dynarmic: Update and enable DYNARMIC_IGNORE_ASSERTS

decoders: Optimize memcpy for the other functions

CMakeLists.txt

@@ -376,7 +376,7 @@ if (ENABLE_SDL2)
     if (YUZU_USE_BUNDLED_SDL2)
         # Detect toolchain and platform
         if ((MSVC_VERSION GREATER_EQUAL 1910 AND MSVC_VERSION LESS 1930) AND ARCHITECTURE_x86_64)
-            set(SDL2_VER "SDL2-2.0.15-prerelease")
+            set(SDL2_VER "SDL2-2.0.16")
         else()
             message(FATAL_ERROR "No bundled SDL2 binaries for your toolchain. Disable YUZU_USE_BUNDLED_SDL2 and provide your own.")
         endif()
@@ -396,7 +396,7 @@ if (ENABLE_SDL2)
     elseif (YUZU_USE_EXTERNAL_SDL2)
         message(STATUS "Using SDL2 from externals.")
     else()
-        find_package(SDL2 2.0.15 REQUIRED)
+        find_package(SDL2 2.0.16 REQUIRED)
 
         # Some installations don't set SDL2_LIBRARIES
         if("${SDL2_LIBRARIES}" STREQUAL "")

externals/CMakeLists.txt

@@ -7,7 +7,9 @@ include(DownloadExternals)
 # xbyak
 if (ARCHITECTURE_x86 OR ARCHITECTURE_x86_64)
     add_library(xbyak INTERFACE)
-    target_include_directories(xbyak SYSTEM INTERFACE ./xbyak/xbyak)
+    file(MAKE_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}/xbyak/include)
+    file(COPY ${CMAKE_CURRENT_SOURCE_DIR}/xbyak/xbyak DESTINATION ${CMAKE_CURRENT_BINARY_DIR}/xbyak/include)
+    target_include_directories(xbyak SYSTEM INTERFACE ${CMAKE_CURRENT_BINARY_DIR}/xbyak/include)
     target_compile_definitions(xbyak INTERFACE XBYAK_NO_OP_NAMES)
 endif()
 
@@ -19,6 +21,7 @@ target_include_directories(catch-single-include INTERFACE catch/single_include)
 if (ARCHITECTURE_x86_64)
     set(DYNARMIC_TESTS OFF)
     set(DYNARMIC_NO_BUNDLED_FMT ON)
+    set(DYNARMIC_IGNORE_ASSERTS ON CACHE BOOL "" FORCE)
     add_subdirectory(dynarmic)
 endif()
 

src/common/settings.h

@@ -75,7 +75,7 @@ public:
      */
     explicit BasicSetting(const Type& default_val, const std::string& name)
         : default_value{default_val}, global{default_val}, label{name} {}
-    ~BasicSetting() = default;
+    virtual ~BasicSetting() = default;
 
     /**
      *  Returns a reference to the setting's value.
@@ -161,7 +161,7 @@ public:
     explicit BasicRangedSetting(const Type& default_val, const Type& min_val, const Type& max_val,
                                 const std::string& name)
         : BasicSetting<Type>{default_val, name}, minimum{min_val}, maximum{max_val} {}
-    ~BasicRangedSetting() = default;
+    virtual ~BasicRangedSetting() = default;
 
     /**
      * Like BasicSetting's SetValue, except value is clamped to the range of the setting.
@@ -208,7 +208,7 @@ public:
      */
     explicit Setting(const Type& default_val, const std::string& name)
         : BasicSetting<Type>(default_val, name) {}
-    ~Setting() = default;
+    virtual ~Setting() = default;
 
