early-access version 1857

README.md

@@ -1,7 +1,7 @@
 yuzu emulator early access
 =============
 
-This is the source code for early-access 1855.
+This is the source code for early-access 1857.
 
 ## Legal Notice
 

src/common/thread_worker.h

@@ -81,7 +81,7 @@ public:
         {
             std::unique_lock lock{queue_mutex};
             requests.emplace(std::move(work));
-            ++work_scherduled;
+            ++work_scheduled;
         }
         condition.notify_one();
     }
@@ -94,7 +94,7 @@ public:
         });
         std::unique_lock lock{queue_mutex};
         wait_condition.wait(lock, [this] {
-            return workers_stopped >= workers_queued || work_done >= work_scherduled;
+            return workers_stopped >= workers_queued || work_done >= work_scheduled;
         });
     }
 
@@ -103,7 +103,7 @@ private:
     std::mutex queue_mutex;
     std::condition_variable_any condition;
     std::condition_variable wait_condition;
-    std::atomic<size_t> work_scherduled{};
+    std::atomic<size_t> work_scheduled{};
     std::atomic<size_t> work_done{};
     std::atomic<size_t> workers_stopped{};
     std::atomic<size_t> workers_queued{};

src/common/unique_function.h

@@ -14,7 +14,7 @@ class UniqueFunction {
     class CallableBase {
     public:
         virtual ~CallableBase() = default;
-        virtual ResultType operator()(Args...) = 0;
+        virtual ResultType operator()(Args&&...) = 0;
     };
 
     template <typename Functor>
@@ -23,7 +23,7 @@ class UniqueFunction {
         Callable(Functor&& functor_) : functor{std::move(functor_)} {}
         ~Callable() override = default;
 
-        ResultType operator()(Args... args) override {
+        ResultType operator()(Args&&... args) override {
             return functor(std::forward<Args>(args)...);
         }
 
@@ -38,25 +38,20 @@ public:
     UniqueFunction(Functor&& functor)
         : callable{std::make_unique<Callable<Functor>>(std::move(functor))} {}
 
-    UniqueFunction& operator=(UniqueFunction<ResultType, Args...>&& rhs) noexcept {
+    UniqueFunction& operator=(UniqueFunction&& rhs) noexcept = default;
-        callable = std::move(rhs.callable);
+    UniqueFunction(UniqueFunction&& rhs) noexcept = default;
-        return *this;
-    }
 
-    UniqueFunction(UniqueFunction<ResultType, Args...>&& rhs) noexcept
+    UniqueFunction& operator=(const UniqueFunction&) = delete;
-        : callable{std::move(rhs.callable)} {}
+    UniqueFunction(const UniqueFunction&) = delete;
 
-    ResultType operator()(Args... args) const {
+    ResultType operator()(Args&&... args) const {
         return (*callable)(std::forward<Args>(args)...);
     }
 
     explicit operator bool() const noexcept {
-        return callable != nullptr;
+        return static_cast<bool>(callable);
     }
 
-    UniqueFunction& operator=(const UniqueFunction<ResultType, Args...>&) = delete;
-    UniqueFunction(const UniqueFunction<ResultType, Args...>&) = delete;
-
 private:
     std::unique_ptr<CallableBase> callable;
 };

src/video_core/engines/fermi_2d.cpp

@@ -7,6 +7,10 @@
 #include "video_core/engines/fermi_2d.h"
 #include "video_core/memory_manager.h"
 #include "video_core/rasterizer_interface.h"
+#include "video_core/surface.h"
+
+using VideoCore::Surface::BytesPerBlock;
+using VideoCore::Surface::PixelFormatFromRenderTargetFormat;
 
 namespace Tegra::Engines {
 
@@ -49,7 +53,7 @@ void Fermi2D::Blit() {
     UNIMPLEMENTED_IF_MSG(regs.clip_enable != 0, "Clipped blit enabled");
 
