Compare commits

..

4 Commits

Author SHA1 Message Date
lizzie b3eb7e47fd s^2 2026-09-02 21:22:23 +00:00
lizzie 6a7c069824 fix linking issues 2026-09-02 11:37:51 +00:00
lizzie 86df49bbc9 remove latency func 2026-09-02 10:58:33 +00:00
lizzie 4e59b71c59 [audio] Nuke cubeb and use SDL3 exclusively
Signed-off-by: lizzie <lizzie@eden-emu.dev>
2026-09-02 10:55:54 +00:00
80 changed files with 533 additions and 2514 deletions
@@ -1,112 +0,0 @@
diff --git a/include/vk_mem_alloc.h b/include/vk_mem_alloc.h
index 8df0364..4856064 100644
--- a/include/vk_mem_alloc.h
+++ b/include/vk_mem_alloc.h
@@ -3017,7 +3017,7 @@ remove them if not needed.
#if defined(__ANDROID_API__) && (__ANDROID_API__ < 16)
#include <cstdlib>
-static void* vma_aligned_alloc(size_t alignment, size_t size)
+static inline void* vma_aligned_alloc(size_t alignment, size_t size)
{
// alignment must be >= sizeof(void*)
if(alignment < sizeof(void*))
@@ -3860,7 +3860,7 @@ Returned value is the found element, if present in the collection or place where
new element with value (key) should be inserted.
*/
template <typename CmpLess, typename IterT, typename KeyT>
-static IterT VmaBinaryFindFirstNotLess(IterT beg, IterT end, const KeyT& key, const CmpLess& cmp)
+static inline IterT VmaBinaryFindFirstNotLess(IterT beg, IterT end, const KeyT& key, const CmpLess& cmp)
{
size_t down = 0;
size_t up = size_t(end - beg);
@@ -3898,7 +3898,7 @@ Warning! O(n^2) complexity. Use only inside VMA_HEAVY_ASSERT.
T must be pointer type, e.g. VmaAllocation, VmaPool.
*/
template<typename T>
-static bool VmaValidatePointerArray(uint32_t count, const T* arr)
+static inline bool VmaValidatePointerArray(uint32_t count, const T* arr)
{
for (uint32_t i = 0; i < count; ++i)
{
@@ -4188,13 +4188,13 @@ static void VmaFree(const VkAllocationCallbacks* pAllocationCallbacks, void* ptr
}
template<typename T>
-static T* VmaAllocate(const VkAllocationCallbacks* pAllocationCallbacks)
+static inline T* VmaAllocate(const VkAllocationCallbacks* pAllocationCallbacks)
{
return (T*)VmaMalloc(pAllocationCallbacks, sizeof(T), VMA_ALIGN_OF(T));
}
template<typename T>
-static T* VmaAllocateArray(const VkAllocationCallbacks* pAllocationCallbacks, size_t count)
+static inline T* VmaAllocateArray(const VkAllocationCallbacks* pAllocationCallbacks, size_t count)
{
return (T*)VmaMalloc(pAllocationCallbacks, sizeof(T) * count, VMA_ALIGN_OF(T));
}
@@ -4204,14 +4204,14 @@ static T* VmaAllocateArray(const VkAllocationCallbacks* pAllocationCallbacks, si
#define vma_new_array(allocator, type, count) new(VmaAllocateArray<type>((allocator), (count)))(type)
template<typename T>
-static void vma_delete(const VkAllocationCallbacks* pAllocationCallbacks, T* ptr)
+static inline void vma_delete(const VkAllocationCallbacks* pAllocationCallbacks, T* ptr)
{
ptr->~T();
VmaFree(pAllocationCallbacks, ptr);
}
template<typename T>
-static void vma_delete_array(const VkAllocationCallbacks* pAllocationCallbacks, T* ptr, size_t count)
+static inline void vma_delete_array(const VkAllocationCallbacks* pAllocationCallbacks, T* ptr, size_t count)
{
if (ptr != VMA_NULL)
{
@@ -4658,13 +4658,13 @@ void VmaVector<T, AllocatorT>::remove(size_t index)
#endif // _VMA_VECTOR_FUNCTIONS
template<typename T, typename allocatorT>
-static void VmaVectorInsert(VmaVector<T, allocatorT>& vec, size_t index, const T& item)
+static inline void VmaVectorInsert(VmaVector<T, allocatorT>& vec, size_t index, const T& item)
{
vec.insert(index, item);
}
template<typename T, typename allocatorT>
-static void VmaVectorRemove(VmaVector<T, allocatorT>& vec, size_t index)
+static inline void VmaVectorRemove(VmaVector<T, allocatorT>& vec, size_t index)
{
vec.remove(index);
}
@@ -10620,19 +10620,19 @@ static void VmaFree(VmaAllocator hAllocator, void* ptr)
}
template<typename T>
-static T* VmaAllocate(VmaAllocator hAllocator)
+static inline T* VmaAllocate(VmaAllocator hAllocator)
{
return (T*)VmaMalloc(hAllocator, sizeof(T), VMA_ALIGN_OF(T));
}
template<typename T>
-static T* VmaAllocateArray(VmaAllocator hAllocator, size_t count)
+static inline T* VmaAllocateArray(VmaAllocator hAllocator, size_t count)
{
return (T*)VmaMalloc(hAllocator, sizeof(T) * count, VMA_ALIGN_OF(T));
}
template<typename T>
-static void vma_delete(VmaAllocator hAllocator, T* ptr)
+static inline void vma_delete(VmaAllocator hAllocator, T* ptr)
{
if(ptr != VMA_NULL)
{
@@ -10642,7 +10642,7 @@ static void vma_delete(VmaAllocator hAllocator, T* ptr)
}
template<typename T>
-static void vma_delete_array(VmaAllocator hAllocator, T* ptr, size_t count)
+static inline void vma_delete_array(VmaAllocator hAllocator, T* ptr, size_t count)
{
if(ptr != VMA_NULL)
{
+1 -9
View File
@@ -3,7 +3,6 @@
cmake_minimum_required(VERSION 3.31)
set(CMAKE_OSX_DEPLOYMENT_TARGET "15.0" CACHE STRING "macOS deployment target")
project(yuzu)
list(APPEND CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/CMakeModules")
@@ -53,7 +52,6 @@ if (YUZU_STATIC_ROOM)
# disable e v e r y t h i n g
set(ENABLE_QT OFF)
set(YUZU_CMD OFF)
set(ENABLE_CUBEB OFF)
set(ENABLE_UPDATE_CHECKER OFF)
set(USE_DISCORD_PRESENCE OFF)
set(BUILD_TESTING OFF)
@@ -215,8 +213,6 @@ if(MSVC)
endif()
# TODO(crueter): Cleanup, each dep that has a bundled option should allow to choose between bundled, external, system
option(ENABLE_CUBEB "Enables the cubeb audio backend" ON)
set(EXT_DEFAULT OFF)
if (MSVC OR ANDROID)
set(EXT_DEFAULT ON)
@@ -513,7 +509,7 @@ endfunction()
# =============================================
if (APPLE)
foreach(fw Carbon Metal Cocoa IOKit CoreVideo CoreMedia Security UniformTypeIdentifiers Foundation)
foreach(fw Carbon Metal Cocoa IOKit CoreVideo CoreMedia Security UniformTypeIdentifiers)
find_library(${fw}_LIBRARY ${fw} REQUIRED)
list(APPEND PLATFORM_LIBRARIES ${${fw}_LIBRARY})
endforeach()
@@ -575,10 +571,6 @@ if (NOT YUZU_STATIC_ROOM)
find_package(DiscordRPC)
endif()
if (ENABLE_CUBEB)
find_package(cubeb)
endif()
if (YUZU_TESTS OR DYNARMIC_TESTS)
find_package(Catch2)
endif()
+10 -17
View File
@@ -1,4 +1,12 @@
{
"": {
"ci": true,
"hash": "9f50d993c39529e022ad456163de91ac5934e16faf8fc348f305355fc0a643c534f87ded707e11bd03bcbc0a1760bd20852b6533a67ac0558a078385181cb184",
"name": "SDL3",
"package": "SDL3",
"repo": "crueter-ci/SDL3",
"version": "3.4.14-1788231389-147a8ee32d"
},
"biscuit": {
"hash": "1229f345b014f7ca544dedb4edb3311e41ba736f9aa9a67f88b5f26f3c983288c6bb6cdedcfb0b8a02c63088a37e6a0d7ba97d9c2a4d721b213916327cffe28a",
"min_version": "0.9.1",
@@ -46,24 +54,12 @@
"repo": "arun11299/cpp-jwt",
"version": "7f24eb4c32"
},
"cubeb": {
"find_args": "CONFIG",
"hash": "8a4bcb2f83ba590f52c66626e895304a73eb61928dbc57777e1822e55378e3568366f17f9da4b80036cc2ef4ea9723c32abf6e7d9bbe00fb03654f0991596ab0",
"options": [
"USE_SANITIZERS OFF",
"BUILD_TESTS OFF",
"BUILD_TOOLS OFF",
"BUNDLE_SPEEX ON"
],
"repo": "mozilla/cubeb",
"version": "fa02160712"
},
"discord-rpc": {
"find_args": "MODULE",
"hash": "8d680b3a16d6f6bf292ad823cf8635595ff986f2a49f758e3009f505b540a9d75194a8cf44cddea75736a8c3e3b5160166e716a0939ce3f18cc463cf648753fe",
"hash": "8213c43dcb0f7d479f5861091d111ed12fbdec1e62e6d729d65a4bc181d82f48a35d5fd3cd5c291f2393ac7c9681eabc6b76609755f55376284c8a8d67e148f3",
"package": "DiscordRPC",
"repo": "eden-emulator/discord-rpc",
"version": "76616d8675"
"version": "0d8b2d6a37"
},
"enet": {
"find_args": "MODULE",
@@ -303,9 +299,6 @@
"find_args": "CONFIG",
"hash": "deb5902ef8db0e329fbd5f3f4385eb0e26bdd9f14f3a2334823fb3fe18f36bc5d235d620d6e5f6fe3551ec3ea7038638899db8778c09f6d5c278f5ff95c3344b",
"package": "VulkanMemoryAllocator",
"patches": [
"0001-macos-clang.patch"
],
"repo": "GPUOpen-LibrariesAndSDKs/VulkanMemoryAllocator",
"version": "v3.3.0"
},
-2
View File
@@ -100,8 +100,6 @@ For reasons unberknownst to any human being, `glslangValidator` will crash upon
For this reason this patch is NOT applied to default on all platforms (for obvious reasons) - instead this is a HaikuOS specific patch, apply with `git apply <absolute path to patch>` after cloning SPIRV-Tools then `make -C build` and add the resulting binary (in `build/StandAlone/glslang`) into PATH.
`cubeb_devel` will also not work, either disable cubeb or uninstall it.
Still will not run flawlessly until `mesa-24` is available. Modify CMakeCache.txt with the `.so` of libGL and libGLESv2 by doing the incredibly difficult task of copy pasting them (`cp /boot/system/lib/libGL.so .`)
If you have `quazip1_qt6_devel`, uninstall it. It may call `Core5Compat` on CMake which is wrongly packaged.
+2 -3
View File
@@ -73,7 +73,6 @@ Certain other dependencies will be fetched by CPM regardless. System packages *c
* [SimpleIni](https://github.com/brofield/simpleini)
* [DiscordRPC](https://github.com/eden-emulator/discord-rpc)
* [cubeb](https://github.com/mozilla/cubeb)
* [libusb](https://github.com/libusb/libusb)
* [VulkanMemoryAllocator](https://github.com/GPUOpen-LibrariesAndSDKs/VulkanMemoryAllocator)
* [sirit](https://github.com/eden-emulator/sirit)
@@ -122,7 +121,7 @@ sudo emerge -a \
dev-util/spirv-tools dev-util/spirv-headers dev-util/vulkan-headers \
dev-util/vulkan-utility-libraries dev-util/glslang \
media-gfx/renderdoc media-libs/libva media-libs/opus media-video/ffmpeg \
media-libs/VulkanMemoryAllocator media-libs/libsdl3 media-libs/cubeb \
media-libs/VulkanMemoryAllocator media-libs/libsdl3 \
net-libs/enet \
sys-libs/zlib \
dev-cpp/nlohmann_json dev-cpp/simpleini dev-cpp/cpp-httplib dev-cpp/cpp-jwt \
@@ -356,7 +355,7 @@ pacman -Syuu --needed --noconfirm $packages
<summary>HaikuOS</summary>
```sh
pkgman install git cmake patch libfmt_devel nlohmann_json lz4_devel opus_devel boost1.90_devel vulkan_devel qt6_base_devel qt6_declarative_devel libsdl3_devel ffmpeg7_devel libx11_devel enet_devel catch2_devel quazip1_qt5_devel qt6_5compat_devel glslang qt6_devel qt6_charts_devel cubeb_devel simpleini quazip_qt6_devel
pkgman install git cmake patch libfmt_devel nlohmann_json lz4_devel opus_devel boost1.90_devel vulkan_devel qt6_base_devel qt6_declarative_devel libsdl3_devel ffmpeg7_devel libx11_devel enet_devel catch2_devel quazip1_qt5_devel qt6_5compat_devel glslang qt6_devel qt6_charts_devel simpleini quazip_qt6_devel
```
[Caveats](./Caveats.md#haikuos).
+1 -1
View File
@@ -4,7 +4,7 @@ Modern game consoles require heavy power to be emulated appropriately. This is w
For example, take a disk write, instead of emulating a proper SD card we instead use the C++ standard library for I/O. Additionally we use the abstractions provided by the `fs` service to "lie" to programs about certain SD card properties. Notably this includes making up sizes for the fake SD card, giving "realistic" values or expected outputs for a given card, and so on. And instead of writing to an actual SD card, the emulator simply writes to a file.
This also means grand part of the emulator consists of just re-implementing firmware but using HLE primitives; for example audio doesn't go to an emulated audio device, but rather gets processed on the fly by a dedicated service and then passed to SDL3/cubeb/etc.
This also means grand part of the emulator consists of just re-implementing firmware but using HLE primitives; for example audio doesn't go to an emulated audio device, but rather gets processed on the fly by a dedicated service and then passed to SDL3 or null backend.
As such, many of the systems implemented are not 100% accurate to the original software, but they're "good enough" to pass as being so. While we do strive to maintain high compatibility (especially with homebrew), there are realistic limitations to these approaches.
-2
View File
@@ -35,8 +35,6 @@ These options control dependencies.
- `ENABLE_WEB_SERVICE` (ON) Enable multiplayer service
- `ENABLE_WIFI_SCAN` (OFF) Enable WiFi scanning (requires iw on Linux) - experimental
- `ENABLE_CUBEB` (ON) Enables the cubeb audio backend
- This option is subject for removal.
- `YUZU_TESTS` (ON) Compile tests - requires Catch2
- `ENABLE_LTO` (OFF) Enable link-time optimization
- Not recommended on Windows
-30
View File
@@ -106,36 +106,6 @@ endif()
# SimpleIni
AddJsonPackage(simpleini)
# Most linux distros don't package cubeb, so enable regardless of cpm settings
if(ENABLE_CUBEB)
AddJsonPackage(cubeb)
if (cubeb_ADDED)
if (NOT MSVC)
if (TARGET speex)
target_compile_options(speex PRIVATE $<$<COMPILE_LANGUAGE:C,CXX>:-Wno-sign-compare>)
endif()
set_target_properties(cubeb PROPERTIES COMPILE_OPTIONS "")
target_compile_options(cubeb INTERFACE
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-implicit-const-int-float-conversion>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-shadow>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-missing-declarations>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-return-type>
$<$<COMPILE_LANGUAGE:C,CXX>:-Wno-uninitialized>
)
else()
target_compile_options(cubeb PRIVATE
/wd4456
/wd4458
)
endif()
endif()
if (NOT TARGET cubeb::cubeb)
add_library(cubeb::cubeb ALIAS cubeb)
endif()
endif()
if (NOT YUZU_USE_BUNDLED_SDL3)
if (NOT WIN32)
# Yuzu itself needs: Atomic Audio Events Joystick Haptic Sensor Threads Timers
-1
View File
@@ -11,7 +11,6 @@
#include <limits>
#include <span>
#include <array>
#include <algorithm>
#include <time.h>
namespace Tz {
+1 -1
View File
@@ -15,7 +15,7 @@ pkgs.mkShellNoCC {
enet libopus vulkan-headers vulkan-utility-libraries
spirv-tools spirv-headers vulkan-loader unzip
glslang python3 httplib cpp-jwt ffmpeg-headless
libusb1 cubeb
libusb1
# eden
qt6.qtbase qt6.qtmultimedia qt6.qtwayland qt6.qttools
qt6.qtwebengine qt6.qt5compat
@@ -30,7 +30,6 @@ enum class BooleanSetting(override val key: String) : AbstractBooleanSetting {
RENDERER_REACTIVE_FLUSHING("use_reactive_flushing"),
ENABLE_BUFFER_HISTORY("enable_buffer_history"),
USE_OPTIMIZED_VERTEX_BUFFERS("use_optimized_vertex_buffers"),
ENABLE_SHADER_PHI_TRACKING("enable_shader_phi_tracking"),
ENABLE_GPU_BUFFER_READBACK("enable_gpu_buffer_readback"),
SYNC_MEMORY_OPERATIONS("sync_memory_operations"),
BUFFER_REORDER_DISABLE("disable_buffer_reorder"),
@@ -927,13 +927,6 @@ abstract class SettingsItem(
descriptionId = R.string.use_optimized_vertex_buffers_description
)
)
put(
SwitchSetting(
BooleanSetting.ENABLE_SHADER_PHI_TRACKING,
titleId = R.string.enable_shader_phi_tracking,
descriptionId = R.string.enable_shader_phi_tracking_description
)
)
put(
SwitchSetting(
BooleanSetting.SYNC_MEMORY_OPERATIONS,
@@ -338,7 +338,6 @@ class SettingsFragmentPresenter(
add(BooleanSetting.ENABLE_BUFFER_HISTORY.key)
add(BooleanSetting.ENABLE_GPU_BUFFER_READBACK.key)
add(BooleanSetting.USE_OPTIMIZED_VERTEX_BUFFERS.key)
add(BooleanSetting.ENABLE_SHADER_PHI_TRACKING.key)
add(HeaderSetting(R.string.hacks))
@@ -58,13 +58,6 @@ class LicensesFragment : Fragment() {
R.string.license_fidelityfx_fsr_copyright,
R.string.license_fidelityfx_fsr_text
),
License(
R.string.license_cubeb,
R.string.license_cubeb_description,
R.string.license_cubeb_link,
R.string.license_cubeb_copyright,
R.string.license_cubeb_text
),
License(
R.string.license_dynarmic,
R.string.license_dynarmic_description,
@@ -799,9 +799,6 @@
<string name="theme_mode_light">روشن</string>
<string name="theme_mode_dark">تاریک</string>
<!-- Audio output engines -->
<string name="cubeb">cubeb</string>
<!-- Anisotropic filtering options -->
<string name="multiplier_x2">x2</string>
<string name="multiplier_x4">x4</string>
@@ -466,13 +466,11 @@
<string-array name="outputEngineEntries">
<item>@string/auto</item>
<item>@string/sdl3</item>
<item>@string/cubeb</item>
<item>@string/string_null</item>
</string-array>
<integer-array name="outputEngineValues">
<item>0</item>
<item>2</item>
<item>1</item>
<item>3</item>
</integer-array>
@@ -570,8 +570,6 @@
<string name="enable_gpu_buffer_readback_description">Preserves GPU-modified buffer data by reading it back before uploads. Some games require this to render certain effects properly. May cause issues if the hardware cannot handle the additional workload.</string>
<string name="use_optimized_vertex_buffers">Optimized Vertex Buffers</string>
<string name="use_optimized_vertex_buffers_description">Enables optimized vertex buffer binding for improved performance. Requires Mesa 26.0+ Turnip drivers/ QCOM drivers. Will crash on older Turnip drivers (25.3 and below).</string>
<string name="enable_shader_phi_tracking">Shader Phi Tracking</string>
<string name="enable_shader_phi_tracking_description">toggle for test on shader phi tracking.</string>
<string name="hacks">Hacks</string>
@@ -1243,7 +1241,6 @@
<!-- Audio output engines -->
<string name="sdl3" translatable="false">SDL3</string>
<string name="cubeb" translatable="false">cubeb</string>
<!-- Anisotropic filtering options -->
<string name="multiplier_x2" translatable="false">x2</string>
@@ -1383,23 +1380,6 @@ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
THE SOFTWARE.