     /**
      * Tells this setting to represent either the global or custom setting when other member
@@ -237,13 +237,13 @@ public:
      *
      * @returns The required value of the setting
      */
-    [[nodiscard]] const Type& GetValue() const override {
+    [[nodiscard]] virtual const Type& GetValue() const override {
         if (use_global) {
             return this->global;
         }
         return custom;
     }
-    [[nodiscard]] const Type& GetValue(bool need_global) const {
+    [[nodiscard]] virtual const Type& GetValue(bool need_global) const {
         if (use_global || need_global) {
             return this->global;
         }
@@ -286,7 +286,7 @@ public:
      *
      * @returns A reference to the current setting value
      */
-    explicit operator const Type&() const override {
+    virtual explicit operator const Type&() const override {
         if (use_global) {
             return this->global;
         }
@@ -318,7 +318,22 @@ public:
         : BasicSetting<Type>{default_val, name},
           BasicRangedSetting<Type>{default_val, min_val, max_val, name}, Setting<Type>{default_val,
                                                                                        name} {}
-    ~RangedSetting() = default;
+    virtual ~RangedSetting() = default;
+
+    // The following are needed to avoid a MSVC bug
+    // (source: https://stackoverflow.com/questions/469508)
+    [[nodiscard]] const Type& GetValue() const override {
+        return Setting<Type>::GetValue();
+    }
+    [[nodiscard]] const Type& GetValue(bool need_global) const override {
+        return Setting<Type>::GetValue(need_global);
+    }
+    explicit operator const Type&() const override {
+        if (this->use_global) {
+            return this->global;
+        }
+        return this->custom;
+    }
 
     /**
      * Like BasicSetting's SetValue, except value is clamped to the range of the setting. Sets the

src/common/threadsafe_queue.h

@@ -46,15 +46,13 @@ public:
         ElementPtr* new_ptr = new ElementPtr();
         write_ptr->next.store(new_ptr, std::memory_order_release);
         write_ptr = new_ptr;
+        ++size;
 
-        const size_t previous_size{size++};
-
-        // Acquire the mutex and then immediately release it as a fence.
+        // cv_mutex must be held or else there will be a missed wakeup if the other thread is in the
+        // line before cv.wait
         // TODO(bunnei): This can be replaced with C++20 waitable atomics when properly supported.
         // See discussion on https://github.com/yuzu-emu/yuzu/pull/3173 for details.
-        if (previous_size == 0) {
-            std::lock_guard lock{cv_mutex};
-        }
+        std::lock_guard lock{cv_mutex};
         cv.notify_one();
     }
 

src/common/x64/xbyak_abi.h

@@ -6,7 +6,7 @@
 
 #include <bitset>
 #include <initializer_list>
-#include <xbyak.h>
+#include <xbyak/xbyak.h>
 #include "common/assert.h"
 
 namespace Common::X64 {

src/common/x64/xbyak_util.h

@@ -5,7 +5,7 @@
 #pragma once
 
 #include <type_traits>
-#include <xbyak.h>
+#include <xbyak/xbyak.h>
 #include "common/x64/xbyak_abi.h"
 
 namespace Common::X64 {

src/input_common/sdl/sdl_impl.cpp

@@ -889,6 +889,9 @@ SDLState::SDLState() {
     RegisterFactory<VibrationDevice>("sdl", vibration_factory);
     RegisterFactory<MotionDevice>("sdl", motion_factory);
 
+    // Disable raw input. When enabled this setting causes SDL to die when a web applet opens
+    SDL_SetHint(SDL_HINT_JOYSTICK_RAWINPUT, "0");
+
     // Enable HIDAPI rumble. This prevents SDL from disabling motion on PS4 and PS5 controllers
     SDL_SetHint(SDL_HINT_JOYSTICK_HIDAPI_PS4_RUMBLE, "1");
     SDL_SetHint(SDL_HINT_JOYSTICK_HIDAPI_PS5_RUMBLE, "1");

src/video_core/command_classes/vic.cpp

@@ -96,12 +96,11 @@ void Vic::Execute() {
         if (!converted_frame_buffer) {
             converted_frame_buffer = AVMallocPtr{static_cast<u8*>(av_malloc(linear_size)), av_free};
         }
-
-        const int converted_stride{frame->width * 4};
+        const std::array<int, 4> converted_stride{frame->width * 4, frame->height * 4, 0, 0};
         u8* const converted_frame_buf_addr{converted_frame_buffer.get()};
 
         sws_scale(scaler_ctx, frame->data, frame->linesize, 0, frame->height,
-                  &converted_frame_buf_addr, &converted_stride);
+                  &converted_frame_buf_addr, converted_stride.data());
 
