early-access version 2649

README.md

@@ -1,7 +1,7 @@
 yuzu emulator early access
 =============
 
-This is the source code for early-access 2648.
+This is the source code for early-access 2649.
 
 ## Legal Notice
 

src/core/hid/emulated_controller.cpp

@@ -1207,4 +1207,12 @@ void EmulatedController::DeleteCallback(int key) {
     }
     callback_list.erase(iterator);
 }
+
+void EmulatedController::RemoveServiceCallbacks() {
+    std::lock_guard lock{mutex};
+    const auto count = std::erase_if(
+        callback_list, [](const auto& callback) { return callback.second.is_npad_service; });
+    LOG_DEBUG(Input, "Elements deleted {}", count);
+}
+
 } // namespace Core::HID

src/core/hid/emulated_controller.h

@@ -335,6 +335,9 @@ public:
      */
     void DeleteCallback(int key);
 
+    /// Removes all callbacks created from npad services
+    void RemoveServiceCallbacks();
+
 private:
     /// creates input devices from params
     void LoadDevices();

src/core/hid/hid_core.cpp

@@ -211,4 +211,17 @@ void HIDCore::UnloadInputDevices() {
     devices->UnloadInput();
 }
 
+void HIDCore::RemoveServiceCallbacks() {
+    player_1->RemoveServiceCallbacks();
+    player_2->RemoveServiceCallbacks();
+    player_3->RemoveServiceCallbacks();
+    player_4->RemoveServiceCallbacks();
+    player_5->RemoveServiceCallbacks();
+    player_6->RemoveServiceCallbacks();
+    player_7->RemoveServiceCallbacks();
+    player_8->RemoveServiceCallbacks();
+    other->RemoveServiceCallbacks();
+    handheld->RemoveServiceCallbacks();
+}
+
 } // namespace Core::HID

src/core/hid/hid_core.h

@@ -61,6 +61,9 @@ public:
     /// Removes all callbacks from input common
     void UnloadInputDevices();
 
+    /// Removes all callbacks from npad services
+    void RemoveServiceCallbacks();
+
     /// Number of emulated controllers
     static constexpr std::size_t available_controllers{10};
 

src/core/hle/kernel/board/nintendo/nx/k_system_control.cpp

@@ -148,9 +148,9 @@ u64 GenerateUniformRange(u64 min, u64 max, F f) {
 } // Anonymous namespace
 
 u64 KSystemControl::GenerateRandomU64() {
-    static std::random_device device;
+    std::random_device device;
-    static std::mt19937 gen(device());
+    std::mt19937 gen(device());
-    static std::uniform_int_distribution<u64> distribution(1, std::numeric_limits<u64>::max());
+    std::uniform_int_distribution<u64> distribution(1, std::numeric_limits<u64>::max());
     return distribution(gen);
 }
 

src/core/hle/kernel/k_auto_object.h

@@ -163,7 +163,7 @@ public:
         do {
             ASSERT(cur_ref_count > 0);
         } while (!m_ref_count.compare_exchange_weak(cur_ref_count, cur_ref_count - 1,
-                                                    std::memory_order_relaxed));
+                                                    std::memory_order_acq_rel));
 
         // If ref count hits zero, destroy the object.
         if (cur_ref_count - 1 == 0) {

src/core/hle/kernel/k_process.h

@@ -422,7 +422,7 @@ private:
     bool is_64bit_process = true;
 
     /// Total running time for the process in ticks.
-    u64 total_process_running_time_ticks = 0;
+    std::atomic<u64> total_process_running_time_ticks = 0;
 
     /// Per-process handle table for storing created object handles in.
     KHandleTable handle_table;

src/core/hle/kernel/k_scheduler_lock.h

@@ -4,6 +4,7 @@
 
 #pragma once
 
+#include <atomic>
 #include "common/assert.h"
 #include "core/hle/kernel/k_spin_lock.h"
 #include "core/hle/kernel/k_thread.h"
@@ -75,7 +76,7 @@ private:
     KernelCore& kernel;
     KAlignedSpinLock spin_lock{};
     s32 lock_count{};
-    KThread* owner_thread{};
+    std::atomic<KThread*> owner_thread{};
 };
 
 } // namespace Kernel

src/core/hle/kernel/k_thread.cpp

@@ -723,7 +723,7 @@ void KThread::UpdateState() {
     ASSERT(kernel.GlobalSchedulerContext().IsLocked());
 
     // Set our suspend flags in state.
-    const auto old_state = thread_state;
+    const ThreadState old_state = thread_state;
     const auto new_state =
         static_cast<ThreadState>(this->GetSuspendFlags()) | (old_state & ThreadState::Mask);
     thread_state = new_state;
@@ -738,7 +738,7 @@ void KThread::Continue() {
     ASSERT(kernel.GlobalSchedulerContext().IsLocked());
 
     // Clear our suspend flags in state.
-    const auto old_state = thread_state;
+    const ThreadState old_state = thread_state;
     thread_state = old_state & ThreadState::Mask;
 
     // Note the state change in scheduler.