</string>
<string name="license_cubeb" translatable="false">cubeb</string>
<string name="license_cubeb_description" translatable="false">Cross platform audio library</string>
<string name="license_cubeb_link" translatable="false">https://github.com/mozilla/cubeb</string>
<string name="license_cubeb_copyright" translatable="false">Copyright © 2011 Mozilla Foundation</string>
<string name="license_cubeb_text" translatable="false">
Permission to use, copy, modify, and distribute this software for any
purpose with or without fee is hereby granted, provided that the above
copyright notice and this permission notice appear in all copies.\n\n
THE SOFTWARE IS PROVIDED \"AS IS\" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
</string>
<string name="license_dynarmic" translatable="false">Dynarmic</string>
<string name="license_dynarmic_description" translatable="false">An ARM dynamic recompiler</string>
+4 -9
View File
@@ -229,15 +229,6 @@ endif()
target_include_directories(audio_core PRIVATE ${OPUS_INCLUDE_DIRS})
target_link_libraries(audio_core PUBLIC common core Opus::opus)
if (ENABLE_CUBEB)
target_sources(audio_core PRIVATE
sink/cubeb_sink.cpp
sink/cubeb_sink.h)
target_link_libraries(audio_core PRIVATE cubeb::cubeb)
target_compile_definitions(audio_core PRIVATE HAVE_CUBEB=1)
endif()
target_sources(audio_core PRIVATE
sink/sdl3_sink.cpp
sink/sdl3_sink.h)
@@ -245,4 +236,8 @@ target_sources(audio_core PRIVATE
target_link_libraries(audio_core PRIVATE SDL3::SDL3)
target_compile_definitions(audio_core PRIVATE HAVE_SDL3)
if (ANDROID)
target_link_libraries(audio_core PRIVATE OpenSLES)
endif()
create_target_directory_groups(audio_core)
-450
View File
@@ -1,450 +0,0 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
#include <span>
#include <vector>
#include "audio_core/common/common.h"
#include "audio_core/sink/cubeb_sink.h"
#include "audio_core/sink/sink_stream.h"
#include "common/logging.h"
#include "common/scope_exit.h"
#include "core/core.h"
#ifdef _WIN32
#include <objbase.h>
#undef CreateEvent
#endif
namespace AudioCore::Sink {
/**
* Cubeb sink stream, responsible for sinking samples to hardware.
*/
class CubebSinkStream final : public SinkStream {
public:
/**
* Create a new sink stream.
*
* @param ctx_ - Cubeb context to create this stream with.
* @param device_channels_ - Number of channels supported by the hardware.
* @param system_channels_ - Number of channels the audio systems expect.
* @param output_device - Cubeb output device id.
* @param input_device - Cubeb input device id.
* @param name_ - Name of this stream.
* @param type_ - Type of this stream.
* @param system_ - Core system.
* @param event - Event used only for audio renderer, signalled on buffer consume.
*/
CubebSinkStream(cubeb* ctx_, u32 device_channels_, u32 system_channels_,
cubeb_devid output_device, cubeb_devid input_device, const std::string& name_,
StreamType type_, Core::System& system_)
: SinkStream(system_, type_), ctx{ctx_} {
#ifdef _WIN32
CoInitializeEx(nullptr, COINIT_MULTITHREADED);
#endif
name = name_;
device_channels = device_channels_;
system_channels = system_channels_;
cubeb_stream_params params{};
params.rate = TargetSampleRate;
params.channels = device_channels;
params.format = CUBEB_SAMPLE_S16LE;
params.prefs = CUBEB_STREAM_PREF_NONE;
switch (params.channels) {
case 1:
params.layout = CUBEB_LAYOUT_MONO;
break;
case 2:
params.layout = CUBEB_LAYOUT_STEREO;
break;
case 6:
params.layout = CUBEB_LAYOUT_3F2_LFE;
break;
}
u32 minimum_latency{0};
const auto latency_error = cubeb_get_min_latency(ctx, &params, &minimum_latency);
if (latency_error != CUBEB_OK) {
LOG_CRITICAL(Audio_Sink, "Error getting minimum latency, error: {}", latency_error);
minimum_latency = TargetSampleCount * 2;
}
minimum_latency = (std::max)(minimum_latency, TargetSampleCount * 2);
LOG_INFO(Service_Audio,
"Opening cubeb stream {} type {} with: rate {} channels {} (system channels {}) "
"latency {}",
name, type, params.rate, params.channels, system_channels, minimum_latency);
auto init_error{0};
if (type == StreamType::In) {
init_error = cubeb_stream_init(ctx, &stream_backend, name.c_str(), input_device,
&params, output_device, nullptr, minimum_latency,
&CubebSinkStream::DataCallback,
&CubebSinkStream::StateCallback, this);
} else {
init_error = cubeb_stream_init(ctx, &stream_backend, name.c_str(), input_device,
nullptr, output_device, &params, minimum_latency,
&CubebSinkStream::DataCallback,
&CubebSinkStream::StateCallback, this);
}
if (init_error != CUBEB_OK) {
LOG_CRITICAL(Audio_Sink, "Error initializing cubeb stream, error: {}", init_error);
return;
}
}
/**
* Destroy the sink stream.
*/
~CubebSinkStream() override {
LOG_DEBUG(Service_Audio, "Destructing cubeb stream {}", name);
if (!ctx) {
return;
}
Finalize();
#ifdef _WIN32
CoUninitialize();
#endif
}
/**
* Finalize the sink stream.
*/
void Finalize() override {
Stop();
cubeb_stream_destroy(stream_backend);
}
/**
* Start the sink stream.
*
* @param resume - Set to true if this is resuming the stream a previously-active stream.
* Default false.
*/
void Start(bool resume = false) override {
if (!ctx || !paused) {
return;
}
paused = false;
if (cubeb_stream_start(stream_backend) != CUBEB_OK) {
LOG_CRITICAL(Audio_Sink, "Error starting cubeb stream");
}
}
/**
* Stop the sink stream.
*/
void Stop() override {
if (!ctx || paused) {
return;
}
SignalPause();
if (cubeb_stream_stop(stream_backend) != CUBEB_OK) {
LOG_CRITICAL(Audio_Sink, "Error stopping cubeb stream");
}
}
private:
/**
* Main callback from Cubeb. Either expects samples from us (audio render/audio out), or will
* provide samples to be copied (audio in).
*
* @param stream - Cubeb-specific data about the stream.
* @param user_data - Custom data pointer passed along, points to a CubebSinkStream.
* @param in_buff - Input buffer to be used if the stream is an input type.
* @param out_buff - Output buffer to be used if the stream is an output type.
* @param num_frames_ - Number of frames of audio in the buffers. Note: Not number of samples.
*/
static long DataCallback([[maybe_unused]] cubeb_stream* stream, void* user_data,
[[maybe_unused]] const void* in_buff, void* out_buff,
long num_frames_) {
auto* impl = static_cast<CubebSinkStream*>(user_data);
if (!impl) {
return -1;
}
const std::size_t num_channels = impl->GetDeviceChannels();
const std::size_t frame_size = num_channels;
const std::size_t num_frames{static_cast<size_t>(num_frames_)};
if (impl->type == StreamType::In) {
std::span<const s16> input_buffer{reinterpret_cast<const s16*>(in_buff),
num_frames * frame_size};
impl->ProcessAudioIn(input_buffer, num_frames);
} else {
std::span<s16> output_buffer{reinterpret_cast<s16*>(out_buff), num_frames * frame_size};
impl->ProcessAudioOutAndRender(output_buffer, num_frames);
}
return num_frames_;
}
/**
* Cubeb callback for if a device state changes. Unused currently.
*
* @param stream - Cubeb-specific data about the stream.
* @param user_data - Custom data pointer passed along, points to a CubebSinkStream.
* @param state - New state of the device.
*/
static void StateCallback(cubeb_stream*, void*, cubeb_state) {}
/// Main Cubeb context
cubeb* ctx{};
/// Cubeb stream backend
cubeb_stream* stream_backend{};
};
CubebSink::CubebSink(std::string_view target_device_name) {
// Cubeb requires COM to be initialized on the thread calling cubeb_init on Windows
#ifdef _WIN32
com_init_result = CoInitializeEx(nullptr, COINIT_MULTITHREADED);
#endif
if (cubeb_init(&ctx, "Eden", nullptr) != CUBEB_OK) {
LOG_CRITICAL(Audio_Sink, "cubeb_init failed");
return;
}
if (target_device_name != auto_device_name && !target_device_name.empty()) {
cubeb_device_collection collection;
if (cubeb_enumerate_devices(ctx, CUBEB_DEVICE_TYPE_OUTPUT, &collection) != CUBEB_OK) {
LOG_WARNING(Audio_Sink, "Audio output device enumeration not supported");
} else {
const auto collection_end{collection.device + collection.count};
const auto device{
std::find_if(collection.device, collection_end, [&](const cubeb_device_info& info) {
return info.friendly_name != nullptr &&
target_device_name == std::string(info.friendly_name);
})};
if (device != collection_end) {
output_device = device->devid;
}
cubeb_device_collection_destroy(ctx, &collection);
}
}
cubeb_get_max_channel_count(ctx, &device_channels);
device_channels = device_channels >= 6U ? 6U : 2U;
}
CubebSink::~CubebSink() {
if (!ctx) {
return;
}
for (auto& sink_stream : sink_streams) {
sink_stream.reset();
}
cubeb_destroy(ctx);
#ifdef _WIN32
if (SUCCEEDED(com_init_result)) {
CoUninitialize();
}
#endif
}
SinkStream* CubebSink::AcquireSinkStream(Core::System& system, u32 system_channels_,
const std::string& name, StreamType type) {
system_channels = system_channels_;
SinkStreamPtr& stream = sink_streams.emplace_back(std::make_unique<CubebSinkStream>(
ctx, device_channels, system_channels, output_device, input_device, name, type, system));
return stream.get();
}
void CubebSink::CloseStream(SinkStream* stream) {
for (size_t i = 0; i < sink_streams.size(); i++) {
if (sink_streams[i].get() == stream) {
sink_streams[i].reset();
sink_streams.erase(sink_streams.begin() + i);
break;
}
}
}
void CubebSink::CloseStreams() {
sink_streams.clear();
}
f32 CubebSink::GetDeviceVolume() const {
if (sink_streams.empty()) {
return 1.0f;
}
return sink_streams[0]->GetDeviceVolume();
}
void CubebSink::SetDeviceVolume(f32 volume) {
for (auto& stream : sink_streams) {
stream->SetDeviceVolume(volume);
}
}
void CubebSink::SetSystemVolume(f32 volume) {
for (auto& stream : sink_streams) {
stream->SetSystemVolume(volume);
}
}
std::vector<std::string> ListCubebSinkDevices(bool capture) {
std::vector<std::string> device_list;
cubeb* ctx;
#ifdef _WIN32
auto com_init_result = CoInitializeEx(nullptr, COINIT_MULTITHREADED);
#endif
if (cubeb_init(&ctx, "Eden Device Enumerator", nullptr) != CUBEB_OK) {
LOG_CRITICAL(Audio_Sink, "cubeb_init failed");
return {};
}
#ifdef _WIN32
if (SUCCEEDED(com_init_result)) {
CoUninitialize();
}
#endif
auto type{capture ? CUBEB_DEVICE_TYPE_INPUT : CUBEB_DEVICE_TYPE_OUTPUT};
cubeb_device_collection collection;
if (cubeb_enumerate_devices(ctx, type, &collection) != CUBEB_OK) {
LOG_WARNING(Audio_Sink, "Audio output device enumeration not supported");
} else {
for (std::size_t i = 0; i < collection.count; i++) {
const cubeb_device_info& device = collection.device[i];
if (device.friendly_name && device.friendly_name[0] != '\0' &&
device.state == CUBEB_DEVICE_STATE_ENABLED) {
device_list.emplace_back(device.friendly_name);
}
}
cubeb_device_collection_destroy(ctx, &collection);
}
cubeb_destroy(ctx);
return device_list;
}
/* REVERSION TO 3833 - function GetCubebLatency REINTRODUCED FROM 3833 - DIABLO 3 FIX */
u32 GetCubebLatency() {
cubeb* ctx;
#ifdef _WIN32
auto com_init_result = CoInitializeEx(nullptr, COINIT_MULTITHREADED);
#endif
// Init cubeb
if (cubeb_init(&ctx, "yuzu Latency Getter", nullptr) != CUBEB_OK) {
LOG_CRITICAL(Audio_Sink, "cubeb_init failed");
// Return a large latency so we choose SDL instead.
return 10000u;
}
#ifdef _WIN32
if (SUCCEEDED(com_init_result)) {
CoUninitialize();
}
#endif
// Get min latency
cubeb_stream_params params{};
params.rate = TargetSampleRate;
params.channels = 2;
params.format = CUBEB_SAMPLE_S16LE;
params.prefs = CUBEB_STREAM_PREF_NONE;
params.layout = CUBEB_LAYOUT_STEREO;
u32 latency{0};
const auto latency_error = cubeb_get_min_latency(ctx, &params, &latency);
if (latency_error != CUBEB_OK) {
LOG_CRITICAL(Audio_Sink, "Error getting minimum latency, error: {}", latency_error);
latency = TargetSampleCount * 2;
}
latency = (std::max)(latency, TargetSampleCount * 2);
cubeb_destroy(ctx);
return latency;
}
// REVERTED back to 3833 - Below namespace section and function IsCubebSuitable() removed, reverting to GetCubebLatency() above. - DIABLO 3 FIX
/*
namespace {
static long TmpDataCallback(cubeb_stream*, void*, const void*, void*, long) {
return TargetSampleCount;
}
static void TmpStateCallback(cubeb_stream*, void*, cubeb_state) {}
} // namespace
bool IsCubebSuitable() {
#if !defined(HAVE_CUBEB)
return false;
#else
cubeb* ctx{nullptr};
#ifdef _WIN32
auto com_init_result = CoInitializeEx(nullptr, COINIT_MULTITHREADED);
#endif
// Init cubeb
if (cubeb_init(&ctx, "Eden Latency Getter", nullptr) != CUBEB_OK) {
LOG_ERROR(Audio_Sink, "Cubeb failed to init, it is not suitable.");
return false;
}
SCOPE_EXIT {
cubeb_destroy(ctx);
};
#ifdef _WIN32
if (SUCCEEDED(com_init_result)) {
CoUninitialize();
}
#endif
// Get min latency
cubeb_stream_params params{};
params.rate = TargetSampleRate;
params.channels = 2;
params.format = CUBEB_SAMPLE_S16LE;
params.prefs = CUBEB_STREAM_PREF_NONE;
params.layout = CUBEB_LAYOUT_STEREO;
u32 latency{0};
const auto latency_error = cubeb_get_min_latency(ctx, &params, &latency);
if (latency_error != CUBEB_OK) {
LOG_ERROR(Audio_Sink, "Cubeb could not get min latency, it is not suitable.");
return false;
}
latency = (std::max)(latency, TargetSampleCount * 2);
// Test opening a device with standard parameters
cubeb_devid output_device{0};
cubeb_devid input_device{0};
std::string name{"Eden test"};
cubeb_stream* stream{nullptr};
if (cubeb_stream_init(ctx, &stream, name.c_str(), input_device, nullptr, output_device, &params,
latency, &TmpDataCallback, &TmpStateCallback, nullptr) != CUBEB_OK) {
LOG_CRITICAL(Audio_Sink, "Cubeb could not open a device, it is not suitable.");
return false;
}
cubeb_stream_stop(stream);
cubeb_stream_destroy(stream);
return true;
#endif
}
*/
} // namespace AudioCore::Sink
-118
View File
@@ -1,118 +0,0 @@
// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
#pragma once
#include <string>
#include <vector>
#include <cubeb/cubeb.h>
#include "audio_core/sink/sink.h"
namespace Core {
class System;
}
namespace AudioCore::Sink {
class SinkStream;
/**
* Cubeb backend sink, holds multiple output streams and is responsible for sinking samples to
* hardware. Used by Audio Render, Audio In and Audio Out.
*/
class CubebSink final : public Sink {
public:
explicit CubebSink(std::string_view device_id);
~CubebSink() override;
/**
* Create a new sink stream.
*
* @param system - Core system.
* @param system_channels - Number of channels the audio system expects.
* May differ from the device's channel count.
* @param name - Name of this stream.
* @param type - Type of this stream, render/in/out.
*
* @return A pointer to the created SinkStream
*/
SinkStream* AcquireSinkStream(Core::System& system, u32 system_channels,
const std::string& name, StreamType type) override;
/**
* Close a given stream.
*
* @param stream - The stream to close.
*/
void CloseStream(SinkStream* stream) override;
/**
* Close all streams.
*/
void CloseStreams() override;
/**
* Get the device volume. Set from calls to the IAudioDevice service.
*
* @return Volume of the device.
*/
f32 GetDeviceVolume() const override;
/**
* Set the device volume. Set from calls to the IAudioDevice service.
*
* @param volume - New volume of the device.
*/
void SetDeviceVolume(f32 volume) override;
/**
* Set the system volume. Comes from the audio system using this stream.
*
* @param volume - New volume of the system.
*/
void SetSystemVolume(f32 volume) override;
private:
/// Backend Cubeb context
cubeb* ctx{};
/// Cubeb id of the actual hardware output device
cubeb_devid output_device{};
/// Cubeb id of the actual hardware input device
cubeb_devid input_device{};
/// Vector of streams managed by this sink
std::vector<SinkStreamPtr> sink_streams{};
#ifdef _WIN32
/// Cubeb required COM to be initialized multi-threaded on Windows
u32 com_init_result = 0;
#endif
};
/**
* Get a list of connected devices from Cubeb.
*
* @param capture - Return input (capture) devices if true, otherwise output devices.
*/
std::vector<std::string> ListCubebSinkDevices(bool capture);
// REVERSION - function GetCubebLatency() reintroduced from EA-3833 - DIABLO 3 FIX
/**
* Get the reported latency for this sink.
*
* @return Minimum latency for this sink.
*/
u32 GetCubebLatency();
/**
* Check if this backend is suitable for use.
* Checks if enabled, its latency, whether it opens successfully, etc.
*
* @return True is this backend is suitable, false otherwise.
