build: use system SoundTouch when available

Checking for both a CMake config file and a pkg-config file is needed
because some distros only ship the latter.

Related to #6833 and #7044

externals/CMakeLists.txt

@@ -69,7 +69,12 @@ if (YUZU_USE_EXTERNAL_SDL2)
 endif()
 
 # SoundTouch
-add_subdirectory(soundtouch)
+find_package(SoundTouch MODULE)
+if (NOT SoundTouch_FOUND)
+    message(STATUS "SoundTouch not found, falling back to externals")
+    add_subdirectory(soundtouch)
+    add_library(SoundTouch::SoundTouch ALIAS SoundTouch)
+endif()
 
 # Cubeb
 if(ENABLE_CUBEB)

externals/find-modules/FindSoundTouch.cmake

@@ -0,0 +1,63 @@
+# SPDX-FileCopyrightText: 2022 Andrea Pappacoda <andrea@pappacoda.it>
+#
+# SPDX-License-Identifier: GPL-2.0-or-later
+
+# This find module looks for SoundTouch with both a CMake Config file and a
+# pkg-config. If the library is found, it checks if it enforces 32 bit
+# floating point samples, and if not it defines SOUNDTOUCH_INTEGER_SAMPLES,
+# just like in the bundled library.
+
+find_package(SoundTouch CONFIG)
+if (SoundTouch_FOUND)
+    set(_st_real_name "SoundTouch::SoundTouch")
+else()
+    find_package(PkgConfig QUIET)
+    if (PKG_CONFIG_FOUND)
+        pkg_search_module(SoundTouch IMPORTED_TARGET soundtouch)
+        if (SoundTouch_FOUND)
+            set_target_properties(PkgConfig::SoundTouch PROPERTIES IMPORTED_GLOBAL True)
+            add_library(SoundTouch::SoundTouch ALIAS PkgConfig::SoundTouch)
+            # Need to set this variable because CMake doesn't allow to add
+            # compile definitions to ALIAS targets
+            set(_st_real_name "PkgConfig::SoundTouch")
+        endif()
+    endif()
+endif()
+
+if (SoundTouch_FOUND)
+    find_path(_st_include_dir "soundtouch/soundtouch_config.h")
+    file(READ "${_st_include_dir}/soundtouch/soundtouch_config.h" _st_config_file)
+
+    # Check if the config file defines SOUNDTOUCH_FLOAT_SAMPLES
+    string(REGEX MATCH "#define[ ]+SOUNDTOUCH_FLOAT_SAMPLES[ ]+1" SoundTouch_FLOAT_SAMPLES ${_st_config_file})
+
+    if (NOT SoundTouch_FLOAT_SAMPLES)
+        target_compile_definitions(${_st_real_name} INTERFACE "SOUNDTOUCH_INTEGER_SAMPLES=1")
+        set(SoundTouch_INTEGER_SAMPLES True)
+    else()
+        # Check if SoundTouch supports SOUNDTOUCH_NO_CONFIG
+        file(READ "${_st_include_dir}/soundtouch/STTypes.h" _st_types_file)
+
+        string(FIND "${_st_types_file}" "SOUNDTOUCH_NO_CONFIG" SoundTouch_NO_CONFIG)
+
+        # if found
+        if (NOT SoundTouch_NO_CONFIG EQUAL "-1")
+            target_compile_definitions(${_st_real_name} INTERFACE "SOUNDTOUCH_NO_CONFIG" "SOUNDTOUCH_INTEGER_SAMPLES=1")
+            set(SoundTouch_INTEGER_SAMPLES True)
+        endif()
+
+        unset(_st_types_file)
+    endif()
+
+    unset(_st_real_name)
+    unset(_st_include_dir)
+    unset(_st_config_file)
+endif()
+
+include(FindPackageHandleStandardArgs)
+find_package_handle_standard_args(SoundTouch
+    REQUIRED_VARS
+        SoundTouch_FOUND
+        SoundTouch_INTEGER_SAMPLES
+    VERSION_VAR SoundTouch_VERSION
+)

src/audio_core/CMakeLists.txt

@@ -63,7 +63,7 @@ if (NOT MSVC)
 endif()
 
 target_link_libraries(audio_core PUBLIC common core)
-target_link_libraries(audio_core PRIVATE SoundTouch)
+target_link_libraries(audio_core PRIVATE SoundTouch::SoundTouch)
 
 if(ENABLE_CUBEB)
     target_link_libraries(audio_core PRIVATE cubeb)

src/core/memory.cpp

@@ -322,7 +322,7 @@ struct Memory::Impl {
         }
 
         if (Settings::IsFastmemEnabled()) {
-            const bool is_read_enable = Settings::IsGPULevelHigh() || !cached;
+            const bool is_read_enable = !Settings::IsGPULevelExtreme() || !cached;
             system.DeviceMemory().buffer.Protect(vaddr, size, is_read_enable, !cached);
         }
 