     const auto& args = regs.pixels_from_memory;
-    const Config config{
+    Config config{
         .operation = regs.operation,
         .filter = args.sample_mode.filter,
         .dst_x0 = args.dst_x0,
@@ -61,7 +65,21 @@ void Fermi2D::Blit() {
         .src_x1 = static_cast<s32>((args.du_dx * args.dst_width + args.src_x0) >> 32),
         .src_y1 = static_cast<s32>((args.dv_dy * args.dst_height + args.src_y0) >> 32),
     };
-    if (!rasterizer->AccelerateSurfaceCopy(regs.src, regs.dst, config)) {
+    Surface src = regs.src;
+    const auto bytes_per_pixel = BytesPerBlock(PixelFormatFromRenderTargetFormat(src.format));
+    const auto need_align_to_pitch =
+        src.linear == Tegra::Engines::Fermi2D::MemoryLayout::Pitch &&
+        static_cast<s32>(src.width) == config.src_x1 &&
+        config.src_x1 > static_cast<s32>(src.pitch / bytes_per_pixel) && config.src_x0 > 0;
+    if (need_align_to_pitch) {
+        auto address = src.Address() + config.src_x0 * bytes_per_pixel;
+        src.addr_upper = static_cast<u32>(address >> 32);
+        src.addr_lower = static_cast<u32>(address);
+        src.width -= config.src_x0;
+        config.src_x1 -= config.src_x0;
+        config.src_x0 = 0;
+    }
+    if (!rasterizer->AccelerateSurfaceCopy(src, regs.dst, config)) {
         UNIMPLEMENTED();
     }
 }

src/video_core/renderer_opengl/util_shaders.cpp

@@ -261,9 +261,9 @@ void UtilShaders::CopyBC4(Image& dst_image, Image& src_image, std::span<const Im
         glUniform3ui(LOC_SRC_OFFSET, copy.src_offset.x, copy.src_offset.y, copy.src_offset.z);
         glUniform3ui(LOC_DST_OFFSET, copy.dst_offset.x, copy.dst_offset.y, copy.dst_offset.z);
         glBindImageTexture(BINDING_INPUT_IMAGE, src_image.StorageHandle(),
-                           copy.src_subresource.base_level, GL_FALSE, 0, GL_READ_ONLY, GL_RG32UI);
+                           copy.src_subresource.base_level, GL_TRUE, 0, GL_READ_ONLY, GL_RG32UI);
         glBindImageTexture(BINDING_OUTPUT_IMAGE, dst_image.StorageHandle(),
-                           copy.dst_subresource.base_level, GL_FALSE, 0, GL_WRITE_ONLY, GL_RGBA8UI);
+                           copy.dst_subresource.base_level, GL_TRUE, 0, GL_WRITE_ONLY, GL_RGBA8UI);
         glDispatchCompute(copy.extent.width, copy.extent.height, copy.extent.depth);
     }
     program_manager.RestoreGuestCompute();

src/video_core/texture_cache/texture_cache.h

@@ -159,9 +159,7 @@ public:
     /// Blit an image with the given parameters
     void BlitImage(const Tegra::Engines::Fermi2D::Surface& dst,
                    const Tegra::Engines::Fermi2D::Surface& src,
-                   const Tegra::Engines::Fermi2D::Config& copy,
+                   const Tegra::Engines::Fermi2D::Config& copy);
-                   std::optional<Region2D> src_region_override = {},
-                   std::optional<Region2D> dst_region_override = {});
 
     /// Invalidate the contents of the color buffer index
     /// These contents become unspecified, the cache can assume aggressive optimizations.
@@ -760,9 +758,7 @@ void TextureCache<P>::UnmapGPUMemory(GPUVAddr gpu_addr, size_t size) {
 template <class P>
 void TextureCache<P>::BlitImage(const Tegra::Engines::Fermi2D::Surface& dst,
                                 const Tegra::Engines::Fermi2D::Surface& src,
-                                const Tegra::Engines::Fermi2D::Config& copy,
+                                const Tegra::Engines::Fermi2D::Config& copy) {
-                                std::optional<Region2D> src_override,
-                                std::optional<Region2D> dst_override) {
     const BlitImages images = GetBlitImages(dst, src);
     const ImageId dst_id = images.dst_id;
     const ImageId src_id = images.src_id;
@@ -773,47 +769,25 @@ void TextureCache<P>::BlitImage(const Tegra::Engines::Fermi2D::Surface& dst,
     const ImageBase& src_image = slot_images[src_id];
 
