth = 0.0f,
+                .maxDepth = 1.0f,
+            });
+        } else if (engine->viewport.scaleX == 0.0f || engine->viewport.scaleY == 0.0f) {
             builder.SetViewport(index, ConvertViewport(engine->viewport0, engine->viewportClip0, engine->windowOrigin, engine->viewportScaleOffsetEnable));
-        else
+        } else {
             builder.SetViewport(index, ConvertViewport(engine->viewport, engine->viewportClip, engine->windowOrigin, engine->viewportScaleOffsetEnable));
+        }
     }
 
     /* Scissor */

app/src/main/cpp/skyline/gpu/interconnect/maxwell_3d/active_state.h

@@ -92,6 +92,7 @@ namespace skyline::gpu::interconnect::maxwell3d {
             const engine::ViewportClip &viewportClip;
             const engine::WindowOrigin &windowOrigin;
             const u32 &viewportScaleOffsetEnable;
+            const engine::SurfaceClip &surfaceClip;
 
             void DirtyBind(DirtyManager &manager, dirty::Handle handle) const;
         };

app/src/main/cpp/skyline/gpu/interconnect/maxwell_3d/packed_pipeline_state.h

@@ -59,6 +59,7 @@ namespace skyline::gpu::interconnect::maxwell3d {
             bool alphaTestEnable : 1;
             bool depthClampEnable : 1; // Use SetDepthClampEnable
             bool dynamicStateActive : 1;
+            bool viewportTransformEnable : 1;
         };
 
         u32 patchSize;

app/src/main/cpp/skyline/gpu/interconnect/maxwell_3d/pipeline_manager.cpp

@@ -192,6 +192,7 @@ namespace skyline::gpu::interconnect::maxwell3d {
                 ConvertCompilerShaderStage(static_cast<PipelineStage>(i)),
                 shaderBinaries[i].binary, shaderBinaries[i].baseOffset,
                 packedState.bindlessTextureConstantBufferSlotSelect,
+                packedState.viewportTransformEnable,
                 [&](u32 index, u32 offset) {
                     size_t shaderStage{i > 0 ? (i - 1) : 0};
                     return constantBuffers[shaderStage][index].Read<int>(ctx.executor, offset);

app/src/main/cpp/skyline/gpu/interconnect/maxwell_3d/pipeline_state.cpp

@@ -444,7 +444,7 @@ namespace skyline::gpu::interconnect::maxwell3d {
 
     /* Global Shader Config State */
     void GlobalShaderConfigState::EngineRegisters::DirtyBind(DirtyManager &manager, dirty::Handle handle) const {
-        manager.Bind(handle, postVtgShaderAttributeSkipMask, bindlessTexture, apiMandatedEarlyZ);
+        manager.Bind(handle, postVtgShaderAttributeSkipMask, bindlessTexture, apiMandatedEarlyZ, viewportScaleOffsetEnable);
     }
 
     GlobalShaderConfigState::GlobalShaderConfigState(const EngineRegisters &engine) : engine{engine} {}
@@ -453,6 +453,7 @@ namespace skyline::gpu::interconnect::maxwell3d {
         packedState.postVtgShaderAttributeSkipMask = engine.postVtgShaderAttributeSkipMask;
         packedState.bindlessTextureConstantBufferSlotSelect = engine.bindlessTexture.constantBufferSlotSelect;
         packedState.apiMandatedEarlyZ = engine.apiMandatedEarlyZ;
+        packedState.viewportTransformEnable = engine.viewportScaleOffsetEnable;
     }
 
     /* Pipeline State */

app/src/main/cpp/skyline/gpu/interconnect/maxwell_3d/pipeline_state.h

@@ -258,6 +258,7 @@ namespace skyline::gpu::interconnect::maxwell3d {
             const std::array<u32, 8> &postVtgShaderAttributeSkipMask;
             const engine_common::BindlessTexture &bindlessTexture;
             const u32 &apiMandatedEarlyZ;
+            const u32 &viewportScaleOffsetEnable;
 
             void DirtyBind(DirtyManager &manager, dirty::Handle handle) const;
         };

app/src/main/cpp/skyline/gpu/shader_manager.cpp

@@ -32,6 +32,7 @@ namespace skyline::gpu {
             .support_float16 = traits.supportsFloat16,
             .support_int64 = traits.supportsInt64,
             .needs_demote_reorder = false,
+            .support_snorm_render_buffer = true
         };
 