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -352,7 +352,7 @@ void RasterizerOpenGL::OnCPUWrite(VAddr addr, u64 size) {
     shader_cache.OnCPUWrite(addr, size);
     {
         std::scoped_lock lock{texture_cache.mutex};
-        texture_cache.WriteMemory(addr, size);
+        texture_cache.CachedWriteMemory(addr, size);
     }
     {
         std::scoped_lock lock{buffer_cache.mutex};
@@ -363,6 +363,10 @@ void RasterizerOpenGL::OnCPUWrite(VAddr addr, u64 size) {
 void RasterizerOpenGL::SyncGuestHost() {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
     shader_cache.SyncGuestHost();
+    {
+        std::scoped_lock lock{texture_cache.mutex};
+        texture_cache.FlushCachedWrites();
+    }
     {
         std::scoped_lock lock{buffer_cache.mutex};
         buffer_cache.FlushCachedWrites();

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -408,7 +408,7 @@ void RasterizerVulkan::OnCPUWrite(VAddr addr, u64 size) {
     pipeline_cache.OnCPUWrite(addr, size);
     {
         std::scoped_lock lock{texture_cache.mutex};
-        texture_cache.WriteMemory(addr, size);
+        texture_cache.CachedWriteMemory(addr, size);
     }
     {
         std::scoped_lock lock{buffer_cache.mutex};
@@ -418,6 +418,10 @@ void RasterizerVulkan::OnCPUWrite(VAddr addr, u64 size) {
 
 void RasterizerVulkan::SyncGuestHost() {
     pipeline_cache.SyncGuestHost();
+    {
+        std::scoped_lock lock{texture_cache.mutex};
+        texture_cache.FlushCachedWrites();
+    }
     {
         std::scoped_lock lock{buffer_cache.mutex};
         buffer_cache.FlushCachedWrites();

src/video_core/texture_cache/image_base.h

@@ -39,6 +39,9 @@ enum class ImageFlagBits : u32 {
     Rescaled = 1 << 13,
     CheckingRescalable = 1 << 14,
     IsRescalable = 1 << 15,
+
+    // Cached CPU
+    CachedCpuModified = 1 << 16, ///< Contents have been modified from the CPU
 };
 DECLARE_ENUM_FLAG_OPERATORS(ImageFlagBits)
 

src/video_core/texture_cache/texture_cache.h

@@ -437,6 +437,23 @@ void TextureCache<P>::WriteMemory(VAddr cpu_addr, size_t size) {
     });
 }
 
+template <class P>
+void TextureCache<P>::CachedWriteMemory(VAddr cpu_addr, size_t size) {
+    const VAddr new_cpu_addr = Common::AlignDown(cpu_addr, CPU_PAGE_SIZE);
+    const size_t new_size = Common::AlignUp(size + cpu_addr - new_cpu_addr, CPU_PAGE_SIZE);
+    ForEachImageInRegion(new_cpu_addr, new_size, [this](ImageId image_id, Image& image) {
+        if (True(image.flags & ImageFlagBits::CachedCpuModified)) {
+            return;
+        }
+        image.flags |= ImageFlagBits::CachedCpuModified;
+        cached_cpu_invalidate.insert(image_id);
+
+        if (True(image.flags & ImageFlagBits::Tracked)) {
+            UntrackImage(image, image_id);
+        }
+    });
+}
+
 template <class P>
 void TextureCache<P>::DownloadMemory(VAddr cpu_addr, size_t size) {
     std::vector<ImageId> images;
@@ -494,6 +511,18 @@ void TextureCache<P>::UnmapGPUMemory(GPUVAddr gpu_addr, size_t size) {
     }
 }
 