*/
// bool IsCubebSuitable(); // REVERTED BACK TO GetCubebLatency() FROM 3833
} // namespace AudioCore::Sink
+60 -138
View File
@@ -25,9 +25,10 @@ namespace {
// See https://github.com/PCSX2/pcsx2/pull/12312
// "SDL and cubeb backends previously resulted in different names for the output which
// caused them be identified as different applications by the OS."
//
// Keep in sync with cubeb_sink.cpp name.
SDL_SetHint("SDL_AUDIO_DEVICE_APP_NAME", "yuzu Latency Getter");
SDL_SetHint("SDL_HINT_AUDIO_DEVICE_STREAM_ROLE", "Game");
// We do our own processing, so just have SDL copy our audio
SDL_SetHint("SDL_HINT_AUDIO_DEVICE_RAW_STREAM", "1");
if (!SDL_InitSubSystem(SDL_INIT_AUDIO)) {
LOG_CRITICAL(Audio_Sink, "SDL_InitSubSystem audio failed: {}", SDL_GetError());
return false;
@@ -82,23 +83,18 @@ public:
SDL_AudioSpec spec{};
spec.freq = TargetSampleRate;
spec.channels = static_cast<u8>(device_channels);
spec.format = SDL_AUDIO_S16;
spec.channels = u8(device_channels);
spec.format = SDL_AUDIO_S16LE;
std::string device_name{output_device};
bool capture{false};
if (type == StreamType::In) {
device_name = input_device;
capture = true;
}
auto const is_capture = (type == StreamType::In);
std::string device_name = is_capture ? input_device : output_device;
const SDL_AudioDeviceID audio_device = device_name.empty()
? (is_capture
? SDL_AUDIO_DEVICE_DEFAULT_RECORDING
: SDL_AUDIO_DEVICE_DEFAULT_PLAYBACK)
: FindAudioDeviceByName(device_name, is_capture);
const SDL_AudioDeviceID audio_device =
device_name.empty() ? (capture ? SDL_AUDIO_DEVICE_DEFAULT_RECORDING
: SDL_AUDIO_DEVICE_DEFAULT_PLAYBACK)
: FindAudioDeviceByName(device_name, capture);
stream = SDL_OpenAudioDeviceStream(audio_device, &spec, &SDLSinkStream::DataCallback,
this);
stream = SDL_OpenAudioDeviceStream(audio_device, &spec, &SDLSinkStream::DataCallback, this);
if (stream == nullptr) {
LOG_CRITICAL(Audio_Sink, "Error opening SDL audio device: {}", SDL_GetError());
@@ -107,13 +103,12 @@ public:
SDL_AudioSpec stream_in{};
SDL_AudioSpec stream_out{};
static_cast<void>(SDL_GetAudioStreamFormat(stream, &stream_in, &stream_out));
void(SDL_GetAudioStreamFormat(stream, &stream_in, &stream_out));
LOG_INFO(Service_Audio,
"Opening SDL stream {} with: rate {} channels {} (system channels {}) "
" format {}",
static_cast<const void*>(stream), stream_out.freq, stream_out.channels,
system_channels, static_cast<int>(stream_out.format));
system_channels, int(stream_out.format));
}
/**
@@ -128,14 +123,12 @@ public:
* Finalize the sink stream.
*/
void Finalize() override {
if (stream == nullptr) {
return;
if (stream != nullptr) {
Stop();
SDL_ClearAudioStream(stream);
SDL_DestroyAudioStream(stream);
stream = nullptr;
}
Stop();
SDL_ClearAudioStream(stream);
SDL_DestroyAudioStream(stream);
stream = nullptr;
}
/**
@@ -148,9 +141,8 @@ public:
if (stream == nullptr || !paused) {
return;
}
paused = false;
static_cast<void>(SDL_ResumeAudioStreamDevice(stream));
void(SDL_ResumeAudioStreamDevice(stream));
}
/**
@@ -161,7 +153,7 @@ public:
return;
}
SignalPause();
static_cast<void>(SDL_PauseAudioStreamDevice(stream));
void(SDL_PauseAudioStreamDevice(stream));
}
private:
@@ -173,51 +165,34 @@ private:
* @param stream - Buffer of samples to be filled or read.
* @param len - Length of the stream in bytes.
*/
static void DataCallback(void* userdata, SDL_AudioStream* stream, int additional_amount,
int total_amount) {
auto* impl = static_cast<SDLSinkStream*>(userdata);
if (!impl) {
return;
}
const std::size_t num_channels = impl->GetDeviceChannels();
const std::size_t frame_size = num_channels;
if (impl->type == StreamType::In) {
const int bytes_available = SDL_GetAudioStreamAvailable(stream);
if (bytes_available <= 0) {
return;
static void DataCallback(void* userdata, SDL_AudioStream* stream, int additional_amount, int total_amount) {
if (auto* impl = static_cast<SDLSinkStream*>(userdata); impl) {
auto const num_channels = impl->GetDeviceChannels();
auto const frame_size = num_channels;
if (impl->type == StreamType::In) {
auto const bytes_available = SDL_GetAudioStreamAvailable(stream);
if (bytes_available > 0) {
std::vector<s16> input(bytes_available / int(sizeof(s16)));
auto const bytes_read = SDL_GetAudioStreamData(stream, input.data(), bytes_available);
if (bytes_read > 0) {
auto const num_frames = std::size_t(bytes_read) / sizeof(s16) / frame_size;
std::span<const s16> input_buffer{input.data(), std::size_t(bytes_read) / sizeof(s16)};
impl->ProcessAudioIn(input_buffer, num_frames);
}
}
} else if (additional_amount > 0 || total_amount > 0) {
auto const bytes_requested = additional_amount > 0 ? additional_amount : total_amount;
std::vector<s16> output(bytes_requested / int(sizeof(s16)));
auto const num_frames = std::size_t(bytes_requested) / sizeof(s16) / frame_size;
std::span<s16> output_buffer{output.data(), output.size()};
impl->ProcessAudioOutAndRender(output_buffer, num_frames);
void(SDL_PutAudioStreamData(stream, output.data(), bytes_requested));
}
std::vector<s16> input(bytes_available / static_cast<int>(sizeof(s16)));
const int bytes_read = SDL_GetAudioStreamData(stream, input.data(), bytes_available);
if (bytes_read <= 0) {
return;
}
const std::size_t num_frames =
static_cast<std::size_t>(bytes_read) / sizeof(s16) / frame_size;
std::span<const s16> input_buffer{input.data(),
static_cast<std::size_t>(bytes_read) / sizeof(s16)};
impl->ProcessAudioIn(input_buffer, num_frames);
} else {
if (additional_amount <= 0 && total_amount <= 0) {
return;
}
const int bytes_requested = additional_amount > 0 ? additional_amount : total_amount;
std::vector<s16> output(bytes_requested / static_cast<int>(sizeof(s16)));
const std::size_t num_frames =
static_cast<std::size_t>(bytes_requested) / sizeof(s16) / frame_size;
std::span<s16> output_buffer{output.data(), output.size()};
impl->ProcessAudioOutAndRender(output_buffer, num_frames);
static_cast<void>(SDL_PutAudioStreamData(stream, output.data(), bytes_requested));
}
}
/// SDL stream attached to an opened input/output device
SDL_AudioStream* stream{};
SDL_AudioStream* stream = nullptr;
};
SDLSink::SDLSink(std::string_view target_device_name) {
@@ -233,11 +208,9 @@ SDLSink::SDLSink(std::string_view target_device_name) {
SDLSink::~SDLSink() = default;
SinkStream* SDLSink::AcquireSinkStream(Core::System& system, u32 system_channels_,
const std::string&, StreamType type) {
SinkStream* SDLSink::AcquireSinkStream(Core::System& system, u32 system_channels_, const std::string&, StreamType type) {
system_channels = system_channels_;
SinkStreamPtr& stream = sink_streams.emplace_back(std::make_unique<SDLSinkStream>(
device_channels, system_channels, output_device, input_device, type, system));
SinkStreamPtr& stream = sink_streams.emplace_back(std::make_unique<SDLSinkStream>(device_channels, system_channels, output_device, input_device, type, system));
return stream.get();
}
@@ -256,11 +229,7 @@ void SDLSink::CloseStreams() {
}
f32 SDLSink::GetDeviceVolume() const {
if (sink_streams.empty()) {
return 1.0f;
}
return sink_streams[0]->GetDeviceVolume();
return sink_streams.empty() ? 1.0f : sink_streams[0]->GetDeviceVolume();
}
void SDLSink::SetDeviceVolume(f32 volume) {
@@ -275,68 +244,21 @@ void SDLSink::SetSystemVolume(f32 volume) {
}
}
std::vector<std::string> ListSDLSinkDevices(bool capture) {
if (!InitializeAudio())
return {}; //no devices
std::vector<std::string> device_list;
int device_count = 0;
SDL_AudioDeviceID* devices =
capture ? SDL_GetAudioRecordingDevices(&device_count)
: SDL_GetAudioPlaybackDevices(&device_count);
if (devices == nullptr) {
return device_list;
}
for (int i = 0; i < device_count; ++i) {
if (const char* name = SDL_GetAudioDeviceName(devices[i])) {
device_list.emplace_back(name);
std::vector<std::string> ListSDLSinkDevices(bool is_capture) {
std::vector<std::string> device_list{};
if (InitializeAudio()) {
int device_count = 0;
SDL_AudioDeviceID* devices = is_capture ? SDL_GetAudioRecordingDevices(&device_count) : SDL_GetAudioPlaybackDevices(&device_count);
if (devices != nullptr) {
for (int i = 0; i < device_count; ++i) {
if (const char* name = SDL_GetAudioDeviceName(devices[i]); name) {
device_list.emplace_back(name);
}
}
SDL_free(devices);
}
}
SDL_free(devices);
return device_list;
}
/* REVERSION to 3833 - function GetSDLLatency() REINTRODUCED FROM 3833 - DIABLO 3 FIX */
u32 GetSDLLatency() {
return TargetSampleCount * 2;
}
// REVERTED back to 3833 - Below function IsSDLSuitable() removed, reverting to GetSDLLatency() above. - DIABLO 3 FIX
/*
bool IsSDLSuitable() {
#if !defined(HAVE_SDL3)
return false;
#else
// Check SDL can init
if (!InitializeAudio()!
return false;
// We can set any latency frequency we want with SDL, so no need to check that.
// Check we can open a device with standard parameters
SDL_AudioSpec spec;
spec.freq = TargetSampleRate;
spec.channels = 2u;
spec.format = AUDIO_S16SYS;
spec.samples = TargetSampleCount * 2;
spec.callback = nullptr;
spec.userdata = nullptr;
SDL_AudioSpec obtained;
auto device = SDL_OpenAudioDevice(nullptr, false, &spec, &obtained, false);
if (device == 0) {
LOG_ERROR(Audio_Sink, "SDL failed to open a device, it is not suitable. Error: {}",
SDL_GetError());
return false;
}
SDL_CloseAudioDevice(device);
return true;
#endif
}
*/
} // namespace AudioCore::Sink
-16
View File
@@ -90,20 +90,4 @@ private:
*/
std::vector<std::string> ListSDLSinkDevices(bool capture);
// REVERSION - function GetSDLLatency() reintroduced from EA-3833 - DIABLO 3 FIX
/**
* Get the reported latency for this sink.
*
* @return Minimum latency for this sink.
*/
u32 GetSDLLatency();
/** REVERTED back to 3833 - Below function IsSDLSuitable() removed, reverting to GetSDLLatency() above. - DIABLO 3 FIX
* Check if this backend is suitable for use.
* Checks if enabled, its latency, whether it opens successfully, etc.
*
* @return True is this backend is suitable, false otherwise.
*/
//bool IsSDLSuitable(); // REVERTED for GetSDLLatency() from EA-3833
} // namespace AudioCore::Sink
+7 -62
View File
@@ -10,9 +10,6 @@
#include <vector>
#include "audio_core/sink/sink_details.h"
#ifdef HAVE_CUBEB
#include "audio_core/sink/cubeb_sink.h"
#endif
#ifdef HAVE_SDL3
#include "audio_core/sink/sdl3_sink.h"
#endif
@@ -34,24 +31,10 @@ struct SinkDetails {
FactoryFn factory;
/// A method to call to list available devices.
ListDevicesFn list_devices;
/// Method to get the latency of this backend - REINTRODUCED FROM 3833 - DIABLO 3 FIX
LatencyFn latency;
/// Check whether this backend is suitable to be used.
/// SuitableFn is_suitable; // REVERTED FOR LatencyFn latency ABOVE - DIABLO 3 FIX
};
// sink_details is ordered in terms of desirability, with the best choice at the top.
constexpr SinkDetails sink_details[] = {
#ifdef HAVE_CUBEB
SinkDetails{
Settings::AudioEngine::Cubeb,
[](std::string_view device_id) -> std::unique_ptr<Sink> {
return std::make_unique<CubebSink>(device_id);
},
&ListCubebSinkDevices,
&GetCubebLatency,
},
#endif
static constexpr SinkDetails sink_details[] = {
#ifdef HAVE_SDL3
SinkDetails{
Settings::AudioEngine::Sdl3,
@@ -59,7 +42,6 @@ constexpr SinkDetails sink_details[] = {
return std::make_unique<SDLSink>(device_id);
},
&ListSDLSinkDevices,
&GetSDLLatency,
},
#endif
SinkDetails{
@@ -68,69 +50,32 @@ constexpr SinkDetails sink_details[] = {
return std::make_unique<NullSink>(device_id);
},
[](bool capture) { return std::vector<std::string>{"null"}; },
[]() { return 0u; },
},
};
const SinkDetails& GetOutputSinkDetails(Settings::AudioEngine sink_id) {
const auto find_backend{[](Settings::AudioEngine id) {
return std::find_if(std::begin(sink_details), std::end(sink_details),
[&id](const auto& sink_detail) { return sink_detail.id == id; });
return std::ranges::find_if(std::begin(sink_details), std::end(sink_details), [&id](const auto& e) {
return e.id == id;
});
}};
auto iter = find_backend(sink_id);
if (sink_id == Settings::AudioEngine::Auto) {
// REVERTED TO 3833 BELOW - DIABLO 3 FIX
/*
// Auto-select a backend. Use the sink details ordering, preferring cubeb first, checking
// that the backend is available and suitable to use.
for (auto& details : sink_details) {
if (details.is_suitable()) {
iter = &details;
break;
}
}
*/ // END REVERTED CODE - DIABLO 3 FIX
// BEGIN REINTRODUCED FROM 3833 - REPLACED CODE BLOCK ABOVE - DIABLO 3 FIX
// Auto-select a backend. Prefer CubeB, but it may report a large minimum latency which
// causes audio issues, in that case go with SDL.
#if defined(HAVE_CUBEB) && defined(HAVE_SDL3)
iter = find_backend(Settings::AudioEngine::Cubeb);
if (iter->latency() > TargetSampleCount * 3) {
if (sink_id == Settings::AudioEngine::Cubeb || sink_id == Settings::AudioEngine::Auto) {
iter = find_backend(Settings::AudioEngine::Sdl3);
}
#else
iter = std::begin(sink_details);
#endif
// END REINTRODUCED SECTION FROM 3833 - DIABLO 3 FIX
LOG_INFO(Service_Audio, "Auto-selecting the {} backend",
Settings::CanonicalizeEnum(iter->id));
/* BEGIN REMOVED - REVERTING BACK TO 3833, this didn't exist at all. - DIABLO 3 FIX
} else {
if (iter != std::end(sink_details) && !iter->is_suitable()) {
LOG_ERROR(Service_Audio, "Selected backend {} is not suitable, falling back to null",
Settings::CanonicalizeEnum(iter->id));
iter = find_backend(Settings::AudioEngine::Null);
} */ // END REMOVED REVERT - DIABLO 3 FIX
LOG_INFO(Service_Audio, "Auto-selecting the {} backend", Settings::CanonicalizeEnum(iter->id));
}
if (iter == std::end(sink_details)) {
LOG_ERROR(Audio, "Invalid sink_id {}", Settings::CanonicalizeEnum(sink_id));
iter = find_backend(Settings::AudioEngine::Null);
}
return *iter;
}
} // Anonymous namespace
std::vector<Settings::AudioEngine> GetSinkIDs() {
std::vector<Settings::AudioEngine> sink_ids(std::size(sink_details));
std::transform(std::begin(sink_details), std::end(sink_details), std::begin(sink_ids),
[](const auto& sink) { return sink.id; });
std::transform(std::begin(sink_details), std::end(sink_details), std::begin(sink_ids), [](const auto& sink) { return sink.id; });
return sink_ids;
}
-1
View File
@@ -5,7 +5,6 @@
// SPDX-License-Identifier: GPL-2.0-or-later
#include <string>
#include <cstdlib>
#include <string_view>
#ifdef _WIN32
#include <llvm/Demangle/Demangle.h>
+1 -5
View File
@@ -1,13 +1,9 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
#pragma once
#include <iterator>
#include <cstring>
#include "common/make_unique_for_overwrite.h"
@@ -65,7 +61,7 @@ public:
void resize(size_type size) {
if (size > buffer_capacity) {
auto new_buffer = Common::make_unique_for_overwrite<T[]>(size);
std::memcpy(new_buffer.get(), buffer.get(), buffer_capacity * sizeof(T));
std::move(buffer.get(), buffer.get() + buffer_capacity, new_buffer.get());
buffer = std::move(new_buffer);
buffer_capacity = size;
}
-8
View File
@@ -592,14 +592,6 @@ struct Values {
true,
true};
SwitchableSetting<bool> enable_shader_phi_tracking{linkage,
true,
"enable_shader_phi_tracking",
Category::RendererAdvanced,
Specialization::Default,
true,
true};
#ifdef __ANDROID__
SwitchableSetting<bool> use_optimized_vertex_buffers{linkage,
false,
-1
View File
@@ -10,7 +10,6 @@
#include <functional>
#include <span>
#include <string>
#include <type_traits>
#include "common/common_types.h"
+3 -3
View File
@@ -6,13 +6,13 @@
#include <mutex>
#include <utility>
#include <type_traits>
#include <boost/asio.hpp>
#include <boost/version.hpp>
#if BOOST_VERSION > 108400 && (!defined(_WINDOWS) && !defined(__ANDROID__)) || defined(YUZU_BOOST_v1)
#define USE_BOOST_v1
#endif
#ifdef USE_BOOST_v1
#include <boost/process/v1/async_pipe.hpp>
#else
@@ -358,7 +358,7 @@ private:
ConnectionState(boost::asio::ip::tcp::socket&& client_socket_, async_pipe signal_pipe_, Kernel::KernelCore& kernel)
: client_socket{std::move(client_socket_)}
, signal_pipe{std::move(signal_pipe_)}
, signal_pipe{signal_pipe_}
, active_thread{kernel, nullptr}
{}
+1 -2
View File
@@ -1,4 +1,4 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
@@ -6,7 +6,6 @@
#pragma once
#include <type_traits>
#include "common/common_funcs.h"
namespace FileSys {
@@ -7,7 +7,6 @@
#pragma once
#include <optional>
#include <type_traits>
#include "common/literals.h"
#include "core/file_sys/fssystem/fs_i_storage.h"
@@ -4,7 +4,6 @@
// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
#include <type_traits>
#include "core/file_sys/errors.h"
#include "core/file_sys/fssystem/fssystem_bucket_tree.h"
#include "core/file_sys/fssystem/fssystem_bucket_tree_utils.h"
@@ -1,13 +1,9 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
#pragma once
#include <mutex>
#include <type_traits>
#include "common/alignment.h"
#include "common/common_funcs.h"
@@ -1,4 +1,4 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
@@ -6,7 +6,6 @@
#pragma once
#include <type_traits>
#include "core/file_sys/errors.h"
#include "core/file_sys/fssystem/fssystem_bucket_tree.h"
#include "core/file_sys/fssystem/fssystem_bucket_tree_utils.h"
@@ -6,8 +6,6 @@
#pragma once
#include <type_traits>
#include <cstddef>
#include "common/literals.h"
#include "core/file_sys/errors.h"
@@ -1,12 +1,8 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
#pragma once
#include <type_traits>
#include "common/alignment.h"
#include "core/file_sys/fssystem/fs_i_storage.h"
#include "core/file_sys/fssystem/fs_types.h"
@@ -6,9 +6,6 @@
#pragma once
#include <type_traits>
#include <array>
#include <cstddef>
#include "core/file_sys/errors.h"
#include "core/file_sys/fssystem/fs_i_storage.h"
#include "core/file_sys/fssystem/fssystem_bucket_tree.h"
@@ -1,15 +1,9 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
#pragma once
#include <optional>
#include <array>
#include <cstddef>
#include <type_traits>
#include "core/file_sys/fssystem/fs_i_storage.h"
#include "core/file_sys/fssystem/fs_types.h"
@@ -1,4 +1,4 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
@@ -6,8 +6,6 @@
#pragma once
#include <type_traits>
#include <cstddef>
#include "core/file_sys/fssystem/fssystem_compression_common.h"
#include "core/file_sys/fssystem/fssystem_nca_header.h"
#include "core/file_sys/vfs/vfs.h"
@@ -1,14 +1,8 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
#pragma once
#include <type_traits>
#include <array>
#include <cstddef>
#include "common/common_funcs.h"
#include "common/common_types.h"
#include "common/literals.h"
-1
View File
@@ -7,7 +7,6 @@
#pragma once
#include <memory>
#include <type_traits>
#include "common/common_funcs.h"
#include "common/page_table.h"
-3
View File
@@ -1021,9 +1021,6 @@ Result KProcess::Run(KernelCore& kernel, s32 priority, size_t stack_size) {
// Suspend for debug, if we should.