         const u32 blk_kind = static_cast<u32>(config.block_linear_kind);
         if (blk_kind != 0) {

src/video_core/macro/macro_jit_x64.h

@@ -6,7 +6,7 @@
 
 #include <array>
 #include <bitset>
-#include <xbyak.h>
+#include <xbyak/xbyak.h>
 #include "common/bit_field.h"
 #include "common/common_types.h"
 #include "common/x64/xbyak_abi.h"

src/video_core/textures/decoders.cpp

@@ -84,34 +84,107 @@ template <bool TO_LINEAR>
 void Swizzle(std::span<u8> output, std::span<const u8> input, u32 bytes_per_pixel, u32 width,
              u32 height, u32 depth, u32 block_height, u32 block_depth, u32 stride_alignment) {
     switch (bytes_per_pixel) {
-    case 1:
-        return SwizzleImpl<TO_LINEAR, 1>(output, input, width, height, depth, block_height,
+#define BPP_CASE(x)                                                                                \
+    case x:                                                                                        \
+        return SwizzleImpl<TO_LINEAR, x>(output, input, width, height, depth, block_height,        \
                                          block_depth, stride_alignment);
-    case 2:
-        return SwizzleImpl<TO_LINEAR, 2>(output, input, width, height, depth, block_height,
-                                         block_depth, stride_alignment);
-    case 3:
-        return SwizzleImpl<TO_LINEAR, 3>(output, input, width, height, depth, block_height,
-                                         block_depth, stride_alignment);
-    case 4:
-        return SwizzleImpl<TO_LINEAR, 4>(output, input, width, height, depth, block_height,
-                                         block_depth, stride_alignment);
-    case 6:
-        return SwizzleImpl<TO_LINEAR, 6>(output, input, width, height, depth, block_height,
-                                         block_depth, stride_alignment);
-    case 8:
-        return SwizzleImpl<TO_LINEAR, 8>(output, input, width, height, depth, block_height,
-                                         block_depth, stride_alignment);
-    case 12:
-        return SwizzleImpl<TO_LINEAR, 12>(output, input, width, height, depth, block_height,
-                                          block_depth, stride_alignment);
-    case 16:
-        return SwizzleImpl<TO_LINEAR, 16>(output, input, width, height, depth, block_height,
-                                          block_depth, stride_alignment);
+        BPP_CASE(1)
+        BPP_CASE(2)
+        BPP_CASE(3)
+        BPP_CASE(4)
+        BPP_CASE(6)
+        BPP_CASE(8)
+        BPP_CASE(12)
+        BPP_CASE(16)
+#undef BPP_CASE
     default:
         UNREACHABLE_MSG("Invalid bytes_per_pixel={}", bytes_per_pixel);
     }
 }
+
+template <u32 BYTES_PER_PIXEL>
+void SwizzleSubrect(u32 subrect_width, u32 subrect_height, u32 source_pitch, u32 swizzled_width,
+                    u8* swizzled_data, const u8* unswizzled_data, u32 block_height_bit,
+                    u32 offset_x, u32 offset_y) {
+    const u32 block_height = 1U << block_height_bit;
+    const u32 image_width_in_gobs =
+        (swizzled_width * BYTES_PER_PIXEL + (GOB_SIZE_X - 1)) / GOB_SIZE_X;
+    for (u32 line = 0; line < subrect_height; ++line) {
+        const u32 dst_y = line + offset_y;
+        const u32 gob_address_y =
+            (dst_y / (GOB_SIZE_Y * block_height)) * GOB_SIZE * block_height * image_width_in_gobs +
+            ((dst_y % (GOB_SIZE_Y * block_height)) / GOB_SIZE_Y) * GOB_SIZE;
+        const auto& table = SWIZZLE_TABLE[dst_y % GOB_SIZE_Y];