         constexpr u32 TegraX1WarpSize{32}; //!< The amount of threads in a warp on the Tegra X1
@@ -90,14 +91,21 @@ namespace skyline::gpu {
         span<u8> binary;
         u32 baseOffset;
         u32 textureBufferIndex;
+        bool viewportTransformEnabled;
         ShaderManager::ConstantBufferRead constantBufferRead;
         ShaderManager::GetTextureType getTextureType;
 
       public:
-        std::vector<ShaderManager::ConstantBufferWord> constantBufferWords;
+        GraphicsEnvironment(const std::array<u32, 8> &postVtgShaderAttributeSkipMask,
-        std::vector<ShaderManager::CachedTextureType> textureTypes;
+                            Shader::Stage pStage,
-
+                            span<u8> pBinary, u32 baseOffset,
-        GraphicsEnvironment(const std::array<u32, 8> &postVtgShaderAttributeSkipMask, Shader::Stage pStage, span<u8> pBinary, u32 baseOffset, u32 textureBufferIndex, ShaderManager::ConstantBufferRead constantBufferRead, ShaderManager::GetTextureType getTextureType) : binary{pBinary}, baseOffset{baseOffset}, textureBufferIndex{textureBufferIndex}, constantBufferRead{std::move(constantBufferRead)}, getTextureType{std::move(getTextureType)} {
+                            u32 textureBufferIndex,
+                            bool viewportTransformEnabled,
+                            ShaderManager::ConstantBufferRead constantBufferRead, ShaderManager::GetTextureType getTextureType)
+            : binary{pBinary}, baseOffset{baseOffset},
+              textureBufferIndex{textureBufferIndex},
+              viewportTransformEnabled{viewportTransformEnabled},
+              constantBufferRead{std::move(constantBufferRead)}, getTextureType{std::move(getTextureType)} {
             gp_passthrough_mask = postVtgShaderAttributeSkipMask;
             stage = pStage;
             sph = *reinterpret_cast<Shader::ProgramHeader *>(binary.data());
@@ -112,15 +120,19 @@ namespace skyline::gpu {
         }
 
         [[nodiscard]] u32 ReadCbufValue(u32 index, u32 offset) final {
-            auto value{constantBufferRead(index, offset)};
+            return constantBufferRead(index, offset);
-            constantBufferWords.emplace_back(index, offset, value);
+        }
-            return value;
+
+        [[nodiscard]] Shader::TexturePixelFormat ReadTexturePixelFormat(u32 handle) final {
+            throw exception("ReadTexturePixelFormat not implemented");
         }
 
         [[nodiscard]] Shader::TextureType ReadTextureType(u32 handle) final {
-            auto type{getTextureType(handle)};
+            return getTextureType(handle);
-            textureTypes.emplace_back(handle, type);
+        }
-            return type;
+
+        [[nodiscard]] u32 ReadViewportTransformState() final {
+            return viewportTransformEnabled ? 1 : 0; // Only relevant for graphics shaders
         }
 
         [[nodiscard]] u32 TextureBoundBuffer() const final {
@@ -156,11 +168,7 @@ namespace skyline::gpu {
         ShaderManager::ConstantBufferRead constantBufferRead;
         ShaderManager::GetTextureType getTextureType;
 
-
       public:
-        std::vector<ShaderManager::ConstantBufferWord> constantBufferWords;
-        std::vector<ShaderManager::CachedTextureType> textureTypes;
-
         ComputeEnvironment(span<u8> pBinary,
                            u32 baseOffset,
                            u32 textureBufferIndex,
@@ -187,15 +195,19 @@ namespace skyline::gpu {
         }
 
         [[nodiscard]] u32 ReadCbufValue(u32 index, u32 offset) final {
-            auto value{constantBufferRead(index, offset)};
+            return constantBufferRead(index, offset);
-            constantBufferWords.emplace_back(index, offset, value);
+        }
-            return value;
+
+        [[nodiscard]] Shader::TexturePixelFormat ReadTexturePixelFormat(u32 handle) final {
+            throw exception("ReadTexturePixelFormat not implemented");
         }
 
         [[nodiscard]] Shader::TextureType ReadTextureType(u32 handle) final {
-            auto type{getTextureType(handle)};
+            return getTextureType(handle);
-            textureTypes.emplace_back(handle, type);
+        }
-            return type;
+
+        [[nodiscard]] u32 ReadViewportTransformState() final {
+            return 0; // Only relevant for graphics shaders
         }
 