if (kernel.System().DebuggerEnabled()) {
LOG_INFO(Debug_GDBStub,
"GDB stub enabled; suspending guest process until a debugger continues execution on port {}",
Settings::values.gdbstub_port.GetValue());
main_thread->RequestSuspend(kernel, SuspendType::Debug);
}
-1
View File
@@ -6,7 +6,6 @@
#pragma once
#include <type_traits>
#include "common/assert.h"
#include "common/bit_field.h"
#include "common/common_funcs.h"
-4
View File
@@ -1,13 +1,9 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2024 yuzu Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
#pragma once
#include <array>
#include <type_traits>
#include <functional>
#include "common/common_funcs.h"
+26 -8
View File
@@ -175,10 +175,19 @@ Result AlbumManager::LoadAlbumScreenShotImage(LoadAlbumScreenShotImageOutput& ou
return ResultIsNotMounted;
}
out_image_output = {};
out_image_output.width = 1280;
out_image_output.height = 720;
out_image_output.attribute.orientation = AlbumImageOrientation::None;
out_image_output = {
.width = 1280,
.height = 720,
.attribute =
{
.unknown_0{},
.orientation = AlbumImageOrientation::None,
.unknown_1{},
.unknown_2{},
.pad163{},
},
.pad179{},
};
std::filesystem::path path;
const auto result = GetFile(path, file_id);
@@ -202,10 +211,19 @@ Result AlbumManager::LoadAlbumScreenShotThumbnail(
return ResultIsNotMounted;
}
out_image_output = {};
out_image_output.width = 320;
out_image_output.height = 180;
out_image_output.attribute.orientation = AlbumImageOrientation::None;
out_image_output = {
.width = 320,
.height = 180,
.attribute =
{
.unknown_0{},
.orientation = AlbumImageOrientation::None,
.unknown_1{},
.unknown_2{},
.pad163{},
},
.pad179{},
};
std::filesystem::path path;
const auto result = GetFile(path, file_id);
+7 -2
View File
@@ -73,8 +73,13 @@ void IScreenShotApplicationService::CaptureAndSaveScreenshot(AlbumReportOption r
Layout::FramebufferLayout layout =
Layout::DefaultFrameLayout(screenshot_width, screenshot_height);
Capture::ScreenShotAttribute attribute{};
attribute.orientation = Capture::AlbumImageOrientation::None;
const Capture::ScreenShotAttribute attribute{
.unknown_0{},
.orientation = Capture::AlbumImageOrientation::None,
.unknown_1{},
.unknown_2{},
.pad163{},
};
renderer.RequestScreenshot(
image_data.data(),
-4
View File
@@ -1,12 +1,8 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
#pragma once
#include <type_traits>
#include "common/common_funcs.h"
#include "common/common_types.h"
+1 -2
View File
@@ -1,4 +1,4 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
@@ -6,7 +6,6 @@
#pragma once
#include <type_traits>
#include <fmt/ranges.h>
#include "common/common_funcs.h"
-1
View File
@@ -8,7 +8,6 @@
#include <array>
#include <chrono>
#include <type_traits>
#include <fmt/ranges.h>
#include "common/common_types.h"
@@ -1,13 +1,10 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
#pragma once
#include <array>
#include <type_traits>
#include "common/common_types.h"
#include "core/hle/service/psc/time/common.h"
@@ -7,7 +7,6 @@
#pragma once
#include <array>
#include <type_traits>
#include "common/bit_field.h"
#include "common/common_funcs.h"
-7
View File
@@ -1,15 +1,8 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
#pragma once
#include <cstddef>
#include <array>
#include <type_traits>
#include "common/bit_field.h"
#include "common/common_funcs.h"
#include "common/common_types.h"
+3
View File
@@ -31,3 +31,6 @@ endif()
if (SOLARIS)
target_link_libraries(network PRIVATE socket nsl)
endif()
if (MINGW)
target_link_libraries(network PRIVATE winmm)
endif()
@@ -267,8 +267,6 @@ std::unique_ptr<TranslationMap> InitializeTranslations(QObject* parent) {
INSERT(Settings, enable_buffer_history, tr("Enable buffer history"),
tr("Enables access to previous buffer states.\nThis option may improve rendering "
"quality and performance consistency in some games."));
INSERT(Settings, enable_shader_phi_tracking, tr("Shader phi tracking"),
tr("toggle for test on shader phi tracking."));
INSERT(Settings, fix_bloom_effects, tr("Fix bloom effects"), tr("Removes bloom in Burnout."));
INSERT(Settings, rescale_hack, tr("Enable Legacy Rescale Pass"),
+51 -53
View File
@@ -6,6 +6,55 @@
add_library(shader_recompiler STATIC
backend/bindings.h
backend/glasm/emit_glasm.cpp
backend/glasm/emit_glasm.h
backend/glasm/emit_glasm_barriers.cpp
backend/glasm/emit_glasm_bitwise_conversion.cpp
backend/glasm/emit_glasm_composite.cpp
backend/glasm/emit_glasm_context_get_set.cpp
backend/glasm/emit_glasm_control_flow.cpp
backend/glasm/emit_glasm_convert.cpp
backend/glasm/emit_glasm_floating_point.cpp
backend/glasm/emit_glasm_image.cpp
backend/glasm/emit_glasm_instructions.h
backend/glasm/emit_glasm_integer.cpp
backend/glasm/emit_glasm_logical.cpp
backend/glasm/emit_glasm_memory.cpp
backend/glasm/emit_glasm_not_implemented.cpp
backend/glasm/emit_glasm_select.cpp
backend/glasm/emit_glasm_shared_memory.cpp
backend/glasm/emit_glasm_special.cpp
backend/glasm/emit_glasm_undefined.cpp
backend/glasm/emit_glasm_warp.cpp
backend/glasm/glasm_emit_context.cpp
backend/glasm/glasm_emit_context.h
backend/glasm/reg_alloc.cpp
backend/glasm/reg_alloc.h
backend/glsl/emit_glsl.cpp
backend/glsl/emit_glsl.h
backend/glsl/emit_glsl_atomic.cpp
backend/glsl/emit_glsl_barriers.cpp
backend/glsl/emit_glsl_bitwise_conversion.cpp
backend/glsl/emit_glsl_composite.cpp
backend/glsl/emit_glsl_context_get_set.cpp
backend/glsl/emit_glsl_control_flow.cpp
backend/glsl/emit_glsl_convert.cpp
backend/glsl/emit_glsl_floating_point.cpp
backend/glsl/emit_glsl_image.cpp
backend/glsl/emit_glsl_instructions.h
backend/glsl/emit_glsl_integer.cpp
backend/glsl/emit_glsl_logical.cpp
backend/glsl/emit_glsl_memory.cpp
backend/glsl/emit_glsl_not_implemented.cpp
backend/glsl/emit_glsl_select.cpp
backend/glsl/emit_glsl_shared_memory.cpp
backend/glsl/emit_glsl_special.cpp
backend/glsl/emit_glsl_undefined.cpp
backend/glsl/emit_glsl_warp.cpp
backend/glsl/glsl_emit_context.cpp
backend/glsl/glsl_emit_context.h
backend/glsl/var_alloc.cpp
backend/glsl/var_alloc.h
backend/spirv/emit_spirv.cpp
backend/spirv/emit_spirv.h
backend/spirv/emit_spirv_atomic.cpp
@@ -190,60 +239,9 @@ add_library(shader_recompiler STATIC
program_header.h
runtime_info.h
shader_info.h
varying_state.h)
varying_state.h
if (ENABLE_OPENGL)
target_sources(shader_recompiler PRIVATE
backend/glasm/emit_glasm.cpp
backend/glasm/emit_glasm.h
backend/glasm/emit_glasm_barriers.cpp
backend/glasm/emit_glasm_bitwise_conversion.cpp
backend/glasm/emit_glasm_composite.cpp
backend/glasm/emit_glasm_context_get_set.cpp
backend/glasm/emit_glasm_control_flow.cpp
backend/glasm/emit_glasm_convert.cpp
backend/glasm/emit_glasm_floating_point.cpp
backend/glasm/emit_glasm_image.cpp
backend/glasm/emit_glasm_instructions.h
backend/glasm/emit_glasm_integer.cpp
backend/glasm/emit_glasm_logical.cpp
backend/glasm/emit_glasm_memory.cpp
backend/glasm/emit_glasm_not_implemented.cpp
backend/glasm/emit_glasm_select.cpp
backend/glasm/emit_glasm_shared_memory.cpp
backend/glasm/emit_glasm_special.cpp
backend/glasm/emit_glasm_undefined.cpp
backend/glasm/emit_glasm_warp.cpp
backend/glasm/glasm_emit_context.cpp
backend/glasm/glasm_emit_context.h
backend/glasm/reg_alloc.cpp
backend/glasm/reg_alloc.h
backend/glsl/emit_glsl.cpp
backend/glsl/emit_glsl.h
backend/glsl/emit_glsl_atomic.cpp
backend/glsl/emit_glsl_barriers.cpp
backend/glsl/emit_glsl_bitwise_conversion.cpp
backend/glsl/emit_glsl_composite.cpp
backend/glsl/emit_glsl_context_get_set.cpp
backend/glsl/emit_glsl_control_flow.cpp
backend/glsl/emit_glsl_convert.cpp
backend/glsl/emit_glsl_floating_point.cpp
backend/glsl/emit_glsl_image.cpp
backend/glsl/emit_glsl_instructions.h
backend/glsl/emit_glsl_integer.cpp
backend/glsl/emit_glsl_logical.cpp
backend/glsl/emit_glsl_memory.cpp
backend/glsl/emit_glsl_not_implemented.cpp
backend/glsl/emit_glsl_select.cpp
backend/glsl/emit_glsl_shared_memory.cpp
backend/glsl/emit_glsl_special.cpp
backend/glsl/emit_glsl_undefined.cpp
backend/glsl/emit_glsl_warp.cpp
backend/glsl/glsl_emit_context.cpp
backend/glsl/glsl_emit_context.h
backend/glsl/var_alloc.cpp
backend/glsl/var_alloc.h)
endif()
)
target_link_libraries(shader_recompiler PUBLIC common fmt::fmt sirit::sirit)
@@ -553,6 +553,17 @@ void GlobalMemoryToStorageBufferPass(IR::Program& program, const HostTranslateIn
}
}
template <typename Descriptors, typename Descriptor, typename Func>
static u32 Add(Descriptors& descriptors, const Descriptor& desc, Func&& pred) {
// TODO: Handle arrays
const auto it{std::ranges::find_if(descriptors, pred)};
if (it != descriptors.end()) {
return static_cast<u32>(std::distance(descriptors.begin(), it));
}
descriptors.push_back(desc);
return static_cast<u32>(descriptors.size()) - 1;
}
void JoinStorageInfo(Info& base, Info& source) {
auto& descriptors = base.storage_buffers_descriptors;
for (auto& desc : source.storage_buffers_descriptors) {
+15 -124
View File
@@ -299,128 +299,23 @@ static inline bool IsTexturePixelFormatIntegerCached(Environment& env,
}
constexpr size_t PHI_TRACK_MAX_DEPTH = 3;
struct PhiTrackState {
boost::container::small_vector<const IR::Inst*, 8> active;
size_t depth{};
};
std::optional<ConstBufferAddr> Track(const IR::Value& value, Environment& env,
const HostTranslateInfo& host_info, PhiTrackState& state);
static inline std::optional<ConstBufferAddr> TrackCached(const IR::Value& v, Environment& env,
const HostTranslateInfo& host_info,
PhiTrackState& state) {
std::optional<ConstBufferAddr> Track(const IR::Value& value, Environment& env, const HostTranslateInfo& host_info);
static inline std::optional<ConstBufferAddr> TrackCached(const IR::Value& v, Environment& env, const HostTranslateInfo& host_info) {
if (const IR::Inst* key = v.InstRecursive()) {
if (auto it = env.track_cache.find(key); it != env.track_cache.end()) return it->second;
auto found = Track(v, env, host_info, state);
auto found = Track(v, env, host_info);
if (found) env.track_cache.emplace(key, *found);
return found;
}
return Track(v, env, host_info, state);
return Track(v, env, host_info);
}
std::optional<ConstBufferAddr> TryGetConstBuffer(const IR::Inst* inst, Environment& env,
const HostTranslateInfo& host_info,
PhiTrackState& state);
std::optional<ConstBufferAddr> TryGetConstBuffer(const IR::Inst* inst, Environment& env, const HostTranslateInfo& host_info);
bool IsSameConstBufferAddr(const ConstBufferAddr& lhs, const ConstBufferAddr& rhs) {
return lhs.index == rhs.index && lhs.offset == rhs.offset &&
lhs.shift_left == rhs.shift_left && lhs.secondary_index == rhs.secondary_index &&
lhs.secondary_offset == rhs.secondary_offset &&
lhs.secondary_shift_left == rhs.secondary_shift_left && lhs.count == rhs.count &&
lhs.has_secondary == rhs.has_secondary && lhs.dynamic_offset == rhs.dynamic_offset;
}
std::optional<ConstBufferAddr> TrackUncached(const IR::Value& value, Environment& env,
const HostTranslateInfo& host_info,
PhiTrackState& state, bool& ambiguous);
std::optional<ConstBufferAddr> TrackPhi(const IR::Inst* phi, Environment& env,
const HostTranslateInfo& host_info, PhiTrackState& state,
bool& ambiguous) {
if (state.depth >= PHI_TRACK_MAX_DEPTH) {
ambiguous = true;
return std::nullopt;
}
if (std::ranges::find(state.active, phi) != state.active.end()) {
return std::nullopt;
}
state.active.push_back(phi);
++state.depth;
std::optional<ConstBufferAddr> agreed;
bool failed = false;
const size_t num_args{phi->NumArgs()};
for (size_t index = 0; index < num_args; ++index) {
const IR::Value arg{phi->Arg(index).Resolve()};
if (arg.IsImmediate()) {
failed = true;
break;
}
const IR::Inst* arg_inst{arg.InstRecursive()};
if (arg_inst == phi) {
continue;
}
if (std::ranges::find(state.active, arg_inst) != state.active.end()) {
continue;
}
bool operand_ambiguous = false;
const std::optional<ConstBufferAddr> operand{
TrackUncached(arg, env, host_info, state, operand_ambiguous)};
if (!operand || operand_ambiguous) {
failed = true;
break;
}
if (!agreed) {
agreed = operand;
continue;
}
if (!IsSameConstBufferAddr(*agreed, *operand)) {
failed = true;
break;
}
}
--state.depth;
state.active.pop_back();
if (failed || !agreed) {
ambiguous = true;
return std::nullopt;
}
return agreed;
}
std::optional<ConstBufferAddr> TrackUncached(const IR::Value& value, Environment& env,
const HostTranslateInfo& host_info,
PhiTrackState& state, bool& ambiguous) {
return IR::BreadthFirstSearch(
value, [&env, &host_info, &state, &ambiguous](
const IR::Inst* inst) -> std::optional<ConstBufferAddr> {
if (inst->GetOpcode() == IR::Opcode::Phi) {
return TrackPhi(inst, env, host_info, state, ambiguous);
}
return TryGetConstBuffer(inst, env, host_info, state);
});
}
std::optional<ConstBufferAddr> Track(const IR::Value& value, Environment& env,
const HostTranslateInfo& host_info, PhiTrackState& state) {
if (!Settings::values.enable_shader_phi_tracking.GetValue()) {
return IR::BreadthFirstSearch(
value, [&env, &host_info, &state](
const IR::Inst* inst) -> std::optional<ConstBufferAddr> {
return TryGetConstBuffer(inst, env, host_info, state);
});
}
bool ambiguous = false;
const std::optional<ConstBufferAddr> result{
TrackUncached(value, env, host_info, state, ambiguous)};
if (ambiguous) {
return std::nullopt;
}
return result;
std::optional<ConstBufferAddr> Track(const IR::Value& value, Environment& env, const HostTranslateInfo& host_info) {
return IR::BreadthFirstSearch(value, [&env, &host_info](const IR::Inst* inst) {
return TryGetConstBuffer(inst, env, host_info);
});
}
std::optional<u32> TryGetConstant(IR::Value& value, Environment& env) {
@@ -444,15 +339,13 @@ std::optional<u32> TryGetConstant(IR::Value& value, Environment& env) {
return ReadCbufCached(env, index_number, offset_number);
}
std::optional<ConstBufferAddr> TryGetConstBuffer(const IR::Inst* inst, Environment& env,
const HostTranslateInfo& host_info,
PhiTrackState& state) {
std::optional<ConstBufferAddr> TryGetConstBuffer(const IR::Inst* inst, Environment& env, const HostTranslateInfo& host_info) {
switch (inst->GetOpcode()) {
default:
return std::nullopt;
case IR::Opcode::BitwiseOr32: {
std::optional lhs{TrackCached(inst->Arg(0), env, host_info, state)};
std::optional rhs{TrackCached(inst->Arg(1), env, host_info, state)};
std::optional lhs{TrackCached(inst->Arg(0), env, host_info)};
std::optional rhs{TrackCached(inst->Arg(1), env, host_info)};
if (!lhs || !rhs) {
return std::nullopt;
}
@@ -482,7 +375,7 @@ std::optional<ConstBufferAddr> TryGetConstBuffer(const IR::Inst* inst, Environme
if (!shift.IsImmediate()) {
return std::nullopt;
}
std::optional lhs{TrackCached(inst->Arg(0), env, host_info, state)};
std::optional lhs{TrackCached(inst->Arg(0), env, host_info)};
if (lhs) {
lhs->shift_left = shift.U32();
}
@@ -510,7 +403,7 @@ std::optional<ConstBufferAddr> TryGetConstBuffer(const IR::Inst* inst, Environme
return std::nullopt;
} while (false);
}
std::optional lhs{TrackCached(op1, env, host_info, state)};
std::optional lhs{TrackCached(op1, env, host_info)};
if (lhs) {
lhs->shift_left = static_cast<u32>(std::countr_zero(op2.U32()));
}
@@ -576,9 +469,7 @@ std::optional<ConstBufferAddr> TryGetConstBuffer(const IR::Inst* inst, Environme
TextureInst MakeInst(Environment& env, IR::Block* block, IR::Inst& inst, const HostTranslateInfo& host_info) {
ConstBufferAddr addr;
if (IsBindless(inst)) {
PhiTrackState state;
const std::optional<ConstBufferAddr> track_addr{
TrackCached(inst.Arg(0), env, host_info, state)};
const std::optional<ConstBufferAddr> track_addr{TrackCached(inst.Arg(0), env, host_info)};
if (!track_addr) {
throw NotImplementedException("Failed to track bindless texture constant buffer");
-3
View File
@@ -20,7 +20,6 @@ add_library(video_core STATIC
buffer_cache/buffer_cache.h
buffer_cache/memory_tracker_base.h
buffer_cache/usage_tracker.h
buffer_cache/virtual_range_cache.h
buffer_cache/word_manager.h
cache_types.h
capture.h
@@ -167,8 +166,6 @@ add_library(video_core STATIC
renderer_vulkan/vk_fence_manager.h
renderer_vulkan/vk_graphics_pipeline.cpp
renderer_vulkan/vk_graphics_pipeline.h
renderer_vulkan/vk_multi_range_buffer.cpp
renderer_vulkan/vk_multi_range_buffer.h
renderer_vulkan/vk_master_semaphore.cpp
renderer_vulkan/vk_master_semaphore.h
renderer_vulkan/vk_pipeline_cache.cpp
+9 -117
View File
@@ -112,13 +112,6 @@ void BufferCache<P>::TickFrame() {
async_buffers_death_ring.clear();
}
template <class P>
void BufferCache<P>::UnmapGPUMemory(size_t as_id, GPUVAddr gpu_addr, size_t size) {
if constexpr (requires { runtime.BindMultiRangeStorageBuffer(u64{}); }) {
virtual_ranges.Unmap(as_id, gpu_addr, size);
}
}
template <class P>
void BufferCache<P>::WriteMemory(DAddr device_addr, u64 size) {
if (memory_tracker.IsRegionGpuModified(device_addr, size)) {
@@ -1005,85 +998,11 @@ void BufferCache<P>::BindHostGraphicsUniformBuffer(size_t stage, u32 index, u32
channel_state->fast_bound_uniform_buffers[stage] &= ~(1u << binding_index);
}
template <class P>
void BufferCache<P>::ResolveMultiRangeStorage(Binding& binding, bool is_written,