+template <class P>
+void TextureCache<P>::FlushCachedWrites() {
+    for (ImageId image_id : cached_cpu_invalidate) {
+        Image& image = slot_images[image_id];
+        if (True(image.flags & ImageFlagBits::CachedCpuModified)) {
+            image.flags &= ~ImageFlagBits::CachedCpuModified;
+            image.flags |= ImageFlagBits::CpuModified;
+        }
+    }
+    cached_cpu_invalidate.clear();
+}
+
 template <class P>
 void TextureCache<P>::BlitImage(const Tegra::Engines::Fermi2D::Surface& dst,
                                 const Tegra::Engines::Fermi2D::Surface& src,
@@ -1560,6 +1589,9 @@ void TextureCache<P>::UnregisterImage(ImageId image_id) {
 template <class P>
 void TextureCache<P>::TrackImage(ImageBase& image, ImageId image_id) {
     ASSERT(False(image.flags & ImageFlagBits::Tracked));
+    if (True(image.flags & ImageFlagBits::CachedCpuModified)) {
+        return;
+    }
     image.flags |= ImageFlagBits::Tracked;
     if (False(image.flags & ImageFlagBits::Sparse)) {
         rasterizer.UpdatePagesCachedCount(image.cpu_addr, image.guest_size_bytes, 1);
@@ -1616,6 +1648,9 @@ void TextureCache<P>::DeleteImage(ImageId image_id, bool immediate_delete) {
         tentative_size = EstimatedDecompressedSize(tentative_size, image.info.format);
     }
     total_used_memory -= Common::AlignUp(tentative_size, 1024);
+    if (True(image.flags & ImageFlagBits::CachedCpuModified)) {
+        cached_cpu_invalidate.erase(image_id);
+    }
     const GPUVAddr gpu_addr = image.gpu_addr;
     const auto alloc_it = image_allocs_table.find(gpu_addr);
     if (alloc_it == image_allocs_table.end()) {
@@ -1782,7 +1817,11 @@ template <class P>
 void TextureCache<P>::PrepareImage(ImageId image_id, bool is_modification, bool invalidate) {
     Image& image = slot_images[image_id];
     if (invalidate) {
-        image.flags &= ~(ImageFlagBits::CpuModified | ImageFlagBits::GpuModified);
+        if (True(image.flags & ImageFlagBits::CachedCpuModified)) {
+            cached_cpu_invalidate.erase(image_id);
+        }
+        image.flags &= ~(ImageFlagBits::CpuModified | ImageFlagBits::GpuModified |
+                         ImageFlagBits::CachedCpuModified);
         if (False(image.flags & ImageFlagBits::Tracked)) {
             TrackImage(image, image_id);
         }

src/video_core/texture_cache/texture_cache_base.h

@@ -8,6 +8,7 @@
 #include <span>
 #include <type_traits>
 #include <unordered_map>
+#include <unordered_set>
 #include <vector>
 #include <queue>
 
@@ -50,6 +51,9 @@ class TextureCache {
     /// Address shift for caching images into a hash table
     static constexpr u64 PAGE_BITS = 20;
 
+    static constexpr u64 CPU_PAGE_BITS = 12;
+    static constexpr u64 CPU_PAGE_SIZE = 1ULL << CPU_PAGE_BITS;
+
     /// Enables debugging features to the texture cache
     static constexpr bool ENABLE_VALIDATION = P::ENABLE_VALIDATION;
     /// Implement blits as copies between framebuffers
@@ -136,6 +140,9 @@ public:
     /// Mark images in a range as modified from the CPU
     void WriteMemory(VAddr cpu_addr, size_t size);
 
+    /// Mark images in a range as modified from the CPU
+    void CachedWriteMemory(VAddr cpu_addr, size_t size);
+
     /// Download contents of host images to guest memory in a region
     void DownloadMemory(VAddr cpu_addr, size_t size);
 
@@ -145,6 +152,8 @@ public:
     /// Remove images in a region
     void UnmapGPUMemory(GPUVAddr gpu_addr, size_t size);
 
+    void FlushCachedWrites();
+
     /// Blit an image with the given parameters
     void BlitImage(const Tegra::Engines::Fermi2D::Surface& dst,
                    const Tegra::Engines::Fermi2D::Surface& src,
@@ -366,6 +375,8 @@ private:
 
     std::unordered_map<ImageId, std::vector<ImageViewId>> sparse_views;
 
+    std::unordered_set<ImageId> cached_cpu_invalidate;
+
     VAddr virtual_invalid_space{};
 
     bool has_deleted_images = false;