     // TODO: Deduplicate
-    const std::optional dst_base = dst_image.TryFindBase(dst.Address());
-    const SubresourceRange dst_range{.base = dst_base.value(), .extent = {1, 1}};
-    const ImageViewInfo dst_view_info(ImageViewType::e2D, images.dst_format, dst_range);
-    const auto [dst_framebuffer_id, dst_view_id] = RenderTargetFromImage(dst_id, dst_view_info);
-    const auto [src_samples_x, src_samples_y] = SamplesLog2(src_image.info.num_samples);
-
-    // out of bounds texture blit checking
-    const bool use_override = src_override.has_value();
-    const s32 src_x0 = copy.src_x0 >> src_samples_x;
-    s32 src_x1 = use_override ? src_override->end.x : copy.src_x1 >> src_samples_x;
-    const s32 src_y0 = copy.src_y0 >> src_samples_y;
-    const s32 src_y1 = copy.src_y1 >> src_samples_y;
-
-    const auto src_width = static_cast<s32>(src_image.info.size.width);
-    const bool width_oob = src_x1 > src_width;
-    const auto width_diff = width_oob ? src_x1 - src_width : 0;
-    if (width_oob) {
-        src_x1 = src_width;
-    }
-
-    const Region2D src_dimensions{
-        Offset2D{.x = src_x0, .y = src_y0},
-        Offset2D{.x = src_x1, .y = src_y1},
-    };
-    const auto src_region = use_override ? *src_override : src_dimensions;
-
     const std::optional src_base = src_image.TryFindBase(src.Address());
     const SubresourceRange src_range{.base = src_base.value(), .extent = {1, 1}};
     const ImageViewInfo src_view_info(ImageViewType::e2D, images.src_format, src_range);
     const auto [src_framebuffer_id, src_view_id] = RenderTargetFromImage(src_id, src_view_info);
-    const auto [dst_samples_x, dst_samples_y] = SamplesLog2(dst_image.info.num_samples);
+    const auto [src_samples_x, src_samples_y] = SamplesLog2(src_image.info.num_samples);
-
+    const Region2D src_region{
-    const s32 dst_x0 = copy.dst_x0 >> dst_samples_x;
+        Offset2D{.x = copy.src_x0 >> src_samples_x, .y = copy.src_y0 >> src_samples_y},
-    const s32 dst_x1 = copy.dst_x1 >> dst_samples_x;
+        Offset2D{.x = copy.src_x1 >> src_samples_x, .y = copy.src_y1 >> src_samples_y},
-    const s32 dst_y0 = copy.dst_y0 >> dst_samples_y;
+    };
-    const s32 dst_y1 = copy.dst_y1 >> dst_samples_y;
+
-    const Region2D dst_dimensions{
+    const std::optional dst_base = dst_image.TryFindBase(dst.Address());
-        Offset2D{.x = dst_x0, .y = dst_y0},
+    const SubresourceRange dst_range{.base = dst_base.value(), .extent = {1, 1}};
-        Offset2D{.x = dst_x1 - width_diff, .y = dst_y1},
+    const ImageViewInfo dst_view_info(ImageViewType::e2D, images.dst_format, dst_range);
+    const auto [dst_framebuffer_id, dst_view_id] = RenderTargetFromImage(dst_id, dst_view_info);
+    const auto [dst_samples_x, dst_samples_y] = SamplesLog2(dst_image.info.num_samples);
+    const Region2D dst_region{
+        Offset2D{.x = copy.dst_x0 >> dst_samples_x, .y = copy.dst_y0 >> dst_samples_y},
+        Offset2D{.x = copy.dst_x1 >> dst_samples_x, .y = copy.dst_y1 >> dst_samples_y},
     };
-    const auto dst_region = use_override ? *dst_override : dst_dimensions;
 
     // Always call this after src_framebuffer_id was queried, as the address might be invalidated.
     Framebuffer* const dst_framebuffer = &slot_framebuffers[dst_framebuffer_id];
@@ -830,21 +804,6 @@ void TextureCache<P>::BlitImage(const Tegra::Engines::Fermi2D::Surface& dst,
         runtime.BlitImage(dst_framebuffer, dst_view, src_view, dst_region, src_region, copy.filter,
                           copy.operation);
     }
-
-    if (width_oob) {
-        // Continue copy of the oob region of the texture on the next row
-        auto oob_src = src;
-        oob_src.height++;
-        const Region2D src_region_override{
-            Offset2D{.x = 0, .y = src_y0 + 1},
-            Offset2D{.x = width_diff, .y = src_y1 + 1},
-        };
-        const Region2D dst_region_override{
-            Offset2D{.x = dst_x1 - width_diff, .y = dst_y0},
-            Offset2D{.x = dst_x1, .y = dst_y1},
-        };
-        BlitImage(dst, oob_src, copy, src_region_override, dst_region_override);
-    }
 }
 
 template <class P>

src/video_core/textures/astc.cpp

@@ -18,15 +18,11 @@
 #include <algorithm>
 #include <cassert>
 #include <cstring>
-#if __cpp_lib_parallel_algorithm
-#include <execution>
-#endif
 #include <span>
 #include <vector>
 
 #include <boost/container/static_vector.hpp>
 
-#include "common/alignment.h"
 #include "common/common_types.h"
 #include "video_core/textures/astc.h"
 