std::vector<MultiRangeSegment>& pool) {
binding.segment_first = 0;
binding.segment_count = 0;
if constexpr (requires { runtime.BindMultiRangeStorageBuffer(u64{}); }) {
if (binding.gpu_addr == 0 || binding.size == 0) {
return;
}
if (is_written && !runtime.PrefersSparseSources()) {
return;
}
const VirtualSegments* found =
virtual_ranges.Query(*gpu_memory, binding.gpu_addr, binding.size);
if (!found || found->size() < 2) {
return;
}
const VirtualSegments segments = *found;
const u32 first = static_cast<u32>(pool.size());
const bool prefer_sparse = runtime.PrefersSparseSources();
for (const VirtualSegment& segment : segments) {
const BufferId buffer_id =
FindBuffer(segment.device_addr, segment.size, prefer_sparse);
if (!buffer_id) {
pool.resize(first);
return;
}
pool.push_back(MultiRangeSegment{
.buffer_id = buffer_id,
.device_addr = segment.device_addr,
.size = segment.size,
});
}
binding.segment_first = first;
binding.segment_count = static_cast<u32>(segments.size());
}
}
template <class P>
bool BufferCache<P>::BindMultiRangeStorage(const Binding& binding, bool is_written,
std::span<const MultiRangeSegment> pool) {
if constexpr (requires { runtime.BindMultiRangeStorageBuffer(u64{}); }) {
if (binding.segment_count < 2) {
return false;
}
if (binding.segment_first + binding.segment_count > pool.size()) {
return false;
}
const u64 key = (static_cast<u64>(gpu_memory->GetID()) << 48) ^ binding.gpu_addr;
runtime.ResetMultiRange();
for (u32 index = 0; index < binding.segment_count; ++index) {
const MultiRangeSegment& segment = pool[binding.segment_first + index];
Buffer& buffer = slot_buffers[segment.buffer_id];
TouchBuffer(buffer, segment.buffer_id);
if (SynchronizeBuffer(buffer, segment.device_addr, segment.size)) {
runtime.InvalidateMultiRange(key);
}
const u32 offset = buffer.Offset(segment.device_addr);
buffer.MarkUsage(offset, segment.size);
if (is_written) {
MarkWrittenBuffer(segment.buffer_id, segment.device_addr, segment.size);
}
runtime.PushMultiRangeSource(buffer, offset, segment.size);
}
return runtime.BindMultiRangeStorageBuffer(key);
} else {
return false;
}
}
template <class P>
void BufferCache<P>::BindHostGraphicsStorageBuffers(size_t stage) {
u32 binding_index = 0;
ForEachEnabledBit(channel_state->enabled_storage_buffers[stage], [&](u32 index) {
const Binding& binding = channel_state->storage_buffers[stage][index];
const bool is_written = ((channel_state->written_storage_buffers[stage] >> index) & 1) != 0;
if (BindMultiRangeStorage(binding, is_written, graphics_segments)) {
return;
}
Buffer& buffer = slot_buffers[binding.buffer_id];
TouchBuffer(buffer, binding.buffer_id);
const u32 size = binding.size;
@@ -1091,6 +1010,7 @@ void BufferCache<P>::BindHostGraphicsStorageBuffers(size_t stage) {
const u32 offset = buffer.Offset(binding.device_addr);
buffer.MarkUsage(offset, size);
const bool is_written = ((channel_state->written_storage_buffers[stage] >> index) & 1) != 0;
if (is_written) {
MarkWrittenBuffer(binding.buffer_id, binding.device_addr, size);
@@ -1219,11 +1139,6 @@ void BufferCache<P>::BindHostComputeStorageBuffers() {
u32 binding_index = 0;
ForEachEnabledBit(channel_state->enabled_compute_storage_buffers, [&](u32 index) {
const Binding& binding = channel_state->compute_storage_buffers[index];
const bool is_written =
((channel_state->written_compute_storage_buffers >> index) & 1) != 0;
if (BindMultiRangeStorage(binding, is_written, compute_segments)) {
return;
}
Buffer& buffer = slot_buffers[binding.buffer_id];
TouchBuffer(buffer, binding.buffer_id);
const u32 size = binding.size;
@@ -1231,6 +1146,8 @@ void BufferCache<P>::BindHostComputeStorageBuffers() {
const u32 offset = buffer.Offset(binding.device_addr);
buffer.MarkUsage(offset, size);
const bool is_written =
((channel_state->written_compute_storage_buffers >> index) & 1) != 0;
if (is_written) {
MarkWrittenBuffer(binding.buffer_id, binding.device_addr, size);
@@ -1276,7 +1193,6 @@ void BufferCache<P>::BindHostComputeTextureBuffers() {
template <class P>
void BufferCache<P>::DoUpdateGraphicsBuffers(bool is_indexed) {
graphics_segments.clear();
BufferOperations([&]() {
if (is_indexed) {
UpdateIndexBuffer();
@@ -1296,7 +1212,6 @@ void BufferCache<P>::DoUpdateGraphicsBuffers(bool is_indexed) {
template <class P>
void BufferCache<P>::DoUpdateComputeBuffers() {
compute_segments.clear();
BufferOperations([&]() {
UpdateComputeUniformBuffers();
UpdateComputeStorageBuffers();
@@ -1439,8 +1354,6 @@ void BufferCache<P>::UpdateStorageBuffers(size_t stage) {
Binding& binding = channel_state->storage_buffers[stage][index];
const BufferId buffer_id = FindBuffer(binding.device_addr, binding.size);
binding.buffer_id = buffer_id;
const bool is_written = ((channel_state->written_storage_buffers[stage] >> index) & 1) != 0;
ResolveMultiRangeStorage(binding, is_written, graphics_segments);
});
}
@@ -1504,9 +1417,6 @@ void BufferCache<P>::UpdateComputeStorageBuffers() {
// Resolve buffer
Binding& binding = channel_state->compute_storage_buffers[index];
binding.buffer_id = FindBuffer(binding.device_addr, binding.size);
const bool is_written =
((channel_state->written_compute_storage_buffers >> index) & 1) != 0;
ResolveMultiRangeStorage(binding, is_written, compute_segments);
});
}
@@ -1530,7 +1440,7 @@ void BufferCache<P>::MarkWrittenBuffer(BufferId buffer_id, DAddr device_addr, u3
}
template <class P>
BufferId BufferCache<P>::FindBuffer(DAddr device_addr, u32 size, bool sparse_compatible) {
BufferId BufferCache<P>::FindBuffer(DAddr device_addr, u32 size) {
if (device_addr == 0) {
return NULL_BUFFER_ID;
}
@@ -1540,18 +1450,10 @@ BufferId BufferCache<P>::FindBuffer(DAddr device_addr, u32 size, bool sparse_com
Buffer& buffer = slot_buffers[buffer_id];
WaitForGpuFenceIfNeeded(buffer);
if (buffer.IsInBounds(device_addr, size)) {
bool usable = true;
if constexpr (requires { buffer.IsSparseCompatible(); }) {
if (sparse_compatible && !buffer.IsSparseCompatible()) {
usable = false;
}
}
if (usable) {
return buffer_id;
}
return buffer_id;
}
}
return CreateBuffer(device_addr, size, sparse_compatible);
return CreateBuffer(device_addr, size);
}
template <class P>
@@ -1673,15 +1575,13 @@ void BufferCache<P>::JoinOverlap(BufferId new_buffer_id, BufferId overlap_id,
}
template <class P>
BufferId BufferCache<P>::CreateBuffer(DAddr device_addr, u32 wanted_size,
bool sparse_compatible) {
BufferId BufferCache<P>::CreateBuffer(DAddr device_addr, u32 wanted_size) {
DAddr device_addr_end = Common::AlignUp(device_addr + wanted_size, CACHING_PAGESIZE);
device_addr = Common::AlignDown(device_addr, CACHING_PAGESIZE);
wanted_size = static_cast<u32>(device_addr_end - device_addr);
const OverlapResult overlap = ResolveOverlaps(device_addr, wanted_size);
const u32 size = static_cast<u32>(overlap.end - overlap.begin);
const BufferId new_buffer_id =
slot_buffers.insert(runtime, overlap.begin, size, sparse_compatible);
const BufferId new_buffer_id = slot_buffers.insert(runtime, overlap.begin, size);
auto& new_buffer = slot_buffers[new_buffer_id];
const size_t size_bytes = new_buffer.SizeBytes();
runtime.ClearBuffer(new_buffer, 0, size_bytes, 0);
@@ -1931,9 +1831,6 @@ void BufferCache<P>::DownloadBufferMemory(Buffer& buffer, DAddr device_addr, u64
template <class P>
void BufferCache<P>::DeleteBuffer(BufferId buffer_id, bool do_not_mark) {
if constexpr (requires { runtime.OnBufferDeleted(slot_buffers[buffer_id]); }) {
runtime.OnBufferDeleted(slot_buffers[buffer_id]);
}
bool dirty_index{false};
boost::container::small_vector<u64, NUM_VERTEX_BUFFERS> dirty_vertex_buffers;
const auto scalar_replace = [buffer_id](Binding& binding) {
@@ -2037,15 +1934,10 @@ Binding BufferCache<P>::StorageBufferBinding(GPUVAddr ssbo_addr, u32 cbuf_index,
// The end address used for size calculation does not need to be aligned
const DAddr cpu_end = Common::AlignUp(*device_addr + size, Core::DEVICE_PAGESIZE);
u32 binding_size = static_cast<u32>(cpu_end - *aligned_device_addr);
if (is_written) {
binding_size = aligned_size;
}
const Binding binding{
.device_addr = *aligned_device_addr,
.size = binding_size,
.size = is_written ? aligned_size : static_cast<u32>(cpu_end - *aligned_device_addr),
.buffer_id = BufferId{},
.gpu_addr = aligned_gpu_addr,
};
return binding;
}
@@ -29,7 +29,6 @@
#include "common/settings.h"
#include "common/slot_vector.h"
#include "video_core/buffer_cache/buffer_base.h"
#include "video_core/buffer_cache/virtual_range_cache.h"
#include "video_core/control/channel_state_cache.h"
#include "video_core/delayed_destruction_ring.h"
#include "video_core/dirty_flags.h"
@@ -84,15 +83,6 @@ struct Binding {
DAddr device_addr{};
u32 size{};
BufferId buffer_id;
GPUVAddr gpu_addr{};
u32 segment_first{};
u32 segment_count{};
};
struct MultiRangeSegment {
BufferId buffer_id;
DAddr device_addr{};
u32 size{};
};
struct TextureBufferBinding : Binding {
@@ -225,14 +215,6 @@ public:
void TickFrame();
bool BindMultiRangeStorage(const Binding& binding, bool is_written,
std::span<const MultiRangeSegment> pool);
void ResolveMultiRangeStorage(Binding& binding, bool is_written,
std::vector<MultiRangeSegment>& pool);
void UnmapGPUMemory(size_t as_id, GPUVAddr gpu_addr, size_t size);
void WriteMemory(DAddr device_addr, u64 size);
void CachedWriteMemory(DAddr device_addr, u64 size);
@@ -432,8 +414,7 @@ private:
void MarkWrittenBuffer(BufferId buffer_id, DAddr device_addr, u32 size);
[[nodiscard]] BufferId FindBuffer(DAddr device_addr, u32 size,
bool sparse_compatible = false);
[[nodiscard]] BufferId FindBuffer(DAddr device_addr, u32 size);
void WaitForGpuFenceIfNeeded(Buffer& buffer);
@@ -441,8 +422,7 @@ private:
void JoinOverlap(BufferId new_buffer_id, BufferId overlap_id, bool accumulate_stream_score);
[[nodiscard]] BufferId CreateBuffer(DAddr device_addr, u32 wanted_size,
bool sparse_compatible = false);
[[nodiscard]] BufferId CreateBuffer(DAddr device_addr, u32 wanted_size);
void Register(BufferId buffer_id);
@@ -534,9 +514,6 @@ private:
};
Common::LeastRecentlyUsedCache<LRUItemParams> lru_cache;
u64 frame_tick = 0;
VirtualRangeCache virtual_ranges;
std::vector<MultiRangeSegment> graphics_segments;
std::vector<MultiRangeSegment> compute_segments;
u64 total_used_memory = 0;
u64 minimum_memory = 0;
u64 critical_memory = 0;
@@ -1,163 +0,0 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
#pragma once
#include <atomic>
#include <limits>
#include <mutex>
#include <optional>
#include <unordered_map>
#include <vector>
#include <boost/container/small_vector.hpp>
#include "common/common_types.h"
#include "video_core/memory_manager.h"
namespace VideoCommon {
struct VirtualSegment {
GPUVAddr gpu_addr;
DAddr device_addr;
u32 size;
};
using VirtualSegments = boost::container::small_vector<VirtualSegment, 8>;
class VirtualRangeCache {
public:
const VirtualSegments* Query(Tegra::MemoryManager& memory, GPUVAddr gpu_addr, u32 size) {
if (has_deferred.load(std::memory_order_acquire)) {
ApplyDeferred();
}
const size_t as_id = memory.GetID();
const u64 key = MakeKey(as_id, gpu_addr);
const auto it = entries.find(key);
if (it != entries.end() && it->second.as_id == as_id &&
it->second.gpu_addr == gpu_addr && it->second.size == size) {
return &it->second.segments;
}
Entry entry;
entry.as_id = as_id;
entry.gpu_addr = gpu_addr;
entry.size = size;
const auto ranges = memory.GetSubmappedRange(gpu_addr, size);
GPUVAddr expected = gpu_addr;
bool contiguous = true;
for (const auto& [range_addr, range_size] : ranges) {
if (range_addr != expected || range_size == 0) {
contiguous = false;
break;
}
const std::optional<DAddr> device_addr = memory.GpuToCpuAddress(range_addr);
if (!device_addr || *device_addr == 0) {
contiguous = false;
break;
}
if (range_size > static_cast<size_t>((std::numeric_limits<u32>::max)())) {
contiguous = false;
break;
}
entry.segments.push_back(VirtualSegment{
.gpu_addr = range_addr,
.device_addr = *device_addr,
.size = static_cast<u32>(range_size),
});
expected += range_size;
}
if (!contiguous || expected != gpu_addr + size) {
entry.segments.clear();
}
const auto result = entries.insert_or_assign(key, std::move(entry));
return &result.first->second.segments;
}
void Unmap(size_t as_id, GPUVAddr gpu_addr, u64 size) {
if (size == 0) {
return;
}
{
std::scoped_lock lock{deferred_mutex};
if (!deferred.empty()) {
DeferredUnmap& last = deferred.back();
if (last.as_id == as_id && last.gpu_addr + last.size == gpu_addr) {
last.size += size;
has_deferred.store(true, std::memory_order_release);
return;
}
}
deferred.push_back(DeferredUnmap{
.as_id = as_id,
.gpu_addr = gpu_addr,
.size = size,
});
}
has_deferred.store(true, std::memory_order_release);
}
void Clear() {
{
std::scoped_lock lock{deferred_mutex};
deferred.clear();
}
has_deferred.store(false, std::memory_order_release);
entries.clear();
}
private:
struct Entry {
size_t as_id{};
GPUVAddr gpu_addr{};
u32 size{};
VirtualSegments segments;
};
struct DeferredUnmap {
size_t as_id;
GPUVAddr gpu_addr;
u64 size;
};
static u64 MakeKey(size_t as_id, GPUVAddr gpu_addr) {
return (static_cast<u64>(as_id) << 48) ^ gpu_addr;
}
void ApplyDeferred() {
std::vector<DeferredUnmap> pending;
{
std::scoped_lock lock{deferred_mutex};
has_deferred.store(false, std::memory_order_release);
pending.swap(deferred);
}
if (pending.empty() || entries.empty()) {
return;
}
for (auto it = entries.begin(); it != entries.end();) {
const Entry& entry = it->second;
const GPUVAddr entry_end = entry.gpu_addr + entry.size;
bool overlaps = false;
for (const DeferredUnmap& unmap : pending) {
if (unmap.as_id != entry.as_id) {
continue;
}
if (entry.gpu_addr < unmap.gpu_addr + unmap.size && unmap.gpu_addr < entry_end) {
overlaps = true;
break;
}
}
if (overlaps) {
it = entries.erase(it);
} else {
++it;
}
}
}
std::unordered_map<u64, Entry> entries;
std::vector<DeferredUnmap> deferred;
std::mutex deferred_mutex;
std::atomic<bool> has_deferred{false};
};
} // namespace VideoCommon
-1
View File
@@ -8,7 +8,6 @@
#include <array>
#include <vector>
#include <type_traits>
#include "common/bit_field.h"
#include "common/common_funcs.h"
-7
View File
@@ -1,15 +1,8 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
#pragma once
#include <array>
#include <cstddef>
#include <type_traits>
#include "common/bit_field.h"
#include "common/common_funcs.h"
#include "common/common_types.h"
-1
View File
@@ -7,7 +7,6 @@
#pragma once
#include <memory>
#include <type_traits>
#include "common/common_types.h"
#include "common/scratch_buffer.h"
@@ -52,7 +52,7 @@ constexpr std::array PROGRAM_LUT{
Buffer::Buffer(BufferCacheRuntime&, VideoCommon::NullBufferParams null_params)
: VideoCommon::BufferBase(null_params) {}
Buffer::Buffer(BufferCacheRuntime& runtime, DAddr cpu_addr_, u64 size_bytes_, bool)
Buffer::Buffer(BufferCacheRuntime& runtime, DAddr cpu_addr_, u64 size_bytes_)
: VideoCommon::BufferBase(cpu_addr_, size_bytes_) {
buffer.Create();
if (runtime.device.HasDebuggingToolAttached()) {
@@ -23,8 +23,7 @@ class BufferCacheRuntime;
class Buffer : public VideoCommon::BufferBase {
public:
explicit Buffer(BufferCacheRuntime&, DAddr cpu_addr, u64 size_bytes,
bool sparse_compatible = false);
explicit Buffer(BufferCacheRuntime&, DAddr cpu_addr, u64 size_bytes);
explicit Buffer(BufferCacheRuntime&, VideoCommon::NullBufferParams);
void ImmediateUpload(size_t offset, std::span<const u8> data) noexcept;
@@ -56,8 +56,7 @@ size_t BytesPerIndex(VkIndexType index_type) {
}
}
vk::Buffer CreateBuffer(const Device& device, const MemoryAllocator& memory_allocator, u64 size,
VkDeviceSize sparse_alignment) {
vk::Buffer CreateBuffer(const Device& device, const MemoryAllocator& memory_allocator, u64 size) {
VkBufferUsageFlags flags =
VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT |
VK_BUFFER_USAGE_UNIFORM_TEXEL_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_TEXEL_BUFFER_BIT |
@@ -83,9 +82,6 @@ vk::Buffer CreateBuffer(const Device& device, const MemoryAllocator& memory_allo
.queueFamilyIndexCount = 0,
.pQueueFamilyIndices = nullptr,
};
if (sparse_alignment > 1) {
return memory_allocator.CreateBuffer(buffer_ci, MemoryUsage::DeviceLocal, sparse_alignment);
}
return memory_allocator.CreateBuffer(buffer_ci, MemoryUsage::DeviceLocal);
}
} // Anonymous namespace
@@ -103,14 +99,10 @@ Buffer::Buffer(BufferCacheRuntime& runtime, VideoCommon::NullBufferParams null_p
}
}
Buffer::Buffer(BufferCacheRuntime& runtime, DAddr cpu_addr_, u64 size_bytes_,
bool sparse_compatible_)
Buffer::Buffer(BufferCacheRuntime& runtime, DAddr cpu_addr_, u64 size_bytes_)
: VideoCommon::BufferBase(cpu_addr_, size_bytes_), device{&runtime.device},
scheduler{&runtime.scheduler},
buffer{CreateBuffer(*device, runtime.memory_allocator, SizeBytes(),
runtime.SparseAlignmentFor(sparse_compatible_))},
tracker{SizeBytes()} {
sparse_compatible = sparse_compatible_;
buffer{CreateBuffer(*device, runtime.memory_allocator, SizeBytes())}, tracker{SizeBytes()} {
if (runtime.device.HasDebuggingToolAttached()) {