+        for (u32 x = 0; x < subrect_width; ++x) {
+            const u32 dst_x = x + offset_x;
+            const u32 gob_address =
+                gob_address_y + (dst_x * BYTES_PER_PIXEL / GOB_SIZE_X) * GOB_SIZE * block_height;
+            const u32 swizzled_offset = gob_address + table[(dst_x * BYTES_PER_PIXEL) % GOB_SIZE_X];
+            const u32 unswizzled_offset = line * source_pitch + x * BYTES_PER_PIXEL;
+
+            const u8* const source_line = unswizzled_data + unswizzled_offset;
+            u8* const dest_addr = swizzled_data + swizzled_offset;
+            std::memcpy(dest_addr, source_line, BYTES_PER_PIXEL);
+        }
+    }
+}
+
+template <u32 BYTES_PER_PIXEL>
+void UnswizzleSubrect(u32 line_length_in, u32 line_count, u32 pitch, u32 width, u32 block_height,
+                      u32 origin_x, u32 origin_y, u8* output, const u8* input) {
+    const u32 stride = width * BYTES_PER_PIXEL;
+    const u32 gobs_in_x = (stride + GOB_SIZE_X - 1) / GOB_SIZE_X;
+    const u32 block_size = gobs_in_x << (GOB_SIZE_SHIFT + block_height);
+
+    const u32 block_height_mask = (1U << block_height) - 1;
+    const u32 x_shift = GOB_SIZE_SHIFT + block_height;
+
+    for (u32 line = 0; line < line_count; ++line) {
+        const u32 src_y = line + origin_y;
+        const auto& table = SWIZZLE_TABLE[src_y % GOB_SIZE_Y];
+
+        const u32 block_y = src_y >> GOB_SIZE_Y_SHIFT;
+        const u32 src_offset_y = (block_y >> block_height) * block_size +
+                                 ((block_y & block_height_mask) << GOB_SIZE_SHIFT);
+        for (u32 column = 0; column < line_length_in; ++column) {
+            const u32 src_x = (column + origin_x) * BYTES_PER_PIXEL;
+            const u32 src_offset_x = (src_x >> GOB_SIZE_X_SHIFT) << x_shift;
+
+            const u32 swizzled_offset = src_offset_y + src_offset_x + table[src_x % GOB_SIZE_X];
+            const u32 unswizzled_offset = line * pitch + column * BYTES_PER_PIXEL;
+
+            std::memcpy(output + unswizzled_offset, input + swizzled_offset, BYTES_PER_PIXEL);
+        }
+    }
+}
+
+template <u32 BYTES_PER_PIXEL>
+void SwizzleSliceToVoxel(u32 line_length_in, u32 line_count, u32 pitch, u32 width, u32 height,
+                         u32 block_height, u32 block_depth, u32 origin_x, u32 origin_y, u8* output,
+                         const u8* input) {
+    UNIMPLEMENTED_IF(origin_x > 0);
+    UNIMPLEMENTED_IF(origin_y > 0);
+
+    const u32 stride = width * BYTES_PER_PIXEL;
+    const u32 gobs_in_x = (stride + GOB_SIZE_X - 1) / GOB_SIZE_X;
+    const u32 block_size = gobs_in_x << (GOB_SIZE_SHIFT + block_height + block_depth);
+
+    const u32 block_height_mask = (1U << block_height) - 1;
+    const u32 x_shift = static_cast<u32>(GOB_SIZE_SHIFT) + block_height + block_depth;
+
+    for (u32 line = 0; line < line_count; ++line) {
+        const auto& table = SWIZZLE_TABLE[line % GOB_SIZE_Y];
+        const u32 block_y = line / GOB_SIZE_Y;
+        const u32 dst_offset_y =
+            (block_y >> block_height) * block_size + (block_y & block_height_mask) * GOB_SIZE;
+        for (u32 x = 0; x < line_length_in; ++x) {
+            const u32 dst_offset =
+                ((x / GOB_SIZE_X) << x_shift) + dst_offset_y + table[x % GOB_SIZE_X];
+            const u32 src_offset = x * BYTES_PER_PIXEL + line * pitch;
+            std::memcpy(output + dst_offset, input + src_offset, BYTES_PER_PIXEL);
+        }
+    }
+}
 } // Anonymous namespace
 