src/core/hle/kernel/k_thread.h

@@ -5,6 +5,7 @@
 #pragma once
 
 #include <array>
+#include <atomic>
 #include <span>
 #include <string>
 #include <utility>
@@ -751,7 +752,7 @@ private:
     KAffinityMask original_physical_affinity_mask{};
     s32 original_physical_ideal_core_id{};
     s32 num_core_migration_disables{};
-    ThreadState thread_state{};
+    std::atomic<ThreadState> thread_state{};
     std::atomic<bool> termination_requested{};
     bool wait_cancelled{};
     bool cancellable{};

src/core/hle/kernel/kernel.cpp

@@ -85,7 +85,7 @@ struct KernelCore::Impl {
 
     void InitializeCores() {
         for (u32 core_id = 0; core_id < Core::Hardware::NUM_CPU_CORES; core_id++) {
-            cores[core_id].Initialize(current_process->Is64BitProcess());
+            cores[core_id].Initialize((*current_process).Is64BitProcess());
             system.Memory().SetCurrentPageTable(*current_process, core_id);
         }
     }
@@ -169,11 +169,11 @@ struct KernelCore::Impl {
 
         // Shutdown all processes.
         if (current_process) {
-            current_process->Finalize();
+            (*current_process).Finalize();
             // current_process->Close();
             // TODO: The current process should be destroyed based on accurate ref counting after
             // calling Close(). Adding a manual Destroy() call instead to avoid a memory leak.
-            current_process->Destroy();
+            (*current_process).Destroy();
             current_process = nullptr;
         }
 
@@ -713,7 +713,7 @@ struct KernelCore::Impl {
 
     // Lists all processes that exist in the current session.
     std::vector<KProcess*> process_list;
-    KProcess* current_process{};
+    std::atomic<KProcess*> current_process{};
     std::unique_ptr<Kernel::GlobalSchedulerContext> global_scheduler_context;
     Kernel::TimeManager time_manager;
 

src/video_core/host_shaders/CMakeLists.txt

@@ -18,6 +18,7 @@ set(SHADER_FILES
     full_screen_triangle.vert
     fxaa.frag
     fxaa.vert
+    opengl_convert_s8d24.comp
     opengl_copy_bc4.comp
     opengl_present.frag
     opengl_present.vert

src/video_core/host_shaders/opengl_convert_s8d24.comp

@@ -0,0 +1,18 @@
+// Copyright 2022 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#version 430 core
+
+layout(local_size_x = 16, local_size_y = 8) in;
+
+layout(binding = 0, rgba8ui) restrict uniform uimage2D destination;
+layout(location = 0) uniform uvec3 size;
+
+void main() {
+    if (any(greaterThanEqual(gl_GlobalInvocationID, size))) {
+        return;
+    }
+    uvec4 components = imageLoad(destination, ivec2(gl_GlobalInvocationID.xy));
+    imageStore(destination, ivec2(gl_GlobalInvocationID.xy), components.wxyz);
+}

src/video_core/renderer_opengl/gl_texture_cache.cpp

@@ -409,8 +409,8 @@ ImageBufferMap::~ImageBufferMap() {
 
 TextureCacheRuntime::TextureCacheRuntime(const Device& device_, ProgramManager& program_manager,
                                          StateTracker& state_tracker_)
-    : device{device_}, state_tracker{state_tracker_},
+    : device{device_}, state_tracker{state_tracker_}, util_shaders(program_manager),
-      util_shaders(program_manager), resolution{Settings::values.resolution_info} {
+      format_conversion_pass{util_shaders}, resolution{Settings::values.resolution_info} {
     static constexpr std::array TARGETS{GL_TEXTURE_1D_ARRAY, GL_TEXTURE_2D_ARRAY, GL_TEXTURE_3D};
     for (size_t i = 0; i < TARGETS.size(); ++i) {
         const GLenum target = TARGETS[i];
@@ -1325,6 +1325,9 @@ Framebuffer::Framebuffer(TextureCacheRuntime& runtime, std::span<ImageView*, NUM
 
 Framebuffer::~Framebuffer() = default;
 
+FormatConversionPass::FormatConversionPass(UtilShaders& util_shaders_)
+    : util_shaders{util_shaders_} {}
+
 void FormatConversionPass::ConvertImage(Image& dst_image, Image& src_image,
                                         std::span<const VideoCommon::ImageCopy> copies) {
     const GLenum dst_target = ImageTarget(dst_image.info);
@@ -1357,6 +1360,12 @@ void FormatConversionPass::ConvertImage(Image& dst_image, Image& src_image,
                             dst_origin.z, region.width, region.height, region.depth,
                             dst_image.GlFormat(), dst_image.GlType(), nullptr);
     }
+
+    // Swap component order of S8D24 to ABGR8 reinterprets
+    if (src_image.info.format == PixelFormat::D24_UNORM_S8_UINT &&
+        dst_image.info.format == PixelFormat::A8B8G8R8_UNORM) {
+        util_shaders.ConvertS8D24(dst_image, copies);
+    }
 }
 