buffer.SetObjectNameEXT(fmt::format("Buffer {:#x}", CpuAddr()).c_str());
}
@@ -356,8 +348,7 @@ BufferCacheRuntime::BufferCacheRuntime(const Device& device_, MemoryAllocator& m
: device{device_}, memory_allocator{memory_allocator_}, scheduler{scheduler_},
staging_pool{staging_pool_}, guest_descriptor_queue{guest_descriptor_queue_},
quad_index_pass(device, scheduler, descriptor_pool, staging_pool,
compute_pass_descriptor_queue),
multi_range_buffers(device_, memory_allocator_, scheduler_) {
compute_pass_descriptor_queue) {
const VkDriverIdKHR driver_id = device.GetDriverID();
limit_dynamic_storage_buffers = driver_id == VK_DRIVER_ID_QUALCOMM_PROPRIETARY ||
driver_id == VK_DRIVER_ID_ARM_PROPRIETARY;
@@ -545,33 +536,6 @@ void BufferCacheRuntime::ClearBuffer(VkBuffer dest_buffer, u32 offset, size_t si
});
}
bool BufferCacheRuntime::BindMultiRangeStorageBuffer(u64 key) {
if (multi_range_sources.empty() || multi_range_total == 0) {
return false;
}
const MultiRangeRef ref = multi_range_buffers.Get(key, multi_range_sources, multi_range_total);
if (ref.handle == VK_NULL_HANDLE) {
return false;
}
if (ref.needs_gather) {
PreCopyBarrier();
VkDeviceSize dst_offset = 0;
for (const MultiRangeSource& source : multi_range_sources) {
const std::array<VideoCommon::BufferCopy, 1> copy{VideoCommon::BufferCopy{
.src_offset = static_cast<u64>(source.offset),
.dst_offset = static_cast<u64>(dst_offset),
.size = static_cast<size_t>(source.size),
}};
CopyBuffer(ref.handle, source.handle, copy, false);
dst_offset += source.size;
}
PostCopyBarrier();
multi_range_buffers.MarkGathered(key);
}
guest_descriptor_queue.AddBuffer(ref.handle, ref.address, 0, ref.size);
return true;
}
void BufferCacheRuntime::BindIndexBuffer(PrimitiveTopology topology, IndexFormat index_format,
u32 base_vertex, u32 num_indices, VkBuffer buffer,
u32 offset, [[maybe_unused]] u32 size) {
@@ -8,14 +8,11 @@
#include <limits>
#include <boost/container/small_vector.hpp>
#include "video_core/buffer_cache/buffer_cache_base.h"
#include "video_core/buffer_cache/memory_tracker_base.h"
#include "video_core/buffer_cache/usage_tracker.h"
#include "video_core/engines/maxwell_3d.h"
#include "video_core/renderer_vulkan/vk_compute_pass.h"
#include "video_core/renderer_vulkan/vk_multi_range_buffer.h"
#include "video_core/renderer_vulkan/vk_staging_buffer_pool.h"
#include "video_core/renderer_vulkan/vk_update_descriptor.h"
#include "video_core/surface.h"
@@ -34,8 +31,7 @@ class BufferCacheRuntime;
class Buffer : public VideoCommon::BufferBase {
public:
explicit Buffer(BufferCacheRuntime&, VideoCommon::NullBufferParams null_params);
explicit Buffer(BufferCacheRuntime& runtime, VAddr cpu_addr_, u64 size_bytes_,
bool sparse_compatible_ = false);
explicit Buffer(BufferCacheRuntime& runtime, VAddr cpu_addr_, u64 size_bytes_);
[[nodiscard]] VkBufferView View(u32 offset, u32 size, VideoCore::Surface::PixelFormat format);
@@ -47,14 +43,6 @@ public:
return device_address;
}
[[nodiscard]] bool IsSparseCompatible() const noexcept {
return sparse_compatible;
}
[[nodiscard]] vk::MemoryLocation Location() const noexcept {
return buffer.Location();
}
[[nodiscard]] bool IsRegionUsed(u64 offset, u64 size) const noexcept {
return tracker.IsUsed(offset, size);
}
@@ -89,7 +77,6 @@ private:
VkDeviceAddress device_address{};
u64 last_usage_tick{};
bool is_null{};
bool sparse_compatible{};
};
class QuadArrayIndexBuffer;
@@ -138,7 +125,7 @@ public:
void PreCopyBarrier();
void CopyBuffer(VkBuffer dst_buffer, VkBuffer src_buffer,
void CopyBuffer(VkBuffer src_buffer, VkBuffer dst_buffer,
std::span<const VideoCommon::BufferCopy> copies, bool barrier,
bool can_reorder_upload = false);
@@ -168,45 +155,6 @@ public:
return ref.mapped_span;
}
[[nodiscard]] VkDeviceSize SparseAlignmentFor(bool sparse_compatible) const noexcept {
if (!sparse_compatible || !multi_range_buffers.UsesSparse()) {
return 0;
}
return multi_range_buffers.BlockSize();
}
[[nodiscard]] bool PrefersSparseSources() const noexcept {
return multi_range_buffers.UsesSparse();
}
void ResetMultiRange() noexcept {
multi_range_sources.clear();
multi_range_total = 0;
}
void PushMultiRangeSource(const Buffer& buffer, u32 offset, u32 size) {
const vk::MemoryLocation location = buffer.Location();
multi_range_sources.push_back(MultiRangeSource{
.handle = buffer.Handle(),
.memory = location.memory,
.memory_offset = location.offset,
.offset = offset,
.size = size,
.memory_type = location.memory_type,
});
multi_range_total += size;
}
bool BindMultiRangeStorageBuffer(u64 key);
void InvalidateMultiRange(u64 key) {
multi_range_buffers.Invalidate(key);
}
void OnBufferDeleted(const Buffer& buffer) {
multi_range_buffers.DropOwner(buffer.Handle());
}
void BindUniformBuffer(const Buffer& buffer, u32 offset, u32 size) {
BindBuffer(buffer, offset, size);
}
@@ -260,10 +208,6 @@ private:
std::unique_ptr<Uint8Pass> uint8_pass;
QuadIndexedPass quad_index_pass;
MultiRangeBufferCache multi_range_buffers;
boost::container::small_vector<MultiRangeSource, 16> multi_range_sources;
VkDeviceSize multi_range_total{};
bool limit_dynamic_storage_buffers = false;
u32 max_dynamic_storage_buffers = (std::numeric_limits<u32>::max)();
};
@@ -1,341 +0,0 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
#include <mutex>
#include "video_core/renderer_vulkan/vk_multi_range_buffer.h"
#include "video_core/renderer_vulkan/vk_scheduler.h"
#include "video_core/vulkan_common/vulkan_device.h"
namespace Vulkan {
MultiRangeBufferCache::MultiRangeBufferCache(const Device& device_,
MemoryAllocator& memory_allocator_,
Scheduler& scheduler_)
: device{device_}, memory_allocator{memory_allocator_}, scheduler{scheduler_} {
sparse_usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT |
VK_BUFFER_USAGE_STORAGE_BUFFER_BIT;
if (device.IsBufferDeviceAddressSupported()) {
sparse_usage |= VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT;
}
if (!device.IsSparseBindingSupported()) {
return;
}
u32 memory_type_bits = 0;
const VkDeviceSize queried = QueryBlockSize(memory_type_bits);
if (queried == 0 || memory_type_bits == 0) {
return;
}
block_size = queried;
sparse_memory_type_bits = memory_type_bits;
use_sparse = true;
}
MultiRangeBufferCache::~MultiRangeBufferCache() {
const VkDevice logical = *device.GetLogical();
const auto& dld = device.GetDispatchLoader();
for (auto& [key, entry] : entries) {
if (entry.sparse_handle != VK_NULL_HANDLE) {
dld.vkDestroyBuffer(logical, entry.sparse_handle, nullptr);
}
}
entries.clear();
for (const Retired& item : retired) {
dld.vkDestroyBuffer(logical, item.handle, nullptr);
}
retired.clear();
}
VkDeviceSize MultiRangeBufferCache::QueryBlockSize(u32& memory_type_bits) const {
const VkDevice logical = *device.GetLogical();
const auto& dld = device.GetDispatchLoader();
const VkBufferCreateInfo probe_ci{
.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
.pNext = nullptr,
.flags = VK_BUFFER_CREATE_SPARSE_BINDING_BIT | VK_BUFFER_CREATE_SPARSE_ALIASED_BIT,
.size = DEFAULT_BLOCK_SIZE,
.usage = sparse_usage,
.sharingMode = VK_SHARING_MODE_EXCLUSIVE,
.queueFamilyIndexCount = 0,
.pQueueFamilyIndices = nullptr,
};
VkBuffer probe{};
if (dld.vkCreateBuffer(logical, &probe_ci, nullptr, &probe) != VK_SUCCESS) {
return 0;
}
const VkBufferMemoryRequirementsInfo2 reqs_info{
.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_REQUIREMENTS_INFO_2,
.pNext = nullptr,
.buffer = probe,
};
VkMemoryRequirements2 reqs2{
.sType = VK_STRUCTURE_TYPE_MEMORY_REQUIREMENTS_2,
.pNext = nullptr,
.memoryRequirements = {},
};
dld.vkGetBufferMemoryRequirements2(logical, &reqs_info, &reqs2);
dld.vkDestroyBuffer(logical, probe, nullptr);
memory_type_bits = reqs2.memoryRequirements.memoryTypeBits;
return reqs2.memoryRequirements.alignment;
}
u64 MultiRangeBufferCache::HashSources(std::span<const MultiRangeSource> sources) const {
u64 hash = 0xcbf29ce484222325ULL;
const auto mix = [&hash](u64 value) {
hash ^= value;
hash *= 0x100000001b3ULL;
};
for (const MultiRangeSource& source : sources) {
mix(reinterpret_cast<u64>(source.handle));
mix(static_cast<u64>(source.offset));
mix(static_cast<u64>(source.size));
}
return hash;
}
bool MultiRangeBufferCache::CanBindSparse(std::span<const MultiRangeSource> sources) const {
if (!UsesSparse()) {
return false;
}
for (const MultiRangeSource& source : sources) {
if (source.memory == VK_NULL_HANDLE) {
return false;
}
if (source.memory_type >= 32) {
return false;
}
if (((sparse_memory_type_bits >> source.memory_type) & 1) == 0) {
return false;
}
const VkDeviceSize memory_offset = source.memory_offset + source.offset;
if ((memory_offset % block_size) != 0) {
return false;
}
if ((source.size % block_size) != 0) {
return false;
}
}
return true;
}
VkBuffer MultiRangeBufferCache::CreateSparse(std::span<const MultiRangeSource> sources,
VkDeviceSize total) {
const VkDevice logical = *device.GetLogical();
const auto& dld = device.GetDispatchLoader();
const VkBufferCreateInfo buffer_ci{
.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
.pNext = nullptr,
.flags = VK_BUFFER_CREATE_SPARSE_BINDING_BIT | VK_BUFFER_CREATE_SPARSE_ALIASED_BIT,
.size = total,
.usage = sparse_usage,
.sharingMode = VK_SHARING_MODE_EXCLUSIVE,
.queueFamilyIndexCount = 0,
.pQueueFamilyIndices = nullptr,
};
VkBuffer handle{};
if (dld.vkCreateBuffer(logical, &buffer_ci, nullptr, &handle) != VK_SUCCESS) {
return VK_NULL_HANDLE;
}
std::vector<VkSparseMemoryBind> binds;
binds.reserve(sources.size());
VkDeviceSize resource_offset = 0;
for (const MultiRangeSource& source : sources) {
binds.push_back(VkSparseMemoryBind{
.resourceOffset = resource_offset,
.size = source.size,
.memory = source.memory,
.memoryOffset = source.memory_offset + source.offset,
.flags = 0,
});
resource_offset += source.size;
}
const VkSparseBufferMemoryBindInfo buffer_bind{
.buffer = handle,
.bindCount = static_cast<u32>(binds.size()),
.pBinds = binds.data(),
};
const VkBindSparseInfo bind_info{
.sType = VK_STRUCTURE_TYPE_BIND_SPARSE_INFO,
.pNext = nullptr,
.waitSemaphoreCount = 0,
.pWaitSemaphores = nullptr,
.bufferBindCount = 1,
.pBufferBinds = &buffer_bind,
.imageOpaqueBindCount = 0,
.pImageOpaqueBinds = nullptr,
.imageBindCount = 0,
.pImageBinds = nullptr,
.signalSemaphoreCount = 0,
.pSignalSemaphores = nullptr,
};
const VkFenceCreateInfo fence_ci{
.sType = VK_STRUCTURE_TYPE_FENCE_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
};
vk::Fence fence = device.GetLogical().CreateFence(fence_ci);
VkResult bind_result = VK_ERROR_UNKNOWN;
{
std::scoped_lock lock{scheduler.submit_mutex};
bind_result = device.GetGraphicsQueue().BindSparse(bind_info, *fence);
}
if (bind_result != VK_SUCCESS) {
dld.vkDestroyBuffer(logical, handle, nullptr);
return VK_NULL_HANDLE;
}
fence.Wait();
return handle;
}
void MultiRangeBufferCache::DestroySparse(VkBuffer handle) {
if (handle == VK_NULL_HANDLE) {
return;
}
retired.push_back(Retired{
.handle = handle,
.tick = scheduler.CurrentTick(),
});
}
void MultiRangeBufferCache::DrainRetired() {
const VkDevice logical = *device.GetLogical();
const auto& dld = device.GetDispatchLoader();
size_t index = 0;
while (index < retired.size()) {
if (scheduler.IsFree(retired[index].tick)) {
dld.vkDestroyBuffer(logical, retired[index].handle, nullptr);
retired[index] = retired.back();
retired.pop_back();
} else {
++index;
}
}
}
MultiRangeRef MultiRangeBufferCache::Get(u64 key, std::span<const MultiRangeSource> sources,
VkDeviceSize total) {
if (sources.empty() || total == 0) {
return MultiRangeRef{};
}
if (!retired.empty()) {
DrainRetired();
}
const u64 geometry = HashSources(sources);
const auto it = entries.find(key);
if (it != entries.end() && it->second.geometry == geometry && it->second.size == total) {
Entry& entry = it->second;
MultiRangeRef ref{
.handle = entry.sparse_handle,
.address = entry.address,
.size = entry.size,
.needs_gather = false,
};
if (entry.sparse_handle == VK_NULL_HANDLE) {
ref.handle = *entry.gathered;
ref.needs_gather = entry.dirty;
}
return ref;
}
if (it != entries.end()) {
DestroySparse(it->second.sparse_handle);
entries.erase(it);
}
Entry entry;
entry.geometry = geometry;
entry.size = total;
if (CanBindSparse(sources)) {
entry.sparse_handle = CreateSparse(sources, total);
if (entry.sparse_handle != VK_NULL_HANDLE) {
entry.owners.reserve(sources.size());
for (const MultiRangeSource& source : sources) {
entry.owners.push_back(source.handle);
}
}
}
if (entry.sparse_handle == VK_NULL_HANDLE) {
VkBufferUsageFlags flags = VK_BUFFER_USAGE_TRANSFER_SRC_BIT |
VK_BUFFER_USAGE_TRANSFER_DST_BIT |
VK_BUFFER_USAGE_STORAGE_BUFFER_BIT;
if (device.IsBufferDeviceAddressSupported()) {
flags |= VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT;
}
const VkBufferCreateInfo gather_ci{
.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.size = total,
.usage = flags,
.sharingMode = VK_SHARING_MODE_EXCLUSIVE,
.queueFamilyIndexCount = 0,
.pQueueFamilyIndices = nullptr,
};
entry.gathered = memory_allocator.CreateBuffer(gather_ci, MemoryUsage::DeviceLocal);
entry.dirty = true;
}
if (device.IsBufferDeviceAddressSupported()) {
VkBuffer address_handle = entry.sparse_handle;
if (address_handle == VK_NULL_HANDLE) {
address_handle = *entry.gathered;
}
entry.address = device.GetLogical().GetBufferDeviceAddress(address_handle);
}
MultiRangeRef ref{
.handle = entry.sparse_handle,
.address = entry.address,
.size = entry.size,
.needs_gather = false,
};
if (entry.sparse_handle == VK_NULL_HANDLE) {
ref.handle = *entry.gathered;
ref.needs_gather = true;
}
entries.emplace(key, std::move(entry));
return ref;
}
void MultiRangeBufferCache::MarkGathered(u64 key) {
const auto it = entries.find(key);
if (it != entries.end()) {
it->second.dirty = false;
}
}
void MultiRangeBufferCache::DropOwner(VkBuffer owner) {
if (owner == VK_NULL_HANDLE) {
return;
}
for (auto it = entries.begin(); it != entries.end();) {
Entry& entry = it->second;
bool owned = false;
for (const VkBuffer handle : entry.owners) {
if (handle == owner) {
owned = true;
break;
}
}
if (owned) {
DestroySparse(entry.sparse_handle);
it = entries.erase(it);
} else {
++it;
}
}
}
void MultiRangeBufferCache::Invalidate(u64 key) {
const auto it = entries.find(key);
if (it != entries.end()) {
it->second.dirty = true;
}
}
void MultiRangeBufferCache::Clear() {
for (auto& [key, entry] : entries) {
DestroySparse(entry.sparse_handle);
}
entries.clear();
}
} // namespace Vulkan
@@ -1,105 +0,0 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
#pragma once
#include <span>
#include <unordered_map>
#include <vector>
#include "common/common_types.h"
#include "video_core/vulkan_common/vulkan_memory_allocator.h"
#include "video_core/vulkan_common/vulkan_wrapper.h"
namespace Vulkan {
class Device;
class Scheduler;
struct MultiRangeSource {
VkBuffer handle{};
VkDeviceMemory memory{};
VkDeviceSize memory_offset{};
VkDeviceSize offset{};
VkDeviceSize size{};
u32 memory_type{};
};
struct MultiRangeRef {
VkBuffer handle{};
VkDeviceAddress address{};
VkDeviceSize size{};
bool needs_gather{};
};
class MultiRangeBufferCache final {
public:
static constexpr VkDeviceSize DEFAULT_BLOCK_SIZE = 64 * 1024;
explicit MultiRangeBufferCache(const Device& device_, MemoryAllocator& memory_allocator_,
Scheduler& scheduler_);
~MultiRangeBufferCache();
MultiRangeBufferCache(const MultiRangeBufferCache&) = delete;
MultiRangeBufferCache& operator=(const MultiRangeBufferCache&) = delete;
[[nodiscard]] bool UsesSparse() const noexcept {
return use_sparse;
}
[[nodiscard]] VkDeviceSize BlockSize() const noexcept {
return block_size;
}
[[nodiscard]] MultiRangeRef Get(u64 key, std::span<const MultiRangeSource> sources,
VkDeviceSize total);
void MarkGathered(u64 key);
void Invalidate(u64 key);
void DropOwner(VkBuffer owner);
void Clear();
private:
struct Retired {
VkBuffer handle{};
u64 tick{};
};
struct Entry {
vk::Buffer gathered;
VkBuffer sparse_handle{};
VkDeviceAddress address{};
VkDeviceSize size{};
u64 geometry{};
bool dirty{true};
std::vector<VkBuffer> owners;
};
[[nodiscard]] u64 HashSources(std::span<const MultiRangeSource> sources) const;
[[nodiscard]] bool CanBindSparse(std::span<const MultiRangeSource> sources) const;
[[nodiscard]] VkBuffer CreateSparse(std::span<const MultiRangeSource> sources,
VkDeviceSize total);
[[nodiscard]] VkDeviceSize QueryBlockSize(u32& memory_type_bits) const;
void DestroySparse(VkBuffer handle);
void DrainRetired();
const Device& device;
MemoryAllocator& memory_allocator;
Scheduler& scheduler;
bool use_sparse{};
VkDeviceSize block_size{DEFAULT_BLOCK_SIZE};
u32 sparse_memory_type_bits{};
VkBufferUsageFlags sparse_usage{};
std::unordered_map<u64, Entry> entries;
std::vector<Retired> retired;
};
} // namespace Vulkan
@@ -338,7 +338,7 @@ void PresentManager::PresentThread(std::stop_token token) {
// By exchanging the lock ownership we take the swapchain lock
// before the queue lock goes out of scope. This way the swapchain
// lock in WaitPresent is guaranteed to occur after here.