         [[nodiscard]] u32 TextureBoundBuffer() const final {
@@ -240,6 +252,14 @@ namespace skyline::gpu {
             throw exception("Not implemented");
         }
 
+        [[nodiscard]] Shader::TexturePixelFormat ReadTexturePixelFormat(u32 handle) final {
+            throw exception("Not implemented");
+        }
+
+        [[nodiscard]] u32 ReadViewportTransformState() final {
+            throw exception("Not implemented");
+        }
+
         [[nodiscard]] u32 TextureBoundBuffer() const final {
             throw exception("Not implemented");
         }
@@ -259,14 +279,15 @@ namespace skyline::gpu {
         void Dump(u64 hash) final {}
     };
 
-    constexpr ShaderManager::ConstantBufferWord::ConstantBufferWord(u32 index, u32 offset, u32 value) : index(index), offset(offset), value(value) {}
+    Shader::IR::Program ShaderManager::ParseGraphicsShader(const std::array<u32, 8> &postVtgShaderAttributeSkipMask,
-
+                                                           Shader::Stage stage,
-    constexpr ShaderManager::CachedTextureType::CachedTextureType(u32 handle, Shader::TextureType type) : handle(handle), type(type) {}
+                                                           span<u8> binary, u32 baseOffset,
-
+                                                           u32 textureConstantBufferIndex,
-    Shader::IR::Program ShaderManager::ParseGraphicsShader(const std::array<u32, 8> &postVtgShaderAttributeSkipMask, Shader::Stage stage, span<u8> binary, u32 baseOffset, u32 textureConstantBufferIndex, const ConstantBufferRead &constantBufferRead, const GetTextureType &getTextureType) {
+                                                           bool viewportTransformEnabled,
+                                                           const ConstantBufferRead &constantBufferRead, const GetTextureType &getTextureType) {
         std::scoped_lock lock{poolMutex};
 
-        GraphicsEnvironment environment{postVtgShaderAttributeSkipMask, stage, binary, baseOffset, textureConstantBufferIndex, constantBufferRead, getTextureType};
+        GraphicsEnvironment environment{postVtgShaderAttributeSkipMask, stage, binary, baseOffset, textureConstantBufferIndex, viewportTransformEnabled, constantBufferRead, getTextureType};
         Shader::Maxwell::Flow::CFG cfg{environment, flowBlockPool, Shader::Maxwell::Location{static_cast<u32>(baseOffset + sizeof(Shader::ProgramHeader))}};
         return  Shader::Maxwell::TranslateProgram(instructionPool, blockPool, environment, cfg, hostTranslateInfo);
     }
@@ -278,7 +299,11 @@ namespace skyline::gpu {
         return Shader::Maxwell::MergeDualVertexPrograms(vertexA, vertexB, env);
     }
 
-    Shader::IR::Program ShaderManager::ParseComputeShader(span<u8> binary, u32 baseOffset, u32 textureConstantBufferIndex, u32 localMemorySize, u32 sharedMemorySize, std::array<u32, 3> workgroupDimensions, const ConstantBufferRead &constantBufferRead, const GetTextureType &getTextureType) {
+    Shader::IR::Program ShaderManager::ParseComputeShader(span<u8> binary, u32 baseOffset,
+                                                          u32 textureConstantBufferIndex,
+                                                          u32 localMemorySize, u32 sharedMemorySize,
+                                                          std::array<u32, 3> workgroupDimensions,
+                                                          const ConstantBufferRead &constantBufferRead, const GetTextureType &getTextureType) {
         std::scoped_lock lock{poolMutex};
 
         ComputeEnvironment environment{binary, baseOffset, textureConstantBufferIndex, localMemorySize, sharedMemorySize, workgroupDimensions, constantBufferRead, getTextureType};

app/src/main/cpp/skyline/gpu/shader_manager.h

@@ -31,37 +31,14 @@ namespace skyline::gpu {
 
       public:
         using ConstantBufferRead = std::function<u32(u32 index, u32 offset)>; //!< A function which reads a constant buffer at the specified offset and returns the value
-
-        /**
-         * @brief A single u32 word from a constant buffer with the offset it was read from, utilized to ensure constant buffer state is consistent
-         */
-        struct ConstantBufferWord {
-            u32 index; //!< The index of the constant buffer
-            u32 offset; //!< The offset of the constant buffer word
-            u32 value; //!< The contents of the word
-
-            constexpr ConstantBufferWord(u32 index, u32 offset, u32 value);
-
-            constexpr bool operator==(const ConstantBufferWord &other) const = default;
-        };
-
         using GetTextureType = std::function<Shader::TextureType(u32 handle)>; //!< A function which determines the type of a texture from its handle by checking the corresponding TIC
 