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_texture_cache.h

@@ -55,13 +55,14 @@ struct FormatProperties {
 
 class FormatConversionPass {
 public:
-    FormatConversionPass() = default;
+    explicit FormatConversionPass(UtilShaders& util_shaders);
     ~FormatConversionPass() = default;
 
     void ConvertImage(Image& dst_image, Image& src_image,
                       std::span<const VideoCommon::ImageCopy> copies);
 
 private:
+    UtilShaders& util_shaders;
     OGLBuffer intermediate_pbo;
     size_t pbo_size{};
 };

src/video_core/renderer_opengl/util_shaders.cpp

@@ -13,6 +13,7 @@
 #include "video_core/host_shaders/astc_decoder_comp.h"
 #include "video_core/host_shaders/block_linear_unswizzle_2d_comp.h"
 #include "video_core/host_shaders/block_linear_unswizzle_3d_comp.h"
+#include "video_core/host_shaders/opengl_convert_s8d24_comp.h"
 #include "video_core/host_shaders/opengl_copy_bc4_comp.h"
 #include "video_core/host_shaders/pitch_unswizzle_comp.h"
 #include "video_core/renderer_opengl/gl_shader_manager.h"
@@ -50,7 +51,8 @@ UtilShaders::UtilShaders(ProgramManager& program_manager_)
       block_linear_unswizzle_2d_program(MakeProgram(BLOCK_LINEAR_UNSWIZZLE_2D_COMP)),
       block_linear_unswizzle_3d_program(MakeProgram(BLOCK_LINEAR_UNSWIZZLE_3D_COMP)),
       pitch_unswizzle_program(MakeProgram(PITCH_UNSWIZZLE_COMP)),
-      copy_bc4_program(MakeProgram(OPENGL_COPY_BC4_COMP)) {
+      copy_bc4_program(MakeProgram(OPENGL_COPY_BC4_COMP)),
+      convert_s8d24_program(MakeProgram(OPENGL_CONVERT_S8D24_COMP)) {
     const auto swizzle_table = Tegra::Texture::MakeSwizzleTable();
     swizzle_table_buffer.Create();
     glNamedBufferStorage(swizzle_table_buffer.handle, sizeof(swizzle_table), &swizzle_table, 0);
@@ -248,6 +250,26 @@ void UtilShaders::CopyBC4(Image& dst_image, Image& src_image, std::span<const Im
     program_manager.RestoreGuestCompute();
 }
 
+void UtilShaders::ConvertS8D24(Image& dst_image, std::span<const ImageCopy> copies) {
+    static constexpr GLuint BINDING_DESTINATION = 0;
+    static constexpr GLuint LOC_SIZE = 0;
+
+    program_manager.BindComputeProgram(convert_s8d24_program.handle);
+    for (const ImageCopy& copy : copies) {
+        ASSERT(copy.src_subresource.base_layer == 0);
+        ASSERT(copy.src_subresource.num_layers == 1);
+        ASSERT(copy.dst_subresource.base_layer == 0);
+        ASSERT(copy.dst_subresource.num_layers == 1);
+
+        glUniform3ui(LOC_SIZE, copy.extent.width, copy.extent.height, copy.extent.depth);
+        glBindImageTexture(BINDING_DESTINATION, dst_image.StorageHandle(),
+                           copy.dst_subresource.base_level, GL_TRUE, 0, GL_READ_WRITE, GL_RGBA8UI);
+        glDispatchCompute(Common::DivCeil(copy.extent.width, 16u),
+                          Common::DivCeil(copy.extent.height, 8u), copy.extent.depth);
+    }
+    program_manager.RestoreGuestCompute();
+}
+
 GLenum StoreFormat(u32 bytes_per_block) {
     switch (bytes_per_block) {
     case 1:

src/video_core/renderer_opengl/util_shaders.h

@@ -39,6 +39,8 @@ public:
     void CopyBC4(Image& dst_image, Image& src_image,
                  std::span<const VideoCommon::ImageCopy> copies);
 
+    void ConvertS8D24(Image& dst_image, std::span<const VideoCommon::ImageCopy> copies);
+
 private:
     ProgramManager& program_manager;
 
@@ -49,6 +51,7 @@ private:
     OGLProgram block_linear_unswizzle_3d_program;
     OGLProgram pitch_unswizzle_program;
     OGLProgram copy_bc4_program;
+    OGLProgram convert_s8d24_program;
 };
 
 GLenum StoreFormat(u32 bytes_per_block);

src/yuzu/main.cpp

@@ -1540,8 +1540,9 @@ void GMainWindow::ShutdownGame() {
     input_subsystem->GetTas()->Stop();
     OnTasStateChanged();
 
-    // Enable all controllers
+    // Enable all controllers types and remove all service callbacks
     system->HIDCore().SetSupportedStyleTag({Core::HID::NpadStyleSet::All});
+    system->HIDCore().RemoveServiceCallbacks();
 
     render_window->removeEventFilter(render_window);
     render_window->setAttribute(Qt::WA_Hover, false);