 void UnswizzleTexture(std::span<u8> output, std::span<const u8> input, u32 bytes_per_pixel,
@@ -131,81 +204,67 @@ void SwizzleTexture(std::span<u8> output, std::span<const u8> input, u32 bytes_p
 void SwizzleSubrect(u32 subrect_width, u32 subrect_height, u32 source_pitch, u32 swizzled_width,
                     u32 bytes_per_pixel, u8* swizzled_data, const u8* unswizzled_data,
                     u32 block_height_bit, u32 offset_x, u32 offset_y) {
-    const u32 block_height = 1U << block_height_bit;
-    const u32 image_width_in_gobs =
-        (swizzled_width * bytes_per_pixel + (GOB_SIZE_X - 1)) / GOB_SIZE_X;
-    for (u32 line = 0; line < subrect_height; ++line) {
-        const u32 dst_y = line + offset_y;
-        const u32 gob_address_y =
-            (dst_y / (GOB_SIZE_Y * block_height)) * GOB_SIZE * block_height * image_width_in_gobs +
-            ((dst_y % (GOB_SIZE_Y * block_height)) / GOB_SIZE_Y) * GOB_SIZE;
-        const auto& table = SWIZZLE_TABLE[dst_y % GOB_SIZE_Y];
-        for (u32 x = 0; x < subrect_width; ++x) {
-            const u32 dst_x = x + offset_x;
-            const u32 gob_address =
-                gob_address_y + (dst_x * bytes_per_pixel / GOB_SIZE_X) * GOB_SIZE * block_height;
-            const u32 swizzled_offset = gob_address + table[(dst_x * bytes_per_pixel) % GOB_SIZE_X];
-            const u32 unswizzled_offset = line * source_pitch + x * bytes_per_pixel;
-
-            const u8* const source_line = unswizzled_data + unswizzled_offset;
-            u8* const dest_addr = swizzled_data + swizzled_offset;
-            std::memcpy(dest_addr, source_line, bytes_per_pixel);
-        }
+    switch (bytes_per_pixel) {
+#define BPP_CASE(x)                                                                                \
+    case x:                                                                                        \
+        return SwizzleSubrect<x>(subrect_width, subrect_height, source_pitch, swizzled_width,      \
+                                 swizzled_data, unswizzled_data, block_height_bit, offset_x,       \
+                                 offset_y);
+        BPP_CASE(1)
+        BPP_CASE(2)
+        BPP_CASE(3)
+        BPP_CASE(4)
+        BPP_CASE(6)
+        BPP_CASE(8)
+        BPP_CASE(12)
+        BPP_CASE(16)
+#undef BPP_CASE
+    default:
+        UNREACHABLE_MSG("Invalid bytes_per_pixel={}", bytes_per_pixel);
     }
 }
 
 void UnswizzleSubrect(u32 line_length_in, u32 line_count, u32 pitch, u32 width, u32 bytes_per_pixel,
                       u32 block_height, u32 origin_x, u32 origin_y, u8* output, const u8* input) {
-    const u32 stride = width * bytes_per_pixel;
-    const u32 gobs_in_x = (stride + GOB_SIZE_X - 1) / GOB_SIZE_X;
-    const u32 block_size = gobs_in_x << (GOB_SIZE_SHIFT + block_height);
-
-    const u32 block_height_mask = (1U << block_height) - 1;
-    const u32 x_shift = GOB_SIZE_SHIFT + block_height;
-
-    for (u32 line = 0; line < line_count; ++line) {
-        const u32 src_y = line + origin_y;
-        const auto& table = SWIZZLE_TABLE[src_y % GOB_SIZE_Y];
-
-        const u32 block_y = src_y >> GOB_SIZE_Y_SHIFT;
-        const u32 src_offset_y = (block_y >> block_height) * block_size +
-                                 ((block_y & block_height_mask) << GOB_SIZE_SHIFT);
-        for (u32 column = 0; column < line_length_in; ++column) {
-            const u32 src_x = (column + origin_x) * bytes_per_pixel;
-            const u32 src_offset_x = (src_x >> GOB_SIZE_X_SHIFT) << x_shift;
-
-            const u32 swizzled_offset = src_offset_y + src_offset_x + table[src_x % GOB_SIZE_X];
-            const u32 unswizzled_offset = line * pitch + column * bytes_per_pixel;
-
-            std::memcpy(output + unswizzled_offset, input + swizzled_offset, bytes_per_pixel);
-        }
+    switch (bytes_per_pixel) {
+#define BPP_CASE(x)                                                                                \
+    case x:                                                                                        \
+        return UnswizzleSubrect<x>(line_length_in, line_count, pitch, width, block_height,         \
+                                   origin_x, origin_y, output, input);
+        BPP_CASE(1)
+        BPP_CASE(2)
+        BPP_CASE(3)
+        BPP_CASE(4)
+        BPP_CASE(6)
+        BPP_CASE(8)
+        BPP_CASE(12)
+        BPP_CASE(16)
+#undef BPP_CASE
+    default:
+        UNREACHABLE_MSG("Invalid bytes_per_pixel={}", bytes_per_pixel);
     }
 }
 