@@ -1554,87 +1550,30 @@ static void DecompressBlock(std::span<const u8, 16> inBuf, const u32 blockWidth,
 
 void Decompress(std::span<const uint8_t> data, uint32_t width, uint32_t height, uint32_t depth,
                 uint32_t block_width, uint32_t block_height, std::span<uint8_t> output) {
-    struct ASTCStrideInfo {
+    u32 block_index = 0;
-        u32 z{};
+    std::size_t depth_offset = 0;
-        u32 index{};
+    for (u32 z = 0; z < depth; z++) {
-    };
+        for (u32 y = 0; y < height; y += block_height) {
-
+            for (u32 x = 0; x < width; x += block_width) {
-    const u32 rows = Common::DivideUp(height, block_height);
-    const u32 cols = Common::DivideUp(width, block_width);
-
-    const u32 num_strides = depth * rows;
-    std::vector<ASTCStrideInfo> astc_strides(num_strides);
-
-    for (u32 z = 0; z < depth; ++z) {
-        for (u32 index = 0; index < rows; ++index) {
-            astc_strides.emplace_back(ASTCStrideInfo{
-                .z{z},
-                .index{index},
-            });
-        }
-    }
-
-    auto decompress_stride = [&](const ASTCStrideInfo& stride) {
-        const u32 y = stride.index * block_height;
-        const u32 depth_offset = stride.z * height * width * 4;
-        for (u32 x_index = 0; x_index < cols; ++x_index) {
-            const u32 block_index = (stride.z * rows * cols) + (stride.index * cols) + x_index;
-            const u32 x = x_index * block_width;
-
                 const std::span<const u8, 16> blockPtr{data.subspan(block_index * 16, 16)};
 
                 // Blocks can be at most 12x12
                 std::array<u32, 12 * 12> uncompData;
                 DecompressBlock(blockPtr, block_width, block_height, uncompData);
 
-            const u32 decompWidth = std::min(block_width, width - x);
+                u32 decompWidth = std::min(block_width, width - x);
-            const u32 decompHeight = std::min(block_height, height - y);
+                u32 decompHeight = std::min(block_height, height - y);
 
                 const std::span<u8> outRow = output.subspan(depth_offset + (y * width + x) * 4);
-
+                for (u32 jj = 0; jj < decompHeight; jj++) {
-            for (u32 h = 0; h < decompHeight; ++h) {
+                    std::memcpy(outRow.data() + jj * width * 4,
-                std::memcpy(outRow.data() + h * width * 4, uncompData.data() + h * block_width,
+                                uncompData.data() + jj * block_width, decompWidth * 4);
-                            decompWidth * 4);
+                }
+                ++block_index;
             }
         }
-    };
+        depth_offset += height * width * 4;
-
+    }
-#if __cpp_lib_parallel_algorithm
-    std::for_each(std::execution::par, astc_strides.cbegin(), astc_strides.cend(),
-                  decompress_stride);
-#else
-    std::for_each(astc_strides.cbegin(), astc_strides.cend(), decompress_stride);
-#endif
-
-    // const u32 rows = Common::DivideUp(height, block_height);
-    // const u32 cols = Common::DivideUp(width, block_width);
-
-    // for (u32 z = 0; z < depth; ++z) {
-    //     const u32 depth_offset = z * height * width * 4;
-    //     for (u32 y_index = 0; y_index < rows; ++y_index) {
-    //         const u32 y = y_index * block_height;
-    //         for (u32 x_index = 0; x_index < cols; ++x_index) {
-    //             const u32 block_index = (z * rows * cols) + (y_index * cols) + x_index;
-    //             const u32 x = x_index * block_width;
-
-    //             const std::span<const u8, 16> blockPtr{data.subspan(block_index * 16, 16)};
-
-    //             // Blocks can be at most 12x12
-    //             std::array<u32, 12 * 12> uncompData;
-    //             DecompressBlock(blockPtr, block_width, block_height, uncompData);
-
-    //             u32 decompWidth = std::min(block_width, width - x);
-    //             u32 decompHeight = std::min(block_height, height - y);
-
-    //             const std::span<u8> outRow = output.subspan(depth_offset + (y * width + x) * 4);
-    //             for (u32 h = 0; h < decompHeight; ++h) {
-    //                 std::memcpy(outRow.data() + h * width * 4, uncompData.data() + h *
-    //                 block_width,
-    //                             decompWidth * 4);
-    //             }
-    //         }
-    //     }
-    // }
 }
 
 } // namespace Tegra::Texture::ASTC