void(std::exchange(lock, std::unique_lock{swapchain_mutex}));
std::exchange(lock, std::unique_lock{swapchain_mutex});
CopyToSwapchain(frame);
// Free the frame for reuse
@@ -819,7 +819,6 @@ void RasterizerVulkan::ModifyGPUMemory(size_t as_id, GPUVAddr addr, u64 size) {
std::scoped_lock lock{texture_cache.mutex};
texture_cache.UnmapGPUMemory(as_id, addr, size);
}
buffer_cache.UnmapGPUMemory(as_id, addr, size);
}
void RasterizerVulkan::SignalFence(std::function<void()>&& func) {
@@ -300,7 +300,7 @@ void Scheduler::WorkerThread(std::stop_token stop_token) {
// Exchange lock ownership so that we take the execution lock before
// the queue lock goes out of scope. This allows us to force execution
// to complete in the next step.
void(std::exchange(lk, std::unique_lock{execution_mutex}));
std::exchange(lk, std::unique_lock{execution_mutex});
// Perform the work, tracking whether the chunk was a submission
// before executing.
@@ -1570,8 +1570,6 @@ void Device::SetupFamilies(VkSurfaceKHR surface) {
}
if (graphics) {
graphics_family = *graphics;
graphics_family_sparse_binding =
(queue_family_properties[*graphics].queueFlags & VK_QUEUE_SPARSE_BINDING_BIT) != 0;
}
if (present) {
present_family = *present;
@@ -317,10 +317,6 @@ public:
return properties.driver.driverID;
}
bool IsSparseBindingSupported() const {
return features.features.sparseBinding && graphics_family_sparse_binding;
}
/// Returns true for tile-based deferred renderers.
bool IsTiler() const {
switch (GetDriverID()) {
@@ -1150,7 +1146,6 @@ private:
bool owns_static_pipeline_cache{};
u32 instance_version{}; ///< Vulkan instance version.
u32 graphics_family{}; ///< Main graphics queue family index.
bool graphics_family_sparse_binding{};
u32 present_family{}; ///< Main present queue family index.
struct Extensions {
@@ -26,374 +26,350 @@
#include "common/settings.h"
namespace Vulkan {
namespace {
namespace {
// Helpers translating MemoryUsage to flags/usage
[[maybe_unused]] VkMemoryPropertyFlags MemoryUsagePropertyFlags(MemoryUsage usage) {
switch (usage) {
case MemoryUsage::DeviceLocal:
return VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT;
case MemoryUsage::Upload:
return VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT |
VK_MEMORY_PROPERTY_HOST_COHERENT_BIT;
case MemoryUsage::Download:
return VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT |
VK_MEMORY_PROPERTY_HOST_COHERENT_BIT |
VK_MEMORY_PROPERTY_HOST_CACHED_BIT;
case MemoryUsage::Stream:
return VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT |
VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT |
VK_MEMORY_PROPERTY_HOST_COHERENT_BIT;
[[maybe_unused]] VkMemoryPropertyFlags MemoryUsagePropertyFlags(MemoryUsage usage) {
switch (usage) {
case MemoryUsage::DeviceLocal:
return VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT;
case MemoryUsage::Upload:
return VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT |
VK_MEMORY_PROPERTY_HOST_COHERENT_BIT;
case MemoryUsage::Download:
return VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT |
VK_MEMORY_PROPERTY_HOST_COHERENT_BIT |
VK_MEMORY_PROPERTY_HOST_CACHED_BIT;
case MemoryUsage::Stream:
return VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT |
VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT |
VK_MEMORY_PROPERTY_HOST_COHERENT_BIT;
}
ASSERT_MSG(false, "Invalid memory usage={}", usage);
return VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT;
}
ASSERT_MSG(false, "Invalid memory usage={}", usage);
return VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT;
}
[[nodiscard]] VkMemoryPropertyFlags MemoryUsagePreferredVmaFlags(MemoryUsage usage) {
if (usage == MemoryUsage::Download) {
return VK_MEMORY_PROPERTY_HOST_CACHED_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT;
[[nodiscard]] VkMemoryPropertyFlags MemoryUsagePreferredVmaFlags(MemoryUsage usage) {
if (usage == MemoryUsage::Download) {
return VK_MEMORY_PROPERTY_HOST_CACHED_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT;
}
return usage != MemoryUsage::DeviceLocal ? VK_MEMORY_PROPERTY_HOST_COHERENT_BIT
: VkMemoryPropertyFlagBits{};
}
return usage != MemoryUsage::DeviceLocal ? VK_MEMORY_PROPERTY_HOST_COHERENT_BIT
: VkMemoryPropertyFlagBits{};
}
[[nodiscard]] VmaAllocationCreateFlags MemoryUsageVmaFlags(MemoryUsage usage) {
switch (usage) {
case MemoryUsage::Upload:
case MemoryUsage::Stream:
return VMA_ALLOCATION_CREATE_MAPPED_BIT |
VMA_ALLOCATION_CREATE_HOST_ACCESS_SEQUENTIAL_WRITE_BIT;
case MemoryUsage::Download:
return VMA_ALLOCATION_CREATE_MAPPED_BIT |
VMA_ALLOCATION_CREATE_HOST_ACCESS_RANDOM_BIT;
case MemoryUsage::DeviceLocal:
return {};
[[nodiscard]] VmaAllocationCreateFlags MemoryUsageVmaFlags(MemoryUsage usage) {
switch (usage) {
case MemoryUsage::Upload:
case MemoryUsage::Stream:
return VMA_ALLOCATION_CREATE_MAPPED_BIT |
VMA_ALLOCATION_CREATE_HOST_ACCESS_SEQUENTIAL_WRITE_BIT;
case MemoryUsage::Download:
return VMA_ALLOCATION_CREATE_MAPPED_BIT |
VMA_ALLOCATION_CREATE_HOST_ACCESS_RANDOM_BIT;
case MemoryUsage::DeviceLocal:
return {};
}
return {};
}
return {};
}
[[nodiscard]] VmaMemoryUsage MemoryUsageVma(MemoryUsage usage) {
switch (usage) {
case MemoryUsage::DeviceLocal:
case MemoryUsage::Stream:
return VMA_MEMORY_USAGE_AUTO_PREFER_DEVICE;
case MemoryUsage::Upload:
case MemoryUsage::Download:
return VMA_MEMORY_USAGE_AUTO_PREFER_HOST;
[[nodiscard]] VmaMemoryUsage MemoryUsageVma(MemoryUsage usage) {
switch (usage) {
case MemoryUsage::DeviceLocal:
case MemoryUsage::Stream:
return VMA_MEMORY_USAGE_AUTO_PREFER_DEVICE;
case MemoryUsage::Upload:
case MemoryUsage::Download:
return VMA_MEMORY_USAGE_AUTO_PREFER_HOST;
}
return VMA_MEMORY_USAGE_AUTO_PREFER_DEVICE;
}
return VMA_MEMORY_USAGE_AUTO_PREFER_DEVICE;
}
} // namespace
// This avoids calling vkGetBufferMemoryRequirements* directly.
template<typename T>
static VkBuffer GetVkHandleFromBuffer(const T &buf) {
if constexpr (requires { static_cast<VkBuffer>(buf); }) {
return static_cast<VkBuffer>(buf);
} else if constexpr (requires {{ buf.GetHandle() } -> std::convertible_to<VkBuffer>; }) {
return buf.GetHandle();
} else if constexpr (requires {{ buf.Handle() } -> std::convertible_to<VkBuffer>; }) {
return buf.Handle();
} else if constexpr (requires {{ buf.vk_handle() } -> std::convertible_to<VkBuffer>; }) {
return buf.vk_handle();
} else {
static_assert(sizeof(T) == 0, "Cannot extract VkBuffer handle from vk::Buffer");
return VK_NULL_HANDLE;
}
}
} // namespace
//MemoryCommit is now VMA-backed
MemoryCommit::MemoryCommit(VmaAllocator alloc, VmaAllocation a,
const VmaAllocationInfo &info) noexcept
: allocator{alloc}, allocation{a}, memory{info.deviceMemory},
offset{info.offset}, size{info.size}, mapped_ptr{info.pMappedData} {
// Log GPU memory allocation
if (GPU::Logging::IsActive() &&
Settings::values.gpu_log_memory_tracking.GetValue()) {
GPU::Logging::GPULogger::GetInstance().LogMemoryAllocation(
reinterpret_cast<uintptr_t>(memory),
static_cast<u64>(size),
0 // Memory property flags (not easily available from VMA)
);
}
}
MemoryCommit::~MemoryCommit() { Release(); }
MemoryCommit::MemoryCommit(MemoryCommit &&rhs) noexcept
: allocator{std::exchange(rhs.allocator, nullptr)},
allocation{std::exchange(rhs.allocation, nullptr)},
memory{std::exchange(rhs.memory, VK_NULL_HANDLE)},
offset{std::exchange(rhs.offset, 0)},
size{std::exchange(rhs.size, 0)},
mapped_ptr{std::exchange(rhs.mapped_ptr, nullptr)} {}
MemoryCommit &MemoryCommit::operator=(MemoryCommit &&rhs) noexcept {
if (this != &rhs) {
Release();
allocator = std::exchange(rhs.allocator, nullptr);
allocation = std::exchange(rhs.allocation, nullptr);
memory = std::exchange(rhs.memory, VK_NULL_HANDLE);
offset = std::exchange(rhs.offset, 0);
size = std::exchange(rhs.size, 0);
mapped_ptr = std::exchange(rhs.mapped_ptr, nullptr);
}
return *this;
}
std::span<u8> MemoryCommit::Map()
{
if (!allocation) return {};
if (!mapped_ptr) {
if (vmaMapMemory(allocator, allocation, &mapped_ptr) != VK_SUCCESS) return {};
}
const size_t n = static_cast<size_t>(std::min<VkDeviceSize>(size,
(std::numeric_limits<size_t>::max)()));
return std::span<u8>{static_cast<u8 *>(mapped_ptr), n};
}
std::span<const u8> MemoryCommit::Map() const
{
if (!allocation) return {};
if (!mapped_ptr) {
void *p = nullptr;
if (vmaMapMemory(allocator, allocation, &p) != VK_SUCCESS) return {};
const_cast<MemoryCommit *>(this)->mapped_ptr = p;
}
const size_t n = static_cast<size_t>(std::min<VkDeviceSize>(size,
(std::numeric_limits<size_t>::max)()));
return std::span<const u8>{static_cast<const u8 *>(mapped_ptr), n};
}
void MemoryCommit::Unmap()
{
if (allocation && mapped_ptr) {
vmaUnmapMemory(allocator, allocation);
mapped_ptr = nullptr;
}
}
void MemoryCommit::Release() {
if (allocation && allocator) {
// Log GPU memory deallocation
MemoryCommit::MemoryCommit(VmaAllocator alloc, VmaAllocation a,
const VmaAllocationInfo &info) noexcept
: allocator{alloc}, allocation{a}, memory{info.deviceMemory},
offset{info.offset}, size{info.size}, mapped_ptr{info.pMappedData} {
// Log GPU memory allocation
if (GPU::Logging::IsActive() &&
Settings::values.gpu_log_memory_tracking.GetValue() &&
memory != VK_NULL_HANDLE) {
GPU::Logging::GPULogger::GetInstance().LogMemoryDeallocation(
reinterpret_cast<uintptr_t>(memory)
Settings::values.gpu_log_memory_tracking.GetValue()) {
GPU::Logging::GPULogger::GetInstance().LogMemoryAllocation(
reinterpret_cast<uintptr_t>(memory),
static_cast<u64>(size),
0 // Memory property flags (not easily available from VMA)
);
}
}
if (mapped_ptr) {
MemoryCommit::~MemoryCommit() { Release(); }
MemoryCommit::MemoryCommit(MemoryCommit &&rhs) noexcept
: allocator{std::exchange(rhs.allocator, nullptr)},
allocation{std::exchange(rhs.allocation, nullptr)},
memory{std::exchange(rhs.memory, VK_NULL_HANDLE)},
offset{std::exchange(rhs.offset, 0)},
size{std::exchange(rhs.size, 0)},
mapped_ptr{std::exchange(rhs.mapped_ptr, nullptr)} {}
MemoryCommit &MemoryCommit::operator=(MemoryCommit &&rhs) noexcept {
if (this != &rhs) {
Release();
allocator = std::exchange(rhs.allocator, nullptr);
allocation = std::exchange(rhs.allocation, nullptr);
memory = std::exchange(rhs.memory, VK_NULL_HANDLE);
offset = std::exchange(rhs.offset, 0);
size = std::exchange(rhs.size, 0);
mapped_ptr = std::exchange(rhs.mapped_ptr, nullptr);
}
return *this;
}
std::span<u8> MemoryCommit::Map()
{
if (!allocation) return {};
if (!mapped_ptr) {
if (vmaMapMemory(allocator, allocation, &mapped_ptr) != VK_SUCCESS) return {};
}
const size_t n = static_cast<size_t>(std::min<VkDeviceSize>(size,
(std::numeric_limits<size_t>::max)()));
return std::span<u8>{static_cast<u8 *>(mapped_ptr), n};
}
std::span<const u8> MemoryCommit::Map() const
{
if (!allocation) return {};
if (!mapped_ptr) {
void *p = nullptr;
if (vmaMapMemory(allocator, allocation, &p) != VK_SUCCESS) return {};
const_cast<MemoryCommit *>(this)->mapped_ptr = p;
}
const size_t n = static_cast<size_t>(std::min<VkDeviceSize>(size,
(std::numeric_limits<size_t>::max)()));
return std::span<const u8>{static_cast<const u8 *>(mapped_ptr), n};
}
void MemoryCommit::Unmap()
{
if (allocation && mapped_ptr) {
vmaUnmapMemory(allocator, allocation);
mapped_ptr = nullptr;
}
vmaFreeMemory(allocator, allocation);
}
allocation = nullptr;
allocator = nullptr;
memory = VK_NULL_HANDLE;
offset = 0;
size = 0;
}
MemoryAllocator::MemoryAllocator(const Device &device_)
: device{device_}, allocator{device.GetAllocator()},
properties{device_.GetPhysical().GetMemoryProperties().memoryProperties},
buffer_image_granularity{
device_.GetPhysical().GetProperties().limits.bufferImageGranularity} {
void MemoryCommit::Release() {
if (allocation && allocator) {
// Log GPU memory deallocation
if (GPU::Logging::IsActive() &&
Settings::values.gpu_log_memory_tracking.GetValue() &&
memory != VK_NULL_HANDLE) {
GPU::Logging::GPULogger::GetInstance().LogMemoryDeallocation(
reinterpret_cast<uintptr_t>(memory)
);
}
// Preserve the previous "RenderDoc small heap" trimming behavior that we had in original vma minus the heap bug
if (device.HasDebuggingToolAttached())
{
using namespace Common::Literals;
ForEachDeviceLocalHostVisibleHeap(device, [this](size_t heap_idx, VkMemoryHeap &heap) {
if (heap.size <= 256_MiB) {
for (u32 t = 0; t < properties.memoryTypeCount; ++t) {
if (properties.memoryTypes[t].heapIndex == heap_idx) {
valid_memory_types &= ~(1u << t);
if (mapped_ptr) {
vmaUnmapMemory(allocator, allocation);
mapped_ptr = nullptr;
}
vmaFreeMemory(allocator, allocation);
}
allocation = nullptr;
allocator = nullptr;
memory = VK_NULL_HANDLE;
offset = 0;
size = 0;
}
MemoryAllocator::MemoryAllocator(const Device &device_)
: device{device_}, allocator{device.GetAllocator()},
properties{device_.GetPhysical().GetMemoryProperties().memoryProperties},
buffer_image_granularity{
device_.GetPhysical().GetProperties().limits.bufferImageGranularity} {
// Preserve the previous "RenderDoc small heap" trimming behavior that we had in original vma minus the heap bug
if (device.HasDebuggingToolAttached())
{
using namespace Common::Literals;
ForEachDeviceLocalHostVisibleHeap(device, [this](size_t heap_idx, VkMemoryHeap &heap) {
if (heap.size <= 256_MiB) {
for (u32 t = 0; t < properties.memoryTypeCount; ++t) {
if (properties.memoryTypes[t].heapIndex == heap_idx) {
valid_memory_types &= ~(1u << t);
}
}
}
}
});
});
}
}
}
MemoryAllocator::~MemoryAllocator() = default;
MemoryAllocator::~MemoryAllocator() = default;
vk::Image MemoryAllocator::CreateImage(const VkImageCreateInfo &ci) const
{
const VmaAllocationCreateInfo alloc_ci = {
.flags = VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT,
.usage = VMA_MEMORY_USAGE_AUTO_PREFER_DEVICE,
vk::Image MemoryAllocator::CreateImage(const VkImageCreateInfo &ci) const
{
const VmaAllocationCreateInfo alloc_ci = {
.flags = VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT,
.usage = VMA_MEMORY_USAGE_AUTO_PREFER_DEVICE,
.requiredFlags = 0,
.preferredFlags = VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT,
.memoryTypeBits = 0,
.pool = VK_NULL_HANDLE,