 void SwizzleSliceToVoxel(u32 line_length_in, u32 line_count, u32 pitch, u32 width, u32 height,
                          u32 bytes_per_pixel, u32 block_height, u32 block_depth, u32 origin_x,
                          u32 origin_y, u8* output, const u8* input) {
-    UNIMPLEMENTED_IF(origin_x > 0);
-    UNIMPLEMENTED_IF(origin_y > 0);
-
-    const u32 stride = width * bytes_per_pixel;
-    const u32 gobs_in_x = (stride + GOB_SIZE_X - 1) / GOB_SIZE_X;
-    const u32 block_size = gobs_in_x << (GOB_SIZE_SHIFT + block_height + block_depth);
-
-    const u32 block_height_mask = (1U << block_height) - 1;
-    const u32 x_shift = static_cast<u32>(GOB_SIZE_SHIFT) + block_height + block_depth;
-
-    for (u32 line = 0; line < line_count; ++line) {
-        const auto& table = SWIZZLE_TABLE[line % GOB_SIZE_Y];
-        const u32 block_y = line / GOB_SIZE_Y;
-        const u32 dst_offset_y =
-            (block_y >> block_height) * block_size + (block_y & block_height_mask) * GOB_SIZE;
-        for (u32 x = 0; x < line_length_in; ++x) {
-            const u32 dst_offset =
-                ((x / GOB_SIZE_X) << x_shift) + dst_offset_y + table[x % GOB_SIZE_X];
-            const u32 src_offset = x * bytes_per_pixel + line * pitch;
-            std::memcpy(output + dst_offset, input + src_offset, bytes_per_pixel);
-        }
+    switch (bytes_per_pixel) {
+#define BPP_CASE(x)                                                                                \
+    case x:                                                                                        \
+        return SwizzleSliceToVoxel<x>(line_length_in, line_count, pitch, width, height,            \
+                                      block_height, block_depth, origin_x, origin_y, output,       \
+                                      input);
+        BPP_CASE(1)
+        BPP_CASE(2)
+        BPP_CASE(3)
+        BPP_CASE(4)
+        BPP_CASE(6)
+        BPP_CASE(8)
+        BPP_CASE(12)
+        BPP_CASE(16)
+#undef BPP_CASE
+    default:
+        UNREACHABLE_MSG("Invalid bytes_per_pixel={}", bytes_per_pixel);
     }
 }
 
@@ -228,7 +287,7 @@ void SwizzleKepler(const u32 width, const u32 height, const u32 dst_x, const u32
             u8* dest_addr = swizzle_data + swizzled_offset;
             count++;
 
-            std::memcpy(dest_addr, source_line, 1);
+            *dest_addr = *source_line;
         }
     }
 }

src/video_core/vulkan_common/vulkan_device.cpp

@@ -1161,8 +1161,7 @@ void Device::CollectPhysicalMemoryInfo() {
 }
 
 void Device::CollectToolingInfo() {
-    if (!ext_tooling_info || true) {
-        // Disabled to work around https://github.com/yuzu-emu/yuzu/issues/6835
+    if (!ext_tooling_info) {
         return;
     }
     const auto vkGetPhysicalDeviceToolPropertiesEXT =