-        struct CachedTextureType {
-            u32 handle;
-            Shader::TextureType type;
-
-            constexpr CachedTextureType(u32 handle, Shader::TextureType type);
-
-            constexpr bool operator==(const CachedTextureType &other) const = default;
-        };
-
         ShaderManager(const DeviceState &state, GPU &gpu);
 
         /**
          * @return A shader program that corresponds to all the supplied state including the current state of the constant buffers
          */
-        Shader::IR::Program ParseGraphicsShader(const std::array<u32, 8> &postVtgShaderAttributeSkipMask, Shader::Stage stage, span<u8> binary, u32 baseOffset, u32 textureConstantBufferIndex, const ConstantBufferRead &constantBufferRead, const GetTextureType &getTextureType);
+        Shader::IR::Program ParseGraphicsShader(const std::array<u32, 8> &postVtgShaderAttributeSkipMask, Shader::Stage stage, span<u8> binary, u32 baseOffset, u32 textureConstantBufferIndex, bool viewportTransformEnabled, const ConstantBufferRead &constantBufferRead, const GetTextureType &getTextureType);
 
         /**
          * @brief Combines the VertexA and VertexB shader programs into a single program

app/src/main/cpp/skyline/soc/gm20b/engines/maxwell_3d.cpp

@@ -24,7 +24,7 @@ namespace skyline::soc::gm20b::engine::maxwell3d {
             .depthStencilRegisters = {*registers.depthTestEnable, *registers.depthWriteEnable, *registers.depthFunc, *registers.depthBoundsTestEnable, *registers.stencilTestEnable, *registers.twoSidedStencilTestEnable, *registers.stencilOps, *registers.stencilBack, *registers.alphaTestEnable, *registers.alphaFunc, *registers.alphaRef},
             .colorBlendRegisters = {*registers.logicOp, *registers.singleCtWriteControl, *registers.ctWrites, *registers.blendStatePerTargetEnable, *registers.blendPerTargets, *registers.blend},
             .transformFeedbackRegisters = {*registers.streamOutputEnable, *registers.streamOutControls, *registers.streamOutLayoutSelect},
-            .globalShaderConfigRegisters = {*registers.postVtgShaderAttributeSkipMask, *registers.bindlessTexture, *registers.apiMandatedEarlyZEnable},
+            .globalShaderConfigRegisters = {*registers.postVtgShaderAttributeSkipMask, *registers.bindlessTexture, *registers.apiMandatedEarlyZEnable, *registers.viewportScaleOffsetEnable},
             .ctSelect = *registers.ctSelect
         };
     }
@@ -35,7 +35,7 @@ namespace skyline::soc::gm20b::engine::maxwell3d {
             .vertexBuffersRegisters = util::MergeInto<REGTYPE(VertexBufferState), type::VertexStreamCount>(*registers.vertexStreams, *registers.vertexStreamLimits),
             .indexBufferRegisters = {*registers.indexBuffer},
             .transformFeedbackBuffersRegisters = util::MergeInto<REGTYPE(TransformFeedbackBufferState), type::StreamOutBufferCount>(*registers.streamOutBuffers, *registers.streamOutputEnable),
-            .viewportsRegisters = util::MergeInto<REGTYPE(ViewportState), type::ViewportCount>(registers.viewports[0], registers.viewportClips[0], *registers.viewports, *registers.viewportClips, *registers.windowOrigin, *registers.viewportScaleOffsetEnable),
+            .viewportsRegisters = util::MergeInto<REGTYPE(ViewportState), type::ViewportCount>(registers.viewports[0], registers.viewportClips[0], *registers.viewports, *registers.viewportClips, *registers.windowOrigin, *registers.viewportScaleOffsetEnable, *registers.surfaceClip),
             .scissorsRegisters = util::MergeInto<REGTYPE(ScissorState), type::ViewportCount>(*registers.scissors),
             .lineWidthRegisters = {*registers.lineWidth, *registers.lineWidthAliased, *registers.aliasedLineWidthEnable},
             .depthBiasRegisters = {*registers.depthBias, *registers.depthBiasClamp, *registers.slopeScaleDepthBias},