.pUserData = nullptr,
.priority = 0.f,
};
VkImage handle{};
VmaAllocation allocation{};
VmaAllocationInfo alloc_info{};
vk::Check(vmaCreateImage(allocator, &ci, &alloc_ci, &handle, &allocation, &alloc_info));
// Log GPU memory allocation for images
if (GPU::Logging::IsActive() &&
Settings::values.gpu_log_memory_tracking.GetValue()) {
GPU::Logging::GPULogger::GetInstance().LogMemoryAllocation(
reinterpret_cast<uintptr_t>(alloc_info.deviceMemory),
static_cast<u64>(alloc_info.size),
VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT
);
}
return vk::Image(handle, ci.usage, *device.GetLogical(), allocator, allocation,
device.GetDispatchLoader());
}
vk::Buffer MemoryAllocator::CreateBuffer(const VkBufferCreateInfo &ci, MemoryUsage usage) const {
// MESA will do memcpy() if not marked as host cached, so just force mark it for most buffers
auto const anv_flags = (usage == MemoryUsage::Stream
&& device.GetDriverID() == VK_DRIVER_ID_INTEL_OPEN_SOURCE_MESA)
? VK_MEMORY_PROPERTY_HOST_CACHED_BIT : 0;
const VmaAllocationCreateInfo alloc_ci = {
.flags = VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT | MemoryUsageVmaFlags(usage),
.usage = MemoryUsageVma(usage),
.requiredFlags = 0,
.preferredFlags = VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT,
.memoryTypeBits = 0,
.preferredFlags = MemoryUsagePreferredVmaFlags(usage) | anv_flags,
.memoryTypeBits = usage == MemoryUsage::Stream ? 0u : valid_memory_types,
.pool = VK_NULL_HANDLE,
.pUserData = nullptr,
.priority = 0.f,
};
};
VkImage handle{};
VmaAllocation allocation{};
VmaAllocationInfo alloc_info{};
vk::Check(vmaCreateImage(allocator, &ci, &alloc_ci, &handle, &allocation, &alloc_info));
VkBuffer handle{};
VmaAllocationInfo alloc_info{};
VmaAllocation allocation{};
VkMemoryPropertyFlags property_flags{};
// Log GPU memory allocation for images
if (GPU::Logging::IsActive() &&
Settings::values.gpu_log_memory_tracking.GetValue()) {
GPU::Logging::GPULogger::GetInstance().LogMemoryAllocation(
reinterpret_cast<uintptr_t>(alloc_info.deviceMemory),
static_cast<u64>(alloc_info.size),
VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT
);
}
vk::Check(vmaCreateBuffer(allocator, &ci, &alloc_ci, &handle, &allocation, &alloc_info));
vmaGetAllocationMemoryProperties(allocator, allocation, &property_flags);
return vk::Image(handle, ci.usage, *device.GetLogical(), allocator, allocation,
device.GetDispatchLoader());
}
vk::Buffer MemoryAllocator::CreateBuffer(const VkBufferCreateInfo &ci, MemoryUsage usage) const {
// MESA will do memcpy() if not marked as host cached, so just force mark it for most buffers
auto const anv_flags = (usage == MemoryUsage::Stream
&& device.GetDriverID() == VK_DRIVER_ID_INTEL_OPEN_SOURCE_MESA)
? VK_MEMORY_PROPERTY_HOST_CACHED_BIT : 0;
const VmaAllocationCreateInfo alloc_ci = {
.flags = VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT | MemoryUsageVmaFlags(usage),
.usage = MemoryUsageVma(usage),
.requiredFlags = 0,
.preferredFlags = MemoryUsagePreferredVmaFlags(usage) | anv_flags,
.memoryTypeBits = usage == MemoryUsage::Stream ? 0u : valid_memory_types,
.pool = VK_NULL_HANDLE,
.pUserData = nullptr,
.priority = 0.f,
};
VkBuffer handle{};
VmaAllocationInfo alloc_info{};
VmaAllocation allocation{};
VkMemoryPropertyFlags property_flags{};
vk::Check(vmaCreateBuffer(allocator, &ci, &alloc_ci, &handle, &allocation, &alloc_info));
vmaGetAllocationMemoryProperties(allocator, allocation, &property_flags);
// Log GPU memory allocation for buffers
if (GPU::Logging::IsActive() &&
Settings::values.gpu_log_memory_tracking.GetValue()) {
GPU::Logging::GPULogger::GetInstance().LogMemoryAllocation(
reinterpret_cast<uintptr_t>(alloc_info.deviceMemory),
static_cast<u64>(alloc_info.size),
property_flags
);
}
u8 *data = reinterpret_cast<u8 *>(alloc_info.pMappedData);
const std::span<u8> mapped_data = data ? std::span<u8>{data, ci.size} : std::span<u8>{};
const bool is_coherent = (property_flags & VK_MEMORY_PROPERTY_HOST_COHERENT_BIT) != 0;
return vk::Buffer(handle, *device.GetLogical(), allocator, allocation, mapped_data,
is_coherent,
device.GetDispatchLoader());
}
vk::Buffer MemoryAllocator::CreateBuffer(const VkBufferCreateInfo &ci, MemoryUsage usage,
VkDeviceSize min_alignment) const {
if (min_alignment <= 1) {
return CreateBuffer(ci, usage);
}
VkMemoryPropertyFlags anv_flags = 0;
if (usage == MemoryUsage::Stream &&
device.GetDriverID() == VK_DRIVER_ID_INTEL_OPEN_SOURCE_MESA) {
anv_flags = VK_MEMORY_PROPERTY_HOST_CACHED_BIT;
}
u32 memory_type_bits = valid_memory_types;
if (usage == MemoryUsage::Stream) {
memory_type_bits = 0u;
}
const VmaAllocationCreateInfo alloc_ci = {
.flags = VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT | MemoryUsageVmaFlags(usage),
.usage = MemoryUsageVma(usage),
.requiredFlags = 0,
.preferredFlags = MemoryUsagePreferredVmaFlags(usage) | anv_flags,
.memoryTypeBits = memory_type_bits,
.pool = VK_NULL_HANDLE,
.pUserData = nullptr,
.priority = 0.f,
};
VkBuffer handle{};
VmaAllocationInfo alloc_info{};
VmaAllocation allocation{};
VkMemoryPropertyFlags property_flags{};
vk::Check(vmaCreateBufferWithAlignment(allocator, &ci, &alloc_ci, min_alignment, &handle,
&allocation, &alloc_info));
vmaGetAllocationMemoryProperties(allocator, allocation, &property_flags);
u8 *data = reinterpret_cast<u8 *>(alloc_info.pMappedData);
std::span<u8> mapped_data{};
if (data) {
mapped_data = std::span<u8>{data, ci.size};
}
const bool is_coherent = (property_flags & VK_MEMORY_PROPERTY_HOST_COHERENT_BIT) != 0;
return vk::Buffer(handle, *device.GetLogical(), allocator, allocation, mapped_data, is_coherent,
device.GetDispatchLoader());
}
MemoryCommit MemoryAllocator::Commit(const VkMemoryRequirements &reqs, MemoryUsage usage)
{
const auto vma_usage = MemoryUsageVma(usage);
VmaAllocationCreateInfo ci{};
ci.flags = VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT | MemoryUsageVmaFlags(usage);
ci.usage = vma_usage;
ci.memoryTypeBits = reqs.memoryTypeBits & valid_memory_types;
ci.requiredFlags = 0;
ci.preferredFlags = MemoryUsagePreferredVmaFlags(usage);
VmaAllocation a{};
VmaAllocationInfo info{};
VkResult res = vmaAllocateMemory(allocator, &reqs, &ci, &a, &info);
if (res != VK_SUCCESS) {
// Relax 1: drop budget constraint
auto ci2 = ci;
ci2.flags &= ~VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT;
res = vmaAllocateMemory(allocator, &reqs, &ci2, &a, &info);
// Relax 2: if we preferred DEVICE_LOCAL, drop that preference
if (res != VK_SUCCESS && (ci.preferredFlags & VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT)) {
auto ci3 = ci2;
ci3.preferredFlags &= ~VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT;
res = vmaAllocateMemory(allocator, &reqs, &ci3, &a, &info);
// Log GPU memory allocation for buffers
if (GPU::Logging::IsActive() &&
Settings::values.gpu_log_memory_tracking.GetValue()) {
GPU::Logging::GPULogger::GetInstance().LogMemoryAllocation(
reinterpret_cast<uintptr_t>(alloc_info.deviceMemory),
static_cast<u64>(alloc_info.size),
property_flags
);
}
u8 *data = reinterpret_cast<u8 *>(alloc_info.pMappedData);
const std::span<u8> mapped_data = data ? std::span<u8>{data, ci.size} : std::span<u8>{};
const bool is_coherent = (property_flags & VK_MEMORY_PROPERTY_HOST_COHERENT_BIT) != 0;
return vk::Buffer(handle, *device.GetLogical(), allocator, allocation, mapped_data,
is_coherent,
device.GetDispatchLoader());
}
vk::Check(res);
return MemoryCommit(allocator, a, info);
}
MemoryCommit MemoryAllocator::Commit(const VkMemoryRequirements &reqs, MemoryUsage usage)
{
const auto vma_usage = MemoryUsageVma(usage);
VmaAllocationCreateInfo ci{};
ci.flags = VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT | MemoryUsageVmaFlags(usage);
ci.usage = vma_usage;
ci.memoryTypeBits = reqs.memoryTypeBits & valid_memory_types;
ci.requiredFlags = 0;
ci.preferredFlags = MemoryUsagePreferredVmaFlags(usage);
MemoryCommit MemoryAllocator::Commit(const vk::Buffer &buffer, MemoryUsage usage) {
// Allocate memory appropriate for this buffer automatically
const auto vma_usage = MemoryUsageVma(usage);
VmaAllocation a{};
VmaAllocationInfo info{};
VmaAllocationCreateInfo ci{};
ci.flags = VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT | MemoryUsageVmaFlags(usage);
ci.usage = vma_usage;
ci.requiredFlags = 0;
ci.preferredFlags = MemoryUsagePreferredVmaFlags(usage);
ci.pool = VK_NULL_HANDLE;
ci.pUserData = nullptr;
ci.priority = 0.0f;
VkResult res = vmaAllocateMemory(allocator, &reqs, &ci, &a, &info);
const VkBuffer raw = *buffer;
if (res != VK_SUCCESS) {
// Relax 1: drop budget constraint
auto ci2 = ci;
ci2.flags &= ~VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT;
res = vmaAllocateMemory(allocator, &reqs, &ci2, &a, &info);
VmaAllocation a{};
VmaAllocationInfo info{};
// Let VMA infer memory requirements from the buffer
VkResult res = vmaAllocateMemoryForBuffer(allocator, raw, &ci, &a, &info);
if (res != VK_SUCCESS) {
auto ci2 = ci;
ci2.flags &= ~VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT;
res = vmaAllocateMemoryForBuffer(allocator, raw, &ci2, &a, &info);
if (res != VK_SUCCESS && (ci.preferredFlags & VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT)) {
auto ci3 = ci2;
ci3.preferredFlags &= ~VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT;
res = vmaAllocateMemoryForBuffer(allocator, raw, &ci3, &a, &info);
// Relax 2: if we preferred DEVICE_LOCAL, drop that preference
if (res != VK_SUCCESS && (ci.preferredFlags & VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT)) {
auto ci3 = ci2;
ci3.preferredFlags &= ~VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT;
res = vmaAllocateMemory(allocator, &reqs, &ci3, &a, &info);
}
}
vk::Check(res);
return MemoryCommit(allocator, a, info);
}
vk::Check(res);
vk::Check(vmaBindBufferMemory2(allocator, a, 0, raw, nullptr));
return MemoryCommit(allocator, a, info);
}
MemoryCommit MemoryAllocator::Commit(const vk::Buffer &buffer, MemoryUsage usage) {
// Allocate memory appropriate for this buffer automatically
const auto vma_usage = MemoryUsageVma(usage);
VmaAllocationCreateInfo ci{};
ci.flags = VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT | MemoryUsageVmaFlags(usage);
ci.usage = vma_usage;
ci.requiredFlags = 0;
ci.preferredFlags = MemoryUsagePreferredVmaFlags(usage);
ci.pool = VK_NULL_HANDLE;
ci.pUserData = nullptr;
ci.priority = 0.0f;
const VkBuffer raw = *buffer;
VmaAllocation a{};
VmaAllocationInfo info{};
// Let VMA infer memory requirements from the buffer
VkResult res = vmaAllocateMemoryForBuffer(allocator, raw, &ci, &a, &info);
if (res != VK_SUCCESS) {
auto ci2 = ci;
ci2.flags &= ~VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT;
res = vmaAllocateMemoryForBuffer(allocator, raw, &ci2, &a, &info);
if (res != VK_SUCCESS && (ci.preferredFlags & VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT)) {
auto ci3 = ci2;
ci3.preferredFlags &= ~VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT;
res = vmaAllocateMemoryForBuffer(allocator, raw, &ci3, &a, &info);
}
}
vk::Check(res);
vk::Check(vmaBindBufferMemory2(allocator, a, 0, raw, nullptr));
return MemoryCommit(allocator, a, info);
}
} // namespace Vulkan
@@ -1,4 +1,4 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2019 yuzu Emulator Project
@@ -107,9 +107,6 @@ namespace Vulkan {
vk::Buffer CreateBuffer(const VkBufferCreateInfo &ci, MemoryUsage usage) const;
vk::Buffer CreateBuffer(const VkBufferCreateInfo &ci, MemoryUsage usage,
VkDeviceSize min_alignment) const;
/**
* Commits a memory with the specified requirements.
*
@@ -229,7 +229,6 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
X(vkGetPipelineExecutableStatisticsKHR);
X(vkGetSemaphoreCounterValue);
X(vkMapMemory);
X(vkQueueBindSparse);
X(vkQueueSubmit);
X(vkQueueSubmit2);
X(vkResetFences);
@@ -540,19 +539,6 @@ void Buffer::SetObjectNameEXT(const char* name) const {
SetObjectName(dld, owner, handle, VK_OBJECT_TYPE_BUFFER, name);
}
MemoryLocation Buffer::Location() const noexcept {
if (!allocation) {
return MemoryLocation{};
}
VmaAllocationInfo info{};
vmaGetAllocationInfo(allocator, allocation, &info);
return MemoryLocation{
.memory = info.deviceMemory,
.offset = info.offset,
.memory_type = info.memoryType,
};
}
void Buffer::Release() const noexcept {
if (handle) {
vmaDestroyBuffer(allocator, handle, allocation);
@@ -345,7 +345,6 @@ struct DeviceDispatch : InstanceDispatch {
PFN_vkGetQueryPoolResults vkGetQueryPoolResults{};
PFN_vkGetSemaphoreCounterValue vkGetSemaphoreCounterValue{};
PFN_vkMapMemory vkMapMemory{};
PFN_vkQueueBindSparse vkQueueBindSparse{};
PFN_vkQueueSubmit vkQueueSubmit{};
PFN_vkQueueSubmit2 vkQueueSubmit2{};
PFN_vkResetFences vkResetFences{};
@@ -741,12 +740,6 @@ private:
const DeviceDispatch* dld = nullptr;
};
struct MemoryLocation {
VkDeviceMemory memory{};
VkDeviceSize offset{};
u32 memory_type{};
};
class Buffer {
public:
explicit Buffer(VkBuffer handle_, VkDevice owner_, VmaAllocator allocator_,
@@ -818,8 +811,6 @@ public:
void SetObjectNameEXT(const char* name) const;
MemoryLocation Location() const noexcept;
private:
void Release() const noexcept;
@@ -852,11 +843,6 @@ public:
return dld->vkQueueSubmit2(queue, submit_infos.size(), submit_infos.data(), fence);
}
VkResult BindSparse(Span<VkBindSparseInfo> bind_infos,
VkFence fence = VK_NULL_HANDLE) const noexcept {
return dld->vkQueueBindSparse(queue, bind_infos.size(), bind_infos.data(), fence);
}
VkResult Present(const VkPresentInfoKHR& present_info) const noexcept {
return dld->vkQueuePresentKHR(queue, &present_info);
}
+1 -1
View File
@@ -4796,6 +4796,6 @@ void VolumeButton::ResetMultiplier() {
#endif
#if !defined(QT_STATICPLUGIN) || defined(__APPLE__)
#define VMA_IMPLEMENTATION 1
#define VMA_IMPLEMENTATION
#include "video_core/vulkan_common/vma.h"
#endif
+1 -1
View File
@@ -65,7 +65,7 @@ else()
endif()
# update cached cpmfile content
string(JSON cpmfile SET "${cpmfile}" "${KEY}" "${new_object}")
string(JSON cpmfile SET "${cpmfile}" "${key}" "${new_object}")
# write cached cpmfile
get_cpmfile